Merge remote-tracking branch 'remotes/kevin/tags/for-upstream' into s…

…taging Block layer patches for 2.11.0-rc3 # gpg: Signature made Wed 29 Nov 2017 15:25:13 GMT # gpg: using RSA key 0x7F09B272C88F2FD6 # gpg: Good signature from "Kevin Wolf <[email protected]>" # Primary key fingerprint: DC3D EB15 9A9A F95D 3D74 56FE 7F09 B272 C88F 2FD6 * remotes/kevin/tags/for-upstream: block/nfs: fix nfs_client_open for filesize greater than 1TB blockjob: reimplement block_job_sleep_ns to allow cancellation blockjob: introduce block_job_do_yield blockjob: remove clock argument from block_job_sleep_ns block: Expect graph changes in bdrv_parent_drained_begin/end blockjob: Remove the job from the list earlier in block_job_unref() QAPI & interop: Clarify events emitted by 'block-job-cancel' qemu-options: Mention locking option of file driver docs: Add image locking subsection iotests: fix 075 and 078 Signed-off-by: Peter Maydell <[email protected]>
hibari0614 · Nov 29, 2017 · 915308b · 915308b
2 parents 844496f + 5591c00
commit 915308b
Show file tree

Hide file tree

Showing 17 changed files with 187 additions and 72 deletions.
diff --git a/block/backup.c b/block/backup.c
@@ -346,9 +346,9 @@ static bool coroutine_fn yield_and_check(BackupBlockJob *job)
         uint64_t delay_ns = ratelimit_calculate_delay(&job->limit,
                                                       job->bytes_read);
         job->bytes_read = 0;
-        block_job_sleep_ns(&job->common, QEMU_CLOCK_REALTIME, delay_ns);
+        block_job_sleep_ns(&job->common, delay_ns);
     } else {
-        block_job_sleep_ns(&job->common, QEMU_CLOCK_REALTIME, 0);
+        block_job_sleep_ns(&job->common, 0);
     }
 
     if (block_job_is_cancelled(&job->common)) {

diff --git a/block/commit.c b/block/commit.c
@@ -174,7 +174,7 @@ static void coroutine_fn commit_run(void *opaque)
         /* 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, QEMU_CLOCK_REALTIME, delay_ns);
+        block_job_sleep_ns(&s->common, delay_ns);
         if (block_job_is_cancelled(&s->common)) {
             break;
         }

diff --git a/block/io.c b/block/io.c
@@ -42,9 +42,9 @@ static int coroutine_fn bdrv_co_do_pwrite_zeroes(BlockDriverState *bs,
 
 void bdrv_parent_drained_begin(BlockDriverState *bs)
 {
-    BdrvChild *c;
+    BdrvChild *c, *next;
 
-    QLIST_FOREACH(c, &bs->parents, next_parent) {
+    QLIST_FOREACH_SAFE(c, &bs->parents, next_parent, next) {
         if (c->role->drained_begin) {
             c->role->drained_begin(c);
         }
@@ -53,9 +53,9 @@ void bdrv_parent_drained_begin(BlockDriverState *bs)
 
 void bdrv_parent_drained_end(BlockDriverState *bs)
 {
-    BdrvChild *c;
+    BdrvChild *c, *next;
 
-    QLIST_FOREACH(c, &bs->parents, next_parent) {
+    QLIST_FOREACH_SAFE(c, &bs->parents, next_parent, next) {
         if (c->role->drained_end) {
             c->role->drained_end(c);
         }

diff --git a/block/mirror.c b/block/mirror.c
@@ -598,7 +598,7 @@ static void mirror_throttle(MirrorBlockJob *s)
 
     if (now - s->last_pause_ns > SLICE_TIME) {
         s->last_pause_ns = now;
-        block_job_sleep_ns(&s->common, QEMU_CLOCK_REALTIME, 0);
+        block_job_sleep_ns(&s->common, 0);
     } else {
         block_job_pause_point(&s->common);
     }
@@ -870,13 +870,13 @@ static void coroutine_fn mirror_run(void *opaque)
         ret = 0;
         trace_mirror_before_sleep(s, cnt, s->synced, delay_ns);
         if (!s->synced) {
-            block_job_sleep_ns(&s->common, QEMU_CLOCK_REALTIME, delay_ns);
+            block_job_sleep_ns(&s->common, delay_ns);
             if (block_job_is_cancelled(&s->common)) {
                 break;
             }
         } else if (!should_complete) {
             delay_ns = (s->in_flight == 0 && cnt == 0 ? SLICE_TIME : 0);
-            block_job_sleep_ns(&s->common, QEMU_CLOCK_REALTIME, delay_ns);
+            block_job_sleep_ns(&s->common, delay_ns);
         }
         s->last_pause_ns = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
     }

diff --git a/block/nfs.c b/block/nfs.c
@@ -1,7 +1,7 @@
 /*
  * QEMU Block driver for native access to files on NFS shares
  *
- * Copyright (c) 2014-2016 Peter Lieven <[email protected]>
+ * Copyright (c) 2014-2017 Peter Lieven <[email protected]>
  *
  * Permission is hereby granted, free of charge, to any person obtaining a copy
  * of this software and associated documentation files (the "Software"), to deal
@@ -496,7 +496,7 @@ static NFSServer *nfs_config(QDict *options, Error **errp)
 static int64_t nfs_client_open(NFSClient *client, QDict *options,
                                int flags, int open_flags, Error **errp)
 {
-    int ret = -EINVAL;
+    int64_t ret = -EINVAL;
     QemuOpts *opts = NULL;
     Error *local_err = NULL;
     struct stat st;
@@ -686,8 +686,7 @@ static QemuOptsList nfs_create_opts = {
 
 static int nfs_file_create(const char *url, QemuOpts *opts, Error **errp)
 {
-    int ret = 0;
-    int64_t total_size = 0;
+    int64_t ret, total_size;
     NFSClient *client = g_new0(NFSClient, 1);
     QDict *options = NULL;
 

diff --git a/block/stream.c b/block/stream.c
@@ -141,7 +141,7 @@ static void coroutine_fn stream_run(void *opaque)
         /* 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, QEMU_CLOCK_REALTIME, delay_ns);
+        block_job_sleep_ns(&s->common, delay_ns);
         if (block_job_is_cancelled(&s->common)) {
             break;
         }

diff --git a/blockjob.c b/blockjob.c
@@ -37,6 +37,26 @@
 #include "qemu/timer.h"
 #include "qapi-event.h"
 
+/* Right now, this mutex is only needed to synchronize accesses to job->busy
+ * and job->sleep_timer, such as concurrent calls to block_job_do_yield and
+ * block_job_enter. */
+static QemuMutex block_job_mutex;
+
+static void block_job_lock(void)
+{
+    qemu_mutex_lock(&block_job_mutex);
+}
+
+static void block_job_unlock(void)
+{
+    qemu_mutex_unlock(&block_job_mutex);
+}
+
+static void __attribute__((__constructor__)) block_job_init(void)
+{
+    qemu_mutex_init(&block_job_mutex);
+}
+
 static void block_job_event_cancelled(BlockJob *job);
 static void block_job_event_completed(BlockJob *job, const char *msg);
 
@@ -152,6 +172,7 @@ void block_job_unref(BlockJob *job)
 {
     if (--job->refcnt == 0) {
         BlockDriverState *bs = blk_bs(job->blk);
+        QLIST_REMOVE(job, job_list);
         bs->job = NULL;
         block_job_remove_all_bdrv(job);
         blk_remove_aio_context_notifier(job->blk,
@@ -160,7 +181,7 @@ void block_job_unref(BlockJob *job)
         blk_unref(job->blk);
         error_free(job->blocker);
         g_free(job->id);
-        QLIST_REMOVE(job, job_list);
+        assert(!timer_pending(&job->sleep_timer));
         g_free(job);
     }
 }
@@ -287,6 +308,13 @@ static void coroutine_fn block_job_co_entry(void *opaque)
     job->driver->start(job);
 }
 
+static void block_job_sleep_timer_cb(void *opaque)
+{
+    BlockJob *job = opaque;
+
+    block_job_enter(job);
+}
+
 void block_job_start(BlockJob *job)
 {
     assert(job && !block_job_started(job) && job->paused &&
@@ -556,7 +584,7 @@ BlockJobInfo *block_job_query(BlockJob *job, Error **errp)
     info->type      = g_strdup(BlockJobType_str(job->driver->job_type));
     info->device    = g_strdup(job->id);
     info->len       = job->len;
-    info->busy      = job->busy;
+    info->busy      = atomic_read(&job->busy);
     info->paused    = job->pause_count > 0;
     info->offset    = job->offset;
     info->speed     = job->speed;
@@ -664,6 +692,9 @@ void *block_job_create(const char *job_id, const BlockJobDriver *driver,
     job->paused        = true;
     job->pause_count   = 1;
     job->refcnt        = 1;
+    aio_timer_init(qemu_get_aio_context(), &job->sleep_timer,
+                   QEMU_CLOCK_REALTIME, SCALE_NS,
+                   block_job_sleep_timer_cb, job);
 
     error_setg(&job->blocker, "block device is in use by block job: %s",
                BlockJobType_str(driver->job_type));
@@ -729,6 +760,26 @@ static bool block_job_should_pause(BlockJob *job)
     return job->pause_count > 0;
 }
 
+/* Yield, and schedule a timer to reenter the coroutine after @ns nanoseconds.
+ * Reentering the job coroutine with block_job_enter() before the timer has
+ * expired is allowed and cancels the timer.
+ *
+ * If @ns is (uint64_t) -1, no timer is scheduled and block_job_enter() must be
+ * called explicitly. */
+static void block_job_do_yield(BlockJob *job, uint64_t ns)
+{
+    block_job_lock();
+    if (ns != -1) {
+        timer_mod(&job->sleep_timer, ns);
+    }
+    job->busy = false;
+    block_job_unlock();
+    qemu_coroutine_yield();
+
+    /* Set by block_job_enter before re-entering the coroutine.  */
+    assert(job->busy);
+}
+
 void coroutine_fn block_job_pause_point(BlockJob *job)
 {
     assert(job && block_job_started(job));
@@ -746,9 +797,7 @@ void coroutine_fn block_job_pause_point(BlockJob *job)
 
     if (block_job_should_pause(job) && !block_job_is_cancelled(job)) {
         job->paused = true;
-        job->busy = false;
-        qemu_coroutine_yield(); /* wait for block_job_resume() */
-        job->busy = true;
+        block_job_do_yield(job, -1);
         job->paused = false;
     }
 
@@ -778,17 +827,25 @@ void block_job_enter(BlockJob *job)
         return;
     }
 
-    if (!job->busy) {
-        bdrv_coroutine_enter(blk_bs(job->blk), job->co);
+    block_job_lock();
+    if (job->busy) {
+        block_job_unlock();
+        return;
     }
+
+    assert(!job->deferred_to_main_loop);
+    timer_del(&job->sleep_timer);
+    job->busy = true;
+    block_job_unlock();
+    aio_co_wake(job->co);
 }
 
 bool block_job_is_cancelled(BlockJob *job)
 {
     return job->cancelled;
 }
 
-void block_job_sleep_ns(BlockJob *job, QEMUClockType type, int64_t ns)
+void block_job_sleep_ns(BlockJob *job, int64_t ns)
 {
     assert(job->busy);
 
@@ -797,13 +854,8 @@ void block_job_sleep_ns(BlockJob *job, QEMUClockType type, int64_t ns)
         return;
     }
 
-    /* We need to leave job->busy set here, because when we have
-     * put a coroutine to 'sleep', we have scheduled it to run in
-     * the future.  We cannot enter that same coroutine again before
-     * it wakes and runs, otherwise we risk double-entry or entry after
-     * completion. */
     if (!block_job_should_pause(job)) {
-        co_aio_sleep_ns(blk_get_aio_context(job->blk), type, ns);
+        block_job_do_yield(job, qemu_clock_get_ns(QEMU_CLOCK_REALTIME) + ns);
     }
 
     block_job_pause_point(job);
@@ -818,11 +870,9 @@ void block_job_yield(BlockJob *job)
         return;
     }
 
-    job->busy = false;
     if (!block_job_should_pause(job)) {
-        qemu_coroutine_yield();
+        block_job_do_yield(job, -1);
     }
-    job->busy = true;
 
     block_job_pause_point(job);
 }

diff --git a/docs/interop/live-block-operations.rst b/docs/interop/live-block-operations.rst
@@ -506,26 +506,40 @@ Again, given our familiar disk image chain::
 
     [A] <-- [B] <-- [C] <-- [D]
 
-The ``drive-mirror`` (and its newer equivalent ``blockdev-mirror``) allows
-you to copy data from the entire chain into a single target image (which
-can be located on a different host).
-
-Once a 'mirror' job has started, there are two possible actions while a
-``drive-mirror`` job is active:
-
-(1) Issuing the command ``block-job-cancel`` after it emits the event
-    ``BLOCK_JOB_CANCELLED``: will (after completing synchronization of
-    the content from the disk image chain to the target image, [E])
-    create a point-in-time (which is at the time of *triggering* the
-    cancel command) copy, contained in image [E], of the the entire disk
+The ``drive-mirror`` (and its newer equivalent ``blockdev-mirror``)
+allows you to copy data from the entire chain into a single target image
+(which can be located on a different host), [E].
+
+.. note::
+
+    When you cancel an in-progress 'mirror' job *before* the source and
+    target are synchronized, ``block-job-cancel`` will emit the event
+    ``BLOCK_JOB_CANCELLED``.  However, note that if you cancel a
+    'mirror' job *after* it has indicated (via the event
+    ``BLOCK_JOB_READY``) that the source and target have reached
+    synchronization, then the event emitted by ``block-job-cancel``
+    changes to ``BLOCK_JOB_COMPLETED``.
+
+    Besides the 'mirror' job, the "active ``block-commit``" is the only
+    other block device job that emits the event ``BLOCK_JOB_READY``.
+    The rest of the block device jobs ('stream', "non-active
+    ``block-commit``", and 'backup') end automatically.
+
+So there are two possible actions to take, after a 'mirror' job has
+emitted the event ``BLOCK_JOB_READY``, indicating that the source and
+target have reached synchronization:
+
+(1) Issuing the command ``block-job-cancel`` (after it emits the event
+    ``BLOCK_JOB_COMPLETED``) will create a point-in-time (which is at
+    the time of *triggering* the cancel command) copy of the entire disk
     image chain (or only the top-most image, depending on the ``sync``
-    mode).
+    mode), contained in the target image [E]. One use case for this is
+    live VM migration with non-shared storage.
 
-(2) Issuing the command ``block-job-complete`` after it emits the event
-    ``BLOCK_JOB_COMPLETED``: will, after completing synchronization of
-    the content, adjust the guest device (i.e. live QEMU) to point to
-    the target image, and, causing all the new writes from this point on
-    to happen there.  One use case for this is live storage migration.
+(2) Issuing the command ``block-job-complete`` (after it emits the event
+    ``BLOCK_JOB_COMPLETED``) will adjust the guest device (i.e. live
+    QEMU) to point to the target image, [E], causing all the new writes
+    from this point on to happen there.
 
 About synchronization modes: The synchronization mode determines
 *which* part of the disk image chain will be copied to the target.

diff --git a/docs/qemu-block-drivers.texi b/docs/qemu-block-drivers.texi
@@ -785,6 +785,42 @@ warning: ssh server @code{ssh.example.com:22} does not support fsync
 With sufficiently new versions of libssh2 and OpenSSH, @code{fsync} is
 supported.
 
+@node disk_image_locking
+@subsection Disk image file locking
+
+By default, QEMU tries to protect image files from unexpected concurrent
+access, as long as it's supported by the block protocol driver and host
+operating system. If multiple QEMU processes (including QEMU emulators and
+utilities) try to open the same image with conflicting accessing modes, all but
+the first one will get an error.
+
+This feature is currently supported by the file protocol on Linux with the Open
+File Descriptor (OFD) locking API, and can be configured to fall back to POSIX
+locking if the POSIX host doesn't support Linux OFD locking.
+
+To explicitly enable image locking, specify "locking=on" in the file protocol
+driver options. If OFD locking is not possible, a warning will be printed and
+the POSIX locking API will be used. In this case there is a risk that the lock
+will get silently lost when doing hot plugging and block jobs, due to the
+shortcomings of the POSIX locking API.
+
+QEMU transparently handles lock handover during shared storage migration.  For
+shared virtual disk images between multiple VMs, the "share-rw" device option
+should be used.
+
+Alternatively, locking can be fully disabled by "locking=off" block device
+option. In the command line, the option is usually in the form of
+"file.locking=off" as the protocol driver is normally placed as a "file" child
+under a format driver. For example:
+
+@code{-blockdev driver=qcow2,file.filename=/path/to/image,file.locking=off,file.driver=file}
+
+To check if image locking is active, check the output of the "lslocks" command
+on host and see if there are locks held by the QEMU process on the image file.
+More than one byte could be locked by the QEMU instance, each byte of which
+reflects a particular permission that is acquired or protected by the running
+block driver.
+
 @c man end
 
 @ignore