On Tue, Dec 22, 2015 at 08:37:08AM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@xxxxxxxxxx>
>
> now that we try to write dirty buffers before we release them, we
> can get buildup of unwritable dirty buffers on the LRU lists, This
> results in the cache shaker repeatedly trying to write out these
> buffers every time the cache fills up. This results in more
> corruption warnings, and takes up a lot of time doing reclaiming
> nothing. This can effectively livelock the processing parts of phase
> 4.
>
> Fix this by not trying to write buffers with corruption errors on
> them. These errors will get cleared when the buffer is re-read and
> fixed and them marked dirty again. At which point, we'll be able to
> write them and so the cache can reclaim them successfully.
>
> Signed-off-by: Dave Chinner <dchinner@xxxxxxxxxx>
> ---
Reviewed-by: Brian Foster <bfoster@xxxxxxxxxx>
> libxfs/rdwr.c | 27 ++++++++++++++++-----------
> 1 file changed, 16 insertions(+), 11 deletions(-)
>
> diff --git a/libxfs/rdwr.c b/libxfs/rdwr.c
> index 0337a21..a1f0029 100644
> --- a/libxfs/rdwr.c
> +++ b/libxfs/rdwr.c
> @@ -1103,7 +1103,6 @@ int
> libxfs_writebufr(xfs_buf_t *bp)
> {
> int fd = libxfs_device_to_fd(bp->b_target->dev);
> - int error = 0;
>
> /*
> * we never write buffers that are marked stale. This indicates they
> @@ -1134,7 +1133,7 @@ libxfs_writebufr(xfs_buf_t *bp)
> }
>
> if (!(bp->b_flags & LIBXFS_B_DISCONTIG)) {
> - error = __write_buf(fd, bp->b_addr, bp->b_bcount,
> + bp->b_error = __write_buf(fd, bp->b_addr, bp->b_bcount,
> LIBXFS_BBTOOFF64(bp->b_bn), bp->b_flags);
> } else {
> int i;
> @@ -1144,11 +1143,10 @@ libxfs_writebufr(xfs_buf_t *bp)
> off64_t offset = LIBXFS_BBTOOFF64(bp->b_map[i].bm_bn);
> int len = BBTOB(bp->b_map[i].bm_len);
>
> - error = __write_buf(fd, buf, len, offset, bp->b_flags);
> - if (error) {
> - bp->b_error = error;
> + bp->b_error = __write_buf(fd, buf, len, offset,
> + bp->b_flags);
> + if (bp->b_error)
> break;
> - }
> buf += len;
> }
> }
> @@ -1157,14 +1155,14 @@ libxfs_writebufr(xfs_buf_t *bp)
> printf("%lx: %s: wrote %u bytes, blkno=%llu(%llu), %p, error %d\n",
> pthread_self(), __FUNCTION__, bp->b_bcount,
> (long long)LIBXFS_BBTOOFF64(bp->b_bn),
> - (long long)bp->b_bn, bp, error);
> + (long long)bp->b_bn, bp, bp->b_error);
> #endif
> - if (!error) {
> + if (!bp->b_error) {
> bp->b_flags |= LIBXFS_B_UPTODATE;
> bp->b_flags &= ~(LIBXFS_B_DIRTY | LIBXFS_B_EXIT |
> LIBXFS_B_UNCHECKED);
> }
> - return error;
> + return bp->b_error;
> }
>
> int
> @@ -1266,15 +1264,22 @@ libxfs_bulkrelse(
> return count;
> }
>
> +/*
> + * When a buffer is marked dirty, the error is cleared. Hence if we are trying
> + * to flush a buffer prior to cache reclaim that has an error on it it means
> + * we've already tried to flush it and it failed. Prevent repeated corruption
> + * errors from being reported by skipping such buffers - when the corruption is
> + * fixed the buffer will be marked dirty again and we can write it again.
> + */
> static int
> libxfs_bflush(
> struct cache_node *node)
> {
> struct xfs_buf *bp = (struct xfs_buf *)node;
>
> - if (bp->b_flags & LIBXFS_B_DIRTY)
> + if (!bp->b_error && bp->b_flags & LIBXFS_B_DIRTY)
> return libxfs_writebufr(bp);
> - return 0;
> + return bp->b_error;
> }
>
> void
> --
> 2.5.0
>
> _______________________________________________
> xfs mailing list
> xfs@xxxxxxxxxxx
> http://oss.sgi.com/mailman/listinfo/xfs
_______________________________________________
xfs mailing list
xfs@xxxxxxxxxxx
http://oss.sgi.com/mailman/listinfo/xfs
