Re: [PATCH 05/22] xfs: scrub in-memory metadata buffers

[Allison Henderson <allison.henderson@oracle.com>]

On 7/20/2017 9:39 PM, Darrick J. Wong wrote:
> From: Darrick J. Wong <darrick.wong@oracle.com>
>
> Call the verifier function for all in-memory metadata buffers, looking
> for memory corruption either due to bad memory or coding bugs.
>
> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
> ---
>  fs/xfs/Makefile         |    1
>  fs/xfs/libxfs/xfs_fs.h  |    3 +
>  fs/xfs/scrub/common.c   |    4 +
>  fs/xfs/scrub/common.h   |    2 +
>  fs/xfs/scrub/metabufs.c |  177 +++++++++++++++++++++++++++++++++++++++++++++++
>  fs/xfs/xfs_trace.h      |    3 +
>  6 files changed, 188 insertions(+), 2 deletions(-)
>  create mode 100644 fs/xfs/scrub/metabufs.c
>
>
> diff --git a/fs/xfs/Makefile b/fs/xfs/Makefile
> index 4e04da9..67cf4ac 100644
> --- a/fs/xfs/Makefile
> +++ b/fs/xfs/Makefile
> @@ -142,5 +142,6 @@ ifeq ($(CONFIG_XFS_ONLINE_SCRUB),y)
>  xfs-y				+= $(addprefix scrub/, \
>  				   btree.o \
>  				   common.o \
> +				   metabufs.o \
>  				   )
>  endif
> diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
> index aeccc99..9fb3c65 100644
> --- a/fs/xfs/libxfs/xfs_fs.h
> +++ b/fs/xfs/libxfs/xfs_fs.h
> @@ -482,7 +482,8 @@ struct xfs_scrub_metadata {
>   * Metadata types and flags for scrub operation.
>   */
>  #define XFS_SCRUB_TYPE_TEST	0	/* dummy to test ioctl */
> -#define XFS_SCRUB_TYPE_MAX	0
> +#define XFS_SCRUB_TYPE_METABUFS	1	/* in-core metadata buffers */
> +#define XFS_SCRUB_TYPE_MAX	1
>
>  /* i: repair this metadata */
>  #define XFS_SCRUB_FLAG_REPAIR		(1 << 0)
> diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
> index 331aa14..e06131f 100644
> --- a/fs/xfs/scrub/common.c
> +++ b/fs/xfs/scrub/common.c
> @@ -610,6 +610,10 @@ static const struct xfs_scrub_meta_fns meta_scrub_fns[] = {
>  		.setup	= xfs_scrub_setup_fs,
>  		.scrub	= xfs_scrub_dummy,
>  	},
> +	{ /* in-memory metadata buffers */
> +		.setup	= xfs_scrub_setup_metabufs,
> +		.scrub	= xfs_scrub_metabufs,
> +	},
>  };
>
>  /* Dispatch metadata scrubbing. */
> diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
> index 15baccb..5f0818c 100644
> --- a/fs/xfs/scrub/common.h
> +++ b/fs/xfs/scrub/common.h
> @@ -198,12 +198,14 @@ int xfs_scrub_ag_btcur_init(struct xfs_scrub_context *sc,
>
>  #define SETUP_FN(name) int name(struct xfs_scrub_context *sc, struct xfs_inode *ip)
>  SETUP_FN(xfs_scrub_setup_fs);
> +SETUP_FN(xfs_scrub_setup_metabufs);
>  #undef SETUP_FN
>
>  /* Metadata scrubbers */
>
>  #define SCRUB_FN(name) int name(struct xfs_scrub_context *sc)
>  SCRUB_FN(xfs_scrub_dummy);
> +SCRUB_FN(xfs_scrub_metabufs);
>  #undef SCRUB_FN
>
>  #endif	/* __XFS_REPAIR_COMMON_H__ */
> diff --git a/fs/xfs/scrub/metabufs.c b/fs/xfs/scrub/metabufs.c
> new file mode 100644
> index 0000000..63faaa6
> --- /dev/null
> +++ b/fs/xfs/scrub/metabufs.c
> @@ -0,0 +1,177 @@
> +/*
> + * Copyright (C) 2017 Oracle.  All Rights Reserved.
> + *
> + * Author: Darrick J. Wong <darrick.wong@oracle.com>
> + *
> + * This program is free software; you can redistribute it and/or
> + * modify it under the terms of the GNU General Public License
> + * as published by the Free Software Foundation; either version 2
> + * of the License, or (at your option) any later version.
> + *
> + * This program is distributed in the hope that it would be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> + * GNU General Public License for more details.
> + *
> + * You should have received a copy of the GNU General Public License
> + * along with this program; if not, write the Free Software Foundation,
> + * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
> + */
> +#include "xfs.h"
> +#include "xfs_fs.h"
> +#include "xfs_shared.h"
> +#include "xfs_format.h"
> +#include "xfs_trans_resv.h"
> +#include "xfs_mount.h"
> +#include "xfs_defer.h"
> +#include "xfs_btree.h"
> +#include "xfs_bit.h"
> +#include "xfs_log_format.h"
> +#include "xfs_trans.h"
> +#include "xfs_trace.h"
> +#include "xfs_sb.h"
> +#include "scrub/common.h"
> +
> +/* We only iterate buffers one by one, so we don't need any setup. */
> +int
> +xfs_scrub_setup_metabufs(
> +	struct xfs_scrub_context	*sc,
> +	struct xfs_inode		*ip)
> +{
> +	return 0;
> +}
> +
> +#define XFS_SCRUB_METABUFS_TOO_MANY_RETRIES	10
> +struct xfs_scrub_metabufs_info {
> +	struct xfs_scrub_context	*sc;
> +	unsigned int			retries;
> +};
> +
> +/* In-memory buffer corruption. */
> +
> +#define XFS_SCRUB_BUF_OP_ERROR_GOTO(label) \
> +	XFS_SCRUB_OP_ERROR_GOTO(smi->sc, \
> +			xfs_daddr_to_agno(smi->sc->mp, bp->b_bn), \
> +			xfs_daddr_to_agbno(smi->sc->mp, bp->b_bn), "buf", \
> +			&error, label)
> +STATIC int
> +xfs_scrub_metabufs_scrub_buf(
> +	struct xfs_scrub_metabufs_info	*smi,
> +	struct xfs_buf			*bp)
> +{
> +	int				olderror;
> +	int				error = 0;
> +
> +	/*
> +	 * We hold the rcu lock during the rhashtable walk, so we can't risk
> +	 * having the log forced due to a stale buffer by xfs_buf_lock.
> +	 */
> +	if (bp->b_flags & XBF_STALE)
> +		return 0;
> +
> +	atomic_inc(&bp->b_hold);
> +	if (!xfs_buf_trylock(bp)) {
> +		if (smi->retries > XFS_SCRUB_METABUFS_TOO_MANY_RETRIES) {
> +			/* We've retried too many times, do what we can. */
> +			XFS_SCRUB_INCOMPLETE(smi->sc, "metabufs", true);
> +			error = 0;
> +		} else {
> +			/* Restart the metabuf scrub from the start. */
> +			smi->retries++;
> +			error = -EAGAIN;
> +		}
> +		goto out_dec;
> +	}
> +
> +	/* Skip this buffer if it's stale, unread, or has no verifiers. */
> +	if ((bp->b_flags & XBF_STALE) ||
> +	    !(bp->b_flags & XBF_DONE) ||
> +	    !bp->b_ops)
> +		goto out_unlock;
> +
> +	/*
> +	 * Run the verifiers to see if the in-memory buffer is bitrotting or
> +	 * otherwise corrupt.  If the buffer doesn't have a log item then
> +	 * it's clean, so call the read verifier.  However, if the buffer
> +	 * has a log item, it is probably dirty.  Checksums will be written
> +	 * when the buffer is about to go out to disk, so call the write
> +	 * verifier to check the structure.
> +	 */
> +	olderror = bp->b_error;
> +	if (bp->b_fspriv)
> +		bp->b_ops->verify_write(bp);
> +	else
> +		bp->b_ops->verify_read(bp);
> +	error = bp->b_error;
> +	bp->b_error = olderror;
> +
> +	/* Mark any corruption errors we might find. */
> +	XFS_SCRUB_BUF_OP_ERROR_GOTO(out_unlock);
> +
> +out_unlock:
> +	xfs_buf_unlock(bp);
> +out_dec:
> +	atomic_dec(&bp->b_hold);
> +	return error;
> +}
> +#undef XFS_SCRUB_BUF_OP_ERROR_GOTO
> +
> +/* Walk the buffer rhashtable and dispatch buffer checking. */
> +STATIC int
> +xfs_scrub_metabufs_walk_rhash(
> +	struct xfs_scrub_metabufs_info	*smi,
> +	struct rhashtable_iter		*iter)
> +{
> +	struct xfs_buf			*bp;
> +	int				error = 0;
> +
> +	do {
> +		if (xfs_scrub_should_terminate(&error))
> +			break;
> +
> +		bp = rhashtable_walk_next(iter);
> +		if (IS_ERR(bp))
> +			return PTR_ERR(bp);
> +		else if (bp == NULL)
> +			return 0;
> +
> +		error = xfs_scrub_metabufs_scrub_buf(smi, bp);
> +	} while (error != 0);
> +
> +	return error;
> +}
> +
> +/* Try to walk the buffers in this AG in order to scrub them. */
> +int
> +xfs_scrub_metabufs(
> +	struct xfs_scrub_context	*sc)
> +{
> +	struct xfs_scrub_metabufs_info	smi;
> +	struct rhashtable_iter		iter;
> +	struct xfs_perag		*pag;
> +	int				error;
> +
> +	smi.sc = sc;
> +	smi.retries = 0;
> +	pag = xfs_perag_get(sc->mp, sc->sm->sm_agno);
> +	rhashtable_walk_enter(&pag->pag_buf_hash, &iter);
> +
> +	while (1) {
> +		if (xfs_scrub_should_terminate(&error))
> +			break;
> +
> +		error = rhashtable_walk_start(&iter);
> +		if (!error) {
> +			error = xfs_scrub_metabufs_walk_rhash(&smi, &iter);
> +			rhashtable_walk_stop(&iter);
> +		}
> +
> +		if (error != -EAGAIN)
> +			break;
> +		cond_resched();
> +	}
I suppose it's unlikely that we end up looping too many times, but do 
you think we should we have a max number of tries just in case?

Rest of the patch looks good.
Reviewed by: Allison Henderson <allison.henderson@oracle.com>

> +
> +	rhashtable_walk_exit(&iter);
> +	xfs_perag_put(pag);
> +	return error;
> +}
> diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
> index d4de29b..036e65c 100644
> --- a/fs/xfs/xfs_trace.h
> +++ b/fs/xfs/xfs_trace.h
> @@ -3312,7 +3312,8 @@ DEFINE_GETFSMAP_EVENT(xfs_getfsmap_mapping);
>
>  /* scrub */
>  #define XFS_SCRUB_TYPE_DESC \
> -	{ XFS_SCRUB_TYPE_TEST,		"dummy" }
> +	{ XFS_SCRUB_TYPE_TEST,		"dummy" }, \
> +	{ XFS_SCRUB_TYPE_METABUFS,	"metabufs" }
>  DECLARE_EVENT_CLASS(xfs_scrub_class,
>  	TP_PROTO(struct xfs_inode *ip, struct xfs_scrub_metadata *sm,
>  		 int error),
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>