[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH] block: Make op blockers recursive
From: |
Fam Zheng |
Subject: |
Re: [Qemu-devel] [PATCH] block: Make op blockers recursive |
Date: |
Fri, 12 Sep 2014 11:48:33 +0800 |
User-agent: |
Mutt/1.5.23 (2014-03-12) |
On Tue, 09/09 14:28, Benoît Canet wrote:
> On Tue, Sep 09, 2014 at 01:56:46PM +0200, Kevin Wolf wrote:
> > Am 22.08.2014 um 18:11 hat Benoît Canet geschrieben:
> > > Since the block layer code is starting to modify the BDS graph right in
> > > the
> > > middle of BDS chains (block-mirror's replace parameter for example) QEMU
> > > needs
> > > to properly block and unblock whole BDS subtrees; recursion is a neat way
> > > to
> > > achieve this task.
> > >
> > > This patch also takes care of modifying the op blockers users.
> > >
> > > Signed-off-by: Benoit Canet <address@hidden>
> > > ---
> > > block.c | 69
> > > ++++++++++++++++++++++++++++++++++++++++++++---
> > > block/blkverify.c | 21 +++++++++++++++
> > > block/commit.c | 3 +++
> > > block/mirror.c | 17 ++++++++----
> > > block/quorum.c | 25 +++++++++++++++++
> > > block/stream.c | 3 +++
> > > block/vmdk.c | 34 +++++++++++++++++++++++
> > > include/block/block.h | 2 +-
> > > include/block/block_int.h | 6 +++++
> > > 9 files changed, 171 insertions(+), 9 deletions(-)
> > >
> > > diff --git a/block.c b/block.c
> > > index 6fa0201..d964b6c 100644
> > > --- a/block.c
> > > +++ b/block.c
> > > @@ -5446,7 +5446,9 @@ bool bdrv_op_is_blocked(BlockDriverState *bs,
> > > BlockOpType op, Error **errp)
> > > return false;
> > > }
> > >
> > > -void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason)
> > > +/* do the real work of blocking a BDS */
> > > +static void bdrv_do_op_block(BlockDriverState *bs, BlockOpType op,
> > > + Error *reason)
> > > {
> > > BdrvOpBlocker *blocker;
> > > assert((int) op >= 0 && op < BLOCK_OP_TYPE_MAX);
> > > @@ -5456,7 +5458,9 @@ void bdrv_op_block(BlockDriverState *bs,
> > > BlockOpType op, Error *reason)
> > > QLIST_INSERT_HEAD(&bs->op_blockers[op], blocker, list);
> > > }
> > >
> > > -void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason)
> > > +/* do the real work of unblocking a BDS */
> > > +static void bdrv_do_op_unblock(BlockDriverState *bs, BlockOpType op,
> > > + Error *reason)
> > > {
> > > BdrvOpBlocker *blocker, *next;
> > > assert((int) op >= 0 && op < BLOCK_OP_TYPE_MAX);
> > > @@ -5468,6 +5472,65 @@ void bdrv_op_unblock(BlockDriverState *bs,
> > > BlockOpType op, Error *reason)
> > > }
> > > }
> > >
> > > +static bool bdrv_op_is_blocked_by(BlockDriverState *bs, BlockOpType op,
> > > + Error *reason)
> > > +{
> > > + BdrvOpBlocker *blocker, *next;
> > > + assert((int) op >= 0 && op < BLOCK_OP_TYPE_MAX);
> > > + QLIST_FOREACH_SAFE(blocker, &bs->op_blockers[op], list, next) {
> >
> > This doesn't actually need the SAFE version.
> >
> > > + if (blocker->reason == reason) {
> > > + return true;
> > > + }
> > > + }
> > > + return false;
> > > +}
> > > +
> > > +/* block recursively a BDS */
> > > +void bdrv_op_block(BlockDriverState *bs, BlockOpType op, Error *reason)
> > > +{
> > > + if (!bs) {
> > > + return;
> > > + }
> > > +
> > > + /* prevent recursion loop */
> > > + if (bdrv_op_is_blocked_by(bs, op, reason)) {
> > > + return;
> > > + }
> > > +
> > > + /* block first for recursion loop protection to work */
> > > + bdrv_do_op_block(bs, op, reason);
> > > +
> > > + bdrv_op_block(bs->file, op, reason);
> > > + bdrv_op_block(bs->backing_hd, op, reason);
> > > +
> > > + if (bs->drv && bs->drv->bdrv_op_recursive_block) {
> > > + bs->drv->bdrv_op_recursive_block(bs, op, reason);
> > > + }
> >
> > Here you block bs->file/bs->backing_hd automatically, no matter whether
> > the block driver implements the callback or not. I'm not sure if we can
> > do it like this for all times, but for now that should be okay.
> >
> > > +}
> > > +
> > > +/* unblock recursively a BDS */
> > > +void bdrv_op_unblock(BlockDriverState *bs, BlockOpType op, Error *reason)
> > > +{
> > > + if (!bs) {
> > > + return;
> > > + }
> > > +
> > > + /* prevent recursion loop */
> > > + if (!bdrv_op_is_blocked_by(bs, op, reason)) {
> > > + return;
> > > + }
> > > +
> > > + /* unblock first for recursion loop protection to work */
> > > + bdrv_do_op_unblock(bs, op, reason);
> > > +
> > > + bdrv_op_unblock(bs->file, op, reason);
> > > + bdrv_op_unblock(bs->backing_hd, op, reason);
> > > +
> > > + if (bs->drv && bs->drv->bdrv_op_recursive_unblock) {
> > > + bs->drv->bdrv_op_recursive_unblock(bs, op, reason);
> > > + }
> > > +}
> > > +
> > > void bdrv_op_block_all(BlockDriverState *bs, Error *reason)
> > > {
> > > int i;
> > > @@ -5848,7 +5911,7 @@ BlockDriverState *check_to_replace_node(const char
> > > *node_name, Error **errp)
> > > return NULL;
> > > }
> > >
> > > - if (bdrv_op_is_blocked(to_replace_bs, BLOCK_OP_TYPE_REPLACE, errp)) {
> > > + if (bdrv_op_is_blocked(to_replace_bs, BLOCK_OP_TYPE_MIRROR_REPLACE,
> > > errp)) {
> >
> > That rename could have been a separate patch.
> >
> > > return NULL;
> > > }
> > >
> > > diff --git a/block/blkverify.c b/block/blkverify.c
> > > index 621b785..75ec3df 100644
> > > --- a/block/blkverify.c
> > > +++ b/block/blkverify.c
> > > @@ -320,6 +320,24 @@ static void
> > > blkverify_attach_aio_context(BlockDriverState *bs,
> > > bdrv_attach_aio_context(s->test_file, new_context);
> > > }
> > >
> > > +static void blkverify_op_recursive_block(BlockDriverState *bs,
> > > BlockOpType op,
> > > + Error *reason)
> > > +{
> > > + BDRVBlkverifyState *s = bs->opaque;
> > > +
> > > + bdrv_op_block(bs->file, op, reason);
> > > + bdrv_op_block(s->test_file, op, reason);
> > > +}
> > > +
> > > +static void blkverify_op_recursive_unblock(BlockDriverState *bs,
> > > BlockOpType op,
> > > + Error *reason)
> > > +{
> > > + BDRVBlkverifyState *s = bs->opaque;
> > > +
> > > + bdrv_op_unblock(bs->file, op, reason);
> > > + bdrv_op_unblock(s->test_file, op, reason);
> > > +}
> >
> > Here you block bs->file again, even though the block.c function has
> > already done it. Nothing bad happens, because bdrv_op_block() simply
> > stops when the blocker is already set, but the code would be nicer if we
> > did block bs->file only in one place. (I'm leaning towards doing it in
> > the drivers, but that's your decision.)
> >
> > > static BlockDriver bdrv_blkverify = {
> > > .format_name = "blkverify",
> > > .protocol_name = "blkverify",
> > > @@ -337,6 +355,9 @@ static BlockDriver bdrv_blkverify = {
> > > .bdrv_attach_aio_context = blkverify_attach_aio_context,
> > > .bdrv_detach_aio_context = blkverify_detach_aio_context,
> > >
> > > + .bdrv_op_recursive_block = blkverify_op_recursive_block,
> > > + .bdrv_op_recursive_unblock = blkverify_op_recursive_unblock,
> > > +
> > > .is_filter = true,
> > > .bdrv_recurse_is_first_non_filter =
> > > blkverify_recurse_is_first_non_filter,
> > > };
> > > diff --git a/block/commit.c b/block/commit.c
> > > index 91517d3..8a122b7 100644
> > > --- a/block/commit.c
> > > +++ b/block/commit.c
> > > @@ -142,6 +142,9 @@ wait:
> > >
> > > if (!block_job_is_cancelled(&s->common) && sector_num == end) {
> > > /* success */
> > > + /* unblock only BDS to be dropped */
> > > + bdrv_op_unblock_all(top, s->common.blocker);
> > > + bdrv_op_block_all(base, s->common.blocker);
> >
> > This suggests that bdrv_op_(un)block_all() doesn't provide the right
> > interface, but we rather want an optional base argument in it, so that
> > you can (un)block a specific part of the backing file chain.
>
> Good idea.
>
> >
> > > ret = bdrv_drop_intermediate(active, top, base,
> > > s->backing_file_str);
> > > }
> > >
> > > diff --git a/block/mirror.c b/block/mirror.c
> > > index 5e7a166..28ed47d 100644
> > > --- a/block/mirror.c
> > > +++ b/block/mirror.c
> > > @@ -324,6 +324,7 @@ static void coroutine_fn mirror_run(void *opaque)
> > > uint64_t last_pause_ns;
> > > BlockDriverInfo bdi;
> > > char backing_filename[1024];
> > > + BlockDriverState *to_replace;
> > > int ret = 0;
> > > int n;
> > >
> > > @@ -512,14 +513,16 @@ immediate_exit:
> > > g_free(s->in_flight_bitmap);
> > > bdrv_release_dirty_bitmap(bs, s->dirty_bitmap);
> > > bdrv_iostatus_disable(s->target);
> > > + to_replace = s->common.bs;
> > > + if (s->to_replace) {
> > > + bdrv_op_unblock_all(s->to_replace, s->replace_blocker);
> > > + to_replace = s->to_replace;
> > > + }
> > > if (s->should_complete && ret == 0) {
> > > - BlockDriverState *to_replace = s->common.bs;
> > > - if (s->to_replace) {
> > > - to_replace = s->to_replace;
> > > - }
> > > if (bdrv_get_flags(s->target) != bdrv_get_flags(to_replace)) {
> > > bdrv_reopen(s->target, bdrv_get_flags(to_replace), NULL);
> > > }
> > > + bdrv_op_unblock_all(to_replace, bs->job->blocker);
> >
> > You need to unblock all of the BDSes that are going to be removed.
> > Aren't you unblocking more than that here?
> >
> > It's not a real problem because block_job_completed() would unblock the
> > rest anyway and its bdrv_op_unblock_all() call is simply ignored now,
> > but it would be nicer to just unblock here what's actually needed.
> >
> > > bdrv_swap(s->target, to_replace);
> > > if (s->common.driver->job_type == BLOCK_JOB_TYPE_COMMIT) {
> > > /* drop the bs loop chain formed by the swap: break the loop
> > > then
> > > @@ -530,7 +533,6 @@ immediate_exit:
> > > }
> > > }
> > > if (s->to_replace) {
> > > - bdrv_op_unblock_all(s->to_replace, s->replace_blocker);
> > > error_free(s->replace_blocker);
> > > bdrv_unref(s->to_replace);
> > > }
> > > @@ -648,6 +650,11 @@ static void mirror_start_job(BlockDriverState *bs,
> > > BlockDriverState *target,
> > > return;
> > > }
> > >
> > > + /* remove BLOCK_OP_TYPE_MIRROR_REPLACE from the list of
> > > + * blocked operations so the replaces parameter can work
> > > + */
> > > + bdrv_op_unblock(bs, BLOCK_OP_TYPE_MIRROR_REPLACE, bs->job->blocker);
> >
> > What purpose does a blocker serve when it's disabled before it is
> > checked? I would only ever expect a bdrv_op_unblock() after some
> > operation on the BDS has finished, but not when starting an operation
> > that needs it.
I agree. It makes no sense if the blocker is also the checker.
>
> BLOCK_OP_TYPE_MIRROR_REPLACE is checked and blocked by block-job-complete
> during the time the mirror finish when an arbitrary node of the graph must be
> replaced.
It seems to me mirror unblocks this operation from the job->blocker when job
starts, and never block it (with the job->blocker) again. It's leaked.
Fam
>
> The start of the mirroring block everything including
> BLOCK_OP_TYPE_MIRROR_REPLACE so this hunk relax the blocking so
> block-job-complete
> can have a chance of being able to block it.
>
> The comment of this hunk seems to be deficient and the code not self
> explanatory.
>
> On the other way maybe block-job-complete is done wrong from the start but
> relaxing BLOCK_OP_TYPE_MIRROR_REPLACE when the mirror start is the only way
> to make block-job-complete work as intended.
>
> Thanks
>
> Benoît