* THE SOFTWARE.
*/
#include "config-host.h"
-#ifdef HOST_BSD
+#ifdef CONFIG_BSD
/* include native header before sys-queue.h */
#include <sys/queue.h>
#endif
#include "block_int.h"
#include "module.h"
-#ifdef HOST_BSD
+#ifdef CONFIG_BSD
#include <sys/types.h>
#include <sys/stat.h>
#include <sys/ioctl.h>
static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
BlockDriverCompletionFunc *cb, void *opaque);
+static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
+ BlockDriverCompletionFunc *cb, void *opaque);
static int bdrv_read_em(BlockDriverState *bs, int64_t sector_num,
uint8_t *buf, int nb_sectors);
static int bdrv_write_em(BlockDriverState *bs, int64_t sector_num,
bdrv->bdrv_read = bdrv_read_em;
bdrv->bdrv_write = bdrv_write_em;
}
+
+ if (!bdrv->bdrv_aio_flush)
+ bdrv->bdrv_aio_flush = bdrv_aio_flush_em;
+
bdrv->next = first_drv;
first_drv = bdrv;
}
filename[1] == ':');
}
-static int is_windows_drive(const char *filename)
+int is_windows_drive(const char *filename)
{
if (is_windows_drive_prefix(filename) &&
filename[2] == '\0')
return NULL;
}
-/* XXX: force raw format if block or character device ? It would
- simplify the BSD case */
+/*
+ * Detect host devices. By convention, /dev/cdrom[N] is always
+ * recognized as a host CDROM.
+ */
+static BlockDriver *find_hdev_driver(const char *filename)
+{
+ int score_max = 0, score;
+ BlockDriver *drv = NULL, *d;
+
+ for (d = first_drv; d; d = d->next) {
+ if (d->bdrv_probe_device) {
+ score = d->bdrv_probe_device(filename);
+ if (score > score_max) {
+ score_max = score;
+ drv = d;
+ }
+ }
+ }
+
+ return drv;
+}
+
static BlockDriver *find_image_format(const char *filename)
{
int ret, score, score_max;
uint8_t buf[2048];
BlockDriverState *bs;
- /* detect host devices. By convention, /dev/cdrom[N] is always
- recognized as a host CDROM */
- if (strstart(filename, "/dev/cdrom", NULL))
- return bdrv_find_format("host_device");
-#ifdef _WIN32
- if (is_windows_drive(filename))
- return bdrv_find_format("host_device");
-#else
- {
- struct stat st;
- if (stat(filename, &st) >= 0 &&
- (S_ISCHR(st.st_mode) || S_ISBLK(st.st_mode))) {
- return bdrv_find_format("host_device");
- }
- }
-#endif
-
drv = find_protocol(filename);
/* no need to test disk image formats for vvfat */
if (drv && strcmp(drv->format_name, "vvfat") == 0)
if (flags & BDRV_O_FILE) {
drv = find_protocol(filename);
} else if (!drv) {
- drv = find_image_format(filename);
+ drv = find_hdev_driver(filename);
+ if (!drv) {
+ drv = find_image_format(filename);
+ }
}
if (!drv) {
ret = -ENOENT;
}
bs->drv = drv;
bs->opaque = qemu_mallocz(drv->instance_size);
+
+ /*
+ * Yes, BDRV_O_NOCACHE aka O_DIRECT means we have to present a
+ * write cache to the guest. We do need the fdatasync to flush
+ * out transactions for block allocations, and we maybe have a
+ * volatile write cache in our backing device to deal with.
+ */
+ if (flags & (BDRV_O_CACHE_WB|BDRV_O_NOCACHE))
+ bs->enable_write_cache = 1;
+
/* Note: for compatibility, we open disk image files as RDWR, and
RDONLY as fallback */
if (!(flags & BDRV_O_FILE))
- open_flags = BDRV_O_RDWR | (flags & BDRV_O_CACHE_MASK);
+ open_flags = BDRV_O_RDWR |
+ (flags & (BDRV_O_CACHE_MASK|BDRV_O_NATIVE_AIO));
else
open_flags = flags & ~(BDRV_O_FILE | BDRV_O_SNAPSHOT);
ret = drv->bdrv_open(bs, filename, open_flags);
return bs->sg;
}
+int bdrv_enable_write_cache(BlockDriverState *bs)
+{
+ return bs->enable_write_cache;
+}
+
/* XXX: no longer used */
void bdrv_set_change_cb(BlockDriverState *bs,
void (*change_cb)(void *opaque), void *opaque)
return drv->bdrv_get_info(bs, bdi);
}
-int bdrv_put_buffer(BlockDriverState *bs, const uint8_t *buf, int64_t pos, int size)
+int bdrv_save_vmstate(BlockDriverState *bs, const uint8_t *buf,
+ int64_t pos, int size)
{
BlockDriver *drv = bs->drv;
if (!drv)
return -ENOMEDIUM;
- if (!drv->bdrv_put_buffer)
+ if (!drv->bdrv_save_vmstate)
return -ENOTSUP;
- return drv->bdrv_put_buffer(bs, buf, pos, size);
+ return drv->bdrv_save_vmstate(bs, buf, pos, size);
}
-int bdrv_get_buffer(BlockDriverState *bs, uint8_t *buf, int64_t pos, int size)
+int bdrv_load_vmstate(BlockDriverState *bs, uint8_t *buf,
+ int64_t pos, int size)
{
BlockDriver *drv = bs->drv;
if (!drv)
return -ENOMEDIUM;
- if (!drv->bdrv_get_buffer)
+ if (!drv->bdrv_load_vmstate)
return -ENOTSUP;
- return drv->bdrv_get_buffer(bs, buf, pos, size);
+ return drv->bdrv_load_vmstate(bs, buf, pos, size);
}
/**************************************************************/
return ret;
}
+
+typedef struct MultiwriteCB {
+ int error;
+ int num_requests;
+ int num_callbacks;
+ struct {
+ BlockDriverCompletionFunc *cb;
+ void *opaque;
+ QEMUIOVector *free_qiov;
+ void *free_buf;
+ } callbacks[];
+} MultiwriteCB;
+
+static void multiwrite_user_cb(MultiwriteCB *mcb)
+{
+ int i;
+
+ for (i = 0; i < mcb->num_callbacks; i++) {
+ mcb->callbacks[i].cb(mcb->callbacks[i].opaque, mcb->error);
+ qemu_free(mcb->callbacks[i].free_qiov);
+ qemu_free(mcb->callbacks[i].free_buf);
+ }
+}
+
+static void multiwrite_cb(void *opaque, int ret)
+{
+ MultiwriteCB *mcb = opaque;
+
+ if (ret < 0) {
+ mcb->error = ret;
+ multiwrite_user_cb(mcb);
+ }
+
+ mcb->num_requests--;
+ if (mcb->num_requests == 0) {
+ if (mcb->error == 0) {
+ multiwrite_user_cb(mcb);
+ }
+ qemu_free(mcb);
+ }
+}
+
+static int multiwrite_req_compare(const void *a, const void *b)
+{
+ return (((BlockRequest*) a)->sector - ((BlockRequest*) b)->sector);
+}
+
+/*
+ * Takes a bunch of requests and tries to merge them. Returns the number of
+ * requests that remain after merging.
+ */
+static int multiwrite_merge(BlockDriverState *bs, BlockRequest *reqs,
+ int num_reqs, MultiwriteCB *mcb)
+{
+ int i, outidx;
+
+ // Sort requests by start sector
+ qsort(reqs, num_reqs, sizeof(*reqs), &multiwrite_req_compare);
+
+ // Check if adjacent requests touch the same clusters. If so, combine them,
+ // filling up gaps with zero sectors.
+ outidx = 0;
+ for (i = 1; i < num_reqs; i++) {
+ int merge = 0;
+ int64_t oldreq_last = reqs[outidx].sector + reqs[outidx].nb_sectors;
+
+ // This handles the cases that are valid for all block drivers, namely
+ // exactly sequential writes and overlapping writes.
+ if (reqs[i].sector <= oldreq_last) {
+ merge = 1;
+ }
+
+ // The block driver may decide that it makes sense to combine requests
+ // even if there is a gap of some sectors between them. In this case,
+ // the gap is filled with zeros (therefore only applicable for yet
+ // unused space in format like qcow2).
+ if (!merge && bs->drv->bdrv_merge_requests) {
+ merge = bs->drv->bdrv_merge_requests(bs, &reqs[outidx], &reqs[i]);
+ }
+
+ if (merge) {
+ size_t size;
+ QEMUIOVector *qiov = qemu_mallocz(sizeof(*qiov));
+ qemu_iovec_init(qiov,
+ reqs[outidx].qiov->niov + reqs[i].qiov->niov + 1);
+
+ // Add the first request to the merged one. If the requests are
+ // overlapping, drop the last sectors of the first request.
+ size = (reqs[i].sector - reqs[outidx].sector) << 9;
+ qemu_iovec_concat(qiov, reqs[outidx].qiov, size);
+
+ // We might need to add some zeros between the two requests
+ if (reqs[i].sector > oldreq_last) {
+ size_t zero_bytes = (reqs[i].sector - oldreq_last) << 9;
+ uint8_t *buf = qemu_blockalign(bs, zero_bytes);
+ memset(buf, 0, zero_bytes);
+ qemu_iovec_add(qiov, buf, zero_bytes);
+ mcb->callbacks[i].free_buf = buf;
+ }
+
+ // Add the second request
+ qemu_iovec_concat(qiov, reqs[i].qiov, reqs[i].qiov->size);
+
+ reqs[outidx].nb_sectors += reqs[i].nb_sectors;
+ reqs[outidx].qiov = qiov;
+
+ mcb->callbacks[i].free_qiov = reqs[outidx].qiov;
+ } else {
+ outidx++;
+ reqs[outidx].sector = reqs[i].sector;
+ reqs[outidx].nb_sectors = reqs[i].nb_sectors;
+ reqs[outidx].qiov = reqs[i].qiov;
+ }
+ }
+
+ return outidx + 1;
+}
+
+/*
+ * Submit multiple AIO write requests at once.
+ *
+ * On success, the function returns 0 and all requests in the reqs array have
+ * been submitted. In error case this function returns -1, and any of the
+ * requests may or may not be submitted yet. In particular, this means that the
+ * callback will be called for some of the requests, for others it won't. The
+ * caller must check the error field of the BlockRequest to wait for the right
+ * callbacks (if error != 0, no callback will be called).
+ *
+ * The implementation may modify the contents of the reqs array, e.g. to merge
+ * requests. However, the fields opaque and error are left unmodified as they
+ * are used to signal failure for a single request to the caller.
+ */
+int bdrv_aio_multiwrite(BlockDriverState *bs, BlockRequest *reqs, int num_reqs)
+{
+ BlockDriverAIOCB *acb;
+ MultiwriteCB *mcb;
+ int i;
+
+ if (num_reqs == 0) {
+ return 0;
+ }
+
+ // Create MultiwriteCB structure
+ mcb = qemu_mallocz(sizeof(*mcb) + num_reqs * sizeof(*mcb->callbacks));
+ mcb->num_requests = 0;
+ mcb->num_callbacks = num_reqs;
+
+ for (i = 0; i < num_reqs; i++) {
+ mcb->callbacks[i].cb = reqs[i].cb;
+ mcb->callbacks[i].opaque = reqs[i].opaque;
+ }
+
+ // Check for mergable requests
+ num_reqs = multiwrite_merge(bs, reqs, num_reqs, mcb);
+
+ // Run the aio requests
+ for (i = 0; i < num_reqs; i++) {
+ acb = bdrv_aio_writev(bs, reqs[i].sector, reqs[i].qiov,
+ reqs[i].nb_sectors, multiwrite_cb, mcb);
+
+ if (acb == NULL) {
+ // We can only fail the whole thing if no request has been
+ // submitted yet. Otherwise we'll wait for the submitted AIOs to
+ // complete and report the error in the callback.
+ if (mcb->num_requests == 0) {
+ reqs[i].error = EIO;
+ goto fail;
+ } else {
+ mcb->error = EIO;
+ break;
+ }
+ } else {
+ mcb->num_requests++;
+ }
+ }
+
+ return 0;
+
+fail:
+ free(mcb);
+ return -1;
+}
+
+BlockDriverAIOCB *bdrv_aio_flush(BlockDriverState *bs,
+ BlockDriverCompletionFunc *cb, void *opaque)
+{
+ BlockDriver *drv = bs->drv;
+
+ if (!drv)
+ return NULL;
+
+ /*
+ * Note that unlike bdrv_flush the driver is reponsible for flushing a
+ * backing image if it exists.
+ */
+ return drv->bdrv_aio_flush(bs, cb, opaque);
+}
+
void bdrv_aio_cancel(BlockDriverAIOCB *acb)
{
acb->pool->cancel(acb);
static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb)
{
BlockDriverAIOCBSync *acb = (BlockDriverAIOCBSync *)blockacb;
- qemu_bh_cancel(acb->bh);
+ qemu_bh_delete(acb->bh);
+ acb->bh = NULL;
qemu_aio_release(acb);
}
qemu_iovec_from_buffer(acb->qiov, acb->bounce, acb->qiov->size);
qemu_vfree(acb->bounce);
acb->common.cb(acb->common.opaque, acb->ret);
-
+ qemu_bh_delete(acb->bh);
+ acb->bh = NULL;
qemu_aio_release(acb);
}
return bdrv_aio_rw_vector(bs, sector_num, qiov, nb_sectors, cb, opaque, 1);
}
+static BlockDriverAIOCB *bdrv_aio_flush_em(BlockDriverState *bs,
+ BlockDriverCompletionFunc *cb, void *opaque)
+{
+ BlockDriverAIOCBSync *acb;
+
+ acb = qemu_aio_get(&bdrv_em_aio_pool, bs, cb, opaque);
+ acb->is_write = 1; /* don't bounce in the completion hadler */
+ acb->qiov = NULL;
+ acb->bounce = NULL;
+ acb->ret = 0;
+
+ if (!acb->bh)
+ acb->bh = qemu_bh_new(bdrv_aio_bh_cb, acb);
+
+ bdrv_flush(bs);
+ qemu_bh_schedule(acb->bh);
+ return &acb->common;
+}
+
/**************************************************************/
/* sync block device emulation */
/**
* If eject_flag is TRUE, eject the media. Otherwise, close the tray
*/
-void bdrv_eject(BlockDriverState *bs, int eject_flag)
+int bdrv_eject(BlockDriverState *bs, int eject_flag)
{
BlockDriver *drv = bs->drv;
int ret;
+ if (bs->locked) {
+ return -EBUSY;
+ }
+
if (!drv || !drv->bdrv_eject) {
ret = -ENOTSUP;
} else {
if (ret == -ENOTSUP) {
if (eject_flag)
bdrv_close(bs);
+ ret = 0;
}
+
+ return ret;
}
int bdrv_is_locked(BlockDriverState *bs)
projects / qemu / blobdiff