2012/12/3 Pavel Shilovsky <piastry@xxxxxxxxxxx>:
> If we have a read oplock and set a read lock in it, we can't write to the
> locked area - so, filemap_fdatawrite may fail with a no information for a
> userspace application even if we request a write to non-locked area. Fix
> this by checking for brlocks, populating the page cache without marking
> affected pages dirty and directly writing to the server.
>
> Also remove CONFIG_CIFS_SMB2 ifdefs because it's suitable for both CIFS
> and SMB2 protocols.
>
> Signed-off-by: Pavel Shilovsky <piastry@xxxxxxxxxxx>
> ---
> fs/cifs/cifsfs.c | 1 +
> fs/cifs/cifsglob.h | 1 +
> fs/cifs/file.c | 82 ++++++++++++++++++++++++++++++++--------------------
> 3 files changed, 53 insertions(+), 31 deletions(-)
>
> diff --git a/fs/cifs/cifsfs.c b/fs/cifs/cifsfs.c
> index c6e32f2..210f0af 100644
> --- a/fs/cifs/cifsfs.c
> +++ b/fs/cifs/cifsfs.c
> @@ -229,6 +229,7 @@ cifs_alloc_inode(struct super_block *sb)
> cifs_set_oplock_level(cifs_inode, 0);
> cifs_inode->delete_pending = false;
> cifs_inode->invalid_mapping = false;
> + cifs_inode->leave_pages_clean = false;
> cifs_inode->vfs_inode.i_blkbits = 14; /* 2**14 = CIFS_MAX_MSGSIZE */
> cifs_inode->server_eof = 0;
> cifs_inode->uniqueid = 0;
> diff --git a/fs/cifs/cifsglob.h b/fs/cifs/cifsglob.h
> index 2ca6f7d..426a2bf 100644
> --- a/fs/cifs/cifsglob.h
> +++ b/fs/cifs/cifsglob.h
> @@ -1037,6 +1037,7 @@ struct cifsInodeInfo {
> bool clientCanCacheAll; /* read and writebehind oplock */
> bool delete_pending; /* DELETE_ON_CLOSE is set */
> bool invalid_mapping; /* pagecache is invalid */
> + bool leave_pages_clean; /* protected by i_mutex, not set pages dirty */
> unsigned long time; /* jiffies of last update of inode */
> u64 server_eof; /* current file size on server -- protected by i_lock */
> u64 uniqueid; /* server inode number */
> diff --git a/fs/cifs/file.c b/fs/cifs/file.c
> index e2fabc9..1bc6e67 100644
> --- a/fs/cifs/file.c
> +++ b/fs/cifs/file.c
> @@ -2109,7 +2109,15 @@ static int cifs_write_end(struct file *file, struct address_space *mapping,
> } else {
> rc = copied;
> pos += copied;
> - set_page_dirty(page);
> + /*
> + * When we use strict cache mode and cifs_strict_writev was run
> + * with level II oplock (indicated by leave_pages_clean field of
> + * CIFS_I(inode)), we can leave pages clean and let
> + * cifs_strict_writev send a new data to the server itself.
> + */
> + if (!CIFS_I(inode)->leave_pages_clean ||
> + !(cifs_sb->mnt_cifs_flags & CIFS_MOUNT_STRICT_IO))
> + set_page_dirty(page);
> }
>
> if (rc > 0) {
> @@ -2460,8 +2468,8 @@ ssize_t cifs_user_writev(struct kiocb *iocb, const struct iovec *iov,
> }
>
> static ssize_t
> -cifs_writev(struct kiocb *iocb, const struct iovec *iov,
> - unsigned long nr_segs, loff_t pos)
> +cifs_pagecache_writev(struct kiocb *iocb, const struct iovec *iov,
> + unsigned long nr_segs, loff_t pos, bool cache_all)
> {
> struct file *file = iocb->ki_filp;
> struct cifsFileInfo *cfile = (struct cifsFileInfo *)file->private_data;
> @@ -2483,8 +2491,12 @@ cifs_writev(struct kiocb *iocb, const struct iovec *iov,
> server->vals->exclusive_lock_type, NULL,
> CIFS_WRITE_OP)) {
> mutex_lock(&inode->i_mutex);
> + if (!cache_all)
> + cinode->leave_pages_clean = true;
> rc = __generic_file_aio_write(iocb, iov, nr_segs,
> - &iocb->ki_pos);
> + &iocb->ki_pos);
> + if (!cache_all)
> + cinode->leave_pages_clean = false;
> mutex_unlock(&inode->i_mutex);
> }
>
> @@ -2511,25 +2523,38 @@ cifs_strict_writev(struct kiocb *iocb, const struct iovec *iov,
> struct cifsFileInfo *cfile = (struct cifsFileInfo *)
> iocb->ki_filp->private_data;
> struct cifs_tcon *tcon = tlink_tcon(cfile->tlink);
> + ssize_t written;
>
> -#ifdef CONFIG_CIFS_SMB2
> - /*
> - * If we have an oplock for read and want to write a data to the file
> - * we need to store it in the page cache and then push it to the server
> - * to be sure the next read will get a valid data.
> - */
> - if (!cinode->clientCanCacheAll && cinode->clientCanCacheRead) {
> - ssize_t written;
> - int rc;
> -
> - written = generic_file_aio_write(iocb, iov, nr_segs, pos);
> - rc = filemap_fdatawrite(inode->i_mapping);
> - if (rc)
> - return (ssize_t)rc;
> -
> - return written;
> + /* we have a read oplock - need to store a data in the page cache */
> + if (cinode->clientCanCacheRead) {
> + /*
> + * We need to store clientCanCacheAll here to prevent race
> + * conditions - this value can be changed during an execution
> + * of generic_file_aio_write. For CIFS it can be changed from
> + * true to false only, but for SMB2 it can be changed both from
> + * true to false and vice versa. So, we can end up with a data
> + * stored in the cache, not marked dirty and not sent to the
> + * server if this value changes its state from false to true
> + * after cifs_write_end.
> + */
> + bool cache_all = cinode->clientCanCacheAll;
> +
> + if (cap_unix(tcon->ses) &&
> + ((cifs_sb->mnt_cifs_flags & CIFS_MOUNT_NOPOSIXBRL) == 0) &&
> + (CIFS_UNIX_FCNTL_CAP & le64_to_cpu(
> + tcon->fsUnixInfo.Capability)))
> + written = generic_file_aio_write(iocb, iov, nr_segs,
> + pos);
> + else
> + written = cifs_pagecache_writev(iocb, iov, nr_segs, pos,
> + cache_all);
> + /*
> + * Errors occured during writing or we have read+write oplock -
> + * no need to flush to the server.
> + */
> + if (written < 0 || cache_all)
> + return written;
> }
> -#endif
>
> /*
> * For non-oplocked files in strict cache mode we need to write the data
> @@ -2537,16 +2562,11 @@ cifs_strict_writev(struct kiocb *iocb, const struct iovec *iov,
> * affected pages because it may cause a error with mandatory locks on
> * these pages but not on the region from pos to ppos+len-1.
> */
> -
> - if (!cinode->clientCanCacheAll)
> - return cifs_user_writev(iocb, iov, nr_segs, pos);
> -
> - if (cap_unix(tcon->ses) &&
> - (CIFS_UNIX_FCNTL_CAP & le64_to_cpu(tcon->fsUnixInfo.Capability)) &&
> - ((cifs_sb->mnt_cifs_flags & CIFS_MOUNT_NOPOSIXBRL) == 0))
> - return generic_file_aio_write(iocb, iov, nr_segs, pos);
> -
> - return cifs_writev(iocb, iov, nr_segs, pos);
> + written = cifs_user_writev(iocb, iov, nr_segs, pos);
> + /* need to restore pos if errors occured */
> + if (written < 0)
> + iocb->ki_pos = pos;
> + return written;
> }
>
> static struct cifs_readdata *
> --
> 1.7.10.4
>
This version fixes race conditions when oplock changes its state from
level II to exclusive during generic_file_aio_write after
cifs_write_end - this end up with a new data stored in the page cache,
marked non-dirty and not sent to the server (possible for SMB2
protocol).
--
Best regards,
Pavel Shilovsky.
--
To unsubscribe from this list: send the line "unsubscribe linux-cifs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
