RE: [Patch v2 1/6] cifs: smbd: Check for iov length on sending the last iov

From: Long Li
Date: Mon Apr 23 2018 - 15:34:45 EST


> Subject: RE: [Patch v2 1/6] cifs: smbd: Check for iov length on sending the last
> iov
>
> > -----Original Message-----
> > From: linux-kernel-owner@xxxxxxxxxxxxxxx
> > <linux-kernel-owner@xxxxxxxxxxxxxxx> On Behalf Of Long Li
> > Sent: Tuesday, April 17, 2018 12:17 PM
> > To: Steve French <sfrench@xxxxxxxxx>; linux-cifs@xxxxxxxxxxxxxxx;
> > samba- technical@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx;
> > linux-rdma@xxxxxxxxxxxxxxx
> > Cc: Long Li <longli@xxxxxxxxxxxxx>; stable@xxxxxxxxxxxxxxx
> > Subject: [Patch v2 1/6] cifs: smbd: Check for iov length on sending
> > the last iov
> >
> > From: Long Li <longli@xxxxxxxxxxxxx>
> >
> > When sending the last iov that breaks into smaller buffers to fit the
> > transfer size, it's necessary to check if this is the last iov.
> >
> > If this is the latest iov, stop and proceed to send pages.
> >
> > Signed-off-by: Long Li <longli@xxxxxxxxxxxxx>
> > Cc: stable@xxxxxxxxxxxxxxx
>
> Reviewed-by: Michael Kelley <mikelley@xxxxxxxxxxxxx>
>
> But a question unrelated to this change arose during my review: At the
> beginning and end of smbd_send(), the field smbd_send_pending is
> incremented and decremented, respectively. The increment/decrement
> are not done as atomic operations. Is this code guaranteed to be single
> threaded? If not, the count could become corrupted, and
> smbd_destroy_rdma_work(), which waits for the count to become zero,
> could hang. A similar question applies to smbd_recv_pending in smbd_recv().

It is safe. The transport sending path is locked by TCP_Server_Info->srv_mutex.

The transport receiving path is single threaded.

>
> > ---
> > fs/cifs/smbdirect.c | 2 ++
> > 1 file changed, 2 insertions(+)
> >
> > diff --git a/fs/cifs/smbdirect.c b/fs/cifs/smbdirect.c index
> > 90e673c..b5c6c0d 100644
> > --- a/fs/cifs/smbdirect.c
> > +++ b/fs/cifs/smbdirect.c
> > @@ -2197,6 +2197,8 @@ int smbd_send(struct smbd_connection *info,
> struct smb_rqst *rqst)
> > goto done;
> > }
> > i++;
> > + if (i == rqst->rq_nvec)
> > + break;
> > }
> > start = i;
> > buflen = 0;
> > --
> > 2.7.4