summaryrefslogtreecommitdiffstats
path: root/sandbox/linux/seccomp
Commit message (Collapse)AuthorAgeFilesLines
* Added missing copyright header.markus@chromium.org2010-03-0822-0/+88
| | | | | | | | BUG=32501 TEST=none Review URL: http://codereview.chromium.org/672011 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@40946 0039d316-1c4b-4281-b951-d872f2087c98
* - Add a custom allocator for STL objects. This fixes sandbox failures thatmarkus@chromium.org2010-03-0812-218/+641
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | were observed on some machines (in particular in 32bit mode). - Some more changes to avoid calling into glibc when we can make a direct system call, instead. These particular call sites were unlikely to cause any problems. But it makes the code easier to audit if we avoid all unnecessary calls into glibc. - In 64bit mode, gettimeofday() is handled by vsyscalls and tends to be cheap. In 32bit mode, it is just a regular system call. Some users rely on being able to call gettimeofday() at a very high rate (up to thousands of consecutive calls). Recognize this system call pattern and optimize for it. - Add debugging option that allows us to warn about expensive system calls. In many cases, these warnings can then be used to optimize the sandboxed application. - Fix compilation on newer versions of gcc. - Changed the x86-32 version of the code that we use when intercepting system calls. Previously, we would use CALL to jump to the set of instructions that we had relocated. But we made the mistake of allowing relocation of instructions that reference %esp. This doesn't work, as CALL modifies the stack. We now avoid using CALL and instead jump directly. On x86-32 that requires the use of a PUSH/RET combination as there is no 32bit wide JMP instruction. The x86-64 version of the code was already written in a way that would avoid this particular problem. (I would like to thank Craig Schlenter for his exceptional detective work in tracking down the root cause of this bug!) - For debugging purposes, injected a really small library (less than 4kB) and discovered that some of our memory map manipulations implicitly relied on mappings to be at least two pages long. Fixed the code that made this incorrect assumption. - For really small libraries, the runtime linker can choose a different more compact layout. Our computation of the ASR offset did not know how to deal with that. Fixed by explicitly looking for a ".text" segment instead of looking for a PT_DYNAMIC section. - Closed a file descriptor that we kept open longer than needed. - Removed some unused code. - Added copyright headers TEST=tested on i386 and x86-64 BUG=36133 Review URL: http://codereview.chromium.org/661438 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@40900 0039d316-1c4b-4281-b951-d872f2087c98
* Be more restrictive when finding file names for libraries that need patching.markus@chromium.org2010-02-241-2/+17
| | | | | | | | | | | | | | | | | | | | | This avoids false positives if the directory name matches one of the well-known library names (e.g. ld). False positives not only result in a performance hit at startup, because we are now trying to instrument libraries that don't actually contain any system calls; but even worse than this, we could try to instrument system calls in the sandboxing code itself. And those system calls are deliberately coded so that they will not get rewritten. Fortunately, none of this is a security problem. If we accidentally rewrite system calls that weren't supposed to be rewritten, we will just crash on startup. TEST=the sandbox now works on the buildbots BUG=36133 Review URL: http://codereview.chromium.org/652188 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@39839 0039d316-1c4b-4281-b951-d872f2087c98
* Explicitly ask for unsigned values when comparing addresses. Not only is thismarkus@chromium.org2010-02-241-2/+2
| | | | | | | | | | | code hard to understand (and possibly broken) otherwise, some versions of GCC complain about the comparison without the cast. TEST=none BUG=none Review URL: http://codereview.chromium.org/657034 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@39838 0039d316-1c4b-4281-b951-d872f2087c98
* Treat calls to lstat() and lstat64() the same as calls to stat(). In practise,markus@chromium.org2010-02-243-17/+72
| | | | | | | | | | | | | | | this means the calls will still be denied. But we now return a correct return code. But more importantly, this change brings the source code in line with the code of the stand-alone opensource sandbox. Wherever possible, we try to keep both code bases identical. TEST=none BUG=none Review URL: http://codereview.chromium.org/657040 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@39837 0039d316-1c4b-4281-b951-d872f2087c98
* seccomp: allow dup/dup2evan@chromium.org2010-02-031-0/+2
| | | | | | | | | | This is needed for opening the renderer<->plugin channel. TEST=flash works in seccomp mode Review URL: http://codereview.chromium.org/563024 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@38037 0039d316-1c4b-4281-b951-d872f2087c98
* linux: make the seccomp sandbox work againevan@chromium.org2010-01-081-12/+54
| | | | | | | | | | | | | | | | | We were hitting a stack overflow on renderer startup, because of the following: When we patch out syscalls, we need a scratch space near (within a 32-bit jump) of the original code. We pick the scratch space as the end of the nearest empty region available before the code we're patching. For the vdso region, the stack lies directly before it and so the region we'd grab was directly before the stack. This meant that as soon as the stack attempted to grow it'd fail because it ran into our patch region, and we'd hit a stack overflow. The fix is to specially note when we're near the stack region, and instead put our scratch space as far away from the stack as possible. Review URL: http://codereview.chromium.org/518071 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@35759 0039d316-1c4b-4281-b951-d872f2087c98
* Allow the seccomp sandbox to be enabled, even if the suid sandbox hasmarkus@chromium.org2009-11-0720-103/+141
| | | | | | | | | | already put a chroot() jail around it. The only tricky part is access to /proc/self/maps, but we can safely pass in an open file descriptor. BUG=26527 Review URL: http://codereview.chromium.org/371047 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@31372 0039d316-1c4b-4281-b951-d872f2087c98
* Only enable the seccomp sandbox, if the machine actually has kernel support formarkus@chromium.org2009-11-043-1/+66
| | | | | | | | | | | | | this feature, and if no other obstacle prevents us from enabling it. Otherwise, we print a warning message and continue running without the sandbox. This is not ideal, but given the non-trivial number of users who might not have seccomp enabled by default, this seems the prudent approach. BUG=26521 Review URL: http://codereview.chromium.org/341092 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@30966 0039d316-1c4b-4281-b951-d872f2087c98
* GCC's optimizer is getting more aggressive. It is no longer goodmarkus@chromium.org2009-10-223-7/+14
| | | | | | | | | | | | | | | | | | | | enough to just pass the address of a structure as an input parameter to assembly code. The assembly code must also mark "memory" as getting clobbered, even if it only wants to read from the structure. This seems to be a result of strict aliasing and the lack of an ability for the assembly code to clearly say which pointers it dereferences. Furthermore, if the assembly code touches the stack (e.g. uses "push"), it must now mark the stack pointer as getting clobbered. Otherwise, GCC assumes that the red zone won't be clobbered, and that it is possible to use the stack pointer as an input register. BUG=none TEST=none Review URL: http://codereview.chromium.org/320008 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@29829 0039d316-1c4b-4281-b951-d872f2087c98
* - found all symbols that we directly access from assembly and marked them as ↵markus@chromium.org2009-10-2113-486/+210
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | internal. This ensures that the linker won't complain about IP relative addressing for symbols that could be overridden at run-time. - avoided using "g" register constraints, as there has been a report of some versions of GCC erroneously generating code that is no longer position independant when this constraint is used. - removed the old code that fork()'s a child to try to extend mappings of libraries at run-time. This code always was somewhat fragile and caused a measurable performance penalty when the sandbox was started. Replaced with code that remapped just the very first page. This can actually be done in a running process without disrupting the use of the libraries. - added a special case for the instrumentation code allowing it to deal with jumps between the VDSO and VSyscalls even if the instructions would normally not be eligible for interception as they are IP relative. After making this change, we can again find sufficiently large code snippets to rewrite them successfully. This is only a concern on x86_64. - fixed a bug that would erroneously look for IP relative addressing on x86_32. It doesn't exist for that architecture. TEST=none BUG=http://code.google.com/p/chromium/issues/detail?id=18337 Review URL: http://codereview.chromium.org/306036 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@29726 0039d316-1c4b-4281-b951-d872f2087c98
* Fix minor compilation issues that can trigger on some platforms (e.g. CentOS)markus@chromium.org2009-09-142-2/+2
| | | | | | | | TEST=none BUG=none Review URL: http://codereview.chromium.org/204012 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@26175 0039d316-1c4b-4281-b951-d872f2087c98
* Despite the futex(2) manual page telling us to include <linux/futex.h>, thismarkus@chromium.org2009-09-142-1/+59
| | | | | | | | | | | | | is neither sufficient nor necessary. The header does not actually include a definition for futex(). And while it does include definitions for useful constants, the version of the file that is shipped by some distributions (e.g Centos) doesn't even compile as it is meant to only be used by the Linux kernel. TEST=none BUG=none Review URL: http://codereview.chromium.org/193104 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@26167 0039d316-1c4b-4281-b951-d872f2087c98
* On Linux, move the passing of filedescriptors to a dedicated socketpair().agl@chromium.org2009-09-041-1/+1
| | | | | | | | | | | | | | | | | (Patch by Markus) This allows the fast path to use read()/write() instead of recvmsg()/sendmsg() which is much cheaper for the Seccomp sandbox. Also, fixed minor seccomp sandbox issues discovered by this change. BUG=19120 ISSUE=164373 http://codereview.chromium.org/177049 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@25518 0039d316-1c4b-4281-b951-d872f2087c98
* Linux: fix seccomp sandbox for GCC 4.4agl@chromium.org2009-08-131-2/+4
| | | | | | | http://codereview.chromium.org/164484 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@23318 0039d316-1c4b-4281-b951-d872f2087c98
* Fix seccomp sandbox for gcc44markus@chromium.org2009-08-122-2/+2
| | | | | | | | | | | | Constness of return values and paramaters were causing compiler errors. BUG=19120 ISSUE=164373 Review URL: http://codereview.chromium.org/164414 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@23202 0039d316-1c4b-4281-b951-d872f2087c98
* Initial version of the Seccomp sandbox. Imported from ↵markus@chromium.org2009-08-1136-0/+11365
http://code.google.com/p/seccompsandbox/ Make the seccomp sandbox dependant on the --enable-seccomp-sandbox flag Review URL: http://codereview.chromium.org/165310 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@23087 0039d316-1c4b-4281-b951-d872f2087c98
generated by cgit v1.1 at 2025-01-22 16:51:12 +0000