summaryrefslogtreecommitdiffstats
path: root/chrome/browser/zygote_main_linux.cc
Commit message (Collapse)AuthorAgeFilesLines
* Linux: don't use GOT patching to intercept localtime(_r)agl@chromium.org2009-08-061-57/+70
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Our current GOT patching code is platform specific and fails to work when V8 is built as a library. Instead we define global functions for those functions which we wish to override. Since we will be first in the dynamic resolution order, the dynamic linker will point callers to our versions of these functions. However, we have the same binary for both the browser and the renderers, which means that our overrides will apply in the browser too. The global |g_am_zygote_or_renderer| is true iff we are in a zygote or renderer process. It's set in ZygoteMain and inherited by the renderers when they fork. (This means that it'll be incorrect for global constructor functions and before ZygoteMain is called - beware). Our replacement functions can check this global and either proxy the call to the browser over the sandbox IPC (http://code.google.com/p/chromium/wiki/LinuxSandboxIPC) or they can use dlsym with RTLD_NEXT to resolve the symbol, ignoring any symbols in the current module. TEST=Run javascript:alert(new Date().getTimezoneOffset()). It shouldn't return 0 unless you're actually in GMT. BUG=16800 http://codereview.chromium.org/165011 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@22636 0039d316-1c4b-4281-b951-d872f2087c98
* Linux: preload ffmpeg in the zygote.agl@chromium.org2009-08-051-0/+7
| | | | | | | | | | | | | | | | | | When using the sandbox, the renderer cannot load the ffmpeg shared objects from disk. So we load them in the zygote before the sandbox is started. (Note that this doesn't get media working, but it's a start.) BUG=18329 http://codereview.chromium.org/159841 (Note: relanding. Original landing in r22392, reverted in r22405 due to check-deps failure.) git-svn-id: svn://svn.chromium.org/chrome/trunk/src@22492 0039d316-1c4b-4281-b951-d872f2087c98
* Revert "Linux: preload ffmpeg in the zygote."agl@chromium.org2009-08-041-7/+0
| | | | | | | This reverts commit r22392. git-svn-id: svn://svn.chromium.org/chrome/trunk/src@22405 0039d316-1c4b-4281-b951-d872f2087c98
* Linux: don't try and read shmmax in the sandbox.agl@chromium.org2009-08-041-0/+3
| | | | | | | | | | | In the sandbox, reading shmmax from proc always fails. Thus, we move the function into base and call it before starting the sandbox. The value is then cached. http://codereview.chromium.org/159843 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@22393 0039d316-1c4b-4281-b951-d872f2087c98
* Linux: preload ffmpeg in the zygote.agl@chromium.org2009-08-041-0/+7
| | | | | | | | | | | | | | | When using the sandbox, the renderer cannot load the ffmpeg shared objects from disk. So we load them in the zygote before the sandbox is started. (Note that this doesn't get media working, but it's a start.) BUG=18329 http://codereview.chromium.org/159841 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@22392 0039d316-1c4b-4281-b951-d872f2087c98
* Assorted fixes for 64-bit.deanm@chromium.org2009-07-281-1/+1
| | | | | | | Review URL: http://codereview.chromium.org/160288 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@21891 0039d316-1c4b-4281-b951-d872f2087c98
* Move setlocale() to zygote_main_linux to make sure that it's calledjshin@chromium.org2009-07-241-0/+8
| | | | | | | | | | | | | | before the sandbox is triggered. BUG=16860 (http://crbug.com/16860 ) TEST=Run Chrome with LC_ALL=en_US.UTF-8. 1. Make sure that you don't see any warning about 'setlocale failing'. 2. Upload a file whose name contains non-ASCII characters in UTF-8 (e.g. "café.txt" or "一二.txt" ) . The filename field in a form should be populated with the name (rather than empty). Review URL: http://codereview.chromium.org/160112 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@21586 0039d316-1c4b-4281-b951-d872f2087c98
* Hide ia32 asm when not building for ARCH_CPU_X86_FAMILYdeanm@chromium.org2009-07-241-0/+3
| | | | | | | | | | | | r21097, 'Linux sandbox: plumb timezone calls through the sandbox' adds some ia32 assembly by the way of a define. This is bad for building for ARM. Patch by Joel Stanley. Review URL: http://codereview.chromium.org/160097 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@21527 0039d316-1c4b-4281-b951-d872f2087c98
* Linux: add support for tm_zone to localtime via the sandbox.agl@chromium.org2009-07-221-4/+15
| | | | | | | | | | | | | | | | Glibc adds a field to struct tm, tm_zone, which contains a pointer to a string for the current time zone. Since this makes struct tm a non-POD data type, it only works for the non-thread-safe localtime() call, not localtime_r(). However, V8 relies on it so this patch adds support for passing it through our sandbox. BUG=17458 http://codereview.chromium.org/159216 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@21311 0039d316-1c4b-4281-b951-d872f2087c98
* linux: add command-line flag to allow debugging of sandboxed processespiman@google.com2009-07-221-4/+12
| | | | | | | Review URL: http://codereview.chromium.org/159124 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@21248 0039d316-1c4b-4281-b951-d872f2087c98
* Linux sandbox: plumb timezone calls through the sandboxagl@chromium.org2009-07-201-1/+87
| | | | | | | | | | | | | | | | | | | | | The localtime (and localtime_r) functions try to access /etc/localtime in the filesystem. For sandboxed renderers, this fails, the the functions default to a UTC timezone. These functions are called from within WebCore and V8 and there's no clean way to patch the source code in place to do a hairpin turn and manage an IPC. Additionally, we cannot overwrite the calls with the usual symbol resolution procedures since the same chrome binary must serve as both the unsandboxed browser and sandboxed renderer. Thus we patch the GOT at runtime if we find, in the zygote process, that we are sandboxed. BUG=16800 TEST=Run javascript:alert(new Date().getTimezoneOffset()) in a sandboxed renderer. The resulting timezone offset (in minutes) should be correct for the local timezone. git-svn-id: svn://svn.chromium.org/chrome/trunk/src@21097 0039d316-1c4b-4281-b951-d872f2087c98
* Linux: reap the sandbox helper process.agl@chromium.org2009-07-161-0/+4
| | | | | | | | | | | TEST=start chrome with a SUID sandbox and check that no zombies persist. http://codereview.chromium.org/149734 (Reland of r20838. Hopefully with the reaping in the correct place this time.) git-svn-id: svn://svn.chromium.org/chrome/trunk/src@20845 0039d316-1c4b-4281-b951-d872f2087c98
* Linux: don't bother passing the chroot directory fd to the zygote.agl@chromium.org2009-07-101-16/+2
| | | | | | | | | | | | | Markus pointed out that the cwd was already shared between the chroot helper process and the zygote, therefore we could avoid some complexity in passing the file descriptor so, also, we could then make the directory mode 0000. http://codereview.chromium.org/155366 BUG=16363 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@20398 0039d316-1c4b-4281-b951-d872f2087c98
* Linux: use a temp directory for the chroot.agl@chromium.org2009-07-101-5/+26
| | | | | | | | | | | | | | | Ubuntu systems (at least) wipe /var/run at boot time, which is deleting our sandbox directory. Instead, we have the SUID helper create a temp directory in /tmp, unlink it and use that for the chroot directory. A file descriptor is passed to the zygote process for it to fchdir into. (Thanks to fta for discussions on this.) BUG=16363 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@20388 0039d316-1c4b-4281-b951-d872f2087c98
* Linux: all the binary to be readable with the SUID sandbox.agl@chromium.org2009-07-101-9/+13
| | | | | | | | | | | | | | | | | | | Previously, we required that the binary be non-readable. This causes the kernel to mark the process as non-dumpable at startup. The thinking was that, although we were putting the renderers into a PID namespace (with the SUID sandbox), they would none the less be in the /same/ PID namespace. So they could ptrace each other unless they were non-dumpable. If the binary was readable, then there would be a window between process startup and the point where we set the non-dumpable flag in which a compromised renderer could ptrace attach. However, now that we have a zygote model, only the (trusted) zygote exists at this point and we can set the non-dumpable flag which is inherited by all our renderer children. git-svn-id: svn://svn.chromium.org/chrome/trunk/src@20383 0039d316-1c4b-4281-b951-d872f2087c98
* Linux: SUID sandbox supportagl@chromium.org2009-07-081-1/+15
| | | | | | | | | | | | | | | | | | | | | * Make processes dumpable when they crash. * Find crashing processes by searching for a socket inode, rather than relying on SCM_CREDENTIALS. The kernel doesn't translate PIDs between PID namespaces with SCM_CREDENTIALS, so we can't use the PID there. * Use a command line flag to the renderer to enable crash dumping. Previously it tried to access the user's home directory for this information. * Search for a sandbox helper binary and, if found, use it. * Include the source for a sandbox helper binary. It's currently not built by default. http://codereview.chromium.org/149230 R=evan,markus BUG=8081 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@20110 0039d316-1c4b-4281-b951-d872f2087c98
* Linux: fix font fallback for charactors missing from the primary font.agl@chromium.org2009-06-261-1/+1
| | | | | | | | | | | | | | | We are using the sandbox host for font fallback now via the WebKit API. However, a patch landed in WebKit which added an assertion about being on the correct thread. Since the sandbox host was using WebString::fromUTF8, we hit the assertion and died. So, we have to implement a WebKitClient for the sandbox host. We cannot just pass in NULL because there's an assertion that it's non-NULL. Also, we cannot pass in a bogus pointer because it actually ends up getting called back. git-svn-id: svn://svn.chromium.org/chrome/trunk/src@19324 0039d316-1c4b-4281-b951-d872f2087c98
* Linux: Enable metrics_service_uitest.cc. Take 2.willchan@chromium.org2009-06-181-10/+35
| | | | | | | | | | Relands r18641, original code review: http://codereview.chromium.org/125268 Expect a crash on Windows. The old method of crashing was flawed on posix (KillProcess just does a SIGTERM). On Windows though, it would terminate the process with the desired exit code, in order to make DidProcessCrash() return true. This process termination does not dump crash information though, since it just forcibly terminates the process, like a SIGKILL on posix. When I switched it to navigate to about:crash though, it actually crashes, and dumps crash information, which the UITest in windows (but not linux/mac) detects. Therefore, until those platforms can detect, we just use #if defined(OS_WIN) around the expected_crashes = 1. Review URL: http://codereview.chromium.org/131007 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@18680 0039d316-1c4b-4281-b951-d872f2087c98
* Revert r18641: "Linux: Enable metrics_service_uitest.cc."willchan@chromium.org2009-06-171-35/+10
| | | | | | | | Broke ui_tests. Review URL: http://codereview.chromium.org/125268 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@18646 0039d316-1c4b-4281-b951-d872f2087c98
* Linux: Enable metrics_service_uitest.cc.willchan@chromium.org2009-06-171-10/+35
| | | | | | | | | Add support for zygotes to call DidProcessCrash on renderers. TEST=metrics_server_uitest.cc Review URL: http://codereview.chromium.org/126264 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@18641 0039d316-1c4b-4281-b951-d872f2087c98
* Linux: Add support for chrooted renderers.agl@chromium.org2009-06-151-2/+55
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | http://code.google.com/p/chromium/wiki/LinuxSandboxIPC Without filesystem access from the renderers, we need another way of dealing with fontconfig and font loading. This add support for: * An "SBX_D" environment variable in the renderers which is used to signal the end of dynamic linking so that the chroot can be enforced. * A sandbox_host process, running outside the sandbox, to deal with fontconfig requests from the renderers. See the wiki page for the reasoning behind making it a separate process. * A new, custom SkFontHost for Skia. Because this is Chrome specific, it will live outside the upstream Skia tree. This FontHost can be configured either to drive fontconfig directly (for the browser process and for any unsandboxed renderers) or to use an IPC system. Since the same SkFontHost has to be linked into both the browser and renderer (they are the same binary), this switch has to be made at run time. Sandbox IPC calls are rare (a couple of dozen at page load time) and add about 50us of overhead for each call. (Reland of r17575 which was reverted in r17577) http://codereview.chromium.org/112074 BUG=8081 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@18405 0039d316-1c4b-4281-b951-d872f2087c98
* Linux: refactor zygote supportagl@chromium.org2009-06-121-0/+168
| | | | | | | | | | | | | | | | | | | http://code.google.com/p/chromium/wiki/LinuxZygote * Move Chrome specific bits out of base * Move away from the idea of reserved file descriptors (which don't really work with zygotes) * Load resources before forking renderers (means that we don't need communication between the zygote process and the renderers) * Make sure that gdb works against the browser again * Make sure that we have different ASLR between the renderers and the browser. http://codereview.chromium.org/119335 (This is a reland. First landed in r18109, reverted in r18112.) git-svn-id: svn://svn.chromium.org/chrome/trunk/src@18291 0039d316-1c4b-4281-b951-d872f2087c98
* Revert 18109, 18111: Windows UI tests failed.agl@chromium.org2009-06-101-168/+0
| | | | git-svn-id: svn://svn.chromium.org/chrome/trunk/src@18112 0039d316-1c4b-4281-b951-d872f2087c98
* Linux: refactor zygote supportagl@chromium.org2009-06-101-0/+168
http://code.google.com/p/chromium/wiki/LinuxZygote * Move Chrome specific bits out of base * Move away from the idea of reserved file descriptors (which don't really work with zygotes) * Load resources before forking renderers (means that we don't need communication between the zygote process and the renderers) * Make sure that gdb works against the browser again * Make sure that we have different ASLR between the renderers and the browser. http://codereview.chromium.org/119335 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@18109 0039d316-1c4b-4281-b951-d872f2087c98
generated by cgit v1.1 at 2025-01-20 22:38:37 +0000