Re: [PATCH] cifs: add credits from unmatched responses/messages

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 






----- Original Message -----
> From: "Pavel Shilovsky" <piastryyy@xxxxxxxxx>
> To: "Ronnie Sahlberg" <lsahlber@xxxxxxxxxx>
> Cc: "linux-cifs" <linux-cifs@xxxxxxxxxxxxxxx>, "Steve French" <smfrench@xxxxxxxxx>, "Pavel Shilovsky"
> <pshilov@xxxxxxxxxxxxx>
> Sent: Wednesday, 6 February, 2019 4:28:38 AM
> Subject: Re: [PATCH] cifs: add credits from unmatched responses/messages
> 
> пн, 4 февр. 2019 г. в 18:58, Ronnie Sahlberg <lsahlber@xxxxxxxxxx>:
> >
> > We should add any credits granted to us from unmatched server responses.
> >
> > Signed-off-by: Ronnie Sahlberg <lsahlber@xxxxxxxxxx>
> > ---
> >  fs/cifs/connect.c  | 22 ++++++++++++++++++++++
> >  fs/cifs/smb2misc.c |  7 -------
> >  2 files changed, 22 insertions(+), 7 deletions(-)
> >
> > diff --git a/fs/cifs/connect.c b/fs/cifs/connect.c
> > index 431ca65b71e7..787d2c39b6c0 100644
> > --- a/fs/cifs/connect.c
> > +++ b/fs/cifs/connect.c
> > @@ -1064,6 +1064,26 @@ cifs_handle_standard(struct TCP_Server_Info *server,
> > struct mid_q_entry *mid)
> >         return 0;
> >  }
> >
> > +static void
> > +smb2_add_credits_from_hdr(char *buffer, struct TCP_Server_Info *server)
> > +{
> > +       struct smb2_sync_hdr *shdr = (struct smb2_sync_hdr *)buffer;
> > +
> > +       /*
> > +        * SMB1 does not use credits.
> > +        */
> > +       if (server->vals->header_preamble_size)
> > +               return;
> > +
> > +       if (shdr->CreditRequest) {
> > +               spin_lock(&server->req_lock);
> > +               server->credits += le16_to_cpu(shdr->CreditRequest);
> > +               spin_unlock(&server->req_lock);
> > +               wake_up(&server->request_q);
> > +       }
> > +}
> > +
> > +
> >  static int
> >  cifs_demultiplex_thread(void *p)
> >  {
> > @@ -1193,6 +1213,7 @@ cifs_demultiplex_thread(void *p)
> >                         } else if (server->ops->is_oplock_break &&
> >                                    server->ops->is_oplock_break(bufs[i],
> >                                                                 server)) {
> > +                               smb2_add_credits_from_hdr(bufs[i], server);
> >                                 cifs_dbg(FYI, "Received oplock break\n");
> >                         } else {
> >                                 cifs_dbg(VFS, "No task to wake, unknown
> >                                 frame "
> > @@ -1204,6 +1225,7 @@ cifs_demultiplex_thread(void *p)
> >                                 if (server->ops->dump_detail)
> >                                         server->ops->dump_detail(bufs[i],
> >                                                                  server);
> > +                               smb2_add_credits_from_hdr(bufs[i], server);
> >                                 cifs_dump_mids(server);
> >  #endif /* CIFS_DEBUG2 */
> >                         }
> > diff --git a/fs/cifs/smb2misc.c b/fs/cifs/smb2misc.c
> > index 7b8b58fb4d3f..6a9c47541c53 100644
> > --- a/fs/cifs/smb2misc.c
> > +++ b/fs/cifs/smb2misc.c
> > @@ -648,13 +648,6 @@ smb2_is_valid_oplock_break(char *buffer, struct
> > TCP_Server_Info *server)
> >         if (rsp->sync_hdr.Command != SMB2_OPLOCK_BREAK)
> >                 return false;
> >
> > -       if (rsp->sync_hdr.CreditRequest) {
> > -               spin_lock(&server->req_lock);
> > -               server->credits +=
> > le16_to_cpu(rsp->sync_hdr.CreditRequest);
> > -               spin_unlock(&server->req_lock);
> > -               wake_up(&server->request_q);
> > -       }
> > -
> >         if (rsp->StructureSize !=
> >                                 smb2_rsp_struct_sizes[SMB2_OPLOCK_BREAK_HE])
> >                                 {
> >                 if (le16_to_cpu(rsp->StructureSize) == 44)
> > --
> > 2.13.6
> >
> 
> So, this will count credits not only for oplock/lease breaks but for
> any unmatched response even if it is unexpected on a wire. Probably
> won't hurt to do so. Have you ever see any evidence of unexpected
> responses during testing? Those cases should be investigated at least.

I have seen the message "No task to wake" occasionally but do not have a way to
trigger/reproduce it yet.


> 
> Reviewed-by: Pavel Shilovsky <pshilov@xxxxxxxxxxxxx>
> 
> --
> Best regards,
> Pavel Shilovsky
> 



[Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux