From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from mailman by lists.gnu.org with tmda-scanned (Exim 4.43) id 1O5aSu-0000Bn-Jz for qemu-devel@nongnu.org; Sat, 24 Apr 2010 04:11:52 -0400 Received: from [140.186.70.92] (port=36039 helo=eggs.gnu.org) by lists.gnu.org with esmtp (Exim 4.43) id 1O5aSs-0000BX-Sc for qemu-devel@nongnu.org; Sat, 24 Apr 2010 04:11:52 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.69) (envelope-from ) id 1O5aSp-0001vl-NG for qemu-devel@nongnu.org; Sat, 24 Apr 2010 04:11:50 -0400 Received: from mtagate7.uk.ibm.com ([194.196.100.167]:39354) by eggs.gnu.org with esmtp (Exim 4.69) (envelope-from ) id 1O5aSp-0001vQ-D5 for qemu-devel@nongnu.org; Sat, 24 Apr 2010 04:11:47 -0400 Received: from d06nrmr1407.portsmouth.uk.ibm.com (d06nrmr1407.portsmouth.uk.ibm.com [9.149.38.185]) by mtagate7.uk.ibm.com (8.13.1/8.13.1) with ESMTP id o3O8Bi3h017314 for ; Sat, 24 Apr 2010 08:11:44 GMT Received: from d06av03.portsmouth.uk.ibm.com (d06av03.portsmouth.uk.ibm.com [9.149.37.213]) by d06nrmr1407.portsmouth.uk.ibm.com (8.13.8/8.13.8/NCO v10.0) with ESMTP id o3O8BiW5716936 for ; Sat, 24 Apr 2010 09:11:44 +0100 Received: from d06av03.portsmouth.uk.ibm.com (loopback [127.0.0.1]) by d06av03.portsmouth.uk.ibm.com (8.12.11.20060308/8.13.3) with ESMTP id o3O8BiH4010588 for ; Sat, 24 Apr 2010 09:11:44 +0100 From: Stefan Hajnoczi Date: Sat, 24 Apr 2010 09:12:13 +0100 Message-Id: <1272096733-6070-2-git-send-email-stefanha@linux.vnet.ibm.com> In-Reply-To: <1272096733-6070-1-git-send-email-stefanha@linux.vnet.ibm.com> References: <1272096733-6070-1-git-send-email-stefanha@linux.vnet.ibm.com> Subject: [Qemu-devel] [PATCH 2/2] qcow2: Implement bdrv_truncate() for growing images List-Id: qemu-devel.nongnu.org List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: qemu-devel@nongnu.org Cc: Kevin Wolf , Stefan Hajnoczi This patch adds the ability to grow qcow2 images in-place using bdrv_truncate(). This enables qemu-img resize command support for qcow2. Snapshots are not supported and bdrv_truncate() will return -ENOTSUP. The notion of resizing an image with snapshots could lead to confusion: users may expect snapshots to remain unchanged, but this is not possible with the current qcow2 on-disk format where the header.size field is global instead of per-snapshot. Others may expect snapshots to change size along with the current image data. I think it is safest to not support snapshots and perhaps add behavior later if there is a consensus. Backing images continue to work. If the image is now larger than its backing image, zeroes are read when accessing beyond the end of the backing image. Signed-off-by: Stefan Hajnoczi --- This applies to kevin/block. block/qcow2-cluster.c | 64 ++++++++++++++++++++++++++++++++++++++--------- block/qcow2-snapshot.c | 2 +- block/qcow2.c | 43 +++++++++++++++++++++++++++++--- block/qcow2.h | 9 ++++++- 4 files changed, 99 insertions(+), 19 deletions(-) diff --git a/block/qcow2-cluster.c b/block/qcow2-cluster.c index c11680d..20c8426 100644 --- a/block/qcow2-cluster.c +++ b/block/qcow2-cluster.c @@ -28,30 +28,39 @@ #include "block_int.h" #include "block/qcow2.h" -int qcow2_grow_l1_table(BlockDriverState *bs, int min_size) +/* + * qcow2_grow_l1_table_common + * + * Grows the L1 table and updates the header on disk. + * + * Setting new_l1_vm_state_index to s->l1_vm_state_index grows the vm state + * area. + * + * Setting new_l1_vm_state_index to the new end of image grows the image data + * area. + * + * Returns 0 on success, -errno in failure case. + */ +static int qcow2_grow_l1_table_common(BlockDriverState *bs, + int new_l1_vm_state_index, + int new_l1_size) { BDRVQcowState *s = bs->opaque; - int new_l1_size, new_l1_size2, ret, i; + int new_l1_size2, ret, i; uint64_t *new_l1_table; int64_t new_l1_table_offset; uint8_t data[12]; - new_l1_size = s->l1_size; - if (min_size <= new_l1_size) - return 0; - if (new_l1_size == 0) { - new_l1_size = 1; - } - while (min_size > new_l1_size) { - new_l1_size = (new_l1_size * 3 + 1) / 2; - } #ifdef DEBUG_ALLOC2 printf("grow l1_table from %d to %d\n", s->l1_size, new_l1_size); #endif new_l1_size2 = sizeof(uint64_t) * new_l1_size; new_l1_table = qemu_mallocz(align_offset(new_l1_size2, 512)); - memcpy(new_l1_table, s->l1_table, s->l1_size * sizeof(uint64_t)); + memcpy(new_l1_table, s->l1_table, s->l1_vm_state_index * sizeof(uint64_t)); + memcpy(&new_l1_table[new_l1_vm_state_index], + &s->l1_table[s->l1_vm_state_index], + (s->l1_size - s->l1_vm_state_index) * sizeof(uint64_t)); /* write new table (align to cluster) */ BLKDBG_EVENT(bs->file, BLKDBG_L1_GROW_ALLOC_TABLE); @@ -83,6 +92,7 @@ int qcow2_grow_l1_table(BlockDriverState *bs, int min_size) s->l1_table_offset = new_l1_table_offset; s->l1_table = new_l1_table; s->l1_size = new_l1_size; + s->l1_vm_state_index = new_l1_vm_state_index; return 0; fail: qemu_free(new_l1_table); @@ -90,6 +100,34 @@ int qcow2_grow_l1_table(BlockDriverState *bs, int min_size) return ret < 0 ? ret : -EIO; } +int qcow2_grow_l1_vm_state(BlockDriverState *bs, int min_size) +{ + BDRVQcowState *s = bs->opaque; + int new_l1_size; + + new_l1_size = s->l1_size; + if (min_size <= new_l1_size) + return 0; + if (new_l1_size == 0) { + new_l1_size = 1; + } + while (min_size > new_l1_size) { + new_l1_size = (new_l1_size * 3 + 1) / 2; + } + + return qcow2_grow_l1_table_common(bs, s->l1_vm_state_index, new_l1_size); +} + +int qcow2_grow_l1_image_data(BlockDriverState *bs, int new_l1_size) +{ + BDRVQcowState *s = bs->opaque; + int new_l1_vm_state_index; + + new_l1_vm_state_index = new_l1_size; + new_l1_size += s->l1_size - s->l1_vm_state_index; + return qcow2_grow_l1_table_common(bs, new_l1_vm_state_index, new_l1_size); +} + void qcow2_l2_cache_reset(BlockDriverState *bs) { BDRVQcowState *s = bs->opaque; @@ -524,7 +562,7 @@ static int get_cluster_table(BlockDriverState *bs, uint64_t offset, l1_index = offset >> (s->l2_bits + s->cluster_bits); if (l1_index >= s->l1_size) { - ret = qcow2_grow_l1_table(bs, l1_index + 1); + ret = qcow2_grow_l1_vm_state(bs, l1_index + 1); if (ret < 0) { return ret; } diff --git a/block/qcow2-snapshot.c b/block/qcow2-snapshot.c index 2a21c17..7f0d810 100644 --- a/block/qcow2-snapshot.c +++ b/block/qcow2-snapshot.c @@ -326,7 +326,7 @@ int qcow2_snapshot_goto(BlockDriverState *bs, const char *snapshot_id) if (qcow2_update_snapshot_refcount(bs, s->l1_table_offset, s->l1_size, -1) < 0) goto fail; - if (qcow2_grow_l1_table(bs, sn->l1_size) < 0) + if (qcow2_grow_l1_vm_state(bs, sn->l1_size) < 0) goto fail; s->l1_size = sn->l1_size; diff --git a/block/qcow2.c b/block/qcow2.c index 4a7ab66..ab622a2 100644 --- a/block/qcow2.c +++ b/block/qcow2.c @@ -140,7 +140,7 @@ static int qcow_read_extensions(BlockDriverState *bs, uint64_t start_offset, static int qcow_open(BlockDriverState *bs, int flags) { BDRVQcowState *s = bs->opaque; - int len, i, shift; + int len, i; QCowHeader header; uint64_t ext_end; @@ -188,8 +188,7 @@ static int qcow_open(BlockDriverState *bs, int flags) /* read the level 1 table */ s->l1_size = header.l1_size; - shift = s->cluster_bits + s->l2_bits; - s->l1_vm_state_index = (header.size + (1LL << shift) - 1) >> shift; + s->l1_vm_state_index = size_to_l1(s, header.size); /* the L1 table must contain at least enough entries to put header.size bytes */ if (s->l1_size < s->l1_vm_state_index) @@ -1095,6 +1094,40 @@ static int qcow_make_empty(BlockDriverState *bs) return 0; } +static int qcow2_truncate(BlockDriverState *bs, int64_t offset) +{ + BDRVQcowState *s = bs->opaque; + int ret, new_l1_size; + + if (offset & 511) { + return -EINVAL; + } + + /* cannot proceed if image has snapshots */ + if (s->nb_snapshots) { + return -ENOTSUP; + } + + /* shrinking is currently not supported */ + if (offset < bs->total_sectors << BDRV_SECTOR_BITS) { + return -ENOTSUP; + } + + new_l1_size = size_to_l1(s, offset); + ret = qcow2_grow_l1_image_data(bs, new_l1_size); + if (ret < 0) { + return ret; + } + + /* write updated header.size */ + offset = cpu_to_be64(offset); + if (bdrv_pwrite(bs->file, offsetof(QCowHeader, size), &offset, + sizeof(uint64_t)) != sizeof(uint64_t)) { + return -EIO; + } + return 0; +} + /* XXX: put compressed sectors first, then all the cluster aligned tables to avoid losing bytes in alignment */ static int qcow_write_compressed(BlockDriverState *bs, int64_t sector_num, @@ -1294,7 +1327,9 @@ static BlockDriver bdrv_qcow2 = { .bdrv_aio_readv = qcow_aio_readv, .bdrv_aio_writev = qcow_aio_writev, .bdrv_aio_flush = qcow_aio_flush, - .bdrv_write_compressed = qcow_write_compressed, + + .bdrv_truncate = qcow2_truncate, + .bdrv_write_compressed = qcow_write_compressed, .bdrv_snapshot_create = qcow2_snapshot_create, .bdrv_snapshot_goto = qcow2_snapshot_goto, diff --git a/block/qcow2.h b/block/qcow2.h index 5bd08db..c328248 100644 --- a/block/qcow2.h +++ b/block/qcow2.h @@ -150,6 +150,12 @@ static inline int size_to_clusters(BDRVQcowState *s, int64_t size) return (size + (s->cluster_size - 1)) >> s->cluster_bits; } +static inline int size_to_l1(BDRVQcowState *s, int64_t size) +{ + int shift = s->cluster_bits + s->l2_bits; + return (size + (1ULL << shift) - 1) >> shift; +} + static inline int64_t align_offset(int64_t offset, int n) { offset = (offset + n - 1) & ~(n - 1); @@ -182,7 +188,8 @@ int qcow2_update_snapshot_refcount(BlockDriverState *bs, int qcow2_check_refcounts(BlockDriverState *bs); /* qcow2-cluster.c functions */ -int qcow2_grow_l1_table(BlockDriverState *bs, int min_size); +int qcow2_grow_l1_vm_state(BlockDriverState *bs, int min_size); +int qcow2_grow_l1_image_data(BlockDriverState *bs, int new_l1_size); void qcow2_l2_cache_reset(BlockDriverState *bs); int qcow2_decompress_cluster(BlockDriverState *bs, uint64_t cluster_offset); void qcow2_encrypt_sectors(BDRVQcowState *s, int64_t sector_num, -- 1.7.0