RFC: extended version of the xfs_bmapi_write retval fix
Hi all,
this series tries to take the xfs_bmapi_wite retval fix a little further, but I'm not entirely happy with the result.
The first patch is the actual return value fix with the comments fixed up as per the comment from Darrick, and a dead check removed I found while developing the rest.
The following patches are random cleanups to get the code in shape for the last patches.
The second to last patch fixes long-standing issues in xfs_bmap_add_extent_delay_real in code paths that haven't been used much or at all.
The last patch changes xfs_bmapi_write to not convert the entire delalloc extent if it hits one. This sounds simpler than it is, because delalloc conversion has been designed to always convert the entire
extent since the initial delalloc commits. I'm not really sure that
putting strain on this code now will do us much good, so I'll just leave these patches out for comments for now and will look into how coding up a loop of delalloc conversion calls in every place that could allocate blocks in the data fork of a regular file and thus hit delalloc so we can compare the approaches.
In the mean time I think that patch 1 is a candidate for 6.9 as it fixes the fuzzer problem.
Diffstat:
libxfs/xfs_attr_remote.c | 1 libxfs/xfs_bmap.c | 120 ++++++++++++++++++++++++++++------------------- libxfs/xfs_da_btree.c | 20 +------ scrub/quota_repair.c | 6 -- scrub/rtbitmap_repair.c | 2 xfs_bmap_util.c | 31 +++++------- xfs_dquot.c | 1 xfs_iomap.c | 8 --- xfs_reflink.c | 14 ----- xfs_rtalloc.c | 2 10 files changed, 93 insertions(+), 112 deletions(-)
PATCH 1/8 xfs: fix error returns from xfs_bmapi_write
xfs_bmapi_write can return 0 without actually returning a mapping in mval in two different cases:
1) when there is absolutely no space available to do an allocation
2) when converting delalloc space, and the allocation is so small that it only covers parts of the delalloc extent before the range requested by the caller
Callers at best can handle one of these cases, but in many cases can't cope with either one. Switch xfs_bmapi_write to always return a mapping or return an error code instead. For case 1) above ENOSPC is the obvious choice which is very much what the callers expect anyway. For case 2) there is no really good error code, so pick a funky one from the SysV streams portfolio.
This fixes the reproducer here:
which uses reserved blocks to create file systems that are gravely out of space and thus cause at least xfs_file_alloc_space to hang and trigger the lack of ENOSPC handling in xfs_dquot_disk_alloc.
Note that this patch does not actually make any caller but xfs_alloc_file_space deal intelligently with case 2) above.
--- fs/xfs/libxfs/xfs_attr_remote.c | 1 - fs/xfs/libxfs/xfs_bmap.c | 46 ++++++++++++++++++++++++++------- fs/xfs/libxfs/xfs_da_btree.c | 20 ++++---------- fs/xfs/scrub/quota_repair.c | 6 ----- fs/xfs/scrub/rtbitmap_repair.c | 2 -- fs/xfs/xfs_bmap_util.c | 31 +++++++++++----------- fs/xfs/xfs_dquot.c | 1 - fs/xfs/xfs_iomap.c | 8 ------ fs/xfs/xfs_reflink.c | 14 ---------- fs/xfs/xfs_rtalloc.c | 2 -- 10 files changed, 57 insertions(+), 74 deletions(-)
fs/xfs/libxfs/xfs_attr_remote.c +0 -1
@@ -626,7 +626,6 @@ xfs_attr_rmtval_set_blk(if (error)return error; -ASSERT (nmap ==1 );ASSERT ((map->br_startblock !=DELAYSTARTBLOCK ) && (map->br_startblock !=HOLESTARTBLOCK ));
fs/xfs/libxfs/xfs_bmap.c +37 -9
84 lines not on this page · raw message
fs/xfs/libxfs/xfs_da_btree.c +5 -15
@@ -2167,8 +2167,8 @@ xfs_da_grow_inode_int(struct xfs_inode *dp = args->dp;int w = args->whichfork;xfs_rfsblock_t nblks = dp->i_nblocks; -struct xfs_bmbt_irec map, *mapp; -int nmap, error, got, i, mapi; +struct xfs_bmbt_irec map, *mapp = ↦ +int nmap, error, got, i, mapi =1 ;/* * Find a spot in the file space to put the new block.
@@ -2184,14 +2184,7 @@ xfs_da_grow_inode_int( error =xfs_bmapi_write (tp, dp, *bno, count,xfs_bmapi_aflag (w)|XFS_BMAPI_METADATA |XFS_BMAPI_CONTIG , args->total, &map, &nmap); -if (error) -return error; - -ASSERT (nmap <=1 ); -if (nmap ==1 ) { - mapp = ↦ - mapi =1 ; - }else if (nmap ==0 && count >1 ) { +if (error == -ENOSPC && count >1 ) {xfs_fileoff_t b;int c;
@@ -2209,16 +2202,13 @@ xfs_da_grow_inode_int( args->total, &mapp[mapi], &nmap);if (error) goto out_free_map; -if (nmap <1 ) -break ; mapi += nmap; b = mapp[mapi -1 ].br_startoff + mapp[mapi -1 ].br_blockcount; } - }else { - mapi =0 ; - mapp =NULL ; } +if (error) + goto out_free_map;/* * Count the blocks we got, make sure it matches the total.
fs/xfs/scrub/quota_repair.c +0 -6
@@ -77,8 +77,6 @@ xrep_quota_item_fill_bmap_hole( irec, &nmaps);if (error)return error; -if (nmaps !=1 ) -return -ENOSPC ; dq->q_blkno =XFS_FSB_TO_DADDR (mp, irec->br_startblock);
@@ -444,10 +442,6 @@ xrep_quota_data_fork( XFS_BMAPI_CONVERT, 0, &nrec, &nmap);if (error) goto out; -if (nmap !=1 ) { - error = -ENOSPC ; - goto out; - }ASSERT (nrec.br_startoff == irec.br_startoff);ASSERT (nrec.br_blockcount == irec.br_blockcount);
fs/xfs/scrub/rtbitmap_repair.c +0 -2
@@ -108,8 +108,6 @@ xrep_rtbitmap_data_mappings( 0, &map, &nmaps);if (error)return error; -if (nmaps !=1 ) -return -EFSCORRUPTED ;/* Commit new extent and all deferred work. */ error =xrep_defer_finish (sc);
fs/xfs/xfs_bmap_util.c +15 -16
@@ -721,33 +721,32 @@ xfs_alloc_file_space(if (error) goto error; - error =xfs_bmapi_write (tp, ip, startoffset_fsb, - allocatesize_fsb,XFS_BMAPI_PREALLOC ,0 , imapp, - &nimaps); -if (error) - goto error; - - ip->i_diflags |=XFS_DIFLAG_PREALLOC ; -xfs_trans_log_inode (tp, ip,XFS_ILOG_CORE ); - - error =xfs_trans_commit (tp); -xfs_iunlock (ip,XFS_ILOCK_EXCL ); -if (error) -break ; -/* * If the allocator cannot find a single free extent large * enough to cover the start block of the requested range, -* xfs_bmapi_write will return 0 but leave *nimaps set to 0. +* xfs_bmapi_write will return -ENOSR. * * In that case we simply need to keep looping with the same * startoffset_fsb so that one of the following allocations * will eventually reach the requested range. */ -if (nimaps) { + error =xfs_bmapi_write (tp, ip, startoffset_fsb, + allocatesize_fsb,XFS_BMAPI_PREALLOC ,0 , imapp, + &nimaps); +if (error) { +if (error != -ENOSR ) + goto error; + error =0 ; + }else { startoffset_fsb += imapp->br_blockcount; allocatesize_fsb -= imapp->br_blockcount; } + + ip->i_diflags |=XFS_DIFLAG_PREALLOC ; +xfs_trans_log_inode (tp, ip,XFS_ILOG_CORE ); + + error =xfs_trans_commit (tp); +xfs_iunlock (ip,XFS_ILOCK_EXCL ); }return error;
fs/xfs/xfs_dquot.c +0 -1
@@ -357,7 +357,6 @@ xfs_dquot_disk_alloc( goto err_cancel;ASSERT (map.br_blockcount ==XFS_DQUOT_CLUSTER_SIZE_FSB ); -ASSERT (nmaps ==1 );ASSERT ((map.br_startblock !=DELAYSTARTBLOCK ) && (map.br_startblock !=HOLESTARTBLOCK ));
fs/xfs/xfs_iomap.c +0 -8
@@ -321,14 +321,6 @@ xfs_iomap_write_direct(if (error) goto out_unlock; -/* -* Copy any maps to caller's array and return any error. -*/ -if (nimaps ==0 ) { - error = -ENOSPC ; - goto out_unlock; - } -if (unlikely (!xfs_valid_startblock (ip, imap->br_startblock))) {xfs_bmap_mark_sick (ip,XFS_DATA_FORK ); error =xfs_alert_fsblock_zero (ip, imap);
fs/xfs/xfs_reflink.c +0 -14
@@ -430,13 +430,6 @@ xfs_reflink_fill_cow_hole(if (error)return error; -/* -* Allocation succeeded but the requested range was not even partially -* satisfied? Bail out! -*/ -if (nimaps ==0 ) -return -ENOSPC ; - convert:return xfs_reflink_convert_unwritten (ip, imap, cmap, convert_now);
@@ -499,13 +492,6 @@ xfs_reflink_fill_delalloc( error =xfs_trans_commit (tp);if (error)return error; - -/* -* Allocation succeeded but the requested range was not even -* partially satisfied? Bail out! -*/ -if (nimaps ==0 ) -return -ENOSPC ; }while (cmap->br_startoff + cmap->br_blockcount <= imap->br_startoff);return xfs_reflink_convert_unwritten (ip, imap, cmap, convert_now);
fs/xfs/xfs_rtalloc.c +0 -2
@@ -709,8 +709,6 @@ xfs_growfs_rt_alloc( nmap =1 ; error =xfs_bmapi_write (tp, ip, oblocks, nblocks - oblocks,XFS_BMAPI_METADATA ,0 , &map, &nmap); -if (!error && nmap <1 ) - error = -ENOSPC ;if (error) goto out_trans_cancel;/*
Signature · 1 line
2.39.2
Re: [PATCH 1/8] xfs: fix error returns from xfs_bmapi_write
Quoted 19 lines · Christoph Hellwig
xfs_bmapi_writecan return 0 without actually returning a mapping in mval in two different cases:1) when there is absolutely no space available to do an allocation
2) when converting delalloc space, and the allocation is so small that it only covers parts of the delalloc extent before the range requested by the callerCallers at best can handle one of these cases, but in many cases can't cope with either one. Switch
xfs_bmapi_writeto always return a mapping or return an error code instead. For case 1) above ENOSPC is the obvious choice which is very much what the callers expect anyway. For case 2) there is no really good error code, so pick a funky one from the SysV streams portfolio.This fixes the reproducer here:
Quoted 8 lines
which uses reserved blocks to create file systems that are gravely out of space and thus cause at least
xfs_file_alloc_spaceto hang and trigger the lack of ENOSPC handling inxfs_dquot_disk_alloc.Note that this patch does not actually make any caller but
xfs_alloc_file_spacedeal intelligently with case 2) above.Signed-off-by Christoph Hellwig <[email protected]>
Hmm. So I think the answer to my questions earlier is that with the series applied, ENOSR should never leak out because bmapi_write calls will always do the piece that the caller asked for, right?
If yes, then
--D
Quoted 339 lines
--- fs/xfs/libxfs/xfs_attr_remote.c | 1 - fs/xfs/libxfs/xfs_bmap.c | 46 ++++++++++++++++++++++++++------- fs/xfs/libxfs/xfs_da_btree.c | 20 ++++---------- fs/xfs/scrub/quota_repair.c | 6 ----- fs/xfs/scrub/rtbitmap_repair.c | 2 -- fs/xfs/xfs_bmap_util.c | 31 +++++++++++----------- fs/xfs/xfs_dquot.c | 1 - fs/xfs/xfs_iomap.c | 8 ------ fs/xfs/xfs_reflink.c | 14 ---------- fs/xfs/xfs_rtalloc.c | 2 -- 10 files changed, 57 insertions(+), 74 deletions(-)fs/xfs/libxfs/xfs_attr_remote.c +0
-1@@ -626,7 +626,6 @@ xfs_attr_rmtval_set_blk(if (error)return error; -ASSERT (nmap ==1 );ASSERT ((map->br_startblock !=DELAYSTARTBLOCK ) && (map->br_startblock !=HOLESTARTBLOCK ));fs/xfs/libxfs/xfs_bmap.c +37
-9@@ -4226,8 +4226,10 @@ xfs_bmapi_allocate( }else { error =xfs_bmap_alloc_userdata (bma); } -if (error || bma->blkno ==NULLFSBLOCK ) +if (error)return error; +if (bma->blkno ==NULLFSBLOCK ) +return -ENOSPC ;if (bma->flags &XFS_BMAPI_ZERO ) { error =xfs_zero_extent (bma->ip, bma->blkno, bma->length);@@ -4406,6 +4408,15 @@ xfs_bmapi_finish(* extent state if necessary. Details behaviour is controlled by the flags * parameter. Only allocates blocks from a single allocation group, to avoid * locking problems. +* +* Returns 0 on success and places the extent mappings in mval. nmaps is used +* as an input/output parameter where the caller specifies the maximum number +* of mappings that may be returned and xfs_bmapi_write passes back the number +* of mappings (including existing mappings) it found. +* +* Returns a negative error code on failure, including -ENOSPC when it could not +* allocate any blocks and -ENOSR when it did allocate blocks to convert a +* delalloc range, but those blocks were before the passed in range. */ int xfs_bmapi_write(@@ -4534,10 +4545,16 @@ xfs_bmapi_write(ASSERT (len >0 );ASSERT (bma.length >0 ); error =xfs_bmapi_allocate (&bma); -if (error) +if (error) { +/* +* If we already allocated space in a previous +* iteration return what we go so far when +* running out of space. +*/ +if (error == -ENOSPC && bma.nallocs) +break ; goto error0; -if (bma.blkno ==NULLFSBLOCK ) -break ; + }/* * If this is a CoW allocation, record the data in @@ -4575,7 +4592,6 @@ xfs_bmapi_write(if (!xfs_iext_next_extent (ifp, &bma.icur, &bma.got)) eof = true; } - *nmap = n; error =xfs_bmap_btree_to_extents (tp, ip, bma.cur, &bma.logflags, whichfork);@@ -4586,7 +4602,22 @@ xfs_bmapi_write( ifp->if_nextents > XFS_IFORK_MAXEXT(ip, whichfork));xfs_bmapi_finish (&bma, whichfork,0 );xfs_bmap_validate_ret (orig_bno, orig_len, orig_flags, orig_mval, - orig_nmap, *nmap); + orig_nmap, n); + +/* +* When converting delayed allocations, xfs_bmapi_allocate ignores +* the passed in bno and always converts from the start of the found +* delalloc extent. +* +* To avoid a successful return with *nmap set to 0, return the magic +* -ENOSR error code for this particular case so that the caller can +* handle it. +*/ +if (!n) { +ASSERT (bma.nallocs >= *nmap); +return -ENOSR ; + } + *nmap = n;return 0 ; error0:xfs_bmapi_finish (&bma, whichfork, error);@@ -4693,9 +4724,6 @@ xfs_bmapi_convert_delalloc(if (error) goto out_finish; - error = -ENOSPC ; -if (WARN_ON_ONCE (bma.blkno ==NULLFSBLOCK )) - goto out_finish;if (WARN_ON_ONCE (!xfs_valid_startblock (ip, bma.got.br_startblock))) {xfs_bmap_mark_sick (ip, whichfork); error = -EFSCORRUPTED ;fs/xfs/libxfs/xfs_da_btree.c +5
-15@@ -2167,8 +2167,8 @@ xfs_da_grow_inode_int(struct xfs_inode *dp = args->dp;int w = args->whichfork;xfs_rfsblock_t nblks = dp->i_nblocks; -struct xfs_bmbt_irec map, *mapp; -int nmap, error, got, i, mapi; +struct xfs_bmbt_irec map, *mapp = ↦ +int nmap, error, got, i, mapi =1 ;/* * Find a spot in the file space to put the new block. @@ -2184,14 +2184,7 @@ xfs_da_grow_inode_int( error =xfs_bmapi_write (tp, dp, *bno, count,xfs_bmapi_aflag (w)|XFS_BMAPI_METADATA |XFS_BMAPI_CONTIG , args->total, &map, &nmap); -if (error) -return error; - -ASSERT (nmap <=1 ); -if (nmap ==1 ) { - mapp = ↦ - mapi =1 ; - }else if (nmap ==0 && count >1 ) { +if (error == -ENOSPC && count >1 ) {xfs_fileoff_t b;int c;@@ -2209,16 +2202,13 @@ xfs_da_grow_inode_int( args->total, &mapp[mapi], &nmap);if (error) goto out_free_map; -if (nmap <1 ) -break ; mapi += nmap; b = mapp[mapi -1 ].br_startoff + mapp[mapi -1 ].br_blockcount; } - }else { - mapi =0 ; - mapp =NULL ; } +if (error) + goto out_free_map;/* * Count the blocks we got, make sure it matches the total. fs/xfs/scrub/quota_repair.c +0
-6@@ -77,8 +77,6 @@ xrep_quota_item_fill_bmap_hole( irec, &nmaps);if (error)return error; -if (nmaps !=1 ) -return -ENOSPC ; dq->q_blkno =XFS_FSB_TO_DADDR (mp, irec->br_startblock);@@ -444,10 +442,6 @@ xrep_quota_data_fork( XFS_BMAPI_CONVERT, 0, &nrec, &nmap);if (error) goto out; -if (nmap !=1 ) { - error = -ENOSPC ; - goto out; - }ASSERT (nrec.br_startoff == irec.br_startoff);ASSERT (nrec.br_blockcount == irec.br_blockcount);fs/xfs/scrub/rtbitmap_repair.c +0
-2@@ -108,8 +108,6 @@ xrep_rtbitmap_data_mappings( 0, &map, &nmaps);if (error)return error; -if (nmaps !=1 ) -return -EFSCORRUPTED ;/* Commit new extent and all deferred work. */ error =xrep_defer_finish (sc);fs/xfs/xfs_bmap_util.c +15
-16@@ -721,33 +721,32 @@ xfs_alloc_file_space(if (error) goto error; - error =xfs_bmapi_write (tp, ip, startoffset_fsb, - allocatesize_fsb,XFS_BMAPI_PREALLOC ,0 , imapp, - &nimaps); -if (error) - goto error; - - ip->i_diflags |=XFS_DIFLAG_PREALLOC ; -xfs_trans_log_inode (tp, ip,XFS_ILOG_CORE ); - - error =xfs_trans_commit (tp); -xfs_iunlock (ip,XFS_ILOCK_EXCL ); -if (error) -break ; -/* * If the allocator cannot find a single free extent large * enough to cover the start block of the requested range, -* xfs_bmapi_write will return 0 but leave *nimaps set to 0. +* xfs_bmapi_write will return -ENOSR. * * In that case we simply need to keep looping with the same * startoffset_fsb so that one of the following allocations * will eventually reach the requested range. */ -if (nimaps) { + error =xfs_bmapi_write (tp, ip, startoffset_fsb, + allocatesize_fsb,XFS_BMAPI_PREALLOC ,0 , imapp, + &nimaps); +if (error) { +if (error != -ENOSR ) + goto error; + error =0 ; + }else { startoffset_fsb += imapp->br_blockcount; allocatesize_fsb -= imapp->br_blockcount; } + + ip->i_diflags |=XFS_DIFLAG_PREALLOC ; +xfs_trans_log_inode (tp, ip,XFS_ILOG_CORE ); + + error =xfs_trans_commit (tp); +xfs_iunlock (ip,XFS_ILOCK_EXCL ); }return error;fs/xfs/xfs_dquot.c +0
-1@@ -357,7 +357,6 @@ xfs_dquot_disk_alloc( goto err_cancel;ASSERT (map.br_blockcount ==XFS_DQUOT_CLUSTER_SIZE_FSB ); -ASSERT (nmaps ==1 );ASSERT ((map.br_startblock !=DELAYSTARTBLOCK ) && (map.br_startblock !=HOLESTARTBLOCK ));fs/xfs/xfs_iomap.c +0
-8@@ -321,14 +321,6 @@ xfs_iomap_write_direct(if (error) goto out_unlock; -/* -* Copy any maps to caller's array and return any error. -*/ -if (nimaps ==0 ) { - error = -ENOSPC ; - goto out_unlock; - } -if (unlikely (!xfs_valid_startblock (ip, imap->br_startblock))) {xfs_bmap_mark_sick (ip,XFS_DATA_FORK ); error =xfs_alert_fsblock_zero (ip, imap);fs/xfs/xfs_reflink.c +0
-14@@ -430,13 +430,6 @@ xfs_reflink_fill_cow_hole(if (error)return error; -/* -* Allocation succeeded but the requested range was not even partially -* satisfied? Bail out! -*/ -if (nimaps ==0 ) -return -ENOSPC ; - convert:return xfs_reflink_convert_unwritten (ip, imap, cmap, convert_now);@@ -499,13 +492,6 @@ xfs_reflink_fill_delalloc( error =xfs_trans_commit (tp);if (error)return error; - -/* -* Allocation succeeded but the requested range was not even -* partially satisfied? Bail out! -*/ -if (nimaps ==0 ) -return -ENOSPC ; }while (cmap->br_startoff + cmap->br_blockcount <= imap->br_startoff);return xfs_reflink_convert_unwritten (ip, imap, cmap, convert_now);fs/xfs/xfs_rtalloc.c +0
-2@@ -709,8 +709,6 @@ xfs_growfs_rt_alloc( nmap =1 ; error =xfs_bmapi_write (tp, ip, oblocks, nblocks - oblocks,XFS_BMAPI_METADATA ,0 , &map, &nmap); -if (!error && nmap <1 ) - error = -ENOSPC ;if (error) goto out_trans_cancel;/* Signature · 1 line
2.39.2
Re: [PATCH 1/8] xfs: fix error returns from xfs_bmapi_write
Quoted 3 lines · Darrick J. Wong
Hmm. So I think the answer to my questions earlier is that with the series applied, ENOSR should never leak out because
bmapi_writecalls will always do the piece that the caller asked for, right?
Almost. With the whole series ENOSR goes away again (but then again I'm a bit skeptical if it opens new unknown cans of worms, see the cover letter).
With just this patch it will leak to user space if we ever hit a case where we completely mishandled this error before. Which we've not observed, but which isn't entirely impossible.
PATCH 2/8 xfs: remove the unusued tmp_logflags variable in xfs_bmapi_allocate
In reply to Christoph Hellwigtmp_logflags is initialized to 0 and then ORed into bma->logflags, which isn't actually doing anything.
--- fs/xfs/libxfs/xfs_bmap.c | 3 --- 1 file changed, 3 deletions(-)
fs/xfs/libxfs/xfs_bmap.c +0 -3
@@ -4191,7 +4191,6 @@ xfs_bmapi_allocate(struct xfs_mount *mp = bma->ip->i_mount;int whichfork =xfs_bmapi_whichfork (bma->flags);struct xfs_ifork *ifp =xfs_ifork_ptr (bma->ip, whichfork); -int tmp_logflags =0 ;int error;ASSERT (bma->length >0 );
@@ -4262,8 +4261,6 @@ xfs_bmapi_allocate( error =xfs_bmap_add_extent_hole_real (bma->tp, bma->ip, whichfork, &bma->icur, &bma->cur, &bma->got, &bma->logflags, bma->flags); - - bma->logflags |= tmp_logflags;if (error)return error;
Signature · 1 line
2.39.2
Re: [PATCH 2/8] xfs: remove the unusued tmp_logflags variable in xfs_bmapi_allocate
Quoted 4 lines · Christoph Hellwig
tmp_logflagsis initialized to 0 and then ORed intobma->logflags, which isn't actually doing anything.Signed-off-by Christoph Hellwig <[email protected]>
Looks good,
--D
Quoted 29 lines
--- fs/xfs/libxfs/xfs_bmap.c | 3 --- 1 file changed, 3 deletions(-)fs/xfs/libxfs/xfs_bmap.c +0
-3@@ -4191,7 +4191,6 @@ xfs_bmapi_allocate(struct xfs_mount *mp = bma->ip->i_mount;int whichfork =xfs_bmapi_whichfork (bma->flags);struct xfs_ifork *ifp =xfs_ifork_ptr (bma->ip, whichfork); -int tmp_logflags =0 ;int error;ASSERT (bma->length >0 );@@ -4262,8 +4261,6 @@ xfs_bmapi_allocate( error =xfs_bmap_add_extent_hole_real (bma->tp, bma->ip, whichfork, &bma->icur, &bma->cur, &bma->got, &bma->logflags, bma->flags); - - bma->logflags |= tmp_logflags;if (error)return error;Signature · 1 line
2.39.2
PATCH 3/8 xfs: lifr a xfs_valid_startblock into xfs_bmapi_allocate
In reply to Christoph Hellwigxfs_bmapi_convert_delalloc has a xfs_valid_startblock check on the block allocated by xfs_bmapi_allocate. Lift it into xfs_bmapi_allocate as we should assert the same for xfs_bmapi_write.
--- fs/xfs/libxfs/xfs_bmap.c | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-)
fs/xfs/libxfs/xfs_bmap.c +5 -6
@@ -4230,6 +4230,11 @@ xfs_bmapi_allocate(if (bma->blkno ==NULLFSBLOCK )return -ENOSPC ; +if (WARN_ON_ONCE (!xfs_valid_startblock (bma->ip, bma->blkno))) { +xfs_bmap_mark_sick (bma->ip, whichfork); +return -EFSCORRUPTED ; + } +if (bma->flags &XFS_BMAPI_ZERO ) { error =xfs_zero_extent (bma->ip, bma->blkno, bma->length);if (error)
@@ -4721,12 +4726,6 @@ xfs_bmapi_convert_delalloc(if (error) goto out_finish; -if (WARN_ON_ONCE (!xfs_valid_startblock (ip, bma.got.br_startblock))) { -xfs_bmap_mark_sick (ip, whichfork); - error = -EFSCORRUPTED ; - goto out_finish; - } -XFS_STATS_ADD (mp, xs_xstrat_bytes,XFS_FSB_TO_B (mp, bma.length));XFS_STATS_INC (mp, xs_xstrat_quick);
Signature · 1 line
2.39.2
Re: [PATCH 3/8] xfs: lifr a xfs_valid_startblock into xfs_bmapi_allocate
Quoted 1 line · Christoph Hellwig
Subject: [PATCH 3/8] xfs: lifr a
xfs_valid_startblockintoxfs_bmapi_allocate
lift
Quoted 5 lines
xfs_bmapi_convert_delallochas axfs_valid_startblockcheck on the block allocated byxfs_bmapi_allocate. Lift it intoxfs_bmapi_allocateas we should assert the same forxfs_bmapi_write.Signed-off-by Christoph Hellwig <[email protected]>
With the typo fixed,
--D
Quoted 37 lines
--- fs/xfs/libxfs/xfs_bmap.c | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-)fs/xfs/libxfs/xfs_bmap.c +5
-6@@ -4230,6 +4230,11 @@ xfs_bmapi_allocate(if (bma->blkno ==NULLFSBLOCK )return -ENOSPC ; +if (WARN_ON_ONCE (!xfs_valid_startblock (bma->ip, bma->blkno))) { +xfs_bmap_mark_sick (bma->ip, whichfork); +return -EFSCORRUPTED ; + } +if (bma->flags &XFS_BMAPI_ZERO ) { error =xfs_zero_extent (bma->ip, bma->blkno, bma->length);if (error)@@ -4721,12 +4726,6 @@ xfs_bmapi_convert_delalloc(if (error) goto out_finish; -if (WARN_ON_ONCE (!xfs_valid_startblock (ip, bma.got.br_startblock))) { -xfs_bmap_mark_sick (ip, whichfork); - error = -EFSCORRUPTED ; - goto out_finish; - } -XFS_STATS_ADD (mp, xs_xstrat_bytes,XFS_FSB_TO_B (mp, bma.length));XFS_STATS_INC (mp, xs_xstrat_quick);Signature · 1 line
2.39.2
PATCH 4/8 xfs: don't open code XFS_FILBLKS_MIN in xfs_bmapi_write
In reply to Christoph HellwigXFS_FILBLKS_MIN uses min_t and thus does the comparison using the correct xfs_filblks_t type. Use it in xfs_bmapi_write and slightly adjust the comment document th potential pitfall to take account of this
--- fs/xfs/libxfs/xfs_bmap.c | 9 +++------ 1 file changed, 3 insertions(+), 6 deletions(-)
fs/xfs/libxfs/xfs_bmap.c +3 -6
@@ -4537,14 +4537,11 @@ xfs_bmapi_write(* allocation length request (which can be 64 bits in * length) and the bma length request, which is * xfs_extlen_t and therefore 32 bits. Hence we have to -* check for 32-bit overflows and handle them here. +* be careful and do the min() using the larger type to +* avoid overflows. */ -if (len > (xfs_filblks_t )XFS_MAX_BMBT_EXTLEN ) - bma.length =XFS_MAX_BMBT_EXTLEN ; -else - bma.length = len; + bma.length =XFS_FILBLKS_MIN (len,XFS_MAX_BMBT_EXTLEN ); -ASSERT (len >0 );ASSERT (bma.length >0 ); error =xfs_bmapi_allocate (&bma);if (error) {
Signature · 1 line
2.39.2
Re: [PATCH 4/8] xfs: don't open code XFS_FILBLKS_MIN in xfs_bmapi_write
Quoted 5 lines · Christoph Hellwig
XFS_FILBLKS_MINusesmin_tand thus does the comparison using the correctxfs_filblks_ttype. Use it inxfs_bmapi_writeand slightly adjust the comment document th potential pitfall to take account of thisSigned-off-by Christoph Hellwig <[email protected]>
Pretty straightforward conversion,
--D
Quoted 30 lines
--- fs/xfs/libxfs/xfs_bmap.c | 9 +++------ 1 file changed, 3 insertions(+), 6 deletions(-)fs/xfs/libxfs/xfs_bmap.c +3
-6@@ -4537,14 +4537,11 @@ xfs_bmapi_write(* allocation length request (which can be 64 bits in * length) and the bma length request, which is * xfs_extlen_t and therefore 32 bits. Hence we have to -* check for 32-bit overflows and handle them here. +* be careful and do the min() using the larger type to +* avoid overflows. */ -if (len > (xfs_filblks_t )XFS_MAX_BMBT_EXTLEN ) - bma.length =XFS_MAX_BMBT_EXTLEN ; -else - bma.length = len; + bma.length =XFS_FILBLKS_MIN (len,XFS_MAX_BMBT_EXTLEN ); -ASSERT (len >0 );ASSERT (bma.length >0 ); error =xfs_bmapi_allocate (&bma);if (error) {Signature · 1 line
2.39.2
PATCH 5/8 xfs: pass the actual offset and len to allocate to xfs_bmapi_allocate
In reply to Christoph Hellwigxfs_bmapi_allocate currently overwrites offset and len when converting delayed allocations, and duplicates the length cap done for non-delalloc allocations. Move all that logic into the callers to avoid duplication and to make the calling conventions more obvious.
--- fs/xfs/libxfs/xfs_bmap.c | 32 ++++++++++++++++++-------------- 1 file changed, 18 insertions(+), 14 deletions(-)
fs/xfs/libxfs/xfs_bmap.c +18 -14
@@ -4194,21 +4194,11 @@ xfs_bmapi_allocate(int error;ASSERT (bma->length >0 ); +ASSERT (bma->length <=XFS_MAX_BMBT_EXTLEN ); -/* -* For the wasdelay case, we could also just allocate the stuff asked -* for in this bmap call but that wouldn't be as good. -*/ if (bma->wasdel) { - bma->length = (xfs_extlen_t )bma->got.br_blockcount; - bma->offset = bma->got.br_startoff;if (!xfs_iext_peek_prev_extent (ifp, &bma->icur, &bma->prev)) bma->prev.br_startoff =NULLFILEOFF ; - }else { - bma->length =XFS_FILBLKS_MIN (bma->length,XFS_MAX_BMBT_EXTLEN ); -if (!bma->eof) - bma->length =XFS_FILBLKS_MIN (bma->length, - bma->got.br_startoff - bma->offset); }if (bma->flags &XFS_BMAPI_CONTIG )
@@ -4542,6 +4532,15 @@ xfs_bmapi_write(*/ bma.length =XFS_FILBLKS_MIN (len,XFS_MAX_BMBT_EXTLEN ); +if (wasdelay) { + bma.offset = bma.got.br_startoff; + bma.length = bma.got.br_blockcount; + }else { +if (!eof) + bma.length =XFS_FILBLKS_MIN (bma.length, + bma.got.br_startoff - bno); + } +ASSERT (bma.length >0 ); error =xfs_bmapi_allocate (&bma);if (error) {
@@ -4694,11 +4693,16 @@ xfs_bmapi_convert_delalloc( bma.tp = tp; bma.ip = ip; bma.wasdel = true; - bma.offset = bma.got.br_startoff; - bma.length =max_t (xfs_filblks_t, bma.got.br_blockcount, -XFS_MAX_BMBT_EXTLEN ); bma.minleft =xfs_bmapi_minleft (tp, ip, whichfork); +/* +* Always allocate convert from the start of the delalloc extent even if +* that is outside the passed in range to create large contiguous +* extents on disk. +*/ + bma.offset = bma.got.br_startoff; + bma.length = bma.got.br_blockcount; +/* * When we're converting the delalloc reservations backing dirty pages * in the page cache, we must be careful about how we create the new
Signature · 1 line
2.39.2
Re: [PATCH 5/8] xfs: pass the actual offset and len to allocate to xfs_bmapi_allocate
Quoted 38 lines · Christoph Hellwig
xfs_bmapi_allocatecurrently overwrites offset and len when converting delayed allocations, and duplicates the length cap done for non-delalloc allocations. Move all that logic into the callers to avoid duplication and to make the calling conventions more obvious.Signed-off-by Christoph Hellwig <[email protected]>--- fs/xfs/libxfs/xfs_bmap.c | 32 ++++++++++++++++++-------------- 1 file changed, 18 insertions(+), 14 deletions(-)fs/xfs/libxfs/xfs_bmap.c +1
-11@@ -4194,21 +4194,11 @@ xfs_bmapi_allocate(int error;ASSERT (bma->length >0 ); +ASSERT (bma->length <=XFS_MAX_BMBT_EXTLEN ); -/* -* For the wasdelay case, we could also just allocate the stuff asked -* for in this bmap call but that wouldn't be as good. -*/ if (bma->wasdel) { - bma->length = (xfs_extlen_t )bma->got.br_blockcount; - bma->offset = bma->got.br_startoff;if (!xfs_iext_peek_prev_extent (ifp, &bma->icur, &bma->prev)) bma->prev.br_startoff =NULLFILEOFF ; - }else { - bma->length =XFS_FILBLKS_MIN (bma->length,XFS_MAX_BMBT_EXTLEN ); -if (!bma->eof) - bma->length =XFS_FILBLKS_MIN (bma->length, - bma->got.br_startoff - bma->offset); }if (bma->flags &XFS_BMAPI_CONTIG )
fs/xfs/libxfs/xfs_bmap.c xfs_bmapi_write( · line 4537
Quoted 6 lines
fs/xfs/libxfs/xfs_bmap.c +3
-0@@ -4542,6 +4532,15 @@ xfs_bmapi_write(*/ bma.length =XFS_FILBLKS_MIN (len,XFS_MAX_BMBT_EXTLEN ); +if (wasdelay) { + bma.offset = bma.got.br_startoff; + bma.length = bma.got.br_blockcount;
This read funny since we'd previously set bma.{offset,length} above, but I guess that preserves the "convert all the delalloc" behavior; and you turn it off in patch 8, right?
--D
Quoted 33 lines
fs/xfs/libxfs/xfs_bmap.c +14
-3+ }else { +if (!eof) + bma.length =XFS_FILBLKS_MIN (bma.length, + bma.got.br_startoff - bno); + } +ASSERT (bma.length >0 ); error =xfs_bmapi_allocate (&bma);if (error) {@@ -4694,11 +4693,16 @@ xfs_bmapi_convert_delalloc( bma.tp = tp; bma.ip = ip; bma.wasdel = true; - bma.offset = bma.got.br_startoff; - bma.length =max_t (xfs_filblks_t, bma.got.br_blockcount, -XFS_MAX_BMBT_EXTLEN ); bma.minleft =xfs_bmapi_minleft (tp, ip, whichfork); +/* +* Always allocate convert from the start of the delalloc extent even if +* that is outside the passed in range to create large contiguous +* extents on disk. +*/ + bma.offset = bma.got.br_startoff; + bma.length = bma.got.br_blockcount; +/* * When we're converting the delalloc reservations backing dirty pages * in the page cache, we must be careful about how we create the new Signature · 1 line
2.39.2
Re: [PATCH 5/8] xfs: pass the actual offset and len to allocate to xfs_bmapi_allocate
Quoted 11 lines · Darrick J. Wong
fs/xfs/libxfs/xfs_bmap.c xfs_bmapi_write( · line 4537
Quoted 7 lines
fs/xfs/libxfs/xfs_bmap.c +3
-0@@ -4542,6 +4532,15 @@ xfs_bmapi_write(*/ bma.length =XFS_FILBLKS_MIN (len,XFS_MAX_BMBT_EXTLEN ); +if (wasdelay) { + bma.offset = bma.got.br_startoff; + bma.length = bma.got.br_blockcount;This read funny since we'd previously set bma.{offset,length} above, but I guess that preserves the "convert all the delalloc" behavior; and you turn it off in patch 8, right?
Yes. That being said I should just move the assignments into the other branch here to make things more obvious.
PATCH 6/8 xfs: remove the xfs_iext_peek_prev_extent call in xfs_bmapi_allocate
In reply to Christoph HellwigBoth callers of xfs_bmapi_allocate already initialize bma->prev, don't redo that in xfs_bmapi_allocate.
--- fs/xfs/libxfs/xfs_bmap.c | 5 ----- 1 file changed, 5 deletions(-)
fs/xfs/libxfs/xfs_bmap.c +0 -5
@@ -4196,11 +4196,6 @@ xfs_bmapi_allocate(ASSERT (bma->length >0 );ASSERT (bma->length <=XFS_MAX_BMBT_EXTLEN ); -if (bma->wasdel) { -if (!xfs_iext_peek_prev_extent (ifp, &bma->icur, &bma->prev)) - bma->prev.br_startoff =NULLFILEOFF ; - } -if (bma->flags &XFS_BMAPI_CONTIG ) bma->minlen = bma->length;else
Signature · 1 line
2.39.2
Re: [PATCH 6/8] xfs: remove the xfs_iext_peek_prev_extent call in xfs_bmapi_allocate
Quoted 4 lines · Christoph Hellwig
Both callers of
xfs_bmapi_allocatealready initializebma->prev, don't redo that inxfs_bmapi_allocate.Signed-off-by Christoph Hellwig <[email protected]>
Looks ok,
--D
Quoted 24 lines
--- fs/xfs/libxfs/xfs_bmap.c | 5 ----- 1 file changed, 5 deletions(-)fs/xfs/libxfs/xfs_bmap.c +0
-5@@ -4196,11 +4196,6 @@ xfs_bmapi_allocate(ASSERT (bma->length >0 );ASSERT (bma->length <=XFS_MAX_BMBT_EXTLEN ); -if (bma->wasdel) { -if (!xfs_iext_peek_prev_extent (ifp, &bma->icur, &bma->prev)) - bma->prev.br_startoff =NULLFILEOFF ; - } -if (bma->flags &XFS_BMAPI_CONTIG ) bma->minlen = bma->length;else Signature · 1 line
2.39.2
PATCH 7/8 xfs: fix xfs_bmap_add_extent_delay_real for partial conversions
In reply to Christoph Hellwigxfs_bmap_add_extent_delay_real takes parts or all of a delalloc extent and converts them to a real extent. It is written to deal with any potential overlap of the to be converted range with the delalloc extent, but it turns out that currently only converting the entire extents, or a part starting at the beginning is actually exercised, as the only caller always tries to convert the entire delalloc extent, and either succeeds or at least progresses partially from the start.
If it only converts a tiny part of a delalloc extent, the indirect block calculation for the new delalloc extent (da_new) might be equivalent to that of the existing delalloc extent (da_new). If this extent conversion now requires allocating an indirect block that gets accounted into da_new, leading to the assert that da_new must be smaller or equal to da_new unless we split the extent to trigger.
Except for the assert that case is actually handled by just trying to allocate more space, as that already handled for the split case (which currently can't be reached at all), so just reusing it should be fine. Except that without dipping into the reserved block pool that would make it a bit too easy to trigger a fs shutdown due to ENOSPC. So in addition to adjusting the assert, also dip into the reserved block pool.
Note that I could only reproduce the assert with a change to only convert the actually asked range instead of the full delalloc extent from xfs_bmapi_write.
--- fs/xfs/libxfs/xfs_bmap.c | 13 +++++++++---- 1 file changed, 9 insertions(+), 4 deletions(-)
fs/xfs/libxfs/xfs_bmap.c +9 -4
@@ -1586,6 +1586,7 @@ xfs_bmap_add_extent_delay_real(if (error) goto done; } +ASSERT (da_new <= da_old);break ;case BMAP_LEFT_FILLING |BMAP_RIGHT_FILLING |BMAP_LEFT_CONTIG :
@@ -1616,6 +1617,7 @@ xfs_bmap_add_extent_delay_real(if (error) goto done; } +ASSERT (da_new <= da_old);break ;case BMAP_LEFT_FILLING |BMAP_RIGHT_FILLING |BMAP_RIGHT_CONTIG :
@@ -1650,6 +1652,7 @@ xfs_bmap_add_extent_delay_real(if (error) goto done; } +ASSERT (da_new <= da_old);break ;case BMAP_LEFT_FILLING |BMAP_RIGHT_FILLING :
@@ -1684,6 +1687,7 @@ xfs_bmap_add_extent_delay_real( goto done; } } +ASSERT (da_new <= da_old);break ;case BMAP_LEFT_FILLING |BMAP_LEFT_CONTIG :
@@ -1722,6 +1726,7 @@ xfs_bmap_add_extent_delay_real(if (error) goto done; } +ASSERT (da_new <= da_old);break ;case BMAP_LEFT_FILLING :
@@ -1812,6 +1817,7 @@ xfs_bmap_add_extent_delay_real(xfs_iext_update_extent (bma->ip, state, &bma->icur, &PREV );xfs_iext_next (ifp, &bma->icur);xfs_iext_update_extent (bma->ip, state, &bma->icur, &RIGHT ); +ASSERT (da_new <= da_old);break ;case BMAP_RIGHT_FILLING :
@@ -1861,6 +1867,7 @@ xfs_bmap_add_extent_delay_real(PREV .br_blockcount = temp;xfs_iext_insert (bma->ip, &bma->icur, &PREV , state);xfs_iext_next (ifp, &bma->icur); +ASSERT (da_new <= da_old);break ;case 0 :
@@ -1983,11 +1990,9 @@ xfs_bmap_add_extent_delay_real( }/* adjust for changes in reserved delayed indirect blocks */ -if (da_new != da_old) { -ASSERT (state ==0 || da_new < da_old); +if (da_new != da_old) error =xfs_mod_fdblocks (mp, (int64_t )(da_old - da_new), - false); - } + da_new > da_old);xfs_bmap_check_leaf_extents (bma->cur, bma->ip, whichfork); done:
Signature · 1 line
2.39.2
Re: [PATCH 7/8] xfs: fix xfs_bmap_add_extent_delay_real for partial conversions
Quoted 11 lines · Christoph Hellwig
xfs_bmap_add_extent_delay_realtakes parts or all of a delalloc extent and converts them to a real extent. It is written to deal with any potential overlap of the to be converted range with the delalloc extent, but it turns out that currently only converting the entire extents, or a part starting at the beginning is actually exercised, as the only caller always tries to convert the entire delalloc extent, and either succeeds or at least progresses partially from the start.If it only converts a tiny part of a delalloc extent, the indirect block calculation for the new delalloc extent (
da_new) might be equivalent to that of the existing delalloc extent (da_new). If this extent conversion now
da_old?
The code changes make sense to me otherwise.
--D
Quoted 98 lines
requires allocating an indirect block that gets accounted into
da_new, leading to the assert thatda_newmust be smaller or equal toda_newunless we split the extent to trigger.Except for the assert that case is actually handled by just trying to allocate more space, as that already handled for the split case (which currently can't be reached at all), so just reusing it should be fine. Except that without dipping into the reserved block pool that would make it a bit too easy to trigger a fs shutdown due to ENOSPC. So in addition to adjusting the assert, also dip into the reserved block pool.
Note that I could only reproduce the assert with a change to only convert the actually asked range instead of the full delalloc extent from
xfs_bmapi_write.Signed-off-by Christoph Hellwig <[email protected]>--- fs/xfs/libxfs/xfs_bmap.c | 13 +++++++++---- 1 file changed, 9 insertions(+), 4 deletions(-)fs/xfs/libxfs/xfs_bmap.c +9
-4@@ -1586,6 +1586,7 @@ xfs_bmap_add_extent_delay_real(if (error) goto done; } +ASSERT (da_new <= da_old);break ;case BMAP_LEFT_FILLING |BMAP_RIGHT_FILLING |BMAP_LEFT_CONTIG :@@ -1616,6 +1617,7 @@ xfs_bmap_add_extent_delay_real(if (error) goto done; } +ASSERT (da_new <= da_old);break ;case BMAP_LEFT_FILLING |BMAP_RIGHT_FILLING |BMAP_RIGHT_CONTIG :@@ -1650,6 +1652,7 @@ xfs_bmap_add_extent_delay_real(if (error) goto done; } +ASSERT (da_new <= da_old);break ;case BMAP_LEFT_FILLING |BMAP_RIGHT_FILLING :@@ -1684,6 +1687,7 @@ xfs_bmap_add_extent_delay_real( goto done; } } +ASSERT (da_new <= da_old);break ;case BMAP_LEFT_FILLING |BMAP_LEFT_CONTIG :@@ -1722,6 +1726,7 @@ xfs_bmap_add_extent_delay_real(if (error) goto done; } +ASSERT (da_new <= da_old);break ;case BMAP_LEFT_FILLING :@@ -1812,6 +1817,7 @@ xfs_bmap_add_extent_delay_real(xfs_iext_update_extent (bma->ip, state, &bma->icur, &PREV );xfs_iext_next (ifp, &bma->icur);xfs_iext_update_extent (bma->ip, state, &bma->icur, &RIGHT ); +ASSERT (da_new <= da_old);break ;case BMAP_RIGHT_FILLING :@@ -1861,6 +1867,7 @@ xfs_bmap_add_extent_delay_real(PREV .br_blockcount = temp;xfs_iext_insert (bma->ip, &bma->icur, &PREV , state);xfs_iext_next (ifp, &bma->icur); +ASSERT (da_new <= da_old);break ;case 0 :@@ -1983,11 +1990,9 @@ xfs_bmap_add_extent_delay_real( }/* adjust for changes in reserved delayed indirect blocks */ -if (da_new != da_old) { -ASSERT (state ==0 || da_new < da_old); +if (da_new != da_old) error =xfs_mod_fdblocks (mp, (int64_t )(da_old - da_new), - false); - } + da_new > da_old);xfs_bmap_check_leaf_extents (bma->cur, bma->ip, whichfork); done:Signature · 1 line
2.39.2
Re: [PATCH 7/8] xfs: fix xfs_bmap_add_extent_delay_real for partial conversions
Quoted 9 lines · Darrick J. Wong
Quoted 8 lines · Christoph Hellwig
xfs_bmap_add_extent_delay_realtakes parts or all of a delalloc extent and converts them to a real extent. It is written to deal with any potential overlap of the to be converted range with the delalloc extent, but it turns out that currently only converting the entire extents, or a part starting at the beginning is actually exercised, as the only caller always tries to convert the entire delalloc extent, and either succeeds or at least progresses partially from the start.
Quoted 5 lines
Quoted 3 lines
If it only converts a tiny part of a delalloc extent, the indirect block calculation for the new delalloc extent (
da_new) might be equivalent to that of the existing delalloc extent (da_new). If this extent conversion nowda_old?
Yes.
PATCH 8/8 xfs: do not allocate the entire delalloc extent in xfs_bmapi_write
In reply to Christoph HellwigWhile trying to convert the entire delalloc extent is a good decision for regular writeback as it leads to larger contigous on-disk extents, but for other callers of xfs_bmapi_write is is rather questionable as it forced them to loop creating new transactions just in case there is no large enough contiguous extent to cover the whole delalloc reservation.
Change xfs_bmapi_write to only allocate the passed in range instead.
--- fs/xfs/libxfs/xfs_bmap.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)
fs/xfs/libxfs/xfs_bmap.c +3 -2
@@ -4533,8 +4533,9 @@ xfs_bmapi_write( bma.length =XFS_FILBLKS_MIN (len,XFS_MAX_BMBT_EXTLEN );if (wasdelay) { - bma.offset = bma.got.br_startoff; - bma.length = bma.got.br_blockcount; + bma.length =XFS_FILBLKS_MIN (bma.length, + bma.got.br_blockcount - + (bno - bma.got.br_startoff)); }else {if (!eof) bma.length = XFS_FILBLKS_MIN(bma.length,
Signature · 1 line
2.39.2
Re: [PATCH 8/8] xfs: do not allocate the entire delalloc extent in xfs_bmapi_write
Quoted 8 lines · Christoph Hellwig
While trying to convert the entire delalloc extent is a good decision for regular writeback as it leads to larger contigous on-disk extents, but for other callers of
xfs_bmapi_writeis is rather questionable as it forced them to loop creating new transactions just in case there is no large enough contiguous extent to cover the whole delalloc reservation.Change
xfs_bmapi_writeto only allocate the passed in range instead.
Looking at this... I guess xfs_map_blocks -> xfs_convert_blocks -> xfs_bmapi_convert_delalloc -> xfs_bmapi_allocate is now how writeback converts delalloc extents before scheduling writeout. This is how the mass-conversions of large da reservations got done before this series, and that's still how it works, right?
Whereas xfs_bmapi_write is for targeted conversions only?
Quoted 1 line
Signed-off-by Christoph Hellwig <[email protected]>
If yes and yes, then:
--D
Quoted 24 lines
--- fs/xfs/libxfs/xfs_bmap.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)fs/xfs/libxfs/xfs_bmap.c +3
-2@@ -4533,8 +4533,9 @@ xfs_bmapi_write( bma.length =XFS_FILBLKS_MIN (len,XFS_MAX_BMBT_EXTLEN );if (wasdelay) { - bma.offset = bma.got.br_startoff; - bma.length = bma.got.br_blockcount; + bma.length =XFS_FILBLKS_MIN (bma.length, + bma.got.br_blockcount - + (bno - bma.got.br_startoff)); }else {if (!eof) bma.length = XFS_FILBLKS_MIN(bma.length,Signature · 1 line
2.39.2
Re: [PATCH 8/8] xfs: do not allocate the entire delalloc extent in xfs_bmapi_write
Quoted 8 lines · Darrick J. Wong
Quoted 7 lines · Christoph Hellwig
While trying to convert the entire delalloc extent is a good decision for regular writeback as it leads to larger contigous on-disk extents, but for other callers of
xfs_bmapi_writeis is rather questionable as it forced them to loop creating new transactions just in case there is no large enough contiguous extent to cover the whole delalloc reservation.
Quoted 7 lines
Quoted 1 line
Change
xfs_bmapi_writeto only allocate the passed in range instead.Looking at this... I guess
xfs_map_blocks->xfs_convert_blocks->xfs_bmapi_convert_delalloc->xfs_bmapi_allocateis now how writeback converts delalloc extents before scheduling writeout. This is how the mass-conversions of large da reservations got done before this series, and that's still how it works, right?
Yes and yes.
Quoted 1 line
Whereas
xfs_bmapi_writeis for targeted conversions only?
targeted is one way to describe it, the other way to look at it is that xfs_bmapi_write is used where the callers want to allocate real (or unwritten) extents for a range, which just happens to convert delalloc as a side-effect as those callers don't want to deal with delalloc extents.
Quoted 5 lines
Quoted 2 lines
Signed-off-by Christoph Hellwig <[email protected]>If yes and yes, then:
Reviewed-by Darrick J. Wong <[email protected]>
So as mentioned in the cover letter I'm quite worried about the new behavior we expose here as we always converted delalloc extents from the start and tried to convert it to the end, and this now changes that. So while the changes looks quite simple they expose a lot of previously untested code and behavior.
It's probably the right thing to do but quite risky, let me know what you think about it.