Closed Bug 1778093 Opened 2 years ago Closed 2 years ago

mach run --debugger=rr no longer works on AMD CPUs

Categories

(Testing :: General, defect)

Default
defect

Tracking

(firefox-esr91 unaffected, firefox-esr102 unaffected, firefox102 unaffected, firefox103 unaffected, firefox104 fixed)

RESOLVED FIXED
104 Branch
Tracking Status
firefox-esr91 --- unaffected
firefox-esr102 --- unaffected
firefox102 --- unaffected
firefox103 --- unaffected
firefox104 --- fixed

People

(Reporter: emilio, Unassigned)

References

(Regression)

Details

(Keywords: regression)

Because we add --disable-avx-512 but we can't use CPUID faulting on those CPUs.

Pernosco works fine on those CPUs with NSS_DISABLE_HW_SHA=1.

$ ./mach run --debugger=rr https://bug1775142.bmoattachments.org/attachment.cgi\?id\=9283277 --verbose                                                                                                                                             
 0:00.44 /usr/local/bin/rr record --disable-avx-512 /home/emilio/src/moz/gecko/obj-debug/dist/bin/firefox https://bug1775142.bmoattachments.org/attachment.cgi?id=9283277 --verbose -no-remote -profile /home/emilio/src/moz/gecko/obj-debug/tmp/profile-default
rr: Saving execution to trace directory `/home/emilio/.local/share/rr/firefox-16'.
[FATAL src/RecordSession.cc:2379:RecordSession() errno: EOPNOTSUPP] CPUID faulting required to disable CPUID features
=== Start rr backtrace:
/usr/local/bin/rr(_ZN2rr13dump_rr_stackEv+0x41)[0x55d2d5e17701]
/usr/local/bin/rr(_ZN2rr15notifying_abortEv+0xe)[0x55d2d5e1775e]
/usr/local/bin/rr(_ZN2rr12FatalOstreamD1Ev+0x4f)[0x55d2d5cfe8af]
/usr/local/bin/rr(_ZN2rr13RecordSessionC2ERKNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEERKSt6vectorIS6_SaIS6_EESD_RKNS_20DisableCPUIDFeaturesENS0_16SyscallBufferingEiNS_7BindCPUES8_PKNS_9TraceUuidEbb+0x235)[0x55d2d5d341c5]
/usr/local/bin/rr(_ZN2rr13RecordSession6createERKSt6vectorINSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEESaIS7_EESB_RKNS_20DisableCPUIDFeaturesENS0_16SyscallBufferingEhNS_7BindCPUERKS7_PKNS_9TraceUuidEbbb+0xcf6)[0x55d2d5d35166]
/usr/local/bin/rr(_ZN2rr13RecordCommand3runERSt6vectorINSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEESaIS7_EE+0x78b)[0x55d2d5d2752b]
/usr/local/bin/rr(main+0x1af)[0x55d2d5c8b35f]
/usr/lib/libc.so.6(+0x29290)[0x7f8ce9e29290]
/usr/lib/libc.so.6(__libc_start_main+0x8a)[0x7f8ce9e2934a]
/usr/local/bin/rr(_start+0x25)[0x55d2d5c8b4b5]
=== End rr backtrace

Should we back out bug 1776771 for now?

Flags: needinfo?(rjesup)

Kyle?

Flags: needinfo?(rjesup) → needinfo?(khuey)

Set release status flags based on info from the regressing bug 1776771

There will be Intel CPUs you can't use this on either (e.g. in certain virtualized environments). You should check /proc/cpuinfo for the cpuid_fault feature before using --disable-avx512

Unrelated but NSS_DISABLE_HW_SHA=1 isn't needed for Pernosco anymore.

Flags: needinfo?(khuey)

Ok, we should back out. I'm on PTO

The regressor has been backed out from autoland and will be merged into central with the next merge. I will leave this bug open in case further investigating is necessary, otherwise it can be closed as fixed.

Flags: needinfo?(emilio)

Let's close for now. Thanks!

Status: NEW → RESOLVED
Closed: 2 years ago
Flags: needinfo?(emilio)
Resolution: --- → FIXED
Target Milestone: --- → 104 Branch
You need to log in before you can comment on or make changes to this bug.