From c3c10f7295421f3b6638716d59cac9531c34adc8 Mon Sep 17 00:00:00 2001 From: Vladimir Sementsov-Ogievskiy Date: Thu, 1 Nov 2018 21:27:37 +0300 Subject: [PATCH] qcow2: aio support for compressed cluster read Allocate buffers locally and release qcow2 lock. Than, reads inside qcow2_co_preadv_compressed may be done in parallel, however all decompression is still done synchronously. Let's improve it in the following commit. Signed-off-by: Vladimir Sementsov-Ogievskiy Reviewed-by: Alberto Garcia Signed-off-by: Kevin Wolf --- block/qcow2.c | 100 +++++++++++++++++++++++++------------------------- block/qcow2.h | 4 -- 2 files changed, 51 insertions(+), 53 deletions(-) diff --git a/block/qcow2.c b/block/qcow2.c index 014aca6492..5d8bac6b62 100644 --- a/block/qcow2.c +++ b/block/qcow2.c @@ -74,6 +74,13 @@ typedef struct { #define QCOW2_EXT_MAGIC_CRYPTO_HEADER 0x0537be77 #define QCOW2_EXT_MAGIC_BITMAPS 0x23852875 +static int coroutine_fn +qcow2_co_preadv_compressed(BlockDriverState *bs, + uint64_t file_cluster_offset, + uint64_t offset, + uint64_t bytes, + QEMUIOVector *qiov); + static int qcow2_probe(const uint8_t *buf, int buf_size, const char *filename) { const QCowHeader *cow_header = (const void *)buf; @@ -1414,7 +1421,6 @@ static int coroutine_fn qcow2_do_open(BlockDriverState *bs, QDict *options, goto fail; } - s->cluster_cache_offset = -1; s->flags = flags; ret = qcow2_refcount_init(bs); @@ -1914,15 +1920,15 @@ static coroutine_fn int qcow2_co_preadv(BlockDriverState *bs, uint64_t offset, break; case QCOW2_CLUSTER_COMPRESSED: - /* add AIO support for compressed blocks ? */ - ret = qcow2_decompress_cluster(bs, cluster_offset); + qemu_co_mutex_unlock(&s->lock); + ret = qcow2_co_preadv_compressed(bs, cluster_offset, + offset, cur_bytes, + &hd_qiov); + qemu_co_mutex_lock(&s->lock); if (ret < 0) { goto fail; } - qemu_iovec_from_buf(&hd_qiov, 0, - s->cluster_cache + offset_in_cluster, - cur_bytes); break; case QCOW2_CLUSTER_NORMAL: @@ -2058,8 +2064,6 @@ static coroutine_fn int qcow2_co_pwritev(BlockDriverState *bs, uint64_t offset, qemu_iovec_init(&hd_qiov, qiov->niov); - s->cluster_cache_offset = -1; /* disable compressed cache */ - qemu_co_mutex_lock(&s->lock); while (bytes != 0) { @@ -2223,8 +2227,6 @@ static void qcow2_close(BlockDriverState *bs) g_free(s->image_backing_file); g_free(s->image_backing_format); - g_free(s->cluster_cache); - qemu_vfree(s->cluster_data); qcow2_refcount_close(bs); qcow2_free_snapshots(bs); } @@ -3401,7 +3403,6 @@ qcow2_co_copy_range_to(BlockDriverState *bs, QCowL2Meta *l2meta = NULL; assert(!bs->encrypted); - s->cluster_cache_offset = -1; /* disable compressed cache */ qemu_co_mutex_lock(&s->lock); @@ -3957,52 +3958,53 @@ fail: return ret; } -int coroutine_fn -qcow2_decompress_cluster(BlockDriverState *bs, uint64_t cluster_offset) +static int coroutine_fn +qcow2_co_preadv_compressed(BlockDriverState *bs, + uint64_t file_cluster_offset, + uint64_t offset, + uint64_t bytes, + QEMUIOVector *qiov) { BDRVQcow2State *s = bs->opaque; - int ret, csize, nb_csectors; + int ret = 0, csize, nb_csectors; uint64_t coffset; + uint8_t *buf, *out_buf; struct iovec iov; QEMUIOVector local_qiov; + int offset_in_cluster = offset_into_cluster(s, offset); - coffset = cluster_offset & s->cluster_offset_mask; - if (s->cluster_cache_offset != coffset) { - nb_csectors = ((cluster_offset >> s->csize_shift) & s->csize_mask) + 1; - csize = nb_csectors * 512 - (coffset & 511); + coffset = file_cluster_offset & s->cluster_offset_mask; + nb_csectors = ((file_cluster_offset >> s->csize_shift) & s->csize_mask) + 1; + csize = nb_csectors * 512 - (coffset & 511); - /* Allocate buffers on first decompress operation, most images are - * uncompressed and the memory overhead can be avoided. The buffers - * are freed in .bdrv_close(). - */ - if (!s->cluster_data) { - /* one more sector for decompressed data alignment */ - s->cluster_data = qemu_try_blockalign(bs->file->bs, - QCOW_MAX_CRYPT_CLUSTERS * s->cluster_size + 512); - if (!s->cluster_data) { - return -ENOMEM; - } - } - if (!s->cluster_cache) { - s->cluster_cache = g_malloc(s->cluster_size); - } - - iov.iov_base = s->cluster_data; - iov.iov_len = csize; - qemu_iovec_init_external(&local_qiov, &iov, 1); - - BLKDBG_EVENT(bs->file, BLKDBG_READ_COMPRESSED); - ret = bdrv_co_preadv(bs->file, coffset, csize, &local_qiov, 0); - if (ret < 0) { - return ret; - } - if (qcow2_decompress(s->cluster_cache, s->cluster_size, - s->cluster_data, csize) < 0) { - return -EIO; - } - s->cluster_cache_offset = coffset; + buf = g_try_malloc(csize); + if (!buf) { + return -ENOMEM; } - return 0; + iov.iov_base = buf; + iov.iov_len = csize; + qemu_iovec_init_external(&local_qiov, &iov, 1); + + out_buf = qemu_blockalign(bs, s->cluster_size); + + BLKDBG_EVENT(bs->file, BLKDBG_READ_COMPRESSED); + ret = bdrv_co_preadv(bs->file, coffset, csize, &local_qiov, 0); + if (ret < 0) { + goto fail; + } + + if (qcow2_decompress(out_buf, s->cluster_size, buf, csize) < 0) { + ret = -EIO; + goto fail; + } + + qemu_iovec_from_buf(qiov, 0, out_buf + offset_in_cluster, bytes); + +fail: + qemu_vfree(out_buf); + g_free(buf); + + return ret; } static int make_completely_empty(BlockDriverState *bs) diff --git a/block/qcow2.h b/block/qcow2.h index 8662b68575..a98d24500b 100644 --- a/block/qcow2.h +++ b/block/qcow2.h @@ -278,9 +278,6 @@ typedef struct BDRVQcow2State { QEMUTimer *cache_clean_timer; unsigned cache_clean_interval; - uint8_t *cluster_cache; - uint8_t *cluster_data; - uint64_t cluster_cache_offset; QLIST_HEAD(QCowClusterAlloc, QCowL2Meta) cluster_allocs; uint64_t *refcount_table; @@ -616,7 +613,6 @@ int qcow2_grow_l1_table(BlockDriverState *bs, uint64_t min_size, bool exact_size); int qcow2_shrink_l1_table(BlockDriverState *bs, uint64_t max_size); int qcow2_write_l1_entry(BlockDriverState *bs, int l1_index); -int qcow2_decompress_cluster(BlockDriverState *bs, uint64_t cluster_offset); int qcow2_encrypt_sectors(BDRVQcow2State *s, int64_t sector_num, uint8_t *buf, int nb_sectors, bool enc, Error **errp);