Closed
Bug 156320
Opened 24 years ago
Closed 24 years ago
trunk topcrash [@ nsRecyclingAllocator::AddToFreeList]
Categories
(Core :: XPCOM, defect, P1)
Tracking
()
RESOLVED
FIXED
mozilla1.0.1
People
(Reporter: jcarpenter0524, Assigned: dp)
Details
(Keywords: crash, topcrash)
Crash Data
Attachments
(1 file)
|
704 bytes,
patch
|
dbaron
:
review+
|
Details | Diff | Splinter Review |
nsRecyclingAllocator::AddToFreeList is a topcrasher for the trunk
Platform List
Windows NT 5.0 build 2195
Windows NT 5.1 build 2600
Windows 98 4.90 build 73010104
Windows 98 4.10 build 67766446
Build Id List
2002070504
2002070408
2002070308
2002070404
2002070204
2002070310
2002070608
2002070508
STACK TRACE:
nsRecyclingAllocator::AddToFreeList
[c:/builds/seamonkey/mozilla/xpcom/ds/nsRecyclingAllocator.cpp line 352]
nsRecyclingAllocator::Free
[c:/builds/seamonkey/mozilla/xpcom/ds/nsRecyclingAllocator.cpp line 229]
zlibFree [c:/builds/seamonkey/mozilla/modules/libjar/nsZipArchive.cpp line 415]
inflate_blocks_free [c:/builds/seamonkey/mozilla/modules/zlib/src/infblock.c
line 378]
inflateEnd [c:/builds/seamonkey/mozilla/modules/zlib/src/inflate.c line 74]
nsZipArchive::InflateItem
[c:/builds/seamonkey/mozilla/modules/libjar/nsZipArchive.cpp line 1454]
nsZipArchive::ReadInit
[c:/builds/seamonkey/mozilla/modules/libjar/nsZipArchive.cpp line 594]
nsJARInputStream::Init
[c:/builds/seamonkey/mozilla/modules/libjar/nsJARInputStream.cpp line 108]
nsJAR::GetInputStream [c:/builds/seamonkey/mozilla/modules/libjar/nsJAR.cpp
line 346]
nsJARChannel::GetInputStream
[c:/builds/seamonkey/mozilla/netwerk/protocol/jar/src/nsJARChannel.cpp line 681]
nsFileTransport::Process
[c:/builds/seamonkey/mozilla/netwerk/base/src/nsFileTransport.cpp line 693]
URLs\COMMENTS:
(7944805) URL: http://slate.msn.com/
(7944805) Comments: crashed. I kept on working though
(8079248) Comments: opening the web browser
(8041697) URL: http://www.gameassault.com
(7980367) URL: www.firearmsmod.com
(7980367) Comments: Blocked an ad reloaded page. Clicked on picture of
the week. Looked at picture closed browser and it crashed.
(8090534) URL: http://www.mozilla.org/
| Reporter | ||
Updated•24 years ago
|
dp and I looked at this -- from the disassembly it looks like |node| is null,
which suggests that we also need to check |mNotUsedList| once the lock is aquired.
Assignee: dougt → dp
(Where |node| is the same as |mNotUsedList|.)
| Assignee | ||
Comment 3•24 years ago
|
||
Thanks dbaron for doing most of the legwork. Accepting...
Severity: normal → major
Status: NEW → ASSIGNED
Priority: -- → P1
Target Milestone: --- → mozilla1.0.1
Comment 4•24 years ago
|
||
With our new "Zero Tolerance for New Crash" Policy we need to fix this ASAP. I
am marking it as blocker. If we leave the crash for longtime we may not be able
to reproduce and the crash will become topcrash in our main release.
Severity: major → blocker
Keywords: zt4newcrash
| Assignee | ||
Comment 5•24 years ago
|
||
This crash is well understood. I have a patch (will attach to bug). This is a
multi threading problem only. Plus only 8 crashes so far in 4 days. I dont think
we should make this a blocker - which will get the tree to not open. I will let
shiva figure out if he wants to change his blocker designation.
I am building testing and hope to checkin by today.
| Assignee | ||
Comment 6•24 years ago
|
||
The crashes happens because of the fact that we check for non-null, wait for
lock, then assume we are still non-null for the not-used-list.
Changed the sequence to acquire-lock, check for non-null, use it.
| Assignee | ||
Comment 7•24 years ago
|
||
dougt / dbaron : could you guys throw in a review ?
Comment on attachment 90558 [details] [diff] [review]
Enter lock before checking for availability in not-used-list
r=dbaron (although if this condition is hit a lot, would it be safe to do the
check both inside and outside the lock?)
Attachment #90558 -
Flags: review+
| Assignee | ||
Comment 9•24 years ago
|
||
That was my initial thought when we looked at this. Then this condition will get
hit only when there is a failover case - when the recycling allocator is already
holding max number of free blocks. That case shouldn't happen too often. That is
how I convinced myself that the simplicity of code and reduction of another
check was better.
| Assignee | ||
Comment 10•24 years ago
|
||
Fix checked in. Let me see the next two days crash report before closing the bug
as i personally couldnt reproduce the bug.
Comment 11•24 years ago
|
||
Removing zt4newcrash. ccing myself.
Severity: blocker → critical
Keywords: zt4newcrash
| Assignee | ||
Comment 12•24 years ago
|
||
No new crashes since we checked this in. It has only been a day. I will mark it
closed and hope to reopen if there are new crashes.
Status: ASSIGNED → RESOLVED
Closed: 24 years ago
Resolution: --- → FIXED
Updated•15 years ago
|
Crash Signature: [@ nsRecyclingAllocator::AddToFreeList]
You need to log in
before you can comment on or make changes to this bug.
Description
•