Linux supports TLS on x86 by allocating slots in the GDT, accessible
via the system calls get/set_thread_area. This allows segment
registers (%gs on x86) to be used to quickly access the TLS.
Previously, we used PTRACE_GETREGSET with the NT_386_TLS regset. This
"register set" provides access to the subarray of the GDT used for TLS.
However, there are multiple slots provided and we don't know which one
is being used by the threading library for the current thread's TLS.
Previously, we were just using the first one, which worked for x86 on
64-bit kernels, but not 32-bit kernels. On 32-bit kernels, the first
slot ended up pointing to the TLS of the main thread.
The authoritative index of the current thread's TLS in the GDT is
given by bits 3-15 of %gs. However, this index cannot be used with
PTRACE_GETREGSET+NT386_TLS because we don't know the location of the
TLS slots in the GDT. PTRACE_GET_THREAD_AREA, however, accepts an
index from the start of the GDT similarly to get/set_thread_area.
Bug: crashpad:30
Change-Id: Ie6dfbdd088c6816fad409812a1a97037d4b38fd7
Reviewed-on: https://chromium-review.googlesource.com/575318
Reviewed-by: Mark Mentovai <mark@chromium.org>
Dynamic linkers use `struct r_debug` and `struct link_map` (defined in
`<link.h>`) to communicate lists of loaded modules to debuggers.
Bug: crashpad:30
Change-Id: Id903a1c199288dd85c34e38710cdb4c6b5fedb5b
Reviewed-on: https://chromium-review.googlesource.com/534853
Reviewed-by: Mark Mentovai <mark@chromium.org>
ELF executables and libraries may be loaded into memory in several
mappings, possibly with holes containing anonymous mappings
or mappings of other files. This method takes an input mapping and
attempts to find the mapping for file offset 0 of the same file.
Bug: crashpad:30
Change-Id: I79abf060b015d58ef0eba54a399a74315d7d2d77
Reviewed-on: https://chromium-review.googlesource.com/565223
Commit-Queue: Joshua Peraza <jperaza@chromium.org>
Reviewed-by: Mark Mentovai <mark@chromium.org>
ee67585e3115 linux: Switch between x86 and x86_64 with the target_arch
GYP variable
Bug: crashpad:30
Change-Id: Ia7860cda42daae698a179b65d22ef7897141de59
Reviewed-on: https://chromium-review.googlesource.com/553557
Reviewed-by: Robert Sesek <rsesek@chromium.org>
Commit-Queue: Mark Mentovai <mark@chromium.org>
These were intended to be enabled previously, but GYP uses “ia32” and
“x64” for x86 and x86_64, and zlib.gyp erroneously used “x86” and
“amd64” instead.
In order to make this work, gcc and clang need -mpclmul to enable the
pclmul extension used by crc_folding.c. The optimized code will only be
used if, at runtime, SSE2, SSE4.2, and PCLMULQDQ support is detected.
Change-Id: Ic709cd2a6c38892083c44c4004573a64b3581eb5
Reviewed-on: https://chromium-review.googlesource.com/553337
Reviewed-by: Robert Sesek <rsesek@chromium.org>
Commit-Queue: Mark Mentovai <mark@chromium.org>
Although GCC will silently accept larger alignments with
__attribute__((aligned())), it warn on alignas() with an alignment
larger than the target’s supported maximum. 8c35d92ae403 switched to
alignas() where possible.
The maxima are at least 128 on x86, x86_64, and arm64, and 64 on arm, in
the common configurations, but may be even larger with certain features
such as AVX enabled. These are ultimately derived from BIGGEST_ALIGNMENT
in gcc/config/*/*.h.
One alignment request in a test specified 1024 as a big alignment
constraint, solely as a test that alignment worked correctly. For this,
it’s perfectly reasonable to limit the alignment request to what GCC
supports on the most constrained target we’ll encounter.
Test: crashapd_util_test AlignedAllocator.AlignedVector
Change-Id: I42af443f437e01228934ab34dc04983742f0ab3f
Reviewed-on: https://chromium-review.googlesource.com/550236
Reviewed-by: Scott Graham <scottmg@chromium.org>
Commit-Queue: Mark Mentovai <mark@chromium.org>
user_fxsr_struct is only used in traditional NDK headers. Unified
headers always use user_fpxregs_struct regardless of API level.
Bug: crashpad:30, b/63025548
Change-Id: Id9d350801e659673b136e6fb8c0cbbbeb6055c4b
Reviewed-on: https://chromium-review.googlesource.com/549376
Reviewed-by: Joshua Peraza <jperaza@chromium.org>
107fb7631788 added the snapshot library as a dependency of
crashpad_util_test. Most of snapshot has not yet been ported to Linux or
Android. snapshot/capture_memory.cc only supports x86 and x86_64, and
will #error on other CPUs. We don’t build for other CPUs on Mac or
Windows, but we do for Android.
To make it easy to run crashpad_util_test on non-x86 again,
conditionally remove capture_memory.cc on Linux and Android.
crashpad_snapshot_test can be enabled for Linux and Android too by
disabling the CrashpadInfoClientOptions tests which require OS support.
There’s not much left in crashpad_snapshot_test currently for Linux
except for CPUContextX86 and ProcessSnapshotMinidump.EmptyFile, but both
pass.
Bug: crashpad:30
Change-Id: Ic19a79932072710c69a296bc0156cbe5656b8cb3
Reviewed-on: https://chromium-review.googlesource.com/549116
Reviewed-by: Joshua Peraza <jperaza@chromium.org>
This folow-up to d2d10d1dc8f3 is for compatibility with 32-bit Android
platforms using NDK API 16.
isinf() is also caught up in the switch.
Change-Id: I652e27061c01afa3dd932f494cc4eeaca4236f40
Reviewed-on: https://chromium-review.googlesource.com/544238
Reviewed-by: Joshua Peraza <jperaza@chromium.org>
TimedWait is implemented using `sem_timedwait` which waits until an
absolute time (time since the epoch) has passed. Previously, the
time to wait (relative to now) was passed without adding the current
time.
Change-Id: I3c169d5b107b8263577c21a8f47dc504058bd708
Reviewed-on: https://chromium-review.googlesource.com/540984
Commit-Queue: Joshua Peraza <jperaza@chromium.org>
Reviewed-by: Mark Mentovai <mark@chromium.org>
Crashpad doesn’t use AVX-512, but when receiving replies to exceptions
forwarded to ReportCrash, may see buffers large enough to contain
AVX-512 thread state. This can result in messages like
“UniversalExceptionRaise: (ipc/rcv) msg too large (0x10004004)”.
I386_THREAD_STATE_MAX has increased from 224 to 614 in the 10.13 SDK,
meaning that the maximum supported size for old_state and new_state in
[mach_]exception_raise_state[_identity]() has increased from 896 to
2,456 bytes. This constant defines the size of the buffer that these
MIG-generated routines will work with. By providing this definition in
compat, the buffer size is increased when building with older SDKs.
Note that on the “send” side, the size of the message given to
mach_msg() will be trimmed to include only the valid part of the state
area based on the stateCnt field, so increasing the value to 614 here
won’t result Crashpad sending messages this large. That would be a
potential interoperability concern with older OS versions.
Bug: crashpad:185, crashpad:190
Change-Id: Ia46091ae46fd6227a17f59eb4bc00914be471aa7
Reviewed-on: https://chromium-review.googlesource.com/541515
Reviewed-by: Robert Sesek <rsesek@chromium.org>
This renames and improves the VariableSizeBitCast helper from
util/linux/auxiliary_vector.* and moves it to misc.
Change-Id: I4bf46f4cfc0e60c900ff9bde467a21ad43c684cd
Reviewed-on: https://chromium-review.googlesource.com/534174
Commit-Queue: Joshua Peraza <jperaza@chromium.org>
Reviewed-by: Mark Mentovai <mark@chromium.org>
This was missed in Crashpad 8c35d92ae403. It syncs with Chromium
16289b3ef759.
Change-Id: I7e92e71fc940e25e751e7487d100b5684bdbf667
Reviewed-on: https://chromium-review.googlesource.com/535577
Commit-Queue: Mark Mentovai <mark@chromium.org>
Reviewed-by: Scott Graham <scottmg@chromium.org>
Reviewed-by: Mark Mentovai <mark@chromium.org>
In 10.13, modules loaded from the dyld shared cache appear with __TEXT
segments that have a nonzero “fileoff” (file offset). Previously, the
fileoff was always 0. Previously, the fileoff for segments in the dyld
shared cache was the actual offset into the shared cache (not 0), but
special consideration was given to __TEXT segments which were forced to
0. See 10.12.4 dyld-433.5/interlinked-dylibs/OptimizerLinkedit.cpp
LinkeditOptimizer<>::updateLoadCommands(). Note the comment there where
the __TEXT segment’s apparent fileoff is set to 0:
// HACK until lldb fixed in: <rdar://problem/20357466>
// DynamicLoaderMacOSXDYLD fixes for Monarch dyld shared cache
Refer also to the lldb commit that references the above,
http://llvm.org/viewvc/llvm-project?view=revision&revision=233714.
Evidently, update_dyld_shared_cache has been revised to no longer apply
this hack in 10.13. Crashpad’s sanity check for __TEXT segments having a
fileoff of 0 is no longer valid, and causes it to reject modules loaded
from the dyld shared cache.
Since this was just a sanity check, remove it entirely.
This caused module information for modules loaded from the dyld shared
cache to be missing from minidumps produced on 10.13, which in turn
prevented symbolization in frames belonging to most system libraries.
For reasons not yet understood, I don’t see this problem in Chrome on
10.13db1 17A264c on a test virtual machine (HFS+ filesystem), although I
do see it on actual hardware (APFS filesystem), and I do see it in
Crashpad’s tests and reduced testcases on both as well.
Bug: crashpad:185, crashpad:189
Test: crashpad_snapshot_test MachOImageReader.Self_DyldImages:ProcessReader.SelfModules:ProcessReader.ChildModules:ProcessTypes.DyldImagesSelf
Change-Id: I8b0a22c55c33ce920804a879f6fab67272f3556e
Reviewed-on: https://chromium-review.googlesource.com/535576
Reviewed-by: Robert Sesek <rsesek@chromium.org>
10.13 introduces two new fields to dyld_all_image_infos. Oddly, it
doesn’t put them in the “reserved” area that was defined in this
structure. This addition made it necessary for the padding problem in
the 32-bit structure previously worked around in Crashpad to be
addressed in the native structure, so Crashpad’s definition is adapted
to match.
This fixes tests on 10.13 that verify that dyld_all_image_infos can be
interpreted correctly.
Note that although the 10.13 SDK includes this structure extension,
numbered version 16, 10.13db1 17A264c continues to use version 15 as
used on 10.12, at least in crashpad_snapshot_test.
Bug: crashpad:185
Test: crashpad_snapshot_test ProcessTypes.DyldImagesSelf
Change-Id: I59a80c85bb234ef698c65a0ac5bbeac5b40fda77
Reviewed-on: https://chromium-review.googlesource.com/535394
Reviewed-by: Robert Sesek <rsesek@chromium.org>
_dyld_get_all_image_infos() was only used in test code in Crashpad.
This addresses two related problems.
When running on 10.13 or later, _dyld_get_all_image_infos() is not
available. It appears to still be implemented in dyld, but its symbol is
now private. This was always known to be an “internal” interface. When
it’s not available, fall back to obtaining the address of the process’
dyld_all_image_infos structure by calling task_info(…, TASK_DYLD_INFO,
…). Note that this is the same thing that the code being tested does,
although the tests are not rendered entirely pointless because the code
being tested consumes dyld_all_image_infos through its own
implementation of an out-of-process reader interface, while the
dyld_all_image_infos data obtained by _dyld_get_all_image_infos() is
handled strictly in-process by ordinary memory reads. This is covered by
bug 187.
When building with the 10.13 SDK, no _dyld_get_all_image_infos symbol is
available to link against. In this case, access the symbol strictly at
runtime via dlopen() if it may be available, or when expecting to only
run on 10.13 and later, don’t even bother looking for this symbol. This
is covered by part of bug 188.
Bug: crashpad:185, crashpad:187, crashpad:188
Change-Id: Ib283e070faf5d1ec35deee420213b53ec24fb1d3
Reviewed-on: https://chromium-review.googlesource.com/534633
Reviewed-by: Robert Sesek <rsesek@chromium.org>
Since Apple closed https://openradar.appspot.com/20239912 without fixing
anything, it looks like we’ll be stuck with these quriky cl_kernels
modules for quite some time. Allow these modules to be tolerated on any
OS version >= 10.10, where they first appeared in a broken state, by
removing the upper bound for the OS version to tolerate with this quirk.
The tolerance was previously expanded to include 10.11 in
cd1f8fa3d2f2c76802952beac71ad85f51bbf771 and 10.12 in
6fe7c5414e46acfa30e8984513bf0896e91b9407. After this third update, this
should hopefully no longer be an annual exercise.
Bug: crashpad:185, crashpad:186
Change-Id: I66d409f2d1638bcf7601b6622f000be245230f34
Reviewed-on: https://chromium-review.googlesource.com/534253
Reviewed-by: Robert Sesek <rsesek@chromium.org>
Commit-Queue: Mark Mentovai <mark@chromium.org>
In the 10.12 SDK, x86_state_hdr from <mach/i386/thread_status.h> was
defined as:
struct x86_state_hdr {
int flavor;
int count;
};
This has changed in the 10.13 SDK to:
struct x86_state_hdr {
uint32_t flavor;
uint32_t count;
};
This triggers signedness mismatch errors where these values are used
with CHECK/DCHECK macros and gtest EXPECT/ASSERT macros.
Compatibility with existing and new SDKs must be maintained, so more
casts must be used.
Bug: crashpad:185, crashpad:188
Change-Id: I8844d6a78520430a8b5b90a35403896c3c6cfa37
Reviewed-on: https://chromium-review.googlesource.com/533375
Reviewed-by: Robert Sesek <rsesek@chromium.org>
Commit-Queue: Mark Mentovai <mark@chromium.org>
Use the standard alignas instead of ALIGNAS in cases where this is
possible too. It’s not currently possible where ALIGNAS may be mixed
with other attributes, although the not-landed
https://codereview.chromium.org/2670873002/ suggests that where ALIGNAS
is mixed with __attribute__((packed)), it’s viable to write “struct
alignas(4) S { /* … */ } __attribute__((packed));”.
This includes an update of mini_chromium to
723e840a2f100a525f7feaad2e93df31d701780a, picking up:
723e840a2f10 Remove ALIGNOF
This tracks upstream https://codereview.chromium.org/2932053002/.
Change-Id: I7ddaf829020ef3be0512f803cecbb7c543294f07
Reviewed-on: https://chromium-review.googlesource.com/533356
Reviewed-by: Scott Graham <scottmg@chromium.org>
Commit-Queue: Mark Mentovai <mark@chromium.org>
ThreadInfo provides a uniform interface to collect register sets or
the thread-local storage address across bitness for x86 and ARM family
architectures. Additionally, ThreadInfo.h defines context structs which
mirror those provided in sys/user.h. This allows tracing across bitness
as the structs in sys/user.h are only provided for a single target
architecture.
Bug: crashpad:30
Change-Id: I91d0d788927bdac5fb630a6ad3c6ea6d3645ef8a
Reviewed-on: https://chromium-review.googlesource.com/494075
Commit-Queue: Joshua Peraza <jperaza@chromium.org>
Reviewed-by: Mark Mentovai <mark@chromium.org>
Drop the text recommending the PolyGerrit UI, since it is now the
default Gerrit UI.
Bug: chromium:717982
Change-Id: I7041ee51670a7a18b510ed7a55045cc2eb09983e
Reviewed-on: https://chromium-review.googlesource.com/494726
Reviewed-by: Scott Graham <scottmg@chromium.org>
The lots-of-regions tests in the MemoryMap test case were very
time-consuming, particularly in debug mode. MemoryMap.MapRunningChild
took as long as 15 seconds on-device (Nexus 5X), and the best result was
in the neighborhood of 7 seconds.
The bulk of the time spent in these tests was in ExpectMappings(), which
calls MemoryMap::FindMapping() in a loop to verify each region. Each
call to FindMapping() traverses the MemoryMap (internally, currently
just a std::vector<>) from the beginning. With the need to verify 4,096
regions, a single call to ExpectMappings() had to perform over 8,000,000
checks to find the regions it needed. In turn, ExpectMappings() is
called once by the SelfLargeMapFile test, and eight times by
MapRunningChild. By reducing the number of regions to 1,024, each call
to ExpectMappings() needs to perform “only” fewer than 600,000 checks.
After this change, MemoryMap.MapRunningChild completes in about a half a
second on-device.
https://crashpad.chromium.org/bug/181 is concerned with implementing a
RangeMap to serve MemoryMap and other similar code. After that’s done,
it, it should be feasible to raise the number of regions used for these
tests again.
Bug: crashpad:30, crashpad:181
Test: crashpad_util_test MemoryMap.SelfLargeMapFile:MemoryMap.MapRunningChild
Change-Id: I8ff88dac72a63c97ac937304b578fbe3b4ebf316
Reviewed-on: https://chromium-review.googlesource.com/494128
Reviewed-by: Joshua Peraza <jperaza@chromium.org>
When the /proc/pid/maps file is not read atomically and the target
process is actively mapping memory, entries can be read multiple times
or missed entirely. This change makes MemoryMap read the whole contents
of the maps file before attempting to parse it as well as check for
duplication/overlap errors, retrying on failure. This change also adds
ptrace attachements to unit tests to reflect actual intended usage.
Bug: crashpad:30
Change-Id: Ie8549548e25c47baa418ee7439d82743f84ff41e
Reviewed-on: https://chromium-review.googlesource.com/491950
Reviewed-by: Joshua Peraza <jperaza@chromium.org>
Reviewed-by: Mark Mentovai <mark@chromium.org>
Commit-Queue: Joshua Peraza <jperaza@chromium.org>
syscall(0) results in SIGSYS on x86_64, but not 32-bit x86. Choose a
high number as a nonexistent syscall number. As of 10.12.4, the highest
known system call number is 521.
Test: crashpad_util_test Signals.Cause*
Change-Id: I82dbd210f0c90fe933898ea0d360b431b10d090e
Reviewed-on: https://chromium-review.googlesource.com/489826
Reviewed-by: Robert Sesek <rsesek@chromium.org>
I opted to leave casts to types that were definitely the same size
alone. reinterpret_cast<uintptr_t>(pointer) and
reinterpret_cast<intptr_t>(pointer) should always be safe, for example.
Casts to other integral types have been replaced with
FromPointerCast<>(), which does zero-extension or sign-extension based
on the target type.
To make it possible to use FromPointerCast<>() with some use sites that
were already using checked_cast<>(), FromPointerCast<>() now uses
check_cast<>() when converting to a narrower type.
Test: crashpad_util_test FromPointerCast*, others
Change-Id: I4a71b4aa2d87f545c75524290a702f5f3138d675
Reviewed-on: https://chromium-review.googlesource.com/489701
Reviewed-by: Scott Graham <scottmg@chromium.org>
Some of the new Linux/Android tests were failing in 32-bit code where
pointers were being casted via reinterpret_cast<>() to LinuxVMAddress,
an unsigned 64-bit type. The behavior of such casts is
implementation-defined, and in this case, sign-extension was being used
to convert the 32-bit pointers to 64 bits, resulting in very large
(unsigned) LinuxVMAddress values that could not possibly refer to proper
addresses in a 32-bit process’ address space.
The offending reinterpret_cast<>() conversions have been replaced with
the new FromPointerCast<>(), which is careful to do sign-extension when
converting to a signed type, and zero-extension when converting to an
unsigned type like LinuxVMAddress.
Bug: crashpad:30
Test: crashpad_util_test FromPointerCast*:MemoryMap.*:ProcessMemory.*
Change-Id: I6f1408dc63369a8740ecd6015d657e4407a7c271
Reviewed-on: https://chromium-review.googlesource.com/488264
Reviewed-by: Joshua Peraza <jperaza@chromium.org>
These tests:
- InitializationState.InitializationState
- InitializationStateDcheckDeathTest.Destroyed_NotUninitialized
- InitializationStateDcheckDeathTest.Destroyed_NotValid
rely on certain behavior from destroyed objects. This is undefined
behavior and we know it, but the whole point of the of
InitializationState and InitializationStateDcheck destructors is to try
to help catch other parts of the program making use of undefined
behavior.
To make it impossible for the memory that formerly hosted these objects
to be repurposed during tests after the objects are destroyed, these
tests that attempt to work with destroyed objects are changed to use
placement new, so that the lifetimes of the objects can be decoupled
from the lifetimes of the buffers.
Test: crashpad_util_test InitializationState*
Change-Id: Ie972a54116c8b90a21a502d3ba13623583dfac06
Reviewed-on: https://chromium-review.googlesource.com/486383
Reviewed-by: Joshua Peraza <jperaza@chromium.org>
aae1e3efb507 CQ config: add gerrit CQAbility verifier.
95da7665b1a3 [win-test] loosen win-driver-target-type test
eb296f67da07 [win] Add support for MS VS2017 (via Registry)
19495aa28282 Update test/no-cpp/gyptest-no-cpp.
a94b02ec68fb Disable a bunch of tests on Mac
ae76d9198630 Clean up gyptest.py
b62d04ff85e6 win,ninja: ninja generator better on windows
8dc77241251e Disable flaky test/copies/gyptest-all under msvs
e8850240a433 Fix MSVC++ 32-on-32 builds after b62d04ff85e6
ffd524cefaad win ninja/make: Always use a native compiler executable
with MSVS 2017
developing.md is updated to call out supported toolchain versions, and
to explain the CDB requirement for end_to_end_tests.py.
Change-Id: Iace68009aa22acec7303ea02a2ded755645ea96c
Reviewed-on: https://chromium-review.googlesource.com/486539
Reviewed-by: Scott Graham <scottmg@chromium.org>
Includes mini_chromium ef0ded8717340c9fe48e8e0f34f3e0e74d10a392.
1d2a024fdb1d android: Use _FILE_OFFSET_BITS after all (undo
dc3d480305b2)
ef0ded871734 win: MSVS 2017 (15)/C++ 14.1/C 19.10 compatibility
Change-Id: I5c814669a0ef8577872bddff9112ce28ec628ba3
Reviewed-on: https://chromium-review.googlesource.com/482639
Commit-Queue: Mark Mentovai <mark@chromium.org>
Reviewed-by: Scott Graham <scottmg@chromium.org>
