[PATCH 06/11] block: intoduce reqlist

[Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>]

Split intersecting-requests functionality out of block-copy to be
reused in copy-before-write filter.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
---
 include/block/reqlist.h |  45 ++++++++++++++++
 block/block-copy.c      | 116 +++++++++++++---------------------------
 block/reqlist.c         |  94 ++++++++++++++++++++++++++++++++
 block/meson.build       |   1 +
 4 files changed, 177 insertions(+), 79 deletions(-)
 create mode 100644 include/block/reqlist.h
 create mode 100644 block/reqlist.c

diff --git a/include/block/reqlist.h b/include/block/reqlist.h
new file mode 100644
index 0000000000..2de86be300
--- /dev/null
+++ b/include/block/reqlist.h
@@ -0,0 +1,45 @@
+/*
+ * block_copy API
+ *
+ * Copyright (C) 2013 Proxmox Server Solutions
+ * Copyright (c) 2021 Virtuozzo International GmbH.
+ *
+ * Authors:
+ *  Dietmar Maurer (dietmar@proxmox.com)
+ *  Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ */
+
+#ifndef REQLIST_H
+#define REQLIST_H
+
+#include "qemu/coroutine.h"
+
+/*
+ * The API is not thread-safe and shouldn't be. The struct is public to be part
+ * of other structures and protected by third-party locks, see
+ * block/block-copy.c for example.
+ */
+
+typedef struct BlockReq {
+    int64_t offset;
+    int64_t bytes;
+
+    CoQueue wait_queue; /* coroutines blocked on this req */
+    QLIST_ENTRY(BlockReq) list;
+} BlockReq;
+
+typedef QLIST_HEAD(, BlockReq) BlockReqList;
+
+void reqlist_init_req(BlockReqList *reqs, BlockReq *req, int64_t offset,
+                      int64_t bytes);
+BlockReq *reqlist_find_conflict(BlockReqList *reqs, int64_t offset,
+                                int64_t bytes);
+bool coroutine_fn reqlist_wait_one(BlockReqList *reqs, int64_t offset,
+                                   int64_t bytes, CoMutex *lock);
+void coroutine_fn reqlist_shrink_req(BlockReq *req, int64_t new_bytes);
+void coroutine_fn reqlist_remove_req(BlockReq *req);
+
+#endif /* REQLIST_H */
diff --git a/block/block-copy.c b/block/block-copy.c
index 5cd291727b..de388b7b96 100644
--- a/block/block-copy.c
+++ b/block/block-copy.c
@@ -17,6 +17,7 @@
 #include "trace.h"
 #include "qapi/error.h"
 #include "block/block-copy.h"
+#include "block/reqlist.h"
 #include "sysemu/block-backend.h"
 #include "qemu/units.h"
 #include "qemu/coroutine.h"
@@ -82,7 +83,6 @@ typedef struct BlockCopyTask {
      */
     BlockCopyState *s;
     BlockCopyCallState *call_state;
-    int64_t offset;
     /*
      * @method can also be set again in the while loop of
      * block_copy_dirty_clusters(), but it is never accessed concurrently
@@ -93,21 +93,17 @@ typedef struct BlockCopyTask {
     BlockCopyMethod method;
 
     /*
-     * Fields whose state changes throughout the execution
-     * Protected by lock in BlockCopyState.
+     * Generally, req is protected by lock in BlockCopyState, Still req.offset
+     * is only set on task creation, so may be read concurrently after creation.
+     * req.bytes is changed at most once, and need only protecting the case of
+     * parallel read while updating @bytes value in block_copy_task_shrink().
      */
-    CoQueue wait_queue; /* coroutines blocked on this task */
-    /*
-     * Only protect the case of parallel read while updating @bytes
-     * value in block_copy_task_shrink().
-     */
-    int64_t bytes;
-    QLIST_ENTRY(BlockCopyTask) list;
+    BlockReq req;
 } BlockCopyTask;
 
 static int64_t task_end(BlockCopyTask *task)
 {
-    return task->offset + task->bytes;
+    return task->req.offset + task->req.bytes;
 }
 
 typedef struct BlockCopyState {
@@ -135,7 +131,7 @@ typedef struct BlockCopyState {
     CoMutex lock;
     int64_t in_flight_bytes;
     BlockCopyMethod method;
-    QLIST_HEAD(, BlockCopyTask) tasks; /* All tasks from all block-copy calls */
+    BlockReqList reqs;
     QLIST_HEAD(, BlockCopyCallState) calls;
     /*
      * skip_unallocated:
@@ -159,42 +155,6 @@ typedef struct BlockCopyState {
     RateLimit rate_limit;
 } BlockCopyState;
 
-/* Called with lock held */
-static BlockCopyTask *find_conflicting_task(BlockCopyState *s,
-                                            int64_t offset, int64_t bytes)
-{
-    BlockCopyTask *t;
-
-    QLIST_FOREACH(t, &s->tasks, list) {
-        if (offset + bytes > t->offset && offset < t->offset + t->bytes) {
-            return t;
-        }
-    }
-
-    return NULL;
-}
-
-/*
- * If there are no intersecting tasks return false. Otherwise, wait for the
- * first found intersecting tasks to finish and return true.
- *
- * Called with lock held. May temporary release the lock.
- * Return value of 0 proves that lock was NOT released.
- */
-static bool coroutine_fn block_copy_wait_one(BlockCopyState *s, int64_t offset,
-                                             int64_t bytes)
-{
-    BlockCopyTask *task = find_conflicting_task(s, offset, bytes);
-
-    if (!task) {
-        return false;
-    }
-
-    qemu_co_queue_wait(&task->wait_queue, &s->lock);
-
-    return true;
-}
-
 /* Called with lock held */
 static int64_t block_copy_chunk_size(BlockCopyState *s)
 {
@@ -238,7 +198,7 @@ block_copy_task_create(BlockCopyState *s, BlockCopyCallState *call_state,
     bytes = QEMU_ALIGN_UP(bytes, s->cluster_size);
 
     /* region is dirty, so no existent tasks possible in it */
-    assert(!find_conflicting_task(s, offset, bytes));
+    assert(!reqlist_find_conflict(&s->reqs, offset, bytes));
 
     bdrv_reset_dirty_bitmap(s->copy_bitmap, offset, bytes);
     s->in_flight_bytes += bytes;
@@ -248,12 +208,9 @@ block_copy_task_create(BlockCopyState *s, BlockCopyCallState *call_state,
         .task.func = block_copy_task_entry,
         .s = s,
         .call_state = call_state,
-        .offset = offset,
-        .bytes = bytes,
         .method = s->method,
     };
-    qemu_co_queue_init(&task->wait_queue);
-    QLIST_INSERT_HEAD(&s->tasks, task, list);
+    reqlist_init_req(&s->reqs, &task->req, offset, bytes);
 
     return task;
 }
@@ -269,34 +226,34 @@ static void coroutine_fn block_copy_task_shrink(BlockCopyTask *task,
                                                 int64_t new_bytes)
 {
     QEMU_LOCK_GUARD(&task->s->lock);
-    if (new_bytes == task->bytes) {
+    if (new_bytes == task->req.bytes) {
         return;
     }
 
-    assert(new_bytes > 0 && new_bytes < task->bytes);
+    assert(new_bytes > 0 && new_bytes < task->req.bytes);
 
-    task->s->in_flight_bytes -= task->bytes - new_bytes;
+    task->s->in_flight_bytes -= task->req.bytes - new_bytes;
     bdrv_set_dirty_bitmap(task->s->copy_bitmap,
-                          task->offset + new_bytes, task->bytes - new_bytes);
+                          task->req.offset + new_bytes,
+                          task->req.bytes - new_bytes);
 
-    task->bytes = new_bytes;
-    qemu_co_queue_restart_all(&task->wait_queue);
+    reqlist_shrink_req(&task->req, new_bytes);
 }
 
 static void coroutine_fn block_copy_task_end(BlockCopyTask *task, int ret)
 {
     QEMU_LOCK_GUARD(&task->s->lock);
-    task->s->in_flight_bytes -= task->bytes;
+    task->s->in_flight_bytes -= task->req.bytes;
     if (ret < 0) {
-        bdrv_set_dirty_bitmap(task->s->copy_bitmap, task->offset, task->bytes);
+        bdrv_set_dirty_bitmap(task->s->copy_bitmap, task->req.offset,
+                              task->req.bytes);
     }
-    QLIST_REMOVE(task, list);
     if (task->s->progress) {
         progress_set_remaining(task->s->progress,
                                bdrv_get_dirty_count(task->s->copy_bitmap) +
                                task->s->in_flight_bytes);
     }
-    qemu_co_queue_restart_all(&task->wait_queue);
+    reqlist_remove_req(&task->req);
 }
 
 void block_copy_state_free(BlockCopyState *s)
@@ -447,7 +404,7 @@ BlockCopyState *block_copy_state_new(BdrvChild *source, BdrvChild *target,
 
     ratelimit_init(&s->rate_limit);
     qemu_co_mutex_init(&s->lock);
-    QLIST_INIT(&s->tasks);
+    QLIST_INIT(&s->reqs);
     QLIST_INIT(&s->calls);
 
     return s;
@@ -480,7 +437,7 @@ static coroutine_fn int block_copy_task_run(AioTaskPool *pool,
 
     aio_task_pool_wait_slot(pool);
     if (aio_task_pool_status(pool) < 0) {
-        co_put_to_shres(task->s->mem, task->bytes);
+        co_put_to_shres(task->s->mem, task->req.bytes);
         block_copy_task_end(task, -ECANCELED);
         g_free(task);
         return -ECANCELED;
@@ -593,7 +550,8 @@ static coroutine_fn int block_copy_task_entry(AioTask *task)
     BlockCopyMethod method = t->method;
     int ret;
 
-    ret = block_copy_do_copy(s, t->offset, t->bytes, &method, &error_is_read);
+    ret = block_copy_do_copy(s, t->req.offset, t->req.bytes, &method,
+                             &error_is_read);
 
     WITH_QEMU_LOCK_GUARD(&s->lock) {
         if (s->method == t->method) {
@@ -606,10 +564,10 @@ static coroutine_fn int block_copy_task_entry(AioTask *task)
                 t->call_state->error_is_read = error_is_read;
             }
         } else if (s->progress) {
-            progress_work_done(s->progress, t->bytes);
+            progress_work_done(s->progress, t->req.bytes);
         }
     }
-    co_put_to_shres(s->mem, t->bytes);
+    co_put_to_shres(s->mem, t->req.bytes);
     block_copy_task_end(t, ret);
 
     return ret;
@@ -768,22 +726,22 @@ block_copy_dirty_clusters(BlockCopyCallState *call_state)
             trace_block_copy_skip_range(s, offset, bytes);
             break;
         }
-        if (task->offset > offset) {
-            trace_block_copy_skip_range(s, offset, task->offset - offset);
+        if (task->req.offset > offset) {
+            trace_block_copy_skip_range(s, offset, task->req.offset - offset);
         }
 
         found_dirty = true;
 
-        ret = block_copy_block_status(s, task->offset, task->bytes,
+        ret = block_copy_block_status(s, task->req.offset, task->req.bytes,
                                       &status_bytes);
         assert(ret >= 0); /* never fail */
-        if (status_bytes < task->bytes) {
+        if (status_bytes < task->req.bytes) {
             block_copy_task_shrink(task, status_bytes);
         }
         if (qatomic_read(&s->skip_unallocated) &&
             !(ret & BDRV_BLOCK_ALLOCATED)) {
             block_copy_task_end(task, 0);
-            trace_block_copy_skip_range(s, task->offset, task->bytes);
+            trace_block_copy_skip_range(s, task->req.offset, task->req.bytes);
             offset = task_end(task);
             bytes = end - offset;
             g_free(task);
@@ -804,11 +762,11 @@ block_copy_dirty_clusters(BlockCopyCallState *call_state)
             }
         }
 
-        ratelimit_calculate_delay(&s->rate_limit, task->bytes);
+        ratelimit_calculate_delay(&s->rate_limit, task->req.bytes);
 
-        trace_block_copy_process(s, task->offset);
+        trace_block_copy_process(s, task->req.offset);
 
-        co_get_from_shres(s->mem, task->bytes);
+        co_get_from_shres(s->mem, task->req.bytes);
 
         offset = task_end(task);
         bytes = end - offset;
@@ -876,8 +834,8 @@ static int coroutine_fn block_copy_common(BlockCopyCallState *call_state)
                  * Check that there is no task we still need to
                  * wait to complete
                  */
-                ret = block_copy_wait_one(s, call_state->offset,
-                                          call_state->bytes);
+                ret = reqlist_wait_one(&s->reqs, call_state->offset,
+                                       call_state->bytes, &s->lock);
                 if (ret == 0) {
                     /*
                      * No pending tasks, but check again the bitmap in this
@@ -885,7 +843,7 @@ static int coroutine_fn block_copy_common(BlockCopyCallState *call_state)
                      * between this and the critical section in
                      * block_copy_dirty_clusters().
                      *
-                     * block_copy_wait_one return value 0 also means that it
+                     * reqlist_wait_one return value 0 also means that it
                      * didn't release the lock. So, we are still in the same
                      * critical section, not interrupted by any concurrent
                      * access to state.
diff --git a/block/reqlist.c b/block/reqlist.c
new file mode 100644
index 0000000000..c41415c16a
--- /dev/null
+++ b/block/reqlist.c
@@ -0,0 +1,94 @@
+/*
+ * block_copy API
+ *
+ * Copyright (C) 2013 Proxmox Server Solutions
+ * Copyright (c) 2021 Virtuozzo International GmbH.
+ *
+ * Authors:
+ *  Dietmar Maurer (dietmar@proxmox.com)
+ *  Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ */
+
+#include "qemu/osdep.h"
+
+#include "block/reqlist.h"
+
+/*
+ * Initialize new request and add it to the list. Caller should be sure that
+ * there are no conflicting requests in the list.
+ */
+void reqlist_init_req(BlockReqList *reqs, BlockReq *req, int64_t offset,
+                      int64_t bytes)
+{
+    assert(!reqlist_find_conflict(reqs, offset, bytes));
+
+    *req = (BlockReq) {
+        .offset = offset,
+        .bytes = bytes,
+    };
+    qemu_co_queue_init(&req->wait_queue);
+    QLIST_INSERT_HEAD(reqs, req, list);
+}
+
+BlockReq *reqlist_find_conflict(BlockReqList *reqs, int64_t offset,
+                                int64_t bytes)
+{
+    BlockReq *r;
+
+    QLIST_FOREACH(r, reqs, list) {
+        if (offset + bytes > r->offset && offset < r->offset + r->bytes) {
+            return r;
+        }
+    }
+
+    return NULL;
+}
+
+/*
+ * If there are no intersecting requests return false. Otherwise, wait for the
+ * first found intersecting request to finish and return true.
+ *
+ * @lock is passed to qemu_co_queue_wait()
+ * False return value proves that lock was NOT released.
+ */
+bool coroutine_fn reqlist_wait_one(BlockReqList *reqs, int64_t offset,
+                                   int64_t bytes, CoMutex *lock)
+{
+    BlockReq *r = reqlist_find_conflict(reqs, offset, bytes);
+
+    if (!r) {
+        return false;
+    }
+
+    qemu_co_queue_wait(&r->wait_queue, lock);
+
+    return true;
+}
+
+/*
+ * Shrink request and wake all waiting coroutines (may be some of them are not
+ * intersecting with shrunk request).
+ */
+void coroutine_fn reqlist_shrink_req(BlockReq *req, int64_t new_bytes)
+{
+    if (new_bytes == req->bytes) {
+        return;
+    }
+
+    assert(new_bytes > 0 && new_bytes < req->bytes);
+
+    req->bytes = new_bytes;
+    qemu_co_queue_restart_all(&req->wait_queue);
+}
+
+/*
+ * Remove request and wake all waiting coroutines. Do not release any memory.
+ */
+void coroutine_fn reqlist_remove_req(BlockReq *req)
+{
+    QLIST_REMOVE(req, list);
+    qemu_co_queue_restart_all(&req->wait_queue);
+}
diff --git a/block/meson.build b/block/meson.build
index 66ee11e62c..c630e9374a 100644
--- a/block/meson.build
+++ b/block/meson.build
@@ -32,6 +32,7 @@ block_ss.add(files(
   'qcow2.c',
   'quorum.c',
   'raw-format.c',
+  'reqlist.c',
   'snapshot.c',
   'throttle-groups.c',
   'throttle.c',
-- 
2.29.2