lkml.org 
[lkml]   [2014]   [Jul]   [25]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
SubjectRe: [PATCH] cpufreq, store_scaling_governor requires policy->rwsem to be held for duration of changing governors
From
On 24 July 2014 23:24, Prarit Bhargava <prarit@redhat.com> wrote:
> The closer I looked at commit 955ef483, the more questions I have. The first
> thing is that it appears that the stacktrace includes function calls that
> are not, and never have been, part of the linux.git tree, ie) the call trace
> shows
>
> [<c0055a79>] lock_acquire+0x61/0xbc
> [<c00fabf1>] sysfs_addrm_finish+0xc1/0x128
> [<c00f9819>] sysfs_hash_and_remove+0x35/0x64
> [<c00fbe6f>] remove_files.isra.0+0x1b/0x24
> [<c00fbea5>] sysfs_remove_group+0x2d/0xa8
> [<c02f9a0b>] cpufreq_governor_interactive+0x13b/0x35c
> [<c02f61df>] __cpufreq_governor+0x2b/0x8c
> [<c02f6579>] __cpufreq_set_policy+0xa9/0xf8
> [<c02f6b75>] store_scaling_governor+0x61/0x100
> [<c02f6f4d>] store+0x39/0x60
> [<c00f9b81>] sysfs_write_file+0xed/0x114
> [<c00b3fd1>] vfs_write+0x65/0xd8
> [<c00b424b>] sys_write+0x2f/0x50
> [<c000cdc1>] ret_fast_syscall+0x1/0x52
>
> and the top part of the stack from cpufreq_governor_interactive() appear to
> be for a driver that is not in the linux.git tree. Google search does show
> that it exists in various android trees, however, my concern is that the "core"
> scaling governors are now broken because of an out tree driver.

Actually the problem did occur with ondemand/conservative as well. And
the problem was, we accessed something from cpufreq sysfs and then were
trying to remove that only and so some recursive locking stuff..

And so nothing interactive governor specific.

The problem looks similar to what's reported here:
http://comments.gmane.org/gmane.linux.power-management.general/47397

And that's how I replied to it:

I have pushed a fix earlier for similar issues:
19c7630 cpufreq: serialize calls to __cpufreq_governor()

but was later reverted by Srivatsa:
56d07db cpufreq: Remove temporary fix for race between CPU hotplug and
sysfs-writes

because we didn't thought about this usecase.

I propose we get that back. I have tested a revert of 56d07db on
my setup and didn't see any crash..

Please see if that works for you as well..

--
viresh


\
 
 \ /
  Last update: 2014-07-25 07:01    [W:0.077 / U:0.252 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site