diff options
author | sra@chromium.org <sra@chromium.org@0039d316-1c4b-4281-b951-d872f2087c98> | 2009-05-08 23:00:29 +0000 |
---|---|---|
committer | sra@chromium.org <sra@chromium.org@0039d316-1c4b-4281-b951-d872f2087c98> | 2009-05-08 23:00:29 +0000 |
commit | 04ca1bc65afb76ea30698c25f24599d20119e3d2 (patch) | |
tree | 2bb65e974d478f4d607b83f6a84a56c01f501ac0 | |
parent | 9adf1dcf3281408560267787eddcc767566b425f (diff) | |
download | chromium_src-04ca1bc65afb76ea30698c25f24599d20119e3d2.zip chromium_src-04ca1bc65afb76ea30698c25f24599d20119e3d2.tar.gz chromium_src-04ca1bc65afb76ea30698c25f24599d20119e3d2.tar.bz2 |
Move Courgette
from src\third_party\courgette
to src\courgette and src\courgette\third_party
Fixed #includes
Added properties to ignore generated files:
C:\c5\src>svn pg svn:ignore courgette
courgette.xcodeproj
courgette.sln
courgette_fuzz.vcproj
courgette_lib.vcproj
courgette_minimal_tool.vcproj
courgette_tool.vcproj
courgette.vcproj
courgette_unittests.vcproj
SConstruct
courgette_fuzz.scons
courgette_lib.scons
courgette_main.scons
courgette_minimal_tool.scons
courgette.scons
courgette_tool.scons
courgette_unittests.scons
Review URL: http://codereview.chromium.org/115062
git-svn-id: svn://svn.chromium.org/chrome/trunk/src@15692 0039d316-1c4b-4281-b951-d872f2087c98
-rw-r--r-- | build/all.gyp | 2 | ||||
-rw-r--r-- | chrome/chrome.sln | 10 | ||||
-rw-r--r-- | courgette/adjustment_method.cc (renamed from third_party/courgette/adjustment_method.cc) | 1406 | ||||
-rw-r--r-- | courgette/adjustment_method.h (renamed from third_party/courgette/adjustment_method.h) | 85 | ||||
-rw-r--r-- | courgette/adjustment_method_unittest.cc (renamed from third_party/courgette/adjustment_method_unittest.cc) | 217 | ||||
-rw-r--r-- | courgette/assembly_program.cc (renamed from third_party/courgette/assembly_program.cc) | 745 | ||||
-rw-r--r-- | courgette/assembly_program.h (renamed from third_party/courgette/assembly_program.h) | 267 | ||||
-rw-r--r-- | courgette/bsdiff_memory_unittest.cc (renamed from third_party/courgette/bsdiff_memory_unittest.cc) | 223 | ||||
-rw-r--r-- | courgette/courgette.gyp (renamed from third_party/courgette/courgette.gyp) | 49 | ||||
-rw-r--r-- | courgette/courgette.h (renamed from third_party/courgette/courgette.h) | 236 | ||||
-rw-r--r-- | courgette/courgette.sln | 111 | ||||
-rw-r--r-- | courgette/courgette_minimal_tool.cc (renamed from third_party/courgette/courgette_minimal_tool.cc) | 100 | ||||
-rw-r--r-- | courgette/courgette_tool.cc (renamed from third_party/courgette/courgette_tool.cc) | 809 | ||||
-rw-r--r-- | courgette/crc.cc (renamed from third_party/courgette/crc.cc) | 44 | ||||
-rw-r--r-- | courgette/crc.h (renamed from third_party/courgette/crc.h) | 34 | ||||
-rw-r--r-- | courgette/difference_estimator.cc (renamed from third_party/courgette/difference_estimator.cc) | 236 | ||||
-rw-r--r-- | courgette/difference_estimator.h (renamed from third_party/courgette/difference_estimator.h) | 116 | ||||
-rw-r--r-- | courgette/difference_estimator_unittest.cc (renamed from third_party/courgette/difference_estimator_unittest.cc) | 116 | ||||
-rw-r--r-- | courgette/disassembler.cc (renamed from third_party/courgette/disassembler.cc) | 872 | ||||
-rw-r--r-- | courgette/disassembler.h (renamed from third_party/courgette/disassembler.h) | 77 | ||||
-rw-r--r-- | courgette/encode_decode_unittest.cc (renamed from third_party/courgette/encode_decode_unittest.cc) | 5 | ||||
-rw-r--r-- | courgette/encoded_program.cc (renamed from third_party/courgette/encoded_program.cc) | 6 | ||||
-rw-r--r-- | courgette/encoded_program.h (renamed from third_party/courgette/encoded_program.h) | 3 | ||||
-rw-r--r-- | courgette/encoded_program_fuzz_unittest.cc (renamed from third_party/courgette/encoded_program_fuzz_unittest.cc) | 471 | ||||
-rw-r--r-- | courgette/encoded_program_unittest.cc (renamed from third_party/courgette/encoded_program_unittest.cc) | 5 | ||||
-rw-r--r-- | courgette/ensemble.cc (renamed from third_party/courgette/ensemble.cc) | 204 | ||||
-rw-r--r-- | courgette/ensemble.h (renamed from third_party/courgette/ensemble.h) | 514 | ||||
-rw-r--r-- | courgette/ensemble_apply.cc (renamed from third_party/courgette/ensemble_apply.cc) | 824 | ||||
-rw-r--r-- | courgette/ensemble_create.cc (renamed from third_party/courgette/ensemble_create.cc) | 764 | ||||
-rw-r--r-- | courgette/image_info.cc (renamed from third_party/courgette/image_info.cc) | 3 | ||||
-rw-r--r-- | courgette/image_info.h (renamed from third_party/courgette/image_info.h) | 1 | ||||
-rw-r--r-- | courgette/image_info_unittest.cc (renamed from third_party/courgette/image_info_unittest.cc) | 3 | ||||
-rw-r--r-- | courgette/memory_monitor.cc (renamed from third_party/courgette/memory_monitor.cc) | 244 | ||||
-rw-r--r-- | courgette/region.h (renamed from third_party/courgette/region.h) | 124 | ||||
-rw-r--r-- | courgette/run_all_unittests.cc (renamed from third_party/courgette/run_all_unittests.cc) | 19 | ||||
-rw-r--r-- | courgette/simple_delta.cc (renamed from third_party/courgette/simple_delta.cc) | 82 | ||||
-rw-r--r-- | courgette/simple_delta.h (renamed from third_party/courgette/simple_delta.h) | 46 | ||||
-rw-r--r-- | courgette/streams.cc (renamed from third_party/courgette/streams.cc) | 4 | ||||
-rw-r--r-- | courgette/streams.h (renamed from third_party/courgette/streams.h) | 2 | ||||
-rw-r--r-- | courgette/streams_unittest.cc (renamed from third_party/courgette/streams_unittest.cc) | 2 | ||||
-rw-r--r-- | courgette/testdata/en-US.dll (renamed from third_party/courgette/testdata/en-US.dll) | bin | 73216 -> 73216 bytes | |||
-rw-r--r-- | courgette/testdata/setup1.exe (renamed from third_party/courgette/testdata/setup1.exe) | bin | 967168 -> 967168 bytes | |||
-rw-r--r-- | courgette/testdata/setup2.exe (renamed from third_party/courgette/testdata/setup2.exe) | bin | 965632 -> 965632 bytes | |||
-rw-r--r-- | courgette/third_party/LICENCE (renamed from third_party/courgette/LICENCE.bsdiff) | 0 | ||||
-rw-r--r-- | courgette/third_party/README.chromium | 23 | ||||
-rw-r--r-- | courgette/third_party/bsdiff.h (renamed from third_party/courgette/bsdiff.h) | 0 | ||||
-rw-r--r-- | courgette/third_party/bsdiff_apply.cc (renamed from third_party/courgette/bsdiff_apply.cc) | 6 | ||||
-rw-r--r-- | courgette/third_party/bsdiff_create.cc (renamed from third_party/courgette/bsdiff_create.cc) | 6 | ||||
-rw-r--r-- | courgette/win32_x86_generator.h (renamed from third_party/courgette/win32_x86_generator.h) | 252 | ||||
-rw-r--r-- | courgette/win32_x86_patcher.h (renamed from third_party/courgette/win32_x86_patcher.h) | 190 |
50 files changed, 4834 insertions, 4724 deletions
diff --git a/build/all.gyp b/build/all.gyp index a2ef584..f2c85a8 100644 --- a/build/all.gyp +++ b/build/all.gyp @@ -59,7 +59,7 @@ ['OS=="win"', { 'dependencies': [ '../sandbox/sandbox.gyp:*', - '../third_party/courgette/courgette.gyp:*', + '../courgette/courgette.gyp:*', '../webkit/activex_shim/activex_shim.gyp:*', '../webkit/activex_shim_dll/activex_shim_dll.gyp:*', 'temp_gyp/breakpad.gyp:*', diff --git a/chrome/chrome.sln b/chrome/chrome.sln index 4be747b..ce4d621 100644 --- a/chrome/chrome.sln +++ b/chrome/chrome.sln @@ -1487,9 +1487,9 @@ Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "tcmalloc", "..\third_party\ EndProject Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "app", "..\app\app.vcproj", "{4631946D-7D5F-44BD-A5A8-504C0A7033BE}" EndProject -Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "courgette_lib", "..\third_party\courgette\courgette_lib.vcproj", "{9A72A362-E617-4205-B9F2-43C6FB280FA1}" +Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "courgette_lib", "..\courgette\courgette_lib.vcproj", "{9A72A362-E617-4205-B9F2-43C6FB280FA1}" EndProject -Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "courgette_fuzz", "..\third_party\courgette\courgette_fuzz.vcproj", "{57C27529-8CA9-4FC3-9C02-DA05B172F785}" +Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "courgette_fuzz", "..\courgette\courgette_fuzz.vcproj", "{57C27529-8CA9-4FC3-9C02-DA05B172F785}" ProjectSection(ProjectDependencies) = postProject {1832A374-8A74-4F9E-B536-69A699B3E165} = {1832A374-8A74-4F9E-B536-69A699B3E165} {8C27D792-2648-4F5E-9ED0-374276327308} = {8C27D792-2648-4F5E-9ED0-374276327308} @@ -1498,7 +1498,7 @@ Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "courgette_fuzz", "..\third_ {F22022F0-2D3B-5610-4E80-C674A8E01C5A} = {F22022F0-2D3B-5610-4E80-C674A8E01C5A} EndProjectSection EndProject -Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "courgette_unittests", "..\third_party\courgette\courgette_unittests.vcproj", "{24309F1A-4035-46F9-A3D8-F47DC4BCC2B8}" +Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "courgette_unittests", "..\courgette\courgette_unittests.vcproj", "{24309F1A-4035-46F9-A3D8-F47DC4BCC2B8}" ProjectSection(ProjectDependencies) = postProject {1832A374-8A74-4F9E-B536-69A699B3E165} = {1832A374-8A74-4F9E-B536-69A699B3E165} {8C27D792-2648-4F5E-9ED0-374276327308} = {8C27D792-2648-4F5E-9ED0-374276327308} @@ -1507,7 +1507,7 @@ Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "courgette_unittests", "..\t {F22022F0-2D3B-5610-4E80-C674A8E01C5A} = {F22022F0-2D3B-5610-4E80-C674A8E01C5A} EndProjectSection EndProject -Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "courgette_tool", "..\third_party\courgette\courgette_tool.vcproj", "{4EA8CE12-9C6F-45E5-9D08-720383FE3685}" +Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "courgette", "..\courgette\courgette.vcproj", "{4EA8CE12-9C6F-45E5-9D08-720383FE3685}" ProjectSection(ProjectDependencies) = postProject {1832A374-8A74-4F9E-B536-69A699B3E165} = {1832A374-8A74-4F9E-B536-69A699B3E165} {8C27D792-2648-4F5E-9ED0-374276327308} = {8C27D792-2648-4F5E-9ED0-374276327308} @@ -1515,7 +1515,7 @@ Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "courgette_tool", "..\third_ {F22022F0-2D3B-5610-4E80-C674A8E01C5A} = {F22022F0-2D3B-5610-4E80-C674A8E01C5A} EndProjectSection EndProject -Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "courgette_minimal_tool", "..\third_party\courgette\courgette_minimal_tool.vcproj", "{EB79415F-2F17-4BDC-AADD-4CA4C2D21B73}" +Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "courgette_minimal_tool", "..\courgette\courgette_minimal_tool.vcproj", "{EB79415F-2F17-4BDC-AADD-4CA4C2D21B73}" ProjectSection(ProjectDependencies) = postProject {1832A374-8A74-4F9E-B536-69A699B3E165} = {1832A374-8A74-4F9E-B536-69A699B3E165} {8C27D792-2648-4F5E-9ED0-374276327308} = {8C27D792-2648-4F5E-9ED0-374276327308} diff --git a/third_party/courgette/adjustment_method.cc b/courgette/adjustment_method.cc index 8f6caef..48e7cc0 100644 --- a/third_party/courgette/adjustment_method.cc +++ b/courgette/adjustment_method.cc @@ -1,703 +1,703 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-#include "third_party/courgette/adjustment_method.h"
-
-#include <algorithm>
-#include <list>
-#include <map>
-#include <set>
-#include <string>
-#include <vector>
-
-#include "base/basictypes.h"
-#include "base/logging.h"
-#include "base/string_util.h"
-
-#include "third_party/courgette/assembly_program.h"
-#include "third_party/courgette/courgette.h"
-#include "third_party/courgette/encoded_program.h"
-#include "third_party/courgette/image_info.h"
-
-namespace courgette {
-
-// We have three discretionary information logging levels for algorithm
-// development. For now just configure with #defines.
-// TODO(sra): make dependent of some configurable setting.
-#define NO_LOG DLOG_IF(INFO, false)
-// #define ALOG1 LOG(INFO)
-// #define ALOG2 LOG(INFO)
-// #define ALOG3 LOG(INFO)
-#define ALOG1 NO_LOG
-#define ALOG2 NO_LOG
-#define ALOG3 NO_LOG
-
-////////////////////////////////////////////////////////////////////////////////
-
-class NullAdjustmentMethod : public AdjustmentMethod {
- bool Adjust(const AssemblyProgram& model, AssemblyProgram* program) {
- return true;
- }
-};
-
-////////////////////////////////////////////////////////////////////////////////
-
-// The purpose of adjustment is to assign indexes to Labels of a program 'p' to
-// make the sequence of indexes similar to a 'model' program 'm'. Labels
-// themselves don't have enough information to do this job, so we work with a
-// LabelInfo surrogate for each label.
-//
-class LabelInfo {
- public:
- Label* label_; // The label that this info a surrogate for.
-
- // Information used only in debugging messages.
- uint32 is_model_ : 1; // Is the label in the model?
- uint32 debug_index_ : 31; // An unique small number for naming the label.
-
- uint32 refs_; // Number of times this Label is referenced.
-
- LabelInfo* assignment_; // Label from other program corresponding to this.
-
- // LabelInfos are in a doubly linked list ordered by address (label_->rva_) so
- // we can quickly find Labels adjacent in address order.
- LabelInfo* next_addr_; // Label(Info) at next highest address.
- LabelInfo* prev_addr_; // Label(Info) at next lowest address.
-
- std::vector<uint32> positions_; // Offsets into the trace of references.
-
- // Just a no-argument constructor and copy constructor. Actual LabelInfo
- // objects are allocated in std::pair structs in a std::map.
- LabelInfo()
- : label_(NULL), is_model_(false), debug_index_(0), refs_(0),
- assignment_(NULL),
- next_addr_(NULL),
- prev_addr_(NULL) {}
-
- private:
- void operator=(const LabelInfo*); // Disallow assignment only.
-
- // Public compiler generated copy constructor is needed to constuct
- // std::pair<Label*, LabelInfo> so that fresh LabelInfos can be allocated
- // inside a std::map.
-};
-
-struct OrderLabelInfoByAddressAscending {
- bool operator()(const LabelInfo* a, const LabelInfo* b) const {
- return a->label_->rva_ < b->label_->rva_;
- }
-};
-
-static std::string ToString(LabelInfo* info) {
- std::string s;
- StringAppendF(&s, "%c%d", "pm"[info->is_model_], info->debug_index_);
- if (info->label_->index_ != Label::kNoIndex)
- StringAppendF(&s, " (%d)", info->label_->index_);
-
- StringAppendF(&s, " #%u", info->refs_);
- return s;
-}
-
-// General graph matching is exponential, essentially trying all permutations.
-// The exponential algorithm can be made faster by avoiding consideration of
-// impossible or unlikely matches. We can make the matching practical by eager
-// matching - by looking for likely matches and commiting to them, and using the
-// committed assignment as the basis for further matching.
-//
-// The basic eager graph-matching assignment is based on several ideas:
-//
-// * The strongest match will be for parts of the program that have not
-// changed. If part of a program has not changed, then the number of
-// references to a label will be the same, and corresponding pairs of
-// adjacent labels will have the same RVA difference.
-//
-// * Some assignments are 'obvious' if you look at the distribution. Example:
-// if both the program and the model have a label that is referred to much
-// more often than the next most refered-to label, it is likely the two
-// labels correspond.
-//
-// * If a label from the program corresponds to a label in the model, it is
-// likely that the labels near the corresponding labels also match. A
-// conservative way of extending the match is to assign only those labels
-// which have exactly the same address offset and reference count.
-//
-// * If two labels correspond, then we can try to match up the references
-// before and after the labels in the reference stream. For this to be
-// practical, the number of references has to be small, e.g. each label has
-// exactly one reference.
-//
-
-// Note: we also tried a completely different approach: random assignment
-// followed by simulated annealing. This produced similar results. The results
-// were not as good for very small differences because the simulated annealing
-// never quite hit the groove. And simulated annealing was several orders of
-// magnitude slower.
-
-
-// TRIE node for suffix strings in the label reference sequence.
-//
-// We dynamically build a trie for both the program and model, growing the trie
-// as necessary. The trie node for a (possibly) empty string of label
-// references contains the distribution of labels following the string. The
-// roots node (for the empty string) thus contains the simple distribution of
-// labels within the label reference stream.
-
-struct Node {
- Node(LabelInfo* in_edge, Node* prev)
- : in_edge_(in_edge), prev_(prev), count_(0),
- in_queue_(false) {
- length_ = 1 + (prev_ ? prev_->length_ : 0);
- }
- LabelInfo* in_edge_; //
- Node* prev_; // Node at shorter length.
- int count_; // Frequency of this path in Trie.
- int length_;
- typedef std::map<LabelInfo*, Node*> Edges;
- Edges edges_;
- std::vector<int> places_; // Indexes into sequence of this item.
- std::list<Node*> edges_in_frequency_order;
-
- bool in_queue_;
- bool Extended() const { return edges_.size() > 0; }
-
- uint32 Weight() const {
- return edges_in_frequency_order.front()->count_;
- }
-};
-
-static std::string ToString(Node* node) {
- std::vector<std::string> prefix;
- for (Node* n = node; n->prev_; n = n->prev_)
- prefix.push_back(ToString(n->in_edge_));
-
- std::string s;
- s += "{";
- const char* sep = "";
- while (!prefix.empty()) {
- s += sep;
- sep = ",";
- s += prefix.back();
- prefix.pop_back();
- }
-
- s += StringPrintf("%u", node->count_);
- s += " @";
- s += Uint64ToString(node->edges_in_frequency_order.size());
- s += "}";
- return s;
-}
-
-typedef std::vector<LabelInfo*> Trace;
-
-struct OrderNodeByCountDecreasing {
- bool operator()(Node* a, Node* b) const {
- if (a->count_ != b->count_)
- return (a->count_) > (b->count_);
- return a->places_.at(0) < b->places_.at(0); // Prefer first occuring.
- }
-};
-
-struct OrderNodeByWeightDecreasing {
- bool operator()(Node* a, Node* b) const {
- // (Maybe tie-break on total count, followed by lowest assigned node indexes
- // in path.)
- uint32 a_weight = a->Weight();
- uint32 b_weight = b->Weight();
- if (a_weight != b_weight)
- return a_weight > b_weight;
- if (a->length_ != b->length_)
- return a->length_ > b->length_; // Prefer longer.
- return a->places_.at(0) < b->places_.at(0); // Prefer first occuring.
- }
-};
-
-typedef std::set<Node*, OrderNodeByWeightDecreasing> NodeQueue;
-
-class AssignmentProblem {
- public:
- AssignmentProblem(const Trace& model,
- const Trace& problem)
- : m_trace_(model),
- p_trace_(problem) {
- }
-
- ~AssignmentProblem() {
- for (size_t i = 0; i < all_nodes_.size(); ++i)
- delete all_nodes_[i];
- }
-
- bool Solve() {
- m_root_ = MakeRootNode(m_trace_);
- p_root_ = MakeRootNode(p_trace_);
- AddToQueue(p_root_);
-
- while (!worklist_.empty()) {
- Node* node = *worklist_.begin();
- node->in_queue_ = false;
- worklist_.erase(node);
- TrySolveNode(node);
- }
-
- ALOG1 << unsolved_.size() << " unsolved items";
- return true;
- }
-
- private:
- void AddToQueue(Node* node) {
- if (node->length_ >= 10) {
- ALOG3 << "Length clipped " << ToString(node->prev_);
- return;
- }
- if (node->in_queue_) {
- LOG(ERROR) << "Double add " << ToString(node);
- return;
- }
- // just to be sure data for prioritizing is available
- ExtendNode(node, p_trace_);
- // SkipCommittedLabels(node);
- if (node->edges_in_frequency_order.empty())
- return;
- node->in_queue_ = true;
- worklist_.insert(node);
- }
-
- void SkipCommittedLabels(Node* node) {
- ExtendNode(node, p_trace_);
- uint32 skipped = 0;
- while (!node->edges_in_frequency_order.empty() &&
- node->edges_in_frequency_order.front()->in_edge_->assignment_) {
- ++skipped;
- node->edges_in_frequency_order.pop_front();
- }
- if (skipped > 0)
- ALOG3 << "Skipped " << skipped << " at " << ToString(node);
- }
-
- void TrySolveNode(Node* p_node) {
- Node* front = p_node->edges_in_frequency_order.front();
- if (front->in_edge_->assignment_) {
- p_node->edges_in_frequency_order.pop_front();
- AddToQueue(front);
- AddToQueue(p_node);
- return;
- }
-
- // Compare frequencies of unassigned edges, and either make
- // assignment(s) or move node to unsolved list
-
- Node* m_node = FindModelNode(p_node);
-
- if (m_node == NULL) {
- ALOG1 << "Can't find model node";
- unsolved_.insert(p_node);
- return;
- }
- ExtendNode(m_node, m_trace_);
-
- // Lets just try greedy
-
- SkipCommittedLabels(m_node);
- if (m_node->edges_in_frequency_order.empty()) {
- ALOG3 << "Punting, no elements left in model vs "
- << p_node->edges_in_frequency_order.size();
- unsolved_.insert(p_node);
- return;
- }
- Node* m_match = m_node->edges_in_frequency_order.front();
- Node* p_match = p_node->edges_in_frequency_order.front();
-
- if (p_match->count_ > 1.1 * m_match->count_ ||
- m_match->count_ > 1.1 * p_match->count_) {
- ALOG2 << "Tricky distribution "
- << p_match->count_ << ":" << m_match->count_ << " "
- << ToString(p_match) << " vs " << ToString(m_match);
- return;
- }
-
- m_node->edges_in_frequency_order.pop_front();
- p_node->edges_in_frequency_order.pop_front();
-
- LabelInfo* p_label_info = p_match->in_edge_;
- LabelInfo* m_label_info = m_match->in_edge_;
- int m_index = p_label_info->label_->index_;
- if (m_index != Label::kNoIndex) {
- ALOG1 << "Cant use unassigned label from model " << m_index;
- unsolved_.insert(p_node);
- return;
- }
-
- Assign(p_label_info, m_label_info);
-
- AddToQueue(p_match); // find matches within new match
- AddToQueue(p_node); // and more matches within this node
- }
-
- void Assign(LabelInfo* p_info, LabelInfo* m_info) {
- AssignOne(p_info, m_info);
- ALOG3 << "Assign " << ToString(p_info) << " := " << ToString(m_info);
- // Now consider unassigned adjacent addresses
- TryExtendAssignment(p_info, m_info);
- }
-
- void AssignOne(LabelInfo* p_info, LabelInfo* m_info) {
- p_info->label_->index_ = m_info->label_->index_;
-
- // Mark as assigned
- m_info->assignment_ = p_info;
- p_info->assignment_ = m_info;
- }
-
- void TryExtendAssignment(LabelInfo* p_info, LabelInfo* m_info) {
- RVA m_rva_base = m_info->label_->rva_;
- RVA p_rva_base = p_info->label_->rva_;
-
- LabelInfo* m_info_next = m_info->next_addr_;
- LabelInfo* p_info_next = p_info->next_addr_;
- for ( ; m_info_next && p_info_next; ) {
- if (m_info_next->assignment_)
- break;
-
- RVA m_rva = m_info_next->label_->rva_;
- RVA p_rva = p_info_next->label_->rva_;
-
- if (m_rva - m_rva_base != p_rva - p_rva_base) {
- // previous label was pointing to something that is different size
- break;
- }
- LabelInfo* m_info_next_next = m_info_next->next_addr_;
- LabelInfo* p_info_next_next = p_info_next->next_addr_;
- if (m_info_next_next && p_info_next_next) {
- RVA m_rva_next = m_info_next_next->label_->rva_;
- RVA p_rva_next = p_info_next_next->label_->rva_;
- if (m_rva_next - m_rva != p_rva_next - p_rva) {
- // Since following labels are no longer in address lockstep, assume
- // this address has a difference.
- break;
- }
- }
-
- // The label has inconsistent numbers of references, it is probably not
- // the same thing.
- if (m_info_next->refs_ != p_info_next->refs_) {
- break;
- }
-
- ALOG3 << " Extending assignment -> "
- << ToString(p_info_next) << " := " << ToString(m_info_next);
-
- AssignOne(p_info_next, m_info_next);
-
- if (p_info_next->refs_ == m_info_next->refs_ &&
- p_info_next->refs_ == 1) {
- TryExtendSequence(p_info_next->positions_[0],
- m_info_next->positions_[0]);
- TryExtendSequenceBackwards(p_info_next->positions_[0],
- m_info_next->positions_[0]);
- }
-
- p_info_next = p_info_next_next;
- m_info_next = m_info_next_next;
- }
-
- LabelInfo* m_info_prev = m_info->prev_addr_;
- LabelInfo* p_info_prev = p_info->prev_addr_;
- for ( ; m_info_prev && p_info_prev; ) {
- if (m_info_prev->assignment_)
- break;
-
- RVA m_rva = m_info_prev->label_->rva_;
- RVA p_rva = p_info_prev->label_->rva_;
-
- if (m_rva - m_rva_base != p_rva - p_rva_base) {
- // previous label was pointing to something that is different size
- break;
- }
- LabelInfo* m_info_prev_prev = m_info_prev->prev_addr_;
- LabelInfo* p_info_prev_prev = p_info_prev->prev_addr_;
-
- // The the label has inconsistent numbers of references, it is
- // probably not the same thing
- if (m_info_prev->refs_ != p_info_prev->refs_) {
- break;
- }
-
- AssignOne(p_info_prev, m_info_prev);
- ALOG3 << " Extending assignment <- " << ToString(p_info_prev) << " := "
- << ToString(m_info_prev);
-
- p_info_prev = p_info_prev_prev;
- m_info_prev = m_info_prev_prev;
- }
- }
-
- uint32 TryExtendSequence(uint32 p_pos_start, uint32 m_pos_start) {
- uint32 p_pos = p_pos_start + 1;
- uint32 m_pos = m_pos_start + 1;
-
- while (p_pos < p_trace_.size() && m_pos < m_trace_.size()) {
- LabelInfo* p_info = p_trace_[p_pos];
- LabelInfo* m_info = m_trace_[m_pos];
-
- // To match, either (1) both are assigned or (2) both are unassigned.
- if ((p_info->assignment_ == NULL) != (m_info->assignment_ == NULL))
- break;
-
- // If they are assigned, it needs to be consistent (same index).
- if (p_info->assignment_ && m_info->assignment_) {
- if (p_info->label_->index_ != m_info->label_->index_)
- break;
- ++p_pos;
- ++m_pos;
- continue;
- }
-
- if (p_info->refs_ != m_info->refs_)
- break;
-
- AssignOne(p_info, m_info);
- ALOG3 << " Extending assignment seq"
- << "[+" << p_pos - p_pos_start << "]"
- << " -> "
- << ToString(p_info) << " := " << ToString(m_info);
-
- ++p_pos;
- ++m_pos;
- }
-
- return p_pos - p_pos_start;
- }
-
- uint32 TryExtendSequenceBackwards(uint32 p_pos_start, uint32 m_pos_start) {
- if (p_pos_start == 0 || m_pos_start == 0)
- return 0;
-
- uint32 p_pos = p_pos_start - 1;
- uint32 m_pos = m_pos_start - 1;
-
- while (p_pos > 0 && m_pos > 0) {
- LabelInfo* p_info = p_trace_[p_pos];
- LabelInfo* m_info = m_trace_[m_pos];
-
- if ((p_info->assignment_ == NULL) != (m_info->assignment_ == NULL))
- break;
-
- if (p_info->assignment_ && m_info->assignment_) {
- if (p_info->label_->index_ != m_info->label_->index_)
- break;
- --p_pos;
- --m_pos;
- continue;
- }
-
- if (p_info->refs_ != m_info->refs_)
- break;
-
- AssignOne(p_info, m_info);
- ALOG3 << " Extending assignment seq"
- << "[-" << p_pos_start - p_pos << "]"
- << " <- "
- << ToString(p_info) << " := " << ToString(m_info);
-
- --p_pos;
- --m_pos;
- }
-
- return p_pos - p_pos_start;
- }
-
- Node* FindModelNode(Node* node) {
- if (node->prev_ == NULL)
- return m_root_;
-
- Node* m_parent = FindModelNode(node->prev_);
- if (m_parent == NULL) {
- return NULL;
- }
-
- ExtendNode(m_parent, m_trace_);
-
- LabelInfo* p_label = node->in_edge_;
- LabelInfo* m_label = p_label->assignment_;
- if (m_label == NULL) {
- ALOG1 << "Expected assigned prefix";
- return NULL;
- }
-
- Node::Edges::iterator e = m_parent->edges_.find(m_label);
- if (e == m_parent->edges_.end()) {
- ALOG2 << "Expected defined edge in parent";
- return NULL;
- }
-
- return e->second;
- }
-
- Node* MakeRootNode(const Trace& trace) {
- Node* node = new Node(NULL, NULL);
- all_nodes_.push_back(node);
- for (size_t i = 0; i < trace.size(); ++i) {
- ++node->count_;
- node->places_.push_back(i);
- }
- return node;
- }
-
- void ExtendNode(Node* node, const Trace& trace) {
- // Make sure trie is filled in at this node.
- if (node->Extended())
- return;
- for (size_t i = 0; i < node->places_.size(); ++i) {
- uint32 index = node->places_.at(i);
- if (index < trace.size()) {
- LabelInfo* item = trace.at(index);
- Node*& slot = node->edges_[item];
- if (slot == NULL) {
- slot = new Node(item, node);
- all_nodes_.push_back(slot);
- node->edges_in_frequency_order.push_back(slot);
- }
- slot->places_.push_back(index + 1);
- ++slot->count_;
- }
- }
- node->edges_in_frequency_order.sort(OrderNodeByCountDecreasing());
- }
-
- const Trace& m_trace_;
- const Trace& p_trace_;
- Node* m_root_;
- Node* p_root_;
-
- NodeQueue worklist_;
- NodeQueue unsolved_;
-
- std::vector<Node*> all_nodes_;
-
- DISALLOW_COPY_AND_ASSIGN(AssignmentProblem);
-};
-
-class GraphAdjuster : public AdjustmentMethod {
- public:
- GraphAdjuster() {}
- ~GraphAdjuster() {}
-
- bool Adjust(const AssemblyProgram& model, AssemblyProgram* program) {
- LOG(INFO) << "GraphAdjuster::Adjust";
- prog_ = program;
- model_ = &model;
- debug_label_index_gen_ = 0;
- return Finish();
- }
-
- bool Finish() {
- prog_->UnassignIndexes();
- CollectTraces(model_, &model_abs32_, &model_rel32_, true);
- CollectTraces(prog_, &prog_abs32_, &prog_rel32_, false);
- Solve(model_abs32_, prog_abs32_);
- Solve(model_rel32_, prog_rel32_);
- prog_->AssignRemainingIndexes();
- return true;
- }
-
- private:
-
- void CollectTraces(const AssemblyProgram* program, Trace* abs32, Trace* rel32,
- bool is_model) {
- const std::vector<Instruction*>& instructions = program->instructions();
- for (size_t i = 0; i < instructions.size(); ++i) {
- Instruction* instruction = instructions.at(i);
- if (Label* label = program->InstructionAbs32Label(instruction))
- ReferenceLabel(abs32, label, is_model);
- if (Label* label = program->InstructionRel32Label(instruction))
- ReferenceLabel(rel32, label, is_model);
- }
- // TODO(sra): we could simply append all the labels in index order to
- // incorporate some costing for entropy (bigger deltas) that will be
- // introduced into the label address table by non-monotonic ordering. This
- // would have some knock-on effects to parts of the algorithm that work on
- // single-occurence labels.
- }
-
- void Solve(const Trace& model, const Trace& problem) {
- LinkLabelInfos(model);
- LinkLabelInfos(problem);
- AssignmentProblem a(model, problem);
- a.Solve();
- }
-
- void LinkLabelInfos(const Trace& trace) {
- typedef std::set<LabelInfo*, OrderLabelInfoByAddressAscending> Ordered;
- Ordered ordered;
- for (Trace::const_iterator p = trace.begin(); p != trace.end(); ++p)
- ordered.insert(*p);
- LabelInfo* prev = NULL;
- for (Ordered::iterator p = ordered.begin(); p != ordered.end(); ++p) {
- LabelInfo* curr = *p;
- if (prev) prev->next_addr_ = curr;
- curr->prev_addr_ = prev;
- prev = curr;
-
- if (curr->positions_.size() != curr->refs_)
- NOTREACHED();
- }
- }
-
- void ReferenceLabel(Trace* trace, Label* label, bool is_model) {
- trace->push_back(MakeLabelInfo(label, is_model, trace->size()));
- }
-
- LabelInfo* MakeLabelInfo(Label* label, bool is_model, uint32 position) {
- LabelInfo& slot = label_infos_[label];
- if (slot.label_ == NULL) {
- slot.label_ = label;
- slot.is_model_ = is_model;
- slot.debug_index_ = ++debug_label_index_gen_;
- }
- slot.positions_.push_back(position);
- ++slot.refs_;
- return &slot;
- }
-
- AssemblyProgram* prog_; // Program to be adjusted, owned by caller.
- const AssemblyProgram* model_; // Program to be mimicked, owned by caller.
-
- Trace model_abs32_;
- Trace model_rel32_;
- Trace prog_abs32_;
- Trace prog_rel32_;
-
- int debug_label_index_gen_;
-
- // Note LabelInfo is allocated inside map, so the LabelInfo lifetimes are
- // managed by the map.
- std::map<Label*, LabelInfo> label_infos_;
-
- private:
- DISALLOW_COPY_AND_ASSIGN(GraphAdjuster);
-};
-
-
-////////////////////////////////////////////////////////////////////////////////
-
-void AdjustmentMethod::Destroy() { delete this; }
-
-AdjustmentMethod* AdjustmentMethod::MakeNullAdjustmentMethod() {
- return new NullAdjustmentMethod();
-}
-
-AdjustmentMethod* AdjustmentMethod::MakeProductionAdjustmentMethod() {
- return new GraphAdjuster();
-}
-
-Status Adjust(const AssemblyProgram& model, AssemblyProgram* program) {
- AdjustmentMethod* method = AdjustmentMethod::MakeProductionAdjustmentMethod();
- bool ok = method->Adjust(model, program);
- method->Destroy();
- if (ok)
- return C_OK;
- else
- return C_ADJUSTMENT_FAILED;
-}
-
-} // namespace courgette
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +#include "courgette/adjustment_method.h" + +#include <algorithm> +#include <list> +#include <map> +#include <set> +#include <string> +#include <vector> + +#include "base/basictypes.h" +#include "base/logging.h" +#include "base/string_util.h" + +#include "courgette/assembly_program.h" +#include "courgette/courgette.h" +#include "courgette/encoded_program.h" +#include "courgette/image_info.h" + +namespace courgette { + +// We have three discretionary information logging levels for algorithm +// development. For now just configure with #defines. +// TODO(sra): make dependent of some configurable setting. +#define NO_LOG DLOG_IF(INFO, false) +// #define ALOG1 LOG(INFO) +// #define ALOG2 LOG(INFO) +// #define ALOG3 LOG(INFO) +#define ALOG1 NO_LOG +#define ALOG2 NO_LOG +#define ALOG3 NO_LOG + +//////////////////////////////////////////////////////////////////////////////// + +class NullAdjustmentMethod : public AdjustmentMethod { + bool Adjust(const AssemblyProgram& model, AssemblyProgram* program) { + return true; + } +}; + +//////////////////////////////////////////////////////////////////////////////// + +// The purpose of adjustment is to assign indexes to Labels of a program 'p' to +// make the sequence of indexes similar to a 'model' program 'm'. Labels +// themselves don't have enough information to do this job, so we work with a +// LabelInfo surrogate for each label. +// +class LabelInfo { + public: + Label* label_; // The label that this info a surrogate for. + + // Information used only in debugging messages. + uint32 is_model_ : 1; // Is the label in the model? + uint32 debug_index_ : 31; // An unique small number for naming the label. + + uint32 refs_; // Number of times this Label is referenced. + + LabelInfo* assignment_; // Label from other program corresponding to this. + + // LabelInfos are in a doubly linked list ordered by address (label_->rva_) so + // we can quickly find Labels adjacent in address order. + LabelInfo* next_addr_; // Label(Info) at next highest address. + LabelInfo* prev_addr_; // Label(Info) at next lowest address. + + std::vector<uint32> positions_; // Offsets into the trace of references. + + // Just a no-argument constructor and copy constructor. Actual LabelInfo + // objects are allocated in std::pair structs in a std::map. + LabelInfo() + : label_(NULL), is_model_(false), debug_index_(0), refs_(0), + assignment_(NULL), + next_addr_(NULL), + prev_addr_(NULL) {} + + private: + void operator=(const LabelInfo*); // Disallow assignment only. + + // Public compiler generated copy constructor is needed to constuct + // std::pair<Label*, LabelInfo> so that fresh LabelInfos can be allocated + // inside a std::map. +}; + +struct OrderLabelInfoByAddressAscending { + bool operator()(const LabelInfo* a, const LabelInfo* b) const { + return a->label_->rva_ < b->label_->rva_; + } +}; + +static std::string ToString(LabelInfo* info) { + std::string s; + StringAppendF(&s, "%c%d", "pm"[info->is_model_], info->debug_index_); + if (info->label_->index_ != Label::kNoIndex) + StringAppendF(&s, " (%d)", info->label_->index_); + + StringAppendF(&s, " #%u", info->refs_); + return s; +} + +// General graph matching is exponential, essentially trying all permutations. +// The exponential algorithm can be made faster by avoiding consideration of +// impossible or unlikely matches. We can make the matching practical by eager +// matching - by looking for likely matches and commiting to them, and using the +// committed assignment as the basis for further matching. +// +// The basic eager graph-matching assignment is based on several ideas: +// +// * The strongest match will be for parts of the program that have not +// changed. If part of a program has not changed, then the number of +// references to a label will be the same, and corresponding pairs of +// adjacent labels will have the same RVA difference. +// +// * Some assignments are 'obvious' if you look at the distribution. Example: +// if both the program and the model have a label that is referred to much +// more often than the next most refered-to label, it is likely the two +// labels correspond. +// +// * If a label from the program corresponds to a label in the model, it is +// likely that the labels near the corresponding labels also match. A +// conservative way of extending the match is to assign only those labels +// which have exactly the same address offset and reference count. +// +// * If two labels correspond, then we can try to match up the references +// before and after the labels in the reference stream. For this to be +// practical, the number of references has to be small, e.g. each label has +// exactly one reference. +// + +// Note: we also tried a completely different approach: random assignment +// followed by simulated annealing. This produced similar results. The results +// were not as good for very small differences because the simulated annealing +// never quite hit the groove. And simulated annealing was several orders of +// magnitude slower. + + +// TRIE node for suffix strings in the label reference sequence. +// +// We dynamically build a trie for both the program and model, growing the trie +// as necessary. The trie node for a (possibly) empty string of label +// references contains the distribution of labels following the string. The +// roots node (for the empty string) thus contains the simple distribution of +// labels within the label reference stream. + +struct Node { + Node(LabelInfo* in_edge, Node* prev) + : in_edge_(in_edge), prev_(prev), count_(0), + in_queue_(false) { + length_ = 1 + (prev_ ? prev_->length_ : 0); + } + LabelInfo* in_edge_; // + Node* prev_; // Node at shorter length. + int count_; // Frequency of this path in Trie. + int length_; + typedef std::map<LabelInfo*, Node*> Edges; + Edges edges_; + std::vector<int> places_; // Indexes into sequence of this item. + std::list<Node*> edges_in_frequency_order; + + bool in_queue_; + bool Extended() const { return edges_.size() > 0; } + + uint32 Weight() const { + return edges_in_frequency_order.front()->count_; + } +}; + +static std::string ToString(Node* node) { + std::vector<std::string> prefix; + for (Node* n = node; n->prev_; n = n->prev_) + prefix.push_back(ToString(n->in_edge_)); + + std::string s; + s += "{"; + const char* sep = ""; + while (!prefix.empty()) { + s += sep; + sep = ","; + s += prefix.back(); + prefix.pop_back(); + } + + s += StringPrintf("%u", node->count_); + s += " @"; + s += Uint64ToString(node->edges_in_frequency_order.size()); + s += "}"; + return s; +} + +typedef std::vector<LabelInfo*> Trace; + +struct OrderNodeByCountDecreasing { + bool operator()(Node* a, Node* b) const { + if (a->count_ != b->count_) + return (a->count_) > (b->count_); + return a->places_.at(0) < b->places_.at(0); // Prefer first occuring. + } +}; + +struct OrderNodeByWeightDecreasing { + bool operator()(Node* a, Node* b) const { + // (Maybe tie-break on total count, followed by lowest assigned node indexes + // in path.) + uint32 a_weight = a->Weight(); + uint32 b_weight = b->Weight(); + if (a_weight != b_weight) + return a_weight > b_weight; + if (a->length_ != b->length_) + return a->length_ > b->length_; // Prefer longer. + return a->places_.at(0) < b->places_.at(0); // Prefer first occuring. + } +}; + +typedef std::set<Node*, OrderNodeByWeightDecreasing> NodeQueue; + +class AssignmentProblem { + public: + AssignmentProblem(const Trace& model, + const Trace& problem) + : m_trace_(model), + p_trace_(problem) { + } + + ~AssignmentProblem() { + for (size_t i = 0; i < all_nodes_.size(); ++i) + delete all_nodes_[i]; + } + + bool Solve() { + m_root_ = MakeRootNode(m_trace_); + p_root_ = MakeRootNode(p_trace_); + AddToQueue(p_root_); + + while (!worklist_.empty()) { + Node* node = *worklist_.begin(); + node->in_queue_ = false; + worklist_.erase(node); + TrySolveNode(node); + } + + ALOG1 << unsolved_.size() << " unsolved items"; + return true; + } + + private: + void AddToQueue(Node* node) { + if (node->length_ >= 10) { + ALOG3 << "Length clipped " << ToString(node->prev_); + return; + } + if (node->in_queue_) { + LOG(ERROR) << "Double add " << ToString(node); + return; + } + // just to be sure data for prioritizing is available + ExtendNode(node, p_trace_); + // SkipCommittedLabels(node); + if (node->edges_in_frequency_order.empty()) + return; + node->in_queue_ = true; + worklist_.insert(node); + } + + void SkipCommittedLabels(Node* node) { + ExtendNode(node, p_trace_); + uint32 skipped = 0; + while (!node->edges_in_frequency_order.empty() && + node->edges_in_frequency_order.front()->in_edge_->assignment_) { + ++skipped; + node->edges_in_frequency_order.pop_front(); + } + if (skipped > 0) + ALOG3 << "Skipped " << skipped << " at " << ToString(node); + } + + void TrySolveNode(Node* p_node) { + Node* front = p_node->edges_in_frequency_order.front(); + if (front->in_edge_->assignment_) { + p_node->edges_in_frequency_order.pop_front(); + AddToQueue(front); + AddToQueue(p_node); + return; + } + + // Compare frequencies of unassigned edges, and either make + // assignment(s) or move node to unsolved list + + Node* m_node = FindModelNode(p_node); + + if (m_node == NULL) { + ALOG1 << "Can't find model node"; + unsolved_.insert(p_node); + return; + } + ExtendNode(m_node, m_trace_); + + // Lets just try greedy + + SkipCommittedLabels(m_node); + if (m_node->edges_in_frequency_order.empty()) { + ALOG3 << "Punting, no elements left in model vs " + << p_node->edges_in_frequency_order.size(); + unsolved_.insert(p_node); + return; + } + Node* m_match = m_node->edges_in_frequency_order.front(); + Node* p_match = p_node->edges_in_frequency_order.front(); + + if (p_match->count_ > 1.1 * m_match->count_ || + m_match->count_ > 1.1 * p_match->count_) { + ALOG2 << "Tricky distribution " + << p_match->count_ << ":" << m_match->count_ << " " + << ToString(p_match) << " vs " << ToString(m_match); + return; + } + + m_node->edges_in_frequency_order.pop_front(); + p_node->edges_in_frequency_order.pop_front(); + + LabelInfo* p_label_info = p_match->in_edge_; + LabelInfo* m_label_info = m_match->in_edge_; + int m_index = p_label_info->label_->index_; + if (m_index != Label::kNoIndex) { + ALOG1 << "Cant use unassigned label from model " << m_index; + unsolved_.insert(p_node); + return; + } + + Assign(p_label_info, m_label_info); + + AddToQueue(p_match); // find matches within new match + AddToQueue(p_node); // and more matches within this node + } + + void Assign(LabelInfo* p_info, LabelInfo* m_info) { + AssignOne(p_info, m_info); + ALOG3 << "Assign " << ToString(p_info) << " := " << ToString(m_info); + // Now consider unassigned adjacent addresses + TryExtendAssignment(p_info, m_info); + } + + void AssignOne(LabelInfo* p_info, LabelInfo* m_info) { + p_info->label_->index_ = m_info->label_->index_; + + // Mark as assigned + m_info->assignment_ = p_info; + p_info->assignment_ = m_info; + } + + void TryExtendAssignment(LabelInfo* p_info, LabelInfo* m_info) { + RVA m_rva_base = m_info->label_->rva_; + RVA p_rva_base = p_info->label_->rva_; + + LabelInfo* m_info_next = m_info->next_addr_; + LabelInfo* p_info_next = p_info->next_addr_; + for ( ; m_info_next && p_info_next; ) { + if (m_info_next->assignment_) + break; + + RVA m_rva = m_info_next->label_->rva_; + RVA p_rva = p_info_next->label_->rva_; + + if (m_rva - m_rva_base != p_rva - p_rva_base) { + // previous label was pointing to something that is different size + break; + } + LabelInfo* m_info_next_next = m_info_next->next_addr_; + LabelInfo* p_info_next_next = p_info_next->next_addr_; + if (m_info_next_next && p_info_next_next) { + RVA m_rva_next = m_info_next_next->label_->rva_; + RVA p_rva_next = p_info_next_next->label_->rva_; + if (m_rva_next - m_rva != p_rva_next - p_rva) { + // Since following labels are no longer in address lockstep, assume + // this address has a difference. + break; + } + } + + // The label has inconsistent numbers of references, it is probably not + // the same thing. + if (m_info_next->refs_ != p_info_next->refs_) { + break; + } + + ALOG3 << " Extending assignment -> " + << ToString(p_info_next) << " := " << ToString(m_info_next); + + AssignOne(p_info_next, m_info_next); + + if (p_info_next->refs_ == m_info_next->refs_ && + p_info_next->refs_ == 1) { + TryExtendSequence(p_info_next->positions_[0], + m_info_next->positions_[0]); + TryExtendSequenceBackwards(p_info_next->positions_[0], + m_info_next->positions_[0]); + } + + p_info_next = p_info_next_next; + m_info_next = m_info_next_next; + } + + LabelInfo* m_info_prev = m_info->prev_addr_; + LabelInfo* p_info_prev = p_info->prev_addr_; + for ( ; m_info_prev && p_info_prev; ) { + if (m_info_prev->assignment_) + break; + + RVA m_rva = m_info_prev->label_->rva_; + RVA p_rva = p_info_prev->label_->rva_; + + if (m_rva - m_rva_base != p_rva - p_rva_base) { + // previous label was pointing to something that is different size + break; + } + LabelInfo* m_info_prev_prev = m_info_prev->prev_addr_; + LabelInfo* p_info_prev_prev = p_info_prev->prev_addr_; + + // The the label has inconsistent numbers of references, it is + // probably not the same thing + if (m_info_prev->refs_ != p_info_prev->refs_) { + break; + } + + AssignOne(p_info_prev, m_info_prev); + ALOG3 << " Extending assignment <- " << ToString(p_info_prev) << " := " + << ToString(m_info_prev); + + p_info_prev = p_info_prev_prev; + m_info_prev = m_info_prev_prev; + } + } + + uint32 TryExtendSequence(uint32 p_pos_start, uint32 m_pos_start) { + uint32 p_pos = p_pos_start + 1; + uint32 m_pos = m_pos_start + 1; + + while (p_pos < p_trace_.size() && m_pos < m_trace_.size()) { + LabelInfo* p_info = p_trace_[p_pos]; + LabelInfo* m_info = m_trace_[m_pos]; + + // To match, either (1) both are assigned or (2) both are unassigned. + if ((p_info->assignment_ == NULL) != (m_info->assignment_ == NULL)) + break; + + // If they are assigned, it needs to be consistent (same index). + if (p_info->assignment_ && m_info->assignment_) { + if (p_info->label_->index_ != m_info->label_->index_) + break; + ++p_pos; + ++m_pos; + continue; + } + + if (p_info->refs_ != m_info->refs_) + break; + + AssignOne(p_info, m_info); + ALOG3 << " Extending assignment seq" + << "[+" << p_pos - p_pos_start << "]" + << " -> " + << ToString(p_info) << " := " << ToString(m_info); + + ++p_pos; + ++m_pos; + } + + return p_pos - p_pos_start; + } + + uint32 TryExtendSequenceBackwards(uint32 p_pos_start, uint32 m_pos_start) { + if (p_pos_start == 0 || m_pos_start == 0) + return 0; + + uint32 p_pos = p_pos_start - 1; + uint32 m_pos = m_pos_start - 1; + + while (p_pos > 0 && m_pos > 0) { + LabelInfo* p_info = p_trace_[p_pos]; + LabelInfo* m_info = m_trace_[m_pos]; + + if ((p_info->assignment_ == NULL) != (m_info->assignment_ == NULL)) + break; + + if (p_info->assignment_ && m_info->assignment_) { + if (p_info->label_->index_ != m_info->label_->index_) + break; + --p_pos; + --m_pos; + continue; + } + + if (p_info->refs_ != m_info->refs_) + break; + + AssignOne(p_info, m_info); + ALOG3 << " Extending assignment seq" + << "[-" << p_pos_start - p_pos << "]" + << " <- " + << ToString(p_info) << " := " << ToString(m_info); + + --p_pos; + --m_pos; + } + + return p_pos - p_pos_start; + } + + Node* FindModelNode(Node* node) { + if (node->prev_ == NULL) + return m_root_; + + Node* m_parent = FindModelNode(node->prev_); + if (m_parent == NULL) { + return NULL; + } + + ExtendNode(m_parent, m_trace_); + + LabelInfo* p_label = node->in_edge_; + LabelInfo* m_label = p_label->assignment_; + if (m_label == NULL) { + ALOG1 << "Expected assigned prefix"; + return NULL; + } + + Node::Edges::iterator e = m_parent->edges_.find(m_label); + if (e == m_parent->edges_.end()) { + ALOG2 << "Expected defined edge in parent"; + return NULL; + } + + return e->second; + } + + Node* MakeRootNode(const Trace& trace) { + Node* node = new Node(NULL, NULL); + all_nodes_.push_back(node); + for (size_t i = 0; i < trace.size(); ++i) { + ++node->count_; + node->places_.push_back(i); + } + return node; + } + + void ExtendNode(Node* node, const Trace& trace) { + // Make sure trie is filled in at this node. + if (node->Extended()) + return; + for (size_t i = 0; i < node->places_.size(); ++i) { + uint32 index = node->places_.at(i); + if (index < trace.size()) { + LabelInfo* item = trace.at(index); + Node*& slot = node->edges_[item]; + if (slot == NULL) { + slot = new Node(item, node); + all_nodes_.push_back(slot); + node->edges_in_frequency_order.push_back(slot); + } + slot->places_.push_back(index + 1); + ++slot->count_; + } + } + node->edges_in_frequency_order.sort(OrderNodeByCountDecreasing()); + } + + const Trace& m_trace_; + const Trace& p_trace_; + Node* m_root_; + Node* p_root_; + + NodeQueue worklist_; + NodeQueue unsolved_; + + std::vector<Node*> all_nodes_; + + DISALLOW_COPY_AND_ASSIGN(AssignmentProblem); +}; + +class GraphAdjuster : public AdjustmentMethod { + public: + GraphAdjuster() {} + ~GraphAdjuster() {} + + bool Adjust(const AssemblyProgram& model, AssemblyProgram* program) { + LOG(INFO) << "GraphAdjuster::Adjust"; + prog_ = program; + model_ = &model; + debug_label_index_gen_ = 0; + return Finish(); + } + + bool Finish() { + prog_->UnassignIndexes(); + CollectTraces(model_, &model_abs32_, &model_rel32_, true); + CollectTraces(prog_, &prog_abs32_, &prog_rel32_, false); + Solve(model_abs32_, prog_abs32_); + Solve(model_rel32_, prog_rel32_); + prog_->AssignRemainingIndexes(); + return true; + } + + private: + + void CollectTraces(const AssemblyProgram* program, Trace* abs32, Trace* rel32, + bool is_model) { + const std::vector<Instruction*>& instructions = program->instructions(); + for (size_t i = 0; i < instructions.size(); ++i) { + Instruction* instruction = instructions.at(i); + if (Label* label = program->InstructionAbs32Label(instruction)) + ReferenceLabel(abs32, label, is_model); + if (Label* label = program->InstructionRel32Label(instruction)) + ReferenceLabel(rel32, label, is_model); + } + // TODO(sra): we could simply append all the labels in index order to + // incorporate some costing for entropy (bigger deltas) that will be + // introduced into the label address table by non-monotonic ordering. This + // would have some knock-on effects to parts of the algorithm that work on + // single-occurence labels. + } + + void Solve(const Trace& model, const Trace& problem) { + LinkLabelInfos(model); + LinkLabelInfos(problem); + AssignmentProblem a(model, problem); + a.Solve(); + } + + void LinkLabelInfos(const Trace& trace) { + typedef std::set<LabelInfo*, OrderLabelInfoByAddressAscending> Ordered; + Ordered ordered; + for (Trace::const_iterator p = trace.begin(); p != trace.end(); ++p) + ordered.insert(*p); + LabelInfo* prev = NULL; + for (Ordered::iterator p = ordered.begin(); p != ordered.end(); ++p) { + LabelInfo* curr = *p; + if (prev) prev->next_addr_ = curr; + curr->prev_addr_ = prev; + prev = curr; + + if (curr->positions_.size() != curr->refs_) + NOTREACHED(); + } + } + + void ReferenceLabel(Trace* trace, Label* label, bool is_model) { + trace->push_back(MakeLabelInfo(label, is_model, trace->size())); + } + + LabelInfo* MakeLabelInfo(Label* label, bool is_model, uint32 position) { + LabelInfo& slot = label_infos_[label]; + if (slot.label_ == NULL) { + slot.label_ = label; + slot.is_model_ = is_model; + slot.debug_index_ = ++debug_label_index_gen_; + } + slot.positions_.push_back(position); + ++slot.refs_; + return &slot; + } + + AssemblyProgram* prog_; // Program to be adjusted, owned by caller. + const AssemblyProgram* model_; // Program to be mimicked, owned by caller. + + Trace model_abs32_; + Trace model_rel32_; + Trace prog_abs32_; + Trace prog_rel32_; + + int debug_label_index_gen_; + + // Note LabelInfo is allocated inside map, so the LabelInfo lifetimes are + // managed by the map. + std::map<Label*, LabelInfo> label_infos_; + + private: + DISALLOW_COPY_AND_ASSIGN(GraphAdjuster); +}; + + +//////////////////////////////////////////////////////////////////////////////// + +void AdjustmentMethod::Destroy() { delete this; } + +AdjustmentMethod* AdjustmentMethod::MakeNullAdjustmentMethod() { + return new NullAdjustmentMethod(); +} + +AdjustmentMethod* AdjustmentMethod::MakeProductionAdjustmentMethod() { + return new GraphAdjuster(); +} + +Status Adjust(const AssemblyProgram& model, AssemblyProgram* program) { + AdjustmentMethod* method = AdjustmentMethod::MakeProductionAdjustmentMethod(); + bool ok = method->Adjust(model, program); + method->Destroy(); + if (ok) + return C_OK; + else + return C_ADJUSTMENT_FAILED; +} + +} // namespace courgette diff --git a/third_party/courgette/adjustment_method.h b/courgette/adjustment_method.h index 2a5b0c7..8ca71e8 100644 --- a/third_party/courgette/adjustment_method.h +++ b/courgette/adjustment_method.h @@ -1,43 +1,42 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-#ifndef COURGETTE_ADJUSTMENT_METHOD_H_
-#define COURGETTE_ADJUSTMENT_METHOD_H_
-
-#include "base/basictypes.h"
-
-namespace courgette {
-
-class AssemblyProgram;
-
-class AdjustmentMethod {
- public:
- // Factory methods for making adjusters.
-
- // Returns the adjustment method used in production.
- static AdjustmentMethod* MakeProductionAdjustmentMethod();
-
- // Returns and adjustement method that makes no adjustments.
- static AdjustmentMethod* MakeNullAdjustmentMethod();
-
-
- // Adjusts |program| to increase similarity to |model|. |program| can be
- // changed in any way provided that it still produces the same output when
- // assembled.
- virtual bool Adjust(const AssemblyProgram& model,
- AssemblyProgram* program) = 0;
-
- // Deletes 'this' adjustment method.
- virtual void Destroy();
-
- protected:
- AdjustmentMethod() {}
- virtual ~AdjustmentMethod() {}
-
- DISALLOW_COPY_AND_ASSIGN(AdjustmentMethod);
-};
-
-} // namespace courgette
-#endif // COURGETTE_ADJUSTMENT_METHOD_H_
-
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +#ifndef COURGETTE_ADJUSTMENT_METHOD_H_ +#define COURGETTE_ADJUSTMENT_METHOD_H_ + +#include "base/basictypes.h" + +namespace courgette { + +class AssemblyProgram; + +class AdjustmentMethod { + public: + // Factory methods for making adjusters. + + // Returns the adjustment method used in production. + static AdjustmentMethod* MakeProductionAdjustmentMethod(); + + // Returns and adjustement method that makes no adjustments. + static AdjustmentMethod* MakeNullAdjustmentMethod(); + + + // Adjusts |program| to increase similarity to |model|. |program| can be + // changed in any way provided that it still produces the same output when + // assembled. + virtual bool Adjust(const AssemblyProgram& model, + AssemblyProgram* program) = 0; + + // Deletes 'this' adjustment method. + virtual void Destroy(); + + protected: + AdjustmentMethod() {} + virtual ~AdjustmentMethod() {} + + DISALLOW_COPY_AND_ASSIGN(AdjustmentMethod); +}; + +} // namespace courgette +#endif // COURGETTE_ADJUSTMENT_METHOD_H_ diff --git a/third_party/courgette/adjustment_method_unittest.cc b/courgette/adjustment_method_unittest.cc index 2048126..661b23d 100644 --- a/third_party/courgette/adjustment_method_unittest.cc +++ b/courgette/adjustment_method_unittest.cc @@ -1,109 +1,108 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-#include <string>
-
-#include "base/path_service.h"
-#include "base/file_util.h"
-#include "base/string_util.h"
-
-#include "third_party/courgette/assembly_program.h"
-#include "third_party/courgette/courgette.h"
-#include "third_party/courgette/streams.h"
-
-#include "testing/gtest/include/gtest/gtest.h"
-
-class AdjustmentMethodTest : public testing::Test {
- public:
- void Test1() const;
-
- private:
- void SetUp() {
- }
-
- void TearDown() {
- }
-
- // Returns one of two similar a simple programs. They differ only in the
- // label assignment, so that it is possible to make them look identical.
- courgette::AssemblyProgram* MakeProgram(int kind) const {
- courgette::AssemblyProgram* prog = new courgette::AssemblyProgram();
- prog->set_image_base(0x00400000);
-
- courgette::Label* labelA = prog->FindOrMakeAbs32Label(0x00410000);
- courgette::Label* labelB = prog->FindOrMakeAbs32Label(0x00410004);
-
- prog->EmitAbs32(labelA);
- prog->EmitAbs32(labelA);
- prog->EmitAbs32(labelB);
- prog->EmitAbs32(labelA);
- prog->EmitAbs32(labelA);
- prog->EmitAbs32(labelB);
-
- if (kind == 0) {
- labelA->index_ = 0;
- labelB->index_ = 1;
- } else {
- labelA->index_ = 1;
- labelB->index_ = 0;
- }
- prog->AssignRemainingIndexes();
-
- return prog;
- }
-
- courgette::AssemblyProgram* MakeProgramA() const { return MakeProgram(0); }
- courgette::AssemblyProgram* MakeProgramB() const { return MakeProgram(1); }
-
- // Returns a string that is the serialized version of |program|.
- // Deletes |program|.
- std::string Serialize(courgette::AssemblyProgram *program) const {
- courgette::EncodedProgram* encoded = NULL;
-
- const courgette::Status encode_status = Encode(program, &encoded);
- EXPECT_EQ(courgette::C_OK, encode_status);
-
- DeleteAssemblyProgram(program);
-
- courgette::SinkStreamSet sinks;
- const courgette::Status write_status = WriteEncodedProgram(encoded, &sinks);
- EXPECT_EQ(courgette::C_OK, write_status);
-
- DeleteEncodedProgram(encoded);
-
- courgette::SinkStream sink;
- bool can_collect = sinks.CopyTo(&sink);
- EXPECT_TRUE(can_collect);
-
- return std::string(reinterpret_cast<const char *>(sink.Buffer()),
- sink.Length());
- }
-};
-
-
-void AdjustmentMethodTest::Test1() const {
- courgette::AssemblyProgram* prog1 = MakeProgramA();
- courgette::AssemblyProgram* prog2 = MakeProgramB();
- std::string s1 = Serialize(prog1);
- std::string s2 = Serialize(prog2);
-
- // Don't use EXPECT_EQ because strings are unprintable.
- EXPECT_FALSE(s1 == s2); // Unadjusted A and B differ.
-
- courgette::AssemblyProgram* prog5 = MakeProgramA();
- courgette::AssemblyProgram* prog6 = MakeProgramB();
- courgette::Status can_adjust = Adjust(*prog5, prog6);
- EXPECT_EQ(courgette::C_OK, can_adjust);
- std::string s5 = Serialize(prog5);
- std::string s6 = Serialize(prog6);
-
- EXPECT_TRUE(s1 == s5); // Adjustment did not change A (prog5)
- EXPECT_TRUE(s5 == s6); // Adjustment did change B into A
-}
-
-
-TEST_F(AdjustmentMethodTest, All) {
- Test1();
-}
-
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +#include <string> + +#include "base/path_service.h" +#include "base/file_util.h" +#include "base/string_util.h" + +#include "courgette/assembly_program.h" +#include "courgette/courgette.h" +#include "courgette/streams.h" + +#include "testing/gtest/include/gtest/gtest.h" + +class AdjustmentMethodTest : public testing::Test { + public: + void Test1() const; + + private: + void SetUp() { + } + + void TearDown() { + } + + // Returns one of two similar a simple programs. They differ only in the + // label assignment, so that it is possible to make them look identical. + courgette::AssemblyProgram* MakeProgram(int kind) const { + courgette::AssemblyProgram* prog = new courgette::AssemblyProgram(); + prog->set_image_base(0x00400000); + + courgette::Label* labelA = prog->FindOrMakeAbs32Label(0x00410000); + courgette::Label* labelB = prog->FindOrMakeAbs32Label(0x00410004); + + prog->EmitAbs32(labelA); + prog->EmitAbs32(labelA); + prog->EmitAbs32(labelB); + prog->EmitAbs32(labelA); + prog->EmitAbs32(labelA); + prog->EmitAbs32(labelB); + + if (kind == 0) { + labelA->index_ = 0; + labelB->index_ = 1; + } else { + labelA->index_ = 1; + labelB->index_ = 0; + } + prog->AssignRemainingIndexes(); + + return prog; + } + + courgette::AssemblyProgram* MakeProgramA() const { return MakeProgram(0); } + courgette::AssemblyProgram* MakeProgramB() const { return MakeProgram(1); } + + // Returns a string that is the serialized version of |program|. + // Deletes |program|. + std::string Serialize(courgette::AssemblyProgram *program) const { + courgette::EncodedProgram* encoded = NULL; + + const courgette::Status encode_status = Encode(program, &encoded); + EXPECT_EQ(courgette::C_OK, encode_status); + + DeleteAssemblyProgram(program); + + courgette::SinkStreamSet sinks; + const courgette::Status write_status = WriteEncodedProgram(encoded, &sinks); + EXPECT_EQ(courgette::C_OK, write_status); + + DeleteEncodedProgram(encoded); + + courgette::SinkStream sink; + bool can_collect = sinks.CopyTo(&sink); + EXPECT_TRUE(can_collect); + + return std::string(reinterpret_cast<const char *>(sink.Buffer()), + sink.Length()); + } +}; + + +void AdjustmentMethodTest::Test1() const { + courgette::AssemblyProgram* prog1 = MakeProgramA(); + courgette::AssemblyProgram* prog2 = MakeProgramB(); + std::string s1 = Serialize(prog1); + std::string s2 = Serialize(prog2); + + // Don't use EXPECT_EQ because strings are unprintable. + EXPECT_FALSE(s1 == s2); // Unadjusted A and B differ. + + courgette::AssemblyProgram* prog5 = MakeProgramA(); + courgette::AssemblyProgram* prog6 = MakeProgramB(); + courgette::Status can_adjust = Adjust(*prog5, prog6); + EXPECT_EQ(courgette::C_OK, can_adjust); + std::string s5 = Serialize(prog5); + std::string s6 = Serialize(prog6); + + EXPECT_TRUE(s1 == s5); // Adjustment did not change A (prog5) + EXPECT_TRUE(s5 == s6); // Adjustment did change B into A +} + + +TEST_F(AdjustmentMethodTest, All) { + Test1(); +} diff --git a/third_party/courgette/assembly_program.cc b/courgette/assembly_program.cc index 178dfdf..e40f38f 100644 --- a/third_party/courgette/assembly_program.cc +++ b/courgette/assembly_program.cc @@ -1,373 +1,372 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-#include "third_party/courgette/assembly_program.h"
-
-#include <memory.h>
-#include <algorithm>
-#include <map>
-#include <set>
-#include <sstream>
-#include <vector>
-
-#include "base/logging.h"
-
-#include "third_party/courgette/courgette.h"
-#include "third_party/courgette/encoded_program.h"
-
-namespace courgette {
-
-// Opcodes of simple assembly language
-enum OP {
- ORIGIN, // ORIGIN <rva> - set current address for assembly.
- MAKERELOCS, // Generates a base relocation table.
- DEFBYTE, // DEFBYTE <value> - emit a byte literal.
- REL32, // REL32 <label> - emit a rel32 encoded reference to 'label'.
- ABS32, // REL32 <label> - emit am abs32 encoded reference to 'label'.
- LAST_OP
-};
-
-// Base class for instructions. Because we have so many instructions we want to
-// keep them as small as possible. For this reason we avoid virtual functions.
-//
-class Instruction {
- public:
- OP op() const { return static_cast<OP>(op_); }
-
- protected:
- explicit Instruction(OP op) : op_(op), info_(0) {}
- Instruction(OP op, unsigned int info) : op_(op), info_(info) {}
-
- uint32 op_ : 4; // A few bits to store the OP code.
- uint32 info_ : 28; // Remaining bits in first word available to subclass.
-
- private:
- DISALLOW_COPY_AND_ASSIGN(Instruction);
-};
-
-namespace {
-
-// Sets the current address for the emitting instructions.
-class OriginInstruction : public Instruction {
- public:
- explicit OriginInstruction(RVA rva) : Instruction(ORIGIN, 0), rva_(rva) {}
- RVA origin_rva() const { return rva_; }
- private:
- RVA rva_;
-};
-
-// Emits an entire base relocation table.
-class MakeRelocsInstruction : public Instruction {
- public:
- MakeRelocsInstruction() : Instruction(MAKERELOCS) {}
-};
-
-// Emits a single byte.
-class ByteInstruction : public Instruction {
- public:
- explicit ByteInstruction(uint8 value) : Instruction(DEFBYTE, value) {}
- uint8 byte_value() const { return info_; }
-};
-
-// A ABS32 to REL32 instruction emits a reference to a label's address.
-class InstructionWithLabel : public Instruction {
- public:
- InstructionWithLabel(OP op, Label* label)
- : Instruction(op, 0), label_(label) {
- if (label == NULL) NOTREACHED();
- }
- Label* label() const { return label_; }
- private:
- Label* label_;
-};
-
-} // namespace
-
-AssemblyProgram::AssemblyProgram()
- : image_base_(0),
- byte_instruction_cache_(NULL) {
-}
-
-static void DeleteContainedLabels(const RVAToLabel& labels) {
- for (RVAToLabel::const_iterator p = labels.begin(); p != labels.end(); ++p)
- delete p->second;
-}
-
-AssemblyProgram::~AssemblyProgram() {
- for (size_t i = 0; i < instructions_.size(); ++i) {
- Instruction* instruction = instructions_[i];
- if (instruction->op() != DEFBYTE) // Will be in byte_instruction_cache_.
- delete instruction;
- }
- if (byte_instruction_cache_) {
- for (size_t i = 0; i < 256; ++i)
- delete byte_instruction_cache_[i];
- delete[] byte_instruction_cache_;
- }
- DeleteContainedLabels(rel32_labels_);
- DeleteContainedLabels(abs32_labels_);
-}
-
-void AssemblyProgram::EmitMakeRelocsInstruction() {
- Emit(new MakeRelocsInstruction());
-}
-
-void AssemblyProgram::EmitOriginInstruction(RVA rva) {
- Emit(new OriginInstruction(rva));
-}
-
-void AssemblyProgram::EmitByteInstruction(uint8 byte) {
- Emit(GetByteInstruction(byte));
-}
-
-void AssemblyProgram::EmitRel32(Label* label) {
- Emit(new InstructionWithLabel(REL32, label));
-}
-
-void AssemblyProgram::EmitAbs32(Label* label) {
- Emit(new InstructionWithLabel(ABS32, label));
-}
-
-Label* AssemblyProgram::FindOrMakeAbs32Label(RVA rva) {
- return FindLabel(rva, &abs32_labels_);
-}
-
-Label* AssemblyProgram::FindOrMakeRel32Label(RVA rva) {
- return FindLabel(rva, &rel32_labels_);
-}
-
-void AssemblyProgram::DefaultAssignIndexes() {
- DefaultAssignIndexes(&abs32_labels_);
- DefaultAssignIndexes(&rel32_labels_);
-}
-
-void AssemblyProgram::UnassignIndexes() {
- UnassignIndexes(&abs32_labels_);
- UnassignIndexes(&rel32_labels_);
-}
-
-void AssemblyProgram::AssignRemainingIndexes() {
- AssignRemainingIndexes(&abs32_labels_);
- AssignRemainingIndexes(&rel32_labels_);
-}
-
-Label* AssemblyProgram::InstructionAbs32Label(
- const Instruction* instruction) const {
- if (instruction->op() == ABS32)
- return static_cast<const InstructionWithLabel*>(instruction)->label();
- return NULL;
-}
-
-Label* AssemblyProgram::InstructionRel32Label(
- const Instruction* instruction) const {
- if (instruction->op() == REL32)
- return static_cast<const InstructionWithLabel*>(instruction)->label();
- return NULL;
-}
-
-Label* AssemblyProgram::FindLabel(RVA rva, RVAToLabel* labels) {
- Label*& slot = (*labels)[rva];
- if (slot == 0) {
- slot = new Label(rva);
- }
- return slot;
-}
-
-void AssemblyProgram::UnassignIndexes(RVAToLabel* labels) {
- for (RVAToLabel::iterator p = labels->begin(); p != labels->end(); ++p) {
- Label* current = p->second;
- current->index_ = Label::kNoIndex;
- }
-}
-
-// DefaultAssignIndexes takes a set of labels and assigns indexes in increasing
-// address order.
-//
-void AssemblyProgram::DefaultAssignIndexes(RVAToLabel* labels) {
- int index = 0;
- for (RVAToLabel::iterator p = labels->begin(); p != labels->end(); ++p) {
- Label* current = p->second;
- if (current->index_ != Label::kNoIndex)
- NOTREACHED();
- current->index_ = index;
- ++index;
- }
-}
-
-// AssignRemainingIndexes assigns indexes to any addresses (labels) that are not
-// yet assigned an index.
-//
-void AssemblyProgram::AssignRemainingIndexes(RVAToLabel* labels) {
- // An address table compresses best when each index is associated with an
- // address that is slight larger than the previous index.
-
- // First see which indexes have not been used. The 'available' vector could
- // grow even bigger, but the number of addresses is a better starting size
- // than empty.
- std::vector<bool> available(labels->size(), true);
- int used = 0;
-
- for (RVAToLabel::iterator p = labels->begin(); p != labels->end(); ++p) {
- size_t index = p->second->index_;
- if (index != Label::kNoIndex) {
- while (index >= available.size())
- available.push_back(true);
- available.at(index) = false;
- ++used;
- }
- }
-
- LOG(INFO) << used << " of " << labels->size() << " labels pre-assigned";
-
- // Are there any unused labels that happen to be adjacent following a used
- // label?
- //
- int fill_forward_count = 0;
- Label* prev = 0;
- for (RVAToLabel::iterator p = labels->begin(); p != labels->end(); ++p) {
- Label* current = p->second;
- if (current->index_ == Label::kNoIndex) {
- size_t index = 0;
- if (prev && prev->index_ != Label::kNoIndex)
- index = prev->index_ + 1;
- if (index < available.size() && available.at(index)) {
- current->index_ = index;
- available.at(index) = false;
- ++fill_forward_count;
- }
- }
- prev = current;
- }
-
- // Are there any unused labels that happen to be adjacent preceeding a used
- // label?
- //
- int fill_backward_count = 0;
- int backward_refs = 0;
- prev = 0;
- for (RVAToLabel::reverse_iterator p = labels->rbegin();
- p != labels->rend();
- ++p) {
- Label* current = p->second;
- if (current->index_ == Label::kNoIndex) {
- int prev_index;
- if (prev)
- prev_index = prev->index_;
- else
- prev_index = available.size();
- if (prev_index != 0 &&
- prev_index != Label::kNoIndex &&
- available.at(prev_index - 1)) {
- current->index_ = prev_index - 1;
- available.at(current->index_) = false;
- ++fill_backward_count;
- }
- }
- prev = current;
- }
-
- // Fill in any remaining indexes
- int fill_infill_count = 0;
- int index = 0;
- for (RVAToLabel::iterator p = labels->begin(); p != labels->end(); ++p) {
- Label* current = p->second;
- if (current->index_ == Label::kNoIndex) {
- while (!available.at(index)) {
- ++index;
- }
- current->index_ = index;
- available.at(index) = false;
- ++index;
- ++fill_infill_count;
- }
- }
-
- LOG(INFO) << " fill"
- << " forward " << fill_forward_count << " "
- << " backward " << fill_backward_count << " "
- << " infill " << fill_infill_count;
-}
-
-typedef void (EncodedProgram::*DefineLabelMethod)(int index, RVA value);
-
-static void DefineLabels(const RVAToLabel& labels,
- EncodedProgram* encoded_format,
- DefineLabelMethod define_label) {
- for (RVAToLabel::const_iterator p = labels.begin(); p != labels.end(); ++p) {
- Label* label = p->second;
- (encoded_format->*define_label)(label->index_, label->rva_);
- }
-}
-
-EncodedProgram* AssemblyProgram::Encode() const {
- EncodedProgram* encoded = new EncodedProgram();
-
- encoded->set_image_base(image_base_);
- DefineLabels(abs32_labels_, encoded, &EncodedProgram::DefineAbs32Label);
- DefineLabels(rel32_labels_, encoded, &EncodedProgram::DefineRel32Label);
- encoded->EndLabels();
-
- for (size_t i = 0; i < instructions_.size(); ++i) {
- Instruction* instruction = instructions_[i];
-
- switch (instruction->op()) {
- case ORIGIN: {
- OriginInstruction* org = static_cast<OriginInstruction*>(instruction);
- encoded->AddOrigin(org->origin_rva());
- break;
- }
- case DEFBYTE: {
- uint8 b = static_cast<ByteInstruction*>(instruction)->byte_value();
- encoded->AddCopy(1, &b);
- break;
- }
- case REL32: {
- Label* label = static_cast<InstructionWithLabel*>(instruction)->label();
- encoded->AddRel32(label->index_);
- break;
- }
- case ABS32: {
- Label* label = static_cast<InstructionWithLabel*>(instruction)->label();
- encoded->AddAbs32(label->index_);
- break;
- }
- case MAKERELOCS: {
- encoded->AddMakeRelocs();
- break;
- }
- default: {
- NOTREACHED() << "Unknown Insn OP kind";
- }
- }
- }
-
- return encoded;
-}
-
-Instruction* AssemblyProgram::GetByteInstruction(uint8 byte) {
- if (!byte_instruction_cache_) {
- byte_instruction_cache_ = new Instruction*[256];
- for (int i = 0; i < 256; ++i) {
- byte_instruction_cache_[i] = new ByteInstruction(static_cast<uint8>(i));
- }
- }
-
- return byte_instruction_cache_[byte];
-}
-
-////////////////////////////////////////////////////////////////////////////////
-
-Status Encode(AssemblyProgram* program, EncodedProgram** output) {
- *output = NULL;
- EncodedProgram *encoded = program->Encode();
- if (encoded) {
- *output = encoded;
- return C_OK;
- } else {
- return C_GENERAL_ERROR;
- }
-}
-
-} // namespace courgette
-
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +#include "courgette/assembly_program.h" + +#include <memory.h> +#include <algorithm> +#include <map> +#include <set> +#include <sstream> +#include <vector> + +#include "base/logging.h" + +#include "courgette/courgette.h" +#include "courgette/encoded_program.h" + +namespace courgette { + +// Opcodes of simple assembly language +enum OP { + ORIGIN, // ORIGIN <rva> - set current address for assembly. + MAKERELOCS, // Generates a base relocation table. + DEFBYTE, // DEFBYTE <value> - emit a byte literal. + REL32, // REL32 <label> - emit a rel32 encoded reference to 'label'. + ABS32, // REL32 <label> - emit am abs32 encoded reference to 'label'. + LAST_OP +}; + +// Base class for instructions. Because we have so many instructions we want to +// keep them as small as possible. For this reason we avoid virtual functions. +// +class Instruction { + public: + OP op() const { return static_cast<OP>(op_); } + + protected: + explicit Instruction(OP op) : op_(op), info_(0) {} + Instruction(OP op, unsigned int info) : op_(op), info_(info) {} + + uint32 op_ : 4; // A few bits to store the OP code. + uint32 info_ : 28; // Remaining bits in first word available to subclass. + + private: + DISALLOW_COPY_AND_ASSIGN(Instruction); +}; + +namespace { + +// Sets the current address for the emitting instructions. +class OriginInstruction : public Instruction { + public: + explicit OriginInstruction(RVA rva) : Instruction(ORIGIN, 0), rva_(rva) {} + RVA origin_rva() const { return rva_; } + private: + RVA rva_; +}; + +// Emits an entire base relocation table. +class MakeRelocsInstruction : public Instruction { + public: + MakeRelocsInstruction() : Instruction(MAKERELOCS) {} +}; + +// Emits a single byte. +class ByteInstruction : public Instruction { + public: + explicit ByteInstruction(uint8 value) : Instruction(DEFBYTE, value) {} + uint8 byte_value() const { return info_; } +}; + +// A ABS32 to REL32 instruction emits a reference to a label's address. +class InstructionWithLabel : public Instruction { + public: + InstructionWithLabel(OP op, Label* label) + : Instruction(op, 0), label_(label) { + if (label == NULL) NOTREACHED(); + } + Label* label() const { return label_; } + private: + Label* label_; +}; + +} // namespace + +AssemblyProgram::AssemblyProgram() + : image_base_(0), + byte_instruction_cache_(NULL) { +} + +static void DeleteContainedLabels(const RVAToLabel& labels) { + for (RVAToLabel::const_iterator p = labels.begin(); p != labels.end(); ++p) + delete p->second; +} + +AssemblyProgram::~AssemblyProgram() { + for (size_t i = 0; i < instructions_.size(); ++i) { + Instruction* instruction = instructions_[i]; + if (instruction->op() != DEFBYTE) // Will be in byte_instruction_cache_. + delete instruction; + } + if (byte_instruction_cache_) { + for (size_t i = 0; i < 256; ++i) + delete byte_instruction_cache_[i]; + delete[] byte_instruction_cache_; + } + DeleteContainedLabels(rel32_labels_); + DeleteContainedLabels(abs32_labels_); +} + +void AssemblyProgram::EmitMakeRelocsInstruction() { + Emit(new MakeRelocsInstruction()); +} + +void AssemblyProgram::EmitOriginInstruction(RVA rva) { + Emit(new OriginInstruction(rva)); +} + +void AssemblyProgram::EmitByteInstruction(uint8 byte) { + Emit(GetByteInstruction(byte)); +} + +void AssemblyProgram::EmitRel32(Label* label) { + Emit(new InstructionWithLabel(REL32, label)); +} + +void AssemblyProgram::EmitAbs32(Label* label) { + Emit(new InstructionWithLabel(ABS32, label)); +} + +Label* AssemblyProgram::FindOrMakeAbs32Label(RVA rva) { + return FindLabel(rva, &abs32_labels_); +} + +Label* AssemblyProgram::FindOrMakeRel32Label(RVA rva) { + return FindLabel(rva, &rel32_labels_); +} + +void AssemblyProgram::DefaultAssignIndexes() { + DefaultAssignIndexes(&abs32_labels_); + DefaultAssignIndexes(&rel32_labels_); +} + +void AssemblyProgram::UnassignIndexes() { + UnassignIndexes(&abs32_labels_); + UnassignIndexes(&rel32_labels_); +} + +void AssemblyProgram::AssignRemainingIndexes() { + AssignRemainingIndexes(&abs32_labels_); + AssignRemainingIndexes(&rel32_labels_); +} + +Label* AssemblyProgram::InstructionAbs32Label( + const Instruction* instruction) const { + if (instruction->op() == ABS32) + return static_cast<const InstructionWithLabel*>(instruction)->label(); + return NULL; +} + +Label* AssemblyProgram::InstructionRel32Label( + const Instruction* instruction) const { + if (instruction->op() == REL32) + return static_cast<const InstructionWithLabel*>(instruction)->label(); + return NULL; +} + +Label* AssemblyProgram::FindLabel(RVA rva, RVAToLabel* labels) { + Label*& slot = (*labels)[rva]; + if (slot == 0) { + slot = new Label(rva); + } + return slot; +} + +void AssemblyProgram::UnassignIndexes(RVAToLabel* labels) { + for (RVAToLabel::iterator p = labels->begin(); p != labels->end(); ++p) { + Label* current = p->second; + current->index_ = Label::kNoIndex; + } +} + +// DefaultAssignIndexes takes a set of labels and assigns indexes in increasing +// address order. +// +void AssemblyProgram::DefaultAssignIndexes(RVAToLabel* labels) { + int index = 0; + for (RVAToLabel::iterator p = labels->begin(); p != labels->end(); ++p) { + Label* current = p->second; + if (current->index_ != Label::kNoIndex) + NOTREACHED(); + current->index_ = index; + ++index; + } +} + +// AssignRemainingIndexes assigns indexes to any addresses (labels) that are not +// yet assigned an index. +// +void AssemblyProgram::AssignRemainingIndexes(RVAToLabel* labels) { + // An address table compresses best when each index is associated with an + // address that is slight larger than the previous index. + + // First see which indexes have not been used. The 'available' vector could + // grow even bigger, but the number of addresses is a better starting size + // than empty. + std::vector<bool> available(labels->size(), true); + int used = 0; + + for (RVAToLabel::iterator p = labels->begin(); p != labels->end(); ++p) { + size_t index = p->second->index_; + if (index != Label::kNoIndex) { + while (index >= available.size()) + available.push_back(true); + available.at(index) = false; + ++used; + } + } + + LOG(INFO) << used << " of " << labels->size() << " labels pre-assigned"; + + // Are there any unused labels that happen to be adjacent following a used + // label? + // + int fill_forward_count = 0; + Label* prev = 0; + for (RVAToLabel::iterator p = labels->begin(); p != labels->end(); ++p) { + Label* current = p->second; + if (current->index_ == Label::kNoIndex) { + size_t index = 0; + if (prev && prev->index_ != Label::kNoIndex) + index = prev->index_ + 1; + if (index < available.size() && available.at(index)) { + current->index_ = index; + available.at(index) = false; + ++fill_forward_count; + } + } + prev = current; + } + + // Are there any unused labels that happen to be adjacent preceeding a used + // label? + // + int fill_backward_count = 0; + int backward_refs = 0; + prev = 0; + for (RVAToLabel::reverse_iterator p = labels->rbegin(); + p != labels->rend(); + ++p) { + Label* current = p->second; + if (current->index_ == Label::kNoIndex) { + int prev_index; + if (prev) + prev_index = prev->index_; + else + prev_index = available.size(); + if (prev_index != 0 && + prev_index != Label::kNoIndex && + available.at(prev_index - 1)) { + current->index_ = prev_index - 1; + available.at(current->index_) = false; + ++fill_backward_count; + } + } + prev = current; + } + + // Fill in any remaining indexes + int fill_infill_count = 0; + int index = 0; + for (RVAToLabel::iterator p = labels->begin(); p != labels->end(); ++p) { + Label* current = p->second; + if (current->index_ == Label::kNoIndex) { + while (!available.at(index)) { + ++index; + } + current->index_ = index; + available.at(index) = false; + ++index; + ++fill_infill_count; + } + } + + LOG(INFO) << " fill" + << " forward " << fill_forward_count << " " + << " backward " << fill_backward_count << " " + << " infill " << fill_infill_count; +} + +typedef void (EncodedProgram::*DefineLabelMethod)(int index, RVA value); + +static void DefineLabels(const RVAToLabel& labels, + EncodedProgram* encoded_format, + DefineLabelMethod define_label) { + for (RVAToLabel::const_iterator p = labels.begin(); p != labels.end(); ++p) { + Label* label = p->second; + (encoded_format->*define_label)(label->index_, label->rva_); + } +} + +EncodedProgram* AssemblyProgram::Encode() const { + EncodedProgram* encoded = new EncodedProgram(); + + encoded->set_image_base(image_base_); + DefineLabels(abs32_labels_, encoded, &EncodedProgram::DefineAbs32Label); + DefineLabels(rel32_labels_, encoded, &EncodedProgram::DefineRel32Label); + encoded->EndLabels(); + + for (size_t i = 0; i < instructions_.size(); ++i) { + Instruction* instruction = instructions_[i]; + + switch (instruction->op()) { + case ORIGIN: { + OriginInstruction* org = static_cast<OriginInstruction*>(instruction); + encoded->AddOrigin(org->origin_rva()); + break; + } + case DEFBYTE: { + uint8 b = static_cast<ByteInstruction*>(instruction)->byte_value(); + encoded->AddCopy(1, &b); + break; + } + case REL32: { + Label* label = static_cast<InstructionWithLabel*>(instruction)->label(); + encoded->AddRel32(label->index_); + break; + } + case ABS32: { + Label* label = static_cast<InstructionWithLabel*>(instruction)->label(); + encoded->AddAbs32(label->index_); + break; + } + case MAKERELOCS: { + encoded->AddMakeRelocs(); + break; + } + default: { + NOTREACHED() << "Unknown Insn OP kind"; + } + } + } + + return encoded; +} + +Instruction* AssemblyProgram::GetByteInstruction(uint8 byte) { + if (!byte_instruction_cache_) { + byte_instruction_cache_ = new Instruction*[256]; + for (int i = 0; i < 256; ++i) { + byte_instruction_cache_[i] = new ByteInstruction(static_cast<uint8>(i)); + } + } + + return byte_instruction_cache_[byte]; +} + +//////////////////////////////////////////////////////////////////////////////// + +Status Encode(AssemblyProgram* program, EncodedProgram** output) { + *output = NULL; + EncodedProgram *encoded = program->Encode(); + if (encoded) { + *output = encoded; + return C_OK; + } else { + return C_GENERAL_ERROR; + } +} + +} // namespace courgette diff --git a/third_party/courgette/assembly_program.h b/courgette/assembly_program.h index 4b48133..e726f81 100644 --- a/third_party/courgette/assembly_program.h +++ b/courgette/assembly_program.h @@ -1,134 +1,133 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-#ifndef COURGETTE_ASSEMBLY_PROGRAM_H_
-#define COURGETTE_ASSEMBLY_PROGRAM_H_
-
-#include <map>
-#include <set>
-#include <vector>
-
-#include "base/basictypes.h"
-
-#include "third_party/courgette/image_info.h"
-
-namespace courgette {
-
-class EncodedProgram;
-class Instruction;
-
-// A Label is a symbolic reference to an address. Unlike a conventional
-// assembly language, we always know the address. The address will later be
-// stored in a table and the Label will be replaced with the index into the
-// table.
-//
-// TODO(sra): Make fields private and add setters and getters.
-class Label {
- public:
- static const int kNoIndex = -1;
- Label() : rva_(0), index_(kNoIndex) {}
- explicit Label(RVA rva) : rva_(rva), index_(kNoIndex) {}
-
- RVA rva_; // Address refered to by the label.
- int index_; // Index of address in address table, kNoIndex until assigned.
-};
-
-typedef std::map<RVA, Label*> RVAToLabel;
-
-// An AssemblyProgram is the result of disassembling an executable file.
-//
-// * The disassembler creates labels in the AssemblyProgram and emits
-// 'Instructions'.
-// * The disassembler then calls DefaultAssignIndexes to assign
-// addresses to positions in the address tables.
-// * [Optional step]
-// * At this point the AssemblyProgram can be converted into an
-// EncodedProgram and serialized to an output stream.
-// * Later, the EncodedProgram can be deserialized and assembled into
-// the original file.
-//
-// The optional step is to modify the AssemblyProgram. One form of modification
-// is to assign indexes in such a way as to make the EncodedProgram for this
-// AssemblyProgram look more like the EncodedProgram for some other
-// AssemblyProgram. The modification process should call UnassignIndexes, do
-// its own assignment, and then call AssignRemainingIndexes to ensure all
-// indexes are assigned.
-//
-class AssemblyProgram {
- public:
- AssemblyProgram();
- ~AssemblyProgram();
-
- void set_image_base(uint64 image_base) { image_base_ = image_base; }
-
- // Instructions will be assembled in the order they are emitted.
-
- // Generates an entire base relocation table.
- void EmitMakeRelocsInstruction();
-
- // Following instruction will be assembled at address 'rva'.
- void EmitOriginInstruction(RVA rva);
-
- // Generates a single byte of data or machine instruction.
- void EmitByteInstruction(uint8 byte);
-
- // Generates 4-byte relative reference to address of 'label'.
- void EmitRel32(Label* label);
-
- // Generates 4-byte absolute reference to address of 'label'.
- void EmitAbs32(Label* label);
-
- Label* FindOrMakeAbs32Label(RVA rva);
- Label* FindOrMakeRel32Label(RVA rva);
-
- void DefaultAssignIndexes();
- void UnassignIndexes();
- void AssignRemainingIndexes();
-
- EncodedProgram* Encode() const;
-
- // Accessor for instruction list.
- const std::vector<Instruction*>& instructions() const {
- return instructions_;
- }
-
- // Returns the label if the instruction contains and absolute address,
- // otherwise returns NULL.
- Label* InstructionAbs32Label(const Instruction* instruction) const;
-
- // Returns the label if the instruction contains and rel32 offset,
- // otherwise returns NULL.
- Label* InstructionRel32Label(const Instruction* instruction) const;
-
-
- private:
- void Emit(Instruction* instruction) { instructions_.push_back(instruction); }
-
- Label* FindLabel(RVA rva, RVAToLabel* labels);
-
- // Helper methods for the public versions.
- static void UnassignIndexes(RVAToLabel* labels);
- static void DefaultAssignIndexes(RVAToLabel* labels);
- static void AssignRemainingIndexes(RVAToLabel* labels);
-
- // Sharing instructions that emit a single byte saves a lot of space.
- Instruction* GetByteInstruction(uint8 byte);
- Instruction** byte_instruction_cache_;
-
- uint64 image_base_; // Desired or mandated base address of image.
-
- std::vector<Instruction*> instructions_; // All the instructions in program.
-
- // These are lookup maps to find the label associated with a given address.
- // We have separate label spaces for addresses referenced by rel32 labels and
- // abs32 labels. This is somewhat arbitrary.
- RVAToLabel rel32_labels_;
- RVAToLabel abs32_labels_;
-
- DISALLOW_COPY_AND_ASSIGN(AssemblyProgram);
-};
-
-} // namespace courgette
-#endif // COURGETTE_ASSEMBLY_PROGRAM_H_
-
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +#ifndef COURGETTE_ASSEMBLY_PROGRAM_H_ +#define COURGETTE_ASSEMBLY_PROGRAM_H_ + +#include <map> +#include <set> +#include <vector> + +#include "base/basictypes.h" + +#include "courgette/image_info.h" + +namespace courgette { + +class EncodedProgram; +class Instruction; + +// A Label is a symbolic reference to an address. Unlike a conventional +// assembly language, we always know the address. The address will later be +// stored in a table and the Label will be replaced with the index into the +// table. +// +// TODO(sra): Make fields private and add setters and getters. +class Label { + public: + static const int kNoIndex = -1; + Label() : rva_(0), index_(kNoIndex) {} + explicit Label(RVA rva) : rva_(rva), index_(kNoIndex) {} + + RVA rva_; // Address refered to by the label. + int index_; // Index of address in address table, kNoIndex until assigned. +}; + +typedef std::map<RVA, Label*> RVAToLabel; + +// An AssemblyProgram is the result of disassembling an executable file. +// +// * The disassembler creates labels in the AssemblyProgram and emits +// 'Instructions'. +// * The disassembler then calls DefaultAssignIndexes to assign +// addresses to positions in the address tables. +// * [Optional step] +// * At this point the AssemblyProgram can be converted into an +// EncodedProgram and serialized to an output stream. +// * Later, the EncodedProgram can be deserialized and assembled into +// the original file. +// +// The optional step is to modify the AssemblyProgram. One form of modification +// is to assign indexes in such a way as to make the EncodedProgram for this +// AssemblyProgram look more like the EncodedProgram for some other +// AssemblyProgram. The modification process should call UnassignIndexes, do +// its own assignment, and then call AssignRemainingIndexes to ensure all +// indexes are assigned. +// +class AssemblyProgram { + public: + AssemblyProgram(); + ~AssemblyProgram(); + + void set_image_base(uint64 image_base) { image_base_ = image_base; } + + // Instructions will be assembled in the order they are emitted. + + // Generates an entire base relocation table. + void EmitMakeRelocsInstruction(); + + // Following instruction will be assembled at address 'rva'. + void EmitOriginInstruction(RVA rva); + + // Generates a single byte of data or machine instruction. + void EmitByteInstruction(uint8 byte); + + // Generates 4-byte relative reference to address of 'label'. + void EmitRel32(Label* label); + + // Generates 4-byte absolute reference to address of 'label'. + void EmitAbs32(Label* label); + + Label* FindOrMakeAbs32Label(RVA rva); + Label* FindOrMakeRel32Label(RVA rva); + + void DefaultAssignIndexes(); + void UnassignIndexes(); + void AssignRemainingIndexes(); + + EncodedProgram* Encode() const; + + // Accessor for instruction list. + const std::vector<Instruction*>& instructions() const { + return instructions_; + } + + // Returns the label if the instruction contains and absolute address, + // otherwise returns NULL. + Label* InstructionAbs32Label(const Instruction* instruction) const; + + // Returns the label if the instruction contains and rel32 offset, + // otherwise returns NULL. + Label* InstructionRel32Label(const Instruction* instruction) const; + + + private: + void Emit(Instruction* instruction) { instructions_.push_back(instruction); } + + Label* FindLabel(RVA rva, RVAToLabel* labels); + + // Helper methods for the public versions. + static void UnassignIndexes(RVAToLabel* labels); + static void DefaultAssignIndexes(RVAToLabel* labels); + static void AssignRemainingIndexes(RVAToLabel* labels); + + // Sharing instructions that emit a single byte saves a lot of space. + Instruction* GetByteInstruction(uint8 byte); + Instruction** byte_instruction_cache_; + + uint64 image_base_; // Desired or mandated base address of image. + + std::vector<Instruction*> instructions_; // All the instructions in program. + + // These are lookup maps to find the label associated with a given address. + // We have separate label spaces for addresses referenced by rel32 labels and + // abs32 labels. This is somewhat arbitrary. + RVAToLabel rel32_labels_; + RVAToLabel abs32_labels_; + + DISALLOW_COPY_AND_ASSIGN(AssemblyProgram); +}; + +} // namespace courgette +#endif // COURGETTE_ASSEMBLY_PROGRAM_H_ diff --git a/third_party/courgette/bsdiff_memory_unittest.cc b/courgette/bsdiff_memory_unittest.cc index e1add46..97da726 100644 --- a/third_party/courgette/bsdiff_memory_unittest.cc +++ b/courgette/bsdiff_memory_unittest.cc @@ -1,112 +1,111 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-#include "third_party/courgette/bsdiff.h"
-
-#include <string>
-
-#include "base/file_util.h"
-#include "base/path_service.h"
-#include "base/string_util.h"
-
-#include "third_party/courgette/courgette.h"
-#include "third_party/courgette/streams.h"
-
-#include "testing/gtest/include/gtest/gtest.h"
-
-class BSDiffMemoryTest : public testing::Test {
- public:
- std::string FileContents(const char *file_name) const;
- void GenerateAndTestPatch(const std::string& a, const std::string& b) const;
-
- private:
- void SetUp() {
- PathService::Get(base::DIR_SOURCE_ROOT, &test_dir_);
- // file_util::AppendToPath(&test_dir_, L"third_party");
- file_util::AppendToPath(&test_dir_, L"courgette");
- file_util::AppendToPath(&test_dir_, L"testdata");
- }
-
- std::wstring test_dir_;
-};
-
-// Reads a test file into a string.
-std::string BSDiffMemoryTest::FileContents(const char *file_name) const {
- std::wstring file_path = test_dir_;
- file_util::AppendToPath(&file_path, UTF8ToWide(file_name));
- std::string file_bytes;
- if (!file_util::ReadFileToString(file_path, &file_bytes)) {
- EXPECT_TRUE(!"Could not read test data");
- }
- return file_bytes;
-}
-
-void BSDiffMemoryTest::GenerateAndTestPatch(const std::string& old_text,
- const std::string& new_text) const {
- courgette::SourceStream old1;
- courgette::SourceStream new1;
- old1.Init(old_text.c_str(), old_text.length());
- new1.Init(new_text.c_str(), new_text.length());
-
- courgette::SinkStream patch1;
- courgette::BSDiffStatus status = CreateBinaryPatch(&old1, &new1, &patch1);
- EXPECT_EQ(courgette::OK, status);
-
- courgette::SourceStream old2;
- courgette::SourceStream patch2;
- old2.Init(old_text.c_str(), old_text.length());
- patch2.Init(patch1);
-
- courgette::SinkStream new2;
- status = ApplyBinaryPatch(&old2, &patch2, &new2);
- EXPECT_EQ(courgette::OK, status);
- EXPECT_EQ(new_text.length(), new2.Length());
- EXPECT_EQ(0, memcmp(new_text.c_str(), new2.Buffer(), new_text.length()));
-}
-
-TEST_F(BSDiffMemoryTest, TestEmpty) {
- GenerateAndTestPatch("", "");
-}
-
-TEST_F(BSDiffMemoryTest, TestEmptyVsNonempty) {
- GenerateAndTestPatch("", "xxx");
-}
-
-TEST_F(BSDiffMemoryTest, TestNonemptyVsEmpty) {
- GenerateAndTestPatch("xxx", "");
-}
-
-TEST_F(BSDiffMemoryTest, TestSmallInputsWithSmallChanges) {
- std::string file1 =
- "I would not, could not, in a box.\n"
- "I could not, would not, with a fox.\n"
- "I will not eat them with a mouse.\n"
- "I will not eat them in a house.\n"
- "I will not eat them here or there.\n"
- "I will not eat them anywhere.\n"
- "I do not eat green eggs and ham.\n"
- "I do not like them, Sam-I-am.\n";
- std::string file2 =
- "I would not, could not, in a BOX.\n"
- "I could not, would not, with a FOX.\n"
- "I will not eat them with a MOUSE.\n"
- "I will not eat them in a HOUSE.\n"
- "I will not eat them in a HOUSE.\n" // Extra line.
- "I will not eat them here or THERE.\n"
- "I will not eat them ANYWHERE.\n"
- "I do not eat green eggs and HAM.\n"
- "I do not like them, Sam-I-am.\n";
- GenerateAndTestPatch(file1, file2);
-}
-
-TEST_F(BSDiffMemoryTest, TestIndenticalDlls) {
- std::string file1 = FileContents("en-US.dll");
- GenerateAndTestPatch(file1, file1);
-}
-
-TEST_F(BSDiffMemoryTest, TestDifferentExes) {
- std::string file1 = FileContents("setup1.exe");
- std::string file2 = FileContents("setup2.exe");
- GenerateAndTestPatch(file1, file2);
-}
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +#include "courgette/third_party/bsdiff.h" + +#include <string> + +#include "base/file_util.h" +#include "base/path_service.h" +#include "base/string_util.h" + +#include "courgette/courgette.h" +#include "courgette/streams.h" + +#include "testing/gtest/include/gtest/gtest.h" + +class BSDiffMemoryTest : public testing::Test { + public: + std::string FileContents(const char *file_name) const; + void GenerateAndTestPatch(const std::string& a, const std::string& b) const; + + private: + void SetUp() { + PathService::Get(base::DIR_SOURCE_ROOT, &test_dir_); + file_util::AppendToPath(&test_dir_, L"courgette"); + file_util::AppendToPath(&test_dir_, L"testdata"); + } + + std::wstring test_dir_; +}; + +// Reads a test file into a string. +std::string BSDiffMemoryTest::FileContents(const char *file_name) const { + std::wstring file_path = test_dir_; + file_util::AppendToPath(&file_path, UTF8ToWide(file_name)); + std::string file_bytes; + if (!file_util::ReadFileToString(file_path, &file_bytes)) { + EXPECT_TRUE(!"Could not read test data"); + } + return file_bytes; +} + +void BSDiffMemoryTest::GenerateAndTestPatch(const std::string& old_text, + const std::string& new_text) const { + courgette::SourceStream old1; + courgette::SourceStream new1; + old1.Init(old_text.c_str(), old_text.length()); + new1.Init(new_text.c_str(), new_text.length()); + + courgette::SinkStream patch1; + courgette::BSDiffStatus status = CreateBinaryPatch(&old1, &new1, &patch1); + EXPECT_EQ(courgette::OK, status); + + courgette::SourceStream old2; + courgette::SourceStream patch2; + old2.Init(old_text.c_str(), old_text.length()); + patch2.Init(patch1); + + courgette::SinkStream new2; + status = ApplyBinaryPatch(&old2, &patch2, &new2); + EXPECT_EQ(courgette::OK, status); + EXPECT_EQ(new_text.length(), new2.Length()); + EXPECT_EQ(0, memcmp(new_text.c_str(), new2.Buffer(), new_text.length())); +} + +TEST_F(BSDiffMemoryTest, TestEmpty) { + GenerateAndTestPatch("", ""); +} + +TEST_F(BSDiffMemoryTest, TestEmptyVsNonempty) { + GenerateAndTestPatch("", "xxx"); +} + +TEST_F(BSDiffMemoryTest, TestNonemptyVsEmpty) { + GenerateAndTestPatch("xxx", ""); +} + +TEST_F(BSDiffMemoryTest, TestSmallInputsWithSmallChanges) { + std::string file1 = + "I would not, could not, in a box.\n" + "I could not, would not, with a fox.\n" + "I will not eat them with a mouse.\n" + "I will not eat them in a house.\n" + "I will not eat them here or there.\n" + "I will not eat them anywhere.\n" + "I do not eat green eggs and ham.\n" + "I do not like them, Sam-I-am.\n"; + std::string file2 = + "I would not, could not, in a BOX.\n" + "I could not, would not, with a FOX.\n" + "I will not eat them with a MOUSE.\n" + "I will not eat them in a HOUSE.\n" + "I will not eat them in a HOUSE.\n" // Extra line. + "I will not eat them here or THERE.\n" + "I will not eat them ANYWHERE.\n" + "I do not eat green eggs and HAM.\n" + "I do not like them, Sam-I-am.\n"; + GenerateAndTestPatch(file1, file2); +} + +TEST_F(BSDiffMemoryTest, TestIndenticalDlls) { + std::string file1 = FileContents("en-US.dll"); + GenerateAndTestPatch(file1, file1); +} + +TEST_F(BSDiffMemoryTest, TestDifferentExes) { + std::string file1 = FileContents("setup1.exe"); + std::string file2 = FileContents("setup2.exe"); + GenerateAndTestPatch(file1, file2); +} diff --git a/third_party/courgette/courgette.gyp b/courgette/courgette.gyp index a55b191..03de462 100644 --- a/third_party/courgette/courgette.gyp +++ b/courgette/courgette.gyp @@ -7,15 +7,15 @@ 'chromium_code': 1, }, 'includes': [ - '../../build/common.gypi', + '../build/common.gypi', ], 'targets': [ { 'target_name': 'courgette_lib', 'type': '<(library)', 'dependencies': [ - '../../base/base.gyp:base', - #'../../third_party/lzma_sdk/7z_C.vcproj', + '../base/base.gyp:base', + #'../third_party/lzma_sdk/7z_C.vcproj', ], 'msvs_guid': '9A72A362-E617-4205-B9F2-43C6FB280FA1', 'sources': [ @@ -23,9 +23,9 @@ 'adjustment_method.h', 'assembly_program.cc', 'assembly_program.h', - 'bsdiff.h', - 'bsdiff_apply.cc', - 'bsdiff_create.cc', + 'third_party/bsdiff.h', + 'third_party/bsdiff_apply.cc', + 'third_party/bsdiff_create.cc', 'courgette.h', 'crc.cc', 'crc.h', @@ -49,38 +49,29 @@ 'win32_x86_generator.h', 'win32_x86_patcher.h', ], - 'include_dirs': [ - #'..', - ], - 'direct_dependent_settings': { - 'include_dirs': [ - # '..', - ], - }, }, - { - 'target_name': 'courgette_tool', + { + 'target_name': 'courgette', 'type': 'executable', 'msvs_guid': '4EA8CE12-9C6F-45E5-9D08-720383FE3685', 'sources': [ 'courgette_tool.cc', - #'memory_monitor.cc', - ], + ], 'dependencies': [ 'courgette_lib', - '../../base/base.gyp:base', + '../base/base.gyp:base', ], }, - { + { 'target_name': 'courgette_minimal_tool', 'type': 'executable', 'msvs_guid': 'EB79415F-2F17-4BDC-AADD-4CA4C2D21B73', 'sources': [ 'courgette_minimal_tool.cc', - ], + ], 'dependencies': [ 'courgette_lib', - '../../base/base.gyp:base', + '../base/base.gyp:base', ], }, { @@ -96,24 +87,24 @@ 'image_info_unittest.cc', 'run_all_unittests.cc', 'streams_unittest.cc', - ], + ], 'dependencies': [ 'courgette_lib', - '../../base/base.gyp:base', - '../../testing/gtest.gyp:gtest', + '../base/base.gyp:base', + '../testing/gtest.gyp:gtest', ], }, - { + { 'target_name': 'courgette_fuzz', 'type': 'executable', 'msvs_guid': '57C27529-8CA9-4FC3-9C02-DA05B172F785', 'sources': [ 'encoded_program_fuzz_unittest.cc', - ], + ], 'dependencies': [ 'courgette_lib', - '../../base/base.gyp:base', - '../../testing/gtest.gyp:gtest', + '../base/base.gyp:base', + '../testing/gtest.gyp:gtest', ], }, ], diff --git a/third_party/courgette/courgette.h b/courgette/courgette.h index 09bb375..0ddded5 100644 --- a/third_party/courgette/courgette.h +++ b/courgette/courgette.h @@ -1,118 +1,118 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-#ifndef COURGETTE_COURGETTE_H_
-#define COURGETTE_COURGETTE_H_
-
-namespace courgette {
-
-// Status codes for Courgette APIs.
-//
-// Client code should only rely on the distintion between C_OK and the other
-// status codes.
-//
-enum Status {
- C_OK = 1, // Successful operation.
-
- C_GENERAL_ERROR = 2, // Error other than listed below.
-
- C_READ_OPEN_ERROR = 3, // Could not open input file for reading.
- C_READ_ERROR = 4, // Could not read from opened input file.
-
- C_WRITE_OPEN_ERROR = 3, // Could not open output file for writing.
- C_WRITE_ERROR = 4, // Could not write to opened output file.
-
- C_BAD_ENSEMBLE_MAGIC = 5, // Ensemble patch has bad magic.
- C_BAD_ENSEMBLE_VERSION = 6, // Ensemble patch has wrong version.
- C_BAD_ENSEMBLE_HEADER = 7, // Ensemble patch has corrupt header.
- C_BAD_ENSEMBLE_CRC = 8, // Ensemble patch has corrupt header.
-
- C_BAD_TRANSFORM = 12, // Transform mis-specified.
- C_BAD_BASE = 13, // Base for transform malformed.
-
- C_BINARY_DIFF_CRC_ERROR = 14, // Internal diff input doesn't have expected
- // CRC.
-
- // Internal errors.
- C_STREAM_ERROR = 20, // Unexpected error from streams.h.
- C_STREAM_NOT_CONSUMED = 21, // Stream has extra data, is expected to be
- // used up.
- C_SERIALIZATION_FAILED = 22, //
- C_DESERIALIZATION_FAILED = 23, //
- C_INPUT_NOT_RECOGNIZED = 24, // Unrecognized input (not an executable).
- C_DISASSEMBLY_FAILED = 25, //
- C_ASSEMBLY_FAILED = 26, //
- C_ADJUSTMENT_FAILED = 27, //
-
-
-};
-
-class SinkStream;
-class SinkStreamSet;
-class SourceStream;
-class SourceStreamSet;
-
-class AssemblyProgram;
-class EncodedProgram;
-
-// Applies the patch to the bytes in |old| and writes the transformed ensemble
-// to |output|.
-// Returns C_OK unless something went wrong.
-Status ApplyEnsemblePatch(SourceStream* old, SourceStream* patch,
- SinkStream* output);
-
-// Applies the patch in |patch_file_name| to the bytes in |old_file_name| and
-// writes the transformed ensemble to |new_file_name|.
-// Returns C_OK unless something went wrong.
-// This function first validates that the patch file has a proper header, so the
-// function can be used to 'try' a patch.
-Status ApplyEnsemblePatch(const wchar_t* old_file_name,
- const wchar_t* patch_file_name,
- const wchar_t* new_file_name);
-
-// Generates a patch that will transform the bytes in |old| into the bytes in
-// |target|.
-// Returns C_OK unless something when wrong (unexpected).
-Status GenerateEnsemblePatch(SourceStream* old, SourceStream* target,
- SinkStream* patch);
-
-// Parses a Windows 32-bit 'Portable Executable' format file from memory,
-// storing the pointer to the AssemblyProgram in |*output|.
-// Returns C_OK if successful, otherwise returns an error status and sets
-// |*output| to NULL.
-Status ParseWin32X86PE(const void* buffer, size_t length,
- AssemblyProgram** output);
-
-// Converts |program| into encoded form, returning it as |*output|.
-// Returns C_OK if succeeded, otherwise returns an error status and
-// sets |*output| to NULL
-Status Encode(AssemblyProgram* program, EncodedProgram** output);
-
-
-// Serializes |encoded| into the stream set.
-// Returns C_OK if succeeded, otherwise returns an error status.
-Status WriteEncodedProgram(EncodedProgram* encoded, SinkStreamSet* sink);
-
-// Assembles |encoded|, emitting the bytes into |buffer|.
-// Returns C_OK if succeeded, otherwise returns an error status and leaves
-// |buffer| in an undefined state.
-Status Assemble(EncodedProgram* encoded, SinkStream* buffer);
-
-// Deserializes program from the stream set.
-// Returns C_OK if succeeded, otherwise returns an error status and
-// sets |*output| to NULL
-Status ReadEncodedProgram(SourceStreamSet* source, EncodedProgram** output);
-
-// Used to free an AssemblyProgram returned by other APIs.
-void DeleteAssemblyProgram(AssemblyProgram* program);
-
-// Used to free an EncodedProgram returned by other APIs.
-void DeleteEncodedProgram(EncodedProgram* encoded);
-
-// Adjusts |program| to look more like |model|.
-//
-Status Adjust(const AssemblyProgram& model, AssemblyProgram *program);
-
-} // namespace courgette
-#endif // COURGETTE_COURGETTE_H_
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +#ifndef COURGETTE_COURGETTE_H_ +#define COURGETTE_COURGETTE_H_ + +namespace courgette { + +// Status codes for Courgette APIs. +// +// Client code should only rely on the distintion between C_OK and the other +// status codes. +// +enum Status { + C_OK = 1, // Successful operation. + + C_GENERAL_ERROR = 2, // Error other than listed below. + + C_READ_OPEN_ERROR = 3, // Could not open input file for reading. + C_READ_ERROR = 4, // Could not read from opened input file. + + C_WRITE_OPEN_ERROR = 3, // Could not open output file for writing. + C_WRITE_ERROR = 4, // Could not write to opened output file. + + C_BAD_ENSEMBLE_MAGIC = 5, // Ensemble patch has bad magic. + C_BAD_ENSEMBLE_VERSION = 6, // Ensemble patch has wrong version. + C_BAD_ENSEMBLE_HEADER = 7, // Ensemble patch has corrupt header. + C_BAD_ENSEMBLE_CRC = 8, // Ensemble patch has corrupt header. + + C_BAD_TRANSFORM = 12, // Transform mis-specified. + C_BAD_BASE = 13, // Base for transform malformed. + + C_BINARY_DIFF_CRC_ERROR = 14, // Internal diff input doesn't have expected + // CRC. + + // Internal errors. + C_STREAM_ERROR = 20, // Unexpected error from streams.h. + C_STREAM_NOT_CONSUMED = 21, // Stream has extra data, is expected to be + // used up. + C_SERIALIZATION_FAILED = 22, // + C_DESERIALIZATION_FAILED = 23, // + C_INPUT_NOT_RECOGNIZED = 24, // Unrecognized input (not an executable). + C_DISASSEMBLY_FAILED = 25, // + C_ASSEMBLY_FAILED = 26, // + C_ADJUSTMENT_FAILED = 27, // + + +}; + +class SinkStream; +class SinkStreamSet; +class SourceStream; +class SourceStreamSet; + +class AssemblyProgram; +class EncodedProgram; + +// Applies the patch to the bytes in |old| and writes the transformed ensemble +// to |output|. +// Returns C_OK unless something went wrong. +Status ApplyEnsemblePatch(SourceStream* old, SourceStream* patch, + SinkStream* output); + +// Applies the patch in |patch_file_name| to the bytes in |old_file_name| and +// writes the transformed ensemble to |new_file_name|. +// Returns C_OK unless something went wrong. +// This function first validates that the patch file has a proper header, so the +// function can be used to 'try' a patch. +Status ApplyEnsemblePatch(const wchar_t* old_file_name, + const wchar_t* patch_file_name, + const wchar_t* new_file_name); + +// Generates a patch that will transform the bytes in |old| into the bytes in +// |target|. +// Returns C_OK unless something when wrong (unexpected). +Status GenerateEnsemblePatch(SourceStream* old, SourceStream* target, + SinkStream* patch); + +// Parses a Windows 32-bit 'Portable Executable' format file from memory, +// storing the pointer to the AssemblyProgram in |*output|. +// Returns C_OK if successful, otherwise returns an error status and sets +// |*output| to NULL. +Status ParseWin32X86PE(const void* buffer, size_t length, + AssemblyProgram** output); + +// Converts |program| into encoded form, returning it as |*output|. +// Returns C_OK if succeeded, otherwise returns an error status and +// sets |*output| to NULL +Status Encode(AssemblyProgram* program, EncodedProgram** output); + + +// Serializes |encoded| into the stream set. +// Returns C_OK if succeeded, otherwise returns an error status. +Status WriteEncodedProgram(EncodedProgram* encoded, SinkStreamSet* sink); + +// Assembles |encoded|, emitting the bytes into |buffer|. +// Returns C_OK if succeeded, otherwise returns an error status and leaves +// |buffer| in an undefined state. +Status Assemble(EncodedProgram* encoded, SinkStream* buffer); + +// Deserializes program from the stream set. +// Returns C_OK if succeeded, otherwise returns an error status and +// sets |*output| to NULL +Status ReadEncodedProgram(SourceStreamSet* source, EncodedProgram** output); + +// Used to free an AssemblyProgram returned by other APIs. +void DeleteAssemblyProgram(AssemblyProgram* program); + +// Used to free an EncodedProgram returned by other APIs. +void DeleteEncodedProgram(EncodedProgram* encoded); + +// Adjusts |program| to look more like |model|. +// +Status Adjust(const AssemblyProgram& model, AssemblyProgram *program); + +} // namespace courgette +#endif // COURGETTE_COURGETTE_H_ diff --git a/courgette/courgette.sln b/courgette/courgette.sln new file mode 100644 index 0000000..38e1f59 --- /dev/null +++ b/courgette/courgette.sln @@ -0,0 +1,111 @@ +Microsoft Visual Studio Solution File, Format Version 9.00
+# Visual Studio 2005
+Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "courgette_lib", "courgette_lib.vcproj", "{9A72A362-E617-4205-B9F2-43C6FB280FA1}"
+EndProject
+Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "courgette_unittests", "courgette_unittests.vcproj", "{24309F1A-4035-46F9-A3D8-F47DC4BCC2B8}"
+ ProjectSection(ProjectDependencies) = postProject
+ {9A72A362-E617-4205-B9F2-43C6FB280FA1} = {9A72A362-E617-4205-B9F2-43C6FB280FA1}
+ {1832A374-8A74-4F9E-B536-69A699B3E165} = {1832A374-8A74-4F9E-B536-69A699B3E165}
+ {BFE8E2A7-3B3B-43B0-A994-3058B852DB8B} = {BFE8E2A7-3B3B-43B0-A994-3058B852DB8B}
+ {F22022F0-2D3B-5610-4E80-C674A8E01C5A} = {F22022F0-2D3B-5610-4E80-C674A8E01C5A}
+ {8C27D792-2648-4F5E-9ED0-374276327308} = {8C27D792-2648-4F5E-9ED0-374276327308}
+ EndProjectSection
+EndProject
+Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "courgette_fuzz", "courgette_fuzz.vcproj", "{57C27529-8CA9-4FC3-9C02-DA05B172F785}"
+ ProjectSection(ProjectDependencies) = postProject
+ {9A72A362-E617-4205-B9F2-43C6FB280FA1} = {9A72A362-E617-4205-B9F2-43C6FB280FA1}
+ {1832A374-8A74-4F9E-B536-69A699B3E165} = {1832A374-8A74-4F9E-B536-69A699B3E165}
+ {BFE8E2A7-3B3B-43B0-A994-3058B852DB8B} = {BFE8E2A7-3B3B-43B0-A994-3058B852DB8B}
+ {F22022F0-2D3B-5610-4E80-C674A8E01C5A} = {F22022F0-2D3B-5610-4E80-C674A8E01C5A}
+ {8C27D792-2648-4F5E-9ED0-374276327308} = {8C27D792-2648-4F5E-9ED0-374276327308}
+ EndProjectSection
+EndProject
+Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "courgette_minimal_tool", "courgette_minimal_tool.vcproj", "{EB79415F-2F17-4BDC-AADD-4CA4C2D21B73}"
+ ProjectSection(ProjectDependencies) = postProject
+ {9A72A362-E617-4205-B9F2-43C6FB280FA1} = {9A72A362-E617-4205-B9F2-43C6FB280FA1}
+ {1832A374-8A74-4F9E-B536-69A699B3E165} = {1832A374-8A74-4F9E-B536-69A699B3E165}
+ {F22022F0-2D3B-5610-4E80-C674A8E01C5A} = {F22022F0-2D3B-5610-4E80-C674A8E01C5A}
+ {8C27D792-2648-4F5E-9ED0-374276327308} = {8C27D792-2648-4F5E-9ED0-374276327308}
+ EndProjectSection
+EndProject
+Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "courgette", "courgette.vcproj", "{4EA8CE12-9C6F-45E5-9D08-720383FE3685}"
+ ProjectSection(ProjectDependencies) = postProject
+ {9A72A362-E617-4205-B9F2-43C6FB280FA1} = {9A72A362-E617-4205-B9F2-43C6FB280FA1}
+ {1832A374-8A74-4F9E-B536-69A699B3E165} = {1832A374-8A74-4F9E-B536-69A699B3E165}
+ {F22022F0-2D3B-5610-4E80-C674A8E01C5A} = {F22022F0-2D3B-5610-4E80-C674A8E01C5A}
+ {8C27D792-2648-4F5E-9ED0-374276327308} = {8C27D792-2648-4F5E-9ED0-374276327308}
+ EndProjectSection
+EndProject
+Project("{2150E333-8FDC-42A3-9474-1A3956D46DE8}") = "dependencies", "dependencies", "{3C050794-ECB2-85EB-D650-18D00BF2A989}"
+EndProject
+Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "icuuc", "..\third_party\icu38\icuuc.vcproj", "{8C27D792-2648-4F5E-9ED0-374276327308}"
+ ProjectSection(ProjectDependencies) = postProject
+ {A0D94973-D355-47A5-A1E2-3456F321F010} = {A0D94973-D355-47A5-A1E2-3456F321F010}
+ EndProjectSection
+EndProject
+Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "icui18n", "..\third_party\icu38\icui18n.vcproj", "{F22022F0-2D3B-5610-4E80-C674A8E01C5A}"
+EndProject
+Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "icudata", "..\third_party\icu38\icudata.vcproj", "{A0D94973-D355-47A5-A1E2-3456F321F010}"
+EndProject
+Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "base", "..\base\base.vcproj", "{1832A374-8A74-4F9E-B536-69A699B3E165}"
+EndProject
+Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "gtest", "..\testing\gtest.vcproj", "{BFE8E2A7-3B3B-43B0-A994-3058B852DB8B}"
+EndProject
+Global
+ GlobalSection(SolutionConfigurationPlatforms) = preSolution
+ Debug|Win32 = Debug|Win32
+ Release|Win32 = Release|Win32
+ EndGlobalSection
+ GlobalSection(ProjectConfigurationPlatforms) = postSolution
+ {1832A374-8A74-4F9E-B536-69A699B3E165}.Debug|Win32.ActiveCfg = Debug|Win32
+ {1832A374-8A74-4F9E-B536-69A699B3E165}.Debug|Win32.Build.0 = Debug|Win32
+ {1832A374-8A74-4F9E-B536-69A699B3E165}.Release|Win32.ActiveCfg = Release|Win32
+ {1832A374-8A74-4F9E-B536-69A699B3E165}.Release|Win32.Build.0 = Release|Win32
+ {24309F1A-4035-46F9-A3D8-F47DC4BCC2B8}.Debug|Win32.ActiveCfg = Debug|Win32
+ {24309F1A-4035-46F9-A3D8-F47DC4BCC2B8}.Debug|Win32.Build.0 = Debug|Win32
+ {24309F1A-4035-46F9-A3D8-F47DC4BCC2B8}.Release|Win32.ActiveCfg = Release|Win32
+ {24309F1A-4035-46F9-A3D8-F47DC4BCC2B8}.Release|Win32.Build.0 = Release|Win32
+ {4EA8CE12-9C6F-45E5-9D08-720383FE3685}.Debug|Win32.ActiveCfg = Debug|Win32
+ {4EA8CE12-9C6F-45E5-9D08-720383FE3685}.Debug|Win32.Build.0 = Debug|Win32
+ {4EA8CE12-9C6F-45E5-9D08-720383FE3685}.Release|Win32.ActiveCfg = Release|Win32
+ {4EA8CE12-9C6F-45E5-9D08-720383FE3685}.Release|Win32.Build.0 = Release|Win32
+ {57C27529-8CA9-4FC3-9C02-DA05B172F785}.Debug|Win32.ActiveCfg = Debug|Win32
+ {57C27529-8CA9-4FC3-9C02-DA05B172F785}.Debug|Win32.Build.0 = Debug|Win32
+ {57C27529-8CA9-4FC3-9C02-DA05B172F785}.Release|Win32.ActiveCfg = Release|Win32
+ {57C27529-8CA9-4FC3-9C02-DA05B172F785}.Release|Win32.Build.0 = Release|Win32
+ {8C27D792-2648-4F5E-9ED0-374276327308}.Debug|Win32.ActiveCfg = Debug|Win32
+ {8C27D792-2648-4F5E-9ED0-374276327308}.Debug|Win32.Build.0 = Debug|Win32
+ {8C27D792-2648-4F5E-9ED0-374276327308}.Release|Win32.ActiveCfg = Release|Win32
+ {8C27D792-2648-4F5E-9ED0-374276327308}.Release|Win32.Build.0 = Release|Win32
+ {9A72A362-E617-4205-B9F2-43C6FB280FA1}.Debug|Win32.ActiveCfg = Debug|Win32
+ {9A72A362-E617-4205-B9F2-43C6FB280FA1}.Debug|Win32.Build.0 = Debug|Win32
+ {9A72A362-E617-4205-B9F2-43C6FB280FA1}.Release|Win32.ActiveCfg = Release|Win32
+ {9A72A362-E617-4205-B9F2-43C6FB280FA1}.Release|Win32.Build.0 = Release|Win32
+ {A0D94973-D355-47A5-A1E2-3456F321F010}.Debug|Win32.ActiveCfg = Debug|Win32
+ {A0D94973-D355-47A5-A1E2-3456F321F010}.Debug|Win32.Build.0 = Debug|Win32
+ {A0D94973-D355-47A5-A1E2-3456F321F010}.Release|Win32.ActiveCfg = Release|Win32
+ {A0D94973-D355-47A5-A1E2-3456F321F010}.Release|Win32.Build.0 = Release|Win32
+ {BFE8E2A7-3B3B-43B0-A994-3058B852DB8B}.Debug|Win32.ActiveCfg = Debug|Win32
+ {BFE8E2A7-3B3B-43B0-A994-3058B852DB8B}.Debug|Win32.Build.0 = Debug|Win32
+ {BFE8E2A7-3B3B-43B0-A994-3058B852DB8B}.Release|Win32.ActiveCfg = Release|Win32
+ {BFE8E2A7-3B3B-43B0-A994-3058B852DB8B}.Release|Win32.Build.0 = Release|Win32
+ {EB79415F-2F17-4BDC-AADD-4CA4C2D21B73}.Debug|Win32.ActiveCfg = Debug|Win32
+ {EB79415F-2F17-4BDC-AADD-4CA4C2D21B73}.Debug|Win32.Build.0 = Debug|Win32
+ {EB79415F-2F17-4BDC-AADD-4CA4C2D21B73}.Release|Win32.ActiveCfg = Release|Win32
+ {EB79415F-2F17-4BDC-AADD-4CA4C2D21B73}.Release|Win32.Build.0 = Release|Win32
+ {F22022F0-2D3B-5610-4E80-C674A8E01C5A}.Debug|Win32.ActiveCfg = Debug|Win32
+ {F22022F0-2D3B-5610-4E80-C674A8E01C5A}.Debug|Win32.Build.0 = Debug|Win32
+ {F22022F0-2D3B-5610-4E80-C674A8E01C5A}.Release|Win32.ActiveCfg = Release|Win32
+ {F22022F0-2D3B-5610-4E80-C674A8E01C5A}.Release|Win32.Build.0 = Release|Win32
+ EndGlobalSection
+ GlobalSection(SolutionProperties) = preSolution
+ HideSolutionNode = FALSE
+ EndGlobalSection
+ GlobalSection(NestedProjects) = preSolution
+ {8C27D792-2648-4F5E-9ED0-374276327308} = {3C050794-ECB2-85EB-D650-18D00BF2A989}
+ {F22022F0-2D3B-5610-4E80-C674A8E01C5A} = {3C050794-ECB2-85EB-D650-18D00BF2A989}
+ {A0D94973-D355-47A5-A1E2-3456F321F010} = {3C050794-ECB2-85EB-D650-18D00BF2A989}
+ {1832A374-8A74-4F9E-B536-69A699B3E165} = {3C050794-ECB2-85EB-D650-18D00BF2A989}
+ {BFE8E2A7-3B3B-43B0-A994-3058B852DB8B} = {3C050794-ECB2-85EB-D650-18D00BF2A989}
+ EndGlobalSection
+EndGlobal
diff --git a/third_party/courgette/courgette_minimal_tool.cc b/courgette/courgette_minimal_tool.cc index de5cd9e..84de068 100644 --- a/third_party/courgette/courgette_minimal_tool.cc +++ b/courgette/courgette_minimal_tool.cc @@ -1,50 +1,50 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-// 'courgette_minimal_tool' is not meant to be a serious command-line tool. It
-// has the minimum logic to apply a Courgette patch to a file. The main purpose
-// is to monitor the code size of the patcher.
-
-#include <string>
-
-#include "base/file_util.h"
-
-#include "third_party/courgette/bsdiff.h"
-#include "third_party/courgette/courgette.h"
-#include "third_party/courgette/streams.h"
-
-void PrintHelp() {
- fprintf(stderr,
- "Usage:\n"
- " courgette_minimal_tool <old-file-input> <patch-file-input>"
- " <new-file-output>\n"
- "\n");
-}
-
-void UsageProblem(const char* message) {
- fprintf(stderr, "%s\n", message);
- PrintHelp();
- exit(1);
-}
-
-void Problem(const char* message) {
- fprintf(stderr, "%s\n", message);
- exit(1);
-}
-
-int wmain(int argc, const wchar_t* argv[]) {
- if (argc != 4)
- UsageProblem("bad args");
-
- courgette::Status status =
- courgette::ApplyEnsemblePatch(argv[1], argv[2], argv[3]);
-
- if (status != courgette::C_OK) {
- if (status == courgette::C_READ_OPEN_ERROR) Problem("Can't open file.");
- if (status == courgette::C_WRITE_OPEN_ERROR) Problem("Can't open file.");
- if (status == courgette::C_READ_ERROR) Problem("Can't read from file.");
- if (status == courgette::C_WRITE_ERROR) Problem("Can't write to file.");
- Problem("patch failed.");
- }
-}
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +// 'courgette_minimal_tool' is not meant to be a serious command-line tool. It +// has the minimum logic to apply a Courgette patch to a file. The main purpose +// is to monitor the code size of the patcher. + +#include <string> + +#include "base/file_util.h" + +#include "courgette/third_party/bsdiff.h" +#include "courgette/courgette.h" +#include "courgette/streams.h" + +void PrintHelp() { + fprintf(stderr, + "Usage:\n" + " courgette_minimal_tool <old-file-input> <patch-file-input>" + " <new-file-output>\n" + "\n"); +} + +void UsageProblem(const char* message) { + fprintf(stderr, "%s\n", message); + PrintHelp(); + exit(1); +} + +void Problem(const char* message) { + fprintf(stderr, "%s\n", message); + exit(1); +} + +int wmain(int argc, const wchar_t* argv[]) { + if (argc != 4) + UsageProblem("bad args"); + + courgette::Status status = + courgette::ApplyEnsemblePatch(argv[1], argv[2], argv[3]); + + if (status != courgette::C_OK) { + if (status == courgette::C_READ_OPEN_ERROR) Problem("Can't open file."); + if (status == courgette::C_WRITE_OPEN_ERROR) Problem("Can't open file."); + if (status == courgette::C_READ_ERROR) Problem("Can't read from file."); + if (status == courgette::C_WRITE_ERROR) Problem("Can't write to file."); + Problem("patch failed."); + } +} diff --git a/third_party/courgette/courgette_tool.cc b/courgette/courgette_tool.cc index 867101a7..c07f2da 100644 --- a/third_party/courgette/courgette_tool.cc +++ b/courgette/courgette_tool.cc @@ -1,405 +1,404 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-#include <vector>
-#include <string>
-
-#include "base/at_exit.h"
-#include "base/basictypes.h"
-#include "base/command_line.h"
-#include "base/file_util.h"
-#include "base/logging.h"
-#include "base/string_util.h"
-
-#include "third_party/courgette/bsdiff.h"
-#include "third_party/courgette/courgette.h"
-#include "third_party/courgette/streams.h"
-
-
-void PrintHelp() {
- fprintf(stderr,
- "Usage:\n"
- " courgette_tool -dis <executable_file> <binary_assembly_file>\n"
- " courgette_tool -asm <binary_assembly_file> <executable_file>\n"
- " courgette_tool -disadj "
- "<executable_file> <reference> <binary_assembly_file>\n"
- " courgette_tool -gen <v1> <v2> <patch>\n"
- " courgette_tool -apply <v1> <patch> <v2>\n"
- "\n");
-}
-
-void UsageProblem(const char* message) {
- fprintf(stderr, "%s", message);
- fprintf(stderr, "\n");
- PrintHelp();
- exit(1);
-}
-
-void Problem(const char* format, ...) {
- va_list args;
- va_start(args, format);
- vfprintf(stderr, format, args);
- fprintf(stderr, "\n");
- va_end(args);
- exit(1);
-}
-
-std::string ReadOrFail(const std::wstring& file_name, const char* kind) {
- FilePath file_path(file_name);
- std::string buffer;
- if (!file_util::ReadFileToString(file_path, &buffer))
- Problem("Can't read %s file.", kind);
- return buffer;
-}
-
-void WriteSinkToFile(const courgette::SinkStream *sink,
- const std::wstring& output_file) {
- FilePath output_path(output_file);
- int count =
- file_util::WriteFile(output_path,
- reinterpret_cast<const char*>(sink->Buffer()),
- sink->Length());
- if (count == -1)
- Problem("Cant write output.");
- if (count != sink->Length())
- Problem("Incomplete write.");
-}
-
-void Disassemble(const std::wstring& input_file,
- const std::wstring& output_file) {
- std::string buffer = ReadOrFail(input_file, "input");
-
- courgette::AssemblyProgram* program = NULL;
- const courgette::Status parse_status =
- courgette::ParseWin32X86PE(buffer.c_str(), buffer.length(), &program);
-
- if (parse_status != courgette::C_OK)
- Problem("Can't parse input.");
-
- courgette::EncodedProgram* encoded = NULL;
- const courgette::Status encode_status = Encode(program, &encoded);
-
- courgette::DeleteAssemblyProgram(program);
-
- if (encode_status != courgette::C_OK)
- Problem("Can't encode program.");
-
- courgette::SinkStreamSet sinks;
-
- const courgette::Status write_status =
- courgette::WriteEncodedProgram(encoded, &sinks);
- if (write_status != courgette::C_OK)
- Problem("Can't serialize encoded program.");
-
- courgette::DeleteEncodedProgram(encoded);
-
- courgette::SinkStream sink;
- sinks.CopyTo(&sink);
-
- WriteSinkToFile(&sink, output_file);
-}
-
-void DisassembleAndAdjust(const std::wstring& program_file,
- const std::wstring& model_file,
- const std::wstring& output_file) {
- std::string program_buffer = ReadOrFail(program_file, "program");
- std::string model_buffer = ReadOrFail(model_file, "reference");
-
- courgette::AssemblyProgram* program = NULL;
- const courgette::Status parse_program_status =
- courgette::ParseWin32X86PE(program_buffer.c_str(),
- program_buffer.length(),
- &program);
- if (parse_program_status != courgette::C_OK)
- Problem("Can't parse program input.");
-
- courgette::AssemblyProgram* model = NULL;
- const courgette::Status parse_model_status =
- courgette::ParseWin32X86PE(model_buffer.c_str(),
- model_buffer.length(),
- &model);
- if (parse_model_status != courgette::C_OK)
- Problem("Can't parse model input.");
-
- const courgette::Status adjust_status = Adjust(*model, program);
- if (adjust_status != courgette::C_OK)
- Problem("Can't adjust program.");
-
- courgette::EncodedProgram* encoded = NULL;
- const courgette::Status encode_status = Encode(program, &encoded);
-
- courgette::DeleteAssemblyProgram(program);
-
- if (encode_status != courgette::C_OK)
- Problem("Can't encode program.");
-
- courgette::SinkStreamSet sinks;
-
- const courgette::Status write_status =
- courgette::WriteEncodedProgram(encoded, &sinks);
- if (write_status != courgette::C_OK)
- Problem("Can't serialize encoded program.");
-
- courgette::DeleteEncodedProgram(encoded);
-
- courgette::SinkStream sink;
- sinks.CopyTo(&sink);
-
- WriteSinkToFile(&sink, output_file);
-}
-
-// Diffs two executable files, write a set of files for the diff, one file per
-// stream of the EncodedProgram format. Each file is the bsdiff between the
-// original file's stream and the new file's stream. This is completely
-// uninteresting to users, but it is handy for seeing how much each which
-// streams are contributing to the final file size. Adjustment is optional.
-void DisassembleAdjustDiff(const std::wstring& model_file,
- const std::wstring& program_file,
- const std::wstring& output_file_root,
- bool adjust) {
- std::string model_buffer = ReadOrFail(model_file, "'old'");
- std::string program_buffer = ReadOrFail(program_file, "'new'");
-
- courgette::AssemblyProgram* model = NULL;
- const courgette::Status parse_model_status =
- courgette::ParseWin32X86PE(model_buffer.c_str(),
- model_buffer.length(),
- &model);
- if (parse_model_status != courgette::C_OK)
- Problem("Can't parse model input.");
-
- courgette::AssemblyProgram* program = NULL;
- const courgette::Status parse_program_status =
- courgette::ParseWin32X86PE(program_buffer.c_str(),
- program_buffer.length(),
- &program);
- if (parse_program_status != courgette::C_OK)
- Problem("Can't parse program input.");
-
- if (adjust) {
- const courgette::Status adjust_status = Adjust(*model, program);
- if (adjust_status != courgette::C_OK)
- Problem("Can't adjust program.");
- }
-
- courgette::EncodedProgram* encoded_program = NULL;
- const courgette::Status encode_program_status =
- Encode(program, &encoded_program);
- courgette::DeleteAssemblyProgram(program);
- if (encode_program_status != courgette::C_OK)
- Problem("Can't encode program.");
-
- courgette::EncodedProgram* encoded_model = NULL;
- const courgette::Status encode_model_status = Encode(model, &encoded_model);
- courgette::DeleteAssemblyProgram(model);
- if (encode_model_status != courgette::C_OK)
- Problem("Can't encode model.");
-
- courgette::SinkStreamSet program_sinks;
- const courgette::Status write_program_status =
- courgette::WriteEncodedProgram(encoded_program, &program_sinks);
- if (write_program_status != courgette::C_OK)
- Problem("Can't serialize encoded program.");
- courgette::DeleteEncodedProgram(encoded_program);
-
- courgette::SinkStreamSet model_sinks;
- const courgette::Status write_model_status =
- courgette::WriteEncodedProgram(encoded_model, &model_sinks);
- if (write_model_status != courgette::C_OK)
- Problem("Can't serialize encoded model.");
- courgette::DeleteEncodedProgram(encoded_program);
-
- for (int i = 0; ; ++i) {
- courgette::SinkStream* old_stream = model_sinks.stream(i);
- courgette::SinkStream* new_stream = program_sinks.stream(i);
- if (old_stream == NULL && new_stream == NULL)
- break;
-
- courgette::SourceStream old_source;
- courgette::SourceStream new_source;
- old_source.Init(*old_stream);
- new_source.Init(*new_stream);
- courgette::SinkStream patch_stream;
- courgette::BSDiffStatus status =
- courgette::CreateBinaryPatch(&old_source, &new_source, &patch_stream);
- if (status != courgette::OK) Problem("-xxx failed.");
-
- WriteSinkToFile(&patch_stream,
- output_file_root + L"-" + IntToWString(i));
- }
-}
-
-void Assemble(const std::wstring& input_file,
- const std::wstring& output_file) {
- std::string buffer = ReadOrFail(input_file, "input");
-
- courgette::SourceStreamSet sources;
- if (!sources.Init(buffer.c_str(), buffer.length()))
- Problem("Bad input file.");
-
- courgette::EncodedProgram* encoded = NULL;
- const courgette::Status read_status = ReadEncodedProgram(&sources, &encoded);
- if (read_status != courgette::C_OK)
- Problem("Bad encoded program.");
-
- courgette::SinkStream sink;
-
- const courgette::Status assemble_status = courgette::Assemble(encoded, &sink);
- if (assemble_status != courgette::C_OK)
- Problem("Can't assemble.");
-
- WriteSinkToFile(&sink, output_file);
-}
-
-void GenerateEnsemblePatch(const std::wstring& old_file,
- const std::wstring& new_file,
- const std::wstring& patch_file) {
- std::string old_buffer = ReadOrFail(old_file, "'old' input");
- std::string new_buffer = ReadOrFail(new_file, "'new' input");
-
- courgette::SourceStream old_stream;
- courgette::SourceStream new_stream;
- old_stream.Init(old_buffer);
- new_stream.Init(new_buffer);
-
- courgette::SinkStream patch_stream;
- courgette::Status status =
- courgette::GenerateEnsemblePatch(&old_stream, &new_stream, &patch_stream);
-
- if (status != courgette::C_OK) Problem("-gen failed.");
-
- WriteSinkToFile(&patch_stream, patch_file);
-}
-
-void ApplyEnsemblePatch(const std::wstring& old_file,
- const std::wstring& patch_file,
- const std::wstring& new_file) {
- std::string old_buffer = ReadOrFail(old_file, "'old' input");
- std::string patch_buffer = ReadOrFail(patch_file, "'patch' input");
-
- courgette::SourceStream old_stream;
- courgette::SourceStream patch_stream;
- old_stream.Init(old_buffer);
- patch_stream.Init(patch_buffer);
- courgette::SinkStream new_stream;
- courgette::Status status =
- courgette::ApplyEnsemblePatch(&old_stream, &patch_stream, &new_stream);
-
- if (status != courgette::C_OK) Problem("-apply failed.");
-
- WriteSinkToFile(&new_stream, new_file);
-}
-
-void GenerateBSDiffPatch(const std::wstring& old_file,
- const std::wstring& new_file,
- const std::wstring& patch_file) {
- std::string old_buffer = ReadOrFail(old_file, "'old' input");
- std::string new_buffer = ReadOrFail(new_file, "'new' input");
-
- courgette::SourceStream old_stream;
- courgette::SourceStream new_stream;
- old_stream.Init(old_buffer);
- new_stream.Init(new_buffer);
-
- courgette::SinkStream patch_stream;
- courgette::BSDiffStatus status =
- courgette::CreateBinaryPatch(&old_stream, &new_stream, &patch_stream);
-
- if (status != courgette::OK) Problem("-genbsdiff failed.");
-
- WriteSinkToFile(&patch_stream, patch_file);
-}
-
-void ApplyBSDiffPatch(const std::wstring& old_file,
- const std::wstring& patch_file,
- const std::wstring& new_file) {
- std::string old_buffer = ReadOrFail(old_file, "'old' input");
- std::string patch_buffer = ReadOrFail(patch_file, "'patch' input");
-
- courgette::SourceStream old_stream;
- courgette::SourceStream patch_stream;
- old_stream.Init(old_buffer);
- patch_stream.Init(patch_buffer);
-
- courgette::SinkStream new_stream;
- courgette::BSDiffStatus status =
- courgette::ApplyBinaryPatch(&old_stream, &patch_stream, &new_stream);
-
- if (status != courgette::OK) Problem("-applybsdiff failed.");
-
- WriteSinkToFile(&new_stream, new_file);
-}
-
-int main(int argc, const char* argv[]) {
- base::AtExitManager at_exit_manager;
- CommandLine::Init(argc, argv);
- const CommandLine& command_line = *CommandLine::ForCurrentProcess();
-
- bool cmd_dis = command_line.HasSwitch(L"dis");
- bool cmd_asm = command_line.HasSwitch(L"asm");
- bool cmd_disadj = command_line.HasSwitch(L"disadj");
- bool cmd_make_patch = command_line.HasSwitch(L"gen");
- bool cmd_apply_patch = command_line.HasSwitch(L"apply");
- bool cmd_make_bsdiff_patch = command_line.HasSwitch(L"genbsdiff");
- bool cmd_apply_bsdiff_patch = command_line.HasSwitch(L"applybsdiff");
- bool cmd_spread_1_adjusted = command_line.HasSwitch(L"gen1a");
- bool cmd_spread_1_unadjusted = command_line.HasSwitch(L"gen1u");
-
- std::vector<std::wstring> values = command_line.GetLooseValues();
-
- // '-repeat=N' is for debugging. Running many iterations can reveal leaks and
- // bugs in cleanup.
- int repeat_count = 1;
- std::wstring repeat_switch = command_line.GetSwitchValue(L"repeat");
- if (!repeat_switch.empty())
- if (!StringToInt(repeat_switch, &repeat_count))
- repeat_count = 1;
-
- if (cmd_dis + cmd_asm + cmd_disadj + cmd_make_patch + cmd_apply_patch +
- cmd_make_bsdiff_patch + cmd_apply_bsdiff_patch +
- cmd_spread_1_adjusted + cmd_spread_1_unadjusted
- != 1)
- UsageProblem(
- "Must have exactly one of:\n"
- " -asm, -dis, -disadj, -gen or -apply, -genbsdiff or -applybsdiff.");
-
- while (repeat_count-- > 0) {
- if (cmd_dis) {
- if (values.size() != 2)
- UsageProblem("-dis <executable_file> <courgette_file>");
- Disassemble(values[0], values[1]);
- } else if (cmd_asm) {
- if (values.size() != 2)
- UsageProblem("-asm <courgette_file_input> <executable_file_output>");
- Assemble(values[0], values[1]);
- } else if (cmd_disadj) {
- if (values.size() != 3)
- UsageProblem("-disadj <executable_file> <model> <courgette_file>");
- DisassembleAndAdjust(values[0], values[1], values[2]);
- } else if (cmd_make_patch) {
- if (values.size() != 3)
- UsageProblem("-gen <old_file> <new_file> <patch_file>");
- GenerateEnsemblePatch(values[0], values[1], values[2]);
- } else if (cmd_apply_patch) {
- if (values.size() != 3)
- UsageProblem("-apply <old_file> <patch_file> <new_file>");
- ApplyEnsemblePatch(values[0], values[1], values[2]);
- } else if (cmd_make_bsdiff_patch) {
- if (values.size() != 3)
- UsageProblem("-genbsdiff <old_file> <new_file> <patch_file>");
- GenerateBSDiffPatch(values[0], values[1], values[2]);
- } else if (cmd_apply_bsdiff_patch) {
- if (values.size() != 3)
- UsageProblem("-applybsdiff <old_file> <patch_file> <new_file>");
- ApplyBSDiffPatch(values[0], values[1], values[2]);
- } else if (cmd_spread_1_adjusted || cmd_spread_1_unadjusted) {
- if (values.size() != 3)
- UsageProblem("-gen1[au] <old_file> <new_file> <patch_files_root>");
- DisassembleAdjustDiff(values[0], values[1], values[2],
- cmd_spread_1_adjusted);
- } else {
- UsageProblem("No operation specified");
- }
- }
-}
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +#include <vector> +#include <string> + +#include "base/at_exit.h" +#include "base/basictypes.h" +#include "base/command_line.h" +#include "base/file_util.h" +#include "base/logging.h" +#include "base/string_util.h" + +#include "courgette/third_party/bsdiff.h" +#include "courgette/courgette.h" +#include "courgette/streams.h" + + +void PrintHelp() { + fprintf(stderr, + "Usage:\n" + " courgette -dis <executable_file> <binary_assembly_file>\n" + " courgette -asm <binary_assembly_file> <executable_file>\n" + " courgette -disadj <executable_file> <reference> <binary_assembly_file>\n" + " courgette -gen <v1> <v2> <patch>\n" + " courgette -apply <v1> <patch> <v2>\n" + "\n"); +} + +void UsageProblem(const char* message) { + fprintf(stderr, "%s", message); + fprintf(stderr, "\n"); + PrintHelp(); + exit(1); +} + +void Problem(const char* format, ...) { + va_list args; + va_start(args, format); + vfprintf(stderr, format, args); + fprintf(stderr, "\n"); + va_end(args); + exit(1); +} + +std::string ReadOrFail(const std::wstring& file_name, const char* kind) { + FilePath file_path(file_name); + std::string buffer; + if (!file_util::ReadFileToString(file_path, &buffer)) + Problem("Can't read %s file.", kind); + return buffer; +} + +void WriteSinkToFile(const courgette::SinkStream *sink, + const std::wstring& output_file) { + FilePath output_path(output_file); + int count = + file_util::WriteFile(output_path, + reinterpret_cast<const char*>(sink->Buffer()), + sink->Length()); + if (count == -1) + Problem("Cant write output."); + if (count != sink->Length()) + Problem("Incomplete write."); +} + +void Disassemble(const std::wstring& input_file, + const std::wstring& output_file) { + std::string buffer = ReadOrFail(input_file, "input"); + + courgette::AssemblyProgram* program = NULL; + const courgette::Status parse_status = + courgette::ParseWin32X86PE(buffer.c_str(), buffer.length(), &program); + + if (parse_status != courgette::C_OK) + Problem("Can't parse input."); + + courgette::EncodedProgram* encoded = NULL; + const courgette::Status encode_status = Encode(program, &encoded); + + courgette::DeleteAssemblyProgram(program); + + if (encode_status != courgette::C_OK) + Problem("Can't encode program."); + + courgette::SinkStreamSet sinks; + + const courgette::Status write_status = + courgette::WriteEncodedProgram(encoded, &sinks); + if (write_status != courgette::C_OK) + Problem("Can't serialize encoded program."); + + courgette::DeleteEncodedProgram(encoded); + + courgette::SinkStream sink; + sinks.CopyTo(&sink); + + WriteSinkToFile(&sink, output_file); +} + +void DisassembleAndAdjust(const std::wstring& program_file, + const std::wstring& model_file, + const std::wstring& output_file) { + std::string program_buffer = ReadOrFail(program_file, "program"); + std::string model_buffer = ReadOrFail(model_file, "reference"); + + courgette::AssemblyProgram* program = NULL; + const courgette::Status parse_program_status = + courgette::ParseWin32X86PE(program_buffer.c_str(), + program_buffer.length(), + &program); + if (parse_program_status != courgette::C_OK) + Problem("Can't parse program input."); + + courgette::AssemblyProgram* model = NULL; + const courgette::Status parse_model_status = + courgette::ParseWin32X86PE(model_buffer.c_str(), + model_buffer.length(), + &model); + if (parse_model_status != courgette::C_OK) + Problem("Can't parse model input."); + + const courgette::Status adjust_status = Adjust(*model, program); + if (adjust_status != courgette::C_OK) + Problem("Can't adjust program."); + + courgette::EncodedProgram* encoded = NULL; + const courgette::Status encode_status = Encode(program, &encoded); + + courgette::DeleteAssemblyProgram(program); + + if (encode_status != courgette::C_OK) + Problem("Can't encode program."); + + courgette::SinkStreamSet sinks; + + const courgette::Status write_status = + courgette::WriteEncodedProgram(encoded, &sinks); + if (write_status != courgette::C_OK) + Problem("Can't serialize encoded program."); + + courgette::DeleteEncodedProgram(encoded); + + courgette::SinkStream sink; + sinks.CopyTo(&sink); + + WriteSinkToFile(&sink, output_file); +} + +// Diffs two executable files, write a set of files for the diff, one file per +// stream of the EncodedProgram format. Each file is the bsdiff between the +// original file's stream and the new file's stream. This is completely +// uninteresting to users, but it is handy for seeing how much each which +// streams are contributing to the final file size. Adjustment is optional. +void DisassembleAdjustDiff(const std::wstring& model_file, + const std::wstring& program_file, + const std::wstring& output_file_root, + bool adjust) { + std::string model_buffer = ReadOrFail(model_file, "'old'"); + std::string program_buffer = ReadOrFail(program_file, "'new'"); + + courgette::AssemblyProgram* model = NULL; + const courgette::Status parse_model_status = + courgette::ParseWin32X86PE(model_buffer.c_str(), + model_buffer.length(), + &model); + if (parse_model_status != courgette::C_OK) + Problem("Can't parse model input."); + + courgette::AssemblyProgram* program = NULL; + const courgette::Status parse_program_status = + courgette::ParseWin32X86PE(program_buffer.c_str(), + program_buffer.length(), + &program); + if (parse_program_status != courgette::C_OK) + Problem("Can't parse program input."); + + if (adjust) { + const courgette::Status adjust_status = Adjust(*model, program); + if (adjust_status != courgette::C_OK) + Problem("Can't adjust program."); + } + + courgette::EncodedProgram* encoded_program = NULL; + const courgette::Status encode_program_status = + Encode(program, &encoded_program); + courgette::DeleteAssemblyProgram(program); + if (encode_program_status != courgette::C_OK) + Problem("Can't encode program."); + + courgette::EncodedProgram* encoded_model = NULL; + const courgette::Status encode_model_status = Encode(model, &encoded_model); + courgette::DeleteAssemblyProgram(model); + if (encode_model_status != courgette::C_OK) + Problem("Can't encode model."); + + courgette::SinkStreamSet program_sinks; + const courgette::Status write_program_status = + courgette::WriteEncodedProgram(encoded_program, &program_sinks); + if (write_program_status != courgette::C_OK) + Problem("Can't serialize encoded program."); + courgette::DeleteEncodedProgram(encoded_program); + + courgette::SinkStreamSet model_sinks; + const courgette::Status write_model_status = + courgette::WriteEncodedProgram(encoded_model, &model_sinks); + if (write_model_status != courgette::C_OK) + Problem("Can't serialize encoded model."); + courgette::DeleteEncodedProgram(encoded_program); + + for (int i = 0; ; ++i) { + courgette::SinkStream* old_stream = model_sinks.stream(i); + courgette::SinkStream* new_stream = program_sinks.stream(i); + if (old_stream == NULL && new_stream == NULL) + break; + + courgette::SourceStream old_source; + courgette::SourceStream new_source; + old_source.Init(*old_stream); + new_source.Init(*new_stream); + courgette::SinkStream patch_stream; + courgette::BSDiffStatus status = + courgette::CreateBinaryPatch(&old_source, &new_source, &patch_stream); + if (status != courgette::OK) Problem("-xxx failed."); + + WriteSinkToFile(&patch_stream, + output_file_root + L"-" + IntToWString(i)); + } +} + +void Assemble(const std::wstring& input_file, + const std::wstring& output_file) { + std::string buffer = ReadOrFail(input_file, "input"); + + courgette::SourceStreamSet sources; + if (!sources.Init(buffer.c_str(), buffer.length())) + Problem("Bad input file."); + + courgette::EncodedProgram* encoded = NULL; + const courgette::Status read_status = ReadEncodedProgram(&sources, &encoded); + if (read_status != courgette::C_OK) + Problem("Bad encoded program."); + + courgette::SinkStream sink; + + const courgette::Status assemble_status = courgette::Assemble(encoded, &sink); + if (assemble_status != courgette::C_OK) + Problem("Can't assemble."); + + WriteSinkToFile(&sink, output_file); +} + +void GenerateEnsemblePatch(const std::wstring& old_file, + const std::wstring& new_file, + const std::wstring& patch_file) { + std::string old_buffer = ReadOrFail(old_file, "'old' input"); + std::string new_buffer = ReadOrFail(new_file, "'new' input"); + + courgette::SourceStream old_stream; + courgette::SourceStream new_stream; + old_stream.Init(old_buffer); + new_stream.Init(new_buffer); + + courgette::SinkStream patch_stream; + courgette::Status status = + courgette::GenerateEnsemblePatch(&old_stream, &new_stream, &patch_stream); + + if (status != courgette::C_OK) Problem("-gen failed."); + + WriteSinkToFile(&patch_stream, patch_file); +} + +void ApplyEnsemblePatch(const std::wstring& old_file, + const std::wstring& patch_file, + const std::wstring& new_file) { + std::string old_buffer = ReadOrFail(old_file, "'old' input"); + std::string patch_buffer = ReadOrFail(patch_file, "'patch' input"); + + courgette::SourceStream old_stream; + courgette::SourceStream patch_stream; + old_stream.Init(old_buffer); + patch_stream.Init(patch_buffer); + courgette::SinkStream new_stream; + courgette::Status status = + courgette::ApplyEnsemblePatch(&old_stream, &patch_stream, &new_stream); + + if (status != courgette::C_OK) Problem("-apply failed."); + + WriteSinkToFile(&new_stream, new_file); +} + +void GenerateBSDiffPatch(const std::wstring& old_file, + const std::wstring& new_file, + const std::wstring& patch_file) { + std::string old_buffer = ReadOrFail(old_file, "'old' input"); + std::string new_buffer = ReadOrFail(new_file, "'new' input"); + + courgette::SourceStream old_stream; + courgette::SourceStream new_stream; + old_stream.Init(old_buffer); + new_stream.Init(new_buffer); + + courgette::SinkStream patch_stream; + courgette::BSDiffStatus status = + courgette::CreateBinaryPatch(&old_stream, &new_stream, &patch_stream); + + if (status != courgette::OK) Problem("-genbsdiff failed."); + + WriteSinkToFile(&patch_stream, patch_file); +} + +void ApplyBSDiffPatch(const std::wstring& old_file, + const std::wstring& patch_file, + const std::wstring& new_file) { + std::string old_buffer = ReadOrFail(old_file, "'old' input"); + std::string patch_buffer = ReadOrFail(patch_file, "'patch' input"); + + courgette::SourceStream old_stream; + courgette::SourceStream patch_stream; + old_stream.Init(old_buffer); + patch_stream.Init(patch_buffer); + + courgette::SinkStream new_stream; + courgette::BSDiffStatus status = + courgette::ApplyBinaryPatch(&old_stream, &patch_stream, &new_stream); + + if (status != courgette::OK) Problem("-applybsdiff failed."); + + WriteSinkToFile(&new_stream, new_file); +} + +int main(int argc, const char* argv[]) { + base::AtExitManager at_exit_manager; + CommandLine::Init(argc, argv); + const CommandLine& command_line = *CommandLine::ForCurrentProcess(); + + bool cmd_dis = command_line.HasSwitch(L"dis"); + bool cmd_asm = command_line.HasSwitch(L"asm"); + bool cmd_disadj = command_line.HasSwitch(L"disadj"); + bool cmd_make_patch = command_line.HasSwitch(L"gen"); + bool cmd_apply_patch = command_line.HasSwitch(L"apply"); + bool cmd_make_bsdiff_patch = command_line.HasSwitch(L"genbsdiff"); + bool cmd_apply_bsdiff_patch = command_line.HasSwitch(L"applybsdiff"); + bool cmd_spread_1_adjusted = command_line.HasSwitch(L"gen1a"); + bool cmd_spread_1_unadjusted = command_line.HasSwitch(L"gen1u"); + + std::vector<std::wstring> values = command_line.GetLooseValues(); + + // '-repeat=N' is for debugging. Running many iterations can reveal leaks and + // bugs in cleanup. + int repeat_count = 1; + std::wstring repeat_switch = command_line.GetSwitchValue(L"repeat"); + if (!repeat_switch.empty()) + if (!StringToInt(repeat_switch, &repeat_count)) + repeat_count = 1; + + if (cmd_dis + cmd_asm + cmd_disadj + cmd_make_patch + cmd_apply_patch + + cmd_make_bsdiff_patch + cmd_apply_bsdiff_patch + + cmd_spread_1_adjusted + cmd_spread_1_unadjusted + != 1) + UsageProblem( + "Must have exactly one of:\n" + " -asm, -dis, -disadj, -gen or -apply, -genbsdiff or -applybsdiff."); + + while (repeat_count-- > 0) { + if (cmd_dis) { + if (values.size() != 2) + UsageProblem("-dis <executable_file> <courgette_file>"); + Disassemble(values[0], values[1]); + } else if (cmd_asm) { + if (values.size() != 2) + UsageProblem("-asm <courgette_file_input> <executable_file_output>"); + Assemble(values[0], values[1]); + } else if (cmd_disadj) { + if (values.size() != 3) + UsageProblem("-disadj <executable_file> <model> <courgette_file>"); + DisassembleAndAdjust(values[0], values[1], values[2]); + } else if (cmd_make_patch) { + if (values.size() != 3) + UsageProblem("-gen <old_file> <new_file> <patch_file>"); + GenerateEnsemblePatch(values[0], values[1], values[2]); + } else if (cmd_apply_patch) { + if (values.size() != 3) + UsageProblem("-apply <old_file> <patch_file> <new_file>"); + ApplyEnsemblePatch(values[0], values[1], values[2]); + } else if (cmd_make_bsdiff_patch) { + if (values.size() != 3) + UsageProblem("-genbsdiff <old_file> <new_file> <patch_file>"); + GenerateBSDiffPatch(values[0], values[1], values[2]); + } else if (cmd_apply_bsdiff_patch) { + if (values.size() != 3) + UsageProblem("-applybsdiff <old_file> <patch_file> <new_file>"); + ApplyBSDiffPatch(values[0], values[1], values[2]); + } else if (cmd_spread_1_adjusted || cmd_spread_1_unadjusted) { + if (values.size() != 3) + UsageProblem("-gen1[au] <old_file> <new_file> <patch_files_root>"); + DisassembleAdjustDiff(values[0], values[1], values[2], + cmd_spread_1_adjusted); + } else { + UsageProblem("No operation specified"); + } + } +} diff --git a/third_party/courgette/crc.cc b/courgette/crc.cc index 6bf51fe..60c8c93 100644 --- a/third_party/courgette/crc.cc +++ b/courgette/crc.cc @@ -1,22 +1,22 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-// Calculate Crc by calling CRC method in LZMA SDK
-
-#include "third_party/courgette/crc.h"
-
-extern "C" {
-#include "third_party/lzma_sdk/7zCrc.h"
-}
-
-namespace courgette {
-
-uint32 CalculateCrc(const uint8* buffer, size_t size) {
- // CrcGenerateTable();
- uint32 crc = 0xffffffffL;
- // crc = ~CrcCalc(buffer, size);
- return crc;
-}
-
-} // namespace
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +// Calculate Crc by calling CRC method in LZMA SDK + +#include "courgette/crc.h" + +extern "C" { +#include "third_party/lzma_sdk/7zCrc.h" +} + +namespace courgette { + +uint32 CalculateCrc(const uint8* buffer, size_t size) { + // CrcGenerateTable(); + uint32 crc = 0xffffffffL; + // crc = ~CrcCalc(buffer, size); + return crc; +} + +} // namespace diff --git a/third_party/courgette/crc.h b/courgette/crc.h index 6f4fbb0..c3662bb 100644 --- a/third_party/courgette/crc.h +++ b/courgette/crc.h @@ -1,17 +1,17 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-#ifndef COURGETTE_CRC_H_
-#define COURGETTE_CRC_H_
-
-#include "base/basictypes.h"
-
-namespace courgette {
-
-// Calculates Crc of the given buffer by calling CRC method in LZMA SDK
-//
-uint32 CalculateCrc(const uint8* buffer, size_t size);
-
-} // namespace courgette
-#endif // COURGETTE_CRC_H_
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +#ifndef COURGETTE_CRC_H_ +#define COURGETTE_CRC_H_ + +#include "base/basictypes.h" + +namespace courgette { + +// Calculates Crc of the given buffer by calling CRC method in LZMA SDK +// +uint32 CalculateCrc(const uint8* buffer, size_t size); + +} // namespace courgette +#endif // COURGETTE_CRC_H_ diff --git a/third_party/courgette/difference_estimator.cc b/courgette/difference_estimator.cc index 943eafb..e591f9b 100644 --- a/third_party/courgette/difference_estimator.cc +++ b/courgette/difference_estimator.cc @@ -1,118 +1,118 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-// We want to measure the similarity of two sequences of bytes as a surrogate
-// for measuring how well a second sequence will compress differentially to the
-// first sequence.
-
-#include "third_party/courgette/difference_estimator.h"
-
-#include "base/hash_tables.h"
-
-namespace courgette {
-
-// Our difference measure is the number of k-tuples that occur in Subject that
-// don't occur in Base.
-const int kTupleSize = 4;
-
-namespace {
-
-COMPILE_ASSERT(kTupleSize >= 4 && kTupleSize <= 8, kTupleSize_between_4_and_8);
-
-size_t HashTuple(const uint8* source) {
- size_t hash1 = *reinterpret_cast<const uint32*>(source);
- size_t hash2 = *reinterpret_cast<const uint32*>(source + kTupleSize - 4);
- size_t hash = ((hash1 * 17 + hash2 * 37) + (hash1 >> 17)) ^ (hash2 >> 23);
- return hash;
-}
-
-bool RegionsEqual(const Region& a, const Region& b) {
- if (a.length() != b.length())
- return false;
- return memcmp(a.start(), b.start(), a.length()) == 0;
-}
-
-} // anonymous namepace
-
-class DifferenceEstimator::Base {
- public:
- explicit Base(const Region& region) : region_(region) { }
-
- void Init() {
- const uint8* start = region_.start();
- const uint8* end = region_.end() - (kTupleSize - 1);
- for (const uint8* p = start; p < end; ++p) {
- size_t hash = HashTuple(p);
- hashes_.insert(hash);
- }
- }
-
- const Region& region() const { return region_; }
-
- private:
- Region region_;
- base::hash_set<size_t> hashes_;
-
- friend class DifferenceEstimator;
- DISALLOW_COPY_AND_ASSIGN(Base);
-};
-
-class DifferenceEstimator::Subject {
- public:
- explicit Subject(const Region& region) : region_(region) {}
-
- const Region& region() const { return region_; }
-
- private:
- Region region_;
-
- DISALLOW_COPY_AND_ASSIGN(Subject);
-};
-
-DifferenceEstimator::DifferenceEstimator() {}
-
-DifferenceEstimator::~DifferenceEstimator() {
- for (size_t i = 0; i < owned_bases_.size(); ++i)
- delete owned_bases_[i];
- for (size_t i = 0; i < owned_subjects_.size(); ++i)
- delete owned_subjects_[i];
-}
-
-DifferenceEstimator::Base* DifferenceEstimator::MakeBase(const Region& region) {
- Base* base = new Base(region);
- base->Init();
- owned_bases_.push_back(base);
- return base;
-}
-
-DifferenceEstimator::Subject* DifferenceEstimator::MakeSubject(
- const Region& region) {
- Subject* subject = new Subject(region);
- owned_subjects_.push_back(subject);
- return subject;
-}
-
-size_t DifferenceEstimator::Measure(Base* base, Subject* subject) {
- size_t mismatches = 0;
- const uint8* start = subject->region().start();
- const uint8* end = subject->region().end() - (kTupleSize - 1);
-
- const uint8* p = start;
- while (p < end) {
- size_t hash = HashTuple(p);
- if (base->hashes_.find(hash) == base->hashes_.end()) {
- ++mismatches;
- }
- p += 1;
- }
-
- if (mismatches == 0) {
- if (RegionsEqual(base->region(), subject->region()))
- return 0;
- }
- ++mismatches; // Guarantee not zero.
- return mismatches;
-}
-
-} // namespace
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +// We want to measure the similarity of two sequences of bytes as a surrogate +// for measuring how well a second sequence will compress differentially to the +// first sequence. + +#include "courgette/difference_estimator.h" + +#include "base/hash_tables.h" + +namespace courgette { + +// Our difference measure is the number of k-tuples that occur in Subject that +// don't occur in Base. +const int kTupleSize = 4; + +namespace { + +COMPILE_ASSERT(kTupleSize >= 4 && kTupleSize <= 8, kTupleSize_between_4_and_8); + +size_t HashTuple(const uint8* source) { + size_t hash1 = *reinterpret_cast<const uint32*>(source); + size_t hash2 = *reinterpret_cast<const uint32*>(source + kTupleSize - 4); + size_t hash = ((hash1 * 17 + hash2 * 37) + (hash1 >> 17)) ^ (hash2 >> 23); + return hash; +} + +bool RegionsEqual(const Region& a, const Region& b) { + if (a.length() != b.length()) + return false; + return memcmp(a.start(), b.start(), a.length()) == 0; +} + +} // anonymous namepace + +class DifferenceEstimator::Base { + public: + explicit Base(const Region& region) : region_(region) { } + + void Init() { + const uint8* start = region_.start(); + const uint8* end = region_.end() - (kTupleSize - 1); + for (const uint8* p = start; p < end; ++p) { + size_t hash = HashTuple(p); + hashes_.insert(hash); + } + } + + const Region& region() const { return region_; } + + private: + Region region_; + base::hash_set<size_t> hashes_; + + friend class DifferenceEstimator; + DISALLOW_COPY_AND_ASSIGN(Base); +}; + +class DifferenceEstimator::Subject { + public: + explicit Subject(const Region& region) : region_(region) {} + + const Region& region() const { return region_; } + + private: + Region region_; + + DISALLOW_COPY_AND_ASSIGN(Subject); +}; + +DifferenceEstimator::DifferenceEstimator() {} + +DifferenceEstimator::~DifferenceEstimator() { + for (size_t i = 0; i < owned_bases_.size(); ++i) + delete owned_bases_[i]; + for (size_t i = 0; i < owned_subjects_.size(); ++i) + delete owned_subjects_[i]; +} + +DifferenceEstimator::Base* DifferenceEstimator::MakeBase(const Region& region) { + Base* base = new Base(region); + base->Init(); + owned_bases_.push_back(base); + return base; +} + +DifferenceEstimator::Subject* DifferenceEstimator::MakeSubject( + const Region& region) { + Subject* subject = new Subject(region); + owned_subjects_.push_back(subject); + return subject; +} + +size_t DifferenceEstimator::Measure(Base* base, Subject* subject) { + size_t mismatches = 0; + const uint8* start = subject->region().start(); + const uint8* end = subject->region().end() - (kTupleSize - 1); + + const uint8* p = start; + while (p < end) { + size_t hash = HashTuple(p); + if (base->hashes_.find(hash) == base->hashes_.end()) { + ++mismatches; + } + p += 1; + } + + if (mismatches == 0) { + if (RegionsEqual(base->region(), subject->region())) + return 0; + } + ++mismatches; // Guarantee not zero. + return mismatches; +} + +} // namespace diff --git a/third_party/courgette/difference_estimator.h b/courgette/difference_estimator.h index bd550ec..295ff93 100644 --- a/third_party/courgette/difference_estimator.h +++ b/courgette/difference_estimator.h @@ -1,58 +1,58 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-// A DifferenceEstimator class provides a means for quickly estimating the
-// difference between two regions of memory.
-
-#ifndef COURGETTE_DIFFERENCE_ESTIMATOR_H_
-#define COURGETTE_DIFFERENCE_ESTIMATOR_H_
-
-#include <vector>
-
-#include "third_party/courgette/region.h"
-
-namespace courgette {
-
-// A DifferenceEstimator simplifies the task of determining which 'Subject' byte
-// strings (stored in regions of memory) are good matches to existing 'Base'
-// regions. The ultimate measure would be to try full differential compression
-// and measure the output size, but an estimate that correlates well with the
-// full compression is more efficient.
-//
-// The measure is asymmetric, if the Subject is a small substring of the Base
-// then it should match very well.
-//
-// The comparison is staged: first make Base and Subject objects for the regions
-// and then call 'Measure' to get the estimate. The staging allows multiple
-// comparisons to be more efficient by precomputing information used in the
-// comparison.
-//
-class DifferenceEstimator {
- public:
- DifferenceEstimator();
- ~DifferenceEstimator();
-
- class Base;
- class Subject;
-
- // This DifferenceEstimator owns the objects returned by MakeBase and
- // MakeSubject. Caller continues to own memory at |region| and must not free
- // it until ~DifferenceEstimator has been called.
- Base* MakeBase(const Region& region);
- Subject* MakeSubject(const Region& region);
-
- // Returns a value correlated with the size of the bsdiff or xdelta difference
- // from |base| to |subject|. Returns zero iff the base and subject regions
- // are bytewise identical.
- size_t Measure(Base* base, Subject* subject);
-
- private:
- std::vector<Base*> owned_bases_;
- std::vector<Subject*> owned_subjects_;
- DISALLOW_COPY_AND_ASSIGN(DifferenceEstimator);
-};
-
-} // namespace
-
-#endif // COURGETTE_DIFFERENCE_ESTIMATOR_H_
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +// A DifferenceEstimator class provides a means for quickly estimating the +// difference between two regions of memory. + +#ifndef COURGETTE_DIFFERENCE_ESTIMATOR_H_ +#define COURGETTE_DIFFERENCE_ESTIMATOR_H_ + +#include <vector> + +#include "courgette/region.h" + +namespace courgette { + +// A DifferenceEstimator simplifies the task of determining which 'Subject' byte +// strings (stored in regions of memory) are good matches to existing 'Base' +// regions. The ultimate measure would be to try full differential compression +// and measure the output size, but an estimate that correlates well with the +// full compression is more efficient. +// +// The measure is asymmetric, if the Subject is a small substring of the Base +// then it should match very well. +// +// The comparison is staged: first make Base and Subject objects for the regions +// and then call 'Measure' to get the estimate. The staging allows multiple +// comparisons to be more efficient by precomputing information used in the +// comparison. +// +class DifferenceEstimator { + public: + DifferenceEstimator(); + ~DifferenceEstimator(); + + class Base; + class Subject; + + // This DifferenceEstimator owns the objects returned by MakeBase and + // MakeSubject. Caller continues to own memory at |region| and must not free + // it until ~DifferenceEstimator has been called. + Base* MakeBase(const Region& region); + Subject* MakeSubject(const Region& region); + + // Returns a value correlated with the size of the bsdiff or xdelta difference + // from |base| to |subject|. Returns zero iff the base and subject regions + // are bytewise identical. + size_t Measure(Base* base, Subject* subject); + + private: + std::vector<Base*> owned_bases_; + std::vector<Subject*> owned_subjects_; + DISALLOW_COPY_AND_ASSIGN(DifferenceEstimator); +}; + +} // namespace + +#endif // COURGETTE_DIFFERENCE_ESTIMATOR_H_ diff --git a/third_party/courgette/difference_estimator_unittest.cc b/courgette/difference_estimator_unittest.cc index 6cc8afd..913752b 100644 --- a/third_party/courgette/difference_estimator_unittest.cc +++ b/courgette/difference_estimator_unittest.cc @@ -1,58 +1,58 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-#include "third_party/courgette/difference_estimator.h"
-
-#include <string>
-
-#include "testing/gtest/include/gtest/gtest.h"
-#include "third_party/courgette/region.h"
-
-using courgette::DifferenceEstimator;
-using courgette::Region;
-
-TEST(DifferenceEstimatorTest, TestSame) {
- static const char kString1[] = "Hello world";
- // kString2 is stack allocated to prevent string sharing.
- const char kString2[] = "Hello world";
- DifferenceEstimator difference_estimator;
- DifferenceEstimator::Base* base =
- difference_estimator.MakeBase(Region(kString1, sizeof(kString1)));
- DifferenceEstimator::Subject* subject =
- difference_estimator.MakeSubject(Region(kString2, sizeof(kString2)));
- EXPECT_EQ(0, difference_estimator.Measure(base, subject));
-}
-
-TEST(DifferenceEstimatorTest, TestDifferent) {
- static const char kString1[] = "Hello world";
- static const char kString2[] = "Hello universe";
- DifferenceEstimator difference_estimator;
- DifferenceEstimator::Base* base =
- difference_estimator.MakeBase(Region(kString1, sizeof(kString1)));
- DifferenceEstimator::Subject* subject =
- difference_estimator.MakeSubject(Region(kString2, sizeof(kString2)));
- EXPECT_EQ(10, difference_estimator.Measure(base, subject));
-}
-
-TEST(DifferenceEstimatorTest, TestDifferentSuperstring) {
- static const char kString1[] = "abcdabcdabcd";
- static const char kString2[] = "abcdabcdabcdabcd";
- DifferenceEstimator difference_estimator;
- DifferenceEstimator::Base* base =
- difference_estimator.MakeBase(Region(kString1, sizeof(kString1)-1));
- DifferenceEstimator::Subject* subject =
- difference_estimator.MakeSubject(Region(kString2, sizeof(kString2)-1));
- EXPECT_EQ(1, difference_estimator.Measure(base, subject));
-}
-
-TEST(DifferenceEstimatorTest, TestDifferentSubstring) {
- static const char kString1[] = "abcdabcdabcdabcd";
- static const char kString2[] = "abcdabcdabcd";
- DifferenceEstimator difference_estimator;
- DifferenceEstimator::Base* base =
- difference_estimator.MakeBase(Region(kString1, sizeof(kString1)-1));
- DifferenceEstimator::Subject* subject =
- difference_estimator.MakeSubject(Region(kString2, sizeof(kString2)-1));
- EXPECT_EQ(1, difference_estimator.Measure(base, subject));
-}
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +#include "courgette/difference_estimator.h" + +#include <string> + +#include "testing/gtest/include/gtest/gtest.h" +#include "courgette/region.h" + +using courgette::DifferenceEstimator; +using courgette::Region; + +TEST(DifferenceEstimatorTest, TestSame) { + static const char kString1[] = "Hello world"; + // kString2 is stack allocated to prevent string sharing. + const char kString2[] = "Hello world"; + DifferenceEstimator difference_estimator; + DifferenceEstimator::Base* base = + difference_estimator.MakeBase(Region(kString1, sizeof(kString1))); + DifferenceEstimator::Subject* subject = + difference_estimator.MakeSubject(Region(kString2, sizeof(kString2))); + EXPECT_EQ(0, difference_estimator.Measure(base, subject)); +} + +TEST(DifferenceEstimatorTest, TestDifferent) { + static const char kString1[] = "Hello world"; + static const char kString2[] = "Hello universe"; + DifferenceEstimator difference_estimator; + DifferenceEstimator::Base* base = + difference_estimator.MakeBase(Region(kString1, sizeof(kString1))); + DifferenceEstimator::Subject* subject = + difference_estimator.MakeSubject(Region(kString2, sizeof(kString2))); + EXPECT_EQ(10, difference_estimator.Measure(base, subject)); +} + +TEST(DifferenceEstimatorTest, TestDifferentSuperstring) { + static const char kString1[] = "abcdabcdabcd"; + static const char kString2[] = "abcdabcdabcdabcd"; + DifferenceEstimator difference_estimator; + DifferenceEstimator::Base* base = + difference_estimator.MakeBase(Region(kString1, sizeof(kString1)-1)); + DifferenceEstimator::Subject* subject = + difference_estimator.MakeSubject(Region(kString2, sizeof(kString2)-1)); + EXPECT_EQ(1, difference_estimator.Measure(base, subject)); +} + +TEST(DifferenceEstimatorTest, TestDifferentSubstring) { + static const char kString1[] = "abcdabcdabcdabcd"; + static const char kString2[] = "abcdabcdabcd"; + DifferenceEstimator difference_estimator; + DifferenceEstimator::Base* base = + difference_estimator.MakeBase(Region(kString1, sizeof(kString1)-1)); + DifferenceEstimator::Subject* subject = + difference_estimator.MakeSubject(Region(kString2, sizeof(kString2)-1)); + EXPECT_EQ(1, difference_estimator.Measure(base, subject)); +} diff --git a/third_party/courgette/disassembler.cc b/courgette/disassembler.cc index 6ebb2da..6395d18 100644 --- a/third_party/courgette/disassembler.cc +++ b/courgette/disassembler.cc @@ -1,436 +1,436 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-#include "third_party/courgette/disassembler.h"
-
-#include <algorithm>
-#include <iostream>
-#include <string>
-#include <vector>
-
-#include "base/basictypes.h"
-#include "base/logging.h"
-
-#include "third_party/courgette/assembly_program.h"
-#include "third_party/courgette/courgette.h"
-#include "third_party/courgette/encoded_program.h"
-#include "third_party/courgette/image_info.h"
-
-// COURGETTE_HISTOGRAM_TARGETS prints out a histogram of how frequently
-// different target addresses are referenced. Purely for debugging.
-#define COURGETTE_HISTOGRAM_TARGETS 0
-
-namespace courgette {
-
-class DisassemblerWin32X86 : public Disassembler {
- public:
- explicit DisassemblerWin32X86(PEInfo* pe_info)
- : pe_info_(pe_info),
- incomplete_disassembly_(false) {
- }
-
- virtual bool Disassemble(AssemblyProgram* target);
-
- virtual void Destroy() { delete this; }
-
- protected:
- PEInfo& pe_info() { return *pe_info_; }
-
- void ParseFile(AssemblyProgram* target);
- bool ParseAbs32Relocs();
- void ParseRel32RelocsFromSections();
- void ParseRel32RelocsFromSection(const Section* section);
-
- void ParseNonSectionFileRegion(uint32 start_file_offset,
- uint32 end_file_offset,
- AssemblyProgram* program);
- void ParseFileRegion(const Section* section,
- uint32 start_file_offset, uint32 end_file_offset,
- AssemblyProgram* program);
-
-#if COURGETTE_HISTOGRAM_TARGETS
- void HistogramTargets(const char* kind, const std::map<RVA, int>& map);
-#endif
-
- PEInfo* pe_info_;
- bool incomplete_disassembly_; // 'true' if can leave out 'uninteresting' bits
-
- std::vector<RVA> abs32_locations_;
- std::vector<RVA> rel32_locations_;
-
-#if COURGETTE_HISTOGRAM_TARGETS
- std::map<RVA, int> abs32_target_rvas_;
- std::map<RVA, int> rel32_target_rvas_;
-#endif
-};
-
-bool DisassemblerWin32X86::Disassemble(AssemblyProgram* target) {
- if (!pe_info().ok())
- return false;
-
- target->set_image_base(pe_info().image_base());
-
- if (!ParseAbs32Relocs())
- return false;
-
- ParseRel32RelocsFromSections();
-
- ParseFile(target);
-
- target->DefaultAssignIndexes();
- return true;
-}
-
-static uint32 Read32LittleEndian(const void* address) {
- return *reinterpret_cast<const uint32*>(address);
-}
-
-bool DisassemblerWin32X86::ParseAbs32Relocs() {
- abs32_locations_.clear();
- if (!pe_info().ParseRelocs(&abs32_locations_))
- return false;
-
- std::sort(abs32_locations_.begin(), abs32_locations_.end());
-
-#if COURGETTE_HISTOGRAM_TARGETS
- for (size_t i = 0; i < abs32_locations_.size(); ++i) {
- RVA rva = abs32_locations_[i];
- // The 4 bytes at the relocation are a reference to some address.
- uint32 target_address = Read32LittleEndian(pe_info().RVAToPointer(rva));
- ++abs32_target_rvas_[target_address - pe_info().image_base()];
- }
-#endif
- return true;
-}
-
-void DisassemblerWin32X86::ParseRel32RelocsFromSections() {
- uint32 file_offset = 0;
- while (file_offset < pe_info().length()) {
- const Section* section = pe_info().FindNextSection(file_offset);
- if (section == NULL)
- break;
- if (file_offset < section->file_offset_of_raw_data)
- file_offset = section->file_offset_of_raw_data;
- ParseRel32RelocsFromSection(section);
- file_offset += section->size_of_raw_data;
- }
- std::sort(rel32_locations_.begin(), rel32_locations_.end());
-
-#if COURGETTE_HISTOGRAM_TARGETS
- LOG(INFO) << "abs32_locations_ " << abs32_locations_.size();
- LOG(INFO) << "rel32_locations_ " << rel32_locations_.size();
- LOG(INFO) << "abs32_target_rvas_ " << abs32_target_rvas_.size();
- LOG(INFO) << "rel32_target_rvas_ " << rel32_target_rvas_.size();
-
- int common = 0;
- std::map<RVA, int>::iterator abs32_iter = abs32_target_rvas_.begin();
- std::map<RVA, int>::iterator rel32_iter = rel32_target_rvas_.begin();
- while (abs32_iter != abs32_target_rvas_.end() &&
- rel32_iter != rel32_target_rvas_.end()) {
- if (abs32_iter->first < rel32_iter->first)
- ++abs32_iter;
- else if (rel32_iter->first < abs32_iter->first)
- ++rel32_iter;
- else {
- ++common;
- ++abs32_iter;
- ++rel32_iter;
- }
- }
- LOG(INFO) << "common " << common;
-#endif
-}
-
-void DisassemblerWin32X86::ParseRel32RelocsFromSection(const Section* section) {
- // TODO(sra): use characteristic.
- bool isCode = strcmp(section->name, ".text") == 0;
- if (!isCode)
- return;
-
- uint32 start_file_offset = section->file_offset_of_raw_data;
- uint32 end_file_offset = start_file_offset + section->size_of_raw_data;
- RVA relocs_start_rva = pe_info().base_relocation_table().address_;
-
- const uint8* start_pointer = pe_info().FileOffsetToPointer(start_file_offset);
- const uint8* end_pointer = pe_info().FileOffsetToPointer(end_file_offset);
-
- RVA start_rva = pe_info().FileOffsetToRVA(start_file_offset);
- RVA end_rva = start_rva + section->virtual_size;
-
- // Quick way to convert from Pointer to RVA within a single Section is to
- // subtract 'pointer_to_rva'.
- const uint8* const adjust_pointer_to_rva = start_pointer - start_rva;
-
- std::vector<RVA>::iterator abs32_pos = abs32_locations_.begin();
-
- // Find the rel32 relocations.
- const uint8* p = start_pointer;
- while (p < end_pointer) {
- RVA current_rva = p - adjust_pointer_to_rva;
- if (current_rva == relocs_start_rva) {
- uint32 relocs_size = pe_info().base_relocation_table().size_;
- if (relocs_size) {
- p += relocs_size;
- continue;
- }
- }
-
- //while (abs32_pos != abs32_locations_.end() && *abs32_pos < current_rva)
- // ++abs32_pos;
-
- // Heuristic discovery of rel32 locations in instruction stream: are the
- // next few bytes the start of an instruction containing a rel32
- // addressing mode?
- const uint8* rel32 = NULL;
-
- if (p + 5 < end_pointer) {
- if (*p == 0xE8 || *p == 0xE9) { // jmp rel32 and call rel32
- rel32 = p + 1;
- }
- }
- if (p + 6 < end_pointer) {
- if (*p == 0x0F && (*(p+1) & 0xF0) == 0x80) { // Jcc long form
- if (p[1] != 0x8A && p[1] != 0x8B) // JPE/JPO unlikely
- rel32 = p + 2;
- }
- }
- if (rel32) {
- RVA rel32_rva = rel32 - adjust_pointer_to_rva;
-
- // Is there an abs32 reloc overlapping the candidate?
- while (abs32_pos != abs32_locations_.end() && *abs32_pos < rel32_rva - 3)
- ++abs32_pos;
- // Now: (*abs32_pos > rel32_rva - 4) i.e. the lowest addressed 4-byte
- // region that could overlap rel32_rva.
- if (abs32_pos != abs32_locations_.end()) {
- if (*abs32_pos < rel32_rva + 4) {
- // Beginning of abs32 reloc is before end of rel32 reloc so they
- // overlap. Skip four bytes past the abs32 reloc.
- p += (*abs32_pos + 4) - current_rva;
- continue;
- }
- }
-
- RVA target_rva = rel32_rva + 4 + Read32LittleEndian(rel32);
- // To be valid, rel32 target must be within image, and within this
- // section.
- if (pe_info().IsValidRVA(target_rva) &&
- start_rva <= target_rva && target_rva < end_rva) {
- rel32_locations_.push_back(rel32_rva);
-#if COURGETTE_HISTOGRAM_TARGETS
- ++rel32_target_rvas_[target_rva];
-#endif
- p += 4;
- continue;
- }
- }
- p += 1;
- }
-}
-
-void DisassemblerWin32X86::ParseFile(AssemblyProgram* program) {
- // Walk all the bytes in the file, whether or not in a section.
- uint32 file_offset = 0;
- while (file_offset < pe_info().length()) {
- const Section* section = pe_info().FindNextSection(file_offset);
- if (section == NULL) {
- // No more sections. There should not be extra stuff following last
- // section.
- // ParseNonSectionFileRegion(file_offset, pe_info().length(), program);
- break;
- }
- if (file_offset < section->file_offset_of_raw_data) {
- uint32 section_start_offset = section->file_offset_of_raw_data;
- ParseNonSectionFileRegion(file_offset, section_start_offset, program);
- file_offset = section_start_offset;
- }
- uint32 end = file_offset + section->size_of_raw_data;
- ParseFileRegion(section, file_offset, end, program);
- file_offset = end;
- }
-
-#if COURGETTE_HISTOGRAM_TARGETS
- HistogramTargets("abs32 relocs", abs32_target_rvas_);
- HistogramTargets("rel32 relocs", rel32_target_rvas_);
-#endif
-}
-
-void DisassemblerWin32X86::ParseNonSectionFileRegion(
- uint32 start_file_offset,
- uint32 end_file_offset,
- AssemblyProgram* program) {
- if (incomplete_disassembly_)
- return;
-
- const uint8* start = pe_info().FileOffsetToPointer(start_file_offset);
- const uint8* end = pe_info().FileOffsetToPointer(end_file_offset);
-
- const uint8* p = start;
-
- while (p < end) {
- program->EmitByteInstruction(*p);
- ++p;
- }
-}
-
-void DisassemblerWin32X86::ParseFileRegion(
- const Section* section,
- uint32 start_file_offset, uint32 end_file_offset,
- AssemblyProgram* program) {
- RVA relocs_start_rva = pe_info().base_relocation_table().address_;
-
- const uint8* start_pointer = pe_info().FileOffsetToPointer(start_file_offset);
- const uint8* end_pointer = pe_info().FileOffsetToPointer(end_file_offset);
-
- RVA start_rva = pe_info().FileOffsetToRVA(start_file_offset);
- RVA end_rva = start_rva + section->virtual_size;
-
- // Quick way to convert from Pointer to RVA within a single Section is to
- // subtract 'pointer_to_rva'.
- const uint8* const adjust_pointer_to_rva = start_pointer - start_rva;
-
- std::vector<RVA>::iterator rel32_pos = rel32_locations_.begin();
- std::vector<RVA>::iterator abs32_pos = abs32_locations_.begin();
-
- program->EmitOriginInstruction(start_rva);
-
- const uint8* p = start_pointer;
-
- while (p < end_pointer) {
- RVA current_rva = p - adjust_pointer_to_rva;
-
- // The base relocation table is usually in the .relocs section, but it could
- // actually be anywhere. Make sure we skip it because we will regenerate it
- // during assembly.
- if (current_rva == relocs_start_rva) {
- program->EmitMakeRelocsInstruction();
- uint32 relocs_size = pe_info().base_relocation_table().size_;
- if (relocs_size) {
- p += relocs_size;
- continue;
- }
- }
-
- while (abs32_pos != abs32_locations_.end() && *abs32_pos < current_rva)
- ++abs32_pos;
-
- if (abs32_pos != abs32_locations_.end() && *abs32_pos == current_rva) {
- uint32 target_address = Read32LittleEndian(p);
- RVA target_rva = target_address - pe_info().image_base();
- // TODO(sra): target could be Label+offset. It is not clear how to guess
- // which it might be. We assume offset==0.
- program->EmitAbs32(program->FindOrMakeAbs32Label(target_rva));
- p += 4;
- continue;
- }
-
- while (rel32_pos != rel32_locations_.end() && *rel32_pos < current_rva)
- ++rel32_pos;
-
- if (rel32_pos != rel32_locations_.end() && *rel32_pos == current_rva) {
- RVA target_rva = current_rva + 4 + Read32LittleEndian(p);
- program->EmitRel32(program->FindOrMakeRel32Label(target_rva));
- p += 4;
- continue;
- }
-
- if (incomplete_disassembly_) {
- if ((abs32_pos == abs32_locations_.end() || end_rva <= *abs32_pos) &&
- (rel32_pos == rel32_locations_.end() || end_rva <= *rel32_pos) &&
- (end_rva <= relocs_start_rva || current_rva >= relocs_start_rva)) {
- // No more relocs in this section, don't bother encoding bytes.
- break;
- }
- }
-
- program->EmitByteInstruction(*p);
- p += 1;
- }
-}
-
-#if COURGETTE_HISTOGRAM_TARGETS
-// Histogram is printed to std::cout. It is purely for debugging the algorithm
-// and is only enabled manually in 'exploration' builds. I don't want to add
-// command-line configuration for this feature because this code has to be
-// small, which means compiled-out.
-void DisassemblerWin32X86::HistogramTargets(const char* kind,
- const std::map<RVA, int>& map) {
- int total = 0;
- std::map<int, std::vector<RVA> > h;
- for (std::map<RVA, int>::const_iterator p = map.begin();
- p != map.end();
- ++p) {
- h[p->second].push_back(p->first);
- total += p->second;
- }
-
- std::cout << total << " " << kind << " to "
- << map.size() << " unique targets" << std::endl;
-
- std::cout << "indegree: #targets-with-indegree (example)" << std::endl;
- const int kFirstN = 15;
- bool someSkipped = false;
- int index = 0;
- for (std::map<int, std::vector<RVA> >::reverse_iterator p = h.rbegin();
- p != h.rend();
- ++p) {
- ++index;
- if (index <= kFirstN || p->first <= 3) {
- if (someSkipped) {
- std::cout << "..." << std::endl;
- }
- size_t count = p->second.size();
- std::cout << std::dec << p->first << ": " << count;
- if (count <= 2) {
- for (size_t i = 0; i < count; ++i)
- std::cout << " " << pe_info().DescribeRVA(p->second[i]);
- }
- std::cout << std::endl;
- someSkipped = false;
- } else {
- someSkipped = true;
- }
- }
-}
-#endif // COURGETTE_HISTOGRAM_TARGETS
-
-Disassembler* Disassembler::MakeDisassemberWin32X86(PEInfo* pe_info) {
- return new DisassemblerWin32X86(pe_info);
-}
-
-////////////////////////////////////////////////////////////////////////////////
-
-Status ParseWin32X86PE(const void* buffer, size_t length,
- AssemblyProgram** output) {
- *output = NULL;
-
- PEInfo* pe_info = new PEInfo();
- pe_info->Init(buffer, length);
-
- if (!pe_info->ParseHeader()) {
- delete pe_info;
- return C_INPUT_NOT_RECOGNIZED;
- }
-
- Disassembler* disassembler = Disassembler::MakeDisassemberWin32X86(pe_info);
- AssemblyProgram* program = new AssemblyProgram();
-
- if (!disassembler->Disassemble(program)) {
- delete program;
- disassembler->Destroy();
- delete pe_info;
- return C_DISASSEMBLY_FAILED;
- }
-
- disassembler->Destroy();
- delete pe_info;
- *output = program;
- return C_OK;
-}
-
-void DeleteAssemblyProgram(AssemblyProgram* program) {
- delete program;
-}
-
-} // namespace courgette
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +#include "courgette/disassembler.h" + +#include <algorithm> +#include <iostream> +#include <string> +#include <vector> + +#include "base/basictypes.h" +#include "base/logging.h" + +#include "courgette/assembly_program.h" +#include "courgette/courgette.h" +#include "courgette/encoded_program.h" +#include "courgette/image_info.h" + +// COURGETTE_HISTOGRAM_TARGETS prints out a histogram of how frequently +// different target addresses are referenced. Purely for debugging. +#define COURGETTE_HISTOGRAM_TARGETS 0 + +namespace courgette { + +class DisassemblerWin32X86 : public Disassembler { + public: + explicit DisassemblerWin32X86(PEInfo* pe_info) + : pe_info_(pe_info), + incomplete_disassembly_(false) { + } + + virtual bool Disassemble(AssemblyProgram* target); + + virtual void Destroy() { delete this; } + + protected: + PEInfo& pe_info() { return *pe_info_; } + + void ParseFile(AssemblyProgram* target); + bool ParseAbs32Relocs(); + void ParseRel32RelocsFromSections(); + void ParseRel32RelocsFromSection(const Section* section); + + void ParseNonSectionFileRegion(uint32 start_file_offset, + uint32 end_file_offset, + AssemblyProgram* program); + void ParseFileRegion(const Section* section, + uint32 start_file_offset, uint32 end_file_offset, + AssemblyProgram* program); + +#if COURGETTE_HISTOGRAM_TARGETS + void HistogramTargets(const char* kind, const std::map<RVA, int>& map); +#endif + + PEInfo* pe_info_; + bool incomplete_disassembly_; // 'true' if can leave out 'uninteresting' bits + + std::vector<RVA> abs32_locations_; + std::vector<RVA> rel32_locations_; + +#if COURGETTE_HISTOGRAM_TARGETS + std::map<RVA, int> abs32_target_rvas_; + std::map<RVA, int> rel32_target_rvas_; +#endif +}; + +bool DisassemblerWin32X86::Disassemble(AssemblyProgram* target) { + if (!pe_info().ok()) + return false; + + target->set_image_base(pe_info().image_base()); + + if (!ParseAbs32Relocs()) + return false; + + ParseRel32RelocsFromSections(); + + ParseFile(target); + + target->DefaultAssignIndexes(); + return true; +} + +static uint32 Read32LittleEndian(const void* address) { + return *reinterpret_cast<const uint32*>(address); +} + +bool DisassemblerWin32X86::ParseAbs32Relocs() { + abs32_locations_.clear(); + if (!pe_info().ParseRelocs(&abs32_locations_)) + return false; + + std::sort(abs32_locations_.begin(), abs32_locations_.end()); + +#if COURGETTE_HISTOGRAM_TARGETS + for (size_t i = 0; i < abs32_locations_.size(); ++i) { + RVA rva = abs32_locations_[i]; + // The 4 bytes at the relocation are a reference to some address. + uint32 target_address = Read32LittleEndian(pe_info().RVAToPointer(rva)); + ++abs32_target_rvas_[target_address - pe_info().image_base()]; + } +#endif + return true; +} + +void DisassemblerWin32X86::ParseRel32RelocsFromSections() { + uint32 file_offset = 0; + while (file_offset < pe_info().length()) { + const Section* section = pe_info().FindNextSection(file_offset); + if (section == NULL) + break; + if (file_offset < section->file_offset_of_raw_data) + file_offset = section->file_offset_of_raw_data; + ParseRel32RelocsFromSection(section); + file_offset += section->size_of_raw_data; + } + std::sort(rel32_locations_.begin(), rel32_locations_.end()); + +#if COURGETTE_HISTOGRAM_TARGETS + LOG(INFO) << "abs32_locations_ " << abs32_locations_.size(); + LOG(INFO) << "rel32_locations_ " << rel32_locations_.size(); + LOG(INFO) << "abs32_target_rvas_ " << abs32_target_rvas_.size(); + LOG(INFO) << "rel32_target_rvas_ " << rel32_target_rvas_.size(); + + int common = 0; + std::map<RVA, int>::iterator abs32_iter = abs32_target_rvas_.begin(); + std::map<RVA, int>::iterator rel32_iter = rel32_target_rvas_.begin(); + while (abs32_iter != abs32_target_rvas_.end() && + rel32_iter != rel32_target_rvas_.end()) { + if (abs32_iter->first < rel32_iter->first) + ++abs32_iter; + else if (rel32_iter->first < abs32_iter->first) + ++rel32_iter; + else { + ++common; + ++abs32_iter; + ++rel32_iter; + } + } + LOG(INFO) << "common " << common; +#endif +} + +void DisassemblerWin32X86::ParseRel32RelocsFromSection(const Section* section) { + // TODO(sra): use characteristic. + bool isCode = strcmp(section->name, ".text") == 0; + if (!isCode) + return; + + uint32 start_file_offset = section->file_offset_of_raw_data; + uint32 end_file_offset = start_file_offset + section->size_of_raw_data; + RVA relocs_start_rva = pe_info().base_relocation_table().address_; + + const uint8* start_pointer = pe_info().FileOffsetToPointer(start_file_offset); + const uint8* end_pointer = pe_info().FileOffsetToPointer(end_file_offset); + + RVA start_rva = pe_info().FileOffsetToRVA(start_file_offset); + RVA end_rva = start_rva + section->virtual_size; + + // Quick way to convert from Pointer to RVA within a single Section is to + // subtract 'pointer_to_rva'. + const uint8* const adjust_pointer_to_rva = start_pointer - start_rva; + + std::vector<RVA>::iterator abs32_pos = abs32_locations_.begin(); + + // Find the rel32 relocations. + const uint8* p = start_pointer; + while (p < end_pointer) { + RVA current_rva = p - adjust_pointer_to_rva; + if (current_rva == relocs_start_rva) { + uint32 relocs_size = pe_info().base_relocation_table().size_; + if (relocs_size) { + p += relocs_size; + continue; + } + } + + //while (abs32_pos != abs32_locations_.end() && *abs32_pos < current_rva) + // ++abs32_pos; + + // Heuristic discovery of rel32 locations in instruction stream: are the + // next few bytes the start of an instruction containing a rel32 + // addressing mode? + const uint8* rel32 = NULL; + + if (p + 5 < end_pointer) { + if (*p == 0xE8 || *p == 0xE9) { // jmp rel32 and call rel32 + rel32 = p + 1; + } + } + if (p + 6 < end_pointer) { + if (*p == 0x0F && (*(p+1) & 0xF0) == 0x80) { // Jcc long form + if (p[1] != 0x8A && p[1] != 0x8B) // JPE/JPO unlikely + rel32 = p + 2; + } + } + if (rel32) { + RVA rel32_rva = rel32 - adjust_pointer_to_rva; + + // Is there an abs32 reloc overlapping the candidate? + while (abs32_pos != abs32_locations_.end() && *abs32_pos < rel32_rva - 3) + ++abs32_pos; + // Now: (*abs32_pos > rel32_rva - 4) i.e. the lowest addressed 4-byte + // region that could overlap rel32_rva. + if (abs32_pos != abs32_locations_.end()) { + if (*abs32_pos < rel32_rva + 4) { + // Beginning of abs32 reloc is before end of rel32 reloc so they + // overlap. Skip four bytes past the abs32 reloc. + p += (*abs32_pos + 4) - current_rva; + continue; + } + } + + RVA target_rva = rel32_rva + 4 + Read32LittleEndian(rel32); + // To be valid, rel32 target must be within image, and within this + // section. + if (pe_info().IsValidRVA(target_rva) && + start_rva <= target_rva && target_rva < end_rva) { + rel32_locations_.push_back(rel32_rva); +#if COURGETTE_HISTOGRAM_TARGETS + ++rel32_target_rvas_[target_rva]; +#endif + p += 4; + continue; + } + } + p += 1; + } +} + +void DisassemblerWin32X86::ParseFile(AssemblyProgram* program) { + // Walk all the bytes in the file, whether or not in a section. + uint32 file_offset = 0; + while (file_offset < pe_info().length()) { + const Section* section = pe_info().FindNextSection(file_offset); + if (section == NULL) { + // No more sections. There should not be extra stuff following last + // section. + // ParseNonSectionFileRegion(file_offset, pe_info().length(), program); + break; + } + if (file_offset < section->file_offset_of_raw_data) { + uint32 section_start_offset = section->file_offset_of_raw_data; + ParseNonSectionFileRegion(file_offset, section_start_offset, program); + file_offset = section_start_offset; + } + uint32 end = file_offset + section->size_of_raw_data; + ParseFileRegion(section, file_offset, end, program); + file_offset = end; + } + +#if COURGETTE_HISTOGRAM_TARGETS + HistogramTargets("abs32 relocs", abs32_target_rvas_); + HistogramTargets("rel32 relocs", rel32_target_rvas_); +#endif +} + +void DisassemblerWin32X86::ParseNonSectionFileRegion( + uint32 start_file_offset, + uint32 end_file_offset, + AssemblyProgram* program) { + if (incomplete_disassembly_) + return; + + const uint8* start = pe_info().FileOffsetToPointer(start_file_offset); + const uint8* end = pe_info().FileOffsetToPointer(end_file_offset); + + const uint8* p = start; + + while (p < end) { + program->EmitByteInstruction(*p); + ++p; + } +} + +void DisassemblerWin32X86::ParseFileRegion( + const Section* section, + uint32 start_file_offset, uint32 end_file_offset, + AssemblyProgram* program) { + RVA relocs_start_rva = pe_info().base_relocation_table().address_; + + const uint8* start_pointer = pe_info().FileOffsetToPointer(start_file_offset); + const uint8* end_pointer = pe_info().FileOffsetToPointer(end_file_offset); + + RVA start_rva = pe_info().FileOffsetToRVA(start_file_offset); + RVA end_rva = start_rva + section->virtual_size; + + // Quick way to convert from Pointer to RVA within a single Section is to + // subtract 'pointer_to_rva'. + const uint8* const adjust_pointer_to_rva = start_pointer - start_rva; + + std::vector<RVA>::iterator rel32_pos = rel32_locations_.begin(); + std::vector<RVA>::iterator abs32_pos = abs32_locations_.begin(); + + program->EmitOriginInstruction(start_rva); + + const uint8* p = start_pointer; + + while (p < end_pointer) { + RVA current_rva = p - adjust_pointer_to_rva; + + // The base relocation table is usually in the .relocs section, but it could + // actually be anywhere. Make sure we skip it because we will regenerate it + // during assembly. + if (current_rva == relocs_start_rva) { + program->EmitMakeRelocsInstruction(); + uint32 relocs_size = pe_info().base_relocation_table().size_; + if (relocs_size) { + p += relocs_size; + continue; + } + } + + while (abs32_pos != abs32_locations_.end() && *abs32_pos < current_rva) + ++abs32_pos; + + if (abs32_pos != abs32_locations_.end() && *abs32_pos == current_rva) { + uint32 target_address = Read32LittleEndian(p); + RVA target_rva = target_address - pe_info().image_base(); + // TODO(sra): target could be Label+offset. It is not clear how to guess + // which it might be. We assume offset==0. + program->EmitAbs32(program->FindOrMakeAbs32Label(target_rva)); + p += 4; + continue; + } + + while (rel32_pos != rel32_locations_.end() && *rel32_pos < current_rva) + ++rel32_pos; + + if (rel32_pos != rel32_locations_.end() && *rel32_pos == current_rva) { + RVA target_rva = current_rva + 4 + Read32LittleEndian(p); + program->EmitRel32(program->FindOrMakeRel32Label(target_rva)); + p += 4; + continue; + } + + if (incomplete_disassembly_) { + if ((abs32_pos == abs32_locations_.end() || end_rva <= *abs32_pos) && + (rel32_pos == rel32_locations_.end() || end_rva <= *rel32_pos) && + (end_rva <= relocs_start_rva || current_rva >= relocs_start_rva)) { + // No more relocs in this section, don't bother encoding bytes. + break; + } + } + + program->EmitByteInstruction(*p); + p += 1; + } +} + +#if COURGETTE_HISTOGRAM_TARGETS +// Histogram is printed to std::cout. It is purely for debugging the algorithm +// and is only enabled manually in 'exploration' builds. I don't want to add +// command-line configuration for this feature because this code has to be +// small, which means compiled-out. +void DisassemblerWin32X86::HistogramTargets(const char* kind, + const std::map<RVA, int>& map) { + int total = 0; + std::map<int, std::vector<RVA> > h; + for (std::map<RVA, int>::const_iterator p = map.begin(); + p != map.end(); + ++p) { + h[p->second].push_back(p->first); + total += p->second; + } + + std::cout << total << " " << kind << " to " + << map.size() << " unique targets" << std::endl; + + std::cout << "indegree: #targets-with-indegree (example)" << std::endl; + const int kFirstN = 15; + bool someSkipped = false; + int index = 0; + for (std::map<int, std::vector<RVA> >::reverse_iterator p = h.rbegin(); + p != h.rend(); + ++p) { + ++index; + if (index <= kFirstN || p->first <= 3) { + if (someSkipped) { + std::cout << "..." << std::endl; + } + size_t count = p->second.size(); + std::cout << std::dec << p->first << ": " << count; + if (count <= 2) { + for (size_t i = 0; i < count; ++i) + std::cout << " " << pe_info().DescribeRVA(p->second[i]); + } + std::cout << std::endl; + someSkipped = false; + } else { + someSkipped = true; + } + } +} +#endif // COURGETTE_HISTOGRAM_TARGETS + +Disassembler* Disassembler::MakeDisassemberWin32X86(PEInfo* pe_info) { + return new DisassemblerWin32X86(pe_info); +} + +//////////////////////////////////////////////////////////////////////////////// + +Status ParseWin32X86PE(const void* buffer, size_t length, + AssemblyProgram** output) { + *output = NULL; + + PEInfo* pe_info = new PEInfo(); + pe_info->Init(buffer, length); + + if (!pe_info->ParseHeader()) { + delete pe_info; + return C_INPUT_NOT_RECOGNIZED; + } + + Disassembler* disassembler = Disassembler::MakeDisassemberWin32X86(pe_info); + AssemblyProgram* program = new AssemblyProgram(); + + if (!disassembler->Disassemble(program)) { + delete program; + disassembler->Destroy(); + delete pe_info; + return C_DISASSEMBLY_FAILED; + } + + disassembler->Destroy(); + delete pe_info; + *output = program; + return C_OK; +} + +void DeleteAssemblyProgram(AssemblyProgram* program) { + delete program; +} + +} // namespace courgette diff --git a/third_party/courgette/disassembler.h b/courgette/disassembler.h index 595412e..fa7c908 100644 --- a/third_party/courgette/disassembler.h +++ b/courgette/disassembler.h @@ -1,39 +1,38 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-#ifndef COURGETTE_DISASSEMBLER_H_
-#define COURGETTE_DISASSEMBLER_H_
-
-#include "base/basictypes.h"
-
-namespace courgette {
-
-class AssemblyProgram;
-class PEInfo;
-
-class Disassembler {
- public:
- // Factory methods for making disassemblers for various kinds of executables.
- // We have only one so far.
-
- static Disassembler* MakeDisassemberWin32X86(PEInfo* pe_info);
-
- // Disassembles the item passed to the factory method into the output
- // parameter 'program'.
- virtual bool Disassemble(AssemblyProgram* program) = 0;
-
- // Deletes 'this' disassembler.
- virtual void Destroy() = 0;
-
- protected:
- Disassembler() {}
- virtual ~Disassembler() {}
-
- private:
- DISALLOW_COPY_AND_ASSIGN(Disassembler);
-};
-
-} // namespace courgette
-#endif // COURGETTE_DISASSEMBLER_H_
-
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +#ifndef COURGETTE_DISASSEMBLER_H_ +#define COURGETTE_DISASSEMBLER_H_ + +#include "base/basictypes.h" + +namespace courgette { + +class AssemblyProgram; +class PEInfo; + +class Disassembler { + public: + // Factory methods for making disassemblers for various kinds of executables. + // We have only one so far. + + static Disassembler* MakeDisassemberWin32X86(PEInfo* pe_info); + + // Disassembles the item passed to the factory method into the output + // parameter 'program'. + virtual bool Disassemble(AssemblyProgram* program) = 0; + + // Deletes 'this' disassembler. + virtual void Destroy() = 0; + + protected: + Disassembler() {} + virtual ~Disassembler() {} + + private: + DISALLOW_COPY_AND_ASSIGN(Disassembler); +}; + +} // namespace courgette +#endif // COURGETTE_DISASSEMBLER_H_ diff --git a/third_party/courgette/encode_decode_unittest.cc b/courgette/encode_decode_unittest.cc index 25a3913..436479d8 100644 --- a/third_party/courgette/encode_decode_unittest.cc +++ b/courgette/encode_decode_unittest.cc @@ -8,8 +8,8 @@ #include "base/file_util.h" #include "base/string_util.h" -#include "third_party/courgette/courgette.h" -#include "third_party/courgette/streams.h" +#include "courgette/courgette.h" +#include "courgette/streams.h" #include "testing/gtest/include/gtest/gtest.h" @@ -20,7 +20,6 @@ class EncodeDecodeTest : public testing::Test { private: void SetUp() { PathService::Get(base::DIR_SOURCE_ROOT, &testdata_dir_); - // file_util::AppendToPath(&testdata_dir_, L"third_party"); file_util::AppendToPath(&testdata_dir_, L"courgette"); file_util::AppendToPath(&testdata_dir_, L"testdata"); } diff --git a/third_party/courgette/encoded_program.cc b/courgette/encoded_program.cc index 0e9f164..f606f96 100644 --- a/third_party/courgette/encoded_program.cc +++ b/courgette/encoded_program.cc @@ -2,7 +2,7 @@ // Use of this source code is governed by a BSD-style license that can be // found in the LICENSE file. -#include "third_party/courgette/encoded_program.h" +#include "courgette/encoded_program.h" #include <algorithm> #include <map> @@ -12,8 +12,8 @@ #include "base/logging.h" #include "base/sys_info.h" -#include "third_party/courgette/courgette.h" -#include "third_party/courgette/streams.h" +#include "courgette/courgette.h" +#include "courgette/streams.h" namespace courgette { diff --git a/third_party/courgette/encoded_program.h b/courgette/encoded_program.h index 7331042..df1119c 100644 --- a/third_party/courgette/encoded_program.h +++ b/courgette/encoded_program.h @@ -8,7 +8,7 @@ #include <vector> #include "base/basictypes.h" -#include "third_party/courgette/image_info.h" +#include "courgette/image_info.h" namespace courgette { @@ -81,4 +81,3 @@ class EncodedProgram { } // namespace courgette #endif // COURGETTE_ENCODED_FORMAT_H_ - diff --git a/third_party/courgette/encoded_program_fuzz_unittest.cc b/courgette/encoded_program_fuzz_unittest.cc index 9cd5020..adc09d9 100644 --- a/third_party/courgette/encoded_program_fuzz_unittest.cc +++ b/courgette/encoded_program_fuzz_unittest.cc @@ -1,236 +1,235 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-// Fuzz testing for EncodedProgram serialized format and assembly.
-//
-// We would like some assurance that if an EncodedProgram is malformed we will
-// not crash. The EncodedProgram could be malformed either due to malicious
-// attack to due to an error in patch generation.
-//
-// We try a lot of arbitrary modifications to the serialized form and make sure
-// that the outcome is not a crash.
-
-#include <string>
-
-#include "base/path_service.h"
-#include "base/file_util.h"
-#include "base/string_util.h"
-#include "base/test_suite.h"
-
-#include "third_party/courgette/courgette.h"
-#include "third_party/courgette/streams.h"
-
-#include "testing/gtest/include/gtest/gtest.h"
-
-class DecodeFuzzTest : public testing::Test {
- public:
- void FuzzExe(const char *) const;
-
- private:
- virtual void SetUp() {
- PathService::Get(base::DIR_SOURCE_ROOT, &testdata_dir_);
- // testdata_dir_ = testdata_dir_.Append(L"third_party");
- testdata_dir_ = testdata_dir_.Append(L"courgette");
- testdata_dir_ = testdata_dir_.Append(L"testdata");
- }
-
- virtual void TearDown() { }
-
- void FuzzByte(const std::string& buffer, const std::string& output,
- size_t index) const;
- void FuzzBits(const std::string& buffer, const std::string& output,
- size_t index, int bits_to_flip) const;
-
- // Returns true if could assemble, false if rejected.
- bool TryAssemble(const std::string& buffer, std::string* output) const;
-
- // Returns contents of |file_name| as uninterprested bytes stored in a string.
- std::string FileContents(const char* file_name) const;
-
- // Full path name of testdata directory
- FilePath testdata_dir_;
-};
-
-// Reads a test file into a string.
-std::string DecodeFuzzTest::FileContents(const char* file_name) const {
- FilePath file_path = testdata_dir_.AppendASCII(file_name);
- std::string file_contents;
- if (!file_util::ReadFileToString(file_path, &file_contents)) {
- EXPECT_TRUE(!"Could not read test data");
- }
- return file_contents;
-}
-
-// Loads an executable and does fuzz testing in the serialized format.
-void DecodeFuzzTest::FuzzExe(const char* file_name) const {
- std::string file1 = FileContents(file_name);
-
- const void* original_buffer = file1.c_str();
- size_t original_length = file1.size();
-
- courgette::AssemblyProgram* program = NULL;
- const courgette::Status parse_status =
- courgette::ParseWin32X86PE(original_buffer, original_length, &program);
- EXPECT_EQ(courgette::C_OK, parse_status);
-
- courgette::EncodedProgram* encoded = NULL;
-
- const courgette::Status encode_status = Encode(program, &encoded);
- EXPECT_EQ(courgette::C_OK, encode_status);
-
- DeleteAssemblyProgram(program);
-
- courgette::SinkStreamSet sinks;
- const courgette::Status write_status = WriteEncodedProgram(encoded, &sinks);
- EXPECT_EQ(courgette::C_OK, write_status);
-
- DeleteEncodedProgram(encoded);
-
- courgette::SinkStream sink;
- bool can_collect = sinks.CopyTo(&sink);
- EXPECT_TRUE(can_collect);
-
- size_t length = sink.Length();
-
- std::string base_buffer(reinterpret_cast<const char*>(sink.Buffer()), length);
- std::string base_output;
- bool ok = TryAssemble(base_buffer, &base_output);
- EXPECT_EQ(true, ok);
-
- // Now we have a good serialized EncodedProgram in |base_buffer|. Time to
- // fuzz.
-
- // More intense fuzzing on the first part because it contains more control
- // information like substeam lengths.
- size_t position = 0;
- for ( ; position < 100 && position < length; position += 1) {
- FuzzByte(base_buffer, base_output, position);
- }
- // We would love to fuzz every position, but it takes too long.
- for ( ; position < length; position += 900) {
- FuzzByte(base_buffer, base_output, position);
- }
-}
-
-// FuzzByte tries to break the EncodedProgram deserializer and assembler. It
-// takes a good serialization of and EncodedProgram, flips some bits, and checks
-// that the behaviour is reasonable. It has testing checks for unreasonable
-// behaviours.
-void DecodeFuzzTest::FuzzByte(const std::string& base_buffer,
- const std::string& base_output,
- size_t index) const {
- printf("Fuzzing position %d\n", static_cast<int>(index));
-
- // The following 10 values are a compromize between run time and coverage of
- // the 255 'wrong' values at this byte position.
-
- // 0xFF flips all the bits.
- FuzzBits(base_buffer, base_output, index, 0xFF);
- // 0x7F flips the most bits without changing Varint32 framing.
- FuzzBits(base_buffer, base_output, index, 0x7F);
- // These all flip one bit.
- FuzzBits(base_buffer, base_output, index, 0x80);
- FuzzBits(base_buffer, base_output, index, 0x40);
- FuzzBits(base_buffer, base_output, index, 0x20);
- FuzzBits(base_buffer, base_output, index, 0x10);
- FuzzBits(base_buffer, base_output, index, 0x08);
- FuzzBits(base_buffer, base_output, index, 0x04);
- FuzzBits(base_buffer, base_output, index, 0x02);
- FuzzBits(base_buffer, base_output, index, 0x01);
-}
-
-// FuzzBits tries to break the EncodedProgram deserializer and assembler. It
-// takes a good serialization of and EncodedProgram, flips some bits, and checks
-// that the behaviour is reasonable.
-//
-// There are EXPECT calls to check for unreasonable behaviour. These are
-// somewhat arbitrary in that the parameters cannot easily be derived from first
-// principles. They may need updating as the serialized format evolves.
-void DecodeFuzzTest::FuzzBits(const std::string& base_buffer,
- const std::string& base_output,
- size_t index, int bits_to_flip) const {
- std::string modified_buffer = base_buffer;
- std::string modified_output;
- modified_buffer[index] ^= bits_to_flip;
-
- bool ok = TryAssemble(modified_buffer, &modified_output);
-
- if (ok) {
- // We normally expect TryAssemble to fail. But sometimes it succeeds.
- // What could have happened? We changed one byte in the serialized form:
- //
- // * If we changed one of the copied bytes, we would see a single byte
- // change in the output.
- // * If we changed an address table element, all the references to that
- // address would be different.
- // * If we changed a copy count, we would run out of data in some stream,
- // or leave data remaining, so should not be here.
- // * If we changed an origin address, it could affect all relocations based
- // off that address. If no relocations were based off the address then
- // there will be no changes.
- // * If we changed an origin address, it could cause some abs32 relocs to
- // shift from one page to the next, changing the number and layout of
- // blocks in the base relocation table.
-
- // Generated length could vary slightly due to base relocation table layout.
- // In the worst case the number of base relocation blocks doubles, approx
- // 12/4096 or 0.3% size of file.
- size_t base_length = base_output.length();
- size_t modified_length = modified_output.length();
- ptrdiff_t diff = base_length - modified_length;
- if (diff < -200 || diff > 200) {
- EXPECT_EQ(base_length, modified_length);
- }
-
- size_t changed_byte_count = 0;
- for (size_t i = 0; i < base_length && i < modified_length; ++i) {
- changed_byte_count += (base_output[i] != modified_output[i]);
- }
-
- if (index > 60) { // Beyond the origin addresses ...
- EXPECT_NE(0, changed_byte_count); // ... we expect some difference.
- }
- // Currently all changes are smaller than this number:
- EXPECT_GE(45000u, changed_byte_count);
- }
-}
-
-bool DecodeFuzzTest::TryAssemble(const std::string& buffer,
- std::string* output) const {
- courgette::EncodedProgram *encoded = NULL;
- bool result = false;
-
- courgette::SourceStreamSet sources;
- bool can_get_source_streams = sources.Init(buffer.c_str(), buffer.length());
- if (can_get_source_streams) {
- const courgette::Status read_status =
- ReadEncodedProgram(&sources, &encoded);
- if (read_status == courgette::C_OK) {
- courgette::SinkStream assembled;
- const courgette::Status assemble_status = Assemble(encoded, &assembled);
-
- if (assemble_status == courgette::C_OK) {
- const void* assembled_buffer = assembled.Buffer();
- size_t assembled_length = assembled.Length();
-
- output->clear();
- output->assign(reinterpret_cast<const char*>(assembled_buffer),
- assembled_length);
- result = true;
- }
- }
- }
-
- DeleteEncodedProgram(encoded);
-
- return result;
-}
-
-TEST_F(DecodeFuzzTest, All) {
- FuzzExe("setup1.exe");
-}
-
-int main(int argc, char** argv) {
- return TestSuite(argc, argv).Run();
-}
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +// Fuzz testing for EncodedProgram serialized format and assembly. +// +// We would like some assurance that if an EncodedProgram is malformed we will +// not crash. The EncodedProgram could be malformed either due to malicious +// attack to due to an error in patch generation. +// +// We try a lot of arbitrary modifications to the serialized form and make sure +// that the outcome is not a crash. + +#include <string> + +#include "base/path_service.h" +#include "base/file_util.h" +#include "base/string_util.h" +#include "base/test_suite.h" + +#include "courgette/courgette.h" +#include "courgette/streams.h" + +#include "testing/gtest/include/gtest/gtest.h" + +class DecodeFuzzTest : public testing::Test { + public: + void FuzzExe(const char *) const; + + private: + virtual void SetUp() { + PathService::Get(base::DIR_SOURCE_ROOT, &testdata_dir_); + testdata_dir_ = testdata_dir_.Append(L"courgette"); + testdata_dir_ = testdata_dir_.Append(L"testdata"); + } + + virtual void TearDown() { } + + void FuzzByte(const std::string& buffer, const std::string& output, + size_t index) const; + void FuzzBits(const std::string& buffer, const std::string& output, + size_t index, int bits_to_flip) const; + + // Returns true if could assemble, false if rejected. + bool TryAssemble(const std::string& buffer, std::string* output) const; + + // Returns contents of |file_name| as uninterprested bytes stored in a string. + std::string FileContents(const char* file_name) const; + + // Full path name of testdata directory + FilePath testdata_dir_; +}; + +// Reads a test file into a string. +std::string DecodeFuzzTest::FileContents(const char* file_name) const { + FilePath file_path = testdata_dir_.AppendASCII(file_name); + std::string file_contents; + if (!file_util::ReadFileToString(file_path, &file_contents)) { + EXPECT_TRUE(!"Could not read test data"); + } + return file_contents; +} + +// Loads an executable and does fuzz testing in the serialized format. +void DecodeFuzzTest::FuzzExe(const char* file_name) const { + std::string file1 = FileContents(file_name); + + const void* original_buffer = file1.c_str(); + size_t original_length = file1.size(); + + courgette::AssemblyProgram* program = NULL; + const courgette::Status parse_status = + courgette::ParseWin32X86PE(original_buffer, original_length, &program); + EXPECT_EQ(courgette::C_OK, parse_status); + + courgette::EncodedProgram* encoded = NULL; + + const courgette::Status encode_status = Encode(program, &encoded); + EXPECT_EQ(courgette::C_OK, encode_status); + + DeleteAssemblyProgram(program); + + courgette::SinkStreamSet sinks; + const courgette::Status write_status = WriteEncodedProgram(encoded, &sinks); + EXPECT_EQ(courgette::C_OK, write_status); + + DeleteEncodedProgram(encoded); + + courgette::SinkStream sink; + bool can_collect = sinks.CopyTo(&sink); + EXPECT_TRUE(can_collect); + + size_t length = sink.Length(); + + std::string base_buffer(reinterpret_cast<const char*>(sink.Buffer()), length); + std::string base_output; + bool ok = TryAssemble(base_buffer, &base_output); + EXPECT_EQ(true, ok); + + // Now we have a good serialized EncodedProgram in |base_buffer|. Time to + // fuzz. + + // More intense fuzzing on the first part because it contains more control + // information like substeam lengths. + size_t position = 0; + for ( ; position < 100 && position < length; position += 1) { + FuzzByte(base_buffer, base_output, position); + } + // We would love to fuzz every position, but it takes too long. + for ( ; position < length; position += 900) { + FuzzByte(base_buffer, base_output, position); + } +} + +// FuzzByte tries to break the EncodedProgram deserializer and assembler. It +// takes a good serialization of and EncodedProgram, flips some bits, and checks +// that the behaviour is reasonable. It has testing checks for unreasonable +// behaviours. +void DecodeFuzzTest::FuzzByte(const std::string& base_buffer, + const std::string& base_output, + size_t index) const { + printf("Fuzzing position %d\n", static_cast<int>(index)); + + // The following 10 values are a compromize between run time and coverage of + // the 255 'wrong' values at this byte position. + + // 0xFF flips all the bits. + FuzzBits(base_buffer, base_output, index, 0xFF); + // 0x7F flips the most bits without changing Varint32 framing. + FuzzBits(base_buffer, base_output, index, 0x7F); + // These all flip one bit. + FuzzBits(base_buffer, base_output, index, 0x80); + FuzzBits(base_buffer, base_output, index, 0x40); + FuzzBits(base_buffer, base_output, index, 0x20); + FuzzBits(base_buffer, base_output, index, 0x10); + FuzzBits(base_buffer, base_output, index, 0x08); + FuzzBits(base_buffer, base_output, index, 0x04); + FuzzBits(base_buffer, base_output, index, 0x02); + FuzzBits(base_buffer, base_output, index, 0x01); +} + +// FuzzBits tries to break the EncodedProgram deserializer and assembler. It +// takes a good serialization of and EncodedProgram, flips some bits, and checks +// that the behaviour is reasonable. +// +// There are EXPECT calls to check for unreasonable behaviour. These are +// somewhat arbitrary in that the parameters cannot easily be derived from first +// principles. They may need updating as the serialized format evolves. +void DecodeFuzzTest::FuzzBits(const std::string& base_buffer, + const std::string& base_output, + size_t index, int bits_to_flip) const { + std::string modified_buffer = base_buffer; + std::string modified_output; + modified_buffer[index] ^= bits_to_flip; + + bool ok = TryAssemble(modified_buffer, &modified_output); + + if (ok) { + // We normally expect TryAssemble to fail. But sometimes it succeeds. + // What could have happened? We changed one byte in the serialized form: + // + // * If we changed one of the copied bytes, we would see a single byte + // change in the output. + // * If we changed an address table element, all the references to that + // address would be different. + // * If we changed a copy count, we would run out of data in some stream, + // or leave data remaining, so should not be here. + // * If we changed an origin address, it could affect all relocations based + // off that address. If no relocations were based off the address then + // there will be no changes. + // * If we changed an origin address, it could cause some abs32 relocs to + // shift from one page to the next, changing the number and layout of + // blocks in the base relocation table. + + // Generated length could vary slightly due to base relocation table layout. + // In the worst case the number of base relocation blocks doubles, approx + // 12/4096 or 0.3% size of file. + size_t base_length = base_output.length(); + size_t modified_length = modified_output.length(); + ptrdiff_t diff = base_length - modified_length; + if (diff < -200 || diff > 200) { + EXPECT_EQ(base_length, modified_length); + } + + size_t changed_byte_count = 0; + for (size_t i = 0; i < base_length && i < modified_length; ++i) { + changed_byte_count += (base_output[i] != modified_output[i]); + } + + if (index > 60) { // Beyond the origin addresses ... + EXPECT_NE(0, changed_byte_count); // ... we expect some difference. + } + // Currently all changes are smaller than this number: + EXPECT_GE(45000u, changed_byte_count); + } +} + +bool DecodeFuzzTest::TryAssemble(const std::string& buffer, + std::string* output) const { + courgette::EncodedProgram *encoded = NULL; + bool result = false; + + courgette::SourceStreamSet sources; + bool can_get_source_streams = sources.Init(buffer.c_str(), buffer.length()); + if (can_get_source_streams) { + const courgette::Status read_status = + ReadEncodedProgram(&sources, &encoded); + if (read_status == courgette::C_OK) { + courgette::SinkStream assembled; + const courgette::Status assemble_status = Assemble(encoded, &assembled); + + if (assemble_status == courgette::C_OK) { + const void* assembled_buffer = assembled.Buffer(); + size_t assembled_length = assembled.Length(); + + output->clear(); + output->assign(reinterpret_cast<const char*>(assembled_buffer), + assembled_length); + result = true; + } + } + } + + DeleteEncodedProgram(encoded); + + return result; +} + +TEST_F(DecodeFuzzTest, All) { + FuzzExe("setup1.exe"); +} + +int main(int argc, char** argv) { + return TestSuite(argc, argv).Run(); +} diff --git a/third_party/courgette/encoded_program_unittest.cc b/courgette/encoded_program_unittest.cc index 9b0025e..aedb838 100644 --- a/third_party/courgette/encoded_program_unittest.cc +++ b/courgette/encoded_program_unittest.cc @@ -2,9 +2,9 @@ // Use of this source code is governed by a BSD-style license that can be // found in the LICENSE file. -#include "third_party/courgette/encoded_program.h" +#include "courgette/encoded_program.h" -#include "third_party/courgette/streams.h" +#include "courgette/streams.h" #include "testing/gtest/include/gtest/gtest.h" TEST(EncodedProgramTest, Test) { @@ -62,4 +62,3 @@ TEST(EncodedProgramTest, Test) { EXPECT_EQ(0, memcmp(assembled_buffer, golden, 8)); } - diff --git a/third_party/courgette/ensemble.cc b/courgette/ensemble.cc index 4c8f013..15bfb3e 100644 --- a/third_party/courgette/ensemble.cc +++ b/courgette/ensemble.cc @@ -1,102 +1,102 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-// *** File comments
-
-#include "third_party/courgette/ensemble.h"
-
-#include "base/basictypes.h"
-#include "base/logging.h"
-#include "base/string_util.h"
-
-#include "third_party/courgette/image_info.h"
-#include "third_party/courgette/region.h"
-#include "third_party/courgette/streams.h"
-#include "third_party/courgette/simple_delta.h"
-
-namespace courgette {
-
-Element::Element(Kind kind, Ensemble* ensemble, const Region& region)
- : kind_(kind), ensemble_(ensemble), region_(region) {
-}
-
-std::string Element::Name() const {
- return ensemble_->name() + "("
- + IntToString(kind()) + ","
- + Uint64ToString(offset_in_ensemble()) + ","
- + Uint64ToString(region().length()) + ")";
-}
-
-// A subclass of Element that has a PEInfo.
-class ElementWinPE : public Element {
- public:
- ElementWinPE(Kind kind, Ensemble* ensemble, const Region& region,
- PEInfo* info)
- : Element(kind, ensemble, region),
- pe_info_(info) {
- }
-
- virtual PEInfo* GetPEInfo() const { return pe_info_; }
-
- protected:
- ~ElementWinPE() { delete pe_info_; }
-
- private:
- PEInfo* pe_info_; // Owned by |this|.
-};
-
-// Scans the Ensemble's region, sniffing out Elements. We assume that the
-// elements do not overlap.
-Status Ensemble::FindEmbeddedElements() {
- size_t length = region_.length();
- const uint8* start = region_.start();
-
- size_t position = 0;
- while (position < length) {
- // Quick test; Windows executables begin with 'MZ'.
- if (start[position] == 'M' &&
- position + 1 < length && start[position + 1] == 'Z') {
- courgette::PEInfo *info = new courgette::PEInfo();
- info->Init(start + position, length - position);
- if (info->ParseHeader()) {
- Region region(start + position, info->length());
-
- if (info->has_text_section()) {
- Element* element = new ElementWinPE(Element::WIN32_X86_WITH_CODE,
- this, region, info);
- owned_elements_.push_back(element);
- elements_.push_back(element);
- position += region.length();
- continue;
- }
-
- // If we had a clever transformation for resource-only executables we
- // should identify the suitable elements here:
- if (!info->has_text_section() && false) {
- Element* element = new ElementWinPE(Element::WIN32_NOCODE,
- this, region, info);
- owned_elements_.push_back(element);
- elements_.push_back(element);
- position += region.length();
- continue;
- }
- }
- delete info;
- }
-
- // This is where to add new formats, e.g. Linux executables, Dalvik
- // executables etc.
-
- // No Element found at current position.
- ++position;
- }
- return C_OK;
-}
-
-Ensemble::~Ensemble() {
- for (size_t i = 0; i < owned_elements_.size(); ++i)
- delete owned_elements_[i];
-}
-
-} // namespace
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +// *** File comments + +#include "courgette/ensemble.h" + +#include "base/basictypes.h" +#include "base/logging.h" +#include "base/string_util.h" + +#include "courgette/image_info.h" +#include "courgette/region.h" +#include "courgette/streams.h" +#include "courgette/simple_delta.h" + +namespace courgette { + +Element::Element(Kind kind, Ensemble* ensemble, const Region& region) + : kind_(kind), ensemble_(ensemble), region_(region) { +} + +std::string Element::Name() const { + return ensemble_->name() + "(" + + IntToString(kind()) + "," + + Uint64ToString(offset_in_ensemble()) + "," + + Uint64ToString(region().length()) + ")"; +} + +// A subclass of Element that has a PEInfo. +class ElementWinPE : public Element { + public: + ElementWinPE(Kind kind, Ensemble* ensemble, const Region& region, + PEInfo* info) + : Element(kind, ensemble, region), + pe_info_(info) { + } + + virtual PEInfo* GetPEInfo() const { return pe_info_; } + + protected: + ~ElementWinPE() { delete pe_info_; } + + private: + PEInfo* pe_info_; // Owned by |this|. +}; + +// Scans the Ensemble's region, sniffing out Elements. We assume that the +// elements do not overlap. +Status Ensemble::FindEmbeddedElements() { + size_t length = region_.length(); + const uint8* start = region_.start(); + + size_t position = 0; + while (position < length) { + // Quick test; Windows executables begin with 'MZ'. + if (start[position] == 'M' && + position + 1 < length && start[position + 1] == 'Z') { + courgette::PEInfo *info = new courgette::PEInfo(); + info->Init(start + position, length - position); + if (info->ParseHeader()) { + Region region(start + position, info->length()); + + if (info->has_text_section()) { + Element* element = new ElementWinPE(Element::WIN32_X86_WITH_CODE, + this, region, info); + owned_elements_.push_back(element); + elements_.push_back(element); + position += region.length(); + continue; + } + + // If we had a clever transformation for resource-only executables we + // should identify the suitable elements here: + if (!info->has_text_section() && false) { + Element* element = new ElementWinPE(Element::WIN32_NOCODE, + this, region, info); + owned_elements_.push_back(element); + elements_.push_back(element); + position += region.length(); + continue; + } + } + delete info; + } + + // This is where to add new formats, e.g. Linux executables, Dalvik + // executables etc. + + // No Element found at current position. + ++position; + } + return C_OK; +} + +Ensemble::~Ensemble() { + for (size_t i = 0; i < owned_elements_.size(); ++i) + delete owned_elements_[i]; +} + +} // namespace diff --git a/third_party/courgette/ensemble.h b/courgette/ensemble.h index 7602c3f..6a58cb0 100644 --- a/third_party/courgette/ensemble.h +++ b/courgette/ensemble.h @@ -1,257 +1,257 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-// The main idea in Courgette is to do patching *under a tranformation*. The
-// input is transformed into a new representation, patching occurs in the new
-// repesentation, and then the tranform is reversed to get the patched data.
-//
-// The idea is applied to pieces (or 'Elements') of the whole (or 'Ensemble').
-// Each of the elements has to go through the same set of steps in lock-step,
-// but there may be many different kinds of elements, which have different
-// transformation.
-//
-// This file declares all the main types involved in creating and applying a
-// patch with this structure.
-
-#ifndef COURGETTE_ENSEMBLE_H_
-#define COURGETTE_ENSEMBLE_H_
-
-#include <vector>
-#include <string>
-
-#include "base/basictypes.h"
-
-#include "third_party/courgette/courgette.h"
-#include "third_party/courgette/region.h"
-#include "third_party/courgette/streams.h"
-
-namespace courgette {
-
-// Forward declarations:
-class Ensemble;
-class PEInfo;
-
-// An Element is a region of an Ensemble with an identifyable kind.
-//
-class Element {
- public:
- enum Kind { WIN32_X86_WITH_CODE, WIN32_NOCODE };
-
- virtual ~Element() {}
-
- Kind kind() const { return kind_; }
- const Region& region() const { return region_; }
-
- // The name is used only for debugging and logging.
- virtual std::string Name() const;
-
- // Returns the byte position of this Element relative to the start of
- // containing Ensemble.
- size_t offset_in_ensemble() const;
-
- // Some subclasses of Element might have a PEInfo.
- virtual PEInfo* GetPEInfo() const { return NULL; }
-
- protected:
- Element(Kind kind, Ensemble* ensemble, const Region& region);
-
- private:
- Kind kind_;
- Ensemble* ensemble_;
- Region region_;
-
- DISALLOW_COPY_AND_ASSIGN(Element);
-};
-
-
-class Ensemble {
- public:
- Ensemble(const Region& region, const char* name)
- : region_(region), name_(name) {}
- ~Ensemble();
-
- const Region& region() const { return region_; }
- const std::string& name() const { return name_; }
-
- // Scans the region to find Elements within the region().
- Status FindEmbeddedElements();
-
- // Returns the elements found by 'FindEmbeddedElements'.
- const std::vector<Element*>& elements() const { return elements_; }
-
-
- private:
- Region region_; // The memory, owned by caller, containing the
- // Ensemble's data.
- std::string name_; // A debugging/logging name for the Ensemble.
-
- std::vector<Element*> elements_; // Embedded elements discovered.
- std::vector<Element*> owned_elements_; // For deallocation.
-
- DISALLOW_COPY_AND_ASSIGN(Ensemble);
-};
-
-inline size_t Element::offset_in_ensemble() const {
- return region().start() - ensemble_->region().start();
-}
-
-// The 'CourgettePatchFile' is class is a 'namespace' for the constants that
-// appear in a Courgette patch file.
-struct CourgettePatchFile {
- //
- // The Courgette patch format interleaves the data for N embedded Elements.
- //
- // Format of a patch file:
- // header:
- // magic
- // version
- // source-checksum
- // target-checksum
- // multiple-streams:
- // stream 0:
- // number-of-transformed-elements (N) - varint32
- // transformation-1-method-id
- // transformation-2-method-id
- // ...
- // transformation-1-initial-parameters
- // transformation-2-initial-parameters
- // ...
- // stream 1:
- // correction:
- // transformation-1-parameters
- // transformation-2-parameters
- // ...
- // stream 2:
- // correction:
- // transformed-element-1
- // transformed-element-2
- // ...
- // stream 3:
- // correction:
- // base-file
- // element-1
- // element-2
- // ...
-
- static const uint32 kMagic = 'C' | ('o' << 8) | ('u' << 16);
-
- static const uint32 kVersion = 20090320;
-
- // Transformation method IDs.
- enum TransformationMethodId {
- T_COURGETTE_WIN32_X86 = 1, // Windows 32 bit 'Portable Executable' x86.
- };
-};
-
-// For any transform you would implement both a TransformationPatcher and a
-// TransformationPatchGenerator.
-//
-// TransformationPatcher is the interface which abstracts out the actual
-// transformation used on an Element. The patching itself happens outside the
-// actions of a TransformationPatcher. There are four steps.
-//
-// The first step is an Init step. The parameters to the Init step identify the
-// element, for example, range of locations within the original ensemble that
-// correspond to the element.
-//
-// PredictTransformParameters, explained below.
-//
-// The two final steps are 'Transform' - to transform the element into a new
-// representation, and to 'Reform' - to transform from the new representation
-// back to the original form.
-//
-// The Transform step takes some parameters. This allows the transform to be
-// customized to the particular element, or to receive some assistance in the
-// analysis required to perform the transform. The transform parameters might
-// be extensive but mostly predicable, so preceeding Transform is a
-// PredictTransformParameters step.
-//
-class TransformationPatcher {
- public:
- virtual ~TransformationPatcher() {}
-
- // First step: provides parameters for the patching. This would at a minimum
- // identify the element within the ensemble being patched.
- virtual Status Init(SourceStream* parameter_stream) = 0;
-
- // Second step: predicts transform parameters.
- virtual Status PredictTransformParameters(
- SinkStreamSet* predicted_parameters) = 0;
-
- // Third step: transforms element from original representation into alternate
- // representation.
- virtual Status Transform(SourceStreamSet* corrected_parameters,
- SinkStreamSet* transformed_element) = 0;
-
- // Final step: transforms element back from alternate representation into
- // original representation.
- virtual Status Reform(SourceStreamSet* transformed_element,
- SinkStream* reformed_element) = 0;
-};
-
-// TransformationPatchGenerator is the interface which abstracts out the actual
-// transformation used (and adjustment used) when differentially compressing one
-// Element from the |new_ensemble| against a corresponding element in the
-// |old_ensemble|.
-//
-// This is not a pure interface. There is a small amount of inheritance
-// implementation for the fields and actions common to all
-// TransformationPatchGenerators.
-//
-// When TransformationPatchGenerator is subclassed, there will be a
-// corresponding subclass of TransformationPatcher.
-//
-class TransformationPatchGenerator {
- public:
- TransformationPatchGenerator(Element* old_element,
- Element* new_element,
- TransformationPatcher* patcher);
-
- virtual ~TransformationPatchGenerator();
-
- // Returns the TransformationMethodId that identies this transformation.
- virtual CourgettePatchFile::TransformationMethodId Kind() = 0;
-
- // Writes the parameters that will be passed to TransformationPatcher::Init.
- virtual Status WriteInitialParameters(SinkStream* parameter_stream) = 0;
-
- // Predicts the transform parameters for the |old_element|. This must match
- // exactly the output that will be produced by the PredictTransformParameters
- // method of the corresponding subclass of TransformationPatcher. This method
- // is not pure. The default implementation delegates to the patcher to
- // guarantee matching output.
- virtual Status PredictTransformParameters(SinkStreamSet* prediction);
-
- // Writes the desired parameters for the transform of the old element from the
- // file representation to the alternate representation.
- virtual Status CorrectedTransformParameters(SinkStreamSet* parameters) = 0;
-
- // Writes both |old_element| and |new_element| in the new representation.
- // |old_corrected_parameters| will match the |corrected_parameters| passed to
- // the Transform method of the corresponding sublcass of
- // TransformationPatcher.
- //
- // The output written to |old_transformed_element| must match exactly the
- // output written by the Transform method of the corresponding subclass of
- // TransformationPatcher.
- virtual Status Transform(SourceStreamSet* old_corrected_parameters,
- SinkStreamSet* old_transformed_element,
- SinkStreamSet* new_transformed_element) = 0;
-
- // Transforms the new transformed_element back from the alternate
- // representation into the original file format. This must match exactly the
- // output that will be produced by the corresponding subclass of
- // TransformationPatcher::Reform. This method is not pure. The default
- // implementation delegates to the patcher.
- virtual Status Reform(SourceStreamSet* transformed_element,
- SinkStream* reformed_element);
-
- protected:
- Element* old_element_;
- Element* new_element_;
- TransformationPatcher* patcher_;
-};
-
-} // namespace
-#endif // COURGETTE_ENSEMBLE_H_
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +// The main idea in Courgette is to do patching *under a tranformation*. The +// input is transformed into a new representation, patching occurs in the new +// repesentation, and then the tranform is reversed to get the patched data. +// +// The idea is applied to pieces (or 'Elements') of the whole (or 'Ensemble'). +// Each of the elements has to go through the same set of steps in lock-step, +// but there may be many different kinds of elements, which have different +// transformation. +// +// This file declares all the main types involved in creating and applying a +// patch with this structure. + +#ifndef COURGETTE_ENSEMBLE_H_ +#define COURGETTE_ENSEMBLE_H_ + +#include <vector> +#include <string> + +#include "base/basictypes.h" + +#include "courgette/courgette.h" +#include "courgette/region.h" +#include "courgette/streams.h" + +namespace courgette { + +// Forward declarations: +class Ensemble; +class PEInfo; + +// An Element is a region of an Ensemble with an identifyable kind. +// +class Element { + public: + enum Kind { WIN32_X86_WITH_CODE, WIN32_NOCODE }; + + virtual ~Element() {} + + Kind kind() const { return kind_; } + const Region& region() const { return region_; } + + // The name is used only for debugging and logging. + virtual std::string Name() const; + + // Returns the byte position of this Element relative to the start of + // containing Ensemble. + size_t offset_in_ensemble() const; + + // Some subclasses of Element might have a PEInfo. + virtual PEInfo* GetPEInfo() const { return NULL; } + + protected: + Element(Kind kind, Ensemble* ensemble, const Region& region); + + private: + Kind kind_; + Ensemble* ensemble_; + Region region_; + + DISALLOW_COPY_AND_ASSIGN(Element); +}; + + +class Ensemble { + public: + Ensemble(const Region& region, const char* name) + : region_(region), name_(name) {} + ~Ensemble(); + + const Region& region() const { return region_; } + const std::string& name() const { return name_; } + + // Scans the region to find Elements within the region(). + Status FindEmbeddedElements(); + + // Returns the elements found by 'FindEmbeddedElements'. + const std::vector<Element*>& elements() const { return elements_; } + + + private: + Region region_; // The memory, owned by caller, containing the + // Ensemble's data. + std::string name_; // A debugging/logging name for the Ensemble. + + std::vector<Element*> elements_; // Embedded elements discovered. + std::vector<Element*> owned_elements_; // For deallocation. + + DISALLOW_COPY_AND_ASSIGN(Ensemble); +}; + +inline size_t Element::offset_in_ensemble() const { + return region().start() - ensemble_->region().start(); +} + +// The 'CourgettePatchFile' is class is a 'namespace' for the constants that +// appear in a Courgette patch file. +struct CourgettePatchFile { + // + // The Courgette patch format interleaves the data for N embedded Elements. + // + // Format of a patch file: + // header: + // magic + // version + // source-checksum + // target-checksum + // multiple-streams: + // stream 0: + // number-of-transformed-elements (N) - varint32 + // transformation-1-method-id + // transformation-2-method-id + // ... + // transformation-1-initial-parameters + // transformation-2-initial-parameters + // ... + // stream 1: + // correction: + // transformation-1-parameters + // transformation-2-parameters + // ... + // stream 2: + // correction: + // transformed-element-1 + // transformed-element-2 + // ... + // stream 3: + // correction: + // base-file + // element-1 + // element-2 + // ... + + static const uint32 kMagic = 'C' | ('o' << 8) | ('u' << 16); + + static const uint32 kVersion = 20090320; + + // Transformation method IDs. + enum TransformationMethodId { + T_COURGETTE_WIN32_X86 = 1, // Windows 32 bit 'Portable Executable' x86. + }; +}; + +// For any transform you would implement both a TransformationPatcher and a +// TransformationPatchGenerator. +// +// TransformationPatcher is the interface which abstracts out the actual +// transformation used on an Element. The patching itself happens outside the +// actions of a TransformationPatcher. There are four steps. +// +// The first step is an Init step. The parameters to the Init step identify the +// element, for example, range of locations within the original ensemble that +// correspond to the element. +// +// PredictTransformParameters, explained below. +// +// The two final steps are 'Transform' - to transform the element into a new +// representation, and to 'Reform' - to transform from the new representation +// back to the original form. +// +// The Transform step takes some parameters. This allows the transform to be +// customized to the particular element, or to receive some assistance in the +// analysis required to perform the transform. The transform parameters might +// be extensive but mostly predicable, so preceeding Transform is a +// PredictTransformParameters step. +// +class TransformationPatcher { + public: + virtual ~TransformationPatcher() {} + + // First step: provides parameters for the patching. This would at a minimum + // identify the element within the ensemble being patched. + virtual Status Init(SourceStream* parameter_stream) = 0; + + // Second step: predicts transform parameters. + virtual Status PredictTransformParameters( + SinkStreamSet* predicted_parameters) = 0; + + // Third step: transforms element from original representation into alternate + // representation. + virtual Status Transform(SourceStreamSet* corrected_parameters, + SinkStreamSet* transformed_element) = 0; + + // Final step: transforms element back from alternate representation into + // original representation. + virtual Status Reform(SourceStreamSet* transformed_element, + SinkStream* reformed_element) = 0; +}; + +// TransformationPatchGenerator is the interface which abstracts out the actual +// transformation used (and adjustment used) when differentially compressing one +// Element from the |new_ensemble| against a corresponding element in the +// |old_ensemble|. +// +// This is not a pure interface. There is a small amount of inheritance +// implementation for the fields and actions common to all +// TransformationPatchGenerators. +// +// When TransformationPatchGenerator is subclassed, there will be a +// corresponding subclass of TransformationPatcher. +// +class TransformationPatchGenerator { + public: + TransformationPatchGenerator(Element* old_element, + Element* new_element, + TransformationPatcher* patcher); + + virtual ~TransformationPatchGenerator(); + + // Returns the TransformationMethodId that identies this transformation. + virtual CourgettePatchFile::TransformationMethodId Kind() = 0; + + // Writes the parameters that will be passed to TransformationPatcher::Init. + virtual Status WriteInitialParameters(SinkStream* parameter_stream) = 0; + + // Predicts the transform parameters for the |old_element|. This must match + // exactly the output that will be produced by the PredictTransformParameters + // method of the corresponding subclass of TransformationPatcher. This method + // is not pure. The default implementation delegates to the patcher to + // guarantee matching output. + virtual Status PredictTransformParameters(SinkStreamSet* prediction); + + // Writes the desired parameters for the transform of the old element from the + // file representation to the alternate representation. + virtual Status CorrectedTransformParameters(SinkStreamSet* parameters) = 0; + + // Writes both |old_element| and |new_element| in the new representation. + // |old_corrected_parameters| will match the |corrected_parameters| passed to + // the Transform method of the corresponding sublcass of + // TransformationPatcher. + // + // The output written to |old_transformed_element| must match exactly the + // output written by the Transform method of the corresponding subclass of + // TransformationPatcher. + virtual Status Transform(SourceStreamSet* old_corrected_parameters, + SinkStreamSet* old_transformed_element, + SinkStreamSet* new_transformed_element) = 0; + + // Transforms the new transformed_element back from the alternate + // representation into the original file format. This must match exactly the + // output that will be produced by the corresponding subclass of + // TransformationPatcher::Reform. This method is not pure. The default + // implementation delegates to the patcher. + virtual Status Reform(SourceStreamSet* transformed_element, + SinkStream* reformed_element); + + protected: + Element* old_element_; + Element* new_element_; + TransformationPatcher* patcher_; +}; + +} // namespace +#endif // COURGETTE_ENSEMBLE_H_ diff --git a/third_party/courgette/ensemble_apply.cc b/courgette/ensemble_apply.cc index 7954ca4..5a9edfd4 100644 --- a/third_party/courgette/ensemble_apply.cc +++ b/courgette/ensemble_apply.cc @@ -1,412 +1,412 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-// This file contains the code to apply a Courgette patch.
-
-#include "third_party/courgette/ensemble.h"
-
-#include "base/basictypes.h"
-#include "base/file_util.h"
-#include "base/logging.h"
-
-#include "third_party/courgette/crc.h"
-#include "third_party/courgette/image_info.h"
-#include "third_party/courgette/region.h"
-#include "third_party/courgette/streams.h"
-#include "third_party/courgette/simple_delta.h"
-#include "third_party/courgette/win32_x86_patcher.h"
-
-namespace courgette {
-
-// EnsemblePatchApplication is all the logic and data required to apply the
-// multi-stage patch.
-class EnsemblePatchApplication {
- public:
- EnsemblePatchApplication();
- ~EnsemblePatchApplication();
-
- Status ReadHeader(SourceStream* header_stream);
-
- Status InitBase(const Region& region);
-
- Status ValidateBase();
-
- Status ReadInitialParameters(SourceStream* initial_parameters);
-
- Status PredictTransformParameters(SinkStreamSet* predicted_parameters);
-
- Status SubpatchTransformParameters(SinkStreamSet* prediction,
- SourceStream* correction,
- SourceStreamSet* corrected_parameters);
-
- Status TransformUp(SourceStreamSet* parameters,
- SinkStreamSet* transformed_elements);
-
- Status SubpatchTransformedElements(SinkStreamSet* elements,
- SourceStream* correction,
- SourceStreamSet* corrected_elements);
-
- Status TransformDown(SourceStreamSet* transformed_elements,
- SinkStream* basic_elements);
-
- Status SubpatchFinalOutput(SourceStream* original,
- SourceStream* correction,
- SinkStream* corrected_ensemble);
-
- private:
- Status SubpatchStreamSets(SinkStreamSet* predicted_items,
- SourceStream* correction,
- SourceStreamSet* corrected_items,
- SinkStream* corrected_items_storage);
-
- Region base_region_; // Location of in-memory copy of 'old' version.
-
- uint32 source_checksum_;
- uint32 target_checksum_;
-
- std::vector<TransformationPatcher*> patchers_;
-
- SinkStream corrected_parameters_storage_;
- SinkStream corrected_elements_storage_;
-
- DISALLOW_COPY_AND_ASSIGN(EnsemblePatchApplication);
-};
-
-EnsemblePatchApplication::EnsemblePatchApplication()
- : source_checksum_(0), target_checksum_(0) {
-}
-
-EnsemblePatchApplication::~EnsemblePatchApplication() {
- for (size_t i = 0; i < patchers_.size(); ++i) {
- delete patchers_[i];
- }
-}
-
-Status EnsemblePatchApplication::ReadHeader(SourceStream* header_stream) {
- uint32 magic;
- if (!header_stream->ReadVarint32(&magic))
- return C_BAD_ENSEMBLE_MAGIC;
-
- if (magic != CourgettePatchFile::kMagic)
- return C_BAD_ENSEMBLE_MAGIC;
-
- uint32 version;
- if (!header_stream->ReadVarint32(&version))
- return C_BAD_ENSEMBLE_VERSION;
-
- if (version != CourgettePatchFile::kVersion)
- return C_BAD_ENSEMBLE_VERSION;
-
- if (!header_stream->ReadVarint32(&source_checksum_))
- return C_BAD_ENSEMBLE_HEADER;
-
- if (!header_stream->ReadVarint32(&target_checksum_))
- return C_BAD_ENSEMBLE_HEADER;
-
- return C_OK;
-}
-
-Status EnsemblePatchApplication::InitBase(const Region& region) {
- base_region_.assign(region);
- return C_OK;
-}
-
-Status EnsemblePatchApplication::ValidateBase() {
- uint32 checksum = CalculateCrc(base_region_.start(), base_region_.length());
- if (source_checksum_ != checksum)
- return C_BAD_ENSEMBLE_CRC;
-
- return C_OK;
-}
-
-Status EnsemblePatchApplication::ReadInitialParameters(
- SourceStream* transformation_parameters) {
- uint32 number_of_transformations = 0;
- if (!transformation_parameters->ReadVarint32(&number_of_transformations))
- return C_BAD_ENSEMBLE_HEADER;
-
- for (size_t i = 0; i < number_of_transformations; ++i) {
- uint32 kind;
- if (!transformation_parameters->ReadVarint32(&kind))
- return C_BAD_ENSEMBLE_HEADER;
-
- if (kind == CourgettePatchFile::T_COURGETTE_WIN32_X86) {
- TransformationPatcher* patcher =
- new CourgetteWin32X86Patcher(base_region_);
- patchers_.push_back(patcher);
- } else {
- return C_BAD_ENSEMBLE_HEADER;
- }
- }
-
- for (size_t i = 0; i < patchers_.size(); ++i) {
- Status status = patchers_[i]->Init(transformation_parameters);
- if (status != C_OK)
- return status;
- }
-
- // All transformation_parameters should have been consumed by the above loop.
- if (!transformation_parameters->Empty())
- return C_BAD_ENSEMBLE_HEADER;
-
- return C_OK;
-}
-
-Status EnsemblePatchApplication::PredictTransformParameters(
- SinkStreamSet* all_predicted_parameters) {
- for (size_t i = 0; i < patchers_.size(); ++i) {
- SinkStreamSet single_predicted_parameters;
- Status status =
- patchers_[i]->PredictTransformParameters(&single_predicted_parameters);
- if (status != C_OK)
- return status;
- if (!all_predicted_parameters->WriteSet(&single_predicted_parameters))
- return C_STREAM_ERROR;
- }
- return C_OK;
-}
-
-Status EnsemblePatchApplication::SubpatchTransformParameters(
- SinkStreamSet* predicted_parameters,
- SourceStream* correction,
- SourceStreamSet* corrected_parameters) {
- return SubpatchStreamSets(predicted_parameters,
- correction,
- corrected_parameters,
- &corrected_parameters_storage_);
-}
-
-Status EnsemblePatchApplication::TransformUp(
- SourceStreamSet* parameters,
- SinkStreamSet* transformed_elements) {
- for (size_t i = 0; i < patchers_.size(); ++i) {
- SourceStreamSet single_parameters;
- if (!parameters->ReadSet(&single_parameters))
- return C_STREAM_ERROR;
- SinkStreamSet single_transformed_element;
- Status status = patchers_[i]->Transform(&single_parameters,
- &single_transformed_element);
- if (status != C_OK)
- return status;
- if (!single_parameters.Empty())
- return C_STREAM_NOT_CONSUMED;
- if (!transformed_elements->WriteSet(&single_transformed_element))
- return C_STREAM_ERROR;
- }
-
- if (!parameters->Empty())
- return C_STREAM_NOT_CONSUMED;
- return C_OK;
-}
-
-Status EnsemblePatchApplication::SubpatchTransformedElements(
- SinkStreamSet* predicted_elements,
- SourceStream* correction,
- SourceStreamSet* corrected_elements) {
- return SubpatchStreamSets(predicted_elements,
- correction,
- corrected_elements,
- &corrected_elements_storage_);
-}
-
-Status EnsemblePatchApplication::TransformDown(
- SourceStreamSet* transformed_elements,
- SinkStream* basic_elements) {
- // Construct blob of original input followed by reformed elements.
-
- // The original input:
- basic_elements->Write(base_region_.start(), base_region_.length());
-
- for (size_t i = 0; i < patchers_.size(); ++i) {
- SourceStreamSet single_corrected_element;
- if (!transformed_elements->ReadSet(&single_corrected_element))
- return C_STREAM_ERROR;
- Status status = patchers_[i]->Reform(&single_corrected_element,
- basic_elements);
- if (status != C_OK)
- return status;
- if (!single_corrected_element.Empty())
- return C_STREAM_NOT_CONSUMED;
- }
-
- if (!transformed_elements->Empty())
- return C_STREAM_NOT_CONSUMED;
-
- return C_OK;
-}
-
-Status EnsemblePatchApplication::SubpatchFinalOutput(
- SourceStream* original,
- SourceStream* correction,
- SinkStream* corrected_ensemble) {
- Status delta_status = ApplySimpleDelta(original, correction,
- corrected_ensemble);
- if (delta_status != C_OK)
- return delta_status;
-
- if (CalculateCrc(corrected_ensemble->Buffer(),
- corrected_ensemble->Length()) != target_checksum_)
- return C_BAD_ENSEMBLE_CRC;
-
- return C_OK;
-}
-
-Status EnsemblePatchApplication::SubpatchStreamSets(
- SinkStreamSet* predicted_items,
- SourceStream* correction,
- SourceStreamSet* corrected_items,
- SinkStream* corrected_items_storage) {
- SinkStream linearized_predicted_items;
- if (!predicted_items->CopyTo(&linearized_predicted_items))
- return C_STREAM_ERROR;
-
- SourceStream prediction;
- prediction.Init(linearized_predicted_items);
-
- Status status = ApplySimpleDelta(&prediction,
- correction,
- corrected_items_storage);
- if (status != C_OK)
- return status;
-
- if (!corrected_items->Init(corrected_items_storage->Buffer(),
- corrected_items_storage->Length()))
- return C_STREAM_ERROR;
-
- return C_OK;
-}
-
-Status ApplyEnsemblePatch(SourceStream* base,
- SourceStream* patch,
- SinkStream* output) {
- Status status;
- EnsemblePatchApplication patch_process;
-
- status = patch_process.ReadHeader(patch);
- if (status != C_OK)
- return status;
-
- status = patch_process.InitBase(Region(base->Buffer(), base->Remaining()));
- if (status != C_OK)
- return status;
-
- status = patch_process.ValidateBase();
- if (status != C_OK)
- return status;
-
- // The rest of the patch stream is a StreamSet.
- SourceStreamSet patch_streams;
- patch_streams.Init(patch);
-
- SourceStream* transformation_descriptions = patch_streams.stream(0);
- SourceStream* parameter_correction = patch_streams.stream(1);
- SourceStream* transformed_elements_correction = patch_streams.stream(2);
- SourceStream* ensemble_correction = patch_streams.stream(3);
-
- status = patch_process.ReadInitialParameters(transformation_descriptions);
- if (status != C_OK)
- return status;
-
- SinkStreamSet predicted_parameters;
- status = patch_process.PredictTransformParameters(&predicted_parameters);
- if (status != C_OK)
- return status;
-
- SourceStreamSet corrected_parameters;
- status = patch_process.SubpatchTransformParameters(&predicted_parameters,
- parameter_correction,
- &corrected_parameters);
- if (status != C_OK)
- return status;
-
- SinkStreamSet transformed_elements;
- status = patch_process.TransformUp(&corrected_parameters,
- &transformed_elements);
- if (status != C_OK)
- return status;
-
- SourceStreamSet corrected_transformed_elements;
- status = patch_process.SubpatchTransformedElements(
- &transformed_elements,
- transformed_elements_correction,
- &corrected_transformed_elements);
- if (status != C_OK)
- return status;
-
- SinkStream original_ensemble_and_corrected_base_elements;
- status = patch_process.TransformDown(
- &corrected_transformed_elements,
- &original_ensemble_and_corrected_base_elements);
- if (status != C_OK)
- return status;
-
- SourceStream final_patch_prediction;
- final_patch_prediction.Init(original_ensemble_and_corrected_base_elements);
- status = patch_process.SubpatchFinalOutput(&final_patch_prediction,
- ensemble_correction, output);
- if (status != C_OK)
- return status;
-
- return C_OK;
-}
-
-Status ApplyEnsemblePatch(const wchar_t* old_file_name,
- const wchar_t* patch_file_name,
- const wchar_t* new_file_name) {
- Status status;
-
- // First read enough of the patch file to validate the header is well-formed.
- // A few varint32 numbers should fit in 100.
- FilePath patch_file_path(patch_file_name);
- const int BIG_ENOUGH_FOR_HEADER = 100;
- char buffer[BIG_ENOUGH_FOR_HEADER];
- int read_count =
- file_util::ReadFile(patch_file_path, buffer, sizeof(buffer));
- if (read_count < 0)
- return C_READ_OPEN_ERROR;
-
- // 'Dry-run' the first step of the patch process to validate format of header.
- SourceStream patch_header_stream;
- patch_header_stream.Init(buffer, read_count);
- EnsemblePatchApplication patch_process;
- status = patch_process.ReadHeader(&patch_header_stream);
- if (status != C_OK)
- return status;
-
- // Header smells good so read the whole patch file for real.
- std::string patch_file_buffer;
- if (!file_util::ReadFileToString(patch_file_path, &patch_file_buffer))
- return C_READ_ERROR;
-
- // Read the old_file.
- FilePath old_file_path(old_file_name);
- std::string old_file_buffer;
- if (!file_util::ReadFileToString(old_file_path, &old_file_buffer))
- return C_READ_ERROR;
-
- // Apply patch on streams.
- SourceStream old_source_stream;
- SourceStream patch_source_stream;
- old_source_stream.Init(old_file_buffer);
- patch_source_stream.Init(patch_file_buffer);
- SinkStream new_sink_stream;
- status = ApplyEnsemblePatch(&old_source_stream, &patch_source_stream,
- &new_sink_stream);
-
- // Write the patched data to |new_file_name|.
- FilePath new_file_path(new_file_name);
- int written =
- file_util::WriteFile(
- new_file_path,
- reinterpret_cast<const char*>(new_sink_stream.Buffer()),
- new_sink_stream.Length());
- if (written == -1)
- return C_WRITE_OPEN_ERROR;
- if (written != new_sink_stream.Length())
- return C_WRITE_ERROR;
-
- return C_OK;
-}
-
-} // namespace
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +// This file contains the code to apply a Courgette patch. + +#include "courgette/ensemble.h" + +#include "base/basictypes.h" +#include "base/file_util.h" +#include "base/logging.h" + +#include "courgette/crc.h" +#include "courgette/image_info.h" +#include "courgette/region.h" +#include "courgette/streams.h" +#include "courgette/simple_delta.h" +#include "courgette/win32_x86_patcher.h" + +namespace courgette { + +// EnsemblePatchApplication is all the logic and data required to apply the +// multi-stage patch. +class EnsemblePatchApplication { + public: + EnsemblePatchApplication(); + ~EnsemblePatchApplication(); + + Status ReadHeader(SourceStream* header_stream); + + Status InitBase(const Region& region); + + Status ValidateBase(); + + Status ReadInitialParameters(SourceStream* initial_parameters); + + Status PredictTransformParameters(SinkStreamSet* predicted_parameters); + + Status SubpatchTransformParameters(SinkStreamSet* prediction, + SourceStream* correction, + SourceStreamSet* corrected_parameters); + + Status TransformUp(SourceStreamSet* parameters, + SinkStreamSet* transformed_elements); + + Status SubpatchTransformedElements(SinkStreamSet* elements, + SourceStream* correction, + SourceStreamSet* corrected_elements); + + Status TransformDown(SourceStreamSet* transformed_elements, + SinkStream* basic_elements); + + Status SubpatchFinalOutput(SourceStream* original, + SourceStream* correction, + SinkStream* corrected_ensemble); + + private: + Status SubpatchStreamSets(SinkStreamSet* predicted_items, + SourceStream* correction, + SourceStreamSet* corrected_items, + SinkStream* corrected_items_storage); + + Region base_region_; // Location of in-memory copy of 'old' version. + + uint32 source_checksum_; + uint32 target_checksum_; + + std::vector<TransformationPatcher*> patchers_; + + SinkStream corrected_parameters_storage_; + SinkStream corrected_elements_storage_; + + DISALLOW_COPY_AND_ASSIGN(EnsemblePatchApplication); +}; + +EnsemblePatchApplication::EnsemblePatchApplication() + : source_checksum_(0), target_checksum_(0) { +} + +EnsemblePatchApplication::~EnsemblePatchApplication() { + for (size_t i = 0; i < patchers_.size(); ++i) { + delete patchers_[i]; + } +} + +Status EnsemblePatchApplication::ReadHeader(SourceStream* header_stream) { + uint32 magic; + if (!header_stream->ReadVarint32(&magic)) + return C_BAD_ENSEMBLE_MAGIC; + + if (magic != CourgettePatchFile::kMagic) + return C_BAD_ENSEMBLE_MAGIC; + + uint32 version; + if (!header_stream->ReadVarint32(&version)) + return C_BAD_ENSEMBLE_VERSION; + + if (version != CourgettePatchFile::kVersion) + return C_BAD_ENSEMBLE_VERSION; + + if (!header_stream->ReadVarint32(&source_checksum_)) + return C_BAD_ENSEMBLE_HEADER; + + if (!header_stream->ReadVarint32(&target_checksum_)) + return C_BAD_ENSEMBLE_HEADER; + + return C_OK; +} + +Status EnsemblePatchApplication::InitBase(const Region& region) { + base_region_.assign(region); + return C_OK; +} + +Status EnsemblePatchApplication::ValidateBase() { + uint32 checksum = CalculateCrc(base_region_.start(), base_region_.length()); + if (source_checksum_ != checksum) + return C_BAD_ENSEMBLE_CRC; + + return C_OK; +} + +Status EnsemblePatchApplication::ReadInitialParameters( + SourceStream* transformation_parameters) { + uint32 number_of_transformations = 0; + if (!transformation_parameters->ReadVarint32(&number_of_transformations)) + return C_BAD_ENSEMBLE_HEADER; + + for (size_t i = 0; i < number_of_transformations; ++i) { + uint32 kind; + if (!transformation_parameters->ReadVarint32(&kind)) + return C_BAD_ENSEMBLE_HEADER; + + if (kind == CourgettePatchFile::T_COURGETTE_WIN32_X86) { + TransformationPatcher* patcher = + new CourgetteWin32X86Patcher(base_region_); + patchers_.push_back(patcher); + } else { + return C_BAD_ENSEMBLE_HEADER; + } + } + + for (size_t i = 0; i < patchers_.size(); ++i) { + Status status = patchers_[i]->Init(transformation_parameters); + if (status != C_OK) + return status; + } + + // All transformation_parameters should have been consumed by the above loop. + if (!transformation_parameters->Empty()) + return C_BAD_ENSEMBLE_HEADER; + + return C_OK; +} + +Status EnsemblePatchApplication::PredictTransformParameters( + SinkStreamSet* all_predicted_parameters) { + for (size_t i = 0; i < patchers_.size(); ++i) { + SinkStreamSet single_predicted_parameters; + Status status = + patchers_[i]->PredictTransformParameters(&single_predicted_parameters); + if (status != C_OK) + return status; + if (!all_predicted_parameters->WriteSet(&single_predicted_parameters)) + return C_STREAM_ERROR; + } + return C_OK; +} + +Status EnsemblePatchApplication::SubpatchTransformParameters( + SinkStreamSet* predicted_parameters, + SourceStream* correction, + SourceStreamSet* corrected_parameters) { + return SubpatchStreamSets(predicted_parameters, + correction, + corrected_parameters, + &corrected_parameters_storage_); +} + +Status EnsemblePatchApplication::TransformUp( + SourceStreamSet* parameters, + SinkStreamSet* transformed_elements) { + for (size_t i = 0; i < patchers_.size(); ++i) { + SourceStreamSet single_parameters; + if (!parameters->ReadSet(&single_parameters)) + return C_STREAM_ERROR; + SinkStreamSet single_transformed_element; + Status status = patchers_[i]->Transform(&single_parameters, + &single_transformed_element); + if (status != C_OK) + return status; + if (!single_parameters.Empty()) + return C_STREAM_NOT_CONSUMED; + if (!transformed_elements->WriteSet(&single_transformed_element)) + return C_STREAM_ERROR; + } + + if (!parameters->Empty()) + return C_STREAM_NOT_CONSUMED; + return C_OK; +} + +Status EnsemblePatchApplication::SubpatchTransformedElements( + SinkStreamSet* predicted_elements, + SourceStream* correction, + SourceStreamSet* corrected_elements) { + return SubpatchStreamSets(predicted_elements, + correction, + corrected_elements, + &corrected_elements_storage_); +} + +Status EnsemblePatchApplication::TransformDown( + SourceStreamSet* transformed_elements, + SinkStream* basic_elements) { + // Construct blob of original input followed by reformed elements. + + // The original input: + basic_elements->Write(base_region_.start(), base_region_.length()); + + for (size_t i = 0; i < patchers_.size(); ++i) { + SourceStreamSet single_corrected_element; + if (!transformed_elements->ReadSet(&single_corrected_element)) + return C_STREAM_ERROR; + Status status = patchers_[i]->Reform(&single_corrected_element, + basic_elements); + if (status != C_OK) + return status; + if (!single_corrected_element.Empty()) + return C_STREAM_NOT_CONSUMED; + } + + if (!transformed_elements->Empty()) + return C_STREAM_NOT_CONSUMED; + + return C_OK; +} + +Status EnsemblePatchApplication::SubpatchFinalOutput( + SourceStream* original, + SourceStream* correction, + SinkStream* corrected_ensemble) { + Status delta_status = ApplySimpleDelta(original, correction, + corrected_ensemble); + if (delta_status != C_OK) + return delta_status; + + if (CalculateCrc(corrected_ensemble->Buffer(), + corrected_ensemble->Length()) != target_checksum_) + return C_BAD_ENSEMBLE_CRC; + + return C_OK; +} + +Status EnsemblePatchApplication::SubpatchStreamSets( + SinkStreamSet* predicted_items, + SourceStream* correction, + SourceStreamSet* corrected_items, + SinkStream* corrected_items_storage) { + SinkStream linearized_predicted_items; + if (!predicted_items->CopyTo(&linearized_predicted_items)) + return C_STREAM_ERROR; + + SourceStream prediction; + prediction.Init(linearized_predicted_items); + + Status status = ApplySimpleDelta(&prediction, + correction, + corrected_items_storage); + if (status != C_OK) + return status; + + if (!corrected_items->Init(corrected_items_storage->Buffer(), + corrected_items_storage->Length())) + return C_STREAM_ERROR; + + return C_OK; +} + +Status ApplyEnsemblePatch(SourceStream* base, + SourceStream* patch, + SinkStream* output) { + Status status; + EnsemblePatchApplication patch_process; + + status = patch_process.ReadHeader(patch); + if (status != C_OK) + return status; + + status = patch_process.InitBase(Region(base->Buffer(), base->Remaining())); + if (status != C_OK) + return status; + + status = patch_process.ValidateBase(); + if (status != C_OK) + return status; + + // The rest of the patch stream is a StreamSet. + SourceStreamSet patch_streams; + patch_streams.Init(patch); + + SourceStream* transformation_descriptions = patch_streams.stream(0); + SourceStream* parameter_correction = patch_streams.stream(1); + SourceStream* transformed_elements_correction = patch_streams.stream(2); + SourceStream* ensemble_correction = patch_streams.stream(3); + + status = patch_process.ReadInitialParameters(transformation_descriptions); + if (status != C_OK) + return status; + + SinkStreamSet predicted_parameters; + status = patch_process.PredictTransformParameters(&predicted_parameters); + if (status != C_OK) + return status; + + SourceStreamSet corrected_parameters; + status = patch_process.SubpatchTransformParameters(&predicted_parameters, + parameter_correction, + &corrected_parameters); + if (status != C_OK) + return status; + + SinkStreamSet transformed_elements; + status = patch_process.TransformUp(&corrected_parameters, + &transformed_elements); + if (status != C_OK) + return status; + + SourceStreamSet corrected_transformed_elements; + status = patch_process.SubpatchTransformedElements( + &transformed_elements, + transformed_elements_correction, + &corrected_transformed_elements); + if (status != C_OK) + return status; + + SinkStream original_ensemble_and_corrected_base_elements; + status = patch_process.TransformDown( + &corrected_transformed_elements, + &original_ensemble_and_corrected_base_elements); + if (status != C_OK) + return status; + + SourceStream final_patch_prediction; + final_patch_prediction.Init(original_ensemble_and_corrected_base_elements); + status = patch_process.SubpatchFinalOutput(&final_patch_prediction, + ensemble_correction, output); + if (status != C_OK) + return status; + + return C_OK; +} + +Status ApplyEnsemblePatch(const wchar_t* old_file_name, + const wchar_t* patch_file_name, + const wchar_t* new_file_name) { + Status status; + + // First read enough of the patch file to validate the header is well-formed. + // A few varint32 numbers should fit in 100. + FilePath patch_file_path(patch_file_name); + const int BIG_ENOUGH_FOR_HEADER = 100; + char buffer[BIG_ENOUGH_FOR_HEADER]; + int read_count = + file_util::ReadFile(patch_file_path, buffer, sizeof(buffer)); + if (read_count < 0) + return C_READ_OPEN_ERROR; + + // 'Dry-run' the first step of the patch process to validate format of header. + SourceStream patch_header_stream; + patch_header_stream.Init(buffer, read_count); + EnsemblePatchApplication patch_process; + status = patch_process.ReadHeader(&patch_header_stream); + if (status != C_OK) + return status; + + // Header smells good so read the whole patch file for real. + std::string patch_file_buffer; + if (!file_util::ReadFileToString(patch_file_path, &patch_file_buffer)) + return C_READ_ERROR; + + // Read the old_file. + FilePath old_file_path(old_file_name); + std::string old_file_buffer; + if (!file_util::ReadFileToString(old_file_path, &old_file_buffer)) + return C_READ_ERROR; + + // Apply patch on streams. + SourceStream old_source_stream; + SourceStream patch_source_stream; + old_source_stream.Init(old_file_buffer); + patch_source_stream.Init(patch_file_buffer); + SinkStream new_sink_stream; + status = ApplyEnsemblePatch(&old_source_stream, &patch_source_stream, + &new_sink_stream); + + // Write the patched data to |new_file_name|. + FilePath new_file_path(new_file_name); + int written = + file_util::WriteFile( + new_file_path, + reinterpret_cast<const char*>(new_sink_stream.Buffer()), + new_sink_stream.Length()); + if (written == -1) + return C_WRITE_OPEN_ERROR; + if (written != new_sink_stream.Length()) + return C_WRITE_ERROR; + + return C_OK; +} + +} // namespace diff --git a/third_party/courgette/ensemble_create.cc b/courgette/ensemble_create.cc index 4d64f1c..6e64474 100644 --- a/third_party/courgette/ensemble_create.cc +++ b/courgette/ensemble_create.cc @@ -1,382 +1,382 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-// The main idea in Courgette is to do patching *under a tranformation*. The
-// input is transformed into a new representation, patching occurs in the new
-// repesentation, and then the tranform is reversed to get the patched data.
-//
-// The idea is applied to pieces (or 'elements') of the whole (or 'ensemble').
-// Each of the elements has to go through the same set of steps in lock-step.
-
-// This file contains the code to create the patch.
-
-
-#include "third_party/courgette/ensemble.h"
-
-#include <vector>
-#include <limits>
-
-#include "base/basictypes.h"
-#include "base/logging.h"
-#include "base/time.h"
-
-#include "third_party/courgette/bsdiff.h"
-#include "third_party/courgette/crc.h"
-#include "third_party/courgette/difference_estimator.h"
-#include "third_party/courgette/image_info.h"
-#include "third_party/courgette/streams.h"
-#include "third_party/courgette/region.h"
-#include "third_party/courgette/simple_delta.h"
-
-#include "third_party/courgette/win32_x86_patcher.h"
-#include "third_party/courgette/win32_x86_generator.h"
-
-namespace courgette {
-
-TransformationPatchGenerator::TransformationPatchGenerator(
- Element* old_element,
- Element* new_element,
- TransformationPatcher* patcher)
- : old_element_(old_element),
- new_element_(new_element),
- patcher_(patcher) {
-}
-
-TransformationPatchGenerator::~TransformationPatchGenerator() {
- delete patcher_;
-}
-
-// The default implementation of PredictTransformParameters delegates to the
-// patcher.
-Status TransformationPatchGenerator::PredictTransformParameters(
- SinkStreamSet* prediction) {
- return patcher_->PredictTransformParameters(prediction);
-}
-
-// The default implementation of Reform delegates to the patcher.
-Status TransformationPatchGenerator::Reform(
- SourceStreamSet* transformed_element,
- SinkStream* reformed_element) {
- return patcher_->Reform(transformed_element, reformed_element);
-}
-
-// Makes a TransformationPatchGenerator of the appropriate variety for the
-// Element kind.
-TransformationPatchGenerator* MakeGenerator(Element* old_element,
- Element* new_element) {
- if (new_element->kind() == Element::WIN32_X86_WITH_CODE) {
- CourgetteWin32X86PatchGenerator* generator =
- new CourgetteWin32X86PatchGenerator(
- old_element,
- new_element,
- new CourgetteWin32X86Patcher(old_element->region()));
- return generator;
- } else {
- LOG(WARNING) << "Unexpected Element::Kind " << old_element->kind();
- return NULL;
- }
-}
-
-// FindGenerators finds TransformationPatchGenerators for the elements of
-// |new_ensemble|. For each element of |new_ensemble| we find the closest
-// matching element from |old_ensemble| and use that as the basis for
-// differential compression. The elements have to be the same kind so as to
-// support transformation into the same kind of 'new representation'.
-//
-Status FindGenerators(Ensemble* old_ensemble, Ensemble* new_ensemble,
- std::vector<TransformationPatchGenerator*>* generators) {
- base::Time start_find_time = base::Time::Now();
- old_ensemble->FindEmbeddedElements();
- new_ensemble->FindEmbeddedElements();
- LOG(INFO) << "done FindEmbeddedElements "
- << (base::Time::Now() - start_find_time).InSecondsF();
-
- std::vector<Element*> old_elements(old_ensemble->elements());
- std::vector<Element*> new_elements(new_ensemble->elements());
-
- LOG(INFO) << "old has " << old_elements.size() << " elements";
- LOG(INFO) << "new has " << new_elements.size() << " elements";
-
- DifferenceEstimator difference_estimator;
- std::vector<DifferenceEstimator::Base*> bases;
-
- base::Time start_bases_time = base::Time::Now();
- for (size_t i = 0; i < old_elements.size(); ++i) {
- bases.push_back(
- difference_estimator.MakeBase(old_elements[i]->region()));
- }
- LOG(INFO) << "done make bases "
- << (base::Time::Now() - start_bases_time).InSecondsF()
- << "s";
-
- for (size_t new_index = 0; new_index < new_elements.size(); ++new_index) {
- Element* new_element = new_elements[new_index];
- DifferenceEstimator::Subject* new_subject =
- difference_estimator.MakeSubject(new_element->region());
-
- // Search through old elements to find the best match.
- //
- // TODO(sra): This is O(N x M), i.e. O(N^2) since old_ensemble and
- // new_ensemble probably have a very similar structure. We can make the
- // search faster by making the comparison provided by DifferenceEstimator
- // more nuanced, returning early if the measured difference is greater than
- // the current best. This will be most effective if we can arrange that the
- // first elements we try to match are likely the 'right' ones. We could
- // prioritize elements that are of a similar size or similar position in the
- // sequence of elements.
- //
- Element* best_old_element = NULL;
- size_t best_difference = std::numeric_limits<size_t>::max();
- for (size_t old_index = 0; old_index < old_elements.size(); ++old_index) {
- Element* old_element = old_elements[old_index];
- // Elements of different kinds are incompatible.
- if (old_element->kind() != new_element->kind())
- continue;
-
- base::Time start_compare = base::Time::Now();
- DifferenceEstimator::Base* old_base = bases[old_index];
- size_t difference = difference_estimator.Measure(old_base, new_subject);
-
- LOG(INFO) << "Compare " << old_element->Name()
- << " to " << new_element->Name()
- << " --> " << difference
- << " in " << (base::Time::Now() - start_compare).InSecondsF()
- << "s";
- if (difference == 0) {
- LOG(INFO) << "Skip " << new_element->Name()
- << " - identical to " << old_element->Name();
- best_difference = 0;
- best_old_element = NULL;
- break;
- }
- if (difference < best_difference) {
- best_difference = difference;
- best_old_element = old_element;
- }
- }
-
- if (best_old_element) {
- LOG(INFO) << "Matched " << best_old_element->Name()
- << " to " << new_element->Name()
- << " --> " << best_difference;
- TransformationPatchGenerator* generator =
- MakeGenerator(best_old_element, new_element);
- if (generator)
- generators->push_back(generator);
- }
- }
-
- LOG(INFO) << "done FindGenerators "
- << "found " << generators->size() << " in "
- << (base::Time::Now() - start_find_time).InSecondsF() << "s";
-
- return C_OK;
-}
-
-void FreeGenerators(std::vector<TransformationPatchGenerator*>* generators) {
- for (size_t i = 0; i < generators->size(); ++i) {
- delete (*generators)[i];
- }
- generators->clear();
-}
-
-////////////////////////////////////////////////////////////////////////////////
-
-Status GenerateEnsemblePatch(SourceStream* base,
- SourceStream* update,
- SinkStream* final_patch) {
- Region old_region(base->Buffer(), base->Remaining());
- Region new_region(update->Buffer(), update->Remaining());
- Ensemble old_ensemble(old_region, "old");
- Ensemble new_ensemble(new_region, "new");
- std::vector<TransformationPatchGenerator*> generators;
- Status generators_status = FindGenerators(&old_ensemble, &new_ensemble,
- &generators);
- if (generators_status != C_OK)
- return generators_status;
-
- SinkStreamSet patch_streams;
-
- SinkStream* tranformation_descriptions = patch_streams.stream(0);
- SinkStream* parameter_correction = patch_streams.stream(1);
- SinkStream* transformed_elements_correction = patch_streams.stream(2);
- SinkStream* ensemble_correction = patch_streams.stream(3);
-
- uint32 number_of_transformations = generators.size();
- tranformation_descriptions->WriteVarint32(number_of_transformations);
-
- for (size_t i = 0; i < number_of_transformations; ++i) {
- CourgettePatchFile::TransformationMethodId kind = generators[i]->Kind();
- tranformation_descriptions->WriteVarint32(kind);
- }
-
- for (size_t i = 0; i < number_of_transformations; ++i) {
- Status status =
- generators[i]->WriteInitialParameters(tranformation_descriptions);
- if (status != C_OK)
- return status;
- }
-
- //
- // Generate sub-patch for parameters.
- //
- SinkStreamSet predicted_parameters_sink;
- SinkStreamSet corrected_parameters_sink;
-
- for (size_t i = 0; i < number_of_transformations; ++i) {
- SinkStreamSet single_predicted_parameters;
- Status status;
- status = generators[i]->PredictTransformParameters(
- &single_predicted_parameters);
- if (status != C_OK)
- return status;
- if (!predicted_parameters_sink.WriteSet(&single_predicted_parameters))
- return C_STREAM_ERROR;
-
- SinkStreamSet single_corrected_parameters;
- status = generators[i]->CorrectedTransformParameters(
- &single_corrected_parameters);
- if (status != C_OK)
- return status;
- if (!corrected_parameters_sink.WriteSet(&single_corrected_parameters))
- return C_STREAM_ERROR;
- }
-
- SinkStream linearized_predicted_parameters;
- SinkStream linearized_corrected_parameters;
-
- if (!predicted_parameters_sink.CopyTo(&linearized_predicted_parameters))
- return C_STREAM_ERROR;
- if (!corrected_parameters_sink.CopyTo(&linearized_corrected_parameters))
- return C_STREAM_ERROR;
-
- SourceStream predicted_parameters_source;
- SourceStream corrected_parameters_source;
- predicted_parameters_source.Init(linearized_predicted_parameters);
- corrected_parameters_source.Init(linearized_corrected_parameters);
-
- Status delta1_status = GenerateSimpleDelta(&predicted_parameters_source,
- &corrected_parameters_source,
- parameter_correction);
- if (delta1_status != C_OK)
- return delta1_status;
-
- //
- // Generate sub-patch for elements.
- //
- corrected_parameters_source.Init(linearized_corrected_parameters);
- SourceStreamSet corrected_parameters_source_set;
- if (!corrected_parameters_source_set.Init(&corrected_parameters_source))
- return C_STREAM_ERROR;
-
- SinkStreamSet predicted_transformed_elements;
- SinkStreamSet corrected_transformed_elements;
-
- for (size_t i = 0; i < number_of_transformations; ++i) {
- SourceStreamSet single_parameters;
- if (!corrected_parameters_source_set.ReadSet(&single_parameters))
- return C_STREAM_ERROR;
- SinkStreamSet single_predicted_transformed_element;
- SinkStreamSet single_corrected_transformed_element;
- Status status = generators[i]->Transform(
- &single_parameters,
- &single_predicted_transformed_element,
- &single_corrected_transformed_element);
- if (status != C_OK)
- return status;
- if (!single_parameters.Empty())
- return C_STREAM_NOT_CONSUMED;
- if (!predicted_transformed_elements.WriteSet(
- &single_predicted_transformed_element))
- return C_STREAM_ERROR;
- if (!corrected_transformed_elements.WriteSet(
- &single_corrected_transformed_element))
- return C_STREAM_ERROR;
- }
-
- if (!corrected_parameters_source_set.Empty())
- return C_STREAM_NOT_CONSUMED;
-
- SinkStream linearized_predicted_transformed_elements;
- SinkStream linearized_corrected_transformed_elements;
-
- if (!predicted_transformed_elements.CopyTo(
- &linearized_predicted_transformed_elements))
- return C_STREAM_ERROR;
- if (!corrected_transformed_elements.CopyTo(
- &linearized_corrected_transformed_elements))
- return C_STREAM_ERROR;
-
- SourceStream predicted_transformed_elements_source;
- SourceStream corrected_transformed_elements_source;
- predicted_transformed_elements_source
- .Init(linearized_predicted_transformed_elements);
- corrected_transformed_elements_source
- .Init(linearized_corrected_transformed_elements);
-
- Status delta2_status =
- GenerateSimpleDelta(&predicted_transformed_elements_source,
- &corrected_transformed_elements_source,
- transformed_elements_correction);
- if (delta2_status != C_OK)
- return delta2_status;
-
- //
- // Generate sub-patch for whole enchilada.
- //
- SinkStream predicted_ensemble;
-
- predicted_ensemble.Write(base->Buffer(), base->Remaining());
-
- SourceStreamSet corrected_transformed_elements_source_set;
- corrected_transformed_elements_source
- .Init(linearized_corrected_transformed_elements);
- if (!corrected_transformed_elements_source_set
- .Init(&corrected_transformed_elements_source))
- return C_STREAM_ERROR;
-
- for (size_t i = 0; i < number_of_transformations; ++i) {
- SourceStreamSet single_corrected_transformed_element;
- if (!corrected_transformed_elements_source_set.ReadSet(
- &single_corrected_transformed_element))
- return C_STREAM_ERROR;
- Status status = generators[i]->Reform(&single_corrected_transformed_element,
- &predicted_ensemble);
- if (status != C_OK)
- return status;
- if (!single_corrected_transformed_element.Empty())
- return C_STREAM_NOT_CONSUMED;
- }
-
- if (!corrected_transformed_elements_source_set.Empty())
- return C_STREAM_NOT_CONSUMED;
-
- FreeGenerators(&generators);
-
- SourceStream predicted_ensemble_source;
- predicted_ensemble_source.Init(predicted_ensemble);
- Status delta3_status = GenerateSimpleDelta(&predicted_ensemble_source,
- update,
- ensemble_correction);
- if (delta3_status != C_OK)
- return delta3_status;
-
- //
- // Final output stream has a header followed by a StreamSet.
- //
- final_patch->WriteVarint32(CourgettePatchFile::kMagic);
- final_patch->WriteVarint32(CourgettePatchFile::kVersion);
-
- final_patch->WriteVarint32(
- CalculateCrc(old_region.start(), old_region.length()));
- final_patch->WriteVarint32(
- CalculateCrc(new_region.start(), new_region.length()));
-
- if (!patch_streams.CopyTo(final_patch))
- return C_STREAM_ERROR;
-
- return C_OK;
-}
-
-} // namespace
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +// The main idea in Courgette is to do patching *under a tranformation*. The +// input is transformed into a new representation, patching occurs in the new +// repesentation, and then the tranform is reversed to get the patched data. +// +// The idea is applied to pieces (or 'elements') of the whole (or 'ensemble'). +// Each of the elements has to go through the same set of steps in lock-step. + +// This file contains the code to create the patch. + + +#include "courgette/ensemble.h" + +#include <vector> +#include <limits> + +#include "base/basictypes.h" +#include "base/logging.h" +#include "base/time.h" + +#include "courgette/third_party/bsdiff.h" +#include "courgette/crc.h" +#include "courgette/difference_estimator.h" +#include "courgette/image_info.h" +#include "courgette/streams.h" +#include "courgette/region.h" +#include "courgette/simple_delta.h" + +#include "courgette/win32_x86_patcher.h" +#include "courgette/win32_x86_generator.h" + +namespace courgette { + +TransformationPatchGenerator::TransformationPatchGenerator( + Element* old_element, + Element* new_element, + TransformationPatcher* patcher) + : old_element_(old_element), + new_element_(new_element), + patcher_(patcher) { +} + +TransformationPatchGenerator::~TransformationPatchGenerator() { + delete patcher_; +} + +// The default implementation of PredictTransformParameters delegates to the +// patcher. +Status TransformationPatchGenerator::PredictTransformParameters( + SinkStreamSet* prediction) { + return patcher_->PredictTransformParameters(prediction); +} + +// The default implementation of Reform delegates to the patcher. +Status TransformationPatchGenerator::Reform( + SourceStreamSet* transformed_element, + SinkStream* reformed_element) { + return patcher_->Reform(transformed_element, reformed_element); +} + +// Makes a TransformationPatchGenerator of the appropriate variety for the +// Element kind. +TransformationPatchGenerator* MakeGenerator(Element* old_element, + Element* new_element) { + if (new_element->kind() == Element::WIN32_X86_WITH_CODE) { + CourgetteWin32X86PatchGenerator* generator = + new CourgetteWin32X86PatchGenerator( + old_element, + new_element, + new CourgetteWin32X86Patcher(old_element->region())); + return generator; + } else { + LOG(WARNING) << "Unexpected Element::Kind " << old_element->kind(); + return NULL; + } +} + +// FindGenerators finds TransformationPatchGenerators for the elements of +// |new_ensemble|. For each element of |new_ensemble| we find the closest +// matching element from |old_ensemble| and use that as the basis for +// differential compression. The elements have to be the same kind so as to +// support transformation into the same kind of 'new representation'. +// +Status FindGenerators(Ensemble* old_ensemble, Ensemble* new_ensemble, + std::vector<TransformationPatchGenerator*>* generators) { + base::Time start_find_time = base::Time::Now(); + old_ensemble->FindEmbeddedElements(); + new_ensemble->FindEmbeddedElements(); + LOG(INFO) << "done FindEmbeddedElements " + << (base::Time::Now() - start_find_time).InSecondsF(); + + std::vector<Element*> old_elements(old_ensemble->elements()); + std::vector<Element*> new_elements(new_ensemble->elements()); + + LOG(INFO) << "old has " << old_elements.size() << " elements"; + LOG(INFO) << "new has " << new_elements.size() << " elements"; + + DifferenceEstimator difference_estimator; + std::vector<DifferenceEstimator::Base*> bases; + + base::Time start_bases_time = base::Time::Now(); + for (size_t i = 0; i < old_elements.size(); ++i) { + bases.push_back( + difference_estimator.MakeBase(old_elements[i]->region())); + } + LOG(INFO) << "done make bases " + << (base::Time::Now() - start_bases_time).InSecondsF() + << "s"; + + for (size_t new_index = 0; new_index < new_elements.size(); ++new_index) { + Element* new_element = new_elements[new_index]; + DifferenceEstimator::Subject* new_subject = + difference_estimator.MakeSubject(new_element->region()); + + // Search through old elements to find the best match. + // + // TODO(sra): This is O(N x M), i.e. O(N^2) since old_ensemble and + // new_ensemble probably have a very similar structure. We can make the + // search faster by making the comparison provided by DifferenceEstimator + // more nuanced, returning early if the measured difference is greater than + // the current best. This will be most effective if we can arrange that the + // first elements we try to match are likely the 'right' ones. We could + // prioritize elements that are of a similar size or similar position in the + // sequence of elements. + // + Element* best_old_element = NULL; + size_t best_difference = std::numeric_limits<size_t>::max(); + for (size_t old_index = 0; old_index < old_elements.size(); ++old_index) { + Element* old_element = old_elements[old_index]; + // Elements of different kinds are incompatible. + if (old_element->kind() != new_element->kind()) + continue; + + base::Time start_compare = base::Time::Now(); + DifferenceEstimator::Base* old_base = bases[old_index]; + size_t difference = difference_estimator.Measure(old_base, new_subject); + + LOG(INFO) << "Compare " << old_element->Name() + << " to " << new_element->Name() + << " --> " << difference + << " in " << (base::Time::Now() - start_compare).InSecondsF() + << "s"; + if (difference == 0) { + LOG(INFO) << "Skip " << new_element->Name() + << " - identical to " << old_element->Name(); + best_difference = 0; + best_old_element = NULL; + break; + } + if (difference < best_difference) { + best_difference = difference; + best_old_element = old_element; + } + } + + if (best_old_element) { + LOG(INFO) << "Matched " << best_old_element->Name() + << " to " << new_element->Name() + << " --> " << best_difference; + TransformationPatchGenerator* generator = + MakeGenerator(best_old_element, new_element); + if (generator) + generators->push_back(generator); + } + } + + LOG(INFO) << "done FindGenerators " + << "found " << generators->size() << " in " + << (base::Time::Now() - start_find_time).InSecondsF() << "s"; + + return C_OK; +} + +void FreeGenerators(std::vector<TransformationPatchGenerator*>* generators) { + for (size_t i = 0; i < generators->size(); ++i) { + delete (*generators)[i]; + } + generators->clear(); +} + +//////////////////////////////////////////////////////////////////////////////// + +Status GenerateEnsemblePatch(SourceStream* base, + SourceStream* update, + SinkStream* final_patch) { + Region old_region(base->Buffer(), base->Remaining()); + Region new_region(update->Buffer(), update->Remaining()); + Ensemble old_ensemble(old_region, "old"); + Ensemble new_ensemble(new_region, "new"); + std::vector<TransformationPatchGenerator*> generators; + Status generators_status = FindGenerators(&old_ensemble, &new_ensemble, + &generators); + if (generators_status != C_OK) + return generators_status; + + SinkStreamSet patch_streams; + + SinkStream* tranformation_descriptions = patch_streams.stream(0); + SinkStream* parameter_correction = patch_streams.stream(1); + SinkStream* transformed_elements_correction = patch_streams.stream(2); + SinkStream* ensemble_correction = patch_streams.stream(3); + + uint32 number_of_transformations = generators.size(); + tranformation_descriptions->WriteVarint32(number_of_transformations); + + for (size_t i = 0; i < number_of_transformations; ++i) { + CourgettePatchFile::TransformationMethodId kind = generators[i]->Kind(); + tranformation_descriptions->WriteVarint32(kind); + } + + for (size_t i = 0; i < number_of_transformations; ++i) { + Status status = + generators[i]->WriteInitialParameters(tranformation_descriptions); + if (status != C_OK) + return status; + } + + // + // Generate sub-patch for parameters. + // + SinkStreamSet predicted_parameters_sink; + SinkStreamSet corrected_parameters_sink; + + for (size_t i = 0; i < number_of_transformations; ++i) { + SinkStreamSet single_predicted_parameters; + Status status; + status = generators[i]->PredictTransformParameters( + &single_predicted_parameters); + if (status != C_OK) + return status; + if (!predicted_parameters_sink.WriteSet(&single_predicted_parameters)) + return C_STREAM_ERROR; + + SinkStreamSet single_corrected_parameters; + status = generators[i]->CorrectedTransformParameters( + &single_corrected_parameters); + if (status != C_OK) + return status; + if (!corrected_parameters_sink.WriteSet(&single_corrected_parameters)) + return C_STREAM_ERROR; + } + + SinkStream linearized_predicted_parameters; + SinkStream linearized_corrected_parameters; + + if (!predicted_parameters_sink.CopyTo(&linearized_predicted_parameters)) + return C_STREAM_ERROR; + if (!corrected_parameters_sink.CopyTo(&linearized_corrected_parameters)) + return C_STREAM_ERROR; + + SourceStream predicted_parameters_source; + SourceStream corrected_parameters_source; + predicted_parameters_source.Init(linearized_predicted_parameters); + corrected_parameters_source.Init(linearized_corrected_parameters); + + Status delta1_status = GenerateSimpleDelta(&predicted_parameters_source, + &corrected_parameters_source, + parameter_correction); + if (delta1_status != C_OK) + return delta1_status; + + // + // Generate sub-patch for elements. + // + corrected_parameters_source.Init(linearized_corrected_parameters); + SourceStreamSet corrected_parameters_source_set; + if (!corrected_parameters_source_set.Init(&corrected_parameters_source)) + return C_STREAM_ERROR; + + SinkStreamSet predicted_transformed_elements; + SinkStreamSet corrected_transformed_elements; + + for (size_t i = 0; i < number_of_transformations; ++i) { + SourceStreamSet single_parameters; + if (!corrected_parameters_source_set.ReadSet(&single_parameters)) + return C_STREAM_ERROR; + SinkStreamSet single_predicted_transformed_element; + SinkStreamSet single_corrected_transformed_element; + Status status = generators[i]->Transform( + &single_parameters, + &single_predicted_transformed_element, + &single_corrected_transformed_element); + if (status != C_OK) + return status; + if (!single_parameters.Empty()) + return C_STREAM_NOT_CONSUMED; + if (!predicted_transformed_elements.WriteSet( + &single_predicted_transformed_element)) + return C_STREAM_ERROR; + if (!corrected_transformed_elements.WriteSet( + &single_corrected_transformed_element)) + return C_STREAM_ERROR; + } + + if (!corrected_parameters_source_set.Empty()) + return C_STREAM_NOT_CONSUMED; + + SinkStream linearized_predicted_transformed_elements; + SinkStream linearized_corrected_transformed_elements; + + if (!predicted_transformed_elements.CopyTo( + &linearized_predicted_transformed_elements)) + return C_STREAM_ERROR; + if (!corrected_transformed_elements.CopyTo( + &linearized_corrected_transformed_elements)) + return C_STREAM_ERROR; + + SourceStream predicted_transformed_elements_source; + SourceStream corrected_transformed_elements_source; + predicted_transformed_elements_source + .Init(linearized_predicted_transformed_elements); + corrected_transformed_elements_source + .Init(linearized_corrected_transformed_elements); + + Status delta2_status = + GenerateSimpleDelta(&predicted_transformed_elements_source, + &corrected_transformed_elements_source, + transformed_elements_correction); + if (delta2_status != C_OK) + return delta2_status; + + // + // Generate sub-patch for whole enchilada. + // + SinkStream predicted_ensemble; + + predicted_ensemble.Write(base->Buffer(), base->Remaining()); + + SourceStreamSet corrected_transformed_elements_source_set; + corrected_transformed_elements_source + .Init(linearized_corrected_transformed_elements); + if (!corrected_transformed_elements_source_set + .Init(&corrected_transformed_elements_source)) + return C_STREAM_ERROR; + + for (size_t i = 0; i < number_of_transformations; ++i) { + SourceStreamSet single_corrected_transformed_element; + if (!corrected_transformed_elements_source_set.ReadSet( + &single_corrected_transformed_element)) + return C_STREAM_ERROR; + Status status = generators[i]->Reform(&single_corrected_transformed_element, + &predicted_ensemble); + if (status != C_OK) + return status; + if (!single_corrected_transformed_element.Empty()) + return C_STREAM_NOT_CONSUMED; + } + + if (!corrected_transformed_elements_source_set.Empty()) + return C_STREAM_NOT_CONSUMED; + + FreeGenerators(&generators); + + SourceStream predicted_ensemble_source; + predicted_ensemble_source.Init(predicted_ensemble); + Status delta3_status = GenerateSimpleDelta(&predicted_ensemble_source, + update, + ensemble_correction); + if (delta3_status != C_OK) + return delta3_status; + + // + // Final output stream has a header followed by a StreamSet. + // + final_patch->WriteVarint32(CourgettePatchFile::kMagic); + final_patch->WriteVarint32(CourgettePatchFile::kVersion); + + final_patch->WriteVarint32( + CalculateCrc(old_region.start(), old_region.length())); + final_patch->WriteVarint32( + CalculateCrc(new_region.start(), new_region.length())); + + if (!patch_streams.CopyTo(final_patch)) + return C_STREAM_ERROR; + + return C_OK; +} + +} // namespace diff --git a/third_party/courgette/image_info.cc b/courgette/image_info.cc index ca615dd..5651f36 100644 --- a/third_party/courgette/image_info.cc +++ b/courgette/image_info.cc @@ -2,7 +2,7 @@ // Use of this source code is governed by a BSD-style license that can be // found in the LICENSE file. -#include "third_party/courgette/image_info.h" +#include "courgette/image_info.h" #include <memory.h> #include <algorithm> @@ -389,4 +389,3 @@ bool PEInfo::ParseRelocs(std::vector<RVA> *relocs) { } } // namespace courgette - diff --git a/third_party/courgette/image_info.h b/courgette/image_info.h index ffdbe97..53a0be7 100644 --- a/third_party/courgette/image_info.h +++ b/courgette/image_info.h @@ -197,4 +197,3 @@ class PEInfo { } // namespace #endif // COURGETTE_IMAGE_INFO_H_ - diff --git a/third_party/courgette/image_info_unittest.cc b/courgette/image_info_unittest.cc index 43ef65f..6d6deb5 100644 --- a/third_party/courgette/image_info_unittest.cc +++ b/courgette/image_info_unittest.cc @@ -2,7 +2,7 @@ // Use of this source code is governed by a BSD-style license that can be // found in the LICENSE file. -#include "third_party/courgette/image_info.h" +#include "courgette/image_info.h" #include <string> @@ -20,7 +20,6 @@ class ImageInfoTest : public testing::Test { private: void SetUp() { PathService::Get(base::DIR_SOURCE_ROOT, &test_dir_); - // file_util::AppendToPath(&test_dir_, L"third_party"); file_util::AppendToPath(&test_dir_, L"courgette"); file_util::AppendToPath(&test_dir_, L"testdata"); } diff --git a/third_party/courgette/memory_monitor.cc b/courgette/memory_monitor.cc index 2077f12..51cb0cc 100644 --- a/third_party/courgette/memory_monitor.cc +++ b/courgette/memory_monitor.cc @@ -1,122 +1,122 @@ -#include <stdio.h>
-#include <map>
-
-#include "base/logging.h"
-#include "base/string_util.h"
-
-bool inH = true;
-struct H {
- H() { inH = false; tick_ = 0; bw_ = 0; d_bw_ = d_tick_ = 0; m_bw_ = 0; mem_ = high_ = 0;}
- ~H() {
- inH = true;
- int i = 0;
- for (M::iterator p = m_.begin(); p != m_.end(); ++p, ++i) {
- size_t s = p->first;
- LOG(INFO) << StringPrintf("%3d %8u: %8u %8u %8u %8u", i, s,
- m_[s], c_[s], h_[s], h_[s] * s);
- }
- LOG(INFO) << "Peak " << fmt(high_);
- }
-
- std::string fmt(size_t s) {
- if (s > 1000000000) return StringPrintf("%.3gG", s/(1000000000.0));
- if (s > 1000000) return StringPrintf("%.3gM", s/(1000000.));
- if (s > 9999) return StringPrintf("%.3gk", s/(1000.));
- return StringPrintf("%d", (int)s);
- }
-
- void tick(size_t w, char sign) {
- d_tick_ += 1;
- d_bw_ += w;
- const size_t T = 4*4*1024;
- const size_t M = 4*1024*1024;
- bool print = false;
- if (d_tick_ >= T) {
- tick_ += (d_tick_/T)*T;
- d_tick_ %= T;
- print = true;
- }
- if (d_bw_ >= M) {
- bw_ += (d_bw_/M) * M;
- d_bw_ %= M;
- print = true;
- }
- if (!print) return;
- std::string o;
- StringAppendF(&o, "%u:", tick_ + d_tick_);
- StringAppendF(&o, " (%c%s)", sign, fmt(w).c_str());
- size_t sum = 0;
- for (M::iterator p = c_.begin(); p != c_.end(); ++p) {
- size_t s = p->first;
- size_t n = p->second;
- if (n) {
- if (s*n >= 64*1024)
- if (n == 1)
- StringAppendF(&o, " %s", fmt(s).c_str());
- else
- StringAppendF(&o, " %s*%u", fmt(s).c_str(), n);
- sum += s*n;
- }
- }
- StringAppendF(&o, " = %s", fmt(sum).c_str());
- LOG(INFO) << o;
- //printf("%s\n", o.c_str());
- if (sum > 200*1024*1024) {
- // __asm int 3;
- m_bw_ = sum;
- }
- }
- void add(size_t s, void *p) {
- if (!inH) {
- inH = true;
- mem_ += s; if (mem_ > high_) high_ = mem_;
- c_[s] += 1;
- m_[s] += 1;
- if (c_[s] > h_[s]) h_[s] = c_[s];
- allocs_[p] = s;
- inH = false;
- tick(s, '+');
- }
- }
-
- void sub(void *p) {
- if (!inH) {
- inH = true;
- size_t s = allocs_[p];
- if (s) {
- mem_ -= s;
- c_[s] -= 1;
- allocs_[p] = 0;
- tick(s, '-');
- }
- inH = false;
- }
- }
-
- typedef std::map<size_t, size_t> M;
- M m_;
- M c_;
- M h_;
-
- size_t bw_;
- size_t d_bw_;
- size_t tick_;
- size_t d_tick_;
- size_t m_bw_;
- size_t mem_;
- size_t high_;
-
- std::map<void*, size_t> allocs_;
-} _H;
-
-void* operator new(size_t s) {
- //printf("%u\n", s);
- void *p = malloc(s);
- _H.add(s, p);
- return p;
-}
-
-void operator delete(void *p) {
- _H.sub(p);
- free(p);
-}
+#include <stdio.h> +#include <map> + +#include "base/logging.h" +#include "base/string_util.h" + +bool inH = true; +struct H { + H() { inH = false; tick_ = 0; bw_ = 0; d_bw_ = d_tick_ = 0; m_bw_ = 0; mem_ = high_ = 0;} + ~H() { + inH = true; + int i = 0; + for (M::iterator p = m_.begin(); p != m_.end(); ++p, ++i) { + size_t s = p->first; + LOG(INFO) << StringPrintf("%3d %8u: %8u %8u %8u %8u", i, s, + m_[s], c_[s], h_[s], h_[s] * s); + } + LOG(INFO) << "Peak " << fmt(high_); + } + + std::string fmt(size_t s) { + if (s > 1000000000) return StringPrintf("%.3gG", s/(1000000000.0)); + if (s > 1000000) return StringPrintf("%.3gM", s/(1000000.)); + if (s > 9999) return StringPrintf("%.3gk", s/(1000.)); + return StringPrintf("%d", (int)s); + } + + void tick(size_t w, char sign) { + d_tick_ += 1; + d_bw_ += w; + const size_t T = 4*4*1024; + const size_t M = 4*1024*1024; + bool print = false; + if (d_tick_ >= T) { + tick_ += (d_tick_/T)*T; + d_tick_ %= T; + print = true; + } + if (d_bw_ >= M) { + bw_ += (d_bw_/M) * M; + d_bw_ %= M; + print = true; + } + if (!print) return; + std::string o; + StringAppendF(&o, "%u:", tick_ + d_tick_); + StringAppendF(&o, " (%c%s)", sign, fmt(w).c_str()); + size_t sum = 0; + for (M::iterator p = c_.begin(); p != c_.end(); ++p) { + size_t s = p->first; + size_t n = p->second; + if (n) { + if (s*n >= 64*1024) + if (n == 1) + StringAppendF(&o, " %s", fmt(s).c_str()); + else + StringAppendF(&o, " %s*%u", fmt(s).c_str(), n); + sum += s*n; + } + } + StringAppendF(&o, " = %s", fmt(sum).c_str()); + LOG(INFO) << o; + //printf("%s\n", o.c_str()); + if (sum > 200*1024*1024) { + // __asm int 3; + m_bw_ = sum; + } + } + void add(size_t s, void *p) { + if (!inH) { + inH = true; + mem_ += s; if (mem_ > high_) high_ = mem_; + c_[s] += 1; + m_[s] += 1; + if (c_[s] > h_[s]) h_[s] = c_[s]; + allocs_[p] = s; + inH = false; + tick(s, '+'); + } + } + + void sub(void *p) { + if (!inH) { + inH = true; + size_t s = allocs_[p]; + if (s) { + mem_ -= s; + c_[s] -= 1; + allocs_[p] = 0; + tick(s, '-'); + } + inH = false; + } + } + + typedef std::map<size_t, size_t> M; + M m_; + M c_; + M h_; + + size_t bw_; + size_t d_bw_; + size_t tick_; + size_t d_tick_; + size_t m_bw_; + size_t mem_; + size_t high_; + + std::map<void*, size_t> allocs_; +} _H; + +void* operator new(size_t s) { + //printf("%u\n", s); + void *p = malloc(s); + _H.add(s, p); + return p; +} + +void operator delete(void *p) { + _H.sub(p); + free(p); +} diff --git a/third_party/courgette/region.h b/courgette/region.h index 00a130f..4aafbca 100644 --- a/third_party/courgette/region.h +++ b/courgette/region.h @@ -1,62 +1,62 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-#ifndef COURGETTE_REGION_H_
-#define COURGETTE_REGION_H_
-
-#include <string>
-
-#include "base/basictypes.h"
-
-namespace courgette {
-
-// A Region is a descriptor for a region of memory. It has a start address and
-// a length measured in bytes. The Region object does not own the memory.
-//
-class Region {
- public:
- // Default constructor: and empty region.
- Region() : start_(NULL), length_(0) {}
-
- // Usual constructor for regions given a |start| address and |length|.
- Region(const void* start, size_t length)
- : start_(static_cast<const uint8*>(start)),
- length_(length) {
- }
-
- // String constructor. Region is owned by the string, so the string should
- // have a lifetime greater than the region.
- explicit Region(const std::string& string)
- : start_(reinterpret_cast<const uint8*>(string.c_str())),
- length_(string.length()) {
- }
-
- // Copy constructor.
- Region(const Region& other) : start_(other.start_), length_(other.length_) {}
-
- // Assignment 'operator' makes |this| region the same as |other|.
- Region& assign(const Region& other) {
- this->start_ = other.start_;
- this->length_ = other.length_;
- return *this;
- }
-
- // Returns the starting address of the region.
- const uint8* start() const { return start_; }
-
- // Returns the length of the region.
- size_t length() const { return length_; }
-
- // Returns the address after the last byte of the region.
- const uint8* end() const { return start_ + length_; }
-
- private:
- const uint8* start_;
- size_t length_;
-
- void operator=(const Region&); // Disallow assignment operator.
-};
-
-} // namespace
-#endif // COURGETTE_REGION_H_
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +#ifndef COURGETTE_REGION_H_ +#define COURGETTE_REGION_H_ + +#include <string> + +#include "base/basictypes.h" + +namespace courgette { + +// A Region is a descriptor for a region of memory. It has a start address and +// a length measured in bytes. The Region object does not own the memory. +// +class Region { + public: + // Default constructor: and empty region. + Region() : start_(NULL), length_(0) {} + + // Usual constructor for regions given a |start| address and |length|. + Region(const void* start, size_t length) + : start_(static_cast<const uint8*>(start)), + length_(length) { + } + + // String constructor. Region is owned by the string, so the string should + // have a lifetime greater than the region. + explicit Region(const std::string& string) + : start_(reinterpret_cast<const uint8*>(string.c_str())), + length_(string.length()) { + } + + // Copy constructor. + Region(const Region& other) : start_(other.start_), length_(other.length_) {} + + // Assignment 'operator' makes |this| region the same as |other|. + Region& assign(const Region& other) { + this->start_ = other.start_; + this->length_ = other.length_; + return *this; + } + + // Returns the starting address of the region. + const uint8* start() const { return start_; } + + // Returns the length of the region. + size_t length() const { return length_; } + + // Returns the address after the last byte of the region. + const uint8* end() const { return start_ + length_; } + + private: + const uint8* start_; + size_t length_; + + void operator=(const Region&); // Disallow assignment operator. +}; + +} // namespace +#endif // COURGETTE_REGION_H_ diff --git a/third_party/courgette/run_all_unittests.cc b/courgette/run_all_unittests.cc index 5a17ba1..b5fc4a6 100644 --- a/third_party/courgette/run_all_unittests.cc +++ b/courgette/run_all_unittests.cc @@ -1,10 +1,9 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-#include "base/test_suite.h"
-
-int main(int argc, char** argv) {
- return TestSuite(argc, argv).Run();
-}
-
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +#include "base/test_suite.h" + +int main(int argc, char** argv) { + return TestSuite(argc, argv).Run(); +} diff --git a/third_party/courgette/simple_delta.cc b/courgette/simple_delta.cc index 8f93095..71038fd 100644 --- a/third_party/courgette/simple_delta.cc +++ b/courgette/simple_delta.cc @@ -1,41 +1,41 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-// Implementation of the byte-level differential compression used internally by
-// Courgette.
-
-#include "third_party/courgette/simple_delta.h"
-
-#include "base/basictypes.h"
-#include "base/logging.h"
-
-#include "third_party/courgette/bsdiff.h"
-
-namespace courgette {
-
-namespace {
-
-Status BSDiffStatusToStatus(BSDiffStatus status) {
- switch (status) {
- case OK: return C_OK;
- case CRC_ERROR: return C_BINARY_DIFF_CRC_ERROR;
- default: return C_GENERAL_ERROR;
- }
-}
-
-}
-
-Status ApplySimpleDelta(SourceStream* old, SourceStream* delta,
- SinkStream* target) {
- return BSDiffStatusToStatus(ApplyBinaryPatch(old, delta, target));
-}
-
-Status GenerateSimpleDelta(SourceStream* old, SourceStream* target,
- SinkStream* delta) {
- LOG(INFO) << "GenerateSimpleDelta "
- << old->Remaining() << " " << target->Remaining();
- return BSDiffStatusToStatus(CreateBinaryPatch(old, target, delta));
-}
-
-} // namespace courgette
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +// Implementation of the byte-level differential compression used internally by +// Courgette. + +#include "courgette/simple_delta.h" + +#include "base/basictypes.h" +#include "base/logging.h" + +#include "courgette/third_party/bsdiff.h" + +namespace courgette { + +namespace { + +Status BSDiffStatusToStatus(BSDiffStatus status) { + switch (status) { + case OK: return C_OK; + case CRC_ERROR: return C_BINARY_DIFF_CRC_ERROR; + default: return C_GENERAL_ERROR; + } +} + +} + +Status ApplySimpleDelta(SourceStream* old, SourceStream* delta, + SinkStream* target) { + return BSDiffStatusToStatus(ApplyBinaryPatch(old, delta, target)); +} + +Status GenerateSimpleDelta(SourceStream* old, SourceStream* target, + SinkStream* delta) { + LOG(INFO) << "GenerateSimpleDelta " + << old->Remaining() << " " << target->Remaining(); + return BSDiffStatusToStatus(CreateBinaryPatch(old, target, delta)); +} + +} // namespace courgette diff --git a/third_party/courgette/simple_delta.h b/courgette/simple_delta.h index acb2227..d35333a 100644 --- a/third_party/courgette/simple_delta.h +++ b/courgette/simple_delta.h @@ -1,23 +1,23 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-// Byte level differential compression algorithm used by Courgette.
-
-#ifndef COURGETTE_SIMPLE_DELTA_H_
-#define COURGETTE_SIMPLE_DELTA_H_
-
-#include "third_party/courgette/courgette.h"
-#include "third_party/courgette/streams.h"
-
-namespace courgette {
-
-Status ApplySimpleDelta(SourceStream* old, SourceStream* delta,
- SinkStream* target);
-
-Status GenerateSimpleDelta(SourceStream* old, SourceStream* target,
- SinkStream* delta);
-
-} // namespace courgette
-
-#endif // COURGETTE_SIMPLE_DELTA_H_
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +// Byte level differential compression algorithm used by Courgette. + +#ifndef COURGETTE_SIMPLE_DELTA_H_ +#define COURGETTE_SIMPLE_DELTA_H_ + +#include "courgette/courgette.h" +#include "courgette/streams.h" + +namespace courgette { + +Status ApplySimpleDelta(SourceStream* old, SourceStream* delta, + SinkStream* target); + +Status GenerateSimpleDelta(SourceStream* old, SourceStream* target, + SinkStream* delta); + +} // namespace courgette + +#endif // COURGETTE_SIMPLE_DELTA_H_ diff --git a/third_party/courgette/streams.cc b/courgette/streams.cc index 9870cd2..365a416 100644 --- a/third_party/courgette/streams.cc +++ b/courgette/streams.cc @@ -16,7 +16,7 @@ // we use it to cluster similar kinds of data which helps to generate longer // common subsequences and repeated sequences. -#include "third_party/courgette/streams.h" +#include "courgette/streams.h" #include <io.h> #include <memory.h> @@ -31,7 +31,7 @@ static const unsigned int kStreamsSerializationFormatVersion = 20090218; // // This is a cut down Varint implementation, implementing only what we use for -// streams. The original implementation is at google3/util/coding/varint.h +// streams. // class Varint { public: diff --git a/third_party/courgette/streams.h b/courgette/streams.h index 4985310..2fb824f 100644 --- a/third_party/courgette/streams.h +++ b/courgette/streams.h @@ -18,7 +18,7 @@ #include "base/basictypes.h" -#include "third_party/courgette/region.h" +#include "courgette/region.h" namespace courgette { diff --git a/third_party/courgette/streams_unittest.cc b/courgette/streams_unittest.cc index 90a6061..dd330ce 100644 --- a/third_party/courgette/streams_unittest.cc +++ b/courgette/streams_unittest.cc @@ -2,7 +2,7 @@ // Use of this source code is governed by a BSD-style license that can be // found in the LICENSE file. -#include "third_party/courgette/streams.h" +#include "courgette/streams.h" #include "testing/gtest/include/gtest/gtest.h" diff --git a/third_party/courgette/testdata/en-US.dll b/courgette/testdata/en-US.dll Binary files differindex f1ad46c..f1ad46c 100644 --- a/third_party/courgette/testdata/en-US.dll +++ b/courgette/testdata/en-US.dll diff --git a/third_party/courgette/testdata/setup1.exe b/courgette/testdata/setup1.exe Binary files differindex da072f5..da072f5 100644 --- a/third_party/courgette/testdata/setup1.exe +++ b/courgette/testdata/setup1.exe diff --git a/third_party/courgette/testdata/setup2.exe b/courgette/testdata/setup2.exe Binary files differindex 577a47f..577a47f 100644 --- a/third_party/courgette/testdata/setup2.exe +++ b/courgette/testdata/setup2.exe diff --git a/third_party/courgette/LICENCE.bsdiff b/courgette/third_party/LICENCE index c146b5b..c146b5b 100644 --- a/third_party/courgette/LICENCE.bsdiff +++ b/courgette/third_party/LICENCE diff --git a/courgette/third_party/README.chromium b/courgette/third_party/README.chromium new file mode 100644 index 0000000..d3037a1 --- /dev/null +++ b/courgette/third_party/README.chromium @@ -0,0 +1,23 @@ +This directory contains an extensively modified version of Colin Percival's
+bsdiff, available in its original form from:
+
+ http://www.daemonology.net/bsdiff/
+
+The basic principles of operation are best understood by reading Colin's
+unpublised paper:
+
+Colin Percival, Naive differences of executable code, http://www.daemonology.net/bsdiff/, 200
+
+The copy on this directory so extensively modified that the binary format is
+incompatible with the original and it cannot be compiled outside the Chromium
+source tree or the Courgette project.
+
+List of changes made to original code:
+ - wrapped functions in 'courgette' namespace
+ - renamed .c files to .cc
+ - added bsdiff.h header file
+ - changed the code to use streams.h from courgette
+ - changed the encoding of numbers to use the 'varint' encoding
+ - reformatted code to be closer to Google coding standards
+ - renamed variables
+ - added comments
diff --git a/third_party/courgette/bsdiff.h b/courgette/third_party/bsdiff.h index bf7fdec..bf7fdec 100644 --- a/third_party/courgette/bsdiff.h +++ b/courgette/third_party/bsdiff.h diff --git a/third_party/courgette/bsdiff_apply.cc b/courgette/third_party/bsdiff_apply.cc index f45ae14..4b6a011 100644 --- a/third_party/courgette/bsdiff_apply.cc +++ b/courgette/third_party/bsdiff_apply.cc @@ -32,10 +32,10 @@ // Use of this source code is governed by a BSD-style license that can be // found in the LICENSE file. -#include "third_party/courgette/bsdiff.h" +#include "courgette/third_party/bsdiff.h" -#include "third_party/courgette/crc.h" -#include "third_party/courgette/streams.h" +#include "courgette/crc.h" +#include "courgette/streams.h" namespace courgette { diff --git a/third_party/courgette/bsdiff_create.cc b/courgette/third_party/bsdiff_create.cc index cdd6cb6..d59dc21 100644 --- a/third_party/courgette/bsdiff_create.cc +++ b/courgette/third_party/bsdiff_create.cc @@ -19,7 +19,7 @@ --Stephen Adams <sra@chromium.org> */ -#include "third_party/courgette/bsdiff.h" +#include "courgette/third_party/bsdiff.h" #include <stdlib.h> #include <algorithm> @@ -29,8 +29,8 @@ #include "base/string_util.h" #include "base/time.h" -#include "third_party/courgette/crc.h" -#include "third_party/courgette/streams.h" +#include "courgette/crc.h" +#include "courgette/streams.h" namespace courgette { diff --git a/third_party/courgette/win32_x86_generator.h b/courgette/win32_x86_generator.h index 8d3e683..3aabe9f 100644 --- a/third_party/courgette/win32_x86_generator.h +++ b/courgette/win32_x86_generator.h @@ -1,126 +1,126 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-// This is the transformation and adjustment for Windows X86 executables.
-
-#ifndef COURGETTE_WIN32_X86_GENERATOR_H_
-#define COURGETTE_WIN32_X86_GENERATOR_H_
-
-#include "base/scoped_ptr.h"
-
-#include "third_party/courgette/ensemble.h"
-
-namespace courgette {
-
-class CourgetteWin32X86PatchGenerator : public TransformationPatchGenerator {
- public:
- CourgetteWin32X86PatchGenerator(Element* old_element,
- Element* new_element,
- CourgetteWin32X86Patcher* patcher)
- : TransformationPatchGenerator(old_element, new_element, patcher) {
- }
-
- CourgettePatchFile::TransformationMethodId Kind() {
- return CourgettePatchFile::T_COURGETTE_WIN32_X86;
- }
-
- Status WriteInitialParameters(SinkStream* parameter_stream) {
- parameter_stream->WriteVarint32(old_element_->offset_in_ensemble());
- parameter_stream->WriteVarint32(old_element_->region().length());
- return C_OK;
- // TODO(sra): Initialize |patcher_| with these parameters.
- }
-
- Status PredictTransformParameters(SinkStreamSet* prediction) {
- return TransformationPatchGenerator::PredictTransformParameters(prediction);
- }
-
- Status CorrectedTransformParameters(SinkStreamSet* parameters) {
- // No code needed to write an 'empty' parameter set.
- return C_OK;
- }
-
- // The format of a transformed_element is a serialized EncodedProgram. We
- // first disassemble the original old and new Elements into AssemblyPrograms.
- // Then we adjust the new AssemblyProgram to make it as much like the old one
- // as possible, before converting the AssemblyPrograms to EncodedPrograms and
- // serializing them.
- Status Transform(SourceStreamSet* corrected_parameters,
- SinkStreamSet* old_transformed_element,
- SinkStreamSet* new_transformed_element) {
- // Don't expect any corrected parameters.
- if (!corrected_parameters->Empty())
- return C_GENERAL_ERROR;
-
- // Generate old version of program using |corrected_parameters|.
- // TODO(sra): refactor to use same code from patcher_.
- AssemblyProgram* old_program = NULL;
- Status old_parse_status =
- ParseWin32X86PE(old_element_->region().start(),
- old_element_->region().length(),
- &old_program);
- if (old_parse_status != C_OK)
- return old_parse_status;
-
- AssemblyProgram* new_program = NULL;
- Status new_parse_status =
- ParseWin32X86PE(new_element_->region().start(),
- new_element_->region().length(),
- &new_program);
- if (new_parse_status != C_OK) {
- DeleteAssemblyProgram(old_program);
- return new_parse_status;
- }
-
- EncodedProgram* old_encoded = NULL;
- Status old_encode_status = Encode(old_program, &old_encoded);
- if (old_encode_status != C_OK) {
- DeleteAssemblyProgram(old_program);
- return old_encode_status;
- }
-
- Status old_write_status =
- WriteEncodedProgram(old_encoded, old_transformed_element);
- DeleteEncodedProgram(old_encoded);
- if (old_write_status != C_OK) {
- DeleteAssemblyProgram(old_program);
- return old_write_status;
- }
-
- Status adjust_status = Adjust(*old_program, new_program);
- DeleteAssemblyProgram(old_program);
- if (adjust_status != C_OK) {
- DeleteAssemblyProgram(new_program);
- return adjust_status;
- }
-
- EncodedProgram* new_encoded = NULL;
- Status new_encode_status = Encode(new_program, &new_encoded);
- DeleteAssemblyProgram(new_program);
- if (new_encode_status != C_OK)
- return new_encode_status;
-
- Status new_write_status =
- WriteEncodedProgram(new_encoded, new_transformed_element);
- DeleteEncodedProgram(new_encoded);
- if (new_write_status != C_OK)
- return new_write_status;
-
- return C_OK;
- }
-
- Status Reform(SourceStreamSet* transformed_element,
- SinkStream* reformed_element) {
- return TransformationPatchGenerator::Reform(transformed_element,
- reformed_element);
- }
-
- private:
- ~CourgetteWin32X86PatchGenerator() { }
-
- DISALLOW_COPY_AND_ASSIGN(CourgetteWin32X86PatchGenerator);
-};
-
-} // namespace courgette
-#endif // COURGETTE_WIN32_X86_GENERATOR_H_
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +// This is the transformation and adjustment for Windows X86 executables. + +#ifndef COURGETTE_WIN32_X86_GENERATOR_H_ +#define COURGETTE_WIN32_X86_GENERATOR_H_ + +#include "base/scoped_ptr.h" + +#include "courgette/ensemble.h" + +namespace courgette { + +class CourgetteWin32X86PatchGenerator : public TransformationPatchGenerator { + public: + CourgetteWin32X86PatchGenerator(Element* old_element, + Element* new_element, + CourgetteWin32X86Patcher* patcher) + : TransformationPatchGenerator(old_element, new_element, patcher) { + } + + CourgettePatchFile::TransformationMethodId Kind() { + return CourgettePatchFile::T_COURGETTE_WIN32_X86; + } + + Status WriteInitialParameters(SinkStream* parameter_stream) { + parameter_stream->WriteVarint32(old_element_->offset_in_ensemble()); + parameter_stream->WriteVarint32(old_element_->region().length()); + return C_OK; + // TODO(sra): Initialize |patcher_| with these parameters. + } + + Status PredictTransformParameters(SinkStreamSet* prediction) { + return TransformationPatchGenerator::PredictTransformParameters(prediction); + } + + Status CorrectedTransformParameters(SinkStreamSet* parameters) { + // No code needed to write an 'empty' parameter set. + return C_OK; + } + + // The format of a transformed_element is a serialized EncodedProgram. We + // first disassemble the original old and new Elements into AssemblyPrograms. + // Then we adjust the new AssemblyProgram to make it as much like the old one + // as possible, before converting the AssemblyPrograms to EncodedPrograms and + // serializing them. + Status Transform(SourceStreamSet* corrected_parameters, + SinkStreamSet* old_transformed_element, + SinkStreamSet* new_transformed_element) { + // Don't expect any corrected parameters. + if (!corrected_parameters->Empty()) + return C_GENERAL_ERROR; + + // Generate old version of program using |corrected_parameters|. + // TODO(sra): refactor to use same code from patcher_. + AssemblyProgram* old_program = NULL; + Status old_parse_status = + ParseWin32X86PE(old_element_->region().start(), + old_element_->region().length(), + &old_program); + if (old_parse_status != C_OK) + return old_parse_status; + + AssemblyProgram* new_program = NULL; + Status new_parse_status = + ParseWin32X86PE(new_element_->region().start(), + new_element_->region().length(), + &new_program); + if (new_parse_status != C_OK) { + DeleteAssemblyProgram(old_program); + return new_parse_status; + } + + EncodedProgram* old_encoded = NULL; + Status old_encode_status = Encode(old_program, &old_encoded); + if (old_encode_status != C_OK) { + DeleteAssemblyProgram(old_program); + return old_encode_status; + } + + Status old_write_status = + WriteEncodedProgram(old_encoded, old_transformed_element); + DeleteEncodedProgram(old_encoded); + if (old_write_status != C_OK) { + DeleteAssemblyProgram(old_program); + return old_write_status; + } + + Status adjust_status = Adjust(*old_program, new_program); + DeleteAssemblyProgram(old_program); + if (adjust_status != C_OK) { + DeleteAssemblyProgram(new_program); + return adjust_status; + } + + EncodedProgram* new_encoded = NULL; + Status new_encode_status = Encode(new_program, &new_encoded); + DeleteAssemblyProgram(new_program); + if (new_encode_status != C_OK) + return new_encode_status; + + Status new_write_status = + WriteEncodedProgram(new_encoded, new_transformed_element); + DeleteEncodedProgram(new_encoded); + if (new_write_status != C_OK) + return new_write_status; + + return C_OK; + } + + Status Reform(SourceStreamSet* transformed_element, + SinkStream* reformed_element) { + return TransformationPatchGenerator::Reform(transformed_element, + reformed_element); + } + + private: + ~CourgetteWin32X86PatchGenerator() { } + + DISALLOW_COPY_AND_ASSIGN(CourgetteWin32X86PatchGenerator); +}; + +} // namespace courgette +#endif // COURGETTE_WIN32_X86_GENERATOR_H_ diff --git a/third_party/courgette/win32_x86_patcher.h b/courgette/win32_x86_patcher.h index 4febad4..9ab53d9 100644 --- a/third_party/courgette/win32_x86_patcher.h +++ b/courgette/win32_x86_patcher.h @@ -1,95 +1,95 @@ -// Copyright (c) 2009 The Chromium Authors. All rights reserved.
-// Use of this source code is governed by a BSD-style license that can be
-// found in the LICENSE file.
-
-// This is the transformation for Windows X86 executables.
-
-#ifndef COURGETTE_WIN32_X86_PATCHER_H_
-#define COURGETTE_WIN32_X86_PATCHER_H_
-
-#include "third_party/courgette/ensemble.h"
-
-namespace courgette {
-
-// CourgetteWin32X86Patcher is a TransformationPatcher for Windows 32-bit
-// executables.
-//
-class CourgetteWin32X86Patcher : public TransformationPatcher {
- public:
- explicit CourgetteWin32X86Patcher(const Region& region)
- : ensemble_region_(region) {
- }
-
- Status Init(SourceStream* parameter_stream) {
- if (!parameter_stream->ReadVarint32(&base_offset_))
- return C_BAD_TRANSFORM;
- if (!parameter_stream->ReadVarint32(&base_length_))
- return C_BAD_TRANSFORM;
-
- if (base_offset_ > ensemble_region_.length())
- return C_BAD_TRANSFORM;
- if (base_length_ > ensemble_region_.length() - base_offset_)
- return C_BAD_TRANSFORM;
-
- return C_OK;
- }
-
- Status PredictTransformParameters(SinkStreamSet* predicted_parameters) {
- // No code needed to write an 'empty' predicted parameter set.
- return C_OK;
- }
-
- Status Transform(SourceStreamSet* corrected_parameters,
- SinkStreamSet* transformed_element) {
- Status status;
- if (!corrected_parameters->Empty())
- return C_GENERAL_ERROR; // Don't expect any corrected parameters.
-
- AssemblyProgram* program = NULL;
- status = ParseWin32X86PE(ensemble_region_.start() + base_offset_,
- base_length_,
- &program);
- if (status != C_OK)
- return status;
-
- EncodedProgram* encoded = NULL;
- status = Encode(program, &encoded);
- DeleteAssemblyProgram(program);
- if (status != C_OK)
- return status;
-
- status = WriteEncodedProgram(encoded, transformed_element);
- DeleteEncodedProgram(encoded);
- if (status != C_OK)
- return status;
-
- return status;
- }
-
- Status Reform(SourceStreamSet* transformed_element,
- SinkStream* reformed_element) {
- Status status;
- EncodedProgram* encoded_program = NULL;
- status = ReadEncodedProgram(transformed_element, &encoded_program);
- if (status != C_OK)
- return status;
-
- status = Assemble(encoded_program, reformed_element);
- DeleteEncodedProgram(encoded_program);
- if (status != C_OK)
- return status;
-
- return C_OK;
- }
-
- private:
- Region ensemble_region_;
-
- uint32 base_offset_;
- uint32 base_length_;
-
- DISALLOW_COPY_AND_ASSIGN(CourgetteWin32X86Patcher);
-};
-
-} // namespace
-#endif // COURGETTE_WIN32_X86_PATCHER_H_
+// Copyright (c) 2009 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +// This is the transformation for Windows X86 executables. + +#ifndef COURGETTE_WIN32_X86_PATCHER_H_ +#define COURGETTE_WIN32_X86_PATCHER_H_ + +#include "courgette/ensemble.h" + +namespace courgette { + +// CourgetteWin32X86Patcher is a TransformationPatcher for Windows 32-bit +// executables. +// +class CourgetteWin32X86Patcher : public TransformationPatcher { + public: + explicit CourgetteWin32X86Patcher(const Region& region) + : ensemble_region_(region) { + } + + Status Init(SourceStream* parameter_stream) { + if (!parameter_stream->ReadVarint32(&base_offset_)) + return C_BAD_TRANSFORM; + if (!parameter_stream->ReadVarint32(&base_length_)) + return C_BAD_TRANSFORM; + + if (base_offset_ > ensemble_region_.length()) + return C_BAD_TRANSFORM; + if (base_length_ > ensemble_region_.length() - base_offset_) + return C_BAD_TRANSFORM; + + return C_OK; + } + + Status PredictTransformParameters(SinkStreamSet* predicted_parameters) { + // No code needed to write an 'empty' predicted parameter set. + return C_OK; + } + + Status Transform(SourceStreamSet* corrected_parameters, + SinkStreamSet* transformed_element) { + Status status; + if (!corrected_parameters->Empty()) + return C_GENERAL_ERROR; // Don't expect any corrected parameters. + + AssemblyProgram* program = NULL; + status = ParseWin32X86PE(ensemble_region_.start() + base_offset_, + base_length_, + &program); + if (status != C_OK) + return status; + + EncodedProgram* encoded = NULL; + status = Encode(program, &encoded); + DeleteAssemblyProgram(program); + if (status != C_OK) + return status; + + status = WriteEncodedProgram(encoded, transformed_element); + DeleteEncodedProgram(encoded); + if (status != C_OK) + return status; + + return status; + } + + Status Reform(SourceStreamSet* transformed_element, + SinkStream* reformed_element) { + Status status; + EncodedProgram* encoded_program = NULL; + status = ReadEncodedProgram(transformed_element, &encoded_program); + if (status != C_OK) + return status; + + status = Assemble(encoded_program, reformed_element); + DeleteEncodedProgram(encoded_program); + if (status != C_OK) + return status; + + return C_OK; + } + + private: + Region ensemble_region_; + + uint32 base_offset_; + uint32 base_length_; + + DISALLOW_COPY_AND_ASSIGN(CourgetteWin32X86Patcher); +}; + +} // namespace +#endif // COURGETTE_WIN32_X86_PATCHER_H_ |