qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
* [Qemu-devel] [PATCH v6] qcow2: Buffer L1 table in snapshot refcount update
@ 2014-11-11 15:27 Max Reitz
  2014-11-20 14:32 ` Max Reitz
  2014-11-28 10:29 ` Stefan Hajnoczi
  0 siblings, 2 replies; 4+ messages in thread
From: Max Reitz @ 2014-11-11 15:27 UTC (permalink / raw)
  To: qemu-devel; +Cc: Kevin Wolf, Zhang Haoyu, Stefan Hajnoczi, Max Reitz

From: Zhang Haoyu <zhanghy@sangfor.com>

Buffer the active L1 table in qcow2_update_snapshot_refcount() in order
to prevent in-place conversion of the L1 table buffer in the
BDRVQcowState to big endian and back, which would lead to data
corruption if that buffer was accessed concurrently. This should not
happen but better being safe than sorry.

Signed-off-by: Zhang Haoyu <zhanghy@sangfor.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
---
v6 for "snapshot: use local variable to bdrv_pwrite_sync L1 table" (I
changed the commit message wording to make it more clear what this patch
does and why we want it).

Changes in v6:
- Only copy the local buffer back into s->l1_table if we are indeed
  accessing the local L1 table
- Use qemu_vfree() instead of g_free()
---
 block/qcow2-refcount.c | 30 ++++++++++++++----------------
 1 file changed, 14 insertions(+), 16 deletions(-)

diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c
index 9afdb40..c0c4a50 100644
--- a/block/qcow2-refcount.c
+++ b/block/qcow2-refcount.c
@@ -877,14 +877,18 @@ int qcow2_update_snapshot_refcount(BlockDriverState *bs,
 {
     BDRVQcowState *s = bs->opaque;
     uint64_t *l1_table, *l2_table, l2_offset, offset, l1_size2;
-    bool l1_allocated = false;
+    bool active_l1 = false;
     int64_t old_offset, old_l2_offset;
     int i, j, l1_modified = 0, nb_csectors, refcount;
     int ret;
 
     l2_table = NULL;
-    l1_table = NULL;
     l1_size2 = l1_size * sizeof(uint64_t);
+    l1_table = qemu_try_blockalign(bs->file, l1_size2);
+    if (l1_table == NULL) {
+        ret = -ENOMEM;
+        goto fail;
+    }
 
     s->cache_discards = true;
 
@@ -892,13 +896,6 @@ int qcow2_update_snapshot_refcount(BlockDriverState *bs,
      * l1_table_offset when it is the current s->l1_table_offset! Be careful
      * when changing this! */
     if (l1_table_offset != s->l1_table_offset) {
-        l1_table = g_try_malloc0(align_offset(l1_size2, 512));
-        if (l1_size2 && l1_table == NULL) {
-            ret = -ENOMEM;
-            goto fail;
-        }
-        l1_allocated = true;
-
         ret = bdrv_pread(bs->file, l1_table_offset, l1_table, l1_size2);
         if (ret < 0) {
             goto fail;
@@ -908,8 +905,8 @@ int qcow2_update_snapshot_refcount(BlockDriverState *bs,
             be64_to_cpus(&l1_table[i]);
     } else {
         assert(l1_size == s->l1_size);
-        l1_table = s->l1_table;
-        l1_allocated = false;
+        memcpy(l1_table, s->l1_table, l1_size2);
+        active_l1 = true;
     }
 
     for(i = 0; i < l1_size; i++) {
@@ -1051,13 +1048,14 @@ fail:
         }
 
         ret = bdrv_pwrite_sync(bs->file, l1_table_offset, l1_table, l1_size2);
-
-        for (i = 0; i < l1_size; i++) {
-            be64_to_cpus(&l1_table[i]);
+        if (active_l1 && ret == 0) {
+            for (i = 0; i < l1_size; i++) {
+                be64_to_cpus(&l1_table[i]);
+            }
+            memcpy(s->l1_table, l1_table, l1_size2);
         }
     }
-    if (l1_allocated)
-        g_free(l1_table);
+    qemu_vfree(l1_table);
     return ret;
 }
 
-- 
1.9.3

^ permalink raw reply related	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2014-11-28 10:30 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2014-11-11 15:27 [Qemu-devel] [PATCH v6] qcow2: Buffer L1 table in snapshot refcount update Max Reitz
2014-11-20 14:32 ` Max Reitz
2014-11-27 15:09   ` Max Reitz
2014-11-28 10:29 ` Stefan Hajnoczi

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).