qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Kevin Wolf <kwolf@redhat.com>
To: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Cc: qemu-block@nongnu.org, armbru@redhat.com, qemu-devel@nongnu.org,
	mreitz@redhat.com, den@openvz.org, jsnow@redhat.com
Subject: Re: [PATCH v2 20/36] block: add bdrv_attach_child_common() transaction action
Date: Wed, 3 Feb 2021 22:01:04 +0100	[thread overview]
Message-ID: <20210203210104.GH5507@merkur.fritz.box> (raw)
In-Reply-To: <20201127144522.29991-21-vsementsov@virtuozzo.com>

Am 27.11.2020 um 15:45 hat Vladimir Sementsov-Ogievskiy geschrieben:
> Split out no-perm part of bdrv_root_attach_child() into separate
> transaction action. bdrv_root_attach_child() now moves to new
> permission update paradigm: first update graph relations then update
> permissions.
> 
> Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
> ---
>  block.c | 162 ++++++++++++++++++++++++++++++++++++++++----------------
>  1 file changed, 117 insertions(+), 45 deletions(-)
> 
> diff --git a/block.c b/block.c
> index f0fcd75555..a7ccbb4fb1 100644
> --- a/block.c
> +++ b/block.c
> @@ -86,6 +86,13 @@ static void bdrv_parent_set_aio_context_ignore(BdrvChild *c, AioContext *ctx,
>                                                 GSList **ignore);
>  static void bdrv_replace_child_noperm(BdrvChild *child,
>                                        BlockDriverState *new_bs);
> +static int bdrv_attach_child_common(BlockDriverState *child_bs,
> +                                    const char *child_name,
> +                                    const BdrvChildClass *child_class,
> +                                    BdrvChildRole child_role,
> +                                    uint64_t perm, uint64_t shared_perm,
> +                                    void *opaque, BdrvChild **child,
> +                                    GSList **tran, Error **errp);

If you added the new code above bdrv_root_attach_child(), we wouldn't
need the forward declaration and the patch would probably be simpler to
read (because it's the first part of bdrv_root_attach_child() that is
factored out).

>  static int bdrv_reopen_prepare(BDRVReopenState *reopen_state, BlockReopenQueue
>                                 *queue, Error **errp);
> @@ -2898,55 +2905,22 @@ BdrvChild *bdrv_root_attach_child(BlockDriverState *child_bs,
>                                    uint64_t perm, uint64_t shared_perm,
>                                    void *opaque, Error **errp)
>  {
> -    BdrvChild *child;
> -    Error *local_err = NULL;
>      int ret;
> -    AioContext *ctx;
> +    BdrvChild *child = NULL;
> +    GSList *tran = NULL;
>  
> -    ret = bdrv_check_update_perm(child_bs, NULL, perm, shared_perm, NULL, errp);
> +    ret = bdrv_attach_child_common(child_bs, child_name, child_class,
> +                                   child_role, perm, shared_perm, opaque,
> +                                   &child, &tran, errp);
>      if (ret < 0) {
> -        bdrv_abort_perm_update(child_bs);
>          bdrv_unref(child_bs);
>          return NULL;
>      }
>  
> -    child = g_new(BdrvChild, 1);
> -    *child = (BdrvChild) {
> -        .bs             = NULL,
> -        .name           = g_strdup(child_name),
> -        .klass          = child_class,
> -        .role           = child_role,
> -        .perm           = perm,
> -        .shared_perm    = shared_perm,
> -        .opaque         = opaque,
> -    };
> -
> -    ctx = bdrv_child_get_parent_aio_context(child);
> -
> -    /* If the AioContexts don't match, first try to move the subtree of
> -     * child_bs into the AioContext of the new parent. If this doesn't work,
> -     * try moving the parent into the AioContext of child_bs instead. */
> -    if (bdrv_get_aio_context(child_bs) != ctx) {
> -        ret = bdrv_try_set_aio_context(child_bs, ctx, &local_err);
> -        if (ret < 0) {
> -            if (bdrv_parent_try_set_aio_context(child, ctx, NULL) == 0) {
> -                ret = 0;
> -                error_free(local_err);
> -                local_err = NULL;
> -            }
> -        }
> -        if (ret < 0) {
> -            error_propagate(errp, local_err);
> -            g_free(child);
> -            bdrv_abort_perm_update(child_bs);
> -            bdrv_unref(child_bs);
> -            return NULL;
> -        }
> -    }
> -
> -    /* This performs the matching bdrv_set_perm() for the above check. */
> -    bdrv_replace_child(child, child_bs);
> +    ret = bdrv_refresh_perms(child_bs, errp);
> +    tran_finalize(tran, ret);
>  
> +    bdrv_unref(child_bs);
>      return child;
>  }
>  
> @@ -2988,16 +2962,114 @@ BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs,
>      return child;
>  }
>  
> -static void bdrv_detach_child(BdrvChild *child)
> +static void bdrv_remove_empty_child(BdrvChild *child)
>  {
> +    assert(!child->bs);
>      QLIST_SAFE_REMOVE(child, next);
> -
> -    bdrv_replace_child(child, NULL);
> -
>      g_free(child->name);
>      g_free(child);
>  }
>  
> +typedef struct BdrvAttachChildCommonState {
> +    BdrvChild **child;
> +    AioContext *old_parent_ctx;
> +    AioContext *old_child_ctx;
> +} BdrvAttachChildCommonState;
> +
> +static void bdrv_attach_child_common_abort(void *opaque)
> +{
> +    BdrvAttachChildCommonState *s = opaque;
> +    BdrvChild *child = *s->child;
> +    BlockDriverState *bs = child->bs;
> +
> +    bdrv_replace_child_noperm(child, NULL);
> +
> +    if (bdrv_get_aio_context(bs) != s->old_child_ctx) {
> +        bdrv_try_set_aio_context(bs, s->old_child_ctx, &error_abort);

Would failure actually be fatal? I think we can ignore it, the node is
in an AioContext that works for it.

> +    }
> +
> +    if (bdrv_child_get_parent_aio_context(child) != s->old_parent_ctx) {
> +        bdrv_parent_try_set_aio_context(child, s->old_parent_ctx,
> +                                        &error_abort);

And the same here.

> +    }
> +
> +    bdrv_unref(bs);
> +    bdrv_remove_empty_child(child);
> +    *s->child = NULL;
> +}
> +
> +static TransactionActionDrv bdrv_attach_child_common_drv = {
> +    .abort = bdrv_attach_child_common_abort,
> +};
> +
> +/*
> + * Common part of attoching bdrv child to bs or to blk or to job
> + */
> +static int bdrv_attach_child_common(BlockDriverState *child_bs,
> +                                    const char *child_name,
> +                                    const BdrvChildClass *child_class,
> +                                    BdrvChildRole child_role,
> +                                    uint64_t perm, uint64_t shared_perm,
> +                                    void *opaque, BdrvChild **child,
> +                                    GSList **tran, Error **errp)
> +{
> +    int ret;
> +    BdrvChild *new_child;
> +    AioContext *parent_ctx;
> +    AioContext *child_ctx = bdrv_get_aio_context(child_bs);
> +
> +    assert(child);
> +    assert(*child == NULL);
> +
> +    new_child = g_new(BdrvChild, 1);
> +    *new_child = (BdrvChild) {
> +        .bs             = NULL,
> +        .name           = g_strdup(child_name),
> +        .klass          = child_class,
> +        .role           = child_role,
> +        .perm           = perm,
> +        .shared_perm    = shared_perm,
> +        .opaque         = opaque,
> +    };
> +
> +    parent_ctx = bdrv_child_get_parent_aio_context(new_child);
> +    if (child_ctx != parent_ctx) {
> +        ret = bdrv_try_set_aio_context(child_bs, parent_ctx, NULL);
> +        if (ret < 0) {
> +            /*
> +             * bdrv_try_set_aio_context_tran don't need rollback after failure,
> +             * so we don't care.
> +             */
> +            ret = bdrv_parent_try_set_aio_context(new_child, child_ctx, errp);
> +        }
> +        if (ret < 0) {
> +            bdrv_remove_empty_child(new_child);
> +            return ret;
> +        }
> +    }

Not sure why you decided to rewrite this block while moving it from
bdrv_root_attach_child().

We're losing the comment above it, and a possible error message is now
related to changing the context of the parent node instead of the newly
added node, which I imagine is less obvious in the general case.

> +    bdrv_ref(child_bs);
> +    bdrv_replace_child_noperm(new_child, child_bs);
> +
> +    *child = new_child;
> +
> +    BdrvAttachChildCommonState *s = g_new(BdrvAttachChildCommonState, 1);
> +    *s = (BdrvAttachChildCommonState) {
> +        .child = child,
> +        .old_parent_ctx = parent_ctx,
> +        .old_child_ctx = child_ctx,
> +    };
> +    tran_prepend(tran, &bdrv_attach_child_common_drv, s);
> +
> +    return 0;
> +}

Kevin



  reply	other threads:[~2021-02-03 21:03 UTC|newest]

Thread overview: 108+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-11-27 14:44 [PATCH v2 00/36] block: update graph permissions update Vladimir Sementsov-Ogievskiy
2020-11-27 14:44 ` [PATCH v2 01/36] tests/test-bdrv-graph-mod: add test_parallel_exclusive_write Vladimir Sementsov-Ogievskiy
2020-11-27 14:44 ` [PATCH v2 02/36] tests/test-bdrv-graph-mod: add test_parallel_perm_update Vladimir Sementsov-Ogievskiy
2021-01-18 14:05   ` Kevin Wolf
2021-01-18 17:13     ` Vladimir Sementsov-Ogievskiy
2020-11-27 14:44 ` [PATCH v2 03/36] block: bdrv_append(): don't consume reference Vladimir Sementsov-Ogievskiy
2021-01-18 14:18   ` Kevin Wolf
2021-01-18 17:21     ` Vladimir Sementsov-Ogievskiy
2021-01-18 17:59       ` Kevin Wolf
2020-11-27 14:44 ` [PATCH v2 04/36] block: bdrv_append(): return status Vladimir Sementsov-Ogievskiy
2020-12-14 15:49   ` Alberto Garcia
2021-01-18 14:32   ` Kevin Wolf
2020-11-27 14:44 ` [PATCH v2 05/36] block: add bdrv_parent_try_set_aio_context Vladimir Sementsov-Ogievskiy
2021-01-18 15:08   ` Kevin Wolf
2021-01-18 17:26     ` Vladimir Sementsov-Ogievskiy
2020-11-27 14:44 ` [PATCH v2 06/36] block: BdrvChildClass: add .get_parent_aio_context handler Vladimir Sementsov-Ogievskiy
2021-01-18 15:13   ` Kevin Wolf
2021-01-18 17:36     ` Vladimir Sementsov-Ogievskiy
2021-01-19 16:38       ` Kevin Wolf
2021-01-22 11:04         ` Vladimir Sementsov-Ogievskiy
2021-01-22 11:18           ` Kevin Wolf
2021-01-22 11:26             ` Vladimir Sementsov-Ogievskiy
2020-11-27 14:44 ` [PATCH v2 07/36] block: drop ctx argument from bdrv_root_attach_child Vladimir Sementsov-Ogievskiy
2020-11-27 14:44 ` [PATCH v2 08/36] block: make bdrv_reopen_{prepare, commit, abort} private Vladimir Sementsov-Ogievskiy via
2020-12-15 17:28   ` Alberto Garcia
2021-01-18 15:24   ` [PATCH v2 08/36] block: make bdrv_reopen_{prepare,commit,abort} private Kevin Wolf
2020-11-27 14:44 ` [PATCH v2 09/36] block: return value from bdrv_replace_node() Vladimir Sementsov-Ogievskiy
2020-12-15 17:30   ` Alberto Garcia
2021-01-18 15:40   ` Kevin Wolf
2020-11-27 14:44 ` [PATCH v2 10/36] util: add transactions.c Vladimir Sementsov-Ogievskiy
2021-01-18 16:50   ` Kevin Wolf
2021-01-18 17:41     ` Vladimir Sementsov-Ogievskiy
2020-11-27 14:44 ` [PATCH v2 11/36] block: bdrv_refresh_perms: check parents compliance Vladimir Sementsov-Ogievskiy
2021-01-19 17:42   ` Kevin Wolf
2021-01-19 18:10     ` Vladimir Sementsov-Ogievskiy
2020-11-27 14:44 ` [PATCH v2 12/36] block: refactor bdrv_child* permission functions Vladimir Sementsov-Ogievskiy
2021-01-19 18:09   ` Kevin Wolf
2021-01-19 18:30     ` Vladimir Sementsov-Ogievskiy
2021-01-20  9:09       ` Kevin Wolf
2021-01-20  9:56         ` Vladimir Sementsov-Ogievskiy
2021-01-20 10:06           ` Kevin Wolf
2020-11-27 14:44 ` [PATCH v2 13/36] block: rewrite bdrv_child_try_set_perm() using bdrv_refresh_perms() Vladimir Sementsov-Ogievskiy
2020-11-27 14:45 ` [PATCH v2 14/36] block: inline bdrv_child_*() permission functions calls Vladimir Sementsov-Ogievskiy
2020-12-16 17:16   ` Alberto Garcia
2020-11-27 14:45 ` [PATCH v2 15/36] block: use topological sort for permission update Vladimir Sementsov-Ogievskiy
2021-01-27 18:38   ` Kevin Wolf
2021-01-28  9:34     ` Vladimir Sementsov-Ogievskiy
2021-01-28 17:13       ` Kevin Wolf
2021-01-28 18:04         ` Vladimir Sementsov-Ogievskiy
2021-02-03 18:38           ` Kevin Wolf
2021-02-04  7:16             ` Vladimir Sementsov-Ogievskiy
2021-03-10 11:08             ` Vladimir Sementsov-Ogievskiy
2021-03-10 11:55               ` Kevin Wolf
2020-11-27 14:45 ` [PATCH v2 16/36] block: add bdrv_drv_set_perm transaction action Vladimir Sementsov-Ogievskiy
2020-11-27 14:45 ` [PATCH v2 17/36] block: add bdrv_list_* permission update functions Vladimir Sementsov-Ogievskiy
2020-11-27 14:45 ` [PATCH v2 18/36] block: add bdrv_replace_child_safe() transaction action Vladimir Sementsov-Ogievskiy
2020-11-27 14:45 ` [PATCH v2 19/36] block: fix bdrv_replace_node_common Vladimir Sementsov-Ogievskiy
2021-02-03 18:23   ` Kevin Wolf
2021-02-04  7:24     ` Vladimir Sementsov-Ogievskiy
2020-11-27 14:45 ` [PATCH v2 20/36] block: add bdrv_attach_child_common() transaction action Vladimir Sementsov-Ogievskiy
2021-02-03 21:01   ` Kevin Wolf [this message]
2021-02-04  7:34     ` Vladimir Sementsov-Ogievskiy
2021-02-04  7:50       ` Kevin Wolf
2020-11-27 14:45 ` [PATCH v2 21/36] block: add bdrv_attach_child_noperm() " Vladimir Sementsov-Ogievskiy
2020-11-27 14:45 ` [PATCH v2 22/36] block: split out bdrv_replace_node_noperm() Vladimir Sementsov-Ogievskiy
2021-02-03 21:16   ` Kevin Wolf
2020-11-27 14:45 ` [PATCH v2 23/36] block: adapt bdrv_append() for inserting filters Vladimir Sementsov-Ogievskiy
2021-02-03 21:33   ` Kevin Wolf
2021-02-04  8:30     ` Vladimir Sementsov-Ogievskiy
2021-02-04  9:05       ` Kevin Wolf
2021-02-04 11:54         ` Vladimir Sementsov-Ogievskiy
2020-11-27 14:45 ` [PATCH v2 24/36] block: add bdrv_remove_backing transaction action Vladimir Sementsov-Ogievskiy
2020-11-27 14:45 ` [PATCH v2 25/36] block: introduce bdrv_drop_filter() Vladimir Sementsov-Ogievskiy
2021-02-04 11:31   ` Kevin Wolf
2021-02-04 12:27     ` Vladimir Sementsov-Ogievskiy
2020-11-27 14:45 ` [PATCH v2 26/36] block/backup-top: drop .active Vladimir Sementsov-Ogievskiy
2021-02-04 12:26   ` Kevin Wolf
2021-02-04 12:33     ` Vladimir Sementsov-Ogievskiy
2021-02-04 13:25       ` Kevin Wolf
2021-02-04 13:46         ` Vladimir Sementsov-Ogievskiy
2021-02-04 14:31           ` Kevin Wolf
2020-11-27 14:45 ` [PATCH v2 27/36] block: drop ignore_children for permission update functions Vladimir Sementsov-Ogievskiy
2020-11-27 14:45 ` [PATCH v2 28/36] block: add bdrv_set_backing_noperm() transaction action Vladimir Sementsov-Ogievskiy
2021-02-05 14:00   ` Kevin Wolf
2021-02-05 16:06     ` Vladimir Sementsov-Ogievskiy
2021-02-05 16:30       ` Kevin Wolf
2021-03-11 18:29         ` Vladimir Sementsov-Ogievskiy
2021-02-05 16:26   ` Kevin Wolf
2021-02-08  9:34     ` Vladimir Sementsov-Ogievskiy
2020-11-27 14:45 ` [PATCH v2 29/36] blockdev: qmp_x_blockdev_reopen: acquire all contexts Vladimir Sementsov-Ogievskiy
2021-02-05 16:01   ` Kevin Wolf
2021-02-05 16:16     ` Vladimir Sementsov-Ogievskiy
2021-02-05 16:36       ` Kevin Wolf
2020-11-27 14:45 ` [PATCH v2 30/36] block: bdrv_reopen_multiple: refresh permissions on updated graph Vladimir Sementsov-Ogievskiy
2021-02-05 17:57   ` Kevin Wolf
2021-02-08 11:21     ` Vladimir Sementsov-Ogievskiy
2021-02-10 14:13       ` Kevin Wolf
2021-02-10 14:38       ` Kevin Wolf
2020-11-27 14:45 ` [PATCH v2 31/36] block: drop unused permission update functions Vladimir Sementsov-Ogievskiy
2020-11-27 14:45 ` [PATCH v2 32/36] block: inline bdrv_check_perm_common() Vladimir Sementsov-Ogievskiy
2020-11-27 14:45 ` [PATCH v2 33/36] block: inline bdrv_replace_child() Vladimir Sementsov-Ogievskiy
2020-11-27 14:45 ` [PATCH v2 34/36] block: refactor bdrv_child_set_perm_safe() transaction action Vladimir Sementsov-Ogievskiy
2021-02-10 14:51   ` Kevin Wolf
2020-11-27 14:45 ` [PATCH v2 35/36] block: rename bdrv_replace_child_safe() to bdrv_replace_child() Vladimir Sementsov-Ogievskiy
2020-11-27 14:45 ` [PATCH v2 36/36] block: refactor bdrv_node_check_perm() Vladimir Sementsov-Ogievskiy
2021-02-10 15:07   ` Kevin Wolf
2021-02-11  9:50     ` Vladimir Sementsov-Ogievskiy
2021-01-09 10:12 ` [PATCH v2 00/36] block: update graph permissions update Vladimir Sementsov-Ogievskiy

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=20210203210104.GH5507@merkur.fritz.box \
    --to=kwolf@redhat.com \
    --cc=armbru@redhat.com \
    --cc=den@openvz.org \
    --cc=jsnow@redhat.com \
    --cc=mreitz@redhat.com \
    --cc=qemu-block@nongnu.org \
    --cc=qemu-devel@nongnu.org \
    --cc=vsementsov@virtuozzo.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).