Re: atomic queue limit updates for stackable devices

Song Liu <[email protected]> Wed, 28 Feb 2024 18:02:33 -0800
Newsgroups gmane.linux.lvm.devel
Message-ID <CAPhsuW5c5o8JS9T8CCGU811jpxrbepjJHnYpDbge+1rT6j8NbA__6740.40188766725$1709172182$gmane$org@mail.gmail.com>
On Wed, Feb 28, 2024 at 11:56=E2=80=AFAM Christoph Hellwig <[email protected]> wro=
te:
>
> On Tue, Feb 27, 2024 at 01:50:19PM -0800, Song Liu wrote:
> > I think we can do something like:
> >
> > make check S=3D<list of test to skip>
> >
> > I don't have a reliable list to skip at the moment, as some of the test=
s
> > fail on some systems but not on others. However, per early report,
> > I guess we can start with the following skip list:
> >
> > shell/integrity-caching.sh
> > shell/lvconvert-raid-reshape-linear_to_raid6-single-type.sh
> > shell/lvconvert-raid-reshape.sh
>
> Thanks.  I've been iterating over it this morning, eventually growing
> to:
>
> make check
> S=3Dshell/integrity-caching.sh,shell/lvconvert-raid-reshape-linear_to_rai=
d6-single-type.sh,shell/lvconvert-raid-reshape.sh,shell/lvconvert-raid-resh=
ape-linear_to_striped-single-type.sh,shell/lvconvert-raid-reshape-linear_to=
_striped.sh,shell/lvchange-raid456.sh,shell/component-raid.sh,shell/lvconve=
rt-raid-reshape-load.sh,shell/lvchange-raid-transient-failures.sh,shell/lvc=
onvert-raid-reshape-striped_to_linear-single-type.sh,shell/lvconvert-raid-r=
eshape-striped_to_linear.sh,shell/lvconvert-raid-reshape-stripes-load-fail.=
sh,shell/lvconvert-raid-reshape-stripes-load-reload.sh,shell/lvconvert-raid=
-reshape-stripes-load.sh,lvconvert-raid-reshape.sh,shell/lvconvert-raid-res=
tripe-linear.sh,shell/lvconvert-raid-status-validation.sh,shell/lvconvert-r=
aid-takeover-linear_to_raid4.sh,shell/lvconvert-raid-takeover-raid4_to_line=
ar.sh,shell/lvconvert-raid-takeover-alloc-failure.sh
>
> before giving up.  I then tried to run the md-6.9 branch that's
> supposed to have the fixes, but I still see the same md_stop_writes
> hangs.

md-6.9 branch doesn't have all the fixes, as some recent fixes
are routed via the md-6.8 branch. You can try on this branch, which
should provide a better base line. The set applies cleanly on this
branch.

https://git.kernel.org/pub/scm/linux/kernel/git/song/md.git/log/?h=3Dmd-6.9=
-for-hch

Thanks,
Song