Re: [PATCH v2] nvme-rdma: remove race conditions from IB signalling

2017-07-06 Thread Marta Rybczynska
> On Wed, Jun 21, 2017 at 05:14:41PM +0200, Marta Rybczynska wrote: >> I do not see this one in nvme-4.13. Can we get it in, please? >> We're seeing the races in our setup and this patch fixes it. > > I've added it. Note that your mail was whitespace damaged, so I had > to apply it manually. I u

Re: [PATCH v2] nvme-rdma: remove race conditions from IB signalling

2017-07-05 Thread Christoph Hellwig
On Wed, Jun 21, 2017 at 05:14:41PM +0200, Marta Rybczynska wrote: > I do not see this one in nvme-4.13. Can we get it in, please? > We're seeing the races in our setup and this patch fixes it. I've added it. Note that your mail was whitespace damaged, so I had to apply it manually. I used that o

Re: [PATCH v2] nvme-rdma: remove race conditions from IB signalling

2017-06-21 Thread Marta Rybczynska
> On Tue, Jun 06, 2017 at 02:59:43PM +0300, Sagi Grimberg wrote: >> Christoph, >> >> Can you place a stable tag on this (4.11+)? > > Added. I do not see this one in nvme-4.13. Can we get it in, please? We're seeing the races in our setup and this patch fixes it. Thanks, Marta

Re: [PATCH v2] nvme-rdma: remove race conditions from IB signalling

2017-06-07 Thread Christoph Hellwig
On Tue, Jun 06, 2017 at 02:59:43PM +0300, Sagi Grimberg wrote: > Christoph, > > Can you place a stable tag on this (4.11+)? Added.

Re: [PATCH v2] nvme-rdma: remove race conditions from IB signalling

2017-06-06 Thread Sagi Grimberg
Christoph, Can you place a stable tag on this (4.11+)?

[PATCH v2] nvme-rdma: remove race conditions from IB signalling

2017-06-06 Thread Marta Rybczynska
This patch improves the way the RDMA IB signalling is done by using atomic operations for the signalling variable. This avoids race conditions on sig_count. The signalling interval changes slightly and is now the largest power of two not larger than queue depth / 2. ilog() usage idea by Bart Van