[linux-cifs-client] [PATCH] F_GETLK request - returning value

[Jeff Layton]

On Sat, 28 Mar 2009 20:51:08 +0300
Pavel Shilovsky <piastry at etersoft.ru> wrote:

> Sorry, my e-mail client broke patch.
> This is correct one.
> 
> diff --git a/fs/cifs/cifssmb.c b/fs/cifs/cifssmb.c
> index 8f0f86d..aa26a3a 100644
> --- a/fs/cifs/cifssmb.c
> +++ b/fs/cifs/cifssmb.c
> @@ -1882,6 +1882,16 @@ CIFSSMBPosixLock(const int xid, struct cifsTconInfo *tcon,
>                         ((char *)&pSMBr->hdr.Protocol + data_offset);
>                 if (parm_data->lock_type == cpu_to_le16(CIFS_UNLCK))
>                         pLockData->fl_type = F_UNLCK;
> +               else {
> +                       if (parm_data->lock_type == cpu_to_le16(CIFS_RDLCK))
> +                               pLockData->fl_type = F_RDLCK;
> +                       else if (parm_data->lock_type == cpu_to_le16(CIFS_WRLCK))
> +                               pLockData->fl_type = F_WRLCK;
> +

Minor nit: the cpu_to_le16 calls here should probably be
__constant_cpu_to_le16 (including the CIFS_UNLCK one).

> +                       pLockData->fl_start = parm_data->start;
> +                       pLockData->fl_end = parm_data->start + parm_data->length - 1;
> +                       pLockData->fl_pid = parm_data->pid;
> +               }
>         }
> 
>  plk_err_exit:
> diff --git a/fs/cifs/file.c b/fs/cifs/file.c
> index 6851043..92474f0 100644
> --- a/fs/cifs/file.c
> +++ b/fs/cifs/file.c
> @@ -896,8 +896,28 @@ int cifs_lock(struct file *file, int cmd, struct file_lock *pfLock)
> 
>                 } else {
>                         /* if rc == ERR_SHARING_VIOLATION ? */
> -                       rc = 0; /* do not change lock type to unlock
> -                                  since range in use */
> +                       rc = 0;
> +
> +                       if (lockType | LOCKING_ANDX_SHARED_LOCK) {
> +                               pfLock->fl_type = F_WRLCK;
> +                       } else {
> +                               rc = CIFSSMBLock(xid, pTcon, netfid, length, pfLock->fl_start,
> +                                        0, 1, lockType | LOCKING_ANDX_SHARED_LOCK, 0 /* wait flag */ );
> +                               if (rc == 0) {
> +                                       rc = CIFSSMBLock(xid, pTcon, netfid, length,
> +                                                pfLock->fl_start, 1 /* numUnlock */ ,
> +                                                0 /* numLock */ , lockType | LOCKING_ANDX_SHARED_LOCK,
> +                                                0 /* wait flag */ );
> +                                       pfLock->fl_type = F_RDLCK;
> +                                       if (rc != 0)
> +                                               cERROR(1, ("Error unlocking previously locked "
> +                                                  "range %d during test of lock", rc));
> +                                       rc = 0;
> +                               } else {
> +                                       pfLock->fl_type = F_WRLCK;
> +                                       rc = 0;
> +                               }
> +                       }
>                 }
> 

It would be nice if these checks were less invasive, but I guess we're
constrained by the CIFS protocol here. Looks fine to me, though to be
consistent with kernel coding style you should try to keep the code
within 80 columns.

-- 
Jeff Layton <jlayton at redhat.com>