On Mon, Jun 24, 2019 at 5:29 PM Holger Hoffstätte
<hol...@applied-asynchrony.com> wrote:
>
> On 6/25/19 12:46 AM, Chris Murphy wrote:
> > Same call trace on ro mount and ro scrub with 5.2.0-rc2+, but also an
> > additional call trace related to zstd. As this is a zstd compressed
> > file system, it might be related.
> >
> > [ 366.319583] ================================
> > [ 366.325036] WARNING: inconsistent lock state
> > [ 366.330615] 5.2.0-0.rc2.git1.2.fc31.x86_64 #1 Tainted: G W
> > [ 366.336202] --------------------------------
> > [ 366.341788] inconsistent {SOFTIRQ-ON-W} -> {IN-SOFTIRQ-W} usage.
> > [ 366.347423] swapper/4/0 [HC0[0]:SC1[1]:HE1:SE0] takes:
> > [ 366.353042] 000000006070e818 (&(&wsm.lock)->rlock){+.?.}, at:
> > zstd_reclaim_timer_fn+0x26/0x170 [btrfs]
>
> Seeing that this is apparently *rc2* (which is insane to ship) the explanation
> for the above is that 5.2-rc2 was tagged on 2019-05-26, but the patch for the
> above problem [1] was committed *after* that, on 2019-05-28.
a. Please relax. I said from the outset it's Fedora Rawhide, they
build mainline rc kernels on Monday, and each weekday there's a plus
version built from whatever's in Linus's tree prior to the next rc.
It's been this way for a very long time. No one puts Rawhide into
production including myself, it's a test machine.
b. That rc2+ kernel above was used today for read-only inspection of
the target Btrfs file system, just because it's what I already had on
hand. That I've been using the rc5+ kernel I cited in the first email
for the past three days, so I'm not sure why the problem starts
suddenly. The problem is with a stale systemd journal.
--
Chris Murphy
