Closed Bug 844156 Opened 12 years ago Closed 12 years ago

Analysis of other threads running when XPC_WN_Helper_NewResolve is crashing

Categories

(Socorro :: Data request, task)

x86
Windows 7
task
Not set
blocker

Tracking

(Not tracked)

RESOLVED FIXED

People

(Reporter: benjamin, Assigned: selenamarie)

References

Details

In bug 830531 we are experiencing a topcrash which correlates with AMD graphics cards/drivers. It is a memory corruption issue, correlates to multi-core machines, and is very difficult to figure out. I have discovered that many of the reports have one or more threads running ATI driver code: (calls from atiumdag.dll). See e.g. https://crash-stats.mozilla.com/report/index/d11cbfaf-f38d-4a08-a470-2401a2130222 thread 29 and thread 30. The report I would like is the following: On the set of crashes from Firefox 19 release with the signature "XPC_WN_Helper_NewResolve" for the last 3 days: * search all the threads to find threads with atiumdag.dll in the stack (a frame signature that starts with "atiumdag.dll@" report: * the total percentage of crashes that have any ATI thread running * A list of breakpad IDs for any reports where WaitForSingleObject is *not* present in the atiumdal.dll thread. e.g. https://crash-stats.mozilla.com/report/index/d11cbfaf-f38d-4a08-a470-2401a2130222 would not be listed because in both thread 29 and thread 30 WaitForSingleObject is present. If it is possible to do this report today, that would be really helpful.
Assignee: nobody → sdeckelmann
Total % of crashes from 2/18-2/22 that have any ATI thread of the crashes with signature "XPC_WN_Helper_NewResolve": 72% (3105 / 4301)
Benjamin, the main signature in that bug is actually "TlsGetValue", wouldn't it be more helpful to gather stats on that one, maybe also restricted to FF19 and Win8 just to make sure?
With help from Selena, I've finished custom runs on this, and it seems like about %75 running the user-mode threads is pretty constant. I found a very small but perhaps interesting set of crashes where the ATI user driver was running: 23c2eb21-1b3a-4079-954f-d4f202130223 2 0x1002 8.982.6.0 26cd5b04-b0da-41fe-a064-f1cb52130224 2 0x1002 8.982.6.0 518102a2-16c7-4e5c-a4dd-db3ed2130223 2 0x1002 8.982.6.0 62d502b6-c423-4040-b33f-99da32130224 2 0x1002 8.982.7.0 acec81cf-4077-4f3f-b89e-332e52130224 2 0x1002 8.982.6.0 bde3ccc8-aecb-4537-a8d9-e82f02130223 2 0x1002 8.982.7.0 bd230d14-98d4-4ad4-874a-9e5d12130224 2 0x1002 8.712.2.1000 f5ff85a0-8810-4d00-b7ed-5c1b62130223 2 0x1002 8.982.6.0 See https://github.com/bsmedberg/socorro-toolbox/compare/LookupATIFrames for the code. I'll close this for now.
Status: NEW → RESOLVED
Closed: 12 years ago
Resolution: --- → FIXED
Blocks: 830531
You need to log in before you can comment on or make changes to this bug.