Re: [RESEND][PATCH] xfs: add online uevent for mount operation

[Dave Chinner <david@fromorbit.com>]

On Thu, Aug 24, 2017 at 10:10:28AM -0700, Darrick J. Wong wrote:
> On Thu, Aug 24, 2017 at 09:41:55PM +0800, Hou Tao wrote:
> > It will be useful if there is a corresponding online uevent after
> > a XFS filesystem has been mounted. A typical usage of the uevent
> > is setting the error configuration for a specific XFS filesystem
> > or all XFS filesystems by using udevd.
> > 
> > The following is an example of udevd rule which will shutdown
> > any XFS filesystem after the filesystem gets any IO error:
> > 
> >     ACTION=="online", SUBSYSTEM=="xfs", DEVPATH=="/fs/xfs/*", \
> >     RUN+="/bin/sh -c 'echo 0 > /sys%p/error/metadata/default/max_retries; \
> > 	    echo 0 > /sys%p/error/metadata/EIO/max_retries; \
> > 	    echo 0 > /sys%p/error/metadata/ENOSPC/max_retries; \
> > 	    echo 0 > /sys%p/error/metadata/ENODEV/max_retries'"
> >
> > Suggested-by: Dave Chinner <david@fromorbit.com>
> > Signed-off-by: Hou Tao <houtao1@huawei.com>
> > ---
> >  fs/xfs/xfs_super.c | 15 +++++++++++++++
> >  1 file changed, 15 insertions(+)
> > 
> > diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c
> > index 38aaacd..695fe85f 100644
> > --- a/fs/xfs/xfs_super.c
> > +++ b/fs/xfs/xfs_super.c
> > @@ -1530,6 +1530,19 @@ xfs_destroy_percpu_counters(
> >  	percpu_counter_destroy(&mp->m_fdblocks);
> >  }
> >  
> > +static void
> > +xfs_fs_uevent(
> > +	struct xfs_mount	*mp,
> > +	enum kobject_action	action)
> > +{
> > +	int err;
> > +
> > +	err = kobject_uevent(&mp->m_kobj.kobject, action);
> 
> It might be useful to send kobject_uevent_env so that we can squeeze in
> things like the fs uuid for easier identification?  I envision a
> /etc/xfs/fs.conf file like this:
> 
> [903db5b1-82cd-4f26-8221-443a4ed0563a]
> error.metadata.EIO.max_retries = 0
> error.fail_at_unmount = 1
> config.cowextsize = 1048576
> config.inherit_noatime = 1
> 
> With a udev helper that uses the fsuuid to find the appropriate section
> of the ini^Wconfig file and load the appropriate values into sysfs.  If
> we ever decide to expose more config knobs through sysfs then they'll
> be automatically supported.

Yup, that's sounds like a good idea.

> > +	if (err)
> > +		xfs_notice(mp, "Sending XFS uevent %d got error %d",
> > +				action, err);
> > +}
> > +
> >  STATIC int
> >  xfs_fs_fill_super(
> >  	struct super_block	*sb,
> > @@ -1667,6 +1680,8 @@ xfs_fs_fill_super(
> >  		goto out_unmount;
> >  	}
> >  
> > +	xfs_fs_uevent(mp, KOBJ_ONLINE);
> 
> That said, I wonder if we ought to move this discussion to fsdevel to
> pull in any /other/ filesystems that have sysfs knobs that they might
> like to be able to persist?  ext4 has a bunch of tunables too...

Nope. We'll never get anywhere if we do that, just like the last
umpteenth times that generic fs events have been discussed. These
are XFS specific kobjs so we can do what we like with them and we
have an XFS specific problem that needs solving....

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com