qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: "Denis V. Lunev" <den@virtuozzo.com>
To: Alexander Ivanov <alexander.ivanov@virtuozzo.com>, qemu-devel@nongnu.org
Cc: qemu-block@nongnu.org, stefanha@redhat.com,
	vsementsov@yandex-team.ru, kwolf@redhat.com, hreitz@redhat.com
Subject: Re: [PATCH v2 3/5] parallels: Add checking and repairing duplicate offsets in BAT
Date: Tue, 31 Jan 2023 16:45:09 +0100	[thread overview]
Message-ID: <d3a981c4-8fe6-3b1c-31c3-ab4abf319e8d@virtuozzo.com> (raw)
In-Reply-To: <20230112150149.1452033-4-alexander.ivanov@virtuozzo.com>

On 1/12/23 16:01, Alexander Ivanov wrote:
> Cluster offsets must be unique among all the BAT entries. Find duplicate
> offsets in the BAT and fix it by copying the content of the relevant
> cluster to a newly allocated cluster and set the new cluster offset to the
> duplicated entry.
>
> Add host_cluster_index() and highest_offset() helpers to deduplicate the
> code.
>
> Move parallels_fix_leak() call to parallels_co_check() to fix both types
> of leak: real corruption and a leak produced by allocate_clusters()
> during deduplication.
>
> Signed-off-by: Alexander Ivanov <alexander.ivanov@virtuozzo.com>
> ---
>   block/parallels.c | 168 +++++++++++++++++++++++++++++++++++++++++-----
>   1 file changed, 151 insertions(+), 17 deletions(-)
>
> diff --git a/block/parallels.c b/block/parallels.c
> index da1e75096c..73e992875a 100644
> --- a/block/parallels.c
> +++ b/block/parallels.c
> @@ -136,6 +136,26 @@ static int cluster_remainder(BDRVParallelsState *s, int64_t sector_num,
>       return MIN(nb_sectors, ret);
>   }
>   
> +static uint32_t host_cluster_index(BDRVParallelsState *s, int64_t off)
> +{
> +    off -= s->header->data_off << BDRV_SECTOR_BITS;
> +    return off / s->cluster_size;
> +}
> +
> +static int64_t highest_offset(BDRVParallelsState *s)
> +{
> +    int64_t off, high_off = 0;
> +    int i;
> +
> +    for (i = 0; i < s->bat_size; i++) {
> +        off = bat2sect(s, i) << BDRV_SECTOR_BITS;
> +        if (off > high_off) {
> +            high_off = off;
> +        }
> +    }
> +    return high_off;
> +}
> +
>   static int64_t block_status(BDRVParallelsState *s, int64_t sector_num,
>                               int nb_sectors, int *pnum)
>   {
> @@ -518,17 +538,9 @@ static int parallels_check_leak(BlockDriverState *bs,
>                                   BdrvCheckMode fix)
>   {
>       BDRVParallelsState *s = bs->opaque;
> -    int64_t off, high_off, count, leak_size;
> -    uint32_t i;
> -    int ret;
> +    int64_t high_off, count, leak_size;
>   
> -    high_off = 0;
> -    for (i = 0; i < s->bat_size; i++) {
> -        off = bat2sect(s, i) << BDRV_SECTOR_BITS;
> -        if (off > high_off) {
> -            high_off = off;
> -        }
> -    }
> +    high_off = highest_offset(s);
>   
>       res->image_end_offset = high_off + s->cluster_size;
>   
> @@ -541,13 +553,6 @@ static int parallels_check_leak(BlockDriverState *bs,
>           return 0;
>       }
>   
> -    if (fix & BDRV_FIX_LEAKS) {
> -        ret = parallels_fix_leak(bs, res);
> -        if (ret < 0) {
> -            return ret;
> -        }
> -    }
> -
>       count = DIV_ROUND_UP(leak_size, s->cluster_size);
>       fprintf(stderr, "%s space leaked at the end of the image %" PRId64 "\n",
>               fix & BDRV_FIX_LEAKS ? "Repairing" : "ERROR", leak_size);
> @@ -560,6 +565,122 @@ static int parallels_check_leak(BlockDriverState *bs,
>       return 0;
>   }
>   
> +static int parallels_check_duplicate(BlockDriverState *bs,
> +                                     BdrvCheckResult *res,
> +                                     BdrvCheckMode *fix)
> +{
> +    BDRVParallelsState *s = bs->opaque;
> +    QEMUIOVector qiov;
> +    int64_t off, high_off, sector;
> +    unsigned long *bitmap;
> +    uint32_t i, bitmap_size, cluster_index;
> +    int n, ret = 0;
> +    uint64_t *buf = NULL;
> +
> +    high_off = highest_offset(s);
> +    if (high_off == 0) {
> +        return 0;
> +    }
> +
> +    /*
> +     * Create a bitmap of used clusters.
> +     * If a bit is set, there is a BAT entry pointing to this cluster.
> +     * Loop through the BAT entries, check bits relevant to an entry offset.
> +     * If bit is set, this entry is duplicated. Otherwise set the bit.
> +     *
> +     * We shouldn't worry about newly allocated clusters outside the image
> +     * because they are created higher then any existing cluster pointed by
> +     * a BAT entry.
> +     */
> +    bitmap_size = host_cluster_index(s, high_off) + 1;
> +    bitmap = bitmap_new(bitmap_size);
> +
> +    buf = qemu_memalign(4096, s->cluster_size);
> +    qemu_iovec_init(&qiov, 0);
> +    qemu_iovec_add(&qiov, buf, s->cluster_size);
> +
> +    for (i = 0; i < s->bat_size; i++) {
> +        off = bat2sect(s, i) << BDRV_SECTOR_BITS;
> +        if (off == 0) {
> +            continue;
> +        }
> +
> +        cluster_index = host_cluster_index(s, off);
> +        if (test_bit(cluster_index, bitmap)) {
> +            /* this cluster duplicates another one */
> +            fprintf(stderr,
> +                    "%s duplicate offset in BAT entry %u\n",
> +                    *fix & BDRV_FIX_ERRORS ? "Repairing" : "ERROR", i);
> +
> +            res->corruptions++;
> +
> +            if (*fix & BDRV_FIX_ERRORS) {
> +                /*
> +                 * Reset the entry and allocate a new cluster
> +                 * for the relevant guest offset. In this way we let
> +                 * the lower layer to place the new cluster properly.
> +                 * Copy the original cluster to the allocated one.
> +                 */
> +                parallels_set_bat_entry(s, i, 0);
> +
> +                ret = bdrv_co_pread(bs->file, off, s->cluster_size, buf, 0);
> +                if (ret < 0) {
> +                    res->check_errors++;
> +                    goto out;
> +                }
> +
> +                sector = (i * s->cluster_size) >> BDRV_SECTOR_BITS;
> +                sector = allocate_clusters(bs, sector, s->tracks, &n);
> +                if (sector < 0) {
> +                    res->check_errors++;
> +                    ret = sector;
> +                    goto out;
> +                }
> +                off = sector << BDRV_SECTOR_BITS;
> +                if (off > high_off) {
> +                    high_off = off;
> +                }
> +
> +                ret = bdrv_co_pwritev(bs->file, off, s->cluster_size, &qiov, 0);
> +                if (ret < 0) {
> +                    res->check_errors++;
> +                    goto out;
> +                }
> +
> +                /*
> +                 * In the future allocate_cluster() will reuse holed offsets
> +                 * inside the image. Keep the used clusters bitmap content
> +                 * consistent for the new allocated clusters too.
> +                 *
> +                 * Note, clusters allocated outside the current image are not
> +                 * considered, and the bitmap size doesn't change.
> +                 */
> +                cluster_index = host_cluster_index(s, off);
> +                if (cluster_index < bitmap_size) {
> +                    bitmap_set(bitmap, cluster_index, 1);
> +                }
> +
> +                /*
> +                 * When new clusters are allocated, file size increases by
> +                 * 128 Mb blocks. We need to truncate the file to the right
> +                 * size. Let the leak fix code make its job.
> +                 */
> +                *fix |= BDRV_FIX_LEAKS;
> +                res->corruptions_fixed++;
> +            }
> +            res->image_end_offset = high_off + s->cluster_size;
> +        } else {
> +            bitmap_set(bitmap, cluster_index, 1);
> +        }
> +    }
> +
> +out:
> +    qemu_iovec_destroy(&qiov);
> +    g_free(buf);
> +    g_free(bitmap);
> +    return ret;
> +}
> +
>   static void parallels_collect_statistics(BlockDriverState *bs,
>                                            BdrvCheckResult *res,
>                                            BdrvCheckMode fix)
> @@ -608,7 +729,20 @@ static int coroutine_fn parallels_co_check(BlockDriverState *bs,
>               return ret;
>           }
>   
> +        ret = parallels_check_duplicate(bs, res, &fix);
> +        if (ret < 0) {
> +            return ret;
> +        }
> +
>           parallels_collect_statistics(bs, res, fix);
> +
> +        if (fix & BDRV_FIX_LEAKS &&
> +            (res->corruptions_fixed || res->leaks_fixed)) {
> +            ret = parallels_fix_leak(bs, res);
> +            if (ret < 0) {
> +                return ret;
> +            }
> +        }
>       }
>   
>       ret = bdrv_co_flush(bs);
I would be more happy if this patch will be split - helpers creation
is better to be separated from functional changes.

Den


  reply	other threads:[~2023-01-31 15:46 UTC|newest]

Thread overview: 12+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-01-12 15:01 [PATCH v2 0/5] parallels: Add duplication check, repair at open, fix bugs Alexander Ivanov
2023-01-12 15:01 ` [PATCH v2 1/5] parallels: Incorrect data end calculation in parallels_open() Alexander Ivanov
2023-01-12 15:01 ` [PATCH v2 2/5] parallels: Split image leak handling to separate check and fix helpers Alexander Ivanov
2023-01-31 10:02   ` Denis V. Lunev
2023-01-12 15:01 ` [PATCH v2 3/5] parallels: Add checking and repairing duplicate offsets in BAT Alexander Ivanov
2023-01-31 15:45   ` Denis V. Lunev [this message]
2023-01-12 15:01 ` [PATCH v2 4/5] parallels: Replace fprintf by qemu_log in check Alexander Ivanov
2023-01-31 15:45   ` Denis V. Lunev
2023-01-12 15:01 ` [PATCH v2 5/5] parallels: Image repairing in parallels_open() Alexander Ivanov
2023-01-31 15:50   ` Denis V. Lunev
2023-01-31 17:41   ` Denis V. Lunev
2023-01-15 16:03 ` [PATCH v2 0/5] parallels: Add duplication check, repair at open, fix bugs Alexander Ivanov

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=d3a981c4-8fe6-3b1c-31c3-ab4abf319e8d@virtuozzo.com \
    --to=den@virtuozzo.com \
    --cc=alexander.ivanov@virtuozzo.com \
    --cc=hreitz@redhat.com \
    --cc=kwolf@redhat.com \
    --cc=qemu-block@nongnu.org \
    --cc=qemu-devel@nongnu.org \
    --cc=stefanha@redhat.com \
    --cc=vsementsov@yandex-team.ru \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).