[Qemu-devel] [PATCH 23/25] raw-win32: add emulated AIO support

[Paolo Bonzini <pbonzini@redhat.com>]

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 block/raw-win32.c | 187 ++++++++++++++++++++++++++++++++++++++++--------------
 1 file modificato, 138 inserzioni(+), 49 rimozioni(-)

diff --git a/block/raw-win32.c b/block/raw-win32.c
index 78c8306..ffd86e3 100644
--- a/block/raw-win32.c
+++ b/block/raw-win32.c
@@ -25,6 +25,10 @@
 #include "qemu-timer.h"
 #include "block_int.h"
 #include "module.h"
+#include "raw-aio.h"
+#include "trace.h"
+#include "thread-pool.h"
+#include "iov.h"
 #include <windows.h>
 #include <winioctl.h>
 
@@ -32,12 +36,127 @@
 #define FTYPE_CD     1
 #define FTYPE_HARDDISK 2
 
+typedef struct RawWin32AIOData {
+    BlockDriverState *bs;
+    HANDLE hfile;
+    struct iovec *aio_iov;
+    int aio_niov;
+    size_t aio_nbytes;
+    off64_t aio_offset;
+    int aio_type;
+} RawWin32AIOData;
+
 typedef struct BDRVRawState {
     HANDLE hfile;
     int type;
     char drive_path[16]; /* format: "d:\" */
 } BDRVRawState;
 
+/*
+ * Read/writes the data to/from a given linear buffer.
+ *
+ * Returns the number of bytes handles or -errno in case of an error. Short
+ * reads are only returned if the end of the file is reached.
+ */
+static size_t handle_aiocb_rw(RawWin32AIOData *aiocb)
+{
+    size_t offset = 0;
+    int i;
+
+    for (i = 0; i < aiocb->aio_niov; i++) {
+        OVERLAPPED ov;
+        DWORD ret, ret_count, len;
+
+        memset(&ov, 0, sizeof(ov));
+        ov.Offset = (aiocb->aio_offset + offset);
+        ov.OffsetHigh = (aiocb->aio_offset + offset) >> 32;
+        len = aiocb->aio_iov[i].iov_len;
+        if (aiocb->aio_type & QEMU_AIO_WRITE) {
+            ret = WriteFile(aiocb->hfile, aiocb->aio_iov[i].iov_base,
+                            len, &ret_count, &ov);
+        } else {
+            ret = ReadFile(aiocb->hfile, aiocb->aio_iov[i].iov_base,
+                           len, &ret_count, &ov);
+        }
+        if (!ret) {
+            ret_count = 0;
+        }
+        if (ret_count != len) {
+            break;
+        }
+        offset += len;
+    }
+
+    return offset;
+}
+
+static int aio_worker(void *arg)
+{
+    RawWin32AIOData *aiocb = arg;
+    ssize_t ret = 0;
+    size_t count;
+
+    switch (aiocb->aio_type & QEMU_AIO_TYPE_MASK) {
+    case QEMU_AIO_READ:
+        count = handle_aiocb_rw(aiocb);
+        if (count < aiocb->aio_nbytes && aiocb->bs->growable) {
+            /* A short read means that we have reached EOF. Pad the buffer
+             * with zeros for bytes after EOF. */
+            iov_memset(aiocb->aio_iov, aiocb->aio_niov, count,
+                      0, aiocb->aio_nbytes - count);
+
+            count = aiocb->aio_nbytes;
+        }
+        if (count == aiocb->aio_nbytes) {
+            ret = 0;
+        } else {
+            ret = -EINVAL;
+        }
+        break;
+    case QEMU_AIO_WRITE:
+        count = handle_aiocb_rw(aiocb);
+        if (count == aiocb->aio_nbytes) {
+            count = 0;
+        } else {
+            count = -EINVAL;
+        }
+        break;
+    case QEMU_AIO_FLUSH:
+        if (!FlushFileBuffers(aiocb->hfile)) {
+            return -EIO;
+        }
+        break;
+    default:
+        fprintf(stderr, "invalid aio request (0x%x)\n", aiocb->aio_type);
+        ret = -EINVAL;
+        break;
+    }
+
+    g_slice_free(RawWin32AIOData, aiocb);
+    return ret;
+}
+
+static BlockDriverAIOCB *paio_submit(BlockDriverState *bs, HANDLE hfile,
+        int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
+        BlockDriverCompletionFunc *cb, void *opaque, int type)
+{
+    RawWin32AIOData *acb = g_slice_new(RawWin32AIOData);
+
+    acb->bs = bs;
+    acb->hfile = hfile;
+    acb->aio_type = type;
+
+    if (qiov) {
+        acb->aio_iov = qiov->iov;
+        acb->aio_niov = qiov->niov;
+    }
+    acb->aio_nbytes = nb_sectors * 512;
+    acb->aio_offset = sector_num * 512;
+
+    trace_paio_submit(acb, opaque, sector_num, nb_sectors, type);
+    return thread_pool_submit_aio(aio_worker, acb, cb, opaque);
+}
+
 int qemu_ftruncate64(int fd, int64_t length)
 {
     LARGE_INTEGER li;
@@ -117,59 +236,29 @@ static int raw_open(BlockDriverState *bs, const char *filename, int flags)
     return 0;
 }
 
-static int raw_read(BlockDriverState *bs, int64_t sector_num,
-                    uint8_t *buf, int nb_sectors)
+static BlockDriverAIOCB *raw_aio_readv(BlockDriverState *bs,
+                         int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
+                         BlockDriverCompletionFunc *cb, void *opaque)
 {
     BDRVRawState *s = bs->opaque;
-    OVERLAPPED ov;
-    DWORD ret_count;
-    int ret;
-    int64_t offset = sector_num * 512;
-    int count = nb_sectors * 512;
-
-    memset(&ov, 0, sizeof(ov));
-    ov.Offset = offset;
-    ov.OffsetHigh = offset >> 32;
-    ret = ReadFile(s->hfile, buf, count, &ret_count, &ov);
-    if (!ret)
-        return ret_count;
-    if (ret_count == count)
-        ret_count = 0;
-    return ret_count;
+    return paio_submit(bs, s->hfile, sector_num, qiov, nb_sectors,
+                       cb, opaque, QEMU_AIO_READ);
 }
 
-static int raw_write(BlockDriverState *bs, int64_t sector_num,
-                     const uint8_t *buf, int nb_sectors)
+static BlockDriverAIOCB *raw_aio_writev(BlockDriverState *bs,
+                          int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
+                          BlockDriverCompletionFunc *cb, void *opaque)
 {
     BDRVRawState *s = bs->opaque;
-    OVERLAPPED ov;
-    DWORD ret_count;
-    int ret;
-    int64_t offset = sector_num * 512;
-    int count = nb_sectors * 512;
-
-    memset(&ov, 0, sizeof(ov));
-    ov.Offset = offset;
-    ov.OffsetHigh = offset >> 32;
-    ret = WriteFile(s->hfile, buf, count, &ret_count, &ov);
-    if (!ret)
-        return ret_count;
-    if (ret_count == count)
-        ret_count = 0;
-    return ret_count;
+    return paio_submit(bs, s->hfile, sector_num, qiov, nb_sectors,
+                       cb, opaque, QEMU_AIO_WRITE);
 }
 
-static int raw_flush(BlockDriverState *bs)
+static BlockDriverAIOCB *raw_aio_flush(BlockDriverState *bs,
+                         BlockDriverCompletionFunc *cb, void *opaque)
 {
     BDRVRawState *s = bs->opaque;
-    int ret;
-
-    ret = FlushFileBuffers(s->hfile);
-    if (ret == 0) {
-        return -EIO;
-    }
-
-    return 0;
+    return paio_submit(bs, s->hfile, 0, NULL, 0, cb, opaque, QEMU_AIO_FLUSH);
 }
 
 static void raw_close(BlockDriverState *bs)
@@ -290,9 +379,9 @@ static BlockDriver bdrv_file = {
     .bdrv_close		= raw_close,
     .bdrv_create	= raw_create,
 
-    .bdrv_read              = raw_read,
-    .bdrv_write             = raw_write,
-    .bdrv_co_flush_to_disk  = raw_flush,
+    .bdrv_aio_readv     = raw_aio_readv,
+    .bdrv_aio_writev    = raw_aio_writev,
+    .bdrv_aio_flush     = raw_aio_flush,
 
     .bdrv_truncate	= raw_truncate,
     .bdrv_getlength	= raw_getlength,
@@ -413,9 +502,9 @@ static BlockDriver bdrv_host_device = {
     .bdrv_close		= raw_close,
     .bdrv_has_zero_init = hdev_has_zero_init,
 
-    .bdrv_read              = raw_read,
-    .bdrv_write             = raw_write,
-    .bdrv_co_flush_to_disk  = raw_flush,
+    .bdrv_aio_readv     = raw_aio_readv,
+    .bdrv_aio_writev    = raw_aio_writev,
+    .bdrv_aio_flush     = raw_aio_flush,
 
     .bdrv_getlength	= raw_getlength,
     .bdrv_get_allocated_file_size
-- 
1.7.12.1