Windows mozilla::mscom::ProcessRuntime::InitInsideApartment Utility crash
Categories
(Core :: IPC: MSCOM, defect)
Tracking
()
People
(Reporter: gerard-majax, Unassigned)
References
Details
Crash Data
With the work in bug 1824080 we are starting to get some telemetry data showing audio decoding happening on the tab process when utility pref is not disabled: this is not something we like.
In https://sql.telemetry.mozilla.org/queries/91032/source we can link between client_id reporting audio decoding on another process than utility and not having disabled utility, and this yields a few records so far (telemetry is in place since less than a week).
The stacks that seems to come out of those are like:
$ fx-crash-sig --verbose < crashping.json
Symbolicated stack:
0 mozilla::mscom::ProcessRuntime::InitInsideApartment()
1 mozilla::mscom::ProcessRuntime::ProcessRuntime(const mozilla::mscom::ProcessRuntime::ProcessCategory)
2 XRE_InitChildProcess(int, char**, XREChildData const*)
3 wmain(int, wchar_t**)
4 __scrt_common_main_seh()
5 BaseThreadInitThunk
6 RtlUserThreadStart
signature: mozilla::mscom::ProcessRuntime::InitInsideApartment
notes: (0)
Searching for that signature, while a bit generic, yields results, e.g., https://crash-stats.mozilla.org/report/index/cd92ac64-01bf-42ae-89c3-7c95b0230308
I believe we might have something blocking proper loading of Utility that might explain at least in some cases why we end up doing audio decoding on non utility process
| Reporter | ||
Updated•3 years ago
|
Updated•3 years ago
|
| Reporter | ||
Comment 1•3 years ago
|
||
Some other are starting to manifest:
Symbolicated stack:
0 RaiseException
1 __delayLoadHelper2(ImgDelayDescr const*, long long (**)())
2 _tailMerge_oleaut32.dll
3 mozilla::mscom::ProcessRuntime::InitInsideApartment()
4 mozilla::mscom::ProcessRuntime::ProcessRuntime(const mozilla::mscom::ProcessRuntime::ProcessCategory)
5 XRE_InitChildProcess(int, char**, XREChildData const*)
6 wmain(int, wchar_t**)
7 __scrt_common_main_seh()
8 BaseThreadInitThunk
9 RtlUserThreadStart
signature: __delayLoadHelper2 | _tailMerge_oleaut32.dll | mozilla::mscom::ProcessRuntime::InitInsideApartment
notes: (0)
Updated•3 years ago
|
| Reporter | ||
Updated•3 years ago
|
| Reporter | ||
Updated•3 years ago
|
| Reporter | ||
Comment 2•3 years ago
|
||
I am starting to wonder how much this is related to bug 1827379: could our failure to load oleaut32 as reported here be related to the failure to preload mozavcodec.dll and mozavutil.dll ?
Comment 3•3 years ago
|
||
Since the crash volume is low (less than 15 per week), the severity is downgraded to S3. Feel free to change it back if you think the bug is still critical.
For more information, please visit BugBot documentation.
Comment 4•2 years ago
•
|
||
I guess the bot doesn't know about crash ping telemetry :-/
| Reporter | ||
Comment 5•2 years ago
|
||
We're getting some crash reports now and I see SbieDll.dll loaded which seems to be https://github.com/sandboxie-plus/Sandboxie/
| Reporter | ||
Comment 6•2 years ago
|
||
Ok, local procmon I see we delay load oleaut32.dll from SetOaNoCache() https://searchfox.org/mozilla-central/rev/962a843f6d96283c45162c788dc72bf217fca7df/ipc/mscom/ProcessRuntime.cpp#144,293
Comment 7•2 years ago
|
||
The bug is linked to a topcrash signature, which matches the following criteria:
- Top 20 desktop browser crashes on beta
- Top 10 content process crashes on beta
For more information, please visit BugBot documentation.
| Reporter | ||
Comment 8•2 years ago
|
||
We seem to have no more report on utility now that bug 1839463 has landed. Let's verify this in a few weeks.
| Reporter | ||
Updated•2 years ago
|
Comment 9•2 years ago
|
||
Based on the topcrash criteria, the crash signatures linked to this bug are not in the topcrash signatures anymore.
For more information, please visit BugBot documentation.
| Reporter | ||
Comment 10•2 years ago
|
||
(In reply to :gerard-majax [PTO 01/08/2023-10/09/2023] from comment #8)
We seem to have no more report on utility now that bug 1839463 has landed. Let's verify this in a few weeks.
It looks like we have no more report on beta for the past month.
Comment 11•2 years ago
|
||
During nightly crash triage I found two more signatures that could be relevant to this bug:
[@ OOM | unknown | __delayLoadHelper2 | _tailMerge_oleaut32.dll | mozilla::mscom::ProcessRuntime::InitInsideApartment][@ __delayLoadHelper2 | _tailMerge_oleaut32.dll | mozilla::mscom::ProcessRuntime::InitInsideApartment]
Only the first one is flagged as an OOM, but the second one could as well be one. Also several crashes under the main signature here are from 32-bit machines with very little address space left, so they could be OOMs too.
Comment 12•2 years ago
|
||
Sorry for removing the keyword earlier but there is a recent change in the ranking, so the bug is again linked to a topcrash signature, which matches the following criteria:
- Top 20 desktop browser crashes on beta
- Top 10 content process crashes on beta
For more information, please visit BugBot documentation.
| Reporter | ||
Comment 13•2 years ago
|
||
(In reply to Gabriele Svelto [:gsvelto] from comment #11)
During nightly crash triage I found two more signatures that could be relevant to this bug:
[@ OOM | unknown | __delayLoadHelper2 | _tailMerge_oleaut32.dll | mozilla::mscom::ProcessRuntime::InitInsideApartment][@ __delayLoadHelper2 | _tailMerge_oleaut32.dll | mozilla::mscom::ProcessRuntime::InitInsideApartment]Only the first one is flagged as an OOM, but the second one could as well be one. Also several crashes under the main signature here are from 32-bit machines with very little address space left, so they could be OOMs too.
Looking at those crashes, they only get reported on Parent and Content processes, not utility
Comment 14•2 years ago
|
||
Based on the topcrash criteria, the crash signatures linked to this bug are not in the topcrash signatures anymore.
For more information, please visit BugBot documentation.
Comment 15•2 years ago
|
||
Crash Ping Telemetry signals 14 clients on release, lowering severity.
| Reporter | ||
Comment 16•2 years ago
|
||
It seems we got rid of all of them for Utility: crash-stats reports only Content processes on this signatures.
We know crash-stats might not be ultra reliable for non content/parent processes, and looking at crash reports from telemetry it also looks to have either disappeared or be so low volume it's not impacting:
- Utility no actor: https://mozilla.github.io/process-top-crashes/utility_release.html / https://mozilla.github.io/process-top-crashes/utility_beta.html / https://mozilla.github.io/process-top-crashes/utility_nightly.html
- Utility WMF: https://mozilla.github.io/process-top-crashes/utility_release_audio-decoder-wmf.html / https://mozilla.github.io/process-top-crashes/utility_beta_audio-decoder-wmf.html / https://mozilla.github.io/process-top-crashes/utility_nightly_audio-decoder-wmf.html
- Utility Audio Generic: https://mozilla.github.io/process-top-crashes/utility_release_audio-decoder-generic.html / https://mozilla.github.io/process-top-crashes/utility_beta_audio-decoder-generic.html / https://mozilla.github.io/process-top-crashes/utility_nightly_audio-decoder-generic.html
Description
•