lkml.org 
[lkml]   [2008]   [Nov]   [21]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: next-20081121:WARNING: at drivers/dma/dmaengine.c:665 dma_async_device_unregister()
On Fri, Nov 21, 2008 at 3:40 AM, Alexander Beregalov
<a.beregalov@gmail.com> wrote:
> Hi
>
> It occured during shutdown process:
>
> sd 0:2:0:0: [sda] Synchronizing SCSI cache
> ioatdma 0000:00:08.0: Removing dma and dca services
> ------------[ cut here ]------------
> WARNING: at drivers/dma/dmaengine.c:665 dma_async_device_unregister+0x90/0xd0()
> dma_async_device_unregister called while 1 clients hold a reference
> Modules linked in:
> Pid: 12246, comm: reboot Tainted: G W 2.6.28-rc5-next-20081121 #8
> Call Trace:
> [<ffffffff80236b7c>] warn_slowpath+0xae/0xd5
> [<ffffffff80298414>] ? add_partial+0x1a/0x50
> [<ffffffff804588e6>] ? dma_async_device_unregister+0x22/0xd0
> [<ffffffff802576c4>] ? trace_hardirqs_on_caller+0x1f/0x153
> [<ffffffff804e4829>] ? __mutex_lock_common+0x371/0x3be
> [<ffffffff804588e6>] ? dma_async_device_unregister+0x22/0xd0
> [<ffffffff802576c4>] ? trace_hardirqs_on_caller+0x1f/0x153
> [<ffffffff80257805>] ? trace_hardirqs_on+0xd/0xf
> [<ffffffff804e4a51>] ? __mutex_unlock_slowpath+0x13c/0x145
> [<ffffffff80458954>] dma_async_device_unregister+0x90/0xd0
> [<ffffffff80459b62>] ioat_dma_remove+0x1c/0xd0
> [<ffffffff80459992>] ioat_shutdown_functionality+0x66/0x74
> [<ffffffff80388ad2>] pci_device_shutdown+0x24/0x38
> [<ffffffff803e042c>] device_shutdown+0x50/0x99
> [<ffffffff80245612>] kernel_restart_prepare+0x27/0x2e
> [<ffffffff80245652>] kernel_restart+0x11/0x44
> [<ffffffff802457af>] sys_reboot+0x11d/0x149
> [<ffffffff8024d3f1>] ? hrtimer_nanosleep+0x104/0x119
> [<ffffffff8024c9f6>] ? hrtimer_wakeup+0x0/0x21
> [<ffffffff804e4ab5>] ? do_nanosleep+0x50/0xb6
> [<ffffffff802576c4>] ? trace_hardirqs_on_caller+0x1f/0x153
> [<ffffffff804e583a>] ? trace_hardirqs_on_thunk+0x3a/0x3f
> [<ffffffff8020b75b>] system_call_fastpath+0x16/0x1b
> ---[ end trace 4eaa2a86a8e2da22 ]---

Looks like the network stack needs to drop its reference to dmaengine
at shutdown. Need to investigate if netdev_exit is the proper
location for this...

> ------------[ cut here ]------------
> WARNING: at drivers/base/core.c:122 device_release+0x66/0x68()
> Device 'dma0chan0' does not have a release() function, it is broken and must be
> fixed.
> Modules linked in:
> Pid: 12246, comm: reboot Tainted: G W 2.6.28-rc5-next-20081121 #8
> Call Trace:
> [<ffffffff80236b7c>] warn_slowpath+0xae/0xd5
> [<ffffffff802e4812>] ? release_sysfs_dirent+0x7d/0x9d
> [<ffffffff802e484a>] ? __sysfs_put+0x18/0x1a
> [<ffffffff802e4aa3>] ? sysfs_addrm_finish+0x233/0x27b
> [<ffffffff802af85f>] ? ifind+0x2c/0x84
> [<ffffffff80257805>] ? trace_hardirqs_on+0xd/0xf
> [<ffffffff804e44a4>] ? mutex_trylock+0x146/0x15a
> [<ffffffff802e4668>] ? sysfs_addrm_start+0x78/0xa5
> [<ffffffff802e4668>] ? sysfs_addrm_start+0x78/0xa5
> [<ffffffff8037657c>] ? kobj_kset_leave+0x1e/0x57
> [<ffffffff803e0b57>] device_release+0x66/0x68
> [<ffffffff8037663e>] kobject_release+0x52/0x68
> [<ffffffff803765ec>] ? kobject_release+0x0/0x68
> [<ffffffff80377429>] kref_put+0x43/0x4f
> [<ffffffff80376546>] kobject_put+0x47/0x4b
> [<ffffffff803e048a>] put_device+0x15/0x17
> [<ffffffff803e12a3>] device_unregister+0x19/0x1e
> [<ffffffff80458968>] dma_async_device_unregister+0xa4/0xd0
> [<ffffffff80459b62>] ioat_dma_remove+0x1c/0xd0
> [<ffffffff80459992>] ioat_shutdown_functionality+0x66/0x74
> [<ffffffff80388ad2>] pci_device_shutdown+0x24/0x38
> [<ffffffff803e042c>] device_shutdown+0x50/0x99
> [<ffffffff80245612>] kernel_restart_prepare+0x27/0x2e
> [<ffffffff80245652>] kernel_restart+0x11/0x44
> [<ffffffff802457af>] sys_reboot+0x11d/0x149
> [<ffffffff8024d3f1>] ? hrtimer_nanosleep+0x104/0x119
> [<ffffffff8024c9f6>] ? hrtimer_wakeup+0x0/0x21
> [<ffffffff804e4ab5>] ? do_nanosleep+0x50/0xb6
> [<ffffffff802576c4>] ? trace_hardirqs_on_caller+0x1f/0x153
> [<ffffffff804e583a>] ? trace_hardirqs_on_thunk+0x3a/0x3f
> [<ffffffff8020b75b>] system_call_fastpath+0x16/0x1b
> ---[ end trace 4eaa2a86a8e2da22 ]---

With the old reference counting scheme we apparently never released
the dma class devices. I'll address this.

Strange that this did not trigger on my iop13xx platform...

Thanks for the report,
Dan


\
 
 \ /
  Last update: 2008-11-21 18:21    [W:1.476 / U:0.080 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site