Opened 4 months ago

Last modified 5 weeks ago

#19181 new bug

[kernel] PANIC: General protection exception after starting 6th instance of GIMP

Reported by: diver Owned by: nobody
Priority: normal Milestone: Unscheduled
Component: System/Kernel Version: R1/beta5
Keywords: Cc:
Blocked By: Blocking:
Platform: All

Description

This is hrev58201, real hw with 16GB RAM.

First 5 instances of GIMP started quite quickly with no problem.
The 6th one crashed the kernel.

Attachments (3)

IMG_20241016_223230.jpg (218.5 KB ) - added by diver 4 months ago.
screenshot2.png (423.5 KB ) - added by diver 4 months ago.
syslog (331.8 KB ) - added by diver 4 months ago.
syslog from hrev58244

Download all attachments as: .zip

Change History (10)

by diver, 4 months ago

Attachment: IMG_20241016_223230.jpg added

comment:1 by waddlesplash, 4 months ago

Does it happen on beta5?

comment:2 by diver, 4 months ago

I guess I could try to downgrade and check again.

comment:3 by diver, 4 months ago

Downgraded to beta5 and couldn't reproduce, however couldn't start more than 7 instances (Out of memory) with plenty of free RAM left.

Then I updated to hrev58244 and couldn't reproduce either but still OOM.

by diver, 4 months ago

Attachment: screenshot2.png added

by diver, 4 months ago

Attachment: syslog added

syslog from hrev58244

comment:4 by waddlesplash, 4 months ago

(Out of memory) with plenty of free RAM left.

That can happen if applications "reserve" far more memory than they actually use, so this isn't a bug in Haiku.

If you can't reproduce the KDL even after testing, then may we should just close as not-reproducible. Though seeing as the KDL looks like a use-after-free, it may actually be unrelated to the recent VM churn...

comment:5 by diver, 4 months ago

After starting several GIMPs starting even small built-in apps (like Terminal) throws Out of memory.

I will try some more to reproduce this GPE.

comment:6 by waddlesplash, 4 months ago

That's what I'd expect. As long as things go back to normal once GIMP is closed, then all's well.

comment:7 by waddlesplash, 5 weeks ago

The only way I can see this happening is if somehow the vnode and its cache were deleted out from under this thread, but I'm not sure how that could occur. At any rate the additions of more assertions to the VFS vnode ref count functions may catch this slightly earlier if it does occur again.

Note: See TracTickets for help on using tickets.