Render/pepper hang debugging: generate dumps for the browser and renderer processes if users kill a…

Render/pepper hang debugging: generate dumps for the browser and renderer processes if users kill a hung pepper plugin.

This CL collects information in order to diagnose inter-process pepper deadlocks.
- We don't crash those renderers or the browser.
- We only do this on the Windows Canary channel for 20% of pepper plugin hangs, if there are not too many renderers (<4).

BUG=147593
TEST=None


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=158717

[yzshen1, 2012-09-25 19:55:43]

Hi, Brett and Eric.

The previous CL (https://codereview.chromium.org/10909241/) for collecting hang
information didn't provide much useful information.

Therefore, I decide to dump all renderers and the browser process! It should be
okay because as I said in the CL description, I only do it on Canary for 20%
hangs, and if there are less than 4 renderers.

Thanks!

[eroman, 2012-09-25 20:10:19]

lgtm

https://codereview.chromium.org/10983028/diff/1/chrome/browser/hang_monitor/h...
File chrome/browser/hang_monitor/hang_crash_dump_win.cc (right):

https://codereview.chromium.org/10983028/diff/1/chrome/browser/hang_monitor/h...
chrome/browser/hang_monitor/hang_crash_dump_win.cc:63: DCHECK(request_dump) <<
"Failed loading DumpForHangDebugging: error " <<
DumpForHangDebugging --> InjectDumpForHangDebugging

https://codereview.chromium.org/10983028/diff/1/chrome/browser/hang_monitor/h...
chrome/browser/hang_monitor/hang_crash_dump_win.cc:70:
WaitForSingleObject(remote_thread, kGenerateDumpTimeoutMS);
[optional] Note that for better performance, you could make this function take a
list of hprocess, and start all the remotethreads before joining.

Probably best to leave it as-is for simplicity, but wanted to mention this
possibility if the timeouts become an issue.

[yzshen1, 2012-09-25 20:24:09]

https://codereview.chromium.org/10983028/diff/1/chrome/browser/hang_monitor/h...
File chrome/browser/hang_monitor/hang_crash_dump_win.cc (right):

https://codereview.chromium.org/10983028/diff/1/chrome/browser/hang_monitor/h...
chrome/browser/hang_monitor/hang_crash_dump_win.cc:63: DCHECK(request_dump) <<
"Failed loading DumpForHangDebugging: error " <<
On 2012/09/25 20:10:19, eroman wrote:
> DumpForHangDebugging --> InjectDumpForHangDebugging

Done.

https://codereview.chromium.org/10983028/diff/1/chrome/browser/hang_monitor/h...
chrome/browser/hang_monitor/hang_crash_dump_win.cc:70:
WaitForSingleObject(remote_thread, kGenerateDumpTimeoutMS);
On 2012/09/25 20:10:19, eroman wrote:
> [optional] Note that for better performance, you could make this function take
a
> list of hprocess, and start all the remotethreads before joining.
> 
> Probably best to leave it as-is for simplicity, but wanted to mention this
> possibility if the timeouts become an issue.

Thanks for the suggestion!
If this code proves to be helpful and we want it to become some more general
debugging facility, I will follow the suggestion and improve it.

[brettw, 2012-09-25 22:30:46]

lgtm

[commit-bot: I haz the power, 2012-09-25 23:20:26]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/yzshen@chromium.org/10983028/12

[commit-bot: I haz the power, 2012-09-26 00:14:52]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/yzshen@chromium.org/10983028/12

[commit-bot: I haz the power, 2012-09-26 02:01:43]

Change committed as 158717