Crash in [@ mozilla::widget::WlLogHandler] with MOZ_CRASH Reason (Raw): "wl_display@1: error 1: invalid method 1 (since 1 < 3), object wl_pointer@44" when unplugging certain USB hardware
Categories
(Core :: Widget: Gtk, defect)
Tracking
()
People
(Reporter: dholbert, Unassigned)
Details
(Keywords: crash)
Crash report: https://crash-stats.mozilla.org/report/index/e5cd5441-06c9-460d-9545-e09970240911
MOZ_CRASH Reason:
wl_display@1: error 1: invalid method 1 (since 1 < 3), object wl_pointer@44
Top 10 frames:
0 libxul.so MOZ_Crash(char const*, int, char const*) mfbt/Assertions.h:317
0 libxul.so mozilla::widget::WlLogHandler(char const*, __va_list_tag*) widget/gtk/nsWaylandDisplay.cpp:398
1 libwayland-client.so.0 libwayland-client.so.0@0x598d
2 libwayland-client.so.0 libwayland-client.so.0@0x5f1d
3 libffi.so.8 ffi_call_unix64 /usr/src/libffi-3.4.6-1build1/src/x86/unix64.S:104
4 libffi.so.8 ffi_call_int /build/libffi-oEu4Vn/libffi-3.4.6/src/x86/ffi64.c:673
5 libffi.so.8 ffi_call /build/libffi-oEu4Vn/libffi-3.4.6/src/x86/ffi64.c:710
6 libwayland-client.so.0 libwayland-client.so.0@0x6bfd
7 libwayland-client.so.0 libwayland-client.so.0@0x7472
8 libwayland-client.so.0 libwayland-client.so.0@0x76ee
I hit this crash with multiple Firefox instances simultaneously (the same Nightly, running side-by-side in 3 separate profiles). This happened when I switched to a different computer using my KVM -- when I came back to this machine, I had 3 crash-reporter-dialogs to greet me, instead of 3 Firefox sessions.
My crashes:
bp-e5cd5441-06c9-460d-9545-e09970240911
bp-4950a830-0ef7-4212-b938-7692e0240911
bp-2827f0eb-4a80-421e-a287-4c77f0240911
This crash signature seems to be tracked in bug 1863047, but I don't know for sure if the issue I'm hitting is the same as the original crash there or not, since I've got a different "MOZ_CRASH Reason" (i.e. Wayland might be falling over in a different way). Hence, filing separately with see-also for now.
I'm running Ubuntu 24.04.1 (upgraded from 22.04 last week, using an in-place upgrade).
CPU is AMD Ryzen Threadripper PRO 3975WX 32-Cores
about:support says my GPU is AMD Radeon Pro WX 3200 Series
Reporter | ||
Comment 1•2 months ago
|
||
(In reply to Daniel Holbert [:dholbert] from comment #0)
This happened when I switched to a different computer using my KVM -- when I came back to this machine, I had 3 crash-reporter-dialogs to greet me, instead of 3 Firefox sessions.
UPDATE: further testing indicates that it's not actually the KVM that's causing me trouble -- it's when I unplug my Anker PowerExpand+ USB Hub (which my keyboard and mouse wireless-dongles are plugged into).
I don't crash when I unplug the dongles themselves from the USB hub. But if I unplug the USB hub itself from the computer, then all of my Firefox Nightly sessions instantly crash.
Reporter | ||
Comment 2•2 months ago
•
|
||
(the USB Hub has an HDMI port on it, which I don't have anything plugged into -- but presumably it still appears like a potential display connection to my machine. I imagine that port disappearing/reappearing is what's causing Wayland [and hence Firefox] trouble here.) EDIT: I suspect this is a red herring; more details in my next comment.
Reporter | ||
Comment 3•2 months ago
|
||
Further testing:
- My only reliable STR: so far I can only repro when I have both my keyboard and mouse wireless-dongles plugged in to the USB hub, and then unplug the USB hub. That causes me to crash 100% reliably. I've tried a bunch of variants, e.g. swapping my keyboard and/or mouse dongle to be directly connected to my machine (not on the USB hub) and then unplugging a dongle or the hub itself; this doesn't make me crash.
- Using my reliable STR, I tracked down this regression range with mozregression (not surprisingly, a Wayland-related change):
https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=791b8e439bbd8375d6b81419a65ca7059bbd00b5&tochange=6770fb6b8d9217b7d75ed7169cb1389c868498ed
--> Marking as a regression from bug 1915596.
Stransky, could you take a look? If you've got any ideas for diagnostics I might run, I'm happy to try what I can.
Reporter | ||
Comment 4•2 months ago
•
|
||
Looking at crash reports with this mozilla::widget::WlLogHandler
signature that have wl_display@1: error 1: invalid method 1
in the MOZ_CRASH Reason Raw" field:
https://crash-stats.mozilla.org/signature/?moz_crash_reason_raw=~wl_display%401%3A%20error%201%3A%20invalid%20method%201&signature=mozilla%3A%3Awidget%3A%3AWlLogHandler&date=%3E%3D2024-08-11T18%3A59%3A00.000Z&date=%3C2024-09-11T18%3A59%3A00.000Z&_columns=date&_columns=product&_columns=version&_columns=build_id&_columns=platform&_columns=reason&_columns=address&_columns=install_time&_columns=startup_crash&_sort=-date&page=1#reports
It looks like we have 73 such results in the past month, 72 of which are from Nightly 132.0a1 (plus one outlier from 126.0 which I'll ignore). The first Nightly crash was submitted on 2024-09-07
which I think matches the timeframe for when bug 1915596 (regressor) landed (it hit mozilla-central on 2024-09-06
).
Crash-stats attributes these 72 Nightly crashes to 43 installations.
Reporter | ||
Comment 5•2 months ago
|
||
So: 72 crashes of with this MOZ_CRASH Reason over 3-4 days (starting 9/7, up through and including today which really we're only halfway through in PDT) means we've got roughly 18-24 crashes per day on Nightly, which is substantial I think. It's not enough to show up as a spike in the crash volume for the general [ mozilla::widget::WlLogHandler ]
crash-signature (which has been 80-200 per day over the past months), but probably enough to result in a spike once this regression hits beta/release.
[Tracking Requested - why for this release]: new crash in Nightly, with reproducible STR on my hardware at least, with identified regressor.
Reporter | ||
Comment 6•2 months ago
|
||
I can try to capture a pernosco trace later on. (rr/pernosco has been giving me some trouble with submissions recently, but hopefully I can get a recording here.) ni=me as a reminder.
Comment 7•2 months ago
|
||
Should be dupe of Bug 1917558.
Reporter | ||
Comment 8•2 months ago
|
||
Aha, thanks! Looks like it. I'll try the patch there and dupe if that fixes it.
Reporter | ||
Comment 9•2 months ago
|
||
yup, this is a dupe. I can reproduce the crash in a local build of mozilla-central tip (169a59fe35f8) but can't repro anymore if I apply the patch from Bug 1917558.
Reporter | ||
Comment 10•2 months ago
|
||
[clearing some flags / bug-associations that I transferred to the dupe-target.]
Reporter | ||
Updated•2 months ago
|
Description
•