Re: [PATCH RESEND 2/10] xfs: Add support FALLOC_FL_COLLAPSE_RANGE for fallocate
Hi Dave.
2014-02-11 8:32 GMT+09:00, Dave Chinner :
> On Sun, Feb 02, 2014 at 02:44:11PM +0900, Namjae Jeon wrote:
>> From: Namjae Jeon
>>
>> Add support FALLOC_FL_COLLAPSE_RANGE for fallocate.
>>
>> Signed-off-by: Namjae Jeon
>> Signed-off-by: Ashish Sangwan
>
> A more detailed description would be nice for the change logs.
> .
Okay, I will update it on next version.
>
>> +while (nexts++ < num_exts &&
>> + *current_ext < XFS_IFORK_NEXTENTS(ip, whichfork)) {
>> +
>> +gotp = xfs_iext_get_ext(ifp, *current_ext);
>> +xfs_bmbt_get_all(gotp, );
>> +startoff = got.br_startoff - offset_shift_fsb;
>> +
>> +/*
>> + * Before shifting extent into hole, make sure that the hole
>> + * is large enough to accomodate the shift.
>> + */
>> +if (*current_ext) {
>> +xfs_bmbt_get_all(xfs_iext_get_ext(ifp,
>> +*current_ext - 1), );
>> +
>> +if (startoff < left.br_startoff + left.br_blockcount)
>> +error = XFS_ERROR(EINVAL);
>> +
>> +} else if (startoff > xfs_bmbt_get_startoff(gotp)) {
>> +/* Hole is at the start but not large enough */
>> +error = XFS_ERROR(EINVAL);
>> +}
>
> This second branch seems wrong to me:
>
> startoff = got.br_startoff - offset_shift_fsb;
> and
> got.br_startoff = xfs_bmbt_get_startoff(gotp)).
>
> I'm not 100% sure what you are trying to check in this case -
> perhaps some basic ascii art to describe the two cases is in order
> here:
>
> leftholegot
> +---+hhh+-+
> LS LE GSGE
> HS HE
>
> The first is checking that GS - offset_shift_fsb is greater than LE.
> i.e the shift doesn't overrun the hole betwenn LE and GS.
>
> leftholegot
> +---+hhh+-+
> LS LE GSGE
> HS HE
> +---+hhh+-+
> LS LE GS' GE'
> HS HE'
>
> The second I can't visualise from the code or comment
When we shift first extent, *current_ext will be 0. So we need to
check that offset_shift_fsb ( Number of blocks to be shifted ) should
be less than starting offset of the first extent.
So, code will be changed more clearly like this.
+ else if (offset_shift_fsb > got.br_startoff) {
+ /* Hole is at the start but not large enough */
+ error = XFS_ERROR(EINVAL);
+ }
And will update comment more clearly.
>
>
>> +
>> +if (error)
>> +goto del_cursor;
>> +
>> +if (cur) {
>> +error = xfs_bmbt_lookup_eq(cur,
>> +got.br_startoff,
>> +got.br_startblock,
>> +got.br_blockcount,
>> +);
>
> Whitespace comment - a more compact form is the typical XFS
> convention if it will fit in 80 columns:
Okay. I will fix it.
>
> error = xfs_bmbt_lookup_eq(cur, got.br_startoff,
> got.br_startblock,
> got.br_blockcount, );
>
>> +if (error)
>> +goto del_cursor;
>> +XFS_WANT_CORRUPTED_GOTO(i == 1, del_cursor);
>> +}
>> +
>> +/* Check if we can merge 2 adjacent extents */
>> +if (*current_ext &&
>> +left.br_startoff + left.br_blockcount == startoff &&
>> +left.br_startblock + left.br_blockcount ==
>> +got.br_startblock &&
>> +left.br_state == got.br_state &&
>> +left.br_blockcount + got.br_blockcount <= MAXEXTLEN) {
>> +blockcount = left.br_blockcount +
>> +xfs_bmbt_get_blockcount(gotp);
>
> got.br_blockcount?
Right. will fix it.
>
>> +xfs_iext_remove(ip, *current_ext, 1, 0);
>> +if (cur) {
>> +error = xfs_btree_delete(cur, );
>> +if (error)
>> +goto del_cursor;
>> +XFS_WANT_CORRUPTED_GOTO(i == 1, del_cursor);
>> +}
>> +XFS_IFORK_NEXT_SET(ip, whichfork,
>> +XFS_IFORK_NEXTENTS(ip, whichfork) - 1);
>> +gotp = xfs_iext_get_ext(ifp, --*current_ext);
>> +xfs_bmbt_get_all(gotp, );
>> +
>> +/* Make cursor point to the extent we will update */
>> +if (cur) {
>> +error =
Re: [PATCH RESEND 2/10] xfs: Add support FALLOC_FL_COLLAPSE_RANGE for fallocate
Hi Dave.
2014-02-11 8:32 GMT+09:00, Dave Chinner da...@fromorbit.com:
On Sun, Feb 02, 2014 at 02:44:11PM +0900, Namjae Jeon wrote:
From: Namjae Jeon namjae.j...@samsung.com
Add support FALLOC_FL_COLLAPSE_RANGE for fallocate.
Signed-off-by: Namjae Jeon namjae.j...@samsung.com
Signed-off-by: Ashish Sangwan a.sang...@samsung.com
A more detailed description would be nice for the change logs.
.
Okay, I will update it on next version.
+while (nexts++ num_exts
+ *current_ext XFS_IFORK_NEXTENTS(ip, whichfork)) {
+
+gotp = xfs_iext_get_ext(ifp, *current_ext);
+xfs_bmbt_get_all(gotp, got);
+startoff = got.br_startoff - offset_shift_fsb;
+
+/*
+ * Before shifting extent into hole, make sure that the hole
+ * is large enough to accomodate the shift.
+ */
+if (*current_ext) {
+xfs_bmbt_get_all(xfs_iext_get_ext(ifp,
+*current_ext - 1), left);
+
+if (startoff left.br_startoff + left.br_blockcount)
+error = XFS_ERROR(EINVAL);
+
+} else if (startoff xfs_bmbt_get_startoff(gotp)) {
+/* Hole is at the start but not large enough */
+error = XFS_ERROR(EINVAL);
+}
This second branch seems wrong to me:
startoff = got.br_startoff - offset_shift_fsb;
and
got.br_startoff = xfs_bmbt_get_startoff(gotp)).
I'm not 100% sure what you are trying to check in this case -
perhaps some basic ascii art to describe the two cases is in order
here:
leftholegot
+---+hhh+-+
LS LE GSGE
HS HE
The first is checking that GS - offset_shift_fsb is greater than LE.
i.e the shift doesn't overrun the hole betwenn LE and GS.
leftholegot
+---+hhh+-+
LS LE GSGE
HS HE
+---+hhh+-+
LS LE GS' GE'
HS HE'
The second I can't visualise from the code or comment
When we shift first extent, *current_ext will be 0. So we need to
check that offset_shift_fsb ( Number of blocks to be shifted ) should
be less than starting offset of the first extent.
So, code will be changed more clearly like this.
+ else if (offset_shift_fsb got.br_startoff) {
+ /* Hole is at the start but not large enough */
+ error = XFS_ERROR(EINVAL);
+ }
And will update comment more clearly.
+
+if (error)
+goto del_cursor;
+
+if (cur) {
+error = xfs_bmbt_lookup_eq(cur,
+got.br_startoff,
+got.br_startblock,
+got.br_blockcount,
+i);
Whitespace comment - a more compact form is the typical XFS
convention if it will fit in 80 columns:
Okay. I will fix it.
error = xfs_bmbt_lookup_eq(cur, got.br_startoff,
got.br_startblock,
got.br_blockcount, i);
+if (error)
+goto del_cursor;
+XFS_WANT_CORRUPTED_GOTO(i == 1, del_cursor);
+}
+
+/* Check if we can merge 2 adjacent extents */
+if (*current_ext
+left.br_startoff + left.br_blockcount == startoff
+left.br_startblock + left.br_blockcount ==
+got.br_startblock
+left.br_state == got.br_state
+left.br_blockcount + got.br_blockcount = MAXEXTLEN) {
+blockcount = left.br_blockcount +
+xfs_bmbt_get_blockcount(gotp);
got.br_blockcount?
Right. will fix it.
+xfs_iext_remove(ip, *current_ext, 1, 0);
+if (cur) {
+error = xfs_btree_delete(cur, i);
+if (error)
+goto del_cursor;
+XFS_WANT_CORRUPTED_GOTO(i == 1, del_cursor);
+}
+XFS_IFORK_NEXT_SET(ip, whichfork,
+XFS_IFORK_NEXTENTS(ip, whichfork) - 1);
+gotp = xfs_iext_get_ext(ifp, --*current_ext);
+xfs_bmbt_get_all(gotp, got);
+
+/* Make cursor point to the extent we will update */
+if (cur) {
+error = xfs_bmbt_lookup_eq(cur,
+got.br_startoff,
+
Re: [PATCH RESEND 2/10] xfs: Add support FALLOC_FL_COLLAPSE_RANGE for fallocate
On Sun, Feb 02, 2014 at 02:44:11PM +0900, Namjae Jeon wrote:
> From: Namjae Jeon
>
> Add support FALLOC_FL_COLLAPSE_RANGE for fallocate.
>
> Signed-off-by: Namjae Jeon
> Signed-off-by: Ashish Sangwan
A more detailed description would be nice for the change logs.
.
> + while (nexts++ < num_exts &&
> +*current_ext < XFS_IFORK_NEXTENTS(ip, whichfork)) {
> +
> + gotp = xfs_iext_get_ext(ifp, *current_ext);
> + xfs_bmbt_get_all(gotp, );
> + startoff = got.br_startoff - offset_shift_fsb;
> +
> + /*
> + * Before shifting extent into hole, make sure that the hole
> + * is large enough to accomodate the shift.
> + */
> + if (*current_ext) {
> + xfs_bmbt_get_all(xfs_iext_get_ext(ifp,
> + *current_ext - 1), );
> +
> + if (startoff < left.br_startoff + left.br_blockcount)
> + error = XFS_ERROR(EINVAL);
> +
> + } else if (startoff > xfs_bmbt_get_startoff(gotp)) {
> + /* Hole is at the start but not large enough */
> + error = XFS_ERROR(EINVAL);
> + }
This second branch seems wrong to me:
startoff = got.br_startoff - offset_shift_fsb;
and
got.br_startoff = xfs_bmbt_get_startoff(gotp)).
I'm not 100% sure what you are trying to check in this case -
perhaps some basic ascii art to describe the two cases is in order
here:
leftholegot
+---+hhh+-+
LS LE GSGE
HS HE
The first is checking that GS - offset_shift_fsb is greater than LE.
i.e the shift doesn't overrun the hole betwenn LE and GS.
leftholegot
+---+hhh+-+
LS LE GSGE
HS HE
+---+hhh+-+
LS LE GS' GE'
HS HE'
The second I can't visualise from the code or comment
> +
> + if (error)
> + goto del_cursor;
> +
> + if (cur) {
> + error = xfs_bmbt_lookup_eq(cur,
> + got.br_startoff,
> + got.br_startblock,
> + got.br_blockcount,
> + );
Whitespace comment - a more compact form is the typical XFS
convention if it will fit in 80 columns:
error = xfs_bmbt_lookup_eq(cur, got.br_startoff,
got.br_startblock,
got.br_blockcount, );
> + if (error)
> + goto del_cursor;
> + XFS_WANT_CORRUPTED_GOTO(i == 1, del_cursor);
> + }
> +
> + /* Check if we can merge 2 adjacent extents */
> + if (*current_ext &&
> + left.br_startoff + left.br_blockcount == startoff &&
> + left.br_startblock + left.br_blockcount ==
> + got.br_startblock &&
> + left.br_state == got.br_state &&
> + left.br_blockcount + got.br_blockcount <= MAXEXTLEN) {
> + blockcount = left.br_blockcount +
> + xfs_bmbt_get_blockcount(gotp);
got.br_blockcount?
> + xfs_iext_remove(ip, *current_ext, 1, 0);
> + if (cur) {
> + error = xfs_btree_delete(cur, );
> + if (error)
> + goto del_cursor;
> + XFS_WANT_CORRUPTED_GOTO(i == 1, del_cursor);
> + }
> + XFS_IFORK_NEXT_SET(ip, whichfork,
> + XFS_IFORK_NEXTENTS(ip, whichfork) - 1);
> + gotp = xfs_iext_get_ext(ifp, --*current_ext);
> + xfs_bmbt_get_all(gotp, );
> +
> + /* Make cursor point to the extent we will update */
> + if (cur) {
> + error = xfs_bmbt_lookup_eq(cur,
> + got.br_startoff,
> + got.br_startblock,
> + got.br_blockcount,
> + );
whitespace.
> + if (error)
> + goto del_cursor;
> + XFS_WANT_CORRUPTED_GOTO(i == 1, del_cursor);
> + }
> +
> + xfs_bmbt_set_blockcount(gotp, blockcount);
> + got.br_blockcount = blockcount;
> + goto bmbt_update;
> +
Re: [PATCH RESEND 2/10] xfs: Add support FALLOC_FL_COLLAPSE_RANGE for fallocate
On Sun, Feb 02, 2014 at 02:44:11PM +0900, Namjae Jeon wrote:
From: Namjae Jeon namjae.j...@samsung.com
Add support FALLOC_FL_COLLAPSE_RANGE for fallocate.
Signed-off-by: Namjae Jeon namjae.j...@samsung.com
Signed-off-by: Ashish Sangwan a.sang...@samsung.com
A more detailed description would be nice for the change logs.
.
+ while (nexts++ num_exts
+*current_ext XFS_IFORK_NEXTENTS(ip, whichfork)) {
+
+ gotp = xfs_iext_get_ext(ifp, *current_ext);
+ xfs_bmbt_get_all(gotp, got);
+ startoff = got.br_startoff - offset_shift_fsb;
+
+ /*
+ * Before shifting extent into hole, make sure that the hole
+ * is large enough to accomodate the shift.
+ */
+ if (*current_ext) {
+ xfs_bmbt_get_all(xfs_iext_get_ext(ifp,
+ *current_ext - 1), left);
+
+ if (startoff left.br_startoff + left.br_blockcount)
+ error = XFS_ERROR(EINVAL);
+
+ } else if (startoff xfs_bmbt_get_startoff(gotp)) {
+ /* Hole is at the start but not large enough */
+ error = XFS_ERROR(EINVAL);
+ }
This second branch seems wrong to me:
startoff = got.br_startoff - offset_shift_fsb;
and
got.br_startoff = xfs_bmbt_get_startoff(gotp)).
I'm not 100% sure what you are trying to check in this case -
perhaps some basic ascii art to describe the two cases is in order
here:
leftholegot
+---+hhh+-+
LS LE GSGE
HS HE
The first is checking that GS - offset_shift_fsb is greater than LE.
i.e the shift doesn't overrun the hole betwenn LE and GS.
leftholegot
+---+hhh+-+
LS LE GSGE
HS HE
+---+hhh+-+
LS LE GS' GE'
HS HE'
The second I can't visualise from the code or comment
+
+ if (error)
+ goto del_cursor;
+
+ if (cur) {
+ error = xfs_bmbt_lookup_eq(cur,
+ got.br_startoff,
+ got.br_startblock,
+ got.br_blockcount,
+ i);
Whitespace comment - a more compact form is the typical XFS
convention if it will fit in 80 columns:
error = xfs_bmbt_lookup_eq(cur, got.br_startoff,
got.br_startblock,
got.br_blockcount, i);
+ if (error)
+ goto del_cursor;
+ XFS_WANT_CORRUPTED_GOTO(i == 1, del_cursor);
+ }
+
+ /* Check if we can merge 2 adjacent extents */
+ if (*current_ext
+ left.br_startoff + left.br_blockcount == startoff
+ left.br_startblock + left.br_blockcount ==
+ got.br_startblock
+ left.br_state == got.br_state
+ left.br_blockcount + got.br_blockcount = MAXEXTLEN) {
+ blockcount = left.br_blockcount +
+ xfs_bmbt_get_blockcount(gotp);
got.br_blockcount?
+ xfs_iext_remove(ip, *current_ext, 1, 0);
+ if (cur) {
+ error = xfs_btree_delete(cur, i);
+ if (error)
+ goto del_cursor;
+ XFS_WANT_CORRUPTED_GOTO(i == 1, del_cursor);
+ }
+ XFS_IFORK_NEXT_SET(ip, whichfork,
+ XFS_IFORK_NEXTENTS(ip, whichfork) - 1);
+ gotp = xfs_iext_get_ext(ifp, --*current_ext);
+ xfs_bmbt_get_all(gotp, got);
+
+ /* Make cursor point to the extent we will update */
+ if (cur) {
+ error = xfs_bmbt_lookup_eq(cur,
+ got.br_startoff,
+ got.br_startblock,
+ got.br_blockcount,
+ i);
whitespace.
+ if (error)
+ goto del_cursor;
+ XFS_WANT_CORRUPTED_GOTO(i == 1, del_cursor);
+ }
+
+ xfs_bmbt_set_blockcount(gotp, blockcount);
+ got.br_blockcount = blockcount;
+ goto bmbt_update;
+ }
+
[PATCH RESEND 2/10] xfs: Add support FALLOC_FL_COLLAPSE_RANGE for fallocate
From: Namjae Jeon
Add support FALLOC_FL_COLLAPSE_RANGE for fallocate.
Signed-off-by: Namjae Jeon
Signed-off-by: Ashish Sangwan
---
fs/xfs/xfs_bmap.c | 195
fs/xfs/xfs_bmap.h |5 ++
fs/xfs/xfs_bmap_util.c | 99
fs/xfs/xfs_bmap_util.h |2 +
fs/xfs/xfs_file.c | 19 -
fs/xfs/xfs_trace.h |1 +
6 files changed, 319 insertions(+), 2 deletions(-)
diff --git a/fs/xfs/xfs_bmap.c b/fs/xfs/xfs_bmap.c
index 3ef11b2..aba3fc9 100644
--- a/fs/xfs/xfs_bmap.c
+++ b/fs/xfs/xfs_bmap.c
@@ -5358,3 +5358,198 @@ error0:
}
return error;
}
+
+/*
+ * Shift extent records to the left to cover a hole.
+ *
+ * The maximum number of extents to be shifted in a single operation
+ * is @num_exts, and @current_ext keeps track of the current extent
+ * index we have shifted. @offset_shift_fsb is the length by which each
+ * extent is shifted. If there is no hole to shift the extents
+ * into, this will be considered invalid operation and we abort immediately.
+ */
+int
+xfs_bmap_shift_extents(
+ struct xfs_trans*tp,
+ struct xfs_inode*ip,
+ int *done,
+ xfs_fileoff_t start_fsb,
+ xfs_fileoff_t offset_shift_fsb,
+ xfs_extnum_t*current_ext,
+ xfs_fsblock_t *firstblock,
+ struct xfs_bmap_free*flist,
+ int num_exts)
+{
+ struct xfs_btree_cur*cur;
+ struct xfs_bmbt_rec_host*gotp;
+ struct xfs_bmbt_irecgot;
+ struct xfs_bmbt_irecleft;
+ struct xfs_mount*mp = ip->i_mount;
+ struct xfs_ifork*ifp;
+ xfs_extnum_tnexts = 0;
+ xfs_fileoff_t startoff;
+ int error = 0;
+ int i;
+ int whichfork = XFS_DATA_FORK;
+ int logflags;
+ xfs_filblks_t blockcount = 0;
+
+ if (unlikely(XFS_TEST_ERROR(
+ (XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_EXTENTS &&
+XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_BTREE),
+mp, XFS_ERRTAG_BMAPIFORMAT, XFS_RANDOM_BMAPIFORMAT))) {
+ XFS_ERROR_REPORT("xfs_bmap_shift_extents",
+XFS_ERRLEVEL_LOW, mp);
+ return XFS_ERROR(EFSCORRUPTED);
+ }
+
+ if (XFS_FORCED_SHUTDOWN(mp))
+ return XFS_ERROR(EIO);
+
+ ASSERT(current_ext != NULL);
+
+ ifp = XFS_IFORK_PTR(ip, whichfork);
+
+ if (!(ifp->if_flags & XFS_IFEXTENTS)) {
+ /* Read in all the extents */
+ error = xfs_iread_extents(tp, ip, whichfork);
+ if (error)
+ return error;
+ }
+
+ /*
+* If *current_ext is 0, we would need to lookup the extent
+* from where we would start shifting and store it in gotp.
+*/
+ if (!*current_ext) {
+ gotp = xfs_iext_bno_to_ext(ifp, start_fsb, current_ext);
+ /*
+* gotp can be null in 2 cases: 1) if there are no extents
+* or 2) start_fsb lies in a hole beyond which there are
+* no extents. Either way, we are done.
+*/
+ if (!gotp) {
+ *done = 1;
+ return 0;
+ }
+ }
+
+ /* We are going to change core inode */
+ logflags = XFS_ILOG_CORE;
+
+ if (ifp->if_flags & XFS_IFBROOT) {
+ cur = xfs_bmbt_init_cursor(mp, tp, ip, whichfork);
+ cur->bc_private.b.firstblock = *firstblock;
+ cur->bc_private.b.flist = flist;
+ cur->bc_private.b.flags = 0;
+ } else {
+ cur = NULL;
+ logflags |= XFS_ILOG_DEXT;
+ }
+
+ while (nexts++ < num_exts &&
+ *current_ext < XFS_IFORK_NEXTENTS(ip, whichfork)) {
+
+ gotp = xfs_iext_get_ext(ifp, *current_ext);
+ xfs_bmbt_get_all(gotp, );
+ startoff = got.br_startoff - offset_shift_fsb;
+
+ /*
+* Before shifting extent into hole, make sure that the hole
+* is large enough to accomodate the shift.
+*/
+ if (*current_ext) {
+ xfs_bmbt_get_all(xfs_iext_get_ext(ifp,
+ *current_ext - 1), );
+
+ if (startoff < left.br_startoff + left.br_blockcount)
+ error = XFS_ERROR(EINVAL);
+
+ } else if (startoff > xfs_bmbt_get_startoff(gotp)) {
+ /* Hole is at the start but not large enough */
+ error =
[PATCH RESEND 2/10] xfs: Add support FALLOC_FL_COLLAPSE_RANGE for fallocate
From: Namjae Jeon namjae.j...@samsung.com
Add support FALLOC_FL_COLLAPSE_RANGE for fallocate.
Signed-off-by: Namjae Jeon namjae.j...@samsung.com
Signed-off-by: Ashish Sangwan a.sang...@samsung.com
---
fs/xfs/xfs_bmap.c | 195
fs/xfs/xfs_bmap.h |5 ++
fs/xfs/xfs_bmap_util.c | 99
fs/xfs/xfs_bmap_util.h |2 +
fs/xfs/xfs_file.c | 19 -
fs/xfs/xfs_trace.h |1 +
6 files changed, 319 insertions(+), 2 deletions(-)
diff --git a/fs/xfs/xfs_bmap.c b/fs/xfs/xfs_bmap.c
index 3ef11b2..aba3fc9 100644
--- a/fs/xfs/xfs_bmap.c
+++ b/fs/xfs/xfs_bmap.c
@@ -5358,3 +5358,198 @@ error0:
}
return error;
}
+
+/*
+ * Shift extent records to the left to cover a hole.
+ *
+ * The maximum number of extents to be shifted in a single operation
+ * is @num_exts, and @current_ext keeps track of the current extent
+ * index we have shifted. @offset_shift_fsb is the length by which each
+ * extent is shifted. If there is no hole to shift the extents
+ * into, this will be considered invalid operation and we abort immediately.
+ */
+int
+xfs_bmap_shift_extents(
+ struct xfs_trans*tp,
+ struct xfs_inode*ip,
+ int *done,
+ xfs_fileoff_t start_fsb,
+ xfs_fileoff_t offset_shift_fsb,
+ xfs_extnum_t*current_ext,
+ xfs_fsblock_t *firstblock,
+ struct xfs_bmap_free*flist,
+ int num_exts)
+{
+ struct xfs_btree_cur*cur;
+ struct xfs_bmbt_rec_host*gotp;
+ struct xfs_bmbt_irecgot;
+ struct xfs_bmbt_irecleft;
+ struct xfs_mount*mp = ip-i_mount;
+ struct xfs_ifork*ifp;
+ xfs_extnum_tnexts = 0;
+ xfs_fileoff_t startoff;
+ int error = 0;
+ int i;
+ int whichfork = XFS_DATA_FORK;
+ int logflags;
+ xfs_filblks_t blockcount = 0;
+
+ if (unlikely(XFS_TEST_ERROR(
+ (XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_EXTENTS
+XFS_IFORK_FORMAT(ip, whichfork) != XFS_DINODE_FMT_BTREE),
+mp, XFS_ERRTAG_BMAPIFORMAT, XFS_RANDOM_BMAPIFORMAT))) {
+ XFS_ERROR_REPORT(xfs_bmap_shift_extents,
+XFS_ERRLEVEL_LOW, mp);
+ return XFS_ERROR(EFSCORRUPTED);
+ }
+
+ if (XFS_FORCED_SHUTDOWN(mp))
+ return XFS_ERROR(EIO);
+
+ ASSERT(current_ext != NULL);
+
+ ifp = XFS_IFORK_PTR(ip, whichfork);
+
+ if (!(ifp-if_flags XFS_IFEXTENTS)) {
+ /* Read in all the extents */
+ error = xfs_iread_extents(tp, ip, whichfork);
+ if (error)
+ return error;
+ }
+
+ /*
+* If *current_ext is 0, we would need to lookup the extent
+* from where we would start shifting and store it in gotp.
+*/
+ if (!*current_ext) {
+ gotp = xfs_iext_bno_to_ext(ifp, start_fsb, current_ext);
+ /*
+* gotp can be null in 2 cases: 1) if there are no extents
+* or 2) start_fsb lies in a hole beyond which there are
+* no extents. Either way, we are done.
+*/
+ if (!gotp) {
+ *done = 1;
+ return 0;
+ }
+ }
+
+ /* We are going to change core inode */
+ logflags = XFS_ILOG_CORE;
+
+ if (ifp-if_flags XFS_IFBROOT) {
+ cur = xfs_bmbt_init_cursor(mp, tp, ip, whichfork);
+ cur-bc_private.b.firstblock = *firstblock;
+ cur-bc_private.b.flist = flist;
+ cur-bc_private.b.flags = 0;
+ } else {
+ cur = NULL;
+ logflags |= XFS_ILOG_DEXT;
+ }
+
+ while (nexts++ num_exts
+ *current_ext XFS_IFORK_NEXTENTS(ip, whichfork)) {
+
+ gotp = xfs_iext_get_ext(ifp, *current_ext);
+ xfs_bmbt_get_all(gotp, got);
+ startoff = got.br_startoff - offset_shift_fsb;
+
+ /*
+* Before shifting extent into hole, make sure that the hole
+* is large enough to accomodate the shift.
+*/
+ if (*current_ext) {
+ xfs_bmbt_get_all(xfs_iext_get_ext(ifp,
+ *current_ext - 1), left);
+
+ if (startoff left.br_startoff + left.br_blockcount)
+ error = XFS_ERROR(EINVAL);
+
+ } else if (startoff xfs_bmbt_get_startoff(gotp)) {
+ /* Hole is at the start but not
