qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH v4 3/5] parallels: Add checking and repairing duplicate offse


From: Alexander Ivanov
Subject: Re: [PATCH v4 3/5] parallels: Add checking and repairing duplicate offsets in BAT
Date: Fri, 28 Apr 2023 13:48:55 +0200
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101 Thunderbird/102.10.0



On 4/27/23 15:46, Mike Maslenkin wrote:
Sorry for the noise again , but I have another note
No, you don't need to apologize. You help me make code better.

On Mon, Apr 24, 2023 at 12:44 PM Alexander Ivanov
<alexander.ivanov@virtuozzo.com> 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() helper 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 | 134 ++++++++++++++++++++++++++++++++++++++++++++--
  1 file changed, 129 insertions(+), 5 deletions(-)

diff --git a/block/parallels.c b/block/parallels.c
index ec89ed894b..3b992e8173 100644
--- a/block/parallels.c
+++ b/block/parallels.c
@@ -136,6 +136,12 @@ 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 block_status(BDRVParallelsState *s, int64_t sector_num,
                              int nb_sectors, int *pnum)
  {
@@ -533,7 +539,6 @@ parallels_check_leak(BlockDriverState *bs, BdrvCheckResult 
*res,
  {
      BDRVParallelsState *s = bs->opaque;
      int64_t count, leak_size;
-    int ret;

      leak_size = parallels_get_leak_size(bs, res);
      if (leak_size < 0) {
@@ -550,16 +555,123 @@ parallels_check_leak(BlockDriverState *bs, 
BdrvCheckResult *res,
              fix & BDRV_FIX_LEAKS ? "Repairing" : "ERROR", leak_size);

      if (fix & BDRV_FIX_LEAKS) {
-        ret = parallels_fix_leak(bs, res);
-        if (ret < 0) {
-            return ret;
-        }
          res->leaks_fixed += count;
      }

      return 0;
  }

+static int parallels_check_duplicate(BlockDriverState *bs,
+                                     BdrvCheckResult *res,
+                                     BdrvCheckMode *fix)
+{
+    BDRVParallelsState *s = bs->opaque;
+    QEMUIOVector qiov;
+    int64_t off, sector;
+    unsigned long *bitmap;
+    uint32_t i, bitmap_size, cluster_index;
+    int n, ret = 0;
+    uint64_t *buf = NULL;
+
+    /*
+     * 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, res->image_end_offset);
+    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;
+                }
I can not understand how index in a BAT table related to s->cluster_size.
Probably there should be "cluster_index" used?
Anyway, looks like both cause uint32 truncation as result of
({i,cluster_index} * s->cluster_size)

Regards,
Mike.

s->cluster_size is the size of cluster in bytes. We shift this value to BDRV_SECTOR_BITS and have the number of sectors. Pay attention that in allocate_clusters() we need not an offset in the image file, but an offset inside image. As for truncation, you are right,
will add a cast to uint64_t.




reply via email to

[Prev in Thread] Current Thread [Next in Thread]