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
Raw File
Tip revision: 9230a0b65b47fe6856c4468ec0175c4987e5bede authored by Dave Chinner on 20 November 2018, 06:50:08 UTC
xfs: delalloc -> unwritten COW fork allocation can go wrong
Tip revision: 9230a0b
blk-map.c
// SPDX-License-Identifier: GPL-2.0
/*
 * Functions related to mapping data to requests
 */
#include <linux/kernel.h>
#include <linux/sched/task_stack.h>
#include <linux/module.h>
#include <linux/bio.h>
#include <linux/blkdev.h>
#include <linux/uio.h>

#include "blk.h"

/*
 * Append a bio to a passthrough request.  Only works if the bio can be merged
 * into the request based on the driver constraints.
 */
int blk_rq_append_bio(struct request *rq, struct bio **bio)
{
	struct bio *orig_bio = *bio;

	blk_queue_bounce(rq->q, bio);

	if (!rq->bio) {
		blk_rq_bio_prep(rq->q, rq, *bio);
	} else {
		if (!ll_back_merge_fn(rq->q, rq, *bio)) {
			if (orig_bio != *bio) {
				bio_put(*bio);
				*bio = orig_bio;
			}
			return -EINVAL;
		}

		rq->biotail->bi_next = *bio;
		rq->biotail = *bio;
		rq->__data_len += (*bio)->bi_iter.bi_size;
	}

	return 0;
}
EXPORT_SYMBOL(blk_rq_append_bio);

static int __blk_rq_unmap_user(struct bio *bio)
{
	int ret = 0;

	if (bio) {
		if (bio_flagged(bio, BIO_USER_MAPPED))
			bio_unmap_user(bio);
		else
			ret = bio_uncopy_user(bio);
	}

	return ret;
}

static int __blk_rq_map_user_iov(struct request *rq,
		struct rq_map_data *map_data, struct iov_iter *iter,
		gfp_t gfp_mask, bool copy)
{
	struct request_queue *q = rq->q;
	struct bio *bio, *orig_bio;
	int ret;

	if (copy)
		bio = bio_copy_user_iov(q, map_data, iter, gfp_mask);
	else
		bio = bio_map_user_iov(q, iter, gfp_mask);

	if (IS_ERR(bio))
		return PTR_ERR(bio);

	bio->bi_opf &= ~REQ_OP_MASK;
	bio->bi_opf |= req_op(rq);

	orig_bio = bio;

	/*
	 * We link the bounce buffer in and could have to traverse it
	 * later so we have to get a ref to prevent it from being freed
	 */
	ret = blk_rq_append_bio(rq, &bio);
	if (ret) {
		__blk_rq_unmap_user(orig_bio);
		return ret;
	}
	bio_get(bio);

	return 0;
}

/**
 * blk_rq_map_user_iov - map user data to a request, for passthrough requests
 * @q:		request queue where request should be inserted
 * @rq:		request to map data to
 * @map_data:   pointer to the rq_map_data holding pages (if necessary)
 * @iter:	iovec iterator
 * @gfp_mask:	memory allocation flags
 *
 * Description:
 *    Data will be mapped directly for zero copy I/O, if possible. Otherwise
 *    a kernel bounce buffer is used.
 *
 *    A matching blk_rq_unmap_user() must be issued at the end of I/O, while
 *    still in process context.
 *
 *    Note: The mapped bio may need to be bounced through blk_queue_bounce()
 *    before being submitted to the device, as pages mapped may be out of
 *    reach. It's the callers responsibility to make sure this happens. The
 *    original bio must be passed back in to blk_rq_unmap_user() for proper
 *    unmapping.
 */
int blk_rq_map_user_iov(struct request_queue *q, struct request *rq,
			struct rq_map_data *map_data,
			const struct iov_iter *iter, gfp_t gfp_mask)
{
	bool copy = false;
	unsigned long align = q->dma_pad_mask | queue_dma_alignment(q);
	struct bio *bio = NULL;
	struct iov_iter i;
	int ret = -EINVAL;

	if (!iter_is_iovec(iter))
		goto fail;

	if (map_data)
		copy = true;
	else if (iov_iter_alignment(iter) & align)
		copy = true;
	else if (queue_virt_boundary(q))
		copy = queue_virt_boundary(q) & iov_iter_gap_alignment(iter);

	i = *iter;
	do {
		ret =__blk_rq_map_user_iov(rq, map_data, &i, gfp_mask, copy);
		if (ret)
			goto unmap_rq;
		if (!bio)
			bio = rq->bio;
	} while (iov_iter_count(&i));

	if (!bio_flagged(bio, BIO_USER_MAPPED))
		rq->rq_flags |= RQF_COPY_USER;
	return 0;

unmap_rq:
	__blk_rq_unmap_user(bio);
fail:
	rq->bio = NULL;
	return ret;
}
EXPORT_SYMBOL(blk_rq_map_user_iov);

int blk_rq_map_user(struct request_queue *q, struct request *rq,
		    struct rq_map_data *map_data, void __user *ubuf,
		    unsigned long len, gfp_t gfp_mask)
{
	struct iovec iov;
	struct iov_iter i;
	int ret = import_single_range(rq_data_dir(rq), ubuf, len, &iov, &i);

	if (unlikely(ret < 0))
		return ret;

	return blk_rq_map_user_iov(q, rq, map_data, &i, gfp_mask);
}
EXPORT_SYMBOL(blk_rq_map_user);

/**
 * blk_rq_unmap_user - unmap a request with user data
 * @bio:	       start of bio list
 *
 * Description:
 *    Unmap a rq previously mapped by blk_rq_map_user(). The caller must
 *    supply the original rq->bio from the blk_rq_map_user() return, since
 *    the I/O completion may have changed rq->bio.
 */
int blk_rq_unmap_user(struct bio *bio)
{
	struct bio *mapped_bio;
	int ret = 0, ret2;

	while (bio) {
		mapped_bio = bio;
		if (unlikely(bio_flagged(bio, BIO_BOUNCED)))
			mapped_bio = bio->bi_private;

		ret2 = __blk_rq_unmap_user(mapped_bio);
		if (ret2 && !ret)
			ret = ret2;

		mapped_bio = bio;
		bio = bio->bi_next;
		bio_put(mapped_bio);
	}

	return ret;
}
EXPORT_SYMBOL(blk_rq_unmap_user);

/**
 * blk_rq_map_kern - map kernel data to a request, for passthrough requests
 * @q:		request queue where request should be inserted
 * @rq:		request to fill
 * @kbuf:	the kernel buffer
 * @len:	length of user data
 * @gfp_mask:	memory allocation flags
 *
 * Description:
 *    Data will be mapped directly if possible. Otherwise a bounce
 *    buffer is used. Can be called multiple times to append multiple
 *    buffers.
 */
int blk_rq_map_kern(struct request_queue *q, struct request *rq, void *kbuf,
		    unsigned int len, gfp_t gfp_mask)
{
	int reading = rq_data_dir(rq) == READ;
	unsigned long addr = (unsigned long) kbuf;
	int do_copy = 0;
	struct bio *bio, *orig_bio;
	int ret;

	if (len > (queue_max_hw_sectors(q) << 9))
		return -EINVAL;
	if (!len || !kbuf)
		return -EINVAL;

	do_copy = !blk_rq_aligned(q, addr, len) || object_is_on_stack(kbuf);
	if (do_copy)
		bio = bio_copy_kern(q, kbuf, len, gfp_mask, reading);
	else
		bio = bio_map_kern(q, kbuf, len, gfp_mask);

	if (IS_ERR(bio))
		return PTR_ERR(bio);

	bio->bi_opf &= ~REQ_OP_MASK;
	bio->bi_opf |= req_op(rq);

	if (do_copy)
		rq->rq_flags |= RQF_COPY_USER;

	orig_bio = bio;
	ret = blk_rq_append_bio(rq, &bio);
	if (unlikely(ret)) {
		/* request is too big */
		bio_put(orig_bio);
		return ret;
	}

	return 0;
}
EXPORT_SYMBOL(blk_rq_map_kern);
back to top