All of lore.kernel.org
 help / color / mirror / Atom feed
From: Fam Zheng <famz@redhat.com>
To: Paolo Bonzini <pbonzini@redhat.com>
Cc: kwolf@redhat.com, jcody@redhat.com, qemu-devel@nongnu.org,
	stefanha@redhat.com
Subject: Re: [Qemu-devel] [PATCH 1/2] block: allow live commit of active image
Date: Mon, 22 Jul 2013 14:48:34 +0800	[thread overview]
Message-ID: <20130722064834.GA1992@T430s.redhat.com> (raw)
In-Reply-To: <51ECD26A.80104@redhat.com>

On Mon, 07/22 08:34, Paolo Bonzini wrote:
> Il 22/07/2013 05:46, Fam Zheng ha scritto:
> > This patch eliminates limitation of committing the active device.
> > 
> > bdrv_drop_intermediate is reimplemented to take pointers to
> > (BlockDriverState *), so it can modify the caller's local pointers to
> > preserve their semantics, while updating active BDS in-place by
> > bdrv_swap active and base: we need data in 'base' as it's the only
> > remaining after commit, but we can't delete 'active' as it's referenced
> > everywhere in the program.
> > 
> > Guest writes to active device during the commit are tracked by dirty map
> > and committed like block-mirror.
> 
> I have only skimmed the patch, but I think this is incomplete.
> Management needs to know the moment when 'active' is not valid anymore,
> thus this job needs to be completed manually with "block-job-complete".

Does management need access to 'active' image outside of QEMU process?
Although original 'active' it is "dropped" by bdrv_drop_intermediate,
the pointers to original 'active' is still valid because 'base' is moved
to this address (with bdrv_swap). I don't know, what is the problem here
for management?

> 
> In fact, I wonder if block/commit.c could reuse most of the code from
> block/mirror.c (basically everything except that bdrv_swap should be
> replaced by bdrv_drop_intermediate).
> 

Hmm, yes, in this case, it is quite similar to mirroring 'active' to
'base' with sync mode top.

> > Signed-off-by: Fam Zheng <famz@redhat.com>
> > ---
> >  block.c               | 102 ++++++++++----------------------
> >  block/commit.c        | 160 ++++++++++++++++++++++++++------------------------
> >  include/block/block.h |   5 +-
> >  3 files changed, 115 insertions(+), 152 deletions(-)
> > 
> > diff --git a/block.c b/block.c
> > index b560241..367e064 100644
> > --- a/block.c
> > +++ b/block.c
> > @@ -2018,18 +2018,11 @@ 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.
> > - *
> > - * Requires that the overlay to 'top' is opened r/w, so that the backing file
> > - * information in 'bs' can be properly updated.
> > + * Drops images above '*base' up to and including '*top', and sets new '*base'
> > + * as backing_hd of top_overlay (the image orignally has 'top' as backing
> > + * file). top_overlay may be NULL if '*top' is active, no such update needed.
> > + * Requires that the top_overlay to 'top' is opened r/w.
> >   *
> >   * E.g., this will convert the following chain:
> >   * bottom <- base <- intermediate <- top <- active
> > @@ -2046,82 +2039,47 @@ typedef struct BlkIntermediateStates {
> >   *
> >   * base <- active
> >   *
> > - * Error conditions:
> > - *  if active == top, that is considered an error
> > + * It also allows active==top, in which case it converts:
> > + *
> > + * base <- intermediate <- active (also top)
> > + *
> > + * to
> > + *
> > + * base == active == top, i.e. only base remains: *top == *base when return.
> >   *
> >   */
> > -int bdrv_drop_intermediate(BlockDriverState *active, BlockDriverState *top,
> > -                           BlockDriverState *base)
> > +int bdrv_drop_intermediate(BlockDriverState *top_overlay,
> > +                           BlockDriverState **top,
> > +                           BlockDriverState **base)
> >  {
> > -    BlockDriverState *intermediate;
> > +    BlockDriverState *pbs;
> >      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) {
> > +    if (!(*top)->drv || !(*base)->drv) {
> >          goto exit;
> >      }
> >  
> > -    new_top_bs = bdrv_find_overlay(active, top);
> > -
> > -    if (new_top_bs == NULL) {
> > -        /* we could not find the image above 'top', this is an error */
> > -        goto exit;
> > +    for (pbs = (*top)->backing_hd; pbs != *base; pbs = base_bs) {
> > +        assert(pbs);
> > +        base_bs = pbs->backing_hd;
> > +        pbs->backing_hd = NULL;
> > +        bdrv_delete(pbs);
> >      }
> >  
> > -    /* 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;
> > -    }
> > +    bdrv_swap(*base, *top);
> >  
> > -    intermediate = top;
> > +    (*base)->backing_hd = NULL;
> > +    bdrv_delete(*base);
> > +    *base = *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;
> > -        }
> > -        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;
> > -    }
> > -
> > -    /* 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;
> > -    }
> > -    new_top_bs->backing_hd = base_bs;
> > -
> > -
> > -    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);
> > +    /* overlay exists when active != top, need to change backing file for it */
> > +    if (top_overlay) {
> > +        ret = bdrv_change_backing_file(top_overlay, (*base)->filename,
> > +                                       (*base)->drv ?
> > +                                            (*base)->drv->format_name : "");
> >      }
> > -    ret = 0;
> > -
> >  exit:
> > -    QSIMPLEQ_FOREACH_SAFE(intermediate_state, &states_to_delete, entry, next) {
> > -        g_free(intermediate_state);
> > -    }
> >      return ret;
> >  }
> >  
> > diff --git a/block/commit.c b/block/commit.c
> > index 2227fc2..c85b188 100644
> > --- a/block/commit.c
> > +++ b/block/commit.c
> > @@ -17,14 +17,13 @@
> >  #include "block/blockjob.h"
> >  #include "qemu/ratelimit.h"
> >  
> > -enum {
> > -    /*
> > -     * Size of data buffer for populating the image file.  This should be large
> > -     * enough to process multiple clusters in a single call, so that populating
> > -     * contiguous regions of the image is efficient.
> > -     */
> > -    COMMIT_BUFFER_SIZE = 512 * 1024, /* in bytes */
> > -};
> > +/*
> > + * Size of data buffer for populating the image file.  This should be large
> > + * enough to process multiple clusters in a single call, so that populating
> > + * contiguous regions of the image is efficient.
> > + */
> > +#define COMMIT_BUFFER_SECTORS 128
> > +#define COMMIT_BUFFER_BYTES (COMMIT_BUFFER_SECTORS * BDRV_SECTOR_SIZE)
> >  
> >  #define SLICE_TIME 100000000ULL /* ns */
> >  
> > @@ -34,6 +33,7 @@ typedef struct CommitBlockJob {
> >      BlockDriverState *active;
> >      BlockDriverState *top;
> >      BlockDriverState *base;
> > +    BlockDriverState *overlay;
> >      BlockdevOnError on_error;
> >      int base_flags;
> >      int orig_overlay_flags;
> > @@ -65,100 +65,109 @@ static void coroutine_fn commit_run(void *opaque)
> >      BlockDriverState *active = s->active;
> >      BlockDriverState *top = s->top;
> >      BlockDriverState *base = s->base;
> > -    BlockDriverState *overlay_bs;
> >      int64_t sector_num, end;
> >      int ret = 0;
> >      int n = 0;
> >      void *buf;
> > -    int bytes_written = 0;
> >      int64_t base_len;
> > +    int64_t next_dirty;
> > +    HBitmapIter hbi;
> >  
> > +    buf = qemu_blockalign(top, COMMIT_BUFFER_BYTES);
> >      ret = s->common.len = bdrv_getlength(top);
> >  
> > -
> >      if (s->common.len < 0) {
> > -        goto exit_restore_reopen;
> > +        goto exit;
> >      }
> >  
> >      ret = base_len = bdrv_getlength(base);
> >      if (base_len < 0) {
> > -        goto exit_restore_reopen;
> > +        goto exit;
> >      }
> >  
> >      if (base_len < s->common.len) {
> >          ret = bdrv_truncate(base, s->common.len);
> >          if (ret) {
> > -            goto exit_restore_reopen;
> > +            goto exit;
> >          }
> >      }
> >  
> >      end = s->common.len >> BDRV_SECTOR_BITS;
> > -    buf = qemu_blockalign(top, COMMIT_BUFFER_SIZE);
> >  
> >      for (sector_num = 0; sector_num < end; sector_num += n) {
> > -        uint64_t delay_ns = 0;
> > -        bool copy;
> >  
> > -wait:
> > -        /* Note that even when no rate limit is applied we need to yield
> > -         * with no pending I/O here so that bdrv_drain_all() returns.
> > -         */
> > -        block_job_sleep_ns(&s->common, rt_clock, delay_ns);
> > -        if (block_job_is_cancelled(&s->common)) {
> > -            break;
> > -        }
> >          /* Copy if allocated above the base */
> >          ret = bdrv_co_is_allocated_above(top, base, sector_num,
> > -                                         COMMIT_BUFFER_SIZE / BDRV_SECTOR_SIZE,
> > +                                         COMMIT_BUFFER_SECTORS,
> >                                           &n);
> > -        copy = (ret == 1);
> > -        trace_commit_one_iteration(s, sector_num, n, ret);
> > -        if (copy) {
> > -            if (s->common.speed) {
> > -                delay_ns = ratelimit_calculate_delay(&s->limit, n);
> > -                if (delay_ns > 0) {
> > -                    goto wait;
> > -                }
> > -            }
> > -            ret = commit_populate(top, base, sector_num, n, buf);
> > -            bytes_written += n * BDRV_SECTOR_SIZE;
> > +        if (ret) {
> > +            bdrv_set_dirty(top, sector_num, n);
> > +        }
> > +    }
> > +
> > +    while (bdrv_get_dirty_count(s->top)) {
> > +        uint64_t delay_ns = 0;
> > +        if (block_job_is_cancelled(&s->common)) {
> > +            goto exit;
> >          }
> > -        if (ret < 0) {
> > -            if (s->on_error == BLOCKDEV_ON_ERROR_STOP ||
> > -                s->on_error == BLOCKDEV_ON_ERROR_REPORT||
> > -                (s->on_error == BLOCKDEV_ON_ERROR_ENOSPC && ret == -ENOSPC)) {
> > -                goto exit_free_buf;
> > -            } else {
> > -                n = 0;
> > -                continue;
> > +
> > +        bdrv_dirty_iter_init(s->top, &hbi);
> > +        for (next_dirty = hbitmap_iter_next(&hbi);
> > +                next_dirty >= 0;
> > +                next_dirty = hbitmap_iter_next(&hbi)) {
> > +            sector_num = next_dirty;
> > +            if (block_job_is_cancelled(&s->common)) {
> > +                goto exit;
> >              }
> > +            delay_ns = ratelimit_calculate_delay(&s->limit,
> > +                                                 COMMIT_BUFFER_SECTORS);
> > +            /* Note that even when no rate limit is applied we need to yield
> > +             * with no pending I/O here so that bdrv_drain_all() returns.
> > +             */
> > +            block_job_sleep_ns(&s->common, rt_clock, delay_ns);
> > +            trace_commit_one_iteration(s, sector_num,
> > +                                       COMMIT_BUFFER_SECTORS, ret);
> > +            ret = commit_populate(top, base, sector_num,
> > +                                  COMMIT_BUFFER_SECTORS, buf);
> > +            if (ret < 0) {
> > +                if (s->on_error == BLOCKDEV_ON_ERROR_STOP ||
> > +                    s->on_error == BLOCKDEV_ON_ERROR_REPORT ||
> > +                    (s->on_error == BLOCKDEV_ON_ERROR_ENOSPC &&
> > +                         ret == -ENOSPC)) {
> > +                    goto exit;
> > +                } else {
> > +                    continue;
> > +                }
> > +            }
> > +            /* Publish progress */
> > +            s->common.offset += COMMIT_BUFFER_BYTES;
> > +            bdrv_reset_dirty(top, sector_num, COMMIT_BUFFER_SECTORS);
> >          }
> > -        /* Publish progress */
> > -        s->common.offset += n * BDRV_SECTOR_SIZE;
> >      }
> >  
> > -    ret = 0;
> > -
> > -    if (!block_job_is_cancelled(&s->common) && sector_num == end) {
> > -        /* success */
> > -        ret = bdrv_drop_intermediate(active, top, base);
> > +    if (!block_job_is_cancelled(&s->common)) {
> > +        /* Drop intermediate: [top, base) */
> > +        ret = bdrv_drop_intermediate(s->overlay, &top, &base);
> > +        s->common.offset = s->common.len;
> >      }
> >  
> > -exit_free_buf:
> > -    qemu_vfree(buf);
> > +    ret = 0;
> > +
> > +exit:
> > +    bdrv_set_dirty_tracking(active, 0);
> >  
> > -exit_restore_reopen:
> >      /* restore base open flags here if appropriate (e.g., change the base back
> >       * to r/o). These reopens do not need to be atomic, since we won't abort
> >       * even on failure here */
> > -    if (s->base_flags != bdrv_get_flags(base)) {
> > +    if (s->overlay && s->base_flags != bdrv_get_flags(base)) {
> >          bdrv_reopen(base, s->base_flags, NULL);
> >      }
> > -    overlay_bs = bdrv_find_overlay(active, top);
> > -    if (overlay_bs && s->orig_overlay_flags != bdrv_get_flags(overlay_bs)) {
> > -        bdrv_reopen(overlay_bs, s->orig_overlay_flags, NULL);
> > +
> > +    if (s->overlay && s->orig_overlay_flags != bdrv_get_flags(s->overlay)) {
> > +        bdrv_reopen(s->overlay, s->orig_overlay_flags, NULL);
> >      }
> >  
> > +    qemu_vfree(buf);
> >      block_job_completed(&s->common, ret);
> >  }
> >  
> > @@ -198,13 +207,6 @@ void commit_start(BlockDriverState *bs, BlockDriverState *base,
> >          return;
> >      }
> >  
> > -    /* Once we support top == active layer, remove this check */
> > -    if (top == bs) {
> > -        error_setg(errp,
> > -                   "Top image as the active layer is currently unsupported");
> > -        return;
> > -    }
> > -
> >      if (top == base) {
> >          error_setg(errp, "Invalid files for merge: top and base are the same");
> >          return;
> > @@ -212,23 +214,20 @@ void commit_start(BlockDriverState *bs, BlockDriverState *base,
> >  
> >      overlay_bs = bdrv_find_overlay(bs, top);
> >  
> > -    if (overlay_bs == NULL) {
> > -        error_setg(errp, "Could not find overlay image for %s:", top->filename);
> > -        return;
> > -    }
> > -
> >      orig_base_flags    = bdrv_get_flags(base);
> > -    orig_overlay_flags = bdrv_get_flags(overlay_bs);
> > +    if (overlay_bs) {
> > +        orig_overlay_flags = bdrv_get_flags(overlay_bs);
> > +        if (!(orig_overlay_flags & BDRV_O_RDWR)) {
> > +            reopen_queue = bdrv_reopen_queue(reopen_queue, overlay_bs,
> > +                    orig_overlay_flags | BDRV_O_RDWR);
> > +        }
> > +    }
> >  
> >      /* convert base & overlay_bs to r/w, if necessary */
> >      if (!(orig_base_flags & BDRV_O_RDWR)) {
> >          reopen_queue = bdrv_reopen_queue(reopen_queue, base,
> >                                           orig_base_flags | BDRV_O_RDWR);
> >      }
> > -    if (!(orig_overlay_flags & BDRV_O_RDWR)) {
> > -        reopen_queue = bdrv_reopen_queue(reopen_queue, overlay_bs,
> > -                                         orig_overlay_flags | BDRV_O_RDWR);
> > -    }
> >      if (reopen_queue) {
> >          bdrv_reopen_multiple(reopen_queue, &local_err);
> >          if (local_err != NULL) {
> > @@ -237,7 +236,6 @@ void commit_start(BlockDriverState *bs, BlockDriverState *base,
> >          }
> >      }
> >  
> > -
> >      s = block_job_create(&commit_job_type, bs, speed, cb, opaque, errp);
> >      if (!s) {
> >          return;
> > @@ -246,13 +244,19 @@ void commit_start(BlockDriverState *bs, BlockDriverState *base,
> >      s->base   = base;
> >      s->top    = top;
> >      s->active = bs;
> > +    s->overlay = overlay_bs;
> >  
> >      s->base_flags          = orig_base_flags;
> > -    s->orig_overlay_flags  = orig_overlay_flags;
> > +    if (overlay_bs) {
> > +        s->orig_overlay_flags  = orig_overlay_flags;
> > +    }
> >  
> >      s->on_error = on_error;
> >      s->common.co = qemu_coroutine_create(commit_run);
> >  
> >      trace_commit_start(bs, base, top, s, s->common.co, opaque);
> > +
> > +    bdrv_set_dirty_tracking(top, COMMIT_BUFFER_BYTES);
> > +
> >      qemu_coroutine_enter(s->common.co, s);
> >  }
> > diff --git a/include/block/block.h b/include/block/block.h
> > index b6b9014..caf2c22 100644
> > --- a/include/block/block.h
> > +++ b/include/block/block.h
> > @@ -197,8 +197,9 @@ int bdrv_commit_all(void);
> >  int bdrv_change_backing_file(BlockDriverState *bs,
> >      const char *backing_file, const char *backing_fmt);
> >  void bdrv_register(BlockDriver *bdrv);
> > -int bdrv_drop_intermediate(BlockDriverState *active, BlockDriverState *top,
> > -                           BlockDriverState *base);
> > +int bdrv_drop_intermediate(BlockDriverState *top_overlay,
> > +                           BlockDriverState **top,
> > +                           BlockDriverState **base);
> >  BlockDriverState *bdrv_find_overlay(BlockDriverState *active,
> >                                      BlockDriverState *bs);
> >  BlockDriverState *bdrv_find_base(BlockDriverState *bs);
> > 
> 

-- 
Fam

  reply	other threads:[~2013-07-22  6:48 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2013-07-22  3:46 [Qemu-devel] [PATCH 0/2] block: allow commit active as top Fam Zheng
2013-07-22  3:46 ` [Qemu-devel] [PATCH 1/2] block: allow live commit of active image Fam Zheng
2013-07-22  6:34   ` Paolo Bonzini
2013-07-22  6:48     ` Fam Zheng [this message]
2013-07-22 11:07       ` Paolo Bonzini
2013-07-23  2:03         ` Fam Zheng
2013-07-23 11:13           ` Paolo Bonzini
2013-07-22  3:46 ` [Qemu-devel] [PATCH 2/2] qemu-iotests: update test cases for commit active Fam Zheng
2013-07-22  6:43 ` [Qemu-devel] [PATCH 0/2] block: allow commit active as top Wenchao Xia
2013-07-22  7:05   ` Fam Zheng

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=20130722064834.GA1992@T430s.redhat.com \
    --to=famz@redhat.com \
    --cc=jcody@redhat.com \
    --cc=kwolf@redhat.com \
    --cc=pbonzini@redhat.com \
    --cc=qemu-devel@nongnu.org \
    --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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.