OSDN Git Service

block: Fix integer promotion error in bdrv_getlength()
authorEric Blake <eblake@redhat.com>
Thu, 5 Nov 2020 15:51:22 +0000 (09:51 -0600)
committerMax Reitz <mreitz@redhat.com>
Mon, 9 Nov 2020 14:44:21 +0000 (15:44 +0100)
Back in 2015, we attempted to fix error reporting for images that
claimed to have more than INT64_MAX/512 sectors, but due to the type
promotions caused by BDRV_SECTOR_SIZE being unsigned, this
inadvertently forces all negative ret values to be slammed into -EFBIG
rather than the original error.  While we're at it, we can avoid the
confusing ?: by spelling the logic more directly.

Fixes: 4a9c9ea0d3
Reported-by: Guoyi Tu <tu.guoyi@h3c.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <20201105155122.60943-1-eblake@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
block.c

diff --git a/block.c b/block.c
index 56bacc9..2fd9321 100644 (file)
--- a/block.c
+++ b/block.c
@@ -5091,8 +5091,13 @@ int64_t bdrv_getlength(BlockDriverState *bs)
 {
     int64_t ret = bdrv_nb_sectors(bs);
 
-    ret = ret > INT64_MAX / BDRV_SECTOR_SIZE ? -EFBIG : ret;
-    return ret < 0 ? ret : ret * BDRV_SECTOR_SIZE;
+    if (ret < 0) {
+        return ret;
+    }
+    if (ret > INT64_MAX / BDRV_SECTOR_SIZE) {
+        return -EFBIG;
+    }
+    return ret * BDRV_SECTOR_SIZE;
 }
 
 /* return 0 as number of sectors if no device present or error */