1659823 - Crash in [@ mozalloc_abort | moz_xmalloc | mozilla::TransportLayerDtls::GetDecryptedPackets]

Status: RESOLVED WORKSFORME

(Core :: WebRTC, defect, P2)

Description

[[:philipp]]

This bug is for crash report bp-dc6346fa-937b-4012-acd1-821000200818.

Top 10 frames of crashing thread:

0 mozglue.dll mozalloc_abort memory/mozalloc/mozalloc_abort.cpp:33
1 mozglue.dll mozalloc_handle_oom memory/mozalloc/mozalloc_oom.cpp:51
2 mozglue.dll moz_xmalloc memory/mozalloc/mozalloc.cpp:54
3 xul.dll mozilla::TransportLayerDtls::GetDecryptedPackets media/mtransport/transportlayerdtls.cpp:1041
4 xul.dll mozilla::TransportLayerDtls::PacketReceived media/mtransport/transportlayerdtls.cpp:1024
5 xul.dll static sigslot::_opaque_connection::emitter<mozilla::MediaPipeline, const std::basic_string<char, std::char_traits<char>, std::allocator<char> >&, const mozilla::MediaPacket&> media/mtransport/sigslot.h:339
6 xul.dll mozilla::TransportLayerIce::IcePacketReceived media/mtransport/transportlayerice.cpp:181
7 xul.dll static sigslot::_opaque_connection::emitter<mozilla::TransportLayerIce, mozilla::NrIceMediaStream*, int, const unsigned char*, int> media/mtransport/sigslot.h:339
8 xul.dll static mozilla::NrIceCtx::msg_recvd media/mtransport/nricectx.cpp:415
9 xul.dll nr_ice_peer_ctx_deliver_packet_maybe media/mtransport/third_party/nICEr/src/ice/ice_peer_ctx.c:840

these content crashes from 32bit users on windows started showing up since firefox 80.0b6.

this would be the small list of changes going into beta 6:

• https://hg.mozilla.org/releases/mozilla-beta/pushloghtml?fromchange=FIREFOX_80_0b5_RELEASE&tochange=6fa6624abc9cee1610dc4ececdd23115edcddd22&full=1
• https://bugzilla.mozilla.org/buglist.cgi?bug_id=1654842%2C1657155%2C1653020%2C1655266%2C1647950%2C1653958%2C1651601%2C1653626&list_id=15380110

Comment 1

[Michael Froman [:mjf]]

:bwc, it looks like this might be related to a couple things you worked on?

Comment 2

[Byron Campen [:bwc]]

So it looks like this bug existed before 80.0b6 (I see crashes in 78 and 79), but certainly increased in frequency after 80.0b6 was released. However, this rate increase happened at the same time on the release channel, right around Aug 10. This date does not correspond with any update on the release channel. I suspect that this change is due to usage patterns, perhaps related to the resumption of school (Aug 10 was a Monday).

Comment 3

[[:philipp]]

...and august 11 would have been windows patch tuesday - apparently win7 users would also still receive security updates (as the crash spike we see here is mostly win7): https://support.microsoft.com/en-us/help/4571719/windows-7-update

Comment 4

[Byron Campen [:bwc]]

Not seeing this happen at all on 81+, and beta 81 has been out there for about a week. Has this been fixed somehow?

Comment 5

[Byron Campen [:bwc]]

So here's the webrtc stuff fixed in 81 (what a strange URL...):

https://bugzilla.mozilla.org/buglist.cgi?target_milestone=81%20Branch&target_milestone=81%20Branch&target_milestone=81%20Branch&target_milestone=81%20Branch&target_milestone=81%20Branch&target_milestone=81%20Branch&target_milestone=81%20Branch&target_milestone=81%20Branch&target_milestone=81%20Branch&target_milestone=81%20Branch&target_milestone=81%20Branch&query_format=advanced&classification=Client%20Software&classification=Developer%20Infrastructure&classification=Components&classification=Server%20Software&classification=Other&component=WebRTC&component=WebRTC%3A%20Audio%2FVideo&component=WebRTC%3A%20Networking&component=WebRTC%3A%20Signaling&list_id=15396831&resolution=FIXED

Comment 6

[Byron Campen [:bwc]]

Here's a better list:

https://bugzilla.mozilla.org/buglist.cgi?resolution=FIXED&f1=cf_status_firefox81&v1=fixed&o2=notequals&classification=Client%20Software&classification=Developer%20Infrastructure&classification=Components&classification=Server%20Software&classification=Other&query_format=advanced&o1=equals&v2=fixed&f2=cf_status_firefox80&component=WebRTC&component=WebRTC%3A%20Audio%2FVideo&component=WebRTC%3A%20Networking&component=WebRTC%3A%20Signaling&list_id=15396845

Comment 7

[Byron Campen [:bwc]]

Maybe bug 1303867 changed some timing, and that is why this is not happening on 81?

Comment 8

[Byron Campen [:bwc]]

I'm just going to NI myself to check back on this in a week. If 81+ looks fine still, we can close.

Comment 9

[[:philipp]]

unfortunately the signature is reappearing now in 81.0b.

Comment 10

[Ryan VanderMeulen [:RyanVM]]

Not spiking until 81.0b7 (and 80.0b6 last cycle) would imply that it's somehow related to the end of EARLY_BETA_OR_EARLIER. Do we have WebRTC features gated behind that flag?

Comment 11

[Byron Campen [:bwc]]

(In reply to Ryan VanderMeulen [:RyanVM] from comment #10)

Not spiking until 81.0b7 (and 80.0b6 last cycle) would imply that it's somehow related to the end of EARLY_BETA_OR_EARLIER. Do we have WebRTC features gated behind that flag?

Socket process is only used in EARLY_BETA_OR_EARLIER; maybe we have some race when not using socket process? An alternative hypothesis is that this is related to bug 1502992, another thing that looks like it is happening only on beta (but is not related to socket process).

Comment 12

[Pascal Chevrel:pascalc]

Only one crash during the 83 beta cycle, no crash so far on 84, maybe the signature changed or something fixed it.

Comment 13

[BugBot [:suhaib / :marco/ :calixte]]

Closing because no crashes reported for 12 weeks.