qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Jeff Cody <jcody@redhat.com>
To: Fam Zheng <famz@redhat.com>
Cc: kwolf@redhat.com, hbrock@redhat.com, qemu-devel@nongnu.org,
	rjones@redhat.com, imain@redhat.com, stefanha@redhat.com,
	pbonzini@redhat.com
Subject: Re: [Qemu-devel] [PATCH v2 08/11] block: simplify bdrv_drop_intermediate
Date: Wed, 24 Jul 2013 19:16:02 -0400	[thread overview]
Message-ID: <20130724231602.GA19811@localhost.localdomain> (raw)
In-Reply-To: <1374054136-28741-9-git-send-email-famz@redhat.com>

On Wed, Jul 17, 2013 at 05:42:13PM +0800, Fam Zheng wrote:
> bdrv_drop_intermediate used a local list to iterate through backing
> chain and delete each BDS. It is simplified while adopting to refcount
> mechanism.
> 

Hi Fam,

The reason for the local list is to keep the BDS deletion
transactional, so it can be rolled back in case of error (see below)

> Signed-off-by: Fam Zheng <famz@redhat.com>
> ---
>  block.c | 71 ++++++++++-------------------------------------------------------
>  1 file changed, 11 insertions(+), 60 deletions(-)
> 
> diff --git a/block.c b/block.c
> index 57a3876..499de22 100644
> --- a/block.c
> +++ b/block.c
> @@ -2027,12 +2027,6 @@ BlockDriverState *bdrv_find_overlay(BlockDriverState *active,
>      return overlay;
>  }
>  
> -typedef struct BlkIntermediateStates {
> -    BlockDriverState *bs;
> -    QSIMPLEQ_ENTRY(BlkIntermediateStates) entry;
> -} BlkIntermediateStates;
> -
> -
>  /*
>   * Drops images above 'base' up to and including 'top', and sets the image
>   * above 'top' to have base as its backing file.
> @@ -2062,15 +2056,9 @@ typedef struct BlkIntermediateStates {
>  int bdrv_drop_intermediate(BlockDriverState *active, BlockDriverState *top,
>                             BlockDriverState *base)
>  {
> -    BlockDriverState *intermediate;
> -    BlockDriverState *base_bs = NULL;
>      BlockDriverState *new_top_bs = NULL;
> -    BlkIntermediateStates *intermediate_state, *next;
>      int ret = -EIO;
>  
> -    QSIMPLEQ_HEAD(states_to_delete, BlkIntermediateStates) states_to_delete;
> -    QSIMPLEQ_INIT(&states_to_delete);
> -
>      if (!top->drv || !base->drv) {
>          goto exit;
>      }
> @@ -2082,58 +2070,21 @@ int bdrv_drop_intermediate(BlockDriverState *active, BlockDriverState *top,
>          goto exit;
>      }
>  
> -    /* special case of new_top_bs->backing_hd already pointing to base - nothing
> -     * to do, no intermediate images */
> -    if (new_top_bs->backing_hd == base) {
> -        ret = 0;
> -        goto exit;
> -    }
> -
> -    intermediate = top;
> -
> -    /* now we will go down through the list, and add each BDS we find
> -     * into our deletion queue, until we hit the 'base'
> -     */
> -    while (intermediate) {
> -        intermediate_state = g_malloc0(sizeof(BlkIntermediateStates));
> -        intermediate_state->bs = intermediate;
> -        QSIMPLEQ_INSERT_TAIL(&states_to_delete, intermediate_state, entry);
> -
> -        if (intermediate->backing_hd == base) {
> -            base_bs = intermediate->backing_hd;
> -            break;
> +    while (new_top_bs->backing_hd && new_top_bs->backing_hd != base) {
> +        BlockDriverState *backing = new_top_bs->backing_hd;
> +        if (backing == NULL) {
> +            goto exit;

If you simplify it until just a while loop that unrefs/deletes the BDS
inside the loop as you navigate the chain, then any error exit leaves
you in a bad state, with a potentially invalid chain.  This is one
such error potential.

>          }
> -        intermediate = intermediate->backing_hd;
> -    }
> -    if (base_bs == NULL) {
> -        /* something went wrong, we did not end at the base. safely
> -         * unravel everything, and exit with error */
> -        goto exit;
> +        new_top_bs->backing_hd = backing->backing_hd;
> +        /* break backing_hd chain before releasing bs, so we don't free all the
> +         * way up the backing chain */
> +        backing->backing_hd = NULL;
> +        bdrv_unref(backing, false);

These two statements, which unlink this BDS from the chain, can't be
undone now, in case of error.

>      }
>  
> -    /* success - we can delete the intermediate states, and link top->base */
> -    ret = bdrv_change_backing_file(new_top_bs, base_bs->filename,
> -                                   base_bs->drv ? base_bs->drv->format_name : "");
> -    if (ret) {
> -        goto exit;
> -    }
> -    if (new_top_bs->backing_hd) {
> -        bdrv_unref(new_top_bs->backing_hd, false);
> -    }
> -    new_top_bs->backing_hd = base_bs;
> -    bdrv_ref(base_bs, false);
> -
> -    QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) {
> -        /* so that bdrv_close() does not recursively close the chain */
> -        intermediate_state->bs->backing_hd = NULL;
> -        bdrv_delete(intermediate_state->bs);
> -    }

The foreach loop over the list was placed such that there were no more
error paths; we were guaranteed at this point to have been able delete
and unchain each intermediate BDS.

> -    ret = 0;
> -
> +    ret = bdrv_change_backing_file(new_top_bs, base->filename,
> +                                   base->drv ? base->drv->format_name : "");

This is effectively another error path that would cause problems, if
ret < 0.

>  exit:
> -    QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) {
> -        g_free(intermediate_state);
> -    }
>      return ret;
>  }
>  
> -- 
> 1.8.3.2
> 
> 

  reply	other threads:[~2013-07-24 23:16 UTC|newest]

Thread overview: 46+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2013-07-17  9:42 [Qemu-devel] [PATCH v2 00/11] Point-in-time snapshot exporting over NBD Fam Zheng
2013-07-17  9:42 ` [Qemu-devel] [PATCH v2 01/11] block: replace in_use with refcnt_soft and refcnt_hard Fam Zheng
2013-07-17 12:26   ` Paolo Bonzini
2013-07-18  4:53     ` Fam Zheng
2013-07-23  9:36   ` Stefan Hajnoczi
2013-07-23 10:32     ` Fam Zheng
2013-07-23 13:34       ` Stefan Hajnoczi
2013-07-24  0:39         ` Fam Zheng
2013-07-24  7:35           ` Stefan Hajnoczi
2013-07-24  7:44             ` Fam Zheng
2013-07-25  7:52               ` Stefan Hajnoczi
2013-07-17  9:42 ` [Qemu-devel] [PATCH v2 02/11] block: use refcnt for bs->backing_hd and bs->file Fam Zheng
2013-07-17  9:42 ` [Qemu-devel] [PATCH v2 03/11] block: use refcnt for drive_init/drive_uninit Fam Zheng
2013-07-17  9:42 ` [Qemu-devel] [PATCH v2 04/11] block: use refcnt for device attach/detach Fam Zheng
2013-07-23  9:44   ` Stefan Hajnoczi
2013-07-17  9:42 ` [Qemu-devel] [PATCH v2 05/11] migration: omit drive ref as we have bdrv_ref now Fam Zheng
2013-07-23  9:49   ` Stefan Hajnoczi
2013-07-17  9:42 ` [Qemu-devel] [PATCH v2 06/11] xen_disk: simplify blk_disconnect with refcnt Fam Zheng
2013-07-23  9:50   ` Stefan Hajnoczi
2013-07-17  9:42 ` [Qemu-devel] [PATCH v2 07/11] block: hold hard reference for backup/mirror target Fam Zheng
2013-07-23  9:52   ` Stefan Hajnoczi
2013-07-25  6:08     ` Fam Zheng
2013-07-25  7:59       ` Stefan Hajnoczi
2013-07-17  9:42 ` [Qemu-devel] [PATCH v2 08/11] block: simplify bdrv_drop_intermediate Fam Zheng
2013-07-24 23:16   ` Jeff Cody [this message]
2013-07-25  1:34     ` Fam Zheng
2013-07-17  9:42 ` [Qemu-devel] [PATCH v2 09/11] block: add assertion to check refcount before deleting Fam Zheng
2013-07-17  9:42 ` [Qemu-devel] [PATCH v2 10/11] block: add option 'backing' to -drive options Fam Zheng
2013-07-17 12:36   ` Paolo Bonzini
2013-07-17 12:58     ` Kevin Wolf
2013-07-17 13:13       ` Paolo Bonzini
2013-07-17 13:48         ` Kevin Wolf
2013-07-17 14:16           ` Paolo Bonzini
2013-07-17 15:09             ` Kevin Wolf
2013-07-17 15:23               ` Paolo Bonzini
2013-07-23 20:07               ` Ian Main
2013-07-22  6:07     ` Fam Zheng
2013-07-23 19:57       ` Ian Main
2013-07-17  9:42 ` [Qemu-devel] [PATCH v2 11/11] qmp: add command 'blockdev-backup' Fam Zheng
2013-07-17 12:44   ` Eric Blake
2013-07-18  4:41     ` Fam Zheng
2013-07-19 10:16       ` Wenchao Xia
2013-07-23 10:10         ` Stefan Hajnoczi
2013-07-19 10:41 ` [Qemu-devel] [PATCH v2 00/11] Point-in-time snapshot exporting over NBD Wenchao Xia
2013-07-23  1:52   ` Wenchao Xia
2013-07-23  6:35     ` Paolo Bonzini

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20130724231602.GA19811@localhost.localdomain \
    --to=jcody@redhat.com \
    --cc=famz@redhat.com \
    --cc=hbrock@redhat.com \
    --cc=imain@redhat.com \
    --cc=kwolf@redhat.com \
    --cc=pbonzini@redhat.com \
    --cc=qemu-devel@nongnu.org \
    --cc=rjones@redhat.com \
    --cc=stefanha@redhat.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).