nV News Forums

 
 

nV News Forums (http://www.nvnews.net/vbulletin/index.php)
-   NVIDIA Linux (http://www.nvnews.net/vbulletin/forumdisplay.php?f=14)
-   -   Linux Real Time Kernel - a lot of BUG: sleeping function called from invalid context (http://www.nvnews.net/vbulletin/showthread.php?t=149155)

dazo 03-20-10 07:16 PM

Linux Real Time Kernel - a lot of BUG: sleeping function called from invalid context
 
1 Attachment(s)
I'm running Fedora 12 with the Planet CCRMA Real Time kernel ... and it dumps regularly (every 10-15sec) the following BUG notification via klog:

Code:

BUG: sleeping function called from invalid context at kernel/rtmutex.c:684
in_atomic(): 1, irqs_disabled(): 1, pid: 9, name: sirq-tasklet/0
Pid: 9, comm: sirq-tasklet/0 Tainted: P          2.6.31.12-1.rt21.1.fc12.ccrma.x86_64.rt #1
Call Trace:
 [<ffffffff81042002>] __might_sleep+0xe6/0xe8
 [<ffffffff81083ade>] rt_spin_lock_fastlock.clone.1+0x34/0x79
 [<ffffffff814514bd>] rt_spin_lock+0xe/0x10
 [<ffffffff810474e3>] __wake_up+0x20/0x46
 [<ffffffffa06b8b38>] nv_post_event+0xc2/0xdd [nvidia]
 [<ffffffffa05d0b4f>] _nv005330rm+0xe7/0x131 [nvidia]
 [<ffffffffa0356118>] ? _nv015804rm+0xac/0xd1 [nvidia]
 [<ffffffffa01f9263>] ? _nv020367rm+0x9a/0xa9 [nvidia]
 [<ffffffffa01f922d>] ? _nv020367rm+0x64/0xa9 [nvidia]
 [<ffffffffa0694a49>] ? _nv010072rm+0x1a5/0x2e9 [nvidia]
 [<ffffffffa04e1d88>] ? _nv010120rm+0x6b/0x73 [nvidia]
 [<ffffffffa04e1d9e>] ? _nv010122rm+0xe/0x13 [nvidia]
 [<ffffffffa030fd85>] ? _nv016414rm+0x184/0x216 [nvidia]
 [<ffffffffa0514aa7>] ? _nv008600rm+0x30f/0x813 [nvidia]
 [<ffffffffa0513fc6>] ? _nv008603rm+0x66/0x90 [nvidia]
 [<ffffffffa05d2bc0>] ? _nv003827rm+0xa0/0xd3 [nvidia]
 [<ffffffffa05d8853>] ? rm_isr_bh+0x5a/0x8d [nvidia]
 [<ffffffffa06b99a9>] ? nv_kern_isr_bh+0x54/0x59 [nvidia]
 [<ffffffff8105aceb>] ? __tasklet_action+0x9a/0x109
 [<ffffffff8105adf0>] ? tasklet_action+0x4a/0x4c
 [<ffffffff8105ba28>] ? ksoftirqd+0x192/0x2ba
 [<ffffffff8105b896>] ? ksoftirqd+0x0/0x2ba
 [<ffffffff810720ff>] ? kthread+0x91/0x99
 [<ffffffff81012f1a>] ? child_rip+0xa/0x20
 [<ffffffff8107206e>] ? kthread+0x0/0x99
 [<ffffffff81012f10>] ? child_rip+0x0/0x20


I believe this is the classical issue with drivers not being tested on real time kernels (stock kernels with the PREEMPT RT kernel patch set), where spin locks are converted to sleeping spin locks - and they can be preempted by real time processes, iirc.

Please fix this ... as it really makes the nVidia driver useless for those of us using laptops with nVidia cards (not too easy to replace the card with a better working card) and the closed source driver on setups for audio/midi workstations (ie. music production).

jma 03-21-10 01:19 AM

Re: Linux Real Time Kernel - a lot of BUG: sleeping function called from invalid cont
 
You are using driver version 190.53 while current is version 195.36.15?

dazo 03-21-10 04:27 AM

Re: Linux Real Time Kernel - a lot of BUG: sleeping function called from invalid cont
 
1 Attachment(s)
Using the updated driver as well, gives the same result.

Code:

BUG: sleeping function called from invalid context at kernel/rtmutex.c:684
in_atomic(): 1, irqs_disabled(): 1, pid: 23, name: sirq-tasklet/1
Pid: 23, comm: sirq-tasklet/1 Tainted: P          2.6.31.12-1.rt21.1.fc12.ccrma.x86_64.rt #1
Call Trace:
 [<ffffffff81042002>] __might_sleep+0xe6/0xe8
 [<ffffffff81083ade>] rt_spin_lock_fastlock.clone.1+0x34/0x79
 [<ffffffff814514bd>] rt_spin_lock+0xe/0x10
 [<ffffffff810474e3>] __wake_up+0x20/0x46
 [<ffffffffa060a115>] nv_post_event+0x140/0x15e [nvidia]
 [<ffffffffa0515e79>] _nv006542rm+0xe4/0x132 [nvidia]
 [<ffffffffa02d08d8>] ? _nv018347rm+0xac/0xd1 [nvidia]
 [<ffffffffa0162473>] ? _nv023383rm+0x9a/0xa9 [nvidia]
 [<ffffffffa016243d>] ? _nv023383rm+0x64/0xa9 [nvidia]
 [<ffffffffa05ea7a9>] ? _nv011709rm+0x1a5/0x2e9 [nvidia]
 [<ffffffffa041ea78>] ? _nv011726rm+0x6b/0x73 [nvidia]
 [<ffffffffa041ea8e>] ? _nv011728rm+0xe/0x13 [nvidia]
 [<ffffffffa02815f8>] ? _nv019143rm+0x184/0x216 [nvidia]
 [<ffffffffa0452c91>] ? _nv009994rm+0x326/0x7ca [nvidia]
 [<ffffffffa04520be>] ? _nv009997rm+0xbe/0xe3 [nvidia]
 [<ffffffffa0517fc0>] ? _nv004505rm+0xa0/0xd3 [nvidia]
 [<ffffffffa051dc03>] ? rm_isr_bh+0x5a/0x8d [nvidia]
 [<ffffffffa060b025>] ? nv_kern_isr_bh+0x55/0x5a [nvidia]
 [<ffffffff8105aceb>] ? __tasklet_action+0x9a/0x109
 [<ffffffff8105adf0>] ? tasklet_action+0x4a/0x4c
 [<ffffffff8105ba28>] ? ksoftirqd+0x192/0x2ba
 [<ffffffff8105b896>] ? ksoftirqd+0x0/0x2ba
 [<ffffffff810720ff>] ? kthread+0x91/0x99
 [<ffffffff81012f1a>] ? child_rip+0xa/0x20
 [<ffffffff8107206e>] ? kthread+0x0/0x99
 [<ffffffff81012f10>] ? child_rip+0x0/0x20


gbil 03-21-10 09:23 AM

Re: Linux Real Time Kernel - a lot of BUG: sleeping function called from invalid cont
 
I'm pretty sure that nvidia doesn't support RT kernels and up to now patches from the community where needed to compile with RT kernels. Before any nasty comments kick in, my point is that I wouldn't expect any support on this issue from official nvidia sources any time soon.

dazo 03-21-10 09:44 AM

Re: Linux Real Time Kernel - a lot of BUG: sleeping function called from invalid cont
 
If that's so, I really regret having bought a computer with hardware components with lousy drivers. And trust me, I won't do that mistake once again - unless some nVidia officials can come up and reject your comment.

From what I've heard from other places where drivers have cause locking issues on RT, it's just some rework needed to be done on the locking part, and then things usually work better - in both RT and non-RT kernels. So if nVidia don't find RT users that important, I'll waive goodbye to nVidia.

gsking1 07-18-10 09:52 PM

Re: Linux Real Time Kernel - a lot of BUG: sleeping function called from invalid cont
 
I would also like to ask for the nvidia drivers to work better on the real time kernels. I am using Fedora 13 with the CCRMA provided kernel (2.6.33.4-rt..) and self installed NVIDIA drivers (see reference below). Most all works fine except when trying to play any videos with VDPAU. Playing video with VDPAU will cause the below system errors, such as by using mplayer or smplayer (tested with mt2, h264, mp4, etc...) . However, using the generic Fedora 13 kernel (2.6.33.6) it works fine. I have experienced this on both the 2.6.31 and 2.6.33 series rt kernels under Fedora 12.

This system is 32bit dual core intel with 9500GT graphics card. Nvidia driver 256.35, but also had this on prior versions 195.xx I think under FC12.

There are not any other problems noticed otherwise and I really like the VDPAU options you've added to linux. Please consider improving this. The rt kernels are great for multi-media audio production and really do a great job. Please feel free to contact me should you have any questions or comments. Thanks, Geoff

Here's a mailing list reference:
http://ccrma-mail.stanford.edu/piper...ne/016961.html

Example system output using dmesg:

BUG: sleeping function called from invalid context at kernel/rtmutex.c:684
pcnt: 1 0 in_atomic(): 1, irqs_disabled(): 1, pid: 10, name: sirq-tasklet/0
Pid: 10, comm: sirq-tasklet/0 Tainted: P 2.6.33.4-106.rt20.1.fc13.ccrma.i686.rtPAE #1
Call Trace:
[<c042f632>] __might_sleep+0xcc/0xd4
[<c04654ab>] rt_spin_lock_fastlock.clone.1+0x26/0x5f
[<c0793442>] rt_spin_lock+0x8/0xa
[<c0433e2d>] __wake_up+0x14/0x34
[<fa7b45e7>] nv_post_event+0xbf/0xd3 [nvidia]
[<fa7906df>] _nv021435rm+0xc8/0x115 [nvidia]
[<fa51e5b3>] ? _nv010212rm+0x97/0xb3 [nvidia]
[<fa51e813>] ? _nv019434rm+0xae/0xbf [nvidia]
[<fa637474>] ? _nv015461rm+0xb81/0x11c5 [nvidia]
[<fa68423e>] ? _nv018492rm+0x4c7/0xacb [nvidia]
[<fa67c375>] ? _nv018471rm+0x95/0xd5 [nvidia]
[<fa7946cd>] ? _nv002115rm+0x59/0x81 [nvidia]
[<fa7998f6>] ? rm_isr_bh+0x52/0x7d [nvidia]
[<fa7b5107>] ? nv_kern_isr_bh+0x45/0x4e [nvidia]
[<c0443bfb>] ? __tasklet_action+0x78/0xd8
[<c0443cdf>] ? tasklet_action+0x40/0x44
[<c0444432>] ? run_ksoftirqd+0x138/0x236
[<c04442fa>] ? run_ksoftirqd+0x0/0x236
[<c0455736>] ? kthread+0x5f/0x64
[<c04556d7>] ? kthread+0x0/0x64
[<c040907e>] ? kernel_thread_helper+0x6/0x10

octapus 10-29-10 01:00 PM

Re: Linux Real Time Kernel - a lot of BUG: sleeping function called from invalid cont
 
The same kernel crash appeared in other cases, and the author of RT patch provided a patch:

http://www.spinics.net/lists/linux-r.../msg05515.html


All times are GMT -5. The time now is 08:01 PM.

Powered by vBulletin® Version 3.7.1
Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Copyright 1998 - 2014, nV News.