Re: "don't bugger nd->seq" seems to break umount sometimes

[Ian Campbell <ian.campbell@citrix.com>]

On Tue, 2015-05-19 at 17:05 +0800, Zefan Li wrote:
> On 2015/5/19 16:33, Ian Campbell wrote:
> > On Thu, 2015-04-30 at 09:15 +0800, Zefan Li wrote:
> >> On 2015/4/29 20:21, Al Viro wrote:
> >>> On Wed, Apr 29, 2015 at 12:45:45PM +0100, Ian Jackson wrote:
> >>>
> >>>> The symptoms are that `umount' fails with EBUSY,
> >>>
> >>>     [lizf: Backported to 3.4:
> >>>      - remove the changes to follow_link() as it doesn't call set_root()]
> >>>
> >>> looks dubious - I don't have -stable in front of me, but set_root() in
> >>> follow_link() had migrated from __vfs_follow_link(), so could you try
> >>> (tr '#' '\t' | ed fs/namei.c) <<'EOF'
> >>> /__vfs_follow_link/
> >>> /set_root/s/^/#/
> >>> i
> >>> ##if (!nd->root.mnt)
> >>> .
> >>> wq
> >>> EOF
> >>> and see if it helps?
> >>> .
> >>>
> >>
> >> My fault. I just checked 3.2.y, and it made the right change.
> > 
> > Is there going to be a 3.4.108 with a fix for this?
> > 
> 
> definitely.

Great. Do you have a timeline in mind?

We have a daily automated test which has been hitting this issue for a
while, I can continue to ignore the associated cron spam of course ;-)

Ian.