#9 — Mouse locks up when starting OpenGL applications after resume from suspend with Zen 2.6.32 kernel
| State | Resolved |
|---|---|
| Version: | 2.6.32-zen |
| Area | Zen-Unstable Releases/GIT |
| Issue type | Bug |
| Severity | Important |
| Submitted by | Michael Marley |
| Submitted on | Nov 09, 2009 |
| Responsible |
—
|
| Target release: | 2.6.32-zen |
Around the time of 2.6.32-rc5-zen1, I started noticing that after resuming from suspend on my Dell Vostro 1500 (Intel C2D 2.00 gHz, Nvidia 8600m GT, Kubuntu 9.10 x86-64), starting all OpenGL applications and a few others (like applications in WINE) would cause the mouse cursor to lock up for a few seconds. I am using version 190.42 of the Nvidia binary driver, but I tried several other versions and these do not fix the problem. Using a vanilla kernel from kernel.org, or the Ubuntu-supplied kernel on my system works fine. I am not using TuxOnIce, but instead the pm-utils software supplied with Ubuntu. If I need to provide any other information, please tell me.
- Steps to reproduce:
- 1. Suspend the computer
2. Resume the computer
3. Start an OpenGL application and watch the mouse cursor lock up
Added by
Michael Marley
on
Nov 09, 2009 12:47 PM
Here is my .config.
Added by
dominic duklas
on
Nov 09, 2009 02:51 PM
Could you please try disabling BFS and using CFS instead, and report back? I can't think of anything else that might cause this...
Issue state:
Unconfirmed
→
Confirmed
Severity:
Medium
→
Important
Target release:
None
→
2.6.32-zen
Added by
Michael Marley
on
Nov 09, 2009 02:52 PM
I have in fact already attempted to try this, but compilation fails when I select CFS.
Added by
Michael Marley
on
Nov 09, 2009 03:51 PM
Here is the exact error:kernel/kthread.o: In function `__crc_kthread_bind':
(*ABS*+0x4054c848): multiple definition of `__crc_kthread_bind'
kernel/kthread.o: In function `kthread_bind':
(.text+0x240): multiple definition of `kthread_bind'
kernel/sched.o:(.text+0x7300): first defined here
make[2]: *** [kernel/built-in.o] Error 1
make[1]: *** [kernel] Error 2
make[1]: *** Waiting for unfinished jobs....
Added by
Michael Marley
on
Nov 09, 2009 06:43 PM
Ok, someone fixed this in a Git commit. (The compile error, not the issue I reported.) I compiled the kernel with CFS, and the problem does not seem to occur anymore. Would this bug be specific to the 2.6.32 BFS patches, or should I report this to Con Kolivas?
Added by
dominic duklas
on
Nov 10, 2009 11:15 PM
Hmm, please try with .31 kernel patched with bfs ( zen-stable or vanilla + bfs) and if it still happens, please report to con. Otherwise tell us and we'll look into it.
Added by
Michael Marley
on
Nov 11, 2009 01:14 AM
I tried the 2.6.31 zen kernel with BFS and I did not experience the problem.
Added by
dominic duklas
on
Nov 11, 2009 05:53 PM
This could be a 2.6.32 regression which bfs exposes (as it did in the past with .30 and .31...). I honestly can't think of anything else that could be causing this - and it's hard to investigate more without some backtraces. Does dmesg show anything special after resume?
Added by
Michael Marley
on
Nov 11, 2009 05:56 PM
No, it doesn't show anything out of the norm. How do I make a backtrace?
Added by
dominic duklas
on
Nov 12, 2009 12:57 AM
Well, for one you need debugging options enabled in the kernel, and a means of getting them when the system freezes. You could use a serial console, loop dumping of dmesg to file every second or so, or try and see whether you can still ssh to the machine.
Added by
Michael Marley
on
Nov 12, 2009 01:22 AM
I didn't clarify something properly. The mouse only seizes up for a few seconds at the most, and then resumes normal operation. I have checked, and there isn't anything pertaining to this event in dmesg, the syslog, the Xorg log, or anyplace else I could find.
Added by
dominic duklas
on
Nov 12, 2009 05:30 PM
Oh. In that case just enable kernel debugging stuff to do with locking, tracers etc, and see whether that makes something show up in dmesg.
Added by
Michael Marley
on
Nov 12, 2009 06:32 PM
Ok, I enabled all the debug options I could find related to locking and tracing, and absolutely nothing else is produced in the dmesg. Additionally, I have noticed that in some cases 2D performance in general is quite crappy after resuming from suspend with 2.6.32 and BFS.
Added by
dominic duklas
on
Nov 13, 2009 01:10 AM
Could you please try current zen.git master? We basically remade master branch. Also, could you try enabling _all_ debugging options available? Maybe it's not due to locking after all...
Added by
Michael Marley
on
Nov 13, 2009 01:55 AM
It is FTBFS:kernel/rcutree.c: In function ‘rcu_init_percpu_data’:
kernel/rcutree.c:1607: error: ‘lastcomp’ undeclared (first use in this function)
kernel/rcutree.c:1607: error: (Each undeclared identifier is reported only once
kernel/rcutree.c:1607: error: for each function it appears in.)
make[2]: *** [kernel/rcutree.o] Error 1
make[1]: *** [kernel] Error 2
make[1]: *** Waiting for unfinished jobs....
Added by
Miguel Botón
on
Nov 13, 2009 02:16 PM
RCU tree fixed in latest tree.
Added by
Michael Marley
on
Nov 13, 2009 03:09 PM
OK, I have compiled the kernel with *all* debug options (except early printk over USB and firewire, which caused FTBFS). It now produces quite a bit of debug output, which I have attached.
Added by
Michael Marley
on
Nov 13, 2009 03:09 PM
Added by
dominic duklas
on
Nov 15, 2009 12:04 AM
Please use cfs for now, we'll see what we can do about this.
Added by
Brandon Berhent
on
Nov 24, 2009 09:40 PM
Hi, the only way we can confirm this is a bfs-induced bug is for you to try a vanilla kernel.I'd ask to get a vanilla 2.6.32-rc8 source, and apply this patch:
http://ck.kolivas.org/[…]/2.6.32-rc8-sched-bfs-311.patch
simply apply it with patch -p1 -i 2.6.32-rc8* in the vanilla tree, then build kernel as normal.
If you experience the same problem (or not), report back with a trace please and we can make a confirmation in the right direction!
Added by
dominic duklas
on
Dec 18, 2009 02:00 AM
Any luck with 2.6.32 zen? I think con fixed this already
Added by
Phuong
on
Jan 01, 2010 04:40 PM
I also have a Dell Vostro 1500 laptop with an 8600M card but I don't have this problem. Using 2.6.32-zen3 on gentoo (x86-64), BFS enabled. Tested with wine + warcraft 3 and opengl, all works fine after resuming from suspend to ram.
Added by
dominic duklas
on
Jan 12, 2010 02:49 PM
since it works for you, and we have no more info i think we can mark it as resolved
Issue state:
Confirmed
→
Resolved
No responses can be added.
If you can, please log in before submitting a reaction.
Forgot your password?
New user?
