mbox series

[RFC,0/5] bh-safe lock removal for SUNRPC

Message ID 20190503111841.4391-1-trond.myklebust@hammerspace.com (mailing list archive)
Headers show
Series bh-safe lock removal for SUNRPC | expand

Message

Trond Myklebust May 3, 2019, 11:18 a.m. UTC
This patchset aims to remove the bh-safe locks on the client side.
At this time it should be seen as a toy/strawman effort in order to
help the community figure out whether or not there are setups out
there that are actually seeing performance bottlenecks resulting
from taking bh-safe locks inside other spinlocks.

Trond Myklebust (5):
  SUNRPC: Replace the queue timer with a delayed work function
  SUNRPC: Replace direct task wakeups from softirq context
  SUNRPC: Remove the bh-safe lock requirement on xprt->transport_lock
  SUNRPC: Remove the bh-safe lock requirement on the
    rpc_wait_queue->lock
  SUNRPC: Reduce the priority of the xprtiod queue

 include/linux/sunrpc/sched.h               |   3 +-
 include/linux/sunrpc/xprtsock.h            |   5 +
 net/sunrpc/sched.c                         |  76 +++++++++-------
 net/sunrpc/xprt.c                          |  61 ++++++-------
 net/sunrpc/xprtrdma/rpc_rdma.c             |   4 +-
 net/sunrpc/xprtrdma/svc_rdma_backchannel.c |   4 +-
 net/sunrpc/xprtrdma/svc_rdma_transport.c   |   8 +-
 net/sunrpc/xprtsock.c                      | 101 +++++++++++++++++----
 8 files changed, 168 insertions(+), 94 deletions(-)

Comments

Chuck Lever III May 6, 2019, 6:22 p.m. UTC | #1
Hi Trond-

> On May 3, 2019, at 7:18 AM, Trond Myklebust <trondmy@gmail.com> wrote:
> 
> This patchset aims to remove the bh-safe locks on the client side.
> At this time it should be seen as a toy/strawman effort in order to
> help the community figure out whether or not there are setups out
> there that are actually seeing performance bottlenecks resulting
> from taking bh-safe locks inside other spinlocks.

What kernel does this patch set apply to? I've tried both v5.0 and
v5.1, but there appear to be some changes that I'm missing. The
first patch does not apply cleanly.


> Trond Myklebust (5):
>  SUNRPC: Replace the queue timer with a delayed work function
>  SUNRPC: Replace direct task wakeups from softirq context
>  SUNRPC: Remove the bh-safe lock requirement on xprt->transport_lock
>  SUNRPC: Remove the bh-safe lock requirement on the
>    rpc_wait_queue->lock
>  SUNRPC: Reduce the priority of the xprtiod queue
> 
> include/linux/sunrpc/sched.h               |   3 +-
> include/linux/sunrpc/xprtsock.h            |   5 +
> net/sunrpc/sched.c                         |  76 +++++++++-------
> net/sunrpc/xprt.c                          |  61 ++++++-------
> net/sunrpc/xprtrdma/rpc_rdma.c             |   4 +-
> net/sunrpc/xprtrdma/svc_rdma_backchannel.c |   4 +-
> net/sunrpc/xprtrdma/svc_rdma_transport.c   |   8 +-
> net/sunrpc/xprtsock.c                      | 101 +++++++++++++++++----
> 8 files changed, 168 insertions(+), 94 deletions(-)
> 
> -- 
> 2.21.0
> 

--
Chuck Lever
Trond Myklebust May 6, 2019, 6:37 p.m. UTC | #2
On Mon, 2019-05-06 at 14:22 -0400, Chuck Lever wrote:
> Hi Trond-
> 
> > On May 3, 2019, at 7:18 AM, Trond Myklebust <trondmy@gmail.com>
> > wrote:
> > 
> > This patchset aims to remove the bh-safe locks on the client side.
> > At this time it should be seen as a toy/strawman effort in order to
> > help the community figure out whether or not there are setups out
> > there that are actually seeing performance bottlenecks resulting
> > from taking bh-safe locks inside other spinlocks.
> 
> What kernel does this patch set apply to? I've tried both v5.0 and
> v5.1, but there appear to be some changes that I'm missing. The
> first patch does not apply cleanly.
> 

It should hopefully apply on top of Anna's linux-next branch.
Chuck Lever III May 6, 2019, 8:02 p.m. UTC | #3
> On May 6, 2019, at 2:37 PM, Trond Myklebust <trondmy@hammerspace.com> wrote:
> 
> On Mon, 2019-05-06 at 14:22 -0400, Chuck Lever wrote:
>> Hi Trond-
>> 
>>> On May 3, 2019, at 7:18 AM, Trond Myklebust <trondmy@gmail.com>
>>> wrote:
>>> 
>>> This patchset aims to remove the bh-safe locks on the client side.
>>> At this time it should be seen as a toy/strawman effort in order to
>>> help the community figure out whether or not there are setups out
>>> there that are actually seeing performance bottlenecks resulting
>>> from taking bh-safe locks inside other spinlocks.
>> 
>> What kernel does this patch set apply to? I've tried both v5.0 and
>> v5.1, but there appear to be some changes that I'm missing. The
>> first patch does not apply cleanly.
>> 
> 
> It should hopefully apply on top of Anna's linux-next branch.

OK, you did mention that to me last week. Sorry for the noise.

--
Chuck Lever