https://github.com/torvalds/linux
Revision 9230a0b65b47fe6856c4468ec0175c4987e5bede authored by Dave Chinner on 20 November 2018, 06:50:08 UTC, committed by Darrick J. Wong on 21 November 2018, 18:10:53 UTC
Long saga. There have been days spent following this through dead end after dead end in multi-GB event traces. This morning, after writing a trace-cmd wrapper that enabled me to be more selective about XFS trace points, I discovered that I could get just enough essential tracepoints enabled that there was a 50:50 chance the fsx config would fail at ~115k ops. If it didn't fail at op 115547, I stopped fsx at op 115548 anyway. That gave me two traces - one where the problem manifested, and one where it didn't. After refining the traces to have the necessary information, I found that in the failing case there was a real extent in the COW fork compared to an unwritten extent in the working case. Walking back through the two traces to the point where the CWO fork extents actually diverged, I found that the bad case had an extra unwritten extent in it. This is likely because the bug it led me to had triggered multiple times in those 115k ops, leaving stray COW extents around. What I saw was a COW delalloc conversion to an unwritten extent (as they should always be through xfs_iomap_write_allocate()) resulted in a /written extent/: xfs_writepage: dev 259:0 ino 0x83 pgoff 0x17000 size 0x79a00 offset 0 length 0 xfs_iext_remove: dev 259:0 ino 0x83 state RC|LF|RF|COW cur 0xffff888247b899c0/2 offset 32 block 152 count 20 flag 1 caller xfs_bmap_add_extent_delay_real xfs_bmap_pre_update: dev 259:0 ino 0x83 state RC|LF|RF|COW cur 0xffff888247b899c0/1 offset 1 block 4503599627239429 count 31 flag 0 caller xfs_bmap_add_extent_delay_real xfs_bmap_post_update: dev 259:0 ino 0x83 state RC|LF|RF|COW cur 0xffff888247b899c0/1 offset 1 block 121 count 51 flag 0 caller xfs_bmap_add_ex Basically, Cow fork before: 0 1 32 52 +H+DDDDDDDDDDDD+UUUUUUUUUUU+ PREV RIGHT COW delalloc conversion allocates: 1 32 +uuuuuuuuuuuu+ NEW And the result according to the xfs_bmap_post_update trace was: 0 1 32 52 +H+wwwwwwwwwwwwwwwwwwwwwwww+ PREV Which is clearly wrong - it should be a merged unwritten extent, not an unwritten extent. That lead me to look at the LEFT_FILLING|RIGHT_FILLING|RIGHT_CONTIG case in xfs_bmap_add_extent_delay_real(), and sure enough, there's the bug. It takes the old delalloc extent (PREV) and adds the length of the RIGHT extent to it, takes the start block from NEW, removes the RIGHT extent and then updates PREV with the new extent. What it fails to do is update PREV.br_state. For delalloc, this is always XFS_EXT_NORM, while in this case we are converting the delayed allocation to unwritten, so it needs to be updated to XFS_EXT_UNWRITTEN. This LF|RF|RC case does not do this, and so the resultant extent is always written. And that's the bug I've been chasing for a week - a bmap btree bug, not a reflink/dedupe/copy_file_range bug, but a BMBT bug introduced with the recent in core extent tree scalability enhancements. Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
1 parent 2c30717
Tip revision: 9230a0b65b47fe6856c4468ec0175c4987e5bede authored by Dave Chinner on 20 November 2018, 06:50:08 UTC
xfs: delalloc -> unwritten COW fork allocation can go wrong
xfs: delalloc -> unwritten COW fork allocation can go wrong
Tip revision: 9230a0b
sg_split.c
/*
* Copyright (C) 2015 Robert Jarzmik <robert.jarzmik@free.fr>
*
* Scatterlist splitting helpers.
*
* This source code is licensed under the GNU General Public License,
* Version 2. See the file COPYING for more details.
*/
#include <linux/scatterlist.h>
#include <linux/slab.h>
struct sg_splitter {
struct scatterlist *in_sg0;
int nents;
off_t skip_sg0;
unsigned int length_last_sg;
struct scatterlist *out_sg;
};
static int sg_calculate_split(struct scatterlist *in, int nents, int nb_splits,
off_t skip, const size_t *sizes,
struct sg_splitter *splitters, bool mapped)
{
int i;
unsigned int sglen;
size_t size = sizes[0], len;
struct sg_splitter *curr = splitters;
struct scatterlist *sg;
for (i = 0; i < nb_splits; i++) {
splitters[i].in_sg0 = NULL;
splitters[i].nents = 0;
}
for_each_sg(in, sg, nents, i) {
sglen = mapped ? sg_dma_len(sg) : sg->length;
if (skip > sglen) {
skip -= sglen;
continue;
}
len = min_t(size_t, size, sglen - skip);
if (!curr->in_sg0) {
curr->in_sg0 = sg;
curr->skip_sg0 = skip;
}
size -= len;
curr->nents++;
curr->length_last_sg = len;
while (!size && (skip + len < sglen) && (--nb_splits > 0)) {
curr++;
size = *(++sizes);
skip += len;
len = min_t(size_t, size, sglen - skip);
curr->in_sg0 = sg;
curr->skip_sg0 = skip;
curr->nents = 1;
curr->length_last_sg = len;
size -= len;
}
skip = 0;
if (!size && --nb_splits > 0) {
curr++;
size = *(++sizes);
}
if (!nb_splits)
break;
}
return (size || !splitters[0].in_sg0) ? -EINVAL : 0;
}
static void sg_split_phys(struct sg_splitter *splitters, const int nb_splits)
{
int i, j;
struct scatterlist *in_sg, *out_sg;
struct sg_splitter *split;
for (i = 0, split = splitters; i < nb_splits; i++, split++) {
in_sg = split->in_sg0;
out_sg = split->out_sg;
for (j = 0; j < split->nents; j++, out_sg++) {
*out_sg = *in_sg;
if (!j) {
out_sg->offset += split->skip_sg0;
out_sg->length -= split->skip_sg0;
} else {
out_sg->offset = 0;
}
sg_dma_address(out_sg) = 0;
sg_dma_len(out_sg) = 0;
in_sg = sg_next(in_sg);
}
out_sg[-1].length = split->length_last_sg;
sg_mark_end(out_sg - 1);
}
}
static void sg_split_mapped(struct sg_splitter *splitters, const int nb_splits)
{
int i, j;
struct scatterlist *in_sg, *out_sg;
struct sg_splitter *split;
for (i = 0, split = splitters; i < nb_splits; i++, split++) {
in_sg = split->in_sg0;
out_sg = split->out_sg;
for (j = 0; j < split->nents; j++, out_sg++) {
sg_dma_address(out_sg) = sg_dma_address(in_sg);
sg_dma_len(out_sg) = sg_dma_len(in_sg);
if (!j) {
sg_dma_address(out_sg) += split->skip_sg0;
sg_dma_len(out_sg) -= split->skip_sg0;
}
in_sg = sg_next(in_sg);
}
sg_dma_len(--out_sg) = split->length_last_sg;
}
}
/**
* sg_split - split a scatterlist into several scatterlists
* @in: the input sg list
* @in_mapped_nents: the result of a dma_map_sg(in, ...), or 0 if not mapped.
* @skip: the number of bytes to skip in the input sg list
* @nb_splits: the number of desired sg outputs
* @split_sizes: the respective size of each output sg list in bytes
* @out: an array where to store the allocated output sg lists
* @out_mapped_nents: the resulting sg lists mapped number of sg entries. Might
* be NULL if sglist not already mapped (in_mapped_nents = 0)
* @gfp_mask: the allocation flag
*
* This function splits the input sg list into nb_splits sg lists, which are
* allocated and stored into out.
* The @in is split into :
* - @out[0], which covers bytes [@skip .. @skip + @split_sizes[0] - 1] of @in
* - @out[1], which covers bytes [@skip + split_sizes[0] ..
* @skip + @split_sizes[0] + @split_sizes[1] -1]
* etc ...
* It will be the caller's duty to kfree() out array members.
*
* Returns 0 upon success, or error code
*/
int sg_split(struct scatterlist *in, const int in_mapped_nents,
const off_t skip, const int nb_splits,
const size_t *split_sizes,
struct scatterlist **out, int *out_mapped_nents,
gfp_t gfp_mask)
{
int i, ret;
struct sg_splitter *splitters;
splitters = kcalloc(nb_splits, sizeof(*splitters), gfp_mask);
if (!splitters)
return -ENOMEM;
ret = sg_calculate_split(in, sg_nents(in), nb_splits, skip, split_sizes,
splitters, false);
if (ret < 0)
goto err;
ret = -ENOMEM;
for (i = 0; i < nb_splits; i++) {
splitters[i].out_sg = kmalloc_array(splitters[i].nents,
sizeof(struct scatterlist),
gfp_mask);
if (!splitters[i].out_sg)
goto err;
}
/*
* The order of these 3 calls is important and should be kept.
*/
sg_split_phys(splitters, nb_splits);
ret = sg_calculate_split(in, in_mapped_nents, nb_splits, skip,
split_sizes, splitters, true);
if (ret < 0)
goto err;
sg_split_mapped(splitters, nb_splits);
for (i = 0; i < nb_splits; i++) {
out[i] = splitters[i].out_sg;
if (out_mapped_nents)
out_mapped_nents[i] = splitters[i].nents;
}
kfree(splitters);
return 0;
err:
for (i = 0; i < nb_splits; i++)
kfree(splitters[i].out_sg);
kfree(splitters);
return ret;
}
EXPORT_SYMBOL(sg_split);
![swh spinner](/static/img/swh-spinner.gif)
Computing file changes ...