Closed Bug 1362984 Opened 9 years ago Closed 9 years ago

Crash in mozilla::net::nsHttpConnectionMgr::OnMsgReclaimConnection

Categories

(Core :: Networking: HTTP, defect)

55 Branch
defect
Not set
normal

Tracking

()

RESOLVED FIXED
mozilla57
Tracking Status
firefox-esr45 --- unaffected
firefox-esr52 --- unaffected
firefox53 --- unaffected
firefox54 --- unaffected
firefox55 --- disabled
firefox56 --- disabled
firefox57 --- fixed

People

(Reporter: dragana, Assigned: dragana)

References

Details

(Keywords: crash, Whiteboard: [necko-active])

Crash Data

No description provided.
Crash Signature: [@ mozilla::net::nsHttpConnectionMgr::OnMsgReclaimConnection]
Crash Signature: [@ mozilla::net::nsHttpConnectionMgr::OnMsgReclaimConnection] → [@ mozilla::net::nsHttpConnectionMgr::OnMsgReclaimConnection] [@ PLDHashTable::Search | nsClassHashtable<T>::Get | mozilla::net::nsHttpConnectionMgr::OnMsgReclaimConnection]
Crash Signature: [@ mozilla::net::nsHttpConnectionMgr::OnMsgReclaimConnection] [@ PLDHashTable::Search | nsClassHashtable<T>::Get | mozilla::net::nsHttpConnectionMgr::OnMsgReclaimConnection] → [@ mozilla::net::nsHttpConnectionMgr::OnMsgReclaimConnection] [@ PLDHashTable::Search | nsClassHashtable<T>::Get | mozilla::net::nsHttpConnectionMgr::OnMsgReclaimConnection] [@ PLDHashTable::Search | nsClassHashtable<T>::Get | mozilla::net::nsHttpConnec…
These are strange crashes. I do not know what is gong wrong. I notice that all have: LavasoftTcpService64.dll
PLDHashTable::Search | nsClassHashtable<T>::Get | mozilla::net::nsHttpConnectionMgr::OnMsgReschedTransaction
Crash Signature: mozilla::net::nsHttpConnectionMgr::GetOrCreateConnectionEntry] [@ PLDHashTable::Search | nsClassHashtable<T>::Get | mozilla::net::nsHttpConnectionMgr::OnMsgProcessPendingQ] → mozilla::net::nsHttpConnectionMgr::GetOrCreateConnectionEntry] [@ PLDHashTable::Search | nsClassHashtable<T>::Get | mozilla::net::nsHttpConnectionMgr::OnMsgProcessPendingQ] [@ PLDHashTable::Search | nsClassHashtable<T>::Get | mozilla::net::nsHttpConnec…
Patrick, do you have a clue how this crashes can happen? there is always a reference to connection and nsHttpConnectionInfo as well... Me staring at the code does not help at the moment :)
Flags: needinfo?(mcmanus)
https://crash-stats.mozilla.com/report/index/9fd61745-cdc0-4b16-8860-0498f0170508 its the first thing that function does.. is it possible we're partially shutdown and related to a looong tfo caused timeout?
I'm not buying my own shutdown suggestion :) lavasoft appears to be some kind of antivirus dll with a free version. I guess installing it (on a vm) makes sense.. my desktop and its vm is out of commission until the end of the week (hardware failure).
Flags: needinfo?(mcmanus)
(In reply to Patrick McManus [:mcmanus] from comment #6) > I'm not buying my own shutdown suggestion :) > > lavasoft appears to be some kind of antivirus dll with a free version. I > guess installing it (on a vm) makes sense.. my desktop and its vm is out of > commission until the end of the week (hardware failure). I have install it, and everything runs normal. No crashes. The important part the "Web and network protection" in this software is not free, so probably I will not be able to reproduce it.
I have upgrade the antivirus program, but still nothings happening :(
This crashes happen after call to: nsHttpConectionMgr::ReclaimConnection This is only called in the destructor of ConnectionHandle a nsAHttpTranaction hold reference to ConnectionHandle and reference is remove then a transaction gets canceled. The TFO patches created some additional connections that sometime get canceled. The patches relay on nsHttpConectionMgr::ReclaimConnection to remove this connections(the connections cannot be reused so they will be removed). I expect that this additional ReclaimConnection cause this crashes. I cannot prove it because ReclaimConnection dispatches a OnMsgReclaimConnection which crashes, so stack does not show from where was this dispatched. I could make a work around to avoid using ReclaimConnection. I need to hold reference to ConnectionHandle and call Reset before destroying it.
Whiteboard: [necko-active[
Whiteboard: [necko-active[ → [necko-active]
Now that the pref is enabled again, have seen a few crashes in [@ mozilla::net::nsHttpConnectionMgr::OnMsgReclaimConnection] using 20170531030204.
Most of this crashes have conn pointer being null. I think this is fixed by bug 1369509.(need to run tests to confirm it) there an interesting crash: https://crash-stats.mozilla.com/report/index/47e0a363-8366-44fd-9371-61bb30170702 looking at stack, the values of nsConnectionEntry is corrupted, length of mActiveConns array is 1946182912(0x74006500) and capacity is 1912632576(0x72007500) which is too high. This is crash with a build from 20170630 where tfo is turned off. This probebly a latent bug (similar crashes occur in many older versions).
I didn't file a new bug, but [@ PLDHashTable::Search | nsRefPtrHashtable<T>::GetWeak | mozilla::net::nsHttpConnectionMgr::OnMsgReclaimConnection is another signature I see in crash stats that looks possibly related to this bug.
Crash Signature: mozilla::net::nsHttpConnectionMgr::OnMsgReschedTransaction] → mozilla::net::nsHttpConnectionMgr::OnMsgReschedTransaction] [@ PLDHashTable::Search | nsRefPtrHashtable<T>::GetWeak | mozilla::net::nsHttpConnectionMgr::OnMsgProcessPendingQ]
Keywords: crash
Fixed by bug 1363372
Status: ASSIGNED → RESOLVED
Closed: 9 years ago
Resolution: --- → FIXED
Depends on: 1363372
Target Milestone: --- → mozilla57
You need to log in before you can comment on or make changes to this bug.