diff mbox

[v2] sheepdog: fix loadvm operation

Message ID 1366822079-6582-1-git-send-email-namei.unix@gmail.com
State New
Headers show

Commit Message

Liu Yuan April 24, 2013, 4:47 p.m. UTC 
From: Liu Yuan <tailai.ly@taobao.com>

Currently the 'loadvm' opertaion works as following:
1. switch to the snapshot
2. mark current working VDI as a snapshot
3. rely on sd_create_branch to create a new working VDI based on the snapshot

This works not the same as other format as QCOW2. For e.g,

qemu > savevm # get a live snapshot snap1
qemu > savevm # snap2
qemu > loadvm 1 # This will steally create snap3 of the working VDI

Which will result in following snapshot chain:

base <-- snap1 <-- snap2 <-- snap3
          ^
          |
      working VDI

snap3 was unnecessarily created and might be annoying users.

This patch discard the unnecessary 'snap3' creation. and implement
rollback(loadvm) operation to the specified snapshot by
1. switch to the snapshot
2. delete working VDI
3. rely on sd_create_branch to create a new working VDI based on the snapshot

The snapshot chain for above example will be:

base <-- snap1 <-- snap2
          ^
          |
      working VDI

Cc: qemu-devel@nongnu.org
Cc: MORITA Kazutaka <morita.kazutaka@lab.ntt.co.jp>
Cc: Kevin Wolf <kwolf@redhat.com>
Cc: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Liu Yuan <tailai.ly@taobao.com>
---
v2:
 - use do_req() because sd_delete isn't in coroutine
 - don't break old behavior if we boot up on the snapshot by using s->reverted
   to indicate if we delete working VDI successfully
 - fix a subtle case that sd_create_branch() isn't called yet while another
   'loadvm' is executed

 block/sheepdog.c |   52 +++++++++++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 51 insertions(+), 1 deletion(-)

Comments

Liu Yuan April 25, 2013, 4:12 a.m. UTC | #1 
On 04/25/2013 12:47 AM, Liu Yuan wrote:
> - don't break old behavior if we boot up on the snapshot by using s->reverted
>    to indicate if we delete working VDI successfully

If we implement 'boot from snapshot' == loadvm snapshot, we don't need
s->reverted. What do you think, Kazutaka? With this idea, qemu -hda
sheepdog:test:id will be virtually the same as qemu -hda sheepdog:test
-loadvm id.

I don't understand the usecase for snapshting current workding VDI and
then restore to the specified snapshot.

Thanks,
Yuan
MORITA Kazutaka April 25, 2013, 8:11 a.m. UTC | #2 
At Thu, 25 Apr 2013 12:12:58 +0800,
Liu Yuan wrote:
> 
> On 04/25/2013 12:47 AM, Liu Yuan wrote:
> > - don't break old behavior if we boot up on the snapshot by using s->reverted
> >    to indicate if we delete working VDI successfully
> 
> If we implement 'boot from snapshot' == loadvm snapshot, we don't need
> s->reverted. What do you think, Kazutaka? With this idea, qemu -hda
> sheepdog:test:id will be virtually the same as qemu -hda sheepdog:test
> -loadvm id.
> 
> I don't understand the usecase for snapshting current workding VDI and
> then restore to the specified snapshot.

Are you suggesting that booting from snapshot should discard the
current state like savevm?  If yes, it looks okay to me.

Thanks,

Kazutaka
Liu Yuan April 25, 2013, 8:27 a.m. UTC | #3 
On 04/25/2013 04:11 PM, MORITA Kazutaka wrote:
> Are you suggesting that booting from snapshot should discard the
> current state like savevm?  If yes, it looks okay to me.

Yeah, this will make code cleaner. I'll update v3.

Thanks
Yuan
diff mbox

Patch

diff --git a/block/sheepdog.c b/block/sheepdog.c
index 2fe0783..43829a7 100644
--- a/block/sheepdog.c
+++ b/block/sheepdog.c
@@ -36,6 +36,7 @@ 
 #define SD_OP_GET_VDI_INFO   0x14
 #define SD_OP_READ_VDIS      0x15
 #define SD_OP_FLUSH_VDI      0x16
+#define SD_OP_DEL_VDI        0x17
 
 #define SD_FLAG_CMD_WRITE    0x01
 #define SD_FLAG_CMD_COW      0x02
@@ -301,6 +302,7 @@  typedef struct BDRVSheepdogState {
     bool is_snapshot;
     uint32_t cache_flags;
     bool discard_supported;
+    bool reverted;
 
     char *host_spec;
     bool is_unix;
@@ -1553,7 +1555,9 @@  static int sd_create_branch(BDRVSheepdogState *s)
 
     buf = g_malloc(SD_INODE_SIZE);
 
-    ret = do_sd_create(s, s->name, s->inode.vdi_size, s->inode.vdi_id, &vid, 1);
+    /* If reverted, we create a working VDI otherwise go snapshot-create */
+    ret = do_sd_create(s, s->name, s->inode.vdi_size, s->inode.vdi_id, &vid,
+                       !s->reverted);
     if (ret) {
         goto out;
     }
@@ -1578,6 +1582,7 @@  static int sd_create_branch(BDRVSheepdogState *s)
     memcpy(&s->inode, buf, sizeof(s->inode));
 
     s->is_snapshot = false;
+    s->reverted = false;
     ret = 0;
     dprintf("%" PRIx32 " was newly created.\n", s->inode.vdi_id);
 
@@ -1869,6 +1874,41 @@  cleanup:
     return ret;
 }
 
+/* Delete current working VDI by the name */
+static int sd_delete(BDRVSheepdogState *s, char *name)
+{
+    unsigned int wlen = SD_MAX_VDI_LEN, rlen = 0;
+    SheepdogVdiReq hdr = {
+        .opcode = SD_OP_DEL_VDI,
+        .vdi_id = s->inode.vdi_id,
+        .data_length = wlen,
+        .flags = SD_FLAG_CMD_WRITE,
+    };
+    SheepdogVdiRsp *rsp = (SheepdogVdiRsp *)&hdr;
+    int fd, ret;
+
+    fd = connect_to_sdog(s);
+    if (fd < 0) {
+        return fd;
+    }
+
+    ret = do_req(fd, (SheepdogReq *)&hdr, name, &wlen, &rlen);
+    closesocket(fd);
+    if (ret || (rsp->result != SD_RES_SUCCESS &&
+                rsp->result != SD_RES_NO_VDI)) {
+        error_report("%s, %s", sd_strerror(rsp->result), name);
+        return -1;
+    }
+
+    return 0;
+}
+
+/*
+ * We implement rollback(loadvm) operation to the specified snapshot by
+ * 1) switch to the snapshot
+ * 2) delete working VDI
+ * 3) rely on sd_create_branch to create a new working VDI based on the snapshot
+ */
 static int sd_snapshot_goto(BlockDriverState *bs, const char *snapshot_id)
 {
     BDRVSheepdogState *s = bs->opaque;
@@ -1924,6 +1964,16 @@  static int sd_snapshot_goto(BlockDriverState *bs, const char *snapshot_id)
 
     s->is_snapshot = true;
 
+    /*
+     * Even If deletion fails, we will just create extra snapshot based on
+     * the workding VDI which was supposed to be deleted. So no need to
+     * false bail out.
+     */
+    ret = sd_delete(s, vdi);
+    if (!ret) {
+        s->reverted = true;
+    }
+
     g_free(buf);
     g_free(old_s);