diff mbox

[RFC] qemu-img: add support for skipping zeroes in input during convert

Message ID 1385131710-8978-1-git-send-email-pl@kamp.de
State New
Headers show

Commit Message

Peter Lieven Nov. 22, 2013, 2:48 p.m. UTC
Signed-off-by: Peter Lieven <pl@kamp.de>
---
 qemu-img.c |   52 ++++++++++++++++++++++++++++++----------------------
 1 file changed, 30 insertions(+), 22 deletions(-)

Comments

Stefan Hajnoczi Dec. 2, 2013, 2:13 p.m. UTC | #1
On Fri, Nov 22, 2013 at 03:48:30PM +0100, Peter Lieven wrote:
> -            /* If the output image is being created as a copy on write image,
> -               assume that sectors which are unallocated in the input image
> -               are present in both the output's and input's base images (no
> -               need to copy them). */
> -            if (out_baseimg) {
> -                ret = bdrv_is_allocated(bs[bs_i], sector_num - bs_offset,
> -                                        n, &n1);

int coroutine_fn bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num,
                                   int nb_sectors, int *pnum)
{
    int64_t ret = bdrv_get_block_status(bs, sector_num, nb_sectors, pnum);
    if (ret < 0) {
        return ret;
    }
    return
        (ret & BDRV_BLOCK_DATA) ||
        ((ret & BDRV_BLOCK_ZERO) && !bdrv_has_zero_init(bs));
}

bdrv_has_zero_init() returns false when bs->backing_hd != NULL.

> +            if (out_baseimg || has_zero_init) {
> +                n = nb_sectors > INT_MAX ? INT_MAX : nb_sectors;
> +                ret = bdrv_get_block_status(bs[bs_i], sector_num - bs_offset,
> +                                            n, &n1);
>                  if (ret < 0) {
> -                    error_report("error while reading metadata for sector "
> -                                 "%" PRId64 ": %s",
> +                    error_report("error while reading block status of sector %" PRId64 ": %s",
>                                   sector_num - bs_offset, strerror(-ret));
>                      goto out;
>                  }
> -                if (!ret) {
> +                /* If the output image is zero initialized, we are not working
> +                 * on a shared base and the input is zero we can skip the next
> +                 * n1 bytes */
> +                if (!out_baseimg && has_zero_init && ret & BDRV_BLOCK_ZERO) {
> +                    sector_num += n1;
> +                    continue;
> +                }
> +                /* If the output image is being created as a copy on write image,
> +                   assume that sectors which are unallocated in the input image
> +                   are present in both the output's and input's base images (no
> +                   need to copy them). */
> +                if (out_baseimg && !(ret & BDRV_BLOCK_DATA)) {
>                      sector_num += n1;
>                      continue;
>                  }

How are these two if statements different from bdrv_is_allocated()?

Stefan
Peter Lieven Dec. 2, 2013, 2:15 p.m. UTC | #2
Hi Stefan,

please have a look at 

[PATCHv2 1.8 0/9] qemu-img convert optimizations

This one here is obsolete.

Peter

Am 02.12.2013 um 15:13 schrieb Stefan Hajnoczi <stefanha@redhat.com>:

> On Fri, Nov 22, 2013 at 03:48:30PM +0100, Peter Lieven wrote:
>> -            /* If the output image is being created as a copy on write image,
>> -               assume that sectors which are unallocated in the input image
>> -               are present in both the output's and input's base images (no
>> -               need to copy them). */
>> -            if (out_baseimg) {
>> -                ret = bdrv_is_allocated(bs[bs_i], sector_num - bs_offset,
>> -                                        n, &n1);
> 
> int coroutine_fn bdrv_is_allocated(BlockDriverState *bs, int64_t sector_num,
>                                   int nb_sectors, int *pnum)
> {
>    int64_t ret = bdrv_get_block_status(bs, sector_num, nb_sectors, pnum);
>    if (ret < 0) {
>        return ret;
>    }
>    return
>        (ret & BDRV_BLOCK_DATA) ||
>        ((ret & BDRV_BLOCK_ZERO) && !bdrv_has_zero_init(bs));
> }
> 
> bdrv_has_zero_init() returns false when bs->backing_hd != NULL.
> 
>> +            if (out_baseimg || has_zero_init) {
>> +                n = nb_sectors > INT_MAX ? INT_MAX : nb_sectors;
>> +                ret = bdrv_get_block_status(bs[bs_i], sector_num - bs_offset,
>> +                                            n, &n1);
>>                 if (ret < 0) {
>> -                    error_report("error while reading metadata for sector "
>> -                                 "%" PRId64 ": %s",
>> +                    error_report("error while reading block status of sector %" PRId64 ": %s",
>>                                  sector_num - bs_offset, strerror(-ret));
>>                     goto out;
>>                 }
>> -                if (!ret) {
>> +                /* If the output image is zero initialized, we are not working
>> +                 * on a shared base and the input is zero we can skip the next
>> +                 * n1 bytes */
>> +                if (!out_baseimg && has_zero_init && ret & BDRV_BLOCK_ZERO) {
>> +                    sector_num += n1;
>> +                    continue;
>> +                }
>> +                /* If the output image is being created as a copy on write image,
>> +                   assume that sectors which are unallocated in the input image
>> +                   are present in both the output's and input's base images (no
>> +                   need to copy them). */
>> +                if (out_baseimg && !(ret & BDRV_BLOCK_DATA)) {
>>                     sector_num += n1;
>>                     continue;
>>                 }
> 
> How are these two if statements different from bdrv_is_allocated()?
> 
> Stefan
diff mbox

Patch

diff --git a/qemu-img.c b/qemu-img.c
index b6b5644..808f8f8 100644
--- a/qemu-img.c
+++ b/qemu-img.c
@@ -1121,8 +1121,9 @@  out3:
 
 static int img_convert(int argc, char **argv)
 {
-    int c, ret = 0, n, n1, bs_n, bs_i, compress, cluster_size,
+    int c, n, n1, bs_n, bs_i, compress, cluster_size,
         cluster_sectors, skip_create;
+    int64_t ret = 0;
     int progress = 0, flags;
     const char *fmt, *out_fmt, *cache, *out_baseimg, *out_filename;
     BlockDriver *drv, *proto_drv;
@@ -1479,11 +1480,6 @@  static int img_convert(int argc, char **argv)
             if (nb_sectors <= 0) {
                 break;
             }
-            if (nb_sectors >= (IO_BUF_SIZE / 512)) {
-                n = (IO_BUF_SIZE / 512);
-            } else {
-                n = nb_sectors;
-            }
 
             while (sector_num - bs_offset >= bs_sectors) {
                 bs_i ++;
@@ -1495,34 +1491,46 @@  static int img_convert(int argc, char **argv)
                    sector_num, bs_i, bs_offset, bs_sectors); */
             }
 
-            if (n > bs_offset + bs_sectors - sector_num) {
-                n = bs_offset + bs_sectors - sector_num;
-            }
-
-            /* If the output image is being created as a copy on write image,
-               assume that sectors which are unallocated in the input image
-               are present in both the output's and input's base images (no
-               need to copy them). */
-            if (out_baseimg) {
-                ret = bdrv_is_allocated(bs[bs_i], sector_num - bs_offset,
-                                        n, &n1);
+            if (out_baseimg || has_zero_init) {
+                n = nb_sectors > INT_MAX ? INT_MAX : nb_sectors;
+                ret = bdrv_get_block_status(bs[bs_i], sector_num - bs_offset,
+                                            n, &n1);
                 if (ret < 0) {
-                    error_report("error while reading metadata for sector "
-                                 "%" PRId64 ": %s",
+                    error_report("error while reading block status of sector %" PRId64 ": %s",
                                  sector_num - bs_offset, strerror(-ret));
                     goto out;
                 }
-                if (!ret) {
+                /* If the output image is zero initialized, we are not working
+                 * on a shared base and the input is zero we can skip the next
+                 * n1 bytes */
+                if (!out_baseimg && has_zero_init && ret & BDRV_BLOCK_ZERO) {
+                    sector_num += n1;
+                    continue;
+                }
+                /* If the output image is being created as a copy on write image,
+                   assume that sectors which are unallocated in the input image
+                   are present in both the output's and input's base images (no
+                   need to copy them). */
+                if (out_baseimg && !(ret & BDRV_BLOCK_DATA)) {
                     sector_num += n1;
                     continue;
                 }
                 /* The next 'n1' sectors are allocated in the input image. Copy
                    only those as they may be followed by unallocated sectors. */
-                n = n1;
+                nb_sectors = n1;
+            }
+
+            if (nb_sectors >= (IO_BUF_SIZE / 512)) {
+                n = (IO_BUF_SIZE / 512);
             } else {
-                n1 = n;
+                n = nb_sectors;
+            }
+
+            if (n > bs_offset + bs_sectors - sector_num) {
+                n = bs_offset + bs_sectors - sector_num;
             }
 
+            n1 = n;
             ret = bdrv_read(bs[bs_i], sector_num - bs_offset, buf, n);
             if (ret < 0) {
                 error_report("error while reading sector %" PRId64 ": %s",