Re: [PATCH 15/13] sysfs: Protect sysfs_refresh_inode with inode mutex.

[ebiederm@xmission.com (Eric W. Biederman)]

Tejun Heo <tj@kernel.org> writes:

> Eric W. Biederman wrote:
>> In general everything that writes to vfs inodes holds the
>> inode mutex, so hold the inode mutex over sysfs_refresh_inode.
>> The sysfs data structures don't need this but it looks like the
>> vfs might.
>> 
>> Signed-off-by: Eric W. Biederman <ebiederm@aristanetworks.com>
>
> Acked-by: Tejun Heo <tj@kernel.org>
>
> Sidenote: Hmmm... Originally, sysfs completely depended on vfs locking
> but with sysfs_dirent separation, the tree structure itself and some
> attributes went under the protection of sysfs_mutex while leaving more
> vfs oriented fields under vfs locking.  This patchset makes sysfs
> lazier so it can't depend on any vfs layer locking.  I think you've
> converted all necessary places while removing dependency on
> dentry/inode from update operations but it might be a good idea to do
> a audit pass over how fields are being protected now.

You raised a good point.  I took a quick second pass through.
I did not see anything I have missed, and I did not change anything
else on the vfs path.

So at the very least I don't expect there are any locking related
regressions.

Eric