Merge branch 'io_schedule'

Currently ZFS doesn't show any I/O time in eg "top" wait% or in
/proc/$pid/stat's blkio_ticks.  Using io_schedule() instead of
schedule() in zio_wait()'s cv_wait() is the correct way to fix
this.

Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #1158
Closes #1175

diff --git a/include/sys/zfs_context.h b/include/sys/zfs_context.h
index ad282c43c4866f7cb0bb526b3da56837605da234..6b00a5d5db8051ad2bba436f1705280ead7822bb 100644 (file)
--- a/include/sys/zfs_context.h
+++ b/include/sys/zfs_context.h
@@ -322,6 +322,7 @@ extern void cv_signal(kcondvar_t *cv);
 extern void cv_broadcast(kcondvar_t *cv);
 #define cv_timedwait_interruptible(cv, mp, at) cv_timedwait(cv, mp, at)
 #define cv_wait_interruptible(cv, mp)          cv_wait(cv, mp)
+#define cv_wait_io(cv, mp)                     cv_wait(cv, mp)
 
 /*
  * kstat creation, installation and deletion

diff --git a/module/zfs/zio.c b/module/zfs/zio.c
index bcfc081d1a01832ade32c7da8f5e61a738d897af..bfb817b7860b0d0117a04a1a676725dded3ef7d8 100644 (file)
--- a/module/zfs/zio.c
+++ b/module/zfs/zio.c
@@ -1305,34 +1305,18 @@ __zio_execute(zio_t *zio)
 int
 zio_wait(zio_t *zio)
 {
-       uint64_t timeout;
        int error;
 
        ASSERT(zio->io_stage == ZIO_STAGE_OPEN);
        ASSERT(zio->io_executor == NULL);
 
        zio->io_waiter = curthread;
-       timeout = ddi_get_lbolt() + (zio_delay_max / MILLISEC * hz);
 
        __zio_execute(zio);
 
        mutex_enter(&zio->io_lock);
-       while (zio->io_executor != NULL) {
-               /*
-                * Wake up periodically to prevent the kernel from complaining
-                * about a blocked task.  However, check zio_delay_max to see
-                * if the I/O has exceeded the timeout and post an ereport.
-                */
-               cv_timedwait_interruptible(&zio->io_cv, &zio->io_lock,
-                   ddi_get_lbolt() + hz);
-
-               if (timeout && (ddi_get_lbolt() > timeout)) {
-                       zio->io_delay = zio_delay_max;
-                       zfs_ereport_post(FM_EREPORT_ZFS_DELAY,
-                           zio->io_spa, zio->io_vd, zio, 0, 0);
-                       timeout = 0;
-               }
-       }
+       while (zio->io_executor != NULL)
+               cv_wait_io(&zio->io_cv, &zio->io_lock);
        mutex_exit(&zio->io_lock);
 
        error = zio->io_error;
@@ -2905,11 +2889,15 @@ zio_done(zio_t *zio)
        vdev_stat_update(zio, zio->io_size);
 
        /*
-        * When an I/O completes but was slow post an ereport.
+        * If this I/O is attached to a particular vdev is slow, exeeding
+        * 30 seconds to complete, post an error described the I/O delay.
+        * We ignore these errors if the device is currently unavailable.
         */
-       if (zio->io_delay >= zio_delay_max)
-               zfs_ereport_post(FM_EREPORT_ZFS_DELAY, zio->io_spa,
-                   zio->io_vd, zio, 0, 0);
+       if (zio->io_delay >= zio_delay_max) {
+               if (zio->io_vd != NULL && !vdev_is_dead(zio->io_vd))
+                       zfs_ereport_post(FM_EREPORT_ZFS_DELAY, zio->io_spa,
+                                         zio->io_vd, zio, 0, 0);
+       }
 
        if (zio->io_error) {
                /*