static ssize_t gfs2_direct_IO(int rw, struct kiocb *iocb,
const struct iovec *iov, loff_t offset,
unsigned long nr_segs)
{
struct file *file = iocb->ki_filp;
struct inode *inode = file->f_mapping->host;
struct gfs2_inode *ip = GFS2_I(inode);
struct gfs2_holder gh;
int rv;
/*
* Deferred lock, even if its a write, since we do no allocation
* on this path. All we need change is atime, and this lock mode
* ensures that other nodes have flushed their buffered read caches
* (i.e. their page cache entries for this inode). We do not,
* unfortunately have the option of only flushing a range like
* the VFS does.
*/
gfs2_holder_init(ip->i_gl, LM_ST_DEFERRED, 0, &gh);
rv = gfs2_glock_nq(&gh);
if (rv)
return rv;
rv = gfs2_ok_for_dio(ip, rw, offset);
if (rv != 1)
goto out; /* dio not valid, fall back to buffered i/o */
rv = __blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
offset, nr_segs, gfs2_get_block_direct,
NULL, NULL, 0);
out:
gfs2_glock_dq_m(1, &gh);
gfs2_holder_uninit(&gh);
return rv;
}
/*
* xfs_file_dio_aio_write - handle direct IO writes
*
* Lock the inode appropriately to prepare for and issue a direct IO write.
* By separating it from the buffered write path we remove all the tricky to
* follow locking changes and looping.
*
* If there are cached pages or we're extending the file, we need IOLOCK_EXCL
* until we're sure the bytes at the new EOF have been zeroed and/or the cached
* pages are flushed out.
*
* In most cases the direct IO writes will be done holding IOLOCK_SHARED
* allowing them to be done in parallel with reads and other direct IO writes.
* However, if the IO is not aligned to filesystem blocks, the direct IO layer
* needs to do sub-block zeroing and that requires serialisation against other
* direct IOs to the same block. In this case we need to serialise the
* submission of the unaligned IOs so that we don't get racing block zeroing in
* the dio layer. To avoid the problem with aio, we also need to wait for
* outstanding IOs to complete so that unwritten extent conversion is completed
* before we try to map the overlapping block. This is currently implemented by
* hitting it with a big hammer (i.e. inode_dio_wait()).
*
* Returns with locks held indicated by @iolock and errors indicated by
* negative return values.
*/
STATIC ssize_t
xfs_file_dio_aio_write(
struct kiocb *iocb,
struct iov_iter *from)
{
struct file *file = iocb->ki_filp;
struct address_space *mapping = file->f_mapping;
struct inode *inode = mapping->host;
struct xfs_inode *ip = XFS_I(inode);
struct xfs_mount *mp = ip->i_mount;
ssize_t ret = 0;
int unaligned_io = 0;
int iolock;
size_t count = iov_iter_count(from);
loff_t end;
struct iov_iter data;
struct xfs_buftarg *target = XFS_IS_REALTIME_INODE(ip) ?
mp->m_rtdev_targp : mp->m_ddev_targp;
/* DIO must be aligned to device logical sector size */
if ((iocb->ki_pos | count) & target->bt_logical_sectormask)
return -EINVAL;
/* "unaligned" here means not aligned to a filesystem block */
if ((iocb->ki_pos & mp->m_blockmask) ||
((iocb->ki_pos + count) & mp->m_blockmask))
unaligned_io = 1;
/*
* We don't need to take an exclusive lock unless there page cache needs
* to be invalidated or unaligned IO is being executed. We don't need to
* consider the EOF extension case here because
* xfs_file_aio_write_checks() will relock the inode as necessary for
* EOF zeroing cases and fill out the new inode size as appropriate.
*/
if (unaligned_io || mapping->nrpages)
iolock = XFS_IOLOCK_EXCL;
else
iolock = XFS_IOLOCK_SHARED;
xfs_rw_ilock(ip, iolock);
/*
* Recheck if there are cached pages that need invalidate after we got
* the iolock to protect against other threads adding new pages while
* we were waiting for the iolock.
*/
if (mapping->nrpages && iolock == XFS_IOLOCK_SHARED) {
xfs_rw_iunlock(ip, iolock);
iolock = XFS_IOLOCK_EXCL;
xfs_rw_ilock(ip, iolock);
}
ret = xfs_file_aio_write_checks(iocb, from, &iolock);
if (ret)
goto out;
count = iov_iter_count(from);
end = iocb->ki_pos + count - 1;
/*
* See xfs_file_dio_aio_read() for why we do a full-file flush here.
*/
if (mapping->nrpages) {
ret = filemap_write_and_wait(VFS_I(ip)->i_mapping);
if (ret)
goto out;
/*
* Invalidate whole pages. This can return an error if we fail
* to invalidate a page, but this should never happen on XFS.
* Warn if it does fail.
*/
ret = invalidate_inode_pages2(VFS_I(ip)->i_mapping);
WARN_ON_ONCE(ret);
ret = 0;
}
/*
* If we are doing unaligned IO, wait for all other IO to drain,
* otherwise demote the lock if we had to flush cached pages
*/
if (unaligned_io)
inode_dio_wait(inode);
else if (iolock == XFS_IOLOCK_EXCL) {
xfs_rw_ilock_demote(ip, XFS_IOLOCK_EXCL);
iolock = XFS_IOLOCK_SHARED;
}
trace_xfs_file_direct_write(ip, count, iocb->ki_pos);
/* If this is a block-aligned directio CoW, remap immediately. */
if (xfs_is_reflink_inode(ip) && !unaligned_io) {
ret = xfs_reflink_allocate_cow_range(ip, iocb->ki_pos, count);
if (ret)
goto out;
}
data = *from;
ret = __blockdev_direct_IO(iocb, inode, target->bt_bdev, &data,
xfs_get_blocks_direct, xfs_end_io_direct_write,
NULL, DIO_ASYNC_EXTEND);
/* see generic_file_direct_write() for why this is necessary */
if (mapping->nrpages) {
invalidate_inode_pages2_range(mapping,
iocb->ki_pos >> PAGE_SHIFT,
end >> PAGE_SHIFT);
}
static ssize_t gfs2_direct_IO(int rw, struct kiocb *iocb,
const struct iovec *iov, loff_t offset,
unsigned long nr_segs)
{
struct file *file = iocb->ki_filp;
struct inode *inode = file->f_mapping->host;
struct address_space *mapping = inode->i_mapping;
struct gfs2_inode *ip = GFS2_I(inode);
struct gfs2_holder gh;
int rv;
/*
* Deferred lock, even if its a write, since we do no allocation
* on this path. All we need change is atime, and this lock mode
* ensures that other nodes have flushed their buffered read caches
* (i.e. their page cache entries for this inode). We do not,
* unfortunately have the option of only flushing a range like
* the VFS does.
*/
gfs2_holder_init(ip->i_gl, LM_ST_DEFERRED, 0, &gh);
rv = gfs2_glock_nq(&gh);
if (rv)
return rv;
rv = gfs2_ok_for_dio(ip, rw, offset);
if (rv != 1)
goto out; /* dio not valid, fall back to buffered i/o */
/*
* Now since we are holding a deferred (CW) lock at this point, you
* might be wondering why this is ever needed. There is a case however
* where we've granted a deferred local lock against a cached exclusive
* glock. That is ok provided all granted local locks are deferred, but
* it also means that it is possible to encounter pages which are
* cached and possibly also mapped. So here we check for that and sort
* them out ahead of the dio. The glock state machine will take care of
* everything else.
*
* If in fact the cached glock state (gl->gl_state) is deferred (CW) in
* the first place, mapping->nr_pages will always be zero.
*/
if (mapping->nrpages) {
loff_t lstart = offset & (PAGE_CACHE_SIZE - 1);
loff_t len = iov_length(iov, nr_segs);
loff_t end = PAGE_ALIGN(offset + len) - 1;
rv = 0;
if (len == 0)
goto out;
if (test_and_clear_bit(GIF_SW_PAGED, &ip->i_flags))
unmap_shared_mapping_range(ip->i_inode.i_mapping, offset, len);
rv = filemap_write_and_wait_range(mapping, lstart, end);
if (rv)
goto out;
if (rw == WRITE)
truncate_inode_pages_range(mapping, lstart, end);
}
rv = __blockdev_direct_IO(rw, iocb, inode, inode->i_sb->s_bdev, iov,
offset, nr_segs, gfs2_get_block_direct,
NULL, NULL, 0);
out:
gfs2_glock_dq(&gh);
gfs2_holder_uninit(&gh);
return rv;
}
STATIC ssize_t
xfs_file_dio_aio_read(
struct kiocb *iocb,
struct iov_iter *to)
{
struct address_space *mapping = iocb->ki_filp->f_mapping;
struct inode *inode = mapping->host;
struct xfs_inode *ip = XFS_I(inode);
loff_t isize = i_size_read(inode);
size_t count = iov_iter_count(to);
struct iov_iter data;
struct xfs_buftarg *target;
ssize_t ret = 0;
trace_xfs_file_direct_read(ip, count, iocb->ki_pos);
if (!count)
return 0; /* skip atime */
if (XFS_IS_REALTIME_INODE(ip))
target = ip->i_mount->m_rtdev_targp;
else
target = ip->i_mount->m_ddev_targp;
/* DIO must be aligned to device logical sector size */
if ((iocb->ki_pos | count) & target->bt_logical_sectormask) {
if (iocb->ki_pos == isize)
return 0;
return -EINVAL;
}
file_accessed(iocb->ki_filp);
/*
* Locking is a bit tricky here. If we take an exclusive lock for direct
* IO, we effectively serialise all new concurrent read IO to this file
* and block it behind IO that is currently in progress because IO in
* progress holds the IO lock shared. We only need to hold the lock
* exclusive to blow away the page cache, so only take lock exclusively
* if the page cache needs invalidation. This allows the normal direct
* IO case of no page cache pages to proceeed concurrently without
* serialisation.
*/
xfs_rw_ilock(ip, XFS_IOLOCK_SHARED);
if (mapping->nrpages) {
xfs_rw_iunlock(ip, XFS_IOLOCK_SHARED);
xfs_rw_ilock(ip, XFS_IOLOCK_EXCL);
/*
* The generic dio code only flushes the range of the particular
* I/O. Because we take an exclusive lock here, this whole
* sequence is considerably more expensive for us. This has a
* noticeable performance impact for any file with cached pages,
* even when outside of the range of the particular I/O.
*
* Hence, amortize the cost of the lock against a full file
* flush and reduce the chances of repeated iolock cycles going
* forward.
*/
if (mapping->nrpages) {
ret = filemap_write_and_wait(mapping);
if (ret) {
xfs_rw_iunlock(ip, XFS_IOLOCK_EXCL);
return ret;
}
/*
* Invalidate whole pages. This can return an error if
* we fail to invalidate a page, but this should never
* happen on XFS. Warn if it does fail.
*/
ret = invalidate_inode_pages2(mapping);
WARN_ON_ONCE(ret);
ret = 0;
}
xfs_rw_ilock_demote(ip, XFS_IOLOCK_EXCL);
}
data = *to;
ret = __blockdev_direct_IO(iocb, inode, target->bt_bdev, &data,
xfs_get_blocks_direct, NULL, NULL, 0);
if (ret >= 0) {
iocb->ki_pos += ret;
iov_iter_advance(to, ret);
}
xfs_rw_iunlock(ip, XFS_IOLOCK_SHARED);
return ret;
}
