Re: [PATCH 0/2] blk-throttle: minor fixes and cleanup

[Shinichiro Kawasaki <shinichiro.kawasaki@wdc.com>]

On Mar 17, 2026 / 16:13, Zizhi Wo wrote:
> 
> 
> 在 2026/3/17 15:37, Shinichiro Kawasaki 写道:
> > On Mar 16, 2026 / 16:44, Zizhi Wo wrote:
> > > From: Zizhi Wo <wozizhi@huawei.com>
> > > 
> > > This series contains two minor patches for blk-throttle.
> > > 
> > > Patch1: simple cleanup
> > > Patch2: fix timer scheduled on tg_flush_bios()
> > > 
> > > Zizhi Wo (2):
> > >    blk-throttle: remove leftover carryover checks in
> > >      tg_within_[bps/iops]_limit
> > >    blk-throttle: fix timer scheduled on wrong service_queue in
> > >      tg_flush_bios
> > 
> > Hello ZiZhi, just FYI. Blktests CI trial run found that this series triggered
> > the failure of the test case throtl/004 with sdebug:
> > 
> > throtl/004 (nullb) (delete disk while IO is throttled)
> > throtl/004 (nullb) (delete disk while IO is throttled)       [passed]
> >      runtime    ...  1.667s
> > throtl/004 (sdebug) (delete disk while IO is throttled)
> > throtl/004 (sdebug) (delete disk while IO is throttled)      [failed]
> >      runtime    ...  3.273s
> >      --- tests/throtl/004.out	2026-03-16 10:29:46.761897543 +0000
> >      +++ /home/fedora/blktests/results/nodev_sdebug/throtl/004.out.bad	2026-03-16 13:31:54.355173644 +0000
> >      @@ -1,3 +1,2 @@
> >       Running throtl/004
> >      -Input/output error
> >       Test complete
> 
> Hi, I'm very sorry for my oversight. I have already identified the issue
> with the first patch in the patch set and replied to it myself. I will
> review more carefully next time. Please ignore the first patch (although
> the throtl test case did pass locally...)

Thanks for the explanation, no problem for me.

> 
> I'd like to ask whether applying only the second patch can pass the
> corresponding tests in your environment?

Sure, I ran throtl/004 on the kernel v7.0-rc4 + the second patch "blk-throttle:
fix timer scheduled on wrong service_queue in tg_flush_bios" only.
Unfortunately, I still see the test case fails for scsi_debug.