diff mbox

[v7,2/3] linux-aio: handling -EAGAIN for !s->io_q.plugged case

Message ID 1417424656-29714-3-git-send-email-ming.lei@canonical.com
State New
Headers show

Commit Message

Ming Lei Dec. 1, 2014, 9:04 a.m. UTC 
Previously -EAGAIN is simply ignored for !s->io_q.plugged case,
and sometimes it is easy to cause -EIO to VM, such as NVME device.

This patch handles -EAGAIN by io queue for !s->io_q.plugged case,
and it will be retried in following aio completion cb.

Most of times, -EAGAIN only happens if there is pending I/O, but
from linux kernel AIO implementation io_submit() might return it
when kmem_cache_alloc(GFP_KERNEL) returns NULL too. So 'pending'
in 'struct qemu_laio_state' is introduced for tracking active IO,
and -EAGAIN is handled when there is pending I/O.

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Ming Lei <ming.lei@canonical.com>
---
 block/linux-aio.c |   32 ++++++++++++++++++++++++--------
 1 file changed, 24 insertions(+), 8 deletions(-)
diff mbox

Patch

diff --git a/block/linux-aio.c b/block/linux-aio.c
index 53c5616..9403b17 100644
--- a/block/linux-aio.c
+++ b/block/linux-aio.c
@@ -56,6 +56,7 @@  typedef struct {
 } LaioQueue;
 
 struct qemu_laio_state {
+    unsigned long pending;
     io_context_t ctx;
     EventNotifier e;
 
@@ -98,6 +99,7 @@  static void qemu_laio_process_completion(struct qemu_laio_state *s,
             }
         }
     }
+    s->pending--;
     laiocb->common.cb(laiocb->common.opaque, ret);
 
     qemu_aio_unref(laiocb);
@@ -179,6 +181,7 @@  static void laio_cancel(BlockAIOCB *blockacb)
         return;
     }
 
+    laiocb->ctx->pending--;
     laiocb->common.cb(laiocb->common.opaque, laiocb->ret);
 }
 
@@ -280,8 +283,13 @@  static int ioq_enqueue(struct qemu_laio_state *s, struct iocb *iocb)
     s->io_q.iocbs[idx++] = iocb;
     s->io_q.idx = idx;
 
-    /* submit immediately if queue depth is above 2/3 */
-    if (idx > s->io_q.size * 2 / 3) {
+    /*
+     * This is reached in two cases: queue not plugged but io_submit
+     * returned -EAGAIN, or queue plugged.  In the latter case, start
+     * submitting some I/O if the queue is getting too full.  In the
+     * former case, instead, wait until an I/O operation is completed.
+     */
+    if (s->io_q.plugged && unlikely(idx > s->io_q.size * 2 / 3)) {
         ioq_submit(s);
     }
 
@@ -346,15 +354,23 @@  BlockAIOCB *laio_submit(BlockDriverState *bs, void *aio_ctx, int fd,
     }
     io_set_eventfd(&laiocb->iocb, event_notifier_get_fd(&s->e));
 
-    if (!s->io_q.plugged) {
-        if (io_submit(s->ctx, 1, &iocbs) < 0) {
-            goto out_free_aiocb;
-        }
-    } else {
-        if (ioq_enqueue(s, iocbs) < 0) {
+    /* Switch to queue mode until -EAGAIN is handled */
+    if (!s->io_q.plugged && !s->io_q.idx) {
+        int ret = io_submit(s->ctx, 1, &iocbs);
+        if (ret >= 0) {
+            return &laiocb->common;
+        } else if (ret != -EAGAIN || (ret == -EAGAIN && !s->pending)) {
             goto out_free_aiocb;
         }
+        /*
+         * In case of -EAGAIN, only queue the req if there is pending
+         * I/O and it is resubmitted in completion of pending I/O
+         */
+    }
+    if (ioq_enqueue(s, iocbs) < 0) {
+        goto out_free_aiocb;
     }
+    s->pending++;
     return &laiocb->common;
 
 out_free_aiocb: