On Wed, Jan 23, 2019 at 11:31 AM Pavel Shilovsky <piastryyy@xxxxxxxxx> wrote:
>
> вс, 20 янв. 2019 г. в 23:43, Ronnie Sahlberg <lsahlber@xxxxxxxxxx>:
> >
> > Signed-off-by: Ronnie Sahlberg <lsahlber@xxxxxxxxxx>
> > ---
> > fs/cifs/smb2misc.c | 7 +++++++
> > 1 file changed, 7 insertions(+)
> >
> > diff --git a/fs/cifs/smb2misc.c b/fs/cifs/smb2misc.c
> > index 6a9c47541c53..ec1372d2f3df 100644
> > --- a/fs/cifs/smb2misc.c
> > +++ b/fs/cifs/smb2misc.c
> > @@ -642,12 +642,19 @@ smb2_is_valid_oplock_break(char *buffer, struct TCP_Server_Info *server)
> > struct cifs_tcon *tcon;
> > struct cifsInodeInfo *cinode;
> > struct cifsFileInfo *cfile;
> > + struct cifs_credits credits;
> >
> > cifs_dbg(FYI, "Checking for oplock break\n");
> >
> > if (rsp->sync_hdr.Command != SMB2_OPLOCK_BREAK)
> > return false;
> >
> > + if (rsp->sync_hdr.CreditRequest) {
> > + credits.value = le16_to_cpu(rsp->sync_hdr.CreditRequest);
> > + credits.instance = server->reconnect_instance;
> > + add_credits(server, &credits, 0);
>
> add_credits() modifies server->in_flight field. Let's do something
> like the following instead:
>
> diff --git a/fs/cifs/smb2misc.c b/fs/cifs/smb2misc.c
> index 6a9c475..7b8b58f 100644
> --- a/fs/cifs/smb2misc.c
> +++ b/fs/cifs/smb2misc.c
> @@ -648,6 +648,13 @@ 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);
> + }
> +
>
> It look like it should apply to stable kernels without problems.
Sounds good, but why the wake_up()
I will resend without wake_up()
>
> --
> Best regards,
> Pavel Shilovsky
