diff options
author | Eric Van Hensbergen <ericvh@ericvh-desktop.austin.ibm.com> | 2007-10-17 21:31:07 +0200 |
---|---|---|
committer | Eric Van Hensbergen <ericvh@ericvh-desktop.austin.ibm.com> | 2007-10-17 21:31:07 +0200 |
commit | 50fd8010673b770f6489c9ee96680c204aefa84a (patch) | |
tree | 734686c1c5be427d72665f2ea63537886717f157 /fs/9p/vfs_file.c | |
parent | 9p: attach-per-user (diff) | |
download | linux-50fd8010673b770f6489c9ee96680c204aefa84a.tar.xz linux-50fd8010673b770f6489c9ee96680c204aefa84a.zip |
9p: soften invalidation in loose_mode
Loose mode in 9p utilizes the page cache without respecting coherency with
the server. Any writes previously invaldiated the entire mapping for a file.
This patch softens the behavior to only invalidate the region of the actual
write.
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>
Diffstat (limited to 'fs/9p/vfs_file.c')
-rw-r--r-- | fs/9p/vfs_file.c | 6 |
1 files changed, 4 insertions, 2 deletions
diff --git a/fs/9p/vfs_file.c b/fs/9p/vfs_file.c index 716691689fd5..ba4b1caa9c43 100644 --- a/fs/9p/vfs_file.c +++ b/fs/9p/vfs_file.c @@ -162,15 +162,17 @@ v9fs_file_write(struct file *filp, const char __user * data, fid = filp->private_data; ret = p9_client_uwrite(fid, data, *offset, count); - if (ret > 0) + if (ret > 0) { + invalidate_inode_pages2_range(inode->i_mapping, *offset, + *offset+ret); *offset += ret; + } if (*offset > inode->i_size) { inode->i_size = *offset; inode->i_blocks = (inode->i_size + 512 - 1) >> 9; } - invalidate_inode_pages2(inode->i_mapping); return ret; } |