1259192 - crash in mozalloc_abort | NS_DebugBreak | mozilla::ipc::FatalError | mozilla::net::PHttpChannelChild::FatalError | mozilla::net::PHttpChannelChild::OnMessageReceived

Status: RESOLVED WORKSFORME

(Core :: Networking, defect, P3)

Description

[Erin Lancaster [:elan]]

This bug was filed from the Socorro interface and is 
report bp-e56984e3-1431-4858-bf44-612e52160312.
=============================================================

Comment 1

[Erin Lancaster [:elan]]

:jduell, e10s top crash can you find an assignee?

Comment 2

[Josh Matthews [:jdm]]

I wonder why the message argument to NS_RUNTIMEABORT wasn't included in the crash report...

Comment 3

[u408661]

Putting this on the backlog for now, Jason, feel free to make active once we have an assignee

Comment 4

[(No longer employed by Mozilla) Aaron Klotz]

(In reply to Josh Matthews [:jdm] from comment #2)
> I wonder why the message argument to NS_RUNTIMEABORT wasn't included in the
> crash report...

That's bug 1236108 and its dependencies.

Comment 5

[Andrew McCreight [:mccr8]]

It looks like this shows up slightly differently on Nightly due to differences in inlining, I assume.

Comment 6

[Andrew McCreight [:mccr8]]

Assuming my line numbers match up, that's crashing here:
            if ((!(Read((&(cacheKey)), (&(msg__)), (&(iter__)))))) {
                FatalError("Error deserializing 'uint32_t'");

Comment 7

[Kan-Ru Chen [:kanru] (UTC+9)]

Signature will be changed to OOM by bug 1258312.
Close for now. If the signature shows up again we could reopen.

Comment 8

[Nicholas Nethercote [inactive]]

Nightly 20160429030215 had 12 occurrences of 
https://crash-stats.mozilla.com/report/index/9edbd88b-a899-467f-a91f-9e4522160430, but there were all from a single user.

Kan-Ru, I'm needinfo'ing you in case you want to take a look.

Comment 9

[Kan-Ru Chen [:kanru] (UTC+9)]

I took a look and found the error is "[Child 1036] ###!!! ABORT: IPDL error [PHttpChannelChild]: \"Error deserializing 'NetAddr'\". abort()ing as a result." on Msg_OnStartRequest

In NeckoMessageUtils.h there is one comment:

    /* If we get here without hitting any of the cases above, there's not much
     * we can do but let the deserializer fail when it gets this message */

https://dxr.mozilla.org/mozilla-central/rev/0a25833062a880f369e6f9f622413a94cc671bf4/netwerk/ipc/NeckoMessageUtils.h#101-102

I'm not sure if we are hitting this path but it seems better to crash in the Write side instead of the Read side.

Comment 10

[Andrew McCreight [:mccr8]]

I see 8 more crashes in the 5/02 build, so clearly this is an ongoing issue.

(In reply to Kan-Ru Chen [:kanru] (UTC+8) from comment #9)
> I'm not sure if we are hitting this path but it seems better to crash in the
> Write side instead of the Read side.

Definitely! Crashing in the Write() could give the information the Necko team needs to fix whatever is going wrong.

Comment 11

[Andrew McCreight [:mccr8]]

On Nightly, there were many crashes on 5-2, and many on 5-3 build, but none since then. However, the 5-2 crashes are all with a single install time, as are those on the 5-3 build, so it seems like this is all due to one or maybe two users.

Comment 12

[Jim Mathies [:jimm]]

On beta 47 this is currently #24 - 

https://crash-stats.mozilla.com/search/?product=Firefox&version=47.0b3&dom_ipc_enabled=!__null__&_facets=signature&_columns=date&_columns=signature&_columns=product&_columns=version&_columns=build_id&_columns=platform#facet-signature

Comment 13

[Andrew McCreight [:mccr8]]

(In reply to Jim Mathies [:jimm] from comment #12)
> On beta 47 this is currently #24 - 

Looking at the 63 crashes in there, I'd say the majority of the crashes have a single install time, suggesting a single unlucky persistent user.

Comment 14

[Firefox Bug Husbandry Bot]

Bulk change to priority: https://bugzilla.mozilla.org/show_bug.cgi?id=1399258

Comment 15

[Firefox Bug Husbandry Bot]

Bulk change to priority: https://bugzilla.mozilla.org/show_bug.cgi?id=1399258

Comment 16

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

Closing because no crashes reported for 12 weeks.