mbox series

[0/7] rwsem enhancement patches for 5.10

Message ID 20220115005945.2125174-1-jaegeuk@google.com
Headers show
Series rwsem enhancement patches for 5.10 | expand

Message

Jaegeuk Kim Jan. 15, 2022, 12:59 a.m. UTC
Per discussion [1], can we merge these patches in 5.10 first?

[1] https://lore.kernel.org/linux-f2fs-devel/CAEe=Sx=6FCvrp_6x2Bqp3YTzep2s=aWdCmP29g7+sGCWkpNvkg@mail.gmail.com/T/#t

Peter Zijlstra (3):
  locking/rwsem: Better collate rwsem_read_trylock()
  locking/rwsem: Introduce rwsem_write_trylock()
  locking/rwsem: Fold __down_{read,write}*()

Waiman Long (4):
  locking/rwsem: Pass the current atomic count to
    rwsem_down_read_slowpath()
  locking/rwsem: Prevent potential lock starvation
  locking/rwsem: Enable reader optimistic lock stealing
  locking/rwsem: Remove reader optimistic spinning

 kernel/locking/lock_events_list.h |   6 +-
 kernel/locking/rwsem.c            | 359 +++++++++---------------------
 2 files changed, 106 insertions(+), 259 deletions(-)

Comments

Waiman Long Jan. 16, 2022, 1:42 a.m. UTC | #1
On 1/14/22 19:59, Jaegeuk Kim wrote:
> Per discussion [1], can we merge these patches in 5.10 first?
>
> [1] https://lore.kernel.org/linux-f2fs-devel/CAEe=Sx=6FCvrp_6x2Bqp3YTzep2s=aWdCmP29g7+sGCWkpNvkg@mail.gmail.com/T/#t
>
> Peter Zijlstra (3):
>    locking/rwsem: Better collate rwsem_read_trylock()
>    locking/rwsem: Introduce rwsem_write_trylock()
>    locking/rwsem: Fold __down_{read,write}*()
>
> Waiman Long (4):
>    locking/rwsem: Pass the current atomic count to
>      rwsem_down_read_slowpath()
>    locking/rwsem: Prevent potential lock starvation
>    locking/rwsem: Enable reader optimistic lock stealing
>    locking/rwsem: Remove reader optimistic spinning
>
>   kernel/locking/lock_events_list.h |   6 +-
>   kernel/locking/rwsem.c            | 359 +++++++++---------------------
>   2 files changed, 106 insertions(+), 259 deletions(-)

Have you actually tested it in your testing environment to verify that 
these patches can address the problem? I suspect they will help, but it 
will be nice if you also include your test results.

Cheers,
Longman
Greg Kroah-Hartman Jan. 16, 2022, 8:56 a.m. UTC | #2
On Fri, Jan 14, 2022 at 04:59:39PM -0800, Jaegeuk Kim wrote:
> Per discussion [1], can we merge these patches in 5.10 first?
> 
> [1] https://lore.kernel.org/linux-f2fs-devel/CAEe=Sx=6FCvrp_6x2Bqp3YTzep2s=aWdCmP29g7+sGCWkpNvkg@mail.gmail.com/T/#t

I do not understand, what "discussion" exactly is there that requires
these changes for older kernels?

What bug is this fixing?

> Peter Zijlstra (3):
>   locking/rwsem: Better collate rwsem_read_trylock()
>   locking/rwsem: Introduce rwsem_write_trylock()
>   locking/rwsem: Fold __down_{read,write}*()
> 
> Waiman Long (4):
>   locking/rwsem: Pass the current atomic count to
>     rwsem_down_read_slowpath()
>   locking/rwsem: Prevent potential lock starvation
>   locking/rwsem: Enable reader optimistic lock stealing
>   locking/rwsem: Remove reader optimistic spinning
> 
>  kernel/locking/lock_events_list.h |   6 +-
>  kernel/locking/rwsem.c            | 359 +++++++++---------------------
>  2 files changed, 106 insertions(+), 259 deletions(-)

And you are positive that there are no follow-on patches needed for
these core changes?  How were they tested?  What now works that did not
work in 5.10?  Why just 5.10?  What about all older kernels?

We need a lot more information here, sorry.

greg k-h
Greg Kroah-Hartman Jan. 24, 2022, 3:08 p.m. UTC | #3
On Sun, Jan 16, 2022 at 09:56:22AM +0100, Greg KH wrote:
> On Fri, Jan 14, 2022 at 04:59:39PM -0800, Jaegeuk Kim wrote:
> > Per discussion [1], can we merge these patches in 5.10 first?
> > 
> > [1] https://lore.kernel.org/linux-f2fs-devel/CAEe=Sx=6FCvrp_6x2Bqp3YTzep2s=aWdCmP29g7+sGCWkpNvkg@mail.gmail.com/T/#t
> 
> I do not understand, what "discussion" exactly is there that requires
> these changes for older kernels?
> 
> What bug is this fixing?
> 
> > Peter Zijlstra (3):
> >   locking/rwsem: Better collate rwsem_read_trylock()
> >   locking/rwsem: Introduce rwsem_write_trylock()
> >   locking/rwsem: Fold __down_{read,write}*()
> > 
> > Waiman Long (4):
> >   locking/rwsem: Pass the current atomic count to
> >     rwsem_down_read_slowpath()
> >   locking/rwsem: Prevent potential lock starvation
> >   locking/rwsem: Enable reader optimistic lock stealing
> >   locking/rwsem: Remove reader optimistic spinning
> > 
> >  kernel/locking/lock_events_list.h |   6 +-
> >  kernel/locking/rwsem.c            | 359 +++++++++---------------------
> >  2 files changed, 106 insertions(+), 259 deletions(-)
> 
> And you are positive that there are no follow-on patches needed for
> these core changes?  How were they tested?  What now works that did not
> work in 5.10?  Why just 5.10?  What about all older kernels?
> 
> We need a lot more information here, sorry.

Given a lack of response, I'm dropping this from my "to review" queue.
If you want these added to a stable kernel, please resend with the
requested information.

thanks,

greg k-h