Age | Commit message (Collapse) | Author | Files | Lines |
|
* Fix GCStress issue on x86 and Arm32
When we write one of the new gc stress instruction, such as INTERRUPT_INSTR_PROTECT_FIRST_RET
we might be writing it in the epilog region on x86 or ARM as a direct call can be the last
instruction before an epilog. (This isn't allowed on x64)
This fix expands the set of instructions we are allowed by IsMarkerInstr() toinclude these
newly added gc stress instructions.
added comment to GetGcMarkerExceptionCode
Fix DAC build
* Code review feedback
* Additional code review chnages
|
|
* Extract ReplaceInstrAfterCall.
* Avoid GCStress when return multireg with pointers.
Determinate when we need to protect the second register and do not cause GCStress in such cases.
* Add a repro test.
* Reenable MethodImplOptionsTests.
* Extract IsGcCoveregeInterruptInstruction.
That changes how we do checks for arm32 in `IsGcCoverageInterrupt`.
* Tolerate direct call to JIT_RareDisableHelper.
x86 ILStubClass:IL_STUB_PInvoke(byref,ref,int,byref):int generates it like:
Generating: N119 ( 4, 7) [000118] ------------ * RETURNTRAP int REG NA
IN0021: cmp dword ptr [0F9BF9F8H], 0
New Basic Block BB10 [0009] created.
IN0022: je L_M6496_BB10
Call: GCvars=00000001 {V01}, gcrefRegs=00000000 {}, byrefRegs=00000000 {}
IN0023: call CORINFO_HELP_STOP_FOR_GC
* Support GC stress protect return 1/2/both Unix x64.
* Fix arm64.
Do not insert GC Stress instrucitons when we can't determinate the exact return kind.
* Fix review1.
* Fix review2.
* Change the test as Andy suggested.
* Fix some typos.
* Replace all SLOT with PBYTE.
* Disable assert that can fail because of multithreading.
|
|
There were two problems:
* The illegal instruction 0xde01 used for INTERRUPT_INSTR_CALL doesn't
generate SIGILL, but SIGTRAP, since this is the code used for
breakpoints.
* The USE_REDIRECT_FOR_GCSTRESS was defined even for FEATURE_PAL for
ARM, which is incorrect and resulted in explicit redirect frame not
being created in DoGcStress and thus the GC stack walk was skipping
managed frames that it should walk.
|
|
This change enables the VM to support multiple versions GCInfo concurrently.
This is necessary in light of upcoming work to add ReturnType and other
modifications to the GCInfo format -- so that existing ReadyToRun images
will continue to run correctly.
The version# is not stored in the GcInfo structure -- because it is
wasteful to store the version once for every method. Instead, it is
tracked per range-section of generated/loaded methods.
The GCInfo version is computed as:
1) The current GCINFO_VERSION for JITted and Ngened images
2) A function of the Ready-to-run major version stored in READYTORUN_HEADER
for ready-to-run images. ReadyToRunJitManager::JitTokenToGCInfoVersion()
provides the GcInfo version for any Method. Currently, there's only one
version of GCInfo.
An abstraction GCInfoToken is added to the GcInfo interface, which tracks the
{GcInfo, Version} pair in-memory. Several GcInfo APIs are
modified to use GCInfoToken in place of GcInfo pointers.
Notes:
1) SOS GcDump: The GCDump API has separate dump routines for Header and the
pointer-liveness information (DumpGCTable and DumpGCHeader) each of which
advance a pointer to the GCInfo block. These APIs are not changed to
recieve a GCInfoToken in place of the GcInfo block pointer. Instead, they
recieve the GcInfo version at the time of construction.
2) Some routines that are specific to x86 gcInfo (ex: crackMethodInfoHdr)
are not yet updated to use versioning, since the development plan is to
update the Non-x86 GcInfo structure first.
3) The x86 specific structs defining GcInfo headers are moved to GcInfoTypes.h,
along with the non-x86 GcInfo type definitions.
|
|
|
|
[tfs-changeset: 1407945]
|