[PATCH v2 0/6] block-copy: memory limit

[PATCH v2 0/6] block-copy: memory limit

[Vladimir Sementsov-Ogievskiy]

I'm going to bring block-status driven, async copying process to
block-copy, to make it fast. The first step is to limit memory usage of
backup, here is it.

v2: [mostly by Max's comments]
Now based on master (Thank you Max!)
01: add Max's r-b
02: add Max's r-b
03: - refactor block_copy_do_copy goto/return
    - add small comment to block_copy_do_copy
04: - a lot of renaming and wording fixes
    - refactor to use "available" instead of "taken"
    - refactor co_get_from_shres
05: rebase on 04 changes
06: drop extra things from max_transfer calculation

Vladimir Sementsov-Ogievskiy (6):
  block/block-copy: allocate buffer in block_copy_with_bounce_buffer
  block/block-copy: limit copy_range_size to 16 MiB
  block/block-copy: refactor copying
  util: introduce SharedResource
  block/block-copy: add memory limit
  block/block-copy: increase buffered copy request

 include/block/block-copy.h        |   5 +-
 include/qemu/co-shared-resource.h |  71 ++++++++++++
 block/block-copy.c                | 182 +++++++++++++++---------------
 util/qemu-co-shared-resource.c    |  76 +++++++++++++
 block/trace-events                |   6 +-
 util/Makefile.objs                |   1 +
 6 files changed, 249 insertions(+), 92 deletions(-)
 create mode 100644 include/qemu/co-shared-resource.h
 create mode 100644 util/qemu-co-shared-resource.c

-- 
2.21.0

[PATCH v2 1/6] block/block-copy: allocate buffer in block_copy_with_bounce_buffer

[Vladimir Sementsov-Ogievskiy]

Move bounce_buffer allocation block_copy_with_bounce_buffer. This
commit simplifies further work on implementing copying by larger chunks
(of different size) and further asynchronous handling of block_copy
iterations (with help of block/aio_task API).

Allocation works fast, a lot faster than disk io, so it's not a problem
that we now allocate/free bounce_buffer more times. And we anyway will
have to allocate several bounce_buffers for parallel execution of loop
iterations in future.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
---
 block/block-copy.c | 21 ++++++++-------------
 1 file changed, 8 insertions(+), 13 deletions(-)

diff --git a/block/block-copy.c b/block/block-copy.c
index 0f76ea1e63..22b0bd7d07 100644
--- a/block/block-copy.c
+++ b/block/block-copy.c
@@ -126,20 +126,17 @@ void block_copy_set_callbacks(
 static int coroutine_fn block_copy_with_bounce_buffer(BlockCopyState *s,
                                                       int64_t start,
                                                       int64_t end,
-                                                      bool *error_is_read,
-                                                      void **bounce_buffer)
+                                                      bool *error_is_read)
 {
     int ret;
     int nbytes;
+    void *bounce_buffer = qemu_blockalign(s->source->bs, s->cluster_size);
 
     assert(QEMU_IS_ALIGNED(start, s->cluster_size));
     bdrv_reset_dirty_bitmap(s->copy_bitmap, start, s->cluster_size);
     nbytes = MIN(s->cluster_size, s->len - start);
-    if (!*bounce_buffer) {
-        *bounce_buffer = qemu_blockalign(s->source->bs, s->cluster_size);
-    }
 
-    ret = bdrv_co_pread(s->source, start, nbytes, *bounce_buffer, 0);
+    ret = bdrv_co_pread(s->source, start, nbytes, bounce_buffer, 0);
     if (ret < 0) {
         trace_block_copy_with_bounce_buffer_read_fail(s, start, ret);
         if (error_is_read) {
@@ -148,7 +145,7 @@ static int coroutine_fn block_copy_with_bounce_buffer(BlockCopyState *s,
         goto fail;
     }
 
-    ret = bdrv_co_pwrite(s->target, start, nbytes, *bounce_buffer,
+    ret = bdrv_co_pwrite(s->target, start, nbytes, bounce_buffer,
                          s->write_flags);
     if (ret < 0) {
         trace_block_copy_with_bounce_buffer_write_fail(s, start, ret);
@@ -158,8 +155,11 @@ static int coroutine_fn block_copy_with_bounce_buffer(BlockCopyState *s,
         goto fail;
     }
 
+    qemu_vfree(bounce_buffer);
+
     return nbytes;
 fail:
+    qemu_vfree(bounce_buffer);
     bdrv_set_dirty_bitmap(s->copy_bitmap, start, s->cluster_size);
     return ret;
 
@@ -271,7 +271,6 @@ int coroutine_fn block_copy(BlockCopyState *s,
 {
     int ret = 0;
     int64_t end = bytes + start; /* bytes */
-    void *bounce_buffer = NULL;
     int64_t status_bytes;
     BlockCopyInFlightReq req;
 
@@ -324,7 +323,7 @@ int coroutine_fn block_copy(BlockCopyState *s,
         }
         if (!s->use_copy_range) {
             ret = block_copy_with_bounce_buffer(s, start, dirty_end,
-                                                error_is_read, &bounce_buffer);
+                                                error_is_read);
         }
         if (ret < 0) {
             break;
@@ -335,10 +334,6 @@ int coroutine_fn block_copy(BlockCopyState *s,
         ret = 0;
     }
 
-    if (bounce_buffer) {
-        qemu_vfree(bounce_buffer);
-    }
-
     block_copy_inflight_req_end(&req);
 
     return ret;
-- 
2.21.0

[PATCH v2 2/6] block/block-copy: limit copy_range_size to 16 MiB

[Vladimir Sementsov-Ogievskiy]

Large copy range may imply memory allocation and large io effort, so
using 2G copy range request may be bad idea. Let's limit it to 16 MiB.
It also helps the following patch to refactor copy-with-offload
fallback to copy-with-bounce-buffer.

Note, that total memory usage of backup is still not limited, it will
be fixed in further commit.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
---
 block/block-copy.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/block/block-copy.c b/block/block-copy.c
index 22b0bd7d07..e37dfbfd03 100644
--- a/block/block-copy.c
+++ b/block/block-copy.c
@@ -18,6 +18,9 @@
 #include "qapi/error.h"
 #include "block/block-copy.h"
 #include "sysemu/block-backend.h"
+#include "qemu/units.h"
+
+#define BLOCK_COPY_MAX_COPY_RANGE (16 * MiB)
 
 static void coroutine_fn block_copy_wait_inflight_reqs(BlockCopyState *s,
                                                        int64_t start,
@@ -70,9 +73,12 @@ BlockCopyState *block_copy_state_new(BdrvChild *source, BdrvChild *target,
 {
     BlockCopyState *s;
     BdrvDirtyBitmap *copy_bitmap;
+
+    /* Ignore BLOCK_COPY_MAX_COPY_RANGE if requested cluster_size is larger */
     uint32_t max_transfer =
-            MIN_NON_ZERO(INT_MAX, MIN_NON_ZERO(source->bs->bl.max_transfer,
-                                               target->bs->bl.max_transfer));
+            MIN_NON_ZERO(MAX(cluster_size, BLOCK_COPY_MAX_COPY_RANGE),
+                         MIN_NON_ZERO(source->bs->bl.max_transfer,
+                                      target->bs->bl.max_transfer));
 
     copy_bitmap = bdrv_create_dirty_bitmap(source->bs, cluster_size, NULL,
                                            errp);
-- 
2.21.0

[PATCH v2 3/6] block/block-copy: refactor copying

[Vladimir Sementsov-Ogievskiy]

Merge copying code into one function block_copy_do_copy, which only
calls bdrv_ io functions and don't do any synchronization (like dirty
bitmap set/reset).

Refactor block_copy() function so that it takes full decision about
size of chunk to be copied and does all the synchronization (checking
intersecting requests, set/reset dirty bitmaps).

It will help:
 - introduce parallel processing of block_copy iterations: we need to
   calculate chunk size, start async chunk copying and go to the next
   iteration
 - simplify synchronization improvement (like memory limiting in
   further commit and reducing critical section (now we lock the whole
   requested range, when actually we need to lock only dirty region
   which we handle at the moment))

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
---
 block/block-copy.c | 118 ++++++++++++++++++++-------------------------
 block/trace-events |   6 +--
 2 files changed, 54 insertions(+), 70 deletions(-)

diff --git a/block/block-copy.c b/block/block-copy.c
index e37dfbfd03..c21db48734 100644
--- a/block/block-copy.c
+++ b/block/block-copy.c
@@ -126,79 +126,64 @@ void block_copy_set_callbacks(
 }
 
 /*
- * Copy range to target with a bounce buffer and return the bytes copied. If
- * error occurred, return a negative error number
+ * block_copy_do_copy
+ *
+ * Do copy of cluser-aligned chunk. @end is allowed to exceed s->len only to
+ * cover last cluster when s->len is not aligned to clusters.
+ *
+ * No sync here: nor bitmap neighter intersecting requests handling, only copy.
+ *
+ * Returns 0 on success.
  */
-static int coroutine_fn block_copy_with_bounce_buffer(BlockCopyState *s,
-                                                      int64_t start,
-                                                      int64_t end,
-                                                      bool *error_is_read)
+static int coroutine_fn block_copy_do_copy(BlockCopyState *s,
+                                           int64_t start, int64_t end,
+                                           bool *error_is_read)
 {
     int ret;
-    int nbytes;
-    void *bounce_buffer = qemu_blockalign(s->source->bs, s->cluster_size);
+    int nbytes = MIN(end, s->len) - start;
+    void *bounce_buffer = NULL;
 
     assert(QEMU_IS_ALIGNED(start, s->cluster_size));
-    bdrv_reset_dirty_bitmap(s->copy_bitmap, start, s->cluster_size);
-    nbytes = MIN(s->cluster_size, s->len - start);
+    assert(QEMU_IS_ALIGNED(end, s->cluster_size));
+    assert(end < s->len || end == QEMU_ALIGN_UP(s->len, s->cluster_size));
+
+    if (s->use_copy_range) {
+        ret = bdrv_co_copy_range(s->source, start, s->target, start, nbytes,
+                                 0, s->write_flags);
+        if (ret < 0) {
+            trace_block_copy_copy_range_fail(s, start, ret);
+            s->use_copy_range = false;
+            /* Fallback to read+write with allocated buffer */
+        } else {
+            goto out;
+        }
+    }
+
+    bounce_buffer = qemu_blockalign(s->source->bs, nbytes);
 
     ret = bdrv_co_pread(s->source, start, nbytes, bounce_buffer, 0);
     if (ret < 0) {
-        trace_block_copy_with_bounce_buffer_read_fail(s, start, ret);
+        trace_block_copy_read_fail(s, start, ret);
         if (error_is_read) {
             *error_is_read = true;
         }
-        goto fail;
+        goto out;
     }
 
     ret = bdrv_co_pwrite(s->target, start, nbytes, bounce_buffer,
                          s->write_flags);
     if (ret < 0) {
-        trace_block_copy_with_bounce_buffer_write_fail(s, start, ret);
+        trace_block_copy_write_fail(s, start, ret);
         if (error_is_read) {
             *error_is_read = false;
         }
-        goto fail;
+        goto out;
     }
 
+out:
     qemu_vfree(bounce_buffer);
 
-    return nbytes;
-fail:
-    qemu_vfree(bounce_buffer);
-    bdrv_set_dirty_bitmap(s->copy_bitmap, start, s->cluster_size);
     return ret;
-
-}
-
-/*
- * Copy range to target and return the bytes copied. If error occurred, return a
- * negative error number.
- */
-static int coroutine_fn block_copy_with_offload(BlockCopyState *s,
-                                                int64_t start,
-                                                int64_t end)
-{
-    int ret;
-    int nr_clusters;
-    int nbytes;
-
-    assert(QEMU_IS_ALIGNED(s->copy_range_size, s->cluster_size));
-    assert(QEMU_IS_ALIGNED(start, s->cluster_size));
-    nbytes = MIN(s->copy_range_size, MIN(end, s->len) - start);
-    nr_clusters = DIV_ROUND_UP(nbytes, s->cluster_size);
-    bdrv_reset_dirty_bitmap(s->copy_bitmap, start,
-                            s->cluster_size * nr_clusters);
-    ret = bdrv_co_copy_range(s->source, start, s->target, start, nbytes,
-                             0, s->write_flags);
-    if (ret < 0) {
-        trace_block_copy_with_offload_fail(s, start, ret);
-        bdrv_set_dirty_bitmap(s->copy_bitmap, start,
-                              s->cluster_size * nr_clusters);
-        return ret;
-    }
-
-    return nbytes;
 }
 
 /*
@@ -294,7 +279,7 @@ int coroutine_fn block_copy(BlockCopyState *s,
     block_copy_inflight_req_begin(s, &req, start, end);
 
     while (start < end) {
-        int64_t dirty_end;
+        int64_t next_zero, chunk_end;
 
         if (!bdrv_dirty_bitmap_get(s->copy_bitmap, start)) {
             trace_block_copy_skip(s, start);
@@ -302,10 +287,15 @@ int coroutine_fn block_copy(BlockCopyState *s,
             continue; /* already copied */
         }
 
-        dirty_end = bdrv_dirty_bitmap_next_zero(s->copy_bitmap, start,
-                                                (end - start));
-        if (dirty_end < 0) {
-            dirty_end = end;
+        chunk_end = MIN(end, start + (s->use_copy_range ?
+                                      s->copy_range_size : s->cluster_size));
+
+        next_zero = bdrv_dirty_bitmap_next_zero(s->copy_bitmap, start,
+                                                chunk_end - start);
+        if (next_zero >= 0) {
+            assert(next_zero > start); /* start is dirty */
+            assert(next_zero < chunk_end); /* no need to do MIN() */
+            chunk_end = next_zero;
         }
 
         if (s->skip_unallocated) {
@@ -316,27 +306,21 @@ int coroutine_fn block_copy(BlockCopyState *s,
                 continue;
             }
             /* Clamp to known allocated region */
-            dirty_end = MIN(dirty_end, start + status_bytes);
+            chunk_end = MIN(chunk_end, start + status_bytes);
         }
 
         trace_block_copy_process(s, start);
 
-        if (s->use_copy_range) {
-            ret = block_copy_with_offload(s, start, dirty_end);
-            if (ret < 0) {
-                s->use_copy_range = false;
-            }
-        }
-        if (!s->use_copy_range) {
-            ret = block_copy_with_bounce_buffer(s, start, dirty_end,
-                                                error_is_read);
-        }
+        bdrv_reset_dirty_bitmap(s->copy_bitmap, start, chunk_end - start);
+
+        ret = block_copy_do_copy(s, start, chunk_end, error_is_read);
         if (ret < 0) {
+            bdrv_set_dirty_bitmap(s->copy_bitmap, start, chunk_end - start);
             break;
         }
 
-        start += ret;
-        s->progress_bytes_callback(ret, s->progress_opaque);
+        s->progress_bytes_callback(chunk_end - start, s->progress_opaque);
+        start = chunk_end;
         ret = 0;
     }
 
diff --git a/block/trace-events b/block/trace-events
index b8d70f5242..ccde15a14c 100644
--- a/block/trace-events
+++ b/block/trace-events
@@ -45,9 +45,9 @@ backup_do_cow_return(void *job, int64_t offset, uint64_t bytes, int ret) "job %p
 block_copy_skip(void *bcs, int64_t start) "bcs %p start %"PRId64
 block_copy_skip_range(void *bcs, int64_t start, uint64_t bytes) "bcs %p start %"PRId64" bytes %"PRId64
 block_copy_process(void *bcs, int64_t start) "bcs %p start %"PRId64
-block_copy_with_bounce_buffer_read_fail(void *bcs, int64_t start, int ret) "bcs %p start %"PRId64" ret %d"
-block_copy_with_bounce_buffer_write_fail(void *bcs, int64_t start, int ret) "bcs %p start %"PRId64" ret %d"
-block_copy_with_offload_fail(void *bcs, int64_t start, int ret) "bcs %p start %"PRId64" ret %d"
+block_copy_copy_range_fail(void *bcs, int64_t start, int ret) "bcs %p start %"PRId64" ret %d"
+block_copy_read_fail(void *bcs, int64_t start, int ret) "bcs %p start %"PRId64" ret %d"
+block_copy_write_fail(void *bcs, int64_t start, int ret) "bcs %p start %"PRId64" ret %d"
 
 # ../blockdev.c
 qmp_block_job_cancel(void *job) "job %p"
-- 
2.21.0

Re: [PATCH v2 3/6] block/block-copy: refactor copying

[Max Reitz]

[-- Attachment #1.1: Type: text/plain, Size: 1152 bytes --]

On 16.10.19 19:09, Vladimir Sementsov-Ogievskiy wrote:
> Merge copying code into one function block_copy_do_copy, which only
> calls bdrv_ io functions and don't do any synchronization (like dirty
> bitmap set/reset).
> 
> Refactor block_copy() function so that it takes full decision about
> size of chunk to be copied and does all the synchronization (checking
> intersecting requests, set/reset dirty bitmaps).
> 
> It will help:
>  - introduce parallel processing of block_copy iterations: we need to
>    calculate chunk size, start async chunk copying and go to the next
>    iteration
>  - simplify synchronization improvement (like memory limiting in
>    further commit and reducing critical section (now we lock the whole
>    requested range, when actually we need to lock only dirty region
>    which we handle at the moment))
> 
> Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
> ---
>  block/block-copy.c | 118 ++++++++++++++++++++-------------------------
>  block/trace-events |   6 +--
>  2 files changed, 54 insertions(+), 70 deletions(-)

Reviewed-by: Max Reitz <mreitz@redhat.com>


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

[PATCH v2 4/6] util: introduce SharedResource

[Vladimir Sementsov-Ogievskiy]

Introduce an API for some shared splittable resource, like memory.
It's going to be used by backup. Backup uses both read/write io and
copy_range. copy_range may consume memory implictly, so the new API is
abstract: it doesn't allocate any real memory by but only hands out
tickets.

The idea is that we have some total amount of something and callers
should wait in coroutine queue if there is not enough of the resource
at the moment.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
---
 include/qemu/co-shared-resource.h | 71 +++++++++++++++++++++++++++++
 util/qemu-co-shared-resource.c    | 76 +++++++++++++++++++++++++++++++
 util/Makefile.objs                |  1 +
 3 files changed, 148 insertions(+)
 create mode 100644 include/qemu/co-shared-resource.h
 create mode 100644 util/qemu-co-shared-resource.c

diff --git a/include/qemu/co-shared-resource.h b/include/qemu/co-shared-resource.h
new file mode 100644
index 0000000000..04c9c3d5be
--- /dev/null
+++ b/include/qemu/co-shared-resource.h
@@ -0,0 +1,71 @@
+/*
+ * Helper functionality for distributing a fixed total amount of
+ * an abstract resource among multiple coroutines.
+ *
+ * Copyright (c) 2019 Virtuozzo International GmbH
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+ * THE SOFTWARE.
+ */
+
+#ifndef QEMU_CO_SHARED_AMOUNT_H
+#define QEMU_CO_SHARED_AMOUNT_H
+
+
+typedef struct SharedResource SharedResource;
+
+/*
+ * Create SharedResource structure
+ *
+ * @total: total amount of some resource to be shared between clients
+ *
+ * Note: this API is not thread-safe.
+ */
+SharedResource *shres_create(uint64_t total);
+
+/*
+ * Release SharedResource structure
+ *
+ * This function may only be called once everything allocated by all
+ * clients has been deallocated.
+ */
+void shres_destroy(SharedResource *s);
+
+/*
+ * Try to allocate an amount of @n.  Return true on success, and false
+ * if there is too little left of the collective resource to fulfill
+ * the request.
+ */
+bool co_try_get_from_shres(SharedResource *s, uint64_t n);
+
+/*
+ * Allocate an amount of @n, and, if necessary, yield until
+ * that becomes possible.
+ */
+void coroutine_fn co_get_from_shres(SharedResource *s, uint64_t n);
+
+/*
+ * Deallocate an amount of @n.  The total amount allocated by a caller
+ * does not need to be deallocated/released with a single call, but may
+ * be split over several calls.  For example, get(4), get(3), and then
+ * put(5), put(2).
+ */
+void coroutine_fn co_put_to_shres(SharedResource *s, uint64_t n);
+
+
+#endif /* QEMU_CO_SHARED_AMOUNT_H */
diff --git a/util/qemu-co-shared-resource.c b/util/qemu-co-shared-resource.c
new file mode 100644
index 0000000000..1c83cd9d29
--- /dev/null
+++ b/util/qemu-co-shared-resource.c
@@ -0,0 +1,76 @@
+/*
+ * Helper functionality for distributing a fixed total amount of
+ * an abstract resource among multiple coroutines.
+ *
+ * Copyright (c) 2019 Virtuozzo International GmbH
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+ * THE SOFTWARE.
+ */
+
+#include "qemu/osdep.h"
+#include "qemu/coroutine.h"
+#include "qemu/co-shared-resource.h"
+
+struct SharedResource {
+    uint64_t total;
+    uint64_t available;
+
+    CoQueue queue;
+};
+
+SharedResource *shres_create(uint64_t total)
+{
+    SharedResource *s = g_new0(SharedResource, 1);
+
+    s->total = s->available = total;
+    qemu_co_queue_init(&s->queue);
+
+    return s;
+}
+
+void shres_destroy(SharedResource *s)
+{
+    assert(s->available == s->total);
+    g_free(s);
+}
+
+bool co_try_get_from_shres(SharedResource *s, uint64_t n)
+{
+    if (s->available >= n) {
+        s->available -= n;
+        return true;
+    }
+
+    return false;
+}
+
+void coroutine_fn co_get_from_shres(SharedResource *s, uint64_t n)
+{
+    assert(n <= s->total);
+    while (!co_try_get_from_shres(s, n)) {
+        qemu_co_queue_wait(&s->queue, NULL);
+    }
+}
+
+void coroutine_fn co_put_to_shres(SharedResource *s, uint64_t n)
+{
+    assert(s->total - s->available >= n);
+    s->available += n;
+    qemu_co_queue_restart_all(&s->queue);
+}
diff --git a/util/Makefile.objs b/util/Makefile.objs
index 41bf59d127..df124af1c5 100644
--- a/util/Makefile.objs
+++ b/util/Makefile.objs
@@ -37,6 +37,7 @@ util-obj-y += rcu.o
 util-obj-$(CONFIG_MEMBARRIER) += sys_membarrier.o
 util-obj-y += qemu-coroutine.o qemu-coroutine-lock.o qemu-coroutine-io.o
 util-obj-y += qemu-coroutine-sleep.o
+util-obj-y += qemu-co-shared-resource.o
 util-obj-y += coroutine-$(CONFIG_COROUTINE_BACKEND).o
 util-obj-y += buffer.o
 util-obj-y += timed-average.o
-- 
2.21.0

Re: [PATCH v2 4/6] util: introduce SharedResource

[Max Reitz]

[-- Attachment #1.1: Type: text/plain, Size: 2760 bytes --]

On 16.10.19 19:09, Vladimir Sementsov-Ogievskiy wrote:
> Introduce an API for some shared splittable resource, like memory.
> It's going to be used by backup. Backup uses both read/write io and
> copy_range. copy_range may consume memory implictly, so the new API is
> abstract: it doesn't allocate any real memory by but only hands out

-by

> tickets.
> 
> The idea is that we have some total amount of something and callers
> should wait in coroutine queue if there is not enough of the resource
> at the moment.
> 
> Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
> ---
>  include/qemu/co-shared-resource.h | 71 +++++++++++++++++++++++++++++
>  util/qemu-co-shared-resource.c    | 76 +++++++++++++++++++++++++++++++
>  util/Makefile.objs                |  1 +
>  3 files changed, 148 insertions(+)
>  create mode 100644 include/qemu/co-shared-resource.h
>  create mode 100644 util/qemu-co-shared-resource.c
> 
> diff --git a/include/qemu/co-shared-resource.h b/include/qemu/co-shared-resource.h
> new file mode 100644
> index 0000000000..04c9c3d5be
> --- /dev/null
> +++ b/include/qemu/co-shared-resource.h
> @@ -0,0 +1,71 @@
> +/*
> + * Helper functionality for distributing a fixed total amount of
> + * an abstract resource among multiple coroutines.
> + *
> + * Copyright (c) 2019 Virtuozzo International GmbH
> + *
> + * Permission is hereby granted, free of charge, to any person obtaining a copy
> + * of this software and associated documentation files (the "Software"), to deal
> + * in the Software without restriction, including without limitation the rights
> + * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
> + * copies of the Software, and to permit persons to whom the Software is
> + * furnished to do so, subject to the following conditions:
> + *
> + * The above copyright notice and this permission notice shall be included in
> + * all copies or substantial portions of the Software.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
> + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
> + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
> + * THE SOFTWARE.
> + */
> +
> +#ifndef QEMU_CO_SHARED_AMOUNT_H
> +#define QEMU_CO_SHARED_AMOUNT_H

You should probably adjust these include guards to match the new filename.

With those bits fixed:

Reviewed-by: Max Reitz <mreitz@redhat.com>


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH v2 4/6] util: introduce SharedResource

[Vladimir Sementsov-Ogievskiy]

22.10.2019 13:23, Max Reitz wrote:
> On 16.10.19 19:09, Vladimir Sementsov-Ogievskiy wrote:
>> Introduce an API for some shared splittable resource, like memory.
>> It's going to be used by backup. Backup uses both read/write io and
>> copy_range. copy_range may consume memory implictly, so the new API is
>> abstract: it doesn't allocate any real memory by but only hands out
> 
> -by
> 
>> tickets.
>>
>> The idea is that we have some total amount of something and callers
>> should wait in coroutine queue if there is not enough of the resource
>> at the moment.
>>
>> Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
>> ---
>>   include/qemu/co-shared-resource.h | 71 +++++++++++++++++++++++++++++
>>   util/qemu-co-shared-resource.c    | 76 +++++++++++++++++++++++++++++++
>>   util/Makefile.objs                |  1 +
>>   3 files changed, 148 insertions(+)
>>   create mode 100644 include/qemu/co-shared-resource.h
>>   create mode 100644 util/qemu-co-shared-resource.c
>>
>> diff --git a/include/qemu/co-shared-resource.h b/include/qemu/co-shared-resource.h
>> new file mode 100644
>> index 0000000000..04c9c3d5be
>> --- /dev/null
>> +++ b/include/qemu/co-shared-resource.h
>> @@ -0,0 +1,71 @@
>> +/*
>> + * Helper functionality for distributing a fixed total amount of
>> + * an abstract resource among multiple coroutines.
>> + *
>> + * Copyright (c) 2019 Virtuozzo International GmbH
>> + *
>> + * Permission is hereby granted, free of charge, to any person obtaining a copy
>> + * of this software and associated documentation files (the "Software"), to deal
>> + * in the Software without restriction, including without limitation the rights
>> + * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
>> + * copies of the Software, and to permit persons to whom the Software is
>> + * furnished to do so, subject to the following conditions:
>> + *
>> + * The above copyright notice and this permission notice shall be included in
>> + * all copies or substantial portions of the Software.
>> + *
>> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
>> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
>> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
>> + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
>> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
>> + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
>> + * THE SOFTWARE.
>> + */
>> +
>> +#ifndef QEMU_CO_SHARED_AMOUNT_H
>> +#define QEMU_CO_SHARED_AMOUNT_H
> 
> You should probably adjust these include guards to match the new filename.
> 
> With those bits fixed:
> 
> Reviewed-by: Max Reitz <mreitz@redhat.com>
> 

Thank for reviewing! I'll fix it and resend.

-- 
Best regards,
Vladimir

[PATCH v2 5/6] block/block-copy: add memory limit

[Vladimir Sementsov-Ogievskiy]

Currently total allocation for parallel requests to block-copy instance
is unlimited. Let's limit it to 128 MiB.

For now block-copy is used only in backup, so actually we limit total
allocation for backup job.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
---
 include/block/block-copy.h | 3 +++
 block/block-copy.c         | 5 +++++
 2 files changed, 8 insertions(+)

diff --git a/include/block/block-copy.h b/include/block/block-copy.h
index e2e135ff1b..edcdf0072d 100644
--- a/include/block/block-copy.h
+++ b/include/block/block-copy.h
@@ -16,6 +16,7 @@
 #define BLOCK_COPY_H
 
 #include "block/block.h"
+#include "qemu/co-shared-resource.h"
 
 typedef struct BlockCopyInFlightReq {
     int64_t start_byte;
@@ -69,6 +70,8 @@ typedef struct BlockCopyState {
      */
     ProgressResetCallbackFunc progress_reset_callback;
     void *progress_opaque;
+
+    SharedResource *mem;
 } BlockCopyState;
 
 BlockCopyState *block_copy_state_new(BdrvChild *source, BdrvChild *target,
diff --git a/block/block-copy.c b/block/block-copy.c
index c21db48734..d5042e46fd 100644
--- a/block/block-copy.c
+++ b/block/block-copy.c
@@ -21,6 +21,7 @@
 #include "qemu/units.h"
 
 #define BLOCK_COPY_MAX_COPY_RANGE (16 * MiB)
+#define BLOCK_COPY_MAX_MEM (128 * MiB)
 
 static void coroutine_fn block_copy_wait_inflight_reqs(BlockCopyState *s,
                                                        int64_t start,
@@ -64,6 +65,7 @@ void block_copy_state_free(BlockCopyState *s)
     }
 
     bdrv_release_dirty_bitmap(s->source->bs, s->copy_bitmap);
+    shres_destroy(s->mem);
     g_free(s);
 }
 
@@ -95,6 +97,7 @@ BlockCopyState *block_copy_state_new(BdrvChild *source, BdrvChild *target,
         .cluster_size = cluster_size,
         .len = bdrv_dirty_bitmap_size(copy_bitmap),
         .write_flags = write_flags,
+        .mem = shres_create(BLOCK_COPY_MAX_MEM),
     };
 
     s->copy_range_size = QEMU_ALIGN_DOWN(max_transfer, cluster_size),
@@ -313,7 +316,9 @@ int coroutine_fn block_copy(BlockCopyState *s,
 
         bdrv_reset_dirty_bitmap(s->copy_bitmap, start, chunk_end - start);
 
+        co_get_from_shres(s->mem, chunk_end - start);
         ret = block_copy_do_copy(s, start, chunk_end, error_is_read);
+        co_put_to_shres(s->mem, chunk_end - start);
         if (ret < 0) {
             bdrv_set_dirty_bitmap(s->copy_bitmap, start, chunk_end - start);
             break;
-- 
2.21.0

Re: [PATCH v2 5/6] block/block-copy: add memory limit

[Max Reitz]

[-- Attachment #1.1: Type: text/plain, Size: 532 bytes --]

On 16.10.19 19:09, Vladimir Sementsov-Ogievskiy wrote:
> Currently total allocation for parallel requests to block-copy instance
> is unlimited. Let's limit it to 128 MiB.
> 
> For now block-copy is used only in backup, so actually we limit total
> allocation for backup job.
> 
> Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
> ---
>  include/block/block-copy.h | 3 +++
>  block/block-copy.c         | 5 +++++
>  2 files changed, 8 insertions(+)

Reviewed-by: Max Reitz <mreitz@redhat.com>


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

[PATCH v2 6/6] block/block-copy: increase buffered copy request

[Vladimir Sementsov-Ogievskiy]

No reason to limit buffered copy to one cluster. Let's allow up to 1
MiB.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
---
 include/block/block-copy.h |  2 +-
 block/block-copy.c         | 48 +++++++++++++++++++++++++-------------
 2 files changed, 33 insertions(+), 17 deletions(-)

diff --git a/include/block/block-copy.h b/include/block/block-copy.h
index edcdf0072d..0a161724d7 100644
--- a/include/block/block-copy.h
+++ b/include/block/block-copy.h
@@ -38,7 +38,7 @@ typedef struct BlockCopyState {
     BdrvDirtyBitmap *copy_bitmap;
     int64_t cluster_size;
     bool use_copy_range;
-    int64_t copy_range_size;
+    int64_t copy_size;
     uint64_t len;
     QLIST_HEAD(, BlockCopyInFlightReq) inflight_reqs;
 
diff --git a/block/block-copy.c b/block/block-copy.c
index d5042e46fd..74eb235972 100644
--- a/block/block-copy.c
+++ b/block/block-copy.c
@@ -21,6 +21,7 @@
 #include "qemu/units.h"
 
 #define BLOCK_COPY_MAX_COPY_RANGE (16 * MiB)
+#define BLOCK_COPY_MAX_BUFFER (1 * MiB)
 #define BLOCK_COPY_MAX_MEM (128 * MiB)
 
 static void coroutine_fn block_copy_wait_inflight_reqs(BlockCopyState *s,
@@ -75,10 +76,8 @@ BlockCopyState *block_copy_state_new(BdrvChild *source, BdrvChild *target,
 {
     BlockCopyState *s;
     BdrvDirtyBitmap *copy_bitmap;
-
-    /* Ignore BLOCK_COPY_MAX_COPY_RANGE if requested cluster_size is larger */
     uint32_t max_transfer =
-            MIN_NON_ZERO(MAX(cluster_size, BLOCK_COPY_MAX_COPY_RANGE),
+            MIN_NON_ZERO(INT_MAX,
                          MIN_NON_ZERO(source->bs->bl.max_transfer,
                                       target->bs->bl.max_transfer));
 
@@ -100,17 +99,28 @@ BlockCopyState *block_copy_state_new(BdrvChild *source, BdrvChild *target,
         .mem = shres_create(BLOCK_COPY_MAX_MEM),
     };
 
-    s->copy_range_size = QEMU_ALIGN_DOWN(max_transfer, cluster_size),
-    /*
-     * Set use_copy_range, consider the following:
-     * 1. Compression is not supported for copy_range.
-     * 2. copy_range does not respect max_transfer (it's a TODO), so we factor
-     *    that in here. If max_transfer is smaller than the job->cluster_size,
-     *    we do not use copy_range (in that case it's zero after aligning down
-     *    above).
-     */
-    s->use_copy_range =
-        !(write_flags & BDRV_REQ_WRITE_COMPRESSED) && s->copy_range_size > 0;
+    if (max_transfer < cluster_size) {
+        /*
+         * copy_range does not respect max_transfer. We don't want to bother
+         * with requests smaller than block-copy cluster size, so fallback to
+         * buffered copying (read and write respect max_transfer on their
+         * behalf).
+         */
+        s->use_copy_range = false;
+        s->copy_size = cluster_size;
+    } else if (write_flags & BDRV_REQ_WRITE_COMPRESSED) {
+        /* Compression is not supported for copy_range */
+        s->use_copy_range = false;
+        s->copy_size = MAX(cluster_size, BLOCK_COPY_MAX_BUFFER);
+    } else {
+        /*
+         * copy_range does not respect max_transfer (it's a TODO), so we factor
+         * that in here.
+         */
+        s->use_copy_range = true;
+        s->copy_size = MIN(MAX(cluster_size, BLOCK_COPY_MAX_COPY_RANGE),
+                           QEMU_ALIGN_DOWN(max_transfer, cluster_size));
+    }
 
     QLIST_INIT(&s->inflight_reqs);
 
@@ -156,12 +166,19 @@ static int coroutine_fn block_copy_do_copy(BlockCopyState *s,
         if (ret < 0) {
             trace_block_copy_copy_range_fail(s, start, ret);
             s->use_copy_range = false;
+            s->copy_size = MAX(s->cluster_size, BLOCK_COPY_MAX_BUFFER);
             /* Fallback to read+write with allocated buffer */
         } else {
             goto out;
         }
     }
 
+    /*
+     * In case of failed copy_range request above, we may proceed with buffered
+     * request larger than BLOCK_COPY_MAX_BUFFER. Still, further requests will
+     * be properly limited, so don't care too much.
+     */
+
     bounce_buffer = qemu_blockalign(s->source->bs, nbytes);
 
     ret = bdrv_co_pread(s->source, start, nbytes, bounce_buffer, 0);
@@ -290,8 +307,7 @@ int coroutine_fn block_copy(BlockCopyState *s,
             continue; /* already copied */
         }
 
-        chunk_end = MIN(end, start + (s->use_copy_range ?
-                                      s->copy_range_size : s->cluster_size));
+        chunk_end = MIN(end, start + s->copy_size);
 
         next_zero = bdrv_dirty_bitmap_next_zero(s->copy_bitmap, start,
                                                 chunk_end - start);
-- 
2.21.0

Re: [PATCH v2 6/6] block/block-copy: increase buffered copy request

[Max Reitz]

[-- Attachment #1.1: Type: text/plain, Size: 436 bytes --]

On 16.10.19 19:09, Vladimir Sementsov-Ogievskiy wrote:
> No reason to limit buffered copy to one cluster. Let's allow up to 1
> MiB.
> 
> Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
> ---
>  include/block/block-copy.h |  2 +-
>  block/block-copy.c         | 48 +++++++++++++++++++++++++-------------
>  2 files changed, 33 insertions(+), 17 deletions(-)
Reviewed-by: Max Reitz <mreitz@redhat.com>


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]