CIFS: Fix write after setting a read lock for read oplock files

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 populating the page cache without marking affected pages dirty
after a successful write directly 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@etersoft.ru>
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Steve French <smfrench@gmail.com>

diff --git a/fs/cifs/cifsfs.c b/fs/cifs/cifsfs.c
index c6e32f22fbd3c93e476ddb5d4e19fb53a95c4832..210f0af83fc449f8667f6482558dda9d5a060654 100644 (file)
--- 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 dfab450a191ea3a61eec18da5b4f8bd2d5477535..aea1eec649116a3891ce00c9ff6d073d7183fc6a 100644 (file)
--- a/fs/cifs/cifsglob.h
+++ b/fs/cifs/cifsglob.h
@@ -1030,6 +1030,7 @@ 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 1b322d041f1ef24bcee43b1206fc3630cf40d4b6..0a6677ba212b11ef73fbf83465c32f6ffe4fa2c1 100644 (file)
--- a/fs/cifs/file.c
+++ b/fs/cifs/file.c
@@ -2103,7 +2103,15 @@ 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 - cifs_strict_writev
+                * sent the 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) {
@@ -2454,8 +2462,8 @@ 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_ex)
 {
        struct file *file = iocb->ki_filp;
        struct cifsFileInfo *cfile = (struct cifsFileInfo *)file->private_data;
@@ -2477,8 +2485,12 @@ 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_ex)
+                       cinode->leave_pages_clean = true;
                rc = __generic_file_aio_write(iocb, iov, nr_segs,
-                                              &iocb->ki_pos);
+                                             &iocb->ki_pos);
+               if (!cache_ex)
+                       cinode->leave_pages_clean = false;
                mutex_unlock(&inode->i_mutex);
        }
 
@@ -2505,42 +2517,62 @@ 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);
-
-#ifdef CONFIG_CIFS_SMB2
+       ssize_t written, written2;
        /*
-        * 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.
+        * 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.
         */
-       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;
+       bool cache_ex = cinode->clientCanCacheAll;
+       bool cache_read = cinode->clientCanCacheRead;
+       int rc;
+       loff_t saved_pos;
 
-               return written;
+       if (cache_ex) {
+               if (cap_unix(tcon->ses) &&
+                   ((cifs_sb->mnt_cifs_flags & CIFS_MOUNT_NOPOSIXBRL) == 0) &&
+                   (CIFS_UNIX_FCNTL_CAP & le64_to_cpu(
+                                               tcon->fsUnixInfo.Capability)))
+                       return generic_file_aio_write(iocb, iov, nr_segs, pos);
+               return cifs_pagecache_writev(iocb, iov, nr_segs, pos, cache_ex);
        }
-#endif
 
        /*
-        * For non-oplocked files in strict cache mode we need to write the data
-        * to the server exactly from the pos to pos+len-1 rather than flush all
-        * 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.
+        * For files without exclusive oplock in strict cache mode we need to
+        * write the data to the server exactly from the pos to pos+len-1 rather
+        * than flush all 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.
         */
+       written = cifs_user_writev(iocb, iov, nr_segs, pos);
+       if (!cache_read || written <= 0)
+               return written;
 
-       if (!cinode->clientCanCacheAll)
-               return cifs_user_writev(iocb, iov, nr_segs, pos);
-
+       saved_pos = iocb->ki_pos;
+       iocb->ki_pos = pos;
+       /* we have a read oplock - need to store a data in the page cache */
        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);
+           ((cifs_sb->mnt_cifs_flags & CIFS_MOUNT_NOPOSIXBRL) == 0) &&
+           (CIFS_UNIX_FCNTL_CAP & le64_to_cpu(
+                                       tcon->fsUnixInfo.Capability)))
+               written2 = generic_file_aio_write(iocb, iov, nr_segs, pos);
+       else
+               written2 = cifs_pagecache_writev(iocb, iov, nr_segs, pos,
+                                                cache_ex);
+       /* errors occured during writing - invalidate the page cache */
+       if (written2 < 0) {
+               rc = cifs_invalidate_mapping(inode);
+               if (rc)
+                       written = (ssize_t)rc;
+               else
+                       iocb->ki_pos = saved_pos;
+       }
+       return written;
 }
 
 static struct cifs_readdata *