kill BMAPI_DEVICE

There is no reason to go into the iomap machinery just to get the right block device for an inode. Instead look at the realtime flag in the inode and grab the right device from the mount structure.

I created a new helper, xfs_find_bdev_for_inode instead of opencoding it because I plan to use it in other places in the future.

Signed-off-by Christoph Hellwig <[email protected]>

Index: linux-2.6-xfs/fs/xfs/linux-2.6/xfs_aops.c
===================================================================
--- linux-2.6-xfs.orig/fs/xfs/linux-2.6/xfs_aops.c 2007-08-23 14:54:
01.000000000 +0200
+++ linux-2.6-xfs/fs/xfs/linux-2.6/xfs_aops.c 2007-08-23 14:59:
55.000000000 +0200

fs/xfs/linux-2.6/xfs_aops.c +16 -11

@@ -107,6 +107,18 @@ xfs_page_trace(
 #define xfs_page_trace(tag, inode, page, pgoff)
 #endif
 
+STATIC struct block_device *
+xfs_find_bdev_for_inode(
+       struct xfs_inode        *ip)
+{
+       struct xfs_mount        *mp = ip->i_mount;
+
+       if (ip->i_d.di_flags & XFS_DIFLAG_REALTIME)
+               return mp->m_rtdev_targp->bt_bdev;
+       else
+               return mp->m_ddev_targp->bt_bdev;
+}
+
 /*
  * Schedule IO completion handling on a xfsdatad if this was
  * the final hold on this ioend. If we are asked to wait,

@@ -1476,28 +1488,21 @@ xfs_vm_direct_IO(
 {
        struct file     *file = iocb->ki_filp;
        struct inode    *inode = file->f_mapping->host;
-       xfs_iomap_t     iomap;
-       int             maps = 1;
-       int             error;
+       struct block_device *bdev;
        ssize_t         ret;
 
-       error = xfs_bmap(XFS_I(inode), offset, 0,
-                               BMAPI_DEVICE, &iomap, &maps);
-       if (error)
-               return -error;
+       bdev = xfs_find_bdev_for_inode(XFS_I(inode));
 
        if (rw == WRITE) {
                iocb->private = xfs_alloc_ioend(inode, IOMAP_UNWRITTEN);
                ret = blockdev_direct_IO_own_locking(rw, iocb, inode,
-                       iomap.iomap_target->bt_bdev,
-                       iov, offset, nr_segs,
+                       bdev, iov, offset, nr_segs,
                        xfs_get_blocks_direct,
                        xfs_end_io_direct);
        } else {
                iocb->private = xfs_alloc_ioend(inode, IOMAP_READ);
                ret = blockdev_direct_IO_no_locking(rw, iocb, inode,
-                       iomap.iomap_target->bt_bdev,
-                       iov, offset, nr_segs,
+                       bdev, iov, offset, nr_segs,
                        xfs_get_blocks_direct,
                        xfs_end_io_direct);
        }

Index: linux-2.6-xfs/fs/xfs/xfs_iomap.c
===================================================================
--- linux-2.6-xfs.orig/fs/xfs/xfs_iomap.c 2007-08-23 14:39:
45.000000000 +0200
+++ linux-2.6-xfs/fs/xfs/xfs_iomap.c 2007-08-23 14:59:55.000000000+0200

fs/xfs/xfs_iomap.c +1 -1

@@ -193,7 +193,7 @@ xfs_iomap(
 
        switch (flags &
                (BMAPI_READ | BMAPI_WRITE | BMAPI_ALLOCATE |
-                BMAPI_UNWRITTEN | BMAPI_DEVICE)) {
+                BMAPI_UNWRITTEN)) {
        case BMAPI_READ:
                xfs_iomap_enter_trace(XFS_IOMAP_READ_ENTER, io, offset,
 count);

                lockmode = XFS_LCK_MAP_SHARED(mp, io);

fs/xfs/xfs_iomap.c +0 -7

@@ -220,13 +220,6 @@ xfs_iomap(
                break;
        case BMAPI_UNWRITTEN:
                goto phase2;
-       case BMAPI_DEVICE:
-               lockmode = XFS_LCK_MAP_SHARED(mp, io);
-               iomapp->iomap_target = io->io_flags & XFS_IOCORE_RT ?
-                       mp->m_rtdev_targp : mp->m_ddev_targp;
-               error = 0;
-               *niomaps = 1;
-               goto out;
        default:
                BUG();
        }

Index: linux-2.6-xfs/fs/xfs/xfs_iomap.h
===================================================================
--- linux-2.6-xfs.orig/fs/xfs/xfs_iomap.h 2007-08-23 14:39:
45.000000000 +0200
+++ linux-2.6-xfs/fs/xfs/xfs_iomap.h 2007-08-23 14:59:55.000000000+0200

(quoted diff) +0 -0

@@ -43,7 +43,6 @@ typedef enum {
        BMAPI_MMAP = (1 << 6),          /* allocate for mmap write */
        BMAPI_SYNC = (1 << 7),          /* sync write to flush delalloc

space */
BMAPI_TRYLOCK = (1 << 8), /* non-blocking request */
- BMAPI_DEVICE = (1 << 9), /* we only want to know the device
*/
} bmapi_flags_t;

XFS_IOCORE_RT | XFS_DIFLAG_REALTIME can be set from an ioctl (xfs_setattr). A directIO without holding ILOCK
in shared in mode can read a wrong value of di_flag for real time decision. As a result, we may pass in-correct device
during directIO as the proposed xfs_find_bdev_for_inode doesn't hold any lock in reading the flags. It is not based
on iocore flags as well.

On a secondary note, XFS_IOCORE_RT was set without holding iolock which seems to an issue. I tend to leave
xfs_bmapi to decide BMAPI_DEVICE to xfs_iomap.

What is the reason why this has to be seperated?

Quoted 1 line · Bhagi rathi

XFS_IOCORE_RT | XFS_DIFLAG_REALTIME can be set from an ioctl

(xfs_setattr).

Quoted 2 lines

A directIO without holding ILOCK
in shared in mode can read a wrong value of di_flag for real time

decision.

Quoted 5 lines

As a result, we may pass in-correct device
during directIO as the proposed xfs_find_bdev_for_inode doesn't hold any lock in reading the flags. It is not based
on iocore flags as well.

Quoted 3 lines

On a secondary note, XFS_IOCORE_RT was set without holding iolock which seems to an issue. I tend to leave
xfs_bmapi to decide BMAPI_DEVICE to xfs_iomap.

The previous locking doesn't help anything - if the value changes during the direct I/O process we are in trouble anyway. Fortunately we always hold the iolock in shared mode over any direct I/O, so taking this lock in ->setattr will fix this pre-existing issue.

What is the reason why this has to be seperated?

Because it does not belong into xfs_iomap. While there is at least some common code between BMAPI_READ, BMAPI_WRITE and BMAPI_ALLOCATE calls so sharing that makes some sense it does not at all for the other two which this patch separates out in preparation of bigger changes in this area.