[Qemu-devel] [PATCH 13/15] qed: intelligent streaming implementation

[Stefan Hajnoczi]

From: Anthony Liguori <address@hidden>

Signed-off-by: Anthony Liguori <address@hidden>
---
 block/qed.c |  248 +++++++++++++++++++++++++++++++++++++++++++++++++++++++----
 block/qed.h |    3 +-
 2 files changed, 234 insertions(+), 17 deletions(-)

diff --git a/block/qed.c b/block/qed.c
index ffdbc2d..f9f7c94 100644
--- a/block/qed.c
+++ b/block/qed.c
@@ -951,9 +951,8 @@ static void qed_aio_write_l1_update(void *opaque, int ret)
 /**
  * Update L2 table with new cluster offsets and write them out
  */
-static void qed_aio_write_l2_update(void *opaque, int ret)
+static void qed_aio_write_l2_update(QEDAIOCB *acb, int ret, uint64_t offset)
 {
-    QEDAIOCB *acb = opaque;
     BDRVQEDState *s = acb_to_s(acb);
     bool need_alloc = acb->find_cluster_ret == QED_CLUSTER_L1;
     int index;
@@ -969,7 +968,7 @@ static void qed_aio_write_l2_update(void *opaque, int ret)
 
     index = qed_l2_index(s, acb->cur_pos);
     qed_update_l2_table(s, acb->request.l2_table->table, index, 
acb->cur_nclusters,
-                         acb->cur_cluster);
+                         offset);
 
     if (need_alloc) {
         /* Write out the whole new L2 table */
@@ -986,6 +985,51 @@ err:
     qed_aio_complete(acb, ret);
 }
 
+static void qed_aio_write_l2_update_cb(void *opaque, int ret)
+{
+    QEDAIOCB *acb = opaque;
+    qed_aio_write_l2_update(acb, ret, acb->cur_cluster);
+}
+
+/**
+ * Determine if we have a zero write to a block of clusters
+ *
+ * We validate that the write is aligned to a cluster boundary, and that it's
+ * a multiple of cluster size with all zeros.
+ */
+static bool qed_is_zero_write(QEDAIOCB *acb)
+{
+    BDRVQEDState *s = acb_to_s(acb);
+    int i;
+
+    if (!qed_offset_is_cluster_aligned(s, acb->cur_pos)) {
+        return false;
+    }
+
+    if (!qed_offset_is_cluster_aligned(s, acb->cur_qiov.size)) {
+        return false;
+    }
+
+    for (i = 0; i < acb->cur_qiov.niov; i++) {
+        struct iovec *iov = &acb->cur_qiov.iov[i];
+        uint64_t *v;
+        int j;
+
+        if ((iov->iov_len & 0x07)) {
+            return false;
+        }
+
+        v = iov->iov_base;
+        for (j = 0; j < iov->iov_len; j += sizeof(v[0])) {
+            if (v[j >> 3]) {
+                return false;
+            }
+        }
+    }
+
+    return true;
+}
+
 /**
  * Flush new data clusters before updating the L2 table
  *
@@ -1000,7 +1044,7 @@ static void qed_aio_write_flush_before_l2_update(void 
*opaque, int ret)
     QEDAIOCB *acb = opaque;
     BDRVQEDState *s = acb_to_s(acb);
 
-    if (!bdrv_aio_flush(s->bs->file, qed_aio_write_l2_update, opaque)) {
+    if (!bdrv_aio_flush(s->bs->file, qed_aio_write_l2_update_cb, opaque)) {
         qed_aio_complete(acb, -EIO);
     }
 }
@@ -1030,7 +1074,7 @@ static void qed_aio_write_main(void *opaque, int ret)
         if (s->bs->backing_hd) {
             next_fn = qed_aio_write_flush_before_l2_update;
         } else {
-            next_fn = qed_aio_write_l2_update;
+            next_fn = qed_aio_write_l2_update_cb;
         }
     }
 
@@ -1096,6 +1140,18 @@ static bool qed_should_set_need_check(BDRVQEDState *s)
     return !(s->header.features & QED_F_NEED_CHECK);
 }
 
+static void qed_aio_write_zero_cluster(void *opaque, int ret)
+{
+    QEDAIOCB *acb = opaque;
+
+    if (ret) {
+        qed_aio_complete(acb, ret);
+        return;
+    }
+
+    qed_aio_write_l2_update(acb, 0, 1);
+}
+
 /**
  * Start an allocating write request or queue it
  *
@@ -1144,6 +1200,7 @@ static bool qed_start_allocating_write(QEDAIOCB *acb)
 static void qed_aio_write_alloc(QEDAIOCB *acb)
 {
     BDRVQEDState *s = acb_to_s(acb);
+    BlockDriverCompletionFunc *cb;
 
     if (!qed_start_allocating_write(acb)) {
         qemu_iovec_reset(&acb->cur_qiov);
@@ -1154,11 +1211,18 @@ static void qed_aio_write_alloc(QEDAIOCB *acb)
             qed_offset_into_cluster(s, acb->cur_pos) + acb->cur_qiov.size);
     acb->cur_cluster = qed_alloc_clusters(s, acb->cur_nclusters);
 
+    cb = qed_aio_write_prefill;
+
+    /* Zero write detection */
+    if ((acb->flags & QED_AIOCB_CHECK_ZERO_WRITE) && qed_is_zero_write(acb)) {
+        cb = qed_aio_write_zero_cluster;
+    }
+
     if (qed_should_set_need_check(s)) {
         s->header.features |= QED_F_NEED_CHECK;
-        qed_write_header(s, qed_aio_write_prefill, acb);
+        qed_write_header(s, cb, acb);
     } else {
-        qed_aio_write_prefill(acb, 0);
+        cb(acb, 0);
     }
 }
 
@@ -1317,11 +1381,11 @@ static void qed_aio_next_io(void *opaque, int ret)
                       io_fn, acb);
 }
 
-static BlockDriverAIOCB *qed_aio_setup(BlockDriverState *bs,
-                                       int64_t sector_num,
-                                       QEMUIOVector *qiov, int nb_sectors,
-                                       BlockDriverCompletionFunc *cb,
-                                       void *opaque, int flags)
+static QEDAIOCB *qed_aio_setup(BlockDriverState *bs,
+                               int64_t sector_num,
+                               QEMUIOVector *qiov, int nb_sectors,
+                               BlockDriverCompletionFunc *cb,
+                               void *opaque, int flags)
 {
     QEDAIOCB *acb = qemu_aio_get(&qed_aio_pool, bs, cb, opaque);
 
@@ -1337,8 +1401,22 @@ static BlockDriverAIOCB *qed_aio_setup(BlockDriverState 
*bs,
     acb->request.l2_table = NULL;
     qemu_iovec_init(&acb->cur_qiov, qiov->niov);
 
+    return acb;
+}
+
+static BlockDriverAIOCB *bdrv_qed_aio_setup(BlockDriverState *bs,
+                                            int64_t sector_num,
+                                            QEMUIOVector *qiov, int nb_sectors,
+                                            BlockDriverCompletionFunc *cb,
+                                            void *opaque, int flags)
+{
+    QEDAIOCB *acb;
+
+    acb = qed_aio_setup(bs, sector_num, qiov, nb_sectors,
+                        cb, opaque, flags);
     /* Start request */
     qed_aio_next_io(acb, 0);
+
     return &acb->common;
 }
 
@@ -1348,9 +1426,15 @@ static BlockDriverAIOCB 
*bdrv_qed_aio_readv(BlockDriverState *bs,
                                             BlockDriverCompletionFunc *cb,
                                             void *opaque)
 {
-    int flags = bs->copy_on_read ? QED_AIOCB_COPY_ON_READ : 0;
+    /* Don't bloat image file in copy-on-read, use zero detection */
+    int flags = QED_AIOCB_CHECK_ZERO_WRITE;
+
+    if (bs->copy_on_read) {
+        flags |= QED_AIOCB_COPY_ON_READ;
+    }
 
-    return qed_aio_setup(bs, sector_num, qiov, nb_sectors, cb, opaque, flags);
+    return bdrv_qed_aio_setup(bs, sector_num, qiov, nb_sectors, cb,
+                              opaque, flags);
 }
 
 static BlockDriverAIOCB *bdrv_qed_aio_writev(BlockDriverState *bs,
@@ -1359,8 +1443,139 @@ static BlockDriverAIOCB 
*bdrv_qed_aio_writev(BlockDriverState *bs,
                                              BlockDriverCompletionFunc *cb,
                                              void *opaque)
 {
-    return qed_aio_setup(bs, sector_num, qiov, nb_sectors, cb,
-                         opaque, QED_AIOCB_WRITE);
+    return bdrv_qed_aio_setup(bs, sector_num, qiov, nb_sectors, cb,
+                              opaque, QED_AIOCB_WRITE);
+}
+
+typedef struct QEDCopyBackingData {
+    QEDAIOCB *acb;
+    uint64_t offset;
+    QEMUIOVector qiov;
+    void *buffer;
+    size_t len;
+    BlockDriverCompletionFunc *cb;
+    void *opaque;
+} QEDCopyBackingData;
+
+static void qed_aio_copy_backing_cb(void *opaque, int ret)
+{
+    QEDCopyBackingData *copy_backing_data = opaque;
+    QEDAIOCB *acb = copy_backing_data->acb;
+
+    if (ret) {
+        ret = -EIO;
+    } else {
+        ret = (acb->end_pos - copy_backing_data->offset) / BDRV_SECTOR_SIZE;
+    }
+
+    copy_backing_data->cb(copy_backing_data->opaque, ret);
+
+    qemu_iovec_destroy(&copy_backing_data->qiov);
+    qemu_vfree(copy_backing_data->buffer);
+    qemu_free(copy_backing_data);
+}
+
+static void qed_copy_backing_find_cluster_cb(void *opaque, int ret,
+                                             uint64_t offset, size_t len);
+
+/**
+ * Perform the next qed_find_cluster() from a BH
+ *
+ * This is necessary because we iterate over each cluster in turn.
+ * qed_find_cluster() may invoke its callback immediately without returning up
+ * the call stack, causing us to overflow the call stack.  By starting each
+ * iteration from a BH we guarantee that a fresh stack is used each time.
+ */
+static void qed_copy_backing_next_cluster_bh(void *opaque)
+{
+    QEDCopyBackingData *copy_backing_data = opaque;
+    QEDAIOCB *acb = copy_backing_data->acb;
+    BDRVQEDState *s = acb_to_s(acb);
+
+    qemu_bh_delete(acb->bh);
+    acb->bh = NULL;
+
+    acb->cur_pos += s->header.cluster_size;
+    acb->end_pos += s->header.cluster_size;
+
+    qed_find_cluster(s, &acb->request, acb->cur_pos,
+                     acb->end_pos - acb->cur_pos,
+                     qed_copy_backing_find_cluster_cb, copy_backing_data);
+}
+
+/**
+ * Search for an unallocated cluster adjusting the current request until we
+ * can use it to read an unallocated cluster.
+ *
+ * Callback from qed_find_cluster().
+ */
+static void qed_copy_backing_find_cluster_cb(void *opaque, int ret,
+                                             uint64_t offset, size_t len)
+{
+    QEDCopyBackingData *copy_backing_data = opaque;
+    QEDAIOCB *acb = copy_backing_data->acb;
+    BDRVQEDState *s = acb_to_s(acb);
+
+    if (ret < 0) {
+        qed_aio_complete(acb, ret);
+        return;
+    }
+
+    if (ret == QED_CLUSTER_FOUND ||
+        ret == QED_CLUSTER_ZERO) {
+        /* proceed to next cluster */
+
+        if (acb->end_pos == s->header.image_size) {
+            qed_aio_complete(acb, 0);
+            return;
+        }
+
+        acb->bh = qemu_bh_new(qed_copy_backing_next_cluster_bh,
+                              copy_backing_data);
+        qemu_bh_schedule(acb->bh);
+    } else {
+        /* found a hole, kick off request */
+        qed_aio_next_io(acb, 0);
+    }
+}
+
+static BlockDriverAIOCB *bdrv_qed_aio_copy_backing(BlockDriverState *bs,
+    int64_t sector_num, BlockDriverCompletionFunc *cb, void *opaque)
+{
+    BDRVQEDState *s = bs->opaque;
+    QEDCopyBackingData *copy_backing_data;
+    QEDAIOCB *acb;
+    uint32_t cluster_size = s->header.cluster_size;
+    uint64_t start_cluster;
+    QEMUIOVector *qiov;
+
+    copy_backing_data = qemu_mallocz(sizeof(*copy_backing_data));
+
+    copy_backing_data->cb = cb;
+    copy_backing_data->opaque = opaque;
+    copy_backing_data->len = cluster_size;
+    copy_backing_data->buffer = qemu_blockalign(s->bs, cluster_size);
+    copy_backing_data->offset = sector_num * BDRV_SECTOR_SIZE;
+
+    start_cluster = qed_start_of_cluster(s, copy_backing_data->offset);
+    sector_num = start_cluster / BDRV_SECTOR_SIZE;
+
+    qiov = &copy_backing_data->qiov;
+    qemu_iovec_init(qiov, 1);
+    qemu_iovec_add(qiov, copy_backing_data->buffer, cluster_size);
+
+    acb = qed_aio_setup(bs, sector_num, qiov,
+                        cluster_size / BDRV_SECTOR_SIZE,
+                        qed_aio_copy_backing_cb, copy_backing_data,
+                        QED_AIOCB_CHECK_ZERO_WRITE |
+                        QED_AIOCB_COPY_ON_READ);
+    copy_backing_data->acb = acb;
+
+    qed_find_cluster(s, &acb->request, acb->cur_pos,
+                     acb->end_pos - acb->cur_pos,
+                     qed_copy_backing_find_cluster_cb, copy_backing_data);
+
+    return &acb->common;
 }
 
 static BlockDriverAIOCB *bdrv_qed_aio_flush(BlockDriverState *bs,
@@ -1527,6 +1742,7 @@ static BlockDriver bdrv_qed = {
     .bdrv_make_empty          = bdrv_qed_make_empty,
     .bdrv_aio_readv           = bdrv_qed_aio_readv,
     .bdrv_aio_writev          = bdrv_qed_aio_writev,
+    .bdrv_aio_copy_backing    = bdrv_qed_aio_copy_backing,
     .bdrv_aio_flush           = bdrv_qed_aio_flush,
     .bdrv_truncate            = bdrv_qed_truncate,
     .bdrv_getlength           = bdrv_qed_getlength,
diff --git a/block/qed.h b/block/qed.h
index 16f4bd9..48c65f7 100644
--- a/block/qed.h
+++ b/block/qed.h
@@ -124,8 +124,9 @@ typedef struct QEDRequest {
 } QEDRequest;
 
 enum {
-    QED_AIOCB_WRITE = 0x0001,           /* read or write? */
+    QED_AIOCB_WRITE = 0x0001,               /* read or write? */
     QED_AIOCB_COPY_ON_READ = 0x0002,
+    QED_AIOCB_CHECK_ZERO_WRITE = 0x0004,    /* detect zeroes? */
 };
 
 typedef struct QEDAIOCB {
-- 
1.7.5.4