Merge branch 'for-linus' of git://oss.sgi.com/xfs/xfs

* 'for-linus' of git://oss.sgi.com/xfs/xfs:
  xfs: fix getbmap vs mmap deadlock
  xfs: a couple getbmap cleanups
  xfs: add more checks to superblock validation
  xfs_file_last_byte() needs to acquire ilock
This commit is contained in:
Linus Torvalds 2009-05-02 16:52:50 -07:00
commit b4348f32da
3 changed files with 108 additions and 89 deletions

View file

@ -5880,7 +5880,7 @@ xfs_getbmap(
void *arg) /* formatter arg */
{
__int64_t bmvend; /* last block requested */
int error; /* return value */
int error = 0; /* return value */
__int64_t fixlen; /* length for -1 case */
int i; /* extent number */
int lock; /* lock state */
@ -5890,39 +5890,18 @@ xfs_getbmap(
int nexleft; /* # of user extents left */
int subnex; /* # of bmapi's can do */
int nmap; /* number of map entries */
struct getbmapx out; /* output structure */
struct getbmapx *out; /* output structure */
int whichfork; /* data or attr fork */
int prealloced; /* this is a file with
* preallocated data space */
int iflags; /* interface flags */
int bmapi_flags; /* flags for xfs_bmapi */
int cur_ext = 0;
mp = ip->i_mount;
iflags = bmv->bmv_iflags;
whichfork = iflags & BMV_IF_ATTRFORK ? XFS_ATTR_FORK : XFS_DATA_FORK;
/* If the BMV_IF_NO_DMAPI_READ interface bit specified, do not
* generate a DMAPI read event. Otherwise, if the DM_EVENT_READ
* bit is set for the file, generate a read event in order
* that the DMAPI application may do its thing before we return
* the extents. Usually this means restoring user file data to
* regions of the file that look like holes.
*
* The "old behavior" (from XFS_IOC_GETBMAP) is to not specify
* BMV_IF_NO_DMAPI_READ so that read events are generated.
* If this were not true, callers of ioctl( XFS_IOC_GETBMAP )
* could misinterpret holes in a DMAPI file as true holes,
* when in fact they may represent offline user data.
*/
if ((iflags & BMV_IF_NO_DMAPI_READ) == 0 &&
DM_EVENT_ENABLED(ip, DM_EVENT_READ) &&
whichfork == XFS_DATA_FORK) {
error = XFS_SEND_DATA(mp, DM_EVENT_READ, ip, 0, 0, 0, NULL);
if (error)
return XFS_ERROR(error);
}
if (whichfork == XFS_ATTR_FORK) {
if (XFS_IFORK_Q(ip)) {
if (ip->i_d.di_aformat != XFS_DINODE_FMT_EXTENTS &&
@ -5936,11 +5915,37 @@ xfs_getbmap(
ip->i_mount);
return XFS_ERROR(EFSCORRUPTED);
}
} else if (ip->i_d.di_format != XFS_DINODE_FMT_EXTENTS &&
ip->i_d.di_format != XFS_DINODE_FMT_BTREE &&
ip->i_d.di_format != XFS_DINODE_FMT_LOCAL)
return XFS_ERROR(EINVAL);
if (whichfork == XFS_DATA_FORK) {
prealloced = 0;
fixlen = 1LL << 32;
} else {
/*
* If the BMV_IF_NO_DMAPI_READ interface bit specified, do
* not generate a DMAPI read event. Otherwise, if the
* DM_EVENT_READ bit is set for the file, generate a read
* event in order that the DMAPI application may do its thing
* before we return the extents. Usually this means restoring
* user file data to regions of the file that look like holes.
*
* The "old behavior" (from XFS_IOC_GETBMAP) is to not specify
* BMV_IF_NO_DMAPI_READ so that read events are generated.
* If this were not true, callers of ioctl(XFS_IOC_GETBMAP)
* could misinterpret holes in a DMAPI file as true holes,
* when in fact they may represent offline user data.
*/
if (DM_EVENT_ENABLED(ip, DM_EVENT_READ) &&
!(iflags & BMV_IF_NO_DMAPI_READ)) {
error = XFS_SEND_DATA(mp, DM_EVENT_READ, ip,
0, 0, 0, NULL);
if (error)
return XFS_ERROR(error);
}
if (ip->i_d.di_format != XFS_DINODE_FMT_EXTENTS &&
ip->i_d.di_format != XFS_DINODE_FMT_BTREE &&
ip->i_d.di_format != XFS_DINODE_FMT_LOCAL)
return XFS_ERROR(EINVAL);
if (xfs_get_extsz_hint(ip) ||
ip->i_d.di_flags & (XFS_DIFLAG_PREALLOC|XFS_DIFLAG_APPEND)){
prealloced = 1;
@ -5949,42 +5954,41 @@ xfs_getbmap(
prealloced = 0;
fixlen = ip->i_size;
}
} else {
prealloced = 0;
fixlen = 1LL << 32;
}
if (bmv->bmv_length == -1) {
fixlen = XFS_FSB_TO_BB(mp, XFS_B_TO_FSB(mp, fixlen));
bmv->bmv_length = MAX( (__int64_t)(fixlen - bmv->bmv_offset),
(__int64_t)0);
} else if (bmv->bmv_length < 0)
return XFS_ERROR(EINVAL);
if (bmv->bmv_length == 0) {
bmv->bmv_length =
max_t(__int64_t, fixlen - bmv->bmv_offset, 0);
} else if (bmv->bmv_length == 0) {
bmv->bmv_entries = 0;
return 0;
} else if (bmv->bmv_length < 0) {
return XFS_ERROR(EINVAL);
}
nex = bmv->bmv_count - 1;
if (nex <= 0)
return XFS_ERROR(EINVAL);
bmvend = bmv->bmv_offset + bmv->bmv_length;
if (bmv->bmv_count > ULONG_MAX / sizeof(struct getbmapx))
return XFS_ERROR(ENOMEM);
out = kmem_zalloc(bmv->bmv_count * sizeof(struct getbmapx), KM_MAYFAIL);
if (!out)
return XFS_ERROR(ENOMEM);
xfs_ilock(ip, XFS_IOLOCK_SHARED);
if (((iflags & BMV_IF_DELALLOC) == 0) &&
(whichfork == XFS_DATA_FORK) &&
(ip->i_delayed_blks || ip->i_size > ip->i_d.di_size)) {
/* xfs_fsize_t last_byte = xfs_file_last_byte(ip); */
error = xfs_flush_pages(ip, (xfs_off_t)0,
-1, 0, FI_REMAPF);
if (error) {
xfs_iunlock(ip, XFS_IOLOCK_SHARED);
return error;
if (whichfork == XFS_DATA_FORK && !(iflags & BMV_IF_DELALLOC)) {
if (ip->i_delayed_blks || ip->i_size > ip->i_d.di_size) {
error = xfs_flush_pages(ip, 0, -1, 0, FI_REMAPF);
if (error)
goto out_unlock_iolock;
}
}
ASSERT(whichfork == XFS_ATTR_FORK || (iflags & BMV_IF_DELALLOC) ||
ip->i_delayed_blks == 0);
ASSERT(ip->i_delayed_blks == 0);
}
lock = xfs_ilock_map_shared(ip);
@ -5995,23 +5999,25 @@ xfs_getbmap(
if (nex > XFS_IFORK_NEXTENTS(ip, whichfork) * 2 + 1)
nex = XFS_IFORK_NEXTENTS(ip, whichfork) * 2 + 1;
bmapi_flags = xfs_bmapi_aflag(whichfork) |
((iflags & BMV_IF_PREALLOC) ? 0 : XFS_BMAPI_IGSTATE);
bmapi_flags = xfs_bmapi_aflag(whichfork);
if (!(iflags & BMV_IF_PREALLOC))
bmapi_flags |= XFS_BMAPI_IGSTATE;
/*
* Allocate enough space to handle "subnex" maps at a time.
*/
error = ENOMEM;
subnex = 16;
map = kmem_alloc(subnex * sizeof(*map), KM_SLEEP);
map = kmem_alloc(subnex * sizeof(*map), KM_MAYFAIL);
if (!map)
goto out_unlock_ilock;
bmv->bmv_entries = 0;
if ((XFS_IFORK_NEXTENTS(ip, whichfork) == 0)) {
if (((iflags & BMV_IF_DELALLOC) == 0) ||
whichfork == XFS_ATTR_FORK) {
error = 0;
goto unlock_and_return;
}
if (XFS_IFORK_NEXTENTS(ip, whichfork) == 0 &&
(whichfork == XFS_ATTR_FORK || !(iflags & BMV_IF_DELALLOC))) {
error = 0;
goto out_free_map;
}
nexleft = nex;
@ -6023,53 +6029,61 @@ xfs_getbmap(
bmapi_flags, NULL, 0, map, &nmap,
NULL, NULL);
if (error)
goto unlock_and_return;
goto out_free_map;
ASSERT(nmap <= subnex);
for (i = 0; i < nmap && nexleft && bmv->bmv_length; i++) {
out.bmv_oflags = 0;
out[cur_ext].bmv_oflags = 0;
if (map[i].br_state == XFS_EXT_UNWRITTEN)
out.bmv_oflags |= BMV_OF_PREALLOC;
out[cur_ext].bmv_oflags |= BMV_OF_PREALLOC;
else if (map[i].br_startblock == DELAYSTARTBLOCK)
out.bmv_oflags |= BMV_OF_DELALLOC;
out.bmv_offset = XFS_FSB_TO_BB(mp, map[i].br_startoff);
out.bmv_length = XFS_FSB_TO_BB(mp, map[i].br_blockcount);
out.bmv_unused1 = out.bmv_unused2 = 0;
out[cur_ext].bmv_oflags |= BMV_OF_DELALLOC;
out[cur_ext].bmv_offset =
XFS_FSB_TO_BB(mp, map[i].br_startoff);
out[cur_ext].bmv_length =
XFS_FSB_TO_BB(mp, map[i].br_blockcount);
out[cur_ext].bmv_unused1 = 0;
out[cur_ext].bmv_unused2 = 0;
ASSERT(((iflags & BMV_IF_DELALLOC) != 0) ||
(map[i].br_startblock != DELAYSTARTBLOCK));
if (map[i].br_startblock == HOLESTARTBLOCK &&
whichfork == XFS_ATTR_FORK) {
/* came to the end of attribute fork */
out.bmv_oflags |= BMV_OF_LAST;
goto unlock_and_return;
} else {
int full = 0; /* user array is full */
if (!xfs_getbmapx_fix_eof_hole(ip, &out,
prealloced, bmvend,
map[i].br_startblock)) {
goto unlock_and_return;
}
/* format results & advance arg */
error = formatter(&arg, &out, &full);
if (error || full)
goto unlock_and_return;
nexleft--;
bmv->bmv_offset =
out.bmv_offset + out.bmv_length;
bmv->bmv_length = MAX((__int64_t)0,
(__int64_t)(bmvend - bmv->bmv_offset));
bmv->bmv_entries++;
out[cur_ext].bmv_oflags |= BMV_OF_LAST;
goto out_free_map;
}
if (!xfs_getbmapx_fix_eof_hole(ip, &out[cur_ext],
prealloced, bmvend,
map[i].br_startblock))
goto out_free_map;
nexleft--;
bmv->bmv_offset =
out[cur_ext].bmv_offset +
out[cur_ext].bmv_length;
bmv->bmv_length =
max_t(__int64_t, 0, bmvend - bmv->bmv_offset);
bmv->bmv_entries++;
cur_ext++;
}
} while (nmap && nexleft && bmv->bmv_length);
unlock_and_return:
out_free_map:
kmem_free(map);
out_unlock_ilock:
xfs_iunlock_map_shared(ip, lock);
out_unlock_iolock:
xfs_iunlock(ip, XFS_IOLOCK_SHARED);
kmem_free(map);
for (i = 0; i < cur_ext; i++) {
int full = 0; /* user array is full */
/* format results & advance arg */
error = formatter(&arg, &out[i], &full);
if (error || full)
break;
}
return error;
}

View file

@ -1258,8 +1258,10 @@ xfs_file_last_byte(
* necessary.
*/
if (ip->i_df.if_flags & XFS_IFEXTENTS) {
xfs_ilock(ip, XFS_ILOCK_SHARED);
error = xfs_bmap_last_offset(NULL, ip, &last_block,
XFS_DATA_FORK);
xfs_iunlock(ip, XFS_ILOCK_SHARED);
if (error) {
last_block = 0;
}

View file

@ -291,14 +291,17 @@ xfs_mount_validate_sb(
sbp->sb_sectsize > XFS_MAX_SECTORSIZE ||
sbp->sb_sectlog < XFS_MIN_SECTORSIZE_LOG ||
sbp->sb_sectlog > XFS_MAX_SECTORSIZE_LOG ||
sbp->sb_sectsize != (1 << sbp->sb_sectlog) ||
sbp->sb_blocksize < XFS_MIN_BLOCKSIZE ||
sbp->sb_blocksize > XFS_MAX_BLOCKSIZE ||
sbp->sb_blocklog < XFS_MIN_BLOCKSIZE_LOG ||
sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG ||
sbp->sb_blocksize != (1 << sbp->sb_blocklog) ||
sbp->sb_inodesize < XFS_DINODE_MIN_SIZE ||
sbp->sb_inodesize > XFS_DINODE_MAX_SIZE ||
sbp->sb_inodelog < XFS_DINODE_MIN_LOG ||
sbp->sb_inodelog > XFS_DINODE_MAX_LOG ||
sbp->sb_inodesize != (1 << sbp->sb_inodelog) ||
(sbp->sb_blocklog - sbp->sb_inodelog != sbp->sb_inopblog) ||
(sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE) ||
(sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE) ||