[Qemu-devel] [PATCH] blk: Add discard=sparse mode

[Samuel Thibault]

By default, on discard requests, the posix block backend punches holes but
re-fallocates them to keep the allocated size intact. In some situations
it is however convenient, when using sparse disk images, to see disk image
sizes shrink on discard requests.

This commit adds a discard=sparse mode which does this, by disabling the
fallocate call.

Signed-off-by: Samuel Thibault <address@hidden>

diff --git a/block.c b/block.c
index b663204f3f..e9cd83210a 100644
--- a/block.c
+++ b/block.c
@@ -665,12 +665,14 @@ static void bdrv_join_options(BlockDriverState *bs, QDict 
*options,
  */
 int bdrv_parse_discard_flags(const char *mode, int *flags)
 {
-    *flags &= ~BDRV_O_UNMAP;
+    *flags &= ~(BDRV_O_UNMAP | BDRV_O_SPARSE);
 
     if (!strcmp(mode, "off") || !strcmp(mode, "ignore")) {
         /* do nothing */
     } else if (!strcmp(mode, "on") || !strcmp(mode, "unmap")) {
         *flags |= BDRV_O_UNMAP;
+    } else if (!strcmp(mode, "sparse")) {
+        *flags |= BDRV_O_UNMAP | BDRV_O_SPARSE;
     } else {
         return -1;
     }
diff --git a/block/file-posix.c b/block/file-posix.c
index 4de1abd023..f9efadc5be 100644
--- a/block/file-posix.c
+++ b/block/file-posix.c
@@ -1057,6 +1057,10 @@ static ssize_t handle_aiocb_write_zeroes(RawPosixAIOData 
*aiocb)
     BDRVRawState *s = aiocb->bs->opaque;
 #endif
 
+#if defined(CONFIG_FALLOCATE_PUNCH_HOLE) || defined(CONFIG_FALLOCATE)
+    int open_flags = aiocb->bs->open_flags;
+#endif
+
     if (aiocb->aio_type & QEMU_AIO_BLKDEV) {
         return handle_aiocb_write_zeroes_block(aiocb);
     }
@@ -1079,7 +1083,7 @@ static ssize_t handle_aiocb_write_zeroes(RawPosixAIOData 
*aiocb)
 #endif
 
 #ifdef CONFIG_FALLOCATE_PUNCH_HOLE
-    if (s->has_discard && s->has_fallocate) {
+    if (s->has_discard && (s->has_fallocate || open_flags & BDRV_O_SPARSE)) {
         int ret = do_fallocate(s->fd,
                                FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE,
                                aiocb->aio_offset, aiocb->aio_nbytes);
@@ -1098,7 +1102,8 @@ static ssize_t handle_aiocb_write_zeroes(RawPosixAIOData 
*aiocb)
 #endif
 
 #ifdef CONFIG_FALLOCATE
-    if (s->has_fallocate && aiocb->aio_offset >= bdrv_getlength(aiocb->bs)) {
+    if (s->has_fallocate && !(open_flags & BDRV_O_SPARSE)
+            && aiocb->aio_offset >= bdrv_getlength(aiocb->bs)) {
         int ret = do_fallocate(s->fd, 0, aiocb->aio_offset, aiocb->aio_nbytes);
         if (ret == 0 || ret != -ENOTSUP) {
             return ret;
diff --git a/include/block/block.h b/include/block/block.h
index bde5ebda18..103313bee0 100644
--- a/include/block/block.h
+++ b/include/block/block.h
@@ -97,6 +97,8 @@ typedef struct HDGeometry {
                                       select an appropriate protocol driver,
                                       ignoring the format layer */
 #define BDRV_O_NO_IO       0x10000 /* don't initialize for I/O */
+#define BDRV_O_SPARSE      0x20000 /* make guest UNMAP/TRIM operations make 
image
+                                      possibly sparse */
 
 #define BDRV_O_CACHE_MASK  (BDRV_O_NOCACHE | BDRV_O_NO_FLUSH)
 
diff --git a/qemu-nbd.texi b/qemu-nbd.texi
index 9a84e81eed..8df4f58ddc 100644
--- a/qemu-nbd.texi
+++ b/qemu-nbd.texi
@@ -63,8 +63,8 @@ and @samp{native} (Linux only).
 @item address@hidden
 Control whether @dfn{discard} (also known as @dfn{trim} or @dfn{unmap})
 requests are ignored or passed to the filesystem.  @var{discard} is one of
address@hidden (or @samp{off}), @samp{unmap} (or @samp{on}).  The default is
address@hidden
address@hidden (or @samp{off}), @samp{unmap} (or @samp{on}), or @samp{sparse}.
+The default is @samp{ignore}.
 @item address@hidden
 Control the automatic conversion of plain zero writes by the OS to
 driver-specific optimized zero write commands.  @var{detect-zeroes} is one of
diff --git a/qemu-options.hx b/qemu-options.hx
index bf458f83c3..f5c7455fd1 100644
--- a/qemu-options.hx
+++ b/qemu-options.hx
@@ -539,7 +539,7 @@ DEF("drive", HAS_ARG, QEMU_OPTION_drive,
     "       [,serial=s][,addr=A][,rerror=ignore|stop|report]\n"
     "       
[,werror=ignore|stop|report|enospc][,id=name][,aio=threads|native]\n"
     "       [,readonly=on|off][,copy-on-read=on|off]\n"
-    "       [,discard=ignore|unmap][,detect-zeroes=on|off|unmap]\n"
+    "       [,discard=ignore|unmap|sparse][,detect-zeroes=on|off|unmap]\n"
     "       [[,bps=b]|[[,bps_rd=r][,bps_wr=w]]]\n"
     "       [[,iops=i]|[[,iops_rd=r][,iops_wr=w]]]\n"
     "       [[,bps_max=bm]|[[,bps_rd_max=rm][,bps_wr_max=wm]]]\n"
@@ -582,7 +582,7 @@ These options have the same definition as they have in 
@option{-hdachs}.
 @item address@hidden
 @var{aio} is "threads", or "native" and selects between pthread based disk I/O 
and native Linux AIO.
 @item address@hidden
address@hidden is one of "ignore" (or "off") or "unmap" (or "on") and controls 
whether @dfn{discard} (also known as @dfn{trim} or @dfn{unmap}) requests are 
ignored or passed to the filesystem.  Some machine types may not support 
discard requests.
address@hidden is one of "ignore" (or "off") or "unmap" (or "on") or "sparse" 
and controls whether @dfn{discard} (also known as @dfn{trim} or @dfn{unmap}) 
requests are ignored or passed to the filesystem.  Some machine types may not 
support discard requests.
 @item address@hidden
 Specify which disk @var{format} will be used rather than detecting
 the format.  Can be used to specify format=raw to avoid interpreting