Merge remote-tracking branch 'remotes/famz/tags/block-and-testing-pul…

…l-request' into staging # gpg: Signature made Mon 17 Jul 2017 04:47:05 BST # gpg: using RSA key 0xCA35624C6A9171C6 # gpg: Good signature from "Fam Zheng <famz@redhat.com>" # gpg: WARNING: This key is not certified with a trusted signature! # gpg: There is no indication that the signature belongs to the owner. # Primary key fingerprint: 5003 7CB7 9706 0F76 F021 AD56 CA35 624C 6A91 71C6 * remotes/famz/tags/block-and-testing-pull-request: travis: add no-TCG build docker.py: Improve subprocess exit code handling docker.py: Drop infile parameter docker: Don't enable networking as a side-effect of DEBUG=1 ssh: support I/O from any AioContext sheepdog: add queue_lock qed: protect table cache with CoMutex qed: introduce bdrv_qed_init_state block: invoke .bdrv_drain callback in coroutine context and from AioContext qed: move tail of qed_aio_write_main to qed_aio_write_{cow, alloc} vvfat: make it thread-safe vpc: make it thread-safe vdi: make it thread-safe coroutine-lock: add qemu_co_rwlock_downgrade and qemu_co_rwlock_upgrade qcow2: call CoQueue APIs under CoMutex Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
qemu · Jul 17, 2017 · 6632f6f · 6632f6f
2 parents acbaa0f + 9783731
commit 6632f6f
Show file tree

Hide file tree

Showing 18 changed files with 355 additions and 161 deletions.
diff --git a/.travis.yml b/.travis.yml
@@ -86,6 +86,9 @@ matrix:
     - env: CONFIG="--enable-trace-backends=ust"
            TEST_CMD=""
       compiler: gcc
+    - env: CONFIG="--disable-tcg"
+           TEST_CMD=""
+      compiler: gcc
     - env: CONFIG=""
       os: osx
       compiler: clang

diff --git a/block/io.c b/block/io.c
@@ -149,16 +149,46 @@ bool bdrv_requests_pending(BlockDriverState *bs)
     return false;
 }
 
+typedef struct {
+    Coroutine *co;
+    BlockDriverState *bs;
+    bool done;
+} BdrvCoDrainData;
+
+static void coroutine_fn bdrv_drain_invoke_entry(void *opaque)
+{
+    BdrvCoDrainData *data = opaque;
+    BlockDriverState *bs = data->bs;
+
+    bs->drv->bdrv_co_drain(bs);
+
+    /* Set data->done before reading bs->wakeup.  */
+    atomic_mb_set(&data->done, true);
+    bdrv_wakeup(bs);
+}
+
+static void bdrv_drain_invoke(BlockDriverState *bs)
+{
+    BdrvCoDrainData data = { .bs = bs, .done = false };
+
+    if (!bs->drv || !bs->drv->bdrv_co_drain) {
+        return;
+    }
+
+    data.co = qemu_coroutine_create(bdrv_drain_invoke_entry, &data);
+    bdrv_coroutine_enter(bs, data.co);
+    BDRV_POLL_WHILE(bs, !data.done);
+}
+
 static bool bdrv_drain_recurse(BlockDriverState *bs)
 {
     BdrvChild *child, *tmp;
     bool waited;
 
     waited = BDRV_POLL_WHILE(bs, atomic_read(&bs->in_flight) > 0);
 
-    if (bs->drv && bs->drv->bdrv_drain) {
-        bs->drv->bdrv_drain(bs);
-    }
+    /* Ensure any pending metadata writes are submitted to bs->file.  */
+    bdrv_drain_invoke(bs);
 
     QLIST_FOREACH_SAFE(child, &bs->children, next, tmp) {
         BlockDriverState *bs = child->bs;
@@ -184,12 +214,6 @@ static bool bdrv_drain_recurse(BlockDriverState *bs)
     return waited;
 }
 
-typedef struct {
-    Coroutine *co;
-    BlockDriverState *bs;
-    bool done;
-} BdrvCoDrainData;
-
 static void bdrv_co_drain_bh_cb(void *opaque)
 {
     BdrvCoDrainData *data = opaque;

diff --git a/block/qcow2.c b/block/qcow2.c
@@ -2025,8 +2025,6 @@ static coroutine_fn int qcow2_co_pwritev(BlockDriverState *bs, uint64_t offset,
     ret = 0;
 
 fail:
-    qemu_co_mutex_unlock(&s->lock);
-
     while (l2meta != NULL) {
         QCowL2Meta *next;
 
@@ -2040,6 +2038,8 @@ static coroutine_fn int qcow2_co_pwritev(BlockDriverState *bs, uint64_t offset,
         l2meta = next;
     }
 
+    qemu_co_mutex_unlock(&s->lock);
+
     qemu_iovec_destroy(&hd_qiov);
     qemu_vfree(cluster_data);
     trace_qcow2_writev_done_req(qemu_coroutine_self(), ret);

diff --git a/block/qed-cluster.c b/block/qed-cluster.c
@@ -85,6 +85,8 @@ static unsigned int qed_count_contiguous_clusters(BDRVQEDState *s,
  *
  * On failure QED_CLUSTER_L2 or QED_CLUSTER_L1 is returned for missing L2 or L1
  * table offset, respectively. len is number of contiguous unallocated bytes.
+ *
+ * Called with table_lock held.
  */
 int coroutine_fn qed_find_cluster(BDRVQEDState *s, QEDRequest *request,
                                   uint64_t pos, size_t *len,
@@ -112,7 +114,6 @@ int coroutine_fn qed_find_cluster(BDRVQEDState *s, QEDRequest *request,
     }
 
     ret = qed_read_l2_table(s, request, l2_offset);
-    qed_acquire(s);
     if (ret) {
         goto out;
     }
@@ -137,6 +138,5 @@ int coroutine_fn qed_find_cluster(BDRVQEDState *s, QEDRequest *request,
 
 out:
     *img_offset = offset;
-    qed_release(s);
     return ret;
 }
diff --git a/block/qed-l2-cache.c b/block/qed-l2-cache.c
@@ -101,6 +101,8 @@ CachedL2Table *qed_alloc_l2_cache_entry(L2TableCache *l2_cache)
 /**
  * Decrease an entry's reference count and free if necessary when the reference
  * count drops to zero.
+ *
+ * Called with table_lock held.
  */
 void qed_unref_l2_cache_entry(CachedL2Table *entry)
 {
@@ -122,6 +124,8 @@ void qed_unref_l2_cache_entry(CachedL2Table *entry)
  *
  * For a cached entry, this function increases the reference count and returns
  * the entry.
+ *
+ * Called with table_lock held.
  */
 CachedL2Table *qed_find_l2_cache_entry(L2TableCache *l2_cache, uint64_t offset)
 {
@@ -150,6 +154,8 @@ CachedL2Table *qed_find_l2_cache_entry(L2TableCache *l2_cache, uint64_t offset)
  * N.B. This function steals a reference to the l2_table from the caller so the
  * caller must obtain a new reference by issuing a call to
  * qed_find_l2_cache_entry().
+ *
+ * Called with table_lock held.
  */
 void qed_commit_l2_cache_entry(L2TableCache *l2_cache, CachedL2Table *l2_table)
 {

diff --git a/block/qed-table.c b/block/qed-table.c
@@ -18,6 +18,7 @@
 #include "qed.h"
 #include "qemu/bswap.h"
 
+/* Called either from qed_check or with table_lock held.  */
 static int qed_read_table(BDRVQEDState *s, uint64_t offset, QEDTable *table)
 {
     QEMUIOVector qiov;
@@ -32,18 +33,22 @@ static int qed_read_table(BDRVQEDState *s, uint64_t offset, QEDTable *table)
 
     trace_qed_read_table(s, offset, table);
 
+    if (qemu_in_coroutine()) {
+        qemu_co_mutex_unlock(&s->table_lock);
+    }
     ret = bdrv_preadv(s->bs->file, offset, &qiov);
+    if (qemu_in_coroutine()) {
+        qemu_co_mutex_lock(&s->table_lock);
+    }
     if (ret < 0) {
         goto out;
     }
 
     /* Byteswap offsets */
-    qed_acquire(s);
     noffsets = qiov.size / sizeof(uint64_t);
     for (i = 0; i < noffsets; i++) {
         table->offsets[i] = le64_to_cpu(table->offsets[i]);
     }
-    qed_release(s);
 
     ret = 0;
 out:
@@ -61,6 +66,8 @@ static int qed_read_table(BDRVQEDState *s, uint64_t offset, QEDTable *table)
  * @index:      Index of first element
  * @n:          Number of elements
  * @flush:      Whether or not to sync to disk
+ *
+ * Called either from qed_check or with table_lock held.
  */
 static int qed_write_table(BDRVQEDState *s, uint64_t offset, QEDTable *table,
                            unsigned int index, unsigned int n, bool flush)
@@ -97,16 +104,20 @@ static int qed_write_table(BDRVQEDState *s, uint64_t offset, QEDTable *table,
     /* Adjust for offset into table */
     offset += start * sizeof(uint64_t);
 
+    if (qemu_in_coroutine()) {
+        qemu_co_mutex_unlock(&s->table_lock);
+    }
     ret = bdrv_pwritev(s->bs->file, offset, &qiov);
+    if (qemu_in_coroutine()) {
+        qemu_co_mutex_lock(&s->table_lock);
+    }
     trace_qed_write_table_cb(s, table, flush, ret);
     if (ret < 0) {
         goto out;
     }
 
     if (flush) {
-        qed_acquire(s);
         ret = bdrv_flush(s->bs);
-        qed_release(s);
         if (ret < 0) {
             goto out;
         }
@@ -123,6 +134,7 @@ int qed_read_l1_table_sync(BDRVQEDState *s)
     return qed_read_table(s, s->header.l1_table_offset, s->l1_table);
 }
 
+/* Called either from qed_check or with table_lock held.  */
 int qed_write_l1_table(BDRVQEDState *s, unsigned int index, unsigned int n)
 {
     BLKDBG_EVENT(s->bs->file, BLKDBG_L1_UPDATE);
@@ -136,6 +148,7 @@ int qed_write_l1_table_sync(BDRVQEDState *s, unsigned int index,
     return qed_write_l1_table(s, index, n);
 }
 
+/* Called either from qed_check or with table_lock held.  */
 int qed_read_l2_table(BDRVQEDState *s, QEDRequest *request, uint64_t offset)
 {
     int ret;
@@ -154,7 +167,6 @@ int qed_read_l2_table(BDRVQEDState *s, QEDRequest *request, uint64_t offset)
     BLKDBG_EVENT(s->bs->file, BLKDBG_L2_LOAD);
     ret = qed_read_table(s, offset, request->l2_table->table);
 
-    qed_acquire(s);
     if (ret) {
         /* can't trust loaded L2 table anymore */
         qed_unref_l2_cache_entry(request->l2_table);
@@ -170,7 +182,6 @@ int qed_read_l2_table(BDRVQEDState *s, QEDRequest *request, uint64_t offset)
         request->l2_table = qed_find_l2_cache_entry(&s->l2_cache, offset);
         assert(request->l2_table != NULL);
     }
-    qed_release(s);
 
     return ret;
 }
@@ -180,6 +191,7 @@ int qed_read_l2_table_sync(BDRVQEDState *s, QEDRequest *request, uint64_t offset
     return qed_read_l2_table(s, request, offset);
 }
 
+/* Called either from qed_check or with table_lock held.  */
 int qed_write_l2_table(BDRVQEDState *s, QEDRequest *request,
                        unsigned int index, unsigned int n, bool flush)
 {