[v2,0/2] Fix potential kernel panic when increase hardware queue

Message

Weiping Zhang April 4, 2020, 1:35 p.m. UTC

Hi Jens,

This patchset fix a potential kernel panic when increase more hardware
queue at runtime.

Patch1 fix a seperate issue, since patch2 depends on it, so I send a
new patchset.

Change since V1:
 * Add second patch to fix kernel panic when update hardware queue

Weiping Zhang (2):
  block: save previous hardware queue count before udpate
  block: alloc map and request for new hardware queue

 block/blk-mq.c | 29 +++++++++++++++++++++++------
 1 file changed, 23 insertions(+), 6 deletions(-)

Comments

Bart Van Assche April 4, 2020, 5:21 p.m. UTC | #1

On 2020-04-04 06:35, Weiping Zhang wrote:
> This patchset fix a potential kernel panic when increase more hardware
> queue at runtime.
> 
> Patch1 fix a seperate issue, since patch2 depends on it, so I send a
> new patchset.
> 
> Change since V1:
>  * Add second patch to fix kernel panic when update hardware queue
> 
> Weiping Zhang (2):
>   block: save previous hardware queue count before udpate
>   block: alloc map and request for new hardware queue

On top of which kernel version have these patches been prepared and
tested? v5.5, v5.6, Jens' for-next branch or perhaps yet another kernel
version? I'm asking this since recently a fix for
blk_mq_realloc_hw_ctxs() has been accepted in Jens' tree.

Thanks,

Bart.

Weiping Zhang April 5, 2020, 2:01 a.m. UTC | #2

Bart Van Assche <[e-mail]> 于2020年4月5日周日 上午1:22写道：
>
> On 2020-04-04 06:35, Weiping Zhang wrote:
> > This patchset fix a potential kernel panic when increase more hardware
> > queue at runtime.
> >
> > Patch1 fix a seperate issue, since patch2 depends on it, so I send a
> > new patchset.
> >
> > Change since V1:
> >  * Add second patch to fix kernel panic when update hardware queue
> >
> > Weiping Zhang (2):
> >   block: save previous hardware queue count before udpate
> >   block: alloc map and request for new hardware queue
>
> On top of which kernel version have these patches been prepared and
> tested? v5.5, v5.6, Jens' for-next branch or perhaps yet another kernel
> version? I'm asking this since recently a fix for
> blk_mq_realloc_hw_ctxs() has been accepted in Jens' tree.
>
Hi Bart,

It's tested on commit "4308a434e" of block-5.7 branch, this branch has
include the commit
"blk-mq: Fix a recently introduced regression in blk_mq_realloc_hw_ctxs()".

Thanks
Weiping