> -----Original Message----- > From: Jason Gunthorpe [mailto:jgg@xxxxxxxx] > Sent: Thursday, March 15, 2018 3:12 PM > To: Parav Pandit <parav@xxxxxxxxxxxx> > Cc: Leon Romanovsky <leon@xxxxxxxxxx>; Doug Ledford > <dledford@xxxxxxxxxx>; RDMA mailing list <linux-rdma@xxxxxxxxxxxxxxx>; > Daniel Jurgens <danielj@xxxxxxxxxxxx>; Mark Bloch <markb@xxxxxxxxxxxx> > Subject: Re: [PATCH rdma-next v1 09/13] IB/cm: Avoid AV ah_attr overwriting > during LAP message handling > > On Thu, Mar 15, 2018 at 06:15:05PM +0000, Parav Pandit wrote: > > Hi Jason, Doug, > > > > > From: linux-rdma-owner@xxxxxxxxxxxxxxx [mailto:linux-rdma- > > > owner@xxxxxxxxxxxxxxx] On Behalf Of Parav Pandit > > > Sent: Thursday, March 15, 2018 11:37 AM > > > To: Jason Gunthorpe <jgg@xxxxxxxx>; Leon Romanovsky > > > <leon@xxxxxxxxxx> > > > Cc: Doug Ledford <dledford@xxxxxxxxxx>; RDMA mailing list <linux- > > > rdma@xxxxxxxxxxxxxxx>; Daniel Jurgens <danielj@xxxxxxxxxxxx>; Mark > > > Bloch <markb@xxxxxxxxxxxx> > > > Subject: RE: [PATCH rdma-next v1 09/13] IB/cm: Avoid AV ah_attr > > > overwriting during LAP message handling > > > > > > > > > > > > > From: Jason Gunthorpe [mailto:jgg@xxxxxxxx] > > > > Sent: Thursday, March 15, 2018 9:30 AM > > > > To: Leon Romanovsky <leon@xxxxxxxxxx> > > > > Cc: Parav Pandit <parav@xxxxxxxxxxxx>; Doug Ledford > > > > <dledford@xxxxxxxxxx>; RDMA mailing list > > > > <linux-rdma@xxxxxxxxxxxxxxx>; Daniel Jurgens > > > > <danielj@xxxxxxxxxxxx>; Mark Bloch <markb@xxxxxxxxxxxx> > > > > Subject: Re: [PATCH rdma-next v1 09/13] IB/cm: Avoid AV ah_attr > > > > overwriting during LAP message handling > > > > > > > > On Thu, Mar 15, 2018 at 10:00:17AM +0200, Leon Romanovsky wrote: > > > > > > > > > > But.. this patch is still complicated, it is hard to see if > > > > > > all the readers are safe or not.. > > > > > > > > > > It was tested, no new bugs were discovered, so it is safe to > > > > > assume that Parav's series are not breaking IB and RoCE and safe > > > > > for inclusion > > > > > > > > Are you kidding? The patch has obvious data races. > > > > > > > > Testing doesn't discover those. > > > > > > > In that case let's get ucm cleanup full plan in place where ucm is > > > removed and APIs only used by ucm also removed. > > > After that I attempt to fix this path if the code exist. > > > > So you can continue to review the rest of the series and drop this patch. > > In subsequent series too, I won't touch LAP code which doesn't have active > user. > > Sure, I was going to bump this patch anyhow.. > > The question is what should you resend since there is a bug here it seems. > > It looks to me like taking the av writes out of the lock was a mistake, so I think > the revised patch should not do av writes out of the lock.. > To my understanding this is only applicable to LAP messages and not applicable to other CM req/rsp messages. In that case we can revert older commit too and let the bug remain as it for RoCE. I do not intent to redesign the ib_cm layer for LAP flow. -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
