qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Max Reitz <mreitz@redhat.com>
To: qemu-block@nongnu.org
Cc: Kevin Wolf <kwolf@redhat.com>,
	Peter Maydell <peter.maydell@linaro.org>,
	qemu-devel@nongnu.org, Max Reitz <mreitz@redhat.com>
Subject: [PULL 31/53] block/block-copy: add ratelimit to block-copy
Date: Tue, 26 Jan 2021 15:19:54 +0100	[thread overview]
Message-ID: <20210126142016.806073-32-mreitz@redhat.com> (raw)
In-Reply-To: <20210126142016.806073-1-mreitz@redhat.com>

From: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>

We are going to directly use one async block-copy operation for backup
job, so we need rate limiter.

We want to maintain current backup behavior: only background copying is
limited and copy-before-write operations only participate in limit
calculation. Therefore we need one rate limiter for block-copy state
and boolean flag for block-copy call state for actual limitation.

Note, that we can't just calculate each chunk in limiter after
successful copying: it will not save us from starting a lot of async
sub-requests which will exceed limit too much. Instead let's use the
following scheme on sub-request creation:
1. If at the moment limit is not exceeded, create the request and
account it immediately.
2. If at the moment limit is already exceeded, drop create sub-request
and handle limit instead (by sleep).
With this approach we'll never exceed the limit more than by one
sub-request (which pretty much matches current backup behavior).

Note also, that if there is in-flight block-copy async call,
block_copy_kick() should be used after set-speed to apply new setup
faster. For that block_copy_kick() published in this patch.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210116214705.822267-7-vsementsov@virtuozzo.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
---
 include/block/block-copy.h |  5 ++++-
 block/backup-top.c         |  2 +-
 block/backup.c             |  2 +-
 block/block-copy.c         | 46 +++++++++++++++++++++++++++++++++++++-
 4 files changed, 51 insertions(+), 4 deletions(-)

diff --git a/include/block/block-copy.h b/include/block/block-copy.h
index 22372aa375..b5a53ad59e 100644
--- a/include/block/block-copy.h
+++ b/include/block/block-copy.h
@@ -41,7 +41,7 @@ int64_t block_copy_reset_unallocated(BlockCopyState *s,
                                      int64_t offset, int64_t *count);
 
 int coroutine_fn block_copy(BlockCopyState *s, int64_t offset, int64_t bytes,
-                            bool *error_is_read);
+                            bool ignore_ratelimit, bool *error_is_read);
 
 /*
  * Run block-copy in a coroutine, create corresponding BlockCopyCallState
@@ -76,6 +76,9 @@ bool block_copy_call_succeeded(BlockCopyCallState *call_state);
 bool block_copy_call_failed(BlockCopyCallState *call_state);
 int block_copy_call_status(BlockCopyCallState *call_state, bool *error_is_read);
 
+void block_copy_set_speed(BlockCopyState *s, uint64_t speed);
+void block_copy_kick(BlockCopyCallState *call_state);
+
 BdrvDirtyBitmap *block_copy_dirty_bitmap(BlockCopyState *s);
 void block_copy_set_skip_unallocated(BlockCopyState *s, bool skip);
 
diff --git a/block/backup-top.c b/block/backup-top.c
index 789acf6965..779956ddc2 100644
--- a/block/backup-top.c
+++ b/block/backup-top.c
@@ -61,7 +61,7 @@ static coroutine_fn int backup_top_cbw(BlockDriverState *bs, uint64_t offset,
     off = QEMU_ALIGN_DOWN(offset, s->cluster_size);
     end = QEMU_ALIGN_UP(offset + bytes, s->cluster_size);
 
-    return block_copy(s->bcs, off, end - off, NULL);
+    return block_copy(s->bcs, off, end - off, true, NULL);
 }
 
 static int coroutine_fn backup_top_co_pdiscard(BlockDriverState *bs,
diff --git a/block/backup.c b/block/backup.c
index 4b07e9115d..09ff5a92ef 100644
--- a/block/backup.c
+++ b/block/backup.c
@@ -72,7 +72,7 @@ static int coroutine_fn backup_do_cow(BackupBlockJob *job,
 
     trace_backup_do_cow_enter(job, start, offset, bytes);
 
-    ret = block_copy(job->bcs, start, end - start, error_is_read);
+    ret = block_copy(job->bcs, start, end - start, true, error_is_read);
 
     trace_backup_do_cow_return(job, offset, bytes, ret);
 
diff --git a/block/block-copy.c b/block/block-copy.c
index 6bf1735b93..fa27450b14 100644
--- a/block/block-copy.c
+++ b/block/block-copy.c
@@ -26,6 +26,7 @@
 #define BLOCK_COPY_MAX_BUFFER (1 * MiB)
 #define BLOCK_COPY_MAX_MEM (128 * MiB)
 #define BLOCK_COPY_MAX_WORKERS 64
+#define BLOCK_COPY_SLICE_TIME 100000000ULL /* ns */
 
 static coroutine_fn int block_copy_task_entry(AioTask *task);
 
@@ -36,6 +37,7 @@ typedef struct BlockCopyCallState {
     int64_t bytes;
     int max_workers;
     int64_t max_chunk;
+    bool ignore_ratelimit;
     BlockCopyAsyncCallbackFunc cb;
     void *cb_opaque;
 
@@ -48,6 +50,7 @@ typedef struct BlockCopyCallState {
     /* State */
     int ret;
     bool finished;
+    QemuCoSleepState *sleep_state;
 
     /* OUT parameters */
     bool error_is_read;
@@ -111,6 +114,9 @@ typedef struct BlockCopyState {
     void *progress_opaque;
 
     SharedResource *mem;
+
+    uint64_t speed;
+    RateLimit rate_limit;
 } BlockCopyState;
 
 static BlockCopyTask *find_conflicting_task(BlockCopyState *s,
@@ -623,6 +629,21 @@ block_copy_dirty_clusters(BlockCopyCallState *call_state)
         }
         task->zeroes = ret & BDRV_BLOCK_ZERO;
 
+        if (s->speed) {
+            if (!call_state->ignore_ratelimit) {
+                uint64_t ns = ratelimit_calculate_delay(&s->rate_limit, 0);
+                if (ns > 0) {
+                    block_copy_task_end(task, -EAGAIN);
+                    g_free(task);
+                    qemu_co_sleep_ns_wakeable(QEMU_CLOCK_REALTIME, ns,
+                                              &call_state->sleep_state);
+                    continue;
+                }
+            }
+
+            ratelimit_calculate_delay(&s->rate_limit, task->bytes);
+        }
+
         trace_block_copy_process(s, task->offset);
 
         co_get_from_shres(s->mem, task->bytes);
@@ -661,6 +682,13 @@ out:
     return ret < 0 ? ret : found_dirty;
 }
 
+void block_copy_kick(BlockCopyCallState *call_state)
+{
+    if (call_state->sleep_state) {
+        qemu_co_sleep_wake(call_state->sleep_state);
+    }
+}
+
 /*
  * block_copy_common
  *
@@ -707,12 +735,13 @@ static int coroutine_fn block_copy_common(BlockCopyCallState *call_state)
 }
 
 int coroutine_fn block_copy(BlockCopyState *s, int64_t start, int64_t bytes,
-                            bool *error_is_read)
+                            bool ignore_ratelimit, bool *error_is_read)
 {
     BlockCopyCallState call_state = {
         .s = s,
         .offset = start,
         .bytes = bytes,
+        .ignore_ratelimit = ignore_ratelimit,
         .max_workers = BLOCK_COPY_MAX_WORKERS,
     };
 
@@ -798,3 +827,18 @@ void block_copy_set_skip_unallocated(BlockCopyState *s, bool skip)
 {
     s->skip_unallocated = skip;
 }
+
+void block_copy_set_speed(BlockCopyState *s, uint64_t speed)
+{
+    s->speed = speed;
+    if (speed > 0) {
+        ratelimit_set_speed(&s->rate_limit, speed, BLOCK_COPY_SLICE_TIME);
+    }
+
+    /*
+     * Note: it's good to kick all call states from here, but it should be done
+     * only from a coroutine, to not crash if s->calls list changed while
+     * entering one call. So for now, the only user of this function kicks its
+     * only one call_state by hand.
+     */
+}
-- 
2.29.2



  parent reply	other threads:[~2021-01-26 14:52 UTC|newest]

Thread overview: 59+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-01-26 14:19 [PULL 00/53] Block patches Max Reitz
2021-01-26 14:19 ` [PULL 01/53] iotests: fix _check_o_direct Max Reitz
2021-01-26 14:19 ` [PULL 02/53] copy-on-read: support preadv/pwritev_part functions Max Reitz
2021-01-26 14:19 ` [PULL 03/53] block: add API function to insert a node Max Reitz
2021-01-26 14:19 ` [PULL 04/53] copy-on-read: add filter drop function Max Reitz
2021-01-26 14:19 ` [PULL 05/53] qapi: add filter-node-name to block-stream Max Reitz
2021-01-26 14:19 ` [PULL 06/53] qapi: copy-on-read filter: add 'bottom' option Max Reitz
2021-01-26 14:19 ` [PULL 07/53] iotests: add #310 to test bottom node in COR driver Max Reitz
2021-01-26 14:19 ` [PULL 08/53] block: include supported_read_flags into BDS structure Max Reitz
2021-01-26 14:19 ` [PULL 09/53] copy-on-read: skip non-guest reads if no copy needed Max Reitz
2021-01-26 14:19 ` [PULL 10/53] stream: rework backing-file changing Max Reitz
2021-01-26 14:19 ` [PULL 11/53] qapi: block-stream: add "bottom" argument Max Reitz
2021-01-26 14:19 ` [PULL 12/53] iotests: 30: prepare to COR filter insertion by stream job Max Reitz
2021-01-26 14:19 ` [PULL 13/53] block/stream: add s->target_bs Max Reitz
2021-01-26 14:19 ` [PULL 14/53] block: apply COR-filter to block-stream jobs Max Reitz
2021-01-28 18:38   ` Philippe Mathieu-Daudé
2021-01-29  5:26     ` Vladimir Sementsov-Ogievskiy
2021-01-29  8:23       ` Max Reitz
2021-01-29  9:23         ` Philippe Mathieu-Daudé
2021-01-26 14:19 ` [PULL 15/53] iotests.py: Assume a couple of variables as given Max Reitz
2021-01-26 14:19 ` [PULL 16/53] iotests/297: Rewrite in Python and extend reach Max Reitz
2021-01-26 14:19 ` [PULL 17/53] iotests: Move try_remove to iotests.py Max Reitz
2021-01-26 14:19 ` [PULL 18/53] iotests/129: Remove test images in tearDown() Max Reitz
2021-01-26 14:19 ` [PULL 19/53] iotests/129: Do not check @busy Max Reitz
2021-01-26 14:19 ` [PULL 20/53] iotests/129: Use throttle node Max Reitz
2021-01-26 14:19 ` [PULL 21/53] iotests/129: Actually test a commit job Max Reitz
2021-01-26 14:19 ` [PULL 22/53] iotests/129: Limit mirror job's buffer size Max Reitz
2021-01-26 14:19 ` [PULL 23/53] iotests/129: Clean up pylint and mypy complaints Max Reitz
2021-01-26 14:19 ` [PULL 24/53] iotests/300: " Max Reitz
2021-01-26 14:19 ` [PULL 25/53] coroutine-sigaltstack: Add SIGUSR2 mutex Max Reitz
2021-01-26 14:19 ` [PULL 26/53] qapi: backup: add perf.use-copy-range parameter Max Reitz
2021-01-26 14:19 ` [PULL 27/53] block/block-copy: More explicit call_state Max Reitz
2021-01-26 14:19 ` [PULL 28/53] block/block-copy: implement block_copy_async Max Reitz
2021-01-26 14:19 ` [PULL 29/53] block/block-copy: add max_chunk and max_workers parameters Max Reitz
2021-01-26 14:19 ` [PULL 30/53] block/block-copy: add list of all call-states Max Reitz
2021-01-26 14:19 ` Max Reitz [this message]
2021-01-26 14:19 ` [PULL 32/53] block/block-copy: add block_copy_cancel Max Reitz
2021-01-26 14:19 ` [PULL 33/53] blockjob: add set_speed to BlockJobDriver Max Reitz
2021-01-26 14:19 ` [PULL 34/53] job: call job_enter from job_pause Max Reitz
2021-01-26 14:19 ` [PULL 35/53] qapi: backup: add max-chunk and max-workers to x-perf struct Max Reitz
2021-01-26 14:19 ` [PULL 36/53] iotests: 56: prepare for backup over block-copy Max Reitz
2021-01-26 14:20 ` [PULL 37/53] iotests/129: Limit backup's max-chunk/max-workers Max Reitz
2021-01-26 14:20 ` [PULL 38/53] iotests: 185: prepare for backup over block-copy Max Reitz
2021-01-26 14:20 ` [PULL 39/53] iotests: 219: " Max Reitz
2021-01-26 14:20 ` [PULL 40/53] iotests: 257: " Max Reitz
2021-01-26 14:20 ` [PULL 41/53] block/block-copy: make progress_bytes_callback optional Max Reitz
2021-01-26 14:20 ` [PULL 42/53] block/backup: drop extra gotos from backup_run() Max Reitz
2021-01-26 14:20 ` [PULL 43/53] backup: move to block-copy Max Reitz
2021-01-26 14:20 ` [PULL 44/53] qapi: backup: disable copy_range by default Max Reitz
2021-01-26 14:20 ` [PULL 45/53] block/block-copy: drop unused block_copy_set_progress_callback() Max Reitz
2021-01-26 14:20 ` [PULL 46/53] block/block-copy: drop unused argument of block_copy() Max Reitz
2021-01-26 14:20 ` [PULL 47/53] simplebench/bench_block_job: use correct shebang line with python3 Max Reitz
2021-01-26 14:20 ` [PULL 48/53] simplebench: bench_block_job: add cmd_options argument Max Reitz
2021-01-26 14:20 ` [PULL 49/53] simplebench: add bench-backup.py Max Reitz
2021-01-26 14:20 ` [PULL 50/53] block: report errno when flock fcntl fails Max Reitz
2021-01-26 14:20 ` [PULL 51/53] iotests: Add test for the regression fixed in c8bf9a9169 Max Reitz
2021-01-26 14:20 ` [PULL 52/53] iotests/118: Drop 'change' test Max Reitz
2021-01-26 14:20 ` [PULL 53/53] iotests/178: Pass value to invalid option Max Reitz
2021-01-27 17:40 ` [PULL 00/53] Block patches Peter Maydell

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=20210126142016.806073-32-mreitz@redhat.com \
    --to=mreitz@redhat.com \
    --cc=kwolf@redhat.com \
    --cc=peter.maydell@linaro.org \
    --cc=qemu-block@nongnu.org \
    --cc=qemu-devel@nongnu.org \
    /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).