On Tue, Dec 22, 2015 at 08:37:04AM +1100, Dave Chinner wrote:
> From: Dave Chinner <dchinner@xxxxxxxxxx>
>
> It operates on a single AG at a time, sequentially, doing inode updates. All the
> data structures accessed and modified are per AG, as are the modification to
> on-disk structures. Hence we can run this phase concurrently across multiple
> AGs.
>
> This is important for large, broken filesystem repairs, where there can be
> millions of inodes that need link counts updated. Once such repair image takes
> more than 45 minutes to run phase 7 as a single threaded operation.
>
> Signed-off-by: Dave Chinner <dchinner@xxxxxxxxxx>
> ---
Reviewed-by: Brian Foster <bfoster@xxxxxxxxxx>
> repair/phase7.c | 77 ++++++++++++++++++++++++++++++++++-------------------
> repair/progress.c | 4 +--
> repair/protos.h | 2 +-
> repair/xfs_repair.c | 2 +-
> 4 files changed, 54 insertions(+), 31 deletions(-)
>
> diff --git a/repair/phase7.c b/repair/phase7.c
> index b1e3a55..91dad02 100644
> --- a/repair/phase7.c
> +++ b/repair/phase7.c
> @@ -26,6 +26,7 @@
> #include "dinode.h"
> #include "versions.h"
> #include "progress.h"
> +#include "threads.h"
>
> /* dinoc is a pointer to the IN-CORE dinode core */
> static void
> @@ -108,45 +109,67 @@ update_inode_nlinks(
> IRELE(ip);
> }
>
> -void
> -phase7(xfs_mount_t *mp)
> +/*
> + * for each ag, look at each inode 1 at a time. If the number of
> + * links is bad, reset it, log the inode core, commit the transaction
> + */
> +static void
> +do_link_updates(
> + struct work_queue *wq,
> + xfs_agnumber_t agno,
> + void *arg)
> {
> ino_tree_node_t *irec;
> - int i;
> int j;
> __uint32_t nrefs;
>
> + irec = findfirst_inode_rec(agno);
> +
> + while (irec != NULL) {
> + for (j = 0; j < XFS_INODES_PER_CHUNK; j++) {
> + ASSERT(is_inode_confirmed(irec, j));
> +
> + if (is_inode_free(irec, j))
> + continue;
> +
> + ASSERT(no_modify || is_inode_reached(irec, j));
> +
> + nrefs = num_inode_references(irec, j);
> + ASSERT(no_modify || nrefs > 0);
> +
> + if (get_inode_disk_nlinks(irec, j) != nrefs)
> + update_inode_nlinks(wq->mp,
> + XFS_AGINO_TO_INO(wq->mp, agno,
> + irec->ino_startnum + j),
> + nrefs);
> + }
> + irec = next_ino_rec(irec);
> + }
> +
> + PROG_RPT_INC(prog_rpt_done[agno], 1);
> +}
> +
> +void
> +phase7(
> + struct xfs_mount *mp,
> + int scan_threads)
> +{
> + struct work_queue wq;
> + int agno;
> +
> if (!no_modify)
> do_log(_("Phase 7 - verify and correct link counts...\n"));
> else
> do_log(_("Phase 7 - verify link counts...\n"));
>
> - /*
> - * for each ag, look at each inode 1 at a time. If the number of
> - * links is bad, reset it, log the inode core, commit the transaction
> - */
> - for (i = 0; i < glob_agcount; i++) {
> - irec = findfirst_inode_rec(i);
> -
> - while (irec != NULL) {
> - for (j = 0; j < XFS_INODES_PER_CHUNK; j++) {
> - ASSERT(is_inode_confirmed(irec, j));
> + set_progress_msg(PROGRESS_FMT_CORR_LINK, (__uint64_t) glob_agcount);
>
> - if (is_inode_free(irec, j))
> - continue;
> + create_work_queue(&wq, mp, scan_threads);
>
> - ASSERT(no_modify || is_inode_reached(irec, j));
> + for (agno = 0; agno < mp->m_sb.sb_agcount; agno++)
> + queue_work(&wq, do_link_updates, agno, NULL);
>
> - nrefs = num_inode_references(irec, j);
> - ASSERT(no_modify || nrefs > 0);
> + destroy_work_queue(&wq);
>
> - if (get_inode_disk_nlinks(irec, j) != nrefs)
> - update_inode_nlinks(mp,
> - XFS_AGINO_TO_INO(mp, i,
> - irec->ino_startnum + j),
> - nrefs);
> - }
> - irec = next_ino_rec(irec);
> - }
> - }
> + print_final_rpt();
> }
> diff --git a/repair/progress.c b/repair/progress.c
> index 418b803..2a09b23 100644
> --- a/repair/progress.c
> +++ b/repair/progress.c
> @@ -75,9 +75,9 @@ progress_rpt_t progress_rpt_reports[] = {
> {FMT2, N_("moving disconnected inodes to lost+found"), /* 12 */
> &rpt_fmts[FMT2], &rpt_types[TYPE_INODE]},
> {FMT1, N_("verify and correct link counts"), /* 13 */
> - &rpt_fmts[FMT1], &rpt_types[TYPE_INODE]},
> + &rpt_fmts[FMT1], &rpt_types[TYPE_AG]},
> {FMT1, N_("verify link counts"), /* 14 */
> - &rpt_fmts[FMT1], &rpt_types[TYPE_INODE]}
> + &rpt_fmts[FMT1], &rpt_types[TYPE_AG]}
> };
>
> pthread_t report_thread;
> diff --git a/repair/protos.h b/repair/protos.h
> index 9d5a2a6..b113aca 100644
> --- a/repair/protos.h
> +++ b/repair/protos.h
> @@ -50,7 +50,7 @@ void phase3(struct xfs_mount *);
> void phase4(struct xfs_mount *);
> void phase5(struct xfs_mount *);
> void phase6(struct xfs_mount *);
> -void phase7(struct xfs_mount *);
> +void phase7(struct xfs_mount *, int);
>
> int verify_set_agheader(struct xfs_mount *, struct xfs_buf *,
> struct xfs_sb *, struct xfs_agf *, struct xfs_agi *,
> diff --git a/repair/xfs_repair.c b/repair/xfs_repair.c
> index 3eaced4..fcdb212 100644
> --- a/repair/xfs_repair.c
> +++ b/repair/xfs_repair.c
> @@ -893,7 +893,7 @@ main(int argc, char **argv)
> phase6(mp);
> timestamp(PHASE_END, 6, NULL);
>
> - phase7(mp);
> + phase7(mp, phase2_threads);
> timestamp(PHASE_END, 7, NULL);
> } else {
> do_warn(
> --
> 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
