Skip to contentSkip to conversation

qemu-io: Fix 'map' output

PATCH In review Kevin Wolf 2 participants 4 messages May 15, 2013 – May 16, 2013

lore
Share & tools
lore

Kevin Wolf

author
#0 lore raw

PATCH 1/1 qemu-io: Fix 'map' output

The output of the 'map' command in qemu-io used to directly resemble bdrv_is_allocated() and could contain many lines for small chunks that all have the same allocation status. After this patch, they will be coalesced into a single output line for a large chunk.

As a side effect, the command gains some error handling.

---
 qemu-io.c | 46 +++++++++++++++++++++++++++++++++++++++++-----
 1 file changed, 41 insertions(+), 5 deletions(-)
qemu-io.c +41 -5
index 475a8bd..5e6680b 100644
@@ -1635,12 +1635,43 @@ static const cmdinfo_t alloc_cmd = {
     .oneline    = "checks if a sector is present in the file",
 };
 
+
+static int map_is_allocated(int64_t sector_num, int64_t nb_sectors, int64_t *pnum)
+{
+    int num, num_checked;
+    int ret, firstret;
+
+    num_checked = MIN(nb_sectors, INT_MAX);
+    ret = bdrv_is_allocated(bs, sector_num, num_checked, &num);
+    if (ret < 0) {
+        return ret;
+    }
+
+    firstret = ret;
+    *pnum = num;
+
+    while (nb_sectors > 0 && ret == firstret) {
+        sector_num += num;
+        nb_sectors -= num;
+
+        num_checked = MIN(nb_sectors, INT_MAX);
+        ret = bdrv_is_allocated(bs, sector_num, num_checked, &num);
+        if (ret == firstret) {
+            *pnum += num;
+        } else {
+            break;
+        }
+    }
+
+    return firstret;
+}
+
 static int map_f(int argc, char **argv)
 {
     int64_t offset;
     int64_t nb_sectors;
     char s1[64];
-    int num, num_checked;
+    int64_t num;
     int ret;
     const char *retstr;
 
@@ -1648,12 +1679,17 @@ static int map_f(int argc, char **argv)
     nb_sectors = bs->total_sectors;
 
     do {
-        num_checked = MIN(nb_sectors, INT_MAX);
-        ret = bdrv_is_allocated(bs, offset, num_checked, &num);
+        ret = map_is_allocated(offset, nb_sectors, &num);
+        if (ret < 0) {
+            error_report("Failed to get allocation status: %s", strerror(-ret));
+            return 0;
+        }
+
         retstr = ret ? "    allocated" : "not allocated";
         cvtstr(offset << 9ULL, s1, sizeof(s1));
-        printf("[% 24" PRId64 "] % 8d/% 8d sectors %s at offset %s (%d)\n",
-               offset << 9ULL, num, num_checked, retstr, s1, ret);
+        printf("[% 24" PRId64 "] % 8" PRId64 "/% 8" PRId64 " sectors %s "
+               "at offset %s (%d)\n",
+               offset << 9ULL, num, nb_sectors, retstr, s1, ret);
 
         offset += num;
         nb_sectors -= num;
Signature · 1 line
1.8.1.4
#1 lore raw
Quoted 15 lines · Kevin Wolf
qemu-io.c +15 -0
+static int map_is_allocated(int64_t sector_num, int64_t nb_sectors, int64_t *pnum)
+{
+    int num, num_checked;
+    int ret, firstret;
+
+    num_checked = MIN(nb_sectors, INT_MAX);
+    ret = bdrv_is_allocated(bs, sector_num, num_checked, &num);
+    if (ret < 0) {
+        return ret;
+    }
+
+    firstret = ret;
+    *pnum = num;
+
+    while (nb_sectors > 0 && ret == firstret) {

qemu-io.c static const cmdinfo_t alloc_cmd = { · line 1663

Quoted 10 lines
qemu-io.c +10 -0
+        sector_num += num;
+        nb_sectors -= num;
+
+        num_checked = MIN(nb_sectors, INT_MAX);
+        ret = bdrv_is_allocated(bs, sector_num, num_checked, &num);
+        if (ret == firstret) {
+            *pnum += num;
+        } else {
+            break;
+        }

The break makes && ret == firstret redundant above. I suggest just while (nb_sectors > 0) { ... } which is easier to read.

Also, if you respin the patch please tweak the commit message. "Coalesce 'map' output" is more specific than "Fix 'map' output" - unless this really fixes a bug which you didn't mention in the commit description.

author
#2 lore raw

Am 16.05.2013 um 11:14 hat Stefan Hajnoczi geschrieben:

Quoted 21 lines · Kevin Wolf
Quoted 20 lines
qemu-io.c +20 -0
+static int map_is_allocated(int64_t sector_num, int64_t nb_sectors, int64_t *pnum)
+{
+    int num, num_checked;
+    int ret, firstret;
+
+    num_checked = MIN(nb_sectors, INT_MAX);
+    ret = bdrv_is_allocated(bs, sector_num, num_checked, &num);
+    if (ret < 0) {
+        return ret;
+    }
+
+    firstret = ret;
+    *pnum = num;
+
+    while (nb_sectors > 0 && ret == firstret) {
+        sector_num += num;
+        nb_sectors -= num;
+
+        num_checked = MIN(nb_sectors, INT_MAX);
+        ret = bdrv_is_allocated(bs, sector_num, num_checked, &num);
Quoted 8 lines

qemu-io.c static const cmdinfo_t alloc_cmd = { · line 1663

Quoted 5 lines
qemu-io.c +5 -0
+        if (ret == firstret) {
+            *pnum += num;
+        } else {
+            break;
+        }

The break makes && ret == firstret redundant above. I suggest just while (nb_sectors > 0) { ... } which is easier to read.

Okay. I wasn't sure which was better. Don't know though how it came that I have both checks now...

Quoted 4 lines

Also, if you respin the patch please tweak the commit message. "Coalesce 'map' output" is more specific than "Fix 'map' output" - unless this really fixes a bug which you didn't mention in the commit description.

I'll change the title. It makes different formats behave the same even if they work in different granularities. I think QED was bitten by this in qemu-iotests somwhere because it could give different results than qcow2, possibly also dependent on timing. Maybe I should mention that as well in the commit message.

Kevin

#3 lore raw
Quoted 39 lines · Kevin Wolf

Am 16.05.2013 um 11:14 hat Stefan Hajnoczi geschrieben:

Quoted 21 lines · Kevin Wolf
Quoted 20 lines
qemu-io.c +20 -0
+static int map_is_allocated(int64_t sector_num, int64_t nb_sectors, int64_t *pnum)
+{
+    int num, num_checked;
+    int ret, firstret;
+
+    num_checked = MIN(nb_sectors, INT_MAX);
+    ret = bdrv_is_allocated(bs, sector_num, num_checked, &num);
+    if (ret < 0) {
+        return ret;
+    }
+
+    firstret = ret;
+    *pnum = num;
+
+    while (nb_sectors > 0 && ret == firstret) {
+        sector_num += num;
+        nb_sectors -= num;
+
+        num_checked = MIN(nb_sectors, INT_MAX);
+        ret = bdrv_is_allocated(bs, sector_num, num_checked, &num);
Quoted 8 lines

qemu-io.c static const cmdinfo_t alloc_cmd = { · line 1663

Quoted 5 lines
qemu-io.c +5 -0
+        if (ret == firstret) {
+            *pnum += num;
+        } else {
+            break;
+        }

The break makes && ret == firstret redundant above. I suggest just while (nb_sectors > 0) { ... } which is easier to read.

Okay. I wasn't sure which was better. Don't know though how it came that I have both checks now...

Quoted 4 lines

Also, if you respin the patch please tweak the commit message. "Coalesce 'map' output" is more specific than "Fix 'map' output" - unless this really fixes a bug which you didn't mention in the commit description.

Quoted 5 lines

I'll change the title. It makes different formats behave the same even if they work in different granularities. I think QED was bitten by this in qemu-iotests somwhere because it could give different results than qcow2, possibly also dependent on timing. Maybe I should mention that as well in the commit message.

Yes, please. I didn't think of that.