static void fvd_store_compact_cancel (FvdAIOCB * acb)
{
if (acb->store.children) {
int i;
for (i = 0; i < acb->store.num_children; i++) {
if (acb->store.children[i].hd_acb) {
bdrv_aio_cancel (acb->store.children[i].hd_acb);
}
}
my_qemu_free (acb->store.children);
}
if (acb->store.one_child.hd_acb) {
bdrv_aio_cancel (acb->store.one_child.hd_acb);
}
if (acb->jcb.hd_acb) {
bdrv_aio_cancel (acb->jcb.hd_acb);
free_journal_sectors (acb->common.bs->opaque);
}
if (acb->jcb.iov.iov_base != NULL) {
my_qemu_vfree (acb->jcb.iov.iov_base);
}
if (acb->jcb.next_wait_for_journal.le_prev) {
QLIST_REMOVE (acb, jcb.next_wait_for_journal);
}
my_qemu_aio_unref (acb);
}
static void dma_aio_cancel(BlockDriverAIOCB *acb)
{
DMAAIOCB *dbs = container_of(acb, DMAAIOCB, common);
if (dbs->acb) {
bdrv_aio_cancel(dbs->acb);
}
}
/* Cancel a pending data transfer. */
static void scsi_cancel_io(SCSIRequest *req)
{
SCSIGenericReq *r = DO_UPCAST(SCSIGenericReq, req, req);
DPRINTF("Cancel tag=0x%x\n", req->tag);
if (r->req.aiocb) {
bdrv_aio_cancel(r->req.aiocb);
}
r->req.aiocb = NULL;
}
static void dma_aio_cancel(BlockDriverAIOCB *acb)
{
DMAAIOCB *dbs = container_of(acb, DMAAIOCB, common);
if (dbs->acb) {
BlockDriverAIOCB *acb = dbs->acb;
dbs->acb = NULL;
dbs->in_cancel = true;
bdrv_aio_cancel(acb);
dbs->in_cancel = false;
}
dbs->common.cb = NULL;
dma_complete(dbs, 0);
}
/* Cancel a pending data transfer. */
static void scsi_cancel_io(SCSIRequest *req)
{
SCSIGenericReq *r = DO_UPCAST(SCSIGenericReq, req, req);
DPRINTF("Cancel tag=0x%x\n", req->tag);
if (r->req.aiocb) {
bdrv_aio_cancel(r->req.aiocb);
/* This reference was left in by scsi_*_data. We take ownership of
* it independent of whether bdrv_aio_cancel completes the request
* or not. */
scsi_req_unref(&r->req);
}
r->req.aiocb = NULL;
}
/* Return FALSE if the submitted request is cancelled. */
static int submit_rand_io (RandomIO * r)
{
BlockDriverAIOCB *acb = NULL;
QDEBUG ("TESTER %03d: %s test%" PRIX64 " sector_num=%" PRId64
" nb_sectors=%d niov=%d\n", r->tester, op_type_str[r->type],
r->uuid, r->sector_num, r->nb_sectors, r->qiov.niov);
printf ("TESTER %03d: %s sector_num=%" PRId64 " nb_sectors=%d niov=%d\n",
r->tester, op_type_str[r->type], r->sector_num, r->nb_sectors,
r->qiov.niov);
int ret;
if (fail_prob <= 0) {
ret = 0;
} else if (random () / (double) RAND_MAX <= fail_prob) {
ret = -EIO;
} else {
ret = 0;
}
/* This affects whether this request will fail or not. */
sim_set_disk_io_return_code (ret);
switch (r->type) {
case OP_READ:
if (!(acb = bdrv_aio_readv (bs, r->sector_num, &r->qiov, r->nb_sectors,
rand_io_cb, r))) {
die ("bdrv_aio_readv\n");
}
break;
case OP_WRITE:
if (!(acb = bdrv_aio_writev (bs, r->sector_num, &r->qiov, r->nb_sectors,
rand_io_cb, r))) {
die ("bdrv_aio_writev\n");
}
break;
case OP_FLUSH:
if (!(acb = bdrv_aio_flush (bs, rand_io_cb, r))) {
die ("bdrv_aio_flush\n");
}
break;
case OP_NULL:
die ("OP_NULL");
break;
}
sim_set_disk_io_return_code (0); /* Reset to no failure state. */
if (r->allow_cancel && cancel_prob > 0 &&
random () / (double) RAND_MAX <= cancel_prob) {
QDEBUG ("TESTER %03d: cancel %s test%" PRIX64 " sector_num=%" PRId64
" nb_sectors=%d niov=%d\n", r->tester, op_type_str[r->type],
r->uuid, r->sector_num, r->nb_sectors, r->qiov.niov);
printf ("TESTER %03d: cancel %s sector_num=%" PRId64
" nb_sectors=%d niov=%d\n", r->tester, op_type_str[r->type],
r->sector_num, r->nb_sectors, r->qiov.niov);
bdrv_aio_cancel (acb);
return FALSE;
} else {
return TRUE;
}
}
/* Store data in the compact image. The argument 'soft_write' means
* the store was caused by copy-on-read or prefetching, which need not
* update metadata immediately. */
static BlockDriverAIOCB *store_data_in_compact_image (FvdAIOCB * acb,
int soft_write,
FvdAIOCB * parent_acb,
BlockDriverState * bs,
int64_t sector_num,
QEMUIOVector * orig_qiov,
const int nb_sectors,
BlockDriverCompletionFunc
* cb, void *opaque)
{
BDRVFvdState *s = bs->opaque;
const uint32_t first_chunk = sector_num / s->chunk_size;
const uint32_t last_chunk = (sector_num + nb_sectors - 1) / s->chunk_size;
int table_dirty = FALSE;
uint32_t chunk;
int64_t start_sec;
/* Check if storag space is allocated. */
for (chunk = first_chunk; chunk <= last_chunk; chunk++) {
if (IS_EMPTY (s->table[chunk])) {
uint32_t id = allocate_chunk (bs);
if (IS_EMPTY (id)) {
return NULL;
}
id |= DIRTY_TABLE;
WRITE_TABLE (s->table[chunk], id);
table_dirty = TRUE;
} else if (IS_DIRTY (s->table[chunk])) {
/* This is possible if a previous soft-write allocated the storage
* space but did not flush the table entry change to the journal
* and hence did not clean the dirty bit. This is also possible
* with two concurrent hard-writes. The first hard-write allocated
* the storage space but has not flushed the table entry change to
* the journal yet and hence the table entry remains dirty. In
* this case, the second hard-write will also try to flush this
* dirty table entry to the journal. The outcome is correct since
* they store the same metadata change in the journal (although
* twice). For this race condition, we prefer to have two writes
* to the journal rather than introducing a locking mechanism,
* because this happens rarely and those two writes to the journal
* are likely to be merged by the kernel into a single write since
* they are likely to update back-to-back sectors in the journal.
* A locking mechanism would be less efficient, because the large
* size of chunks would cause unnecessary locking due to ``false
* sharing'' of a chunk by two writes. */
table_dirty = TRUE;
}
}
const int update_table = (!soft_write && table_dirty);
size_t iov_left;
uint8_t *iov_buf;
int nb, iov_index, nqiov, niov;
uint32_t prev;
if (first_chunk == last_chunk) {
goto handle_one_continuous_region;
}
/* Count the number of qiov and iov needed to cover the continuous regions
* of the compact image. */
iov_left = orig_qiov->iov[0].iov_len;
iov_buf = orig_qiov->iov[0].iov_base;
iov_index = 0;
nqiov = 0;
niov = 0;
prev = READ_TABLE (s->table[first_chunk]);
/* Data in the first chunk. */
nb = s->chunk_size - (sector_num % s->chunk_size);
for (chunk = first_chunk + 1; chunk <= last_chunk; chunk++) {
uint32_t current = READ_TABLE (s->table[chunk]);
int64_t data_size;
if (chunk < last_chunk) {
data_size = s->chunk_size;
} else {
data_size = (sector_num + nb_sectors) % s->chunk_size;
if (data_size == 0) {
data_size = s->chunk_size;
}
}
if (current == prev + 1) {
nb += data_size; /* Continue the previous region. */
} else {
/* Terminate the previous region. */
niov +=
count_iov (orig_qiov->iov, &iov_index, &iov_buf, &iov_left,
nb * 512);
nqiov++;
nb = data_size; /* Data in the new region. */
}
prev = current;
}
if (nqiov == 0) {
handle_one_continuous_region:
/* A simple case. All data can be written out in one qiov and no new
* chunks are allocated. */
start_sec = READ_TABLE (s->table[first_chunk]) * s->chunk_size +
(sector_num % s->chunk_size);
if (!update_table && !acb) {
if (parent_acb) {
QDEBUG ("STORE: acb%llu-%p "
"store_directly_without_table_update\n",
parent_acb->uuid, parent_acb);
}
return bdrv_aio_writev (s->fvd_data, s->data_offset + start_sec,
orig_qiov, nb_sectors, cb, opaque);
}
if (!acb && !(acb = init_store_acb (soft_write, orig_qiov, bs,
sector_num, nb_sectors, parent_acb, cb, opaque))) {
return NULL;
}
QDEBUG ("STORE: acb%llu-%p store_directly sector_num=%" PRId64
" nb_sectors=%d\n", acb->uuid, acb, acb->sector_num,
acb->nb_sectors);
acb->store.update_table = update_table;
acb->store.num_children = 1;
acb->store.one_child.hd_acb =
bdrv_aio_writev (s->fvd_data, s->data_offset + start_sec, orig_qiov,
nb_sectors, finish_store_data_in_compact_image,
&acb->store.one_child);
if (acb->store.one_child.hd_acb) {
acb->store.one_child.acb = acb;
return &acb->common;
} else {
my_qemu_aio_unref (acb);
return NULL;
}
}
/* qiov for the last continuous region. */
niov += count_iov (orig_qiov->iov, &iov_index, &iov_buf,
&iov_left, nb * 512);
nqiov++;
ASSERT (iov_index == orig_qiov->niov - 1 && iov_left == 0);
/* Need to submit multiple requests to the lower layer. */
if (!acb && !(acb = init_store_acb (soft_write, orig_qiov, bs, sector_num,
nb_sectors, parent_acb, cb, opaque))) {
return NULL;
}
acb->store.update_table = update_table;
acb->store.num_children = nqiov;
if (!parent_acb) {
QDEBUG ("STORE: acb%llu-%p start sector_num=%" PRId64
" nb_sectors=%d\n", acb->uuid, acb, acb->sector_num,
acb->nb_sectors);
}
/* Allocate memory and create multiple requests. */
const size_t metadata_size = nqiov * (sizeof (CompactChildCB) +
sizeof (QEMUIOVector))
+ niov * sizeof (struct iovec);
acb->store.children = (CompactChildCB *) my_qemu_malloc (metadata_size);
QEMUIOVector *q = (QEMUIOVector *) (acb->store.children + nqiov);
struct iovec *v = (struct iovec *) (q + nqiov);
start_sec = READ_TABLE (s->table[first_chunk]) * s->chunk_size +
(sector_num % s->chunk_size);
nqiov = 0;
iov_index = 0;
iov_left = orig_qiov->iov[0].iov_len;
iov_buf = orig_qiov->iov[0].iov_base;
prev = READ_TABLE (s->table[first_chunk]);
/* Data in the first chunk. */
if (first_chunk == last_chunk) {
nb = nb_sectors;
}
else {
nb = s->chunk_size - (sector_num % s->chunk_size);
}
for (chunk = first_chunk + 1; chunk <= last_chunk; chunk++) {
uint32_t current = READ_TABLE (s->table[chunk]);
int64_t data_size;
if (chunk < last_chunk) {
data_size = s->chunk_size;
} else {
data_size = (sector_num + nb_sectors) % s->chunk_size;
if (data_size == 0) {
data_size = s->chunk_size;
}
}
if (current == prev + 1) {
nb += data_size; /* Continue the previous region. */
} else {
/* Terminate the previous continuous region. */
niov = setup_iov (orig_qiov->iov, v, &iov_index,
&iov_buf, &iov_left, nb * 512);
qemu_iovec_init_external (q, v, niov);
QDEBUG ("STORE: acb%llu-%p create_child %d sector_num=%" PRId64
" nb_sectors=%d niov=%d\n", acb->uuid, acb, nqiov,
start_sec, q->size / 512, q->niov);
acb->store.children[nqiov].hd_acb =
bdrv_aio_writev (s->fvd_data, s->data_offset + start_sec, q,
q->size / 512,
finish_store_data_in_compact_image,
&acb->store.children[nqiov]);
if (!acb->store.children[nqiov].hd_acb) {
goto fail;
}
acb->store.children[nqiov].acb = acb;
v += niov;
q++;
nqiov++;
start_sec = current * s->chunk_size; /* Begin of the new region. */
nb = data_size; /* Data in the new region. */
}
prev = current;
}
/* Requst for the last chunk. */
niov = setup_iov (orig_qiov->iov, v, &iov_index, &iov_buf,
&iov_left, nb * 512);
ASSERT (iov_index == orig_qiov->niov - 1 && iov_left == 0);
qemu_iovec_init_external (q, v, niov);
QDEBUG ("STORE: acb%llu-%p create_child_last %d sector_num=%" PRId64
" nb_sectors=%d niov=%d\n", acb->uuid, acb, nqiov, start_sec,
q->size / 512, q->niov);
acb->store.children[nqiov].hd_acb =
bdrv_aio_writev (s->fvd_data, s->data_offset + start_sec, q,
q->size / 512, finish_store_data_in_compact_image,
&acb->store.children[nqiov]);
if (acb->store.children[nqiov].hd_acb) {
acb->store.children[nqiov].acb = acb;
return &acb->common;
}
int i;
fail:
QDEBUG ("STORE: acb%llu-%p failed\n", acb->uuid, acb);
for (i = 0; i < nqiov; i++) {
bdrv_aio_cancel (acb->store.children[i].hd_acb);
}
my_qemu_free (acb->store.children);
my_qemu_aio_unref (acb);
return NULL;
}
