X-Git-Url: https://repo.jachan.dev/qemu.git/blobdiff_plain/5d5de250056b0972cde2e88133db702960a32b72..7d37435bd5801bb49e1c4b550fedd1c5fe143131:/block/file-posix.c

diff --git a/block/file-posix.c b/block/file-posix.c
index 27be94cfe5..8aee7a3fb8 100644
--- a/block/file-posix.c
+++ b/block/file-posix.c
@@ -1152,8 +1152,10 @@ static int hdev_probe_geometry(BlockDriverState *bs, HDGeometry *geo)
 }
 #endif
 
-static ssize_t handle_aiocb_ioctl(RawPosixAIOData *aiocb)
+#if defined(__linux__)
+static int handle_aiocb_ioctl(void *opaque)
 {
+    RawPosixAIOData *aiocb = opaque;
     int ret;
 
     ret = ioctl(aiocb->aio_fildes, aiocb->ioctl.cmd, aiocb->ioctl.buf);
@@ -1163,9 +1165,11 @@ static ssize_t handle_aiocb_ioctl(RawPosixAIOData *aiocb)
 
     return 0;
 }
+#endif /* linux */
 
-static ssize_t handle_aiocb_flush(RawPosixAIOData *aiocb)
+static int handle_aiocb_flush(void *opaque)
 {
+    RawPosixAIOData *aiocb = opaque;
     BDRVRawState *s = aiocb->bs->opaque;
     int ret;
 
@@ -1299,8 +1303,9 @@ static ssize_t handle_aiocb_rw_linear(RawPosixAIOData *aiocb, char *buf)
     return offset;
 }
 
-static ssize_t handle_aiocb_rw(RawPosixAIOData *aiocb)
+static int handle_aiocb_rw(void *opaque)
 {
+    RawPosixAIOData *aiocb = opaque;
     ssize_t nbytes;
     char *buf;
 
@@ -1310,7 +1315,8 @@ static ssize_t handle_aiocb_rw(RawPosixAIOData *aiocb)
          * we can just use plain pread/pwrite without any problems.
          */
         if (aiocb->io.niov == 1) {
-            return handle_aiocb_rw_linear(aiocb, aiocb->io.iov->iov_base);
+            nbytes = handle_aiocb_rw_linear(aiocb, aiocb->io.iov->iov_base);
+            goto out;
         }
         /*
          * We have more than one iovec, and all are properly aligned.
@@ -1322,7 +1328,7 @@ static ssize_t handle_aiocb_rw(RawPosixAIOData *aiocb)
             nbytes = handle_aiocb_rw_vector(aiocb);
             if (nbytes == aiocb->aio_nbytes ||
                 (nbytes < 0 && nbytes != -ENOSYS)) {
-                return nbytes;
+                goto out;
             }
             preadv_present = false;
         }
@@ -1340,7 +1346,8 @@ static ssize_t handle_aiocb_rw(RawPosixAIOData *aiocb)
      */
     buf = qemu_try_blockalign(aiocb->bs, aiocb->aio_nbytes);
     if (buf == NULL) {
-        return -ENOMEM;
+        nbytes = -ENOMEM;
+        goto out;
     }
 
     if (aiocb->aio_type & QEMU_AIO_WRITE) {
@@ -1374,7 +1381,21 @@ static ssize_t handle_aiocb_rw(RawPosixAIOData *aiocb)
     }
     qemu_vfree(buf);
 
-    return nbytes;
+out:
+    if (nbytes == aiocb->aio_nbytes) {
+        return 0;
+    } else if (nbytes >= 0 && nbytes < aiocb->aio_nbytes) {
+        if (aiocb->aio_type & QEMU_AIO_WRITE) {
+            return -EINVAL;
+        } else {
+            iov_memset(aiocb->io.iov, aiocb->io.niov, nbytes,
+                      0, aiocb->aio_nbytes - nbytes);
+            return 0;
+        }
+    } else {
+        assert(nbytes < 0);
+        return nbytes;
+    }
 }
 
 #ifdef CONFIG_XFS
@@ -1464,8 +1485,9 @@ static ssize_t handle_aiocb_write_zeroes_block(RawPosixAIOData *aiocb)
     return ret;
 }
 
-static ssize_t handle_aiocb_write_zeroes(RawPosixAIOData *aiocb)
+static int handle_aiocb_write_zeroes(void *opaque)
 {
+    RawPosixAIOData *aiocb = opaque;
 #if defined(CONFIG_FALLOCATE) || defined(CONFIG_XFS)
     BDRVRawState *s = aiocb->bs->opaque;
 #endif
@@ -1529,8 +1551,9 @@ static ssize_t handle_aiocb_write_zeroes(RawPosixAIOData *aiocb)
     return -ENOTSUP;
 }
 
-static ssize_t handle_aiocb_write_zeroes_unmap(RawPosixAIOData *aiocb)
+static int handle_aiocb_write_zeroes_unmap(void *opaque)
 {
+    RawPosixAIOData *aiocb = opaque;
     BDRVRawState *s G_GNUC_UNUSED = aiocb->bs->opaque;
     int ret;
 
@@ -1572,8 +1595,9 @@ static off_t copy_file_range(int in_fd, off_t *in_off, int out_fd,
 }
 #endif
 
-static ssize_t handle_aiocb_copy_range(RawPosixAIOData *aiocb)
+static int handle_aiocb_copy_range(void *opaque)
 {
+    RawPosixAIOData *aiocb = opaque;
     uint64_t bytes = aiocb->aio_nbytes;
     off_t in_off = aiocb->aio_offset;
     off_t out_off = aiocb->copy_range.aio_offset2;
@@ -1605,8 +1629,9 @@ static ssize_t handle_aiocb_copy_range(RawPosixAIOData *aiocb)
     return 0;
 }
 
-static ssize_t handle_aiocb_discard(RawPosixAIOData *aiocb)
+static int handle_aiocb_discard(void *opaque)
 {
+    RawPosixAIOData *aiocb = opaque;
     int ret = -EOPNOTSUPP;
     BDRVRawState *s = aiocb->bs->opaque;
 
@@ -1645,8 +1670,9 @@ static ssize_t handle_aiocb_discard(RawPosixAIOData *aiocb)
     return ret;
 }
 
-static int handle_aiocb_truncate(RawPosixAIOData *aiocb)
+static int handle_aiocb_truncate(void *opaque)
 {
+    RawPosixAIOData *aiocb = opaque;
     int result = 0;
     int64_t current_length = 0;
     char *buf = NULL;
@@ -1765,65 +1791,6 @@ out:
     return result;
 }
 
-static int aio_worker(void *arg)
-{
-    RawPosixAIOData *aiocb = arg;
-    ssize_t ret = 0;
-
-    switch (aiocb->aio_type & QEMU_AIO_TYPE_MASK) {
-    case QEMU_AIO_READ:
-        ret = handle_aiocb_rw(aiocb);
-        if (ret >= 0 && ret < aiocb->aio_nbytes) {
-            iov_memset(aiocb->io.iov, aiocb->io.niov, ret,
-                      0, aiocb->aio_nbytes - ret);
-
-            ret = aiocb->aio_nbytes;
-        }
-        if (ret == aiocb->aio_nbytes) {
-            ret = 0;
-        } else if (ret >= 0 && ret < aiocb->aio_nbytes) {
-            ret = -EINVAL;
-        }
-        break;
-    case QEMU_AIO_WRITE:
-        ret = handle_aiocb_rw(aiocb);
-        if (ret == aiocb->aio_nbytes) {
-            ret = 0;
-        } else if (ret >= 0 && ret < aiocb->aio_nbytes) {
-            ret = -EINVAL;
-        }
-        break;
-    case QEMU_AIO_FLUSH:
-        ret = handle_aiocb_flush(aiocb);
-        break;
-    case QEMU_AIO_IOCTL:
-        ret = handle_aiocb_ioctl(aiocb);
-        break;
-    case QEMU_AIO_DISCARD:
-        ret = handle_aiocb_discard(aiocb);
-        break;
-    case QEMU_AIO_WRITE_ZEROES:
-        ret = handle_aiocb_write_zeroes(aiocb);
-        break;
-    case QEMU_AIO_WRITE_ZEROES | QEMU_AIO_DISCARD:
-        ret = handle_aiocb_write_zeroes_unmap(aiocb);
-        break;
-    case QEMU_AIO_COPY_RANGE:
-        ret = handle_aiocb_copy_range(aiocb);
-        break;
-    case QEMU_AIO_TRUNCATE:
-        ret = handle_aiocb_truncate(aiocb);
-        break;
-    default:
-        error_report("invalid aio request (0x%x)", aiocb->aio_type);
-        ret = -EINVAL;
-        break;
-    }
-
-    g_free(aiocb);
-    return ret;
-}
-
 static int coroutine_fn raw_thread_pool_submit(BlockDriverState *bs,
                                                ThreadPoolFunc func, void *arg)
 {
@@ -1832,44 +1799,11 @@ static int coroutine_fn raw_thread_pool_submit(BlockDriverState *bs,
     return thread_pool_submit_co(pool, func, arg);
 }
 
-static int paio_submit_co_full(BlockDriverState *bs, int fd,
-                               int64_t offset, int fd2, int64_t offset2,
-                               QEMUIOVector *qiov,
-                               int bytes, int type)
-{
-    RawPosixAIOData *acb = g_new(RawPosixAIOData, 1);
-
-    acb->bs = bs;
-    acb->aio_type = type;
-    acb->aio_fildes = fd;
-
-    acb->aio_nbytes = bytes;
-    acb->aio_offset = offset;
-
-    if (qiov) {
-        acb->io.iov = qiov->iov;
-        acb->io.niov = qiov->niov;
-        assert(qiov->size == bytes);
-    } else {
-        acb->copy_range.aio_fd2 = fd2;
-        acb->copy_range.aio_offset2 = offset2;
-    }
-
-    trace_file_paio_submit_co(offset, bytes, type);
-    return raw_thread_pool_submit(bs, aio_worker, acb);
-}
-
-static inline int paio_submit_co(BlockDriverState *bs, int fd,
-                                 int64_t offset, QEMUIOVector *qiov,
-                                 int bytes, int type)
-{
-    return paio_submit_co_full(bs, fd, offset, -1, 0, qiov, bytes, type);
-}
-
 static int coroutine_fn raw_co_prw(BlockDriverState *bs, uint64_t offset,
                                    uint64_t bytes, QEMUIOVector *qiov, int type)
 {
     BDRVRawState *s = bs->opaque;
+    RawPosixAIOData acb;
 
     if (fd_open(bs) < 0)
         return -EIO;
@@ -1892,7 +1826,20 @@ static int coroutine_fn raw_co_prw(BlockDriverState *bs, uint64_t offset,
         }
     }
 
-    return paio_submit_co(bs, s->fd, offset, qiov, bytes, type);
+    acb = (RawPosixAIOData) {
+        .bs             = bs,
+        .aio_fildes     = s->fd,
+        .aio_type       = type,
+        .aio_offset     = offset,
+        .aio_nbytes     = bytes,
+        .io             = {
+            .iov            = qiov->iov,
+            .niov           = qiov->niov,
+        },
+    };
+
+    assert(qiov->size == bytes);
+    return raw_thread_pool_submit(bs, handle_aiocb_rw, &acb);
 }
 
 static int coroutine_fn raw_co_preadv(BlockDriverState *bs, uint64_t offset,
@@ -1935,6 +1882,7 @@ static void raw_aio_unplug(BlockDriverState *bs)
 static int raw_co_flush_to_disk(BlockDriverState *bs)
 {
     BDRVRawState *s = bs->opaque;
+    RawPosixAIOData acb;
     int ret;
 
     ret = fd_open(bs);
@@ -1942,7 +1890,13 @@ static int raw_co_flush_to_disk(BlockDriverState *bs)
         return ret;
     }
 
-    return paio_submit_co(bs, s->fd, 0, NULL, 0, QEMU_AIO_FLUSH);
+    acb = (RawPosixAIOData) {
+        .bs             = bs,
+        .aio_fildes     = s->fd,
+        .aio_type       = QEMU_AIO_FLUSH,
+    };
+
+    return raw_thread_pool_submit(bs, handle_aiocb_flush, &acb);
 }
 
 static void raw_aio_attach_aio_context(BlockDriverState *bs,
@@ -1981,9 +1935,9 @@ static int coroutine_fn
 raw_regular_truncate(BlockDriverState *bs, int fd, int64_t offset,
                      PreallocMode prealloc, Error **errp)
 {
-    RawPosixAIOData *acb = g_new(RawPosixAIOData, 1);
+    RawPosixAIOData acb;
 
-    *acb = (RawPosixAIOData) {
+    acb = (RawPosixAIOData) {
         .bs             = bs,
         .aio_fildes     = fd,
         .aio_type       = QEMU_AIO_TRUNCATE,
@@ -1994,7 +1948,7 @@ raw_regular_truncate(BlockDriverState *bs, int fd, int64_t offset,
         },
     };
 
-    return raw_thread_pool_submit(bs, aio_worker, acb);
+    return raw_thread_pool_submit(bs, handle_aiocb_truncate, &acb);
 }
 
 static int coroutine_fn raw_co_truncate(BlockDriverState *bs, int64_t offset,
@@ -2129,7 +2083,7 @@ again:
 #endif
     if (!fstat(fd, &sb) && (S_IFCHR & sb.st_mode)) {
 #ifdef DIOCGMEDIASIZE
-	if (ioctl(fd, DIOCGMEDIASIZE, (off_t *)&size))
+        if (ioctl(fd, DIOCGMEDIASIZE, (off_t *)&size))
 #elif defined(DIOCGPART)
         {
                 struct partinfo pi;
@@ -2625,25 +2579,67 @@ static void coroutine_fn raw_co_invalidate_cache(BlockDriverState *bs,
 }
 
 static coroutine_fn int
-raw_co_pdiscard(BlockDriverState *bs, int64_t offset, int bytes)
+raw_do_pdiscard(BlockDriverState *bs, int64_t offset, int bytes, bool blkdev)
 {
     BDRVRawState *s = bs->opaque;
+    RawPosixAIOData acb;
+
+    acb = (RawPosixAIOData) {
+        .bs             = bs,
+        .aio_fildes     = s->fd,
+        .aio_type       = QEMU_AIO_DISCARD,
+        .aio_offset     = offset,
+        .aio_nbytes     = bytes,
+    };
+
+    if (blkdev) {
+        acb.aio_type |= QEMU_AIO_BLKDEV;
+    }
 
-    return paio_submit_co(bs, s->fd, offset, NULL, bytes, QEMU_AIO_DISCARD);
+    return raw_thread_pool_submit(bs, handle_aiocb_discard, &acb);
 }
 
-static int coroutine_fn raw_co_pwrite_zeroes(
-    BlockDriverState *bs, int64_t offset,
-    int bytes, BdrvRequestFlags flags)
+static coroutine_fn int
+raw_co_pdiscard(BlockDriverState *bs, int64_t offset, int bytes)
+{
+    return raw_do_pdiscard(bs, offset, bytes, false);
+}
+
+static int coroutine_fn
+raw_do_pwrite_zeroes(BlockDriverState *bs, int64_t offset, int bytes,
+                     BdrvRequestFlags flags, bool blkdev)
 {
     BDRVRawState *s = bs->opaque;
-    int operation = QEMU_AIO_WRITE_ZEROES;
+    RawPosixAIOData acb;
+    ThreadPoolFunc *handler;
+
+    acb = (RawPosixAIOData) {
+        .bs             = bs,
+        .aio_fildes     = s->fd,
+        .aio_type       = QEMU_AIO_WRITE_ZEROES,
+        .aio_offset     = offset,
+        .aio_nbytes     = bytes,
+    };
+
+    if (blkdev) {
+        acb.aio_type |= QEMU_AIO_BLKDEV;
+    }
 
     if (flags & BDRV_REQ_MAY_UNMAP) {
-        operation |= QEMU_AIO_DISCARD;
+        acb.aio_type |= QEMU_AIO_DISCARD;
+        handler = handle_aiocb_write_zeroes_unmap;
+    } else {
+        handler = handle_aiocb_write_zeroes;
     }
 
-    return paio_submit_co(bs, s->fd, offset, NULL, bytes, operation);
+    return raw_thread_pool_submit(bs, handler, &acb);
+}
+
+static int coroutine_fn raw_co_pwrite_zeroes(
+    BlockDriverState *bs, int64_t offset,
+    int bytes, BdrvRequestFlags flags)
+{
+    return raw_do_pwrite_zeroes(bs, offset, bytes, flags, false);
 }
 
 static int raw_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
@@ -2714,6 +2710,7 @@ static int coroutine_fn raw_co_copy_range_to(BlockDriverState *bs,
                                              BdrvRequestFlags read_flags,
                                              BdrvRequestFlags write_flags)
 {
+    RawPosixAIOData acb;
     BDRVRawState *s = bs->opaque;
     BDRVRawState *src_s;
 
@@ -2726,8 +2723,20 @@ static int coroutine_fn raw_co_copy_range_to(BlockDriverState *bs,
     if (fd_open(src->bs) < 0 || fd_open(dst->bs) < 0) {
         return -EIO;
     }
-    return paio_submit_co_full(bs, src_s->fd, src_offset, s->fd, dst_offset,
-                               NULL, bytes, QEMU_AIO_COPY_RANGE);
+
+    acb = (RawPosixAIOData) {
+        .bs             = bs,
+        .aio_type       = QEMU_AIO_COPY_RANGE,
+        .aio_fildes     = src_s->fd,
+        .aio_offset     = src_offset,
+        .aio_nbytes     = bytes,
+        .copy_range     = {
+            .aio_fd2        = s->fd,
+            .aio_offset2    = dst_offset,
+        },
+    };
+
+    return raw_thread_pool_submit(bs, handle_aiocb_copy_range, &acb);
 }
 
 BlockDriver bdrv_file = {
@@ -3075,36 +3084,39 @@ hdev_open_Mac_error:
 }
 
 #if defined(__linux__)
-
-static BlockAIOCB *hdev_aio_ioctl(BlockDriverState *bs,
-        unsigned long int req, void *buf,
-        BlockCompletionFunc *cb, void *opaque)
+static int coroutine_fn
+hdev_co_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
 {
     BDRVRawState *s = bs->opaque;
-    RawPosixAIOData *acb;
-    ThreadPool *pool;
+    RawPosixAIOData acb;
+    int ret;
 
-    if (fd_open(bs) < 0)
-        return NULL;
+    ret = fd_open(bs);
+    if (ret < 0) {
+        return ret;
+    }
 
     if (req == SG_IO && s->pr_mgr) {
         struct sg_io_hdr *io_hdr = buf;
         if (io_hdr->cmdp[0] == PERSISTENT_RESERVE_OUT ||
             io_hdr->cmdp[0] == PERSISTENT_RESERVE_IN) {
             return pr_manager_execute(s->pr_mgr, bdrv_get_aio_context(bs),
-                                      s->fd, io_hdr, cb, opaque);
+                                      s->fd, io_hdr);
         }
     }
 
-    acb = g_new(RawPosixAIOData, 1);
-    acb->bs = bs;
-    acb->aio_type = QEMU_AIO_IOCTL;
-    acb->aio_fildes = s->fd;
-    acb->aio_offset = 0;
-    acb->ioctl.buf = buf;
-    acb->ioctl.cmd = req;
-    pool = aio_get_thread_pool(bdrv_get_aio_context(bs));
-    return thread_pool_submit_aio(pool, aio_worker, acb, cb, opaque);
+    acb = (RawPosixAIOData) {
+        .bs         = bs,
+        .aio_type   = QEMU_AIO_IOCTL,
+        .aio_fildes = s->fd,
+        .aio_offset = 0,
+        .ioctl      = {
+            .buf        = buf,
+            .cmd        = req,
+        },
+    };
+
+    return raw_thread_pool_submit(bs, handle_aiocb_ioctl, &acb);
 }
 #endif /* linux */
 
@@ -3121,22 +3133,18 @@ static int fd_open(BlockDriverState *bs)
 static coroutine_fn int
 hdev_co_pdiscard(BlockDriverState *bs, int64_t offset, int bytes)
 {
-    BDRVRawState *s = bs->opaque;
     int ret;
 
     ret = fd_open(bs);
     if (ret < 0) {
         return ret;
     }
-    return paio_submit_co(bs, s->fd, offset, NULL, bytes,
-                          QEMU_AIO_DISCARD | QEMU_AIO_BLKDEV);
+    return raw_do_pdiscard(bs, offset, bytes, true);
 }
 
 static coroutine_fn int hdev_co_pwrite_zeroes(BlockDriverState *bs,
     int64_t offset, int bytes, BdrvRequestFlags flags)
 {
-    BDRVRawState *s = bs->opaque;
-    int operation = QEMU_AIO_WRITE_ZEROES | QEMU_AIO_BLKDEV;
     int rc;
 
     rc = fd_open(bs);
@@ -3144,11 +3152,7 @@ static coroutine_fn int hdev_co_pwrite_zeroes(BlockDriverState *bs,
         return rc;
     }
 
-    if (flags & BDRV_REQ_MAY_UNMAP) {
-        operation |= QEMU_AIO_DISCARD;
-    }
-
-    return paio_submit_co(bs, s->fd, offset, NULL, bytes, operation);
+    return raw_do_pwrite_zeroes(bs, offset, bytes, flags, true);
 }
 
 static int coroutine_fn hdev_co_create_opts(const char *filename, QemuOpts *opts,
@@ -3253,7 +3257,7 @@ static BlockDriver bdrv_host_device = {
 
     /* generic scsi device */
 #ifdef __linux__
-    .bdrv_aio_ioctl     = hdev_aio_ioctl,
+    .bdrv_co_ioctl          = hdev_co_ioctl,
 #endif
 };
 
@@ -3375,7 +3379,7 @@ static BlockDriver bdrv_host_cdrom = {
     .bdrv_lock_medium   = cdrom_lock_medium,
 
     /* generic scsi device */
-    .bdrv_aio_ioctl     = hdev_aio_ioctl,
+    .bdrv_co_ioctl      = hdev_co_ioctl,
 };
 #endif /* __linux__ */