2010/10/26 Jeff Layton <jlayton@xxxxxxxxxx>:
> On Tue, 26 Oct 2010 17:03:49 +0400
> Pavel Shilovsky <piastryyy@xxxxxxxxx> wrote:
>
>> Modify cifs_file_aio_write and cifs_write_end to let the client works with
>> strict cache mode.
>>
>
> Not very descriptive of the logic here. Care to explain why you changed
> things the way you did?
I have explained this logic in the patch number 1 that adds
strictcache mount option (it includes README changes).
>
>> Signed-off-by: Pavel Shilovsky <piastryyy@xxxxxxxxx>
>> ---
>> fs/cifs/cifsfs.c | 35 ++++++++++++++++++++++++++++++-----
>> fs/cifs/file.c | 23 ++++++++++++++++++++++-
>> 2 files changed, 52 insertions(+), 6 deletions(-)
>>
>> diff --git a/fs/cifs/cifsfs.c b/fs/cifs/cifsfs.c
>> index 21e0f47..85042e4 100644
>> --- a/fs/cifs/cifsfs.c
>> +++ b/fs/cifs/cifsfs.c
>> @@ -603,12 +603,37 @@ static ssize_t cifs_file_aio_read(struct kiocb
>> *iocb, const struct iovec *iov,
>> static ssize_t cifs_file_aio_write(struct kiocb *iocb, const struct iovec *iov,
>> unsigned long nr_segs, loff_t pos)
>> {
>> - struct inode *inode = iocb->ki_filp->f_path.dentry->d_inode;
>> - ssize_t written;
>> + struct inode *inode;
>> + struct cifs_sb_info *cifs_sb;
>> + ssize_t written, cache_written;
>> + loff_t saved_pos;
>> +
>> + inode = iocb->ki_filp->f_path.dentry->d_inode;
>> +
>> + if (CIFS_I(inode)->clientCanCacheAll)
>> + return generic_file_aio_write(iocb, iov, nr_segs, pos);
>> +
>> + cifs_sb = CIFS_SB(iocb->ki_filp->f_path.dentry->d_sb);
>> +
>> + if ((cifs_sb->mnt_cifs_flags & CIFS_MOUNT_STRICT_IO) == 0) {
>> + written = generic_file_aio_write(iocb, iov, nr_segs, pos);
>> + filemap_write_and_wait(inode->i_mapping);
>> + return written;
>> + }
>
> That doesn't look right. CIFS_MOUNT_STRICT_IO is false. You're
> doing a "normal" aio write (fine), and then calling
> filemap_write_and_wait to sync it out? Why isn't
> filemap_fdatawrite sufficient in the non-strict case?
>
> Also, filemap_write_and_wait can return an error and you're
> ignoring it here.
>
It's the same we have now - so, nothing changes. I try not to affect
existing mode and only to add changes in the strict mode.
>> +
>> + saved_pos = pos;
>> + written = cifs_user_write(iocb->ki_filp, iov->iov_base,
>> + iov->iov_len, &pos);
>> +
>> + if (written > 0) {
>> + cache_written = generic_file_aio_write(iocb, iov,
>> + nr_segs, saved_pos);
>> + if (cache_written != written)
>> + cERROR(1, "Cache written and server written data "
>> + "lengths are different");
>> + } else
>> + iocb->ki_pos = pos;
>>
> ^^^^^
> What exactly is this doing? It looks like you're writing the
> same data to the server twice?
This code writes the data in cifs_user_write but not in
generic_file_aio_write - the client doesn't write anything or sets
pages dirty (to let writepages flush it) in cifs_write_end. So, that's
why this code writes to the server ony once.
>
>> - written = generic_file_aio_write(iocb, iov, nr_segs, pos);
>> - if (!CIFS_I(inode)->clientCanCacheAll)
>> - filemap_fdatawrite(inode->i_mapping);
>> return written;
>> }
>>
>> diff --git a/fs/cifs/file.c b/fs/cifs/file.c
>> index 02a045e..a4d4b3a 100644
>> --- a/fs/cifs/file.c
>> +++ b/fs/cifs/file.c
>> @@ -1578,11 +1578,31 @@ static int cifs_write_end(struct file *file,
>> struct address_space *mapping,
>> struct page *page, void *fsdata)
>> {
>> int rc;
>> - struct inode *inode = mapping->host;
>> + struct inode *inode;
>> + struct cifs_sb_info *cifs_sb;
>> +
>> + inode = mapping->host;
>> + cifs_sb = CIFS_SB(inode->i_sb);
>>
>> cFYI(1, "write_end for page %p from pos %lld with %d bytes",
>> page, pos, copied);
>>
>> + if (cifs_sb->mnt_cifs_flags & CIFS_MOUNT_STRICT_IO) {
>> + rc = copied;
>> + pos += copied;
>> +
>> + if (CIFS_I(inode)->clientCanCacheAll) {
>> + SetPageUptodate(page);
>> + set_page_dirty(page);
>> + }
>> +
>> + /* if we don't have an exclusive oplock the page data was
>> + previously written to the server in cifs_file_aio_write,
>> + so we don't need to do it again - goto exit */
>> +
>> + goto exit;
>> + }
>> +
>
> Oof. So you're only conditionally setting the uptodate and dirty bits
> on the pages after dirtying them? What exactly is going to flush these
> out to the server?
At first I am going to set only dirty flag but the kernel proves the
error in __set_page_dirty_nobuffers on the line:
WARN_ON_ONCE(!PagePrivate(page) && !PageUptodate(page));
and I add setting the page up to date. We flush the data to the server
through writepages (that flushs all dirty pages).
>
>> if (PageChecked(page)) {
>> if (copied == len)
>> SetPageUptodate(page);
>> @@ -1614,6 +1634,7 @@ static int cifs_write_end(struct file *file,
>> struct address_space *mapping,
>> set_page_dirty(page);
>> }
>>
>> +exit:
>> if (rc > 0) {
>> spin_lock(&inode->i_lock);
>> if (pos > inode->i_size)
--
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
