The following changes since commit bb9bf94b3e8926553290bc9a7cb84315af422086: Merge remote-tracking branch 'remotes/ehabkost/tags/machine-next-pull-request' into staging (2018-12-11 19:18:58 +0000) are available in the Git repository at: git://repo.or.cz/qemu/kevin.git tags/for-upstream for you to fetch changes up to 14a8882c0d6660684a0ec7e90cc75d4f38b07f0d: iotests: make 235 work on s390 (and others) (2018-12-12 12:22:31 +0100) ---------------------------------------------------------------- Block layer patches: - qcow2: Decompression worker threads - dmg: lzfse compression support - file-posix: Simplify delegation to worker thread - Don't pass flags to bdrv_reopen_queue() - iotests: make 235 work on s390 (and others) ---------------------------------------------------------------- Alberto Garcia (15): block: Add bdrv_reopen_set_read_only() block: Use bdrv_reopen_set_read_only() in bdrv_backing_update_filename() block: Use bdrv_reopen_set_read_only() in commit_start/complete() block: Use bdrv_reopen_set_read_only() in bdrv_commit() block: Use bdrv_reopen_set_read_only() in stream_start/complete() block: Use bdrv_reopen_set_read_only() in qmp_change_backing_file() block: Use bdrv_reopen_set_read_only() in external_snapshot_commit() block: Use bdrv_reopen_set_read_only() in the mirror driver block: Drop bdrv_reopen() qemu-io: Put flag changes in the options QDict in reopen_f() block: Clean up reopen_backing_file() in block/replication.c block: Remove flags parameter from bdrv_reopen_queue() block: Stop passing flags to bdrv_reopen_queue_child() block: Remove assertions from update_flags_from_options() block: Assert that flags are up-to-date in bdrv_reopen_prepare() Christian Borntraeger (1): iotests: make 235 work on s390 (and others) Julio Faracco (4): block: adding lzfse decompressing support as a module. configure: adding support to lzfse library. dmg: including dmg-lzfse module inside dmg block driver. dmg: exchanging hardcoded dmg UDIF block types to enum. Kevin Wolf (12): file-posix: Reorganise RawPosixAIOData file-posix: Factor out raw_thread_pool_submit() file-posix: Avoid aio_worker() for QEMU_AIO_TRUNCATE file-posix: Avoid aio_worker() for QEMU_AIO_COPY_RANGE file-posix: Avoid aio_worker() for QEMU_AIO_WRITE_ZEROES file-posix: Avoid aio_worker() for QEMU_AIO_DISCARD file-posix: Avoid aio_worker() for QEMU_AIO_FLUSH file-posix: Move read/write operation logic out of aio_worker() file-posix: Avoid aio_worker() for QEMU_AIO_READ/WRITE file-posix: Remove paio_submit_co() file-posix: Switch to .bdrv_co_ioctl file-posix: Avoid aio_worker() for QEMU_AIO_IOCTL Vladimir Sementsov-Ogievskiy (9): block/replication: drop extra synchronization block/backup: drop unused synchronization interface qcow2: use Z_OK instead of 0 for deflateInit2 return code check qcow2: make more generic interface for qcow2_compress qcow2: move decompression from qcow2-cluster.c to qcow2.c qcow2: refactor decompress_buffer qcow2: use byte-based read in qcow2_decompress_cluster qcow2: aio support for compressed cluster read qcow2: do decompression in threads configure | 31 ++++ block/dmg.h | 3 + block/qcow2.h | 4 - include/block/block.h | 6 +- include/block/block_backup.h | 13 -- include/scsi/pr-manager.h | 8 +- block.c | 89 +++++----- block/backup.c | 38 +---- block/commit.c | 23 +-- block/dmg-lzfse.c | 49 ++++++ block/dmg.c | 65 ++++++-- block/file-posix.c | 380 ++++++++++++++++++++++--------------------- block/mirror.c | 19 ++- block/qcow2-cluster.c | 70 -------- block/qcow2.c | 170 ++++++++++++++++--- block/replication.c | 67 +++----- block/stream.c | 20 +-- blockdev.c | 11 +- qemu-io-cmds.c | 29 +++- scsi/pr-manager.c | 21 +-- block/Makefile.objs | 2 + scsi/trace-events | 2 +- tests/qemu-iotests/133 | 18 ++ tests/qemu-iotests/133.out | 15 ++ tests/qemu-iotests/235 | 4 +- 25 files changed, 658 insertions(+), 499 deletions(-) create mode 100644 block/dmg-lzfse.c From: Julio Faracco <jcfaracco@gmail.com> QEMU dmg support includes zlib and bzip2, but it does not contains lzfse support. This commit adds the source file to extend compression support for new DMGs. Signed-off-by: Julio Faracco <jcfaracco@gmail.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/dmg-lzfse.c | 49 +++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 49 insertions(+) create mode 100644 block/dmg-lzfse.c diff --git a/block/dmg-lzfse.c b/block/dmg-lzfse.c new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/block/dmg-lzfse.c @@ -XXX,XX +XXX,XX @@ +/* + * DMG lzfse uncompression + * + * Copyright (c) 2018 Julio Cesar Faracco + * + * Permission is hereby granted, free of charge, to any person obtaining a copy + * of this software and associated documentation files (the "Software"), to deal + * in the Software without restriction, including without limitation the rights + * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell + * copies of the Software, and to permit persons to whom the Software is + * furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN + * THE SOFTWARE. + */ +#include "qemu/osdep.h" +#include "qemu-common.h" +#include "dmg.h" +#include <lzfse.h> + +static int dmg_uncompress_lzfse_do(char *next_in, unsigned int avail_in, + char *next_out, unsigned int avail_out) +{ + size_t out_size = lzfse_decode_buffer((uint8_t *) next_out, avail_out, + (uint8_t *) next_in, avail_in, + NULL); + + /* We need to decode the single chunk only. */ + /* So, out_size == avail_out is not an error here. */ + if (out_size > 0) { + return out_size; + } + return -1; +} + +__attribute__((constructor)) +static void dmg_lzfse_init(void) +{ + assert(!dmg_uncompress_lzfse); + dmg_uncompress_lzfse = dmg_uncompress_lzfse_do; +} -- 2.19.2 From: Julio Faracco <jcfaracco@gmail.com> This commit includes the support to lzfse opensource library. With this library dmg block driver can decompress images with this type of compression inside. Signed-off-by: Julio Faracco <jcfaracco@gmail.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- configure | 31 +++++++++++++++++++++++++++++++ block/Makefile.objs | 2 ++ 2 files changed, 33 insertions(+) diff --git a/configure b/configure index XXXXXXX..XXXXXXX 100755 --- a/configure +++ b/configure @@ -XXX,XX +XXX,XX @@ capstone="" lzo="" snappy="" bzip2="" +lzfse="" guest_agent="" guest_agent_with_vss="no" guest_agent_ntddscsi="no" @@ -XXX,XX +XXX,XX @@ for opt do ;; --enable-bzip2) bzip2="yes" ;; + --enable-lzfse) lzfse="yes" + ;; + --disable-lzfse) lzfse="no" + ;; --enable-guest-agent) guest_agent="yes" ;; --disable-guest-agent) guest_agent="no" @@ -XXX,XX +XXX,XX @@ disabled with --disable-FEATURE, default is enabled if available: snappy support of snappy compression library bzip2 support of bzip2 compression library (for reading bzip2-compressed dmg images) + lzfse support of lzfse compression library + (for reading lzfse-compressed dmg images) seccomp seccomp support coroutine-pool coroutine freelist (better performance) glusterfs GlusterFS backend @@ -XXX,XX +XXX,XX @@ EOF fi fi +########################################## +# lzfse check + +if test "$lzfse" != "no" ; then + cat > $TMPC << EOF +#include <lzfse.h> +int main(void) { lzfse_decode_scratch_size(); return 0; } +EOF + if compile_prog "" "-llzfse" ; then + lzfse="yes" + else + if test "$lzfse" = "yes"; then + feature_not_found "lzfse" "Install lzfse devel" + fi + lzfse="no" + fi +fi + ########################################## # libseccomp check @@ -XXX,XX +XXX,XX @@ echo "Live block migration $live_block_migration" echo "lzo support $lzo" echo "snappy support $snappy" echo "bzip2 support $bzip2" +echo "lzfse support $lzfse" echo "NUMA host support $numa" echo "libxml2 $libxml2" echo "tcmalloc support $tcmalloc" @@ -XXX,XX +XXX,XX @@ if test "$bzip2" = "yes" ; then echo "BZIP2_LIBS=-lbz2" >> $config_host_mak fi +if test "$lzfse" = "yes" ; then + echo "CONFIG_LZFSE=y" >> $config_host_mak + echo "LZFSE_LIBS=-llzfse" >> $config_host_mak +fi + if test "$libiscsi" = "yes" ; then echo "CONFIG_LIBISCSI=m" >> $config_host_mak echo "LIBISCSI_CFLAGS=$libiscsi_cflags" >> $config_host_mak diff --git a/block/Makefile.objs b/block/Makefile.objs index XXXXXXX..XXXXXXX 100644 --- a/block/Makefile.objs +++ b/block/Makefile.objs @@ -XXX,XX +XXX,XX @@ ssh.o-libs := $(LIBSSH2_LIBS) block-obj-dmg-bz2-$(CONFIG_BZIP2) += dmg-bz2.o block-obj-$(if $(CONFIG_DMG),m,n) += $(block-obj-dmg-bz2-y) dmg-bz2.o-libs := $(BZIP2_LIBS) +block-obj-$(if $(CONFIG_LZFSE),m,n) += dmg-lzfse.o +dmg-lzfse.o-libs := $(LZFSE_LIBS) qcow.o-libs := -lz linux-aio.o-libs := -laio parallels.o-cflags := $(LIBXML2_CFLAGS) -- 2.19.2 From: Julio Faracco <jcfaracco@gmail.com> This commit includes the support to new module dmg-lzfse into dmg block driver. It includes the support for block type ULFO (0x80000007). Signed-off-by: Julio Faracco <jcfaracco@gmail.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/dmg.h | 3 +++ block/dmg.c | 28 ++++++++++++++++++++++++++++ 2 files changed, 31 insertions(+) diff --git a/block/dmg.h b/block/dmg.h index XXXXXXX..XXXXXXX 100644 --- a/block/dmg.h +++ b/block/dmg.h @@ -XXX,XX +XXX,XX @@ typedef struct BDRVDMGState { extern int (*dmg_uncompress_bz2)(char *next_in, unsigned int avail_in, char *next_out, unsigned int avail_out); +extern int (*dmg_uncompress_lzfse)(char *next_in, unsigned int avail_in, + char *next_out, unsigned int avail_out); + #endif diff --git a/block/dmg.c b/block/dmg.c index XXXXXXX..XXXXXXX 100644 --- a/block/dmg.c +++ b/block/dmg.c @@ -XXX,XX +XXX,XX @@ int (*dmg_uncompress_bz2)(char *next_in, unsigned int avail_in, char *next_out, unsigned int avail_out); +int (*dmg_uncompress_lzfse)(char *next_in, unsigned int avail_in, + char *next_out, unsigned int avail_out); + enum { /* Limit chunk sizes to prevent unreasonable amounts of memory being used * or truncating when converting to 32-bit types @@ -XXX,XX +XXX,XX @@ static void update_max_chunk_size(BDRVDMGState *s, uint32_t chunk, switch (s->types[chunk]) { case 0x80000005: /* zlib compressed */ case 0x80000006: /* bzip2 compressed */ + case 0x80000007: /* lzfse compressed */ compressed_size = s->lengths[chunk]; uncompressed_sectors = s->sectorcounts[chunk]; break; @@ -XXX,XX +XXX,XX @@ static bool dmg_is_known_block_type(uint32_t entry_type) return true; case 0x80000006: /* bzip2 */ return !!dmg_uncompress_bz2; + case 0x80000007: /* lzfse */ + return !!dmg_uncompress_lzfse; default: return false; } @@ -XXX,XX +XXX,XX @@ static int dmg_open(BlockDriverState *bs, QDict *options, int flags, } block_module_load_one("dmg-bz2"); + block_module_load_one("dmg-lzfse"); s->n_chunks = 0; s->offsets = s->lengths = s->sectors = s->sectorcounts = NULL; @@ -XXX,XX +XXX,XX @@ static inline int dmg_read_chunk(BlockDriverState *bs, uint64_t sector_num) return ret; } break; + case 0x80000007: + if (!dmg_uncompress_lzfse) { + break; + } + /* we need to buffer, because only the chunk as whole can be + * inflated. */ + ret = bdrv_pread(bs->file, s->offsets[chunk], + s->compressed_chunk, s->lengths[chunk]); + if (ret != s->lengths[chunk]) { + return -1; + } + + ret = dmg_uncompress_lzfse((char *)s->compressed_chunk, + (unsigned int) s->lengths[chunk], + (char *)s->uncompressed_chunk, + (unsigned int) + (512 * s->sectorcounts[chunk])); + if (ret < 0) { + return ret; + } + break; case 1: /* copy */ ret = bdrv_pread(bs->file, s->offsets[chunk], s->uncompressed_chunk, s->lengths[chunk]); -- 2.19.2 From: Julio Faracco <jcfaracco@gmail.com> This change is better to understand what kind of block type is being handled by the code. Using a syntax similar to the DMG documentation is easier than tracking all hex values assigned to a block type. Signed-off-by: Julio Faracco <jcfaracco@gmail.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/dmg.c | 43 ++++++++++++++++++++++++++++--------------- 1 file changed, 28 insertions(+), 15 deletions(-) diff --git a/block/dmg.c b/block/dmg.c index XXXXXXX..XXXXXXX 100644 --- a/block/dmg.c +++ b/block/dmg.c @@ -XXX,XX +XXX,XX @@ enum { DMG_SECTORCOUNTS_MAX = DMG_LENGTHS_MAX / 512, }; +enum { + /* DMG Block Type */ + UDZE = 0, /* Zeroes */ + UDRW, /* RAW type */ + UDIG, /* Ignore */ + UDCO = 0x80000004, + UDZO, + UDBZ, + ULFO, + UDCM = 0x7ffffffe, /* Comments */ + UDLE /* Last Entry */ +}; + static int dmg_probe(const uint8_t *buf, int buf_size, const char *filename) { int len; @@ -XXX,XX +XXX,XX @@ static void update_max_chunk_size(BDRVDMGState *s, uint32_t chunk, uint32_t uncompressed_sectors = 0; switch (s->types[chunk]) { - case 0x80000005: /* zlib compressed */ - case 0x80000006: /* bzip2 compressed */ - case 0x80000007: /* lzfse compressed */ + case UDZO: /* zlib compressed */ + case UDBZ: /* bzip2 compressed */ + case ULFO: /* lzfse compressed */ compressed_size = s->lengths[chunk]; uncompressed_sectors = s->sectorcounts[chunk]; break; - case 1: /* copy */ + case UDRW: /* copy */ uncompressed_sectors = DIV_ROUND_UP(s->lengths[chunk], 512); break; - case 2: /* zero */ + case UDIG: /* zero */ /* as the all-zeroes block may be large, it is treated specially: the * sector is not copied from a large buffer, a simple memset is used * instead. Therefore uncompressed_sectors does not need to be set. */ @@ -XXX,XX +XXX,XX @@ typedef struct DmgHeaderState { static bool dmg_is_known_block_type(uint32_t entry_type) { switch (entry_type) { - case 0x00000001: /* uncompressed */ - case 0x00000002: /* zeroes */ - case 0x80000005: /* zlib */ + case UDRW: /* uncompressed */ + case UDIG: /* zeroes */ + case UDZO: /* zlib */ return true; - case 0x80000006: /* bzip2 */ + case UDBZ: /* bzip2 */ return !!dmg_uncompress_bz2; - case 0x80000007: /* lzfse */ + case ULFO: /* lzfse */ return !!dmg_uncompress_lzfse; default: return false; @@ -XXX,XX +XXX,XX @@ static inline int dmg_read_chunk(BlockDriverState *bs, uint64_t sector_num) s->current_chunk = s->n_chunks; switch (s->types[chunk]) { /* block entry type */ - case 0x80000005: { /* zlib compressed */ + case UDZO: { /* zlib compressed */ /* we need to buffer, because only the chunk as whole can be * inflated. */ ret = bdrv_pread(bs->file, s->offsets[chunk], @@ -XXX,XX +XXX,XX @@ static inline int dmg_read_chunk(BlockDriverState *bs, uint64_t sector_num) return -1; } break; } - case 0x80000006: /* bzip2 compressed */ + case UDBZ: /* bzip2 compressed */ if (!dmg_uncompress_bz2) { break; } @@ -XXX,XX +XXX,XX @@ static inline int dmg_read_chunk(BlockDriverState *bs, uint64_t sector_num) return ret; } break; - case 0x80000007: + case ULFO: if (!dmg_uncompress_lzfse) { break; } @@ -XXX,XX +XXX,XX @@ static inline int dmg_read_chunk(BlockDriverState *bs, uint64_t sector_num) return ret; } break; - case 1: /* copy */ + case UDRW: /* copy */ ret = bdrv_pread(bs->file, s->offsets[chunk], s->uncompressed_chunk, s->lengths[chunk]); if (ret != s->lengths[chunk]) { return -1; } break; - case 2: /* zero */ + case UDIG: /* zero */ /* see dmg_read, it is treated specially. No buffer needs to be * pre-filled, the zeroes can be set directly. */ break; -- 2.19.2 From: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> After commit f8d59dfb40 "block/backup: fix fleecing scheme: use serialized writes" fleecing (specifically reading from backup target, when backup source is in backing chain of backup target) is safe, because all backup-job writes to target are serialized. Therefore we don't need additional synchronization for these reads. Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/replication.c | 24 +----------------------- 1 file changed, 1 insertion(+), 23 deletions(-) diff --git a/block/replication.c b/block/replication.c index XXXXXXX..XXXXXXX 100644 --- a/block/replication.c +++ b/block/replication.c @@ -XXX,XX +XXX,XX @@ static coroutine_fn int replication_co_readv(BlockDriverState *bs, QEMUIOVector *qiov) { BDRVReplicationState *s = bs->opaque; - BdrvChild *child = s->secondary_disk; - BlockJob *job = NULL; - CowRequest req; int ret; if (s->mode == REPLICATION_MODE_PRIMARY) { @@ -XXX,XX +XXX,XX @@ static coroutine_fn int replication_co_readv(BlockDriverState *bs, return ret; } - if (child && child->bs) { - job = child->bs->job; - } - - if (job) { - uint64_t remaining_bytes = remaining_sectors * BDRV_SECTOR_SIZE; - - backup_wait_for_overlapping_requests(child->bs->job, - sector_num * BDRV_SECTOR_SIZE, - remaining_bytes); - backup_cow_request_begin(&req, child->bs->job, - sector_num * BDRV_SECTOR_SIZE, - remaining_bytes); - ret = bdrv_co_preadv(bs->file, sector_num * BDRV_SECTOR_SIZE, - remaining_bytes, qiov, 0); - backup_cow_request_end(&req); - goto out; - } - ret = bdrv_co_preadv(bs->file, sector_num * BDRV_SECTOR_SIZE, remaining_sectors * BDRV_SECTOR_SIZE, qiov, 0); -out: + return replication_return_value(s, ret); } -- 2.19.2 From: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- include/block/block_backup.h | 13 ------------ block/backup.c | 38 +++++++----------------------------- 2 files changed, 7 insertions(+), 44 deletions(-) diff --git a/include/block/block_backup.h b/include/block/block_backup.h index XXXXXXX..XXXXXXX 100644 --- a/include/block/block_backup.h +++ b/include/block/block_backup.h @@ -XXX,XX +XXX,XX @@ #include "block/block_int.h" -typedef struct CowRequest { - int64_t start_byte; - int64_t end_byte; - QLIST_ENTRY(CowRequest) list; - CoQueue wait_queue; /* coroutines blocked on this request */ -} CowRequest; - -void backup_wait_for_overlapping_requests(BlockJob *job, int64_t offset, - uint64_t bytes); -void backup_cow_request_begin(CowRequest *req, BlockJob *job, - int64_t offset, uint64_t bytes); -void backup_cow_request_end(CowRequest *req); - void backup_do_checkpoint(BlockJob *job, Error **errp); #endif diff --git a/block/backup.c b/block/backup.c index XXXXXXX..XXXXXXX 100644 --- a/block/backup.c +++ b/block/backup.c @@ -XXX,XX +XXX,XX @@ #define BACKUP_CLUSTER_SIZE_DEFAULT (1 << 16) +typedef struct CowRequest { + int64_t start_byte; + int64_t end_byte; + QLIST_ENTRY(CowRequest) list; + CoQueue wait_queue; /* coroutines blocked on this request */ +} CowRequest; + typedef struct BackupBlockJob { BlockJob common; BlockBackend *target; @@ -XXX,XX +XXX,XX @@ void backup_do_checkpoint(BlockJob *job, Error **errp) hbitmap_set(backup_job->copy_bitmap, 0, len); } -void backup_wait_for_overlapping_requests(BlockJob *job, int64_t offset, - uint64_t bytes) -{ - BackupBlockJob *backup_job = container_of(job, BackupBlockJob, common); - int64_t start, end; - - assert(block_job_driver(job) == &backup_job_driver); - - start = QEMU_ALIGN_DOWN(offset, backup_job->cluster_size); - end = QEMU_ALIGN_UP(offset + bytes, backup_job->cluster_size); - wait_for_overlapping_requests(backup_job, start, end); -} - -void backup_cow_request_begin(CowRequest *req, BlockJob *job, - int64_t offset, uint64_t bytes) -{ - BackupBlockJob *backup_job = container_of(job, BackupBlockJob, common); - int64_t start, end; - - assert(block_job_driver(job) == &backup_job_driver); - - start = QEMU_ALIGN_DOWN(offset, backup_job->cluster_size); - end = QEMU_ALIGN_UP(offset + bytes, backup_job->cluster_size); - cow_request_begin(req, backup_job, start, end); -} - -void backup_cow_request_end(CowRequest *req) -{ - cow_request_end(req); -} - static void backup_drain(BlockJob *job) { BackupBlockJob *s = container_of(job, BackupBlockJob, common); -- 2.19.2 From: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> Use appropriate macro, corresponding to deflateInit2 spec. Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> Reviewed-by: Alberto Garcia <berto@igalia.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/qcow2.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/block/qcow2.c b/block/qcow2.c index XXXXXXX..XXXXXXX 100644 --- a/block/qcow2.c +++ b/block/qcow2.c @@ -XXX,XX +XXX,XX @@ static ssize_t qcow2_compress(void *dest, const void *src, size_t size) memset(&strm, 0, sizeof(strm)); ret = deflateInit2(&strm, Z_DEFAULT_COMPRESSION, Z_DEFLATED, -12, 9, Z_DEFAULT_STRATEGY); - if (ret != 0) { + if (ret != Z_OK) { return -2; } -- 2.19.2 From: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> Give explicit size both for source and destination buffers, to make it similar with decompression path and than cleanly reuse parameter structure for decompression threads. Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> Reviewed-by: Alberto Garcia <berto@igalia.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/qcow2.c | 30 ++++++++++++++++++------------ 1 file changed, 18 insertions(+), 12 deletions(-) diff --git a/block/qcow2.c b/block/qcow2.c index XXXXXXX..XXXXXXX 100644 --- a/block/qcow2.c +++ b/block/qcow2.c @@ -XXX,XX +XXX,XX @@ fail: /* * qcow2_compress() * - * @dest - destination buffer, at least of @size-1 bytes - * @src - source buffer, @size bytes + * @dest - destination buffer, @dest_size bytes + * @src - source buffer, @src_size bytes * * Returns: compressed size on success - * -1 if compression is inefficient + * -1 destination buffer is not enough to store compressed data * -2 on any other error */ -static ssize_t qcow2_compress(void *dest, const void *src, size_t size) +static ssize_t qcow2_compress(void *dest, size_t dest_size, + const void *src, size_t src_size) { ssize_t ret; z_stream strm; @@ -XXX,XX +XXX,XX @@ static ssize_t qcow2_compress(void *dest, const void *src, size_t size) /* strm.next_in is not const in old zlib versions, such as those used on * OpenBSD/NetBSD, so cast the const away */ - strm.avail_in = size; + strm.avail_in = src_size; strm.next_in = (void *) src; - strm.avail_out = size - 1; + strm.avail_out = dest_size; strm.next_out = dest; ret = deflate(&strm, Z_FINISH); if (ret == Z_STREAM_END) { - ret = size - 1 - strm.avail_out; + ret = dest_size - strm.avail_out; } else { ret = (ret == Z_OK ? -1 : -2); } @@ -XXX,XX +XXX,XX @@ static ssize_t qcow2_compress(void *dest, const void *src, size_t size) typedef struct Qcow2CompressData { void *dest; + size_t dest_size; const void *src; - size_t size; + size_t src_size; ssize_t ret; } Qcow2CompressData; @@ -XXX,XX +XXX,XX @@ static int qcow2_compress_pool_func(void *opaque) { Qcow2CompressData *data = opaque; - data->ret = qcow2_compress(data->dest, data->src, data->size); + data->ret = qcow2_compress(data->dest, data->dest_size, + data->src, data->src_size); return 0; } @@ -XXX,XX +XXX,XX @@ static void qcow2_compress_complete(void *opaque, int ret) /* See qcow2_compress definition for parameters description */ static ssize_t qcow2_co_compress(BlockDriverState *bs, - void *dest, const void *src, size_t size) + void *dest, size_t dest_size, + const void *src, size_t src_size) { BDRVQcow2State *s = bs->opaque; BlockAIOCB *acb; ThreadPool *pool = aio_get_thread_pool(bdrv_get_aio_context(bs)); Qcow2CompressData arg = { .dest = dest, + .dest_size = dest_size, .src = src, - .size = size, + .src_size = src_size, }; while (s->nb_compress_threads >= MAX_COMPRESS_THREADS) { @@ -XXX,XX +XXX,XX @@ qcow2_co_pwritev_compressed(BlockDriverState *bs, uint64_t offset, out_buf = g_malloc(s->cluster_size); - out_len = qcow2_co_compress(bs, out_buf, buf, s->cluster_size); + out_len = qcow2_co_compress(bs, out_buf, s->cluster_size - 1, + buf, s->cluster_size); if (out_len == -2) { ret = -EINVAL; goto fail; -- 2.19.2 From: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> Compression is done in threads in qcow2.c. We want to do decompression in the same way, so, firstly, move it to the same file. The only change is braces around if-body in decompress_buffer, to satisfy checkpatch. Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> Reviewed-by: Alberto Garcia <berto@igalia.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/qcow2-cluster.c | 70 ------------------------------------------ block/qcow2.c | 71 +++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 71 insertions(+), 70 deletions(-) diff --git a/block/qcow2-cluster.c b/block/qcow2-cluster.c index XXXXXXX..XXXXXXX 100644 --- a/block/qcow2-cluster.c +++ b/block/qcow2-cluster.c @@ -XXX,XX +XXX,XX @@ again: return 0; } -static int decompress_buffer(uint8_t *out_buf, int out_buf_size, - const uint8_t *buf, int buf_size) -{ - z_stream strm1, *strm = &strm1; - int ret, out_len; - - memset(strm, 0, sizeof(*strm)); - - strm->next_in = (uint8_t *)buf; - strm->avail_in = buf_size; - strm->next_out = out_buf; - strm->avail_out = out_buf_size; - - ret = inflateInit2(strm, -12); - if (ret != Z_OK) - return -1; - ret = inflate(strm, Z_FINISH); - out_len = strm->next_out - out_buf; - if ((ret != Z_STREAM_END && ret != Z_BUF_ERROR) || - out_len != out_buf_size) { - inflateEnd(strm); - return -1; - } - inflateEnd(strm); - return 0; -} - -int qcow2_decompress_cluster(BlockDriverState *bs, uint64_t cluster_offset) -{ - BDRVQcow2State *s = bs->opaque; - int ret, csize, nb_csectors, sector_offset; - uint64_t coffset; - - coffset = cluster_offset & s->cluster_offset_mask; - if (s->cluster_cache_offset != coffset) { - nb_csectors = ((cluster_offset >> s->csize_shift) & s->csize_mask) + 1; - sector_offset = coffset & 511; - csize = nb_csectors * 512 - sector_offset; - - /* 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); - } - - BLKDBG_EVENT(bs->file, BLKDBG_READ_COMPRESSED); - ret = bdrv_read(bs->file, coffset >> 9, s->cluster_data, - nb_csectors); - if (ret < 0) { - return ret; - } - if (decompress_buffer(s->cluster_cache, s->cluster_size, - s->cluster_data + sector_offset, csize) < 0) { - return -EIO; - } - s->cluster_cache_offset = coffset; - } - return 0; -} - /* * This discards as many clusters of nb_clusters as possible at once (i.e. * all clusters in the same L2 slice) and returns the number of discarded diff --git a/block/qcow2.c b/block/qcow2.c index XXXXXXX..XXXXXXX 100644 --- a/block/qcow2.c +++ b/block/qcow2.c @@ -XXX,XX +XXX,XX @@ static ssize_t qcow2_compress(void *dest, size_t dest_size, return ret; } +static int decompress_buffer(uint8_t *out_buf, int out_buf_size, + const uint8_t *buf, int buf_size) +{ + z_stream strm1, *strm = &strm1; + int ret, out_len; + + memset(strm, 0, sizeof(*strm)); + + strm->next_in = (uint8_t *)buf; + strm->avail_in = buf_size; + strm->next_out = out_buf; + strm->avail_out = out_buf_size; + + ret = inflateInit2(strm, -12); + if (ret != Z_OK) { + return -1; + } + ret = inflate(strm, Z_FINISH); + out_len = strm->next_out - out_buf; + if ((ret != Z_STREAM_END && ret != Z_BUF_ERROR) || + out_len != out_buf_size) { + inflateEnd(strm); + return -1; + } + inflateEnd(strm); + return 0; +} + #define MAX_COMPRESS_THREADS 4 typedef struct Qcow2CompressData { @@ -XXX,XX +XXX,XX @@ fail: return ret; } +int qcow2_decompress_cluster(BlockDriverState *bs, uint64_t cluster_offset) +{ + BDRVQcow2State *s = bs->opaque; + int ret, csize, nb_csectors, sector_offset; + uint64_t coffset; + + coffset = cluster_offset & s->cluster_offset_mask; + if (s->cluster_cache_offset != coffset) { + nb_csectors = ((cluster_offset >> s->csize_shift) & s->csize_mask) + 1; + sector_offset = coffset & 511; + csize = nb_csectors * 512 - sector_offset; + + /* 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); + } + + BLKDBG_EVENT(bs->file, BLKDBG_READ_COMPRESSED); + ret = bdrv_read(bs->file, coffset >> 9, s->cluster_data, + nb_csectors); + if (ret < 0) { + return ret; + } + if (decompress_buffer(s->cluster_cache, s->cluster_size, + s->cluster_data + sector_offset, csize) < 0) { + return -EIO; + } + s->cluster_cache_offset = coffset; + } + return 0; +} + static int make_completely_empty(BlockDriverState *bs) { BDRVQcow2State *s = bs->opaque; -- 2.19.2 From: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> - make it look more like a pair of qcow2_compress - rename the function and its parameters - drop extra out_len variable, check filling of output buffer by strm structure itself - fix code style - add some documentation Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> Reviewed-by: Alberto Garcia <berto@igalia.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/qcow2.c | 56 ++++++++++++++++++++++++++++++++------------------- 1 file changed, 35 insertions(+), 21 deletions(-) diff --git a/block/qcow2.c b/block/qcow2.c index XXXXXXX..XXXXXXX 100644 --- a/block/qcow2.c +++ b/block/qcow2.c @@ -XXX,XX +XXX,XX @@ static ssize_t qcow2_compress(void *dest, size_t dest_size, return ret; } -static int decompress_buffer(uint8_t *out_buf, int out_buf_size, - const uint8_t *buf, int buf_size) +/* + * qcow2_decompress() + * + * Decompress some data (not more than @src_size bytes) to produce exactly + * @dest_size bytes. + * + * @dest - destination buffer, @dest_size bytes + * @src - source buffer, @src_size bytes + * + * Returns: 0 on success + * -1 on fail + */ +static ssize_t qcow2_decompress(void *dest, size_t dest_size, + const void *src, size_t src_size) { - z_stream strm1, *strm = &strm1; - int ret, out_len; - - memset(strm, 0, sizeof(*strm)); + int ret = 0; + z_stream strm; - strm->next_in = (uint8_t *)buf; - strm->avail_in = buf_size; - strm->next_out = out_buf; - strm->avail_out = out_buf_size; + memset(&strm, 0, sizeof(strm)); + strm.avail_in = src_size; + strm.next_in = (void *) src; + strm.avail_out = dest_size; + strm.next_out = dest; - ret = inflateInit2(strm, -12); + ret = inflateInit2(&strm, -12); if (ret != Z_OK) { return -1; } - ret = inflate(strm, Z_FINISH); - out_len = strm->next_out - out_buf; - if ((ret != Z_STREAM_END && ret != Z_BUF_ERROR) || - out_len != out_buf_size) { - inflateEnd(strm); - return -1; + + ret = inflate(&strm, Z_FINISH); + if ((ret != Z_STREAM_END && ret != Z_BUF_ERROR) || strm.avail_out != 0) { + /* We approve Z_BUF_ERROR because we need @dest buffer to be filled, but + * @src buffer may be processed partly (because in qcow2 we know size of + * compressed data with precision of one sector) */ + ret = -1; } - inflateEnd(strm); - return 0; + + inflateEnd(&strm); + + return ret; } #define MAX_COMPRESS_THREADS 4 @@ -XXX,XX +XXX,XX @@ int qcow2_decompress_cluster(BlockDriverState *bs, uint64_t cluster_offset) if (ret < 0) { return ret; } - if (decompress_buffer(s->cluster_cache, s->cluster_size, - s->cluster_data + sector_offset, csize) < 0) { + if (qcow2_decompress(s->cluster_cache, s->cluster_size, + s->cluster_data + sector_offset, csize) < 0) { return -EIO; } s->cluster_cache_offset = coffset; -- 2.19.2 From: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> We are gradually moving away from sector-based interfaces, towards byte-based. Get rid of it here too. Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/qcow2.c | 19 ++++++++++++------- 1 file changed, 12 insertions(+), 7 deletions(-) diff --git a/block/qcow2.c b/block/qcow2.c index XXXXXXX..XXXXXXX 100644 --- a/block/qcow2.c +++ b/block/qcow2.c @@ -XXX,XX +XXX,XX @@ fail: return ret; } -int qcow2_decompress_cluster(BlockDriverState *bs, uint64_t cluster_offset) +int coroutine_fn +qcow2_decompress_cluster(BlockDriverState *bs, uint64_t cluster_offset) { BDRVQcow2State *s = bs->opaque; - int ret, csize, nb_csectors, sector_offset; + int ret, csize, nb_csectors; uint64_t coffset; + struct iovec iov; + QEMUIOVector local_qiov; coffset = cluster_offset & s->cluster_offset_mask; if (s->cluster_cache_offset != coffset) { nb_csectors = ((cluster_offset >> s->csize_shift) & s->csize_mask) + 1; - sector_offset = coffset & 511; - csize = nb_csectors * 512 - sector_offset; + 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 @@ -XXX,XX +XXX,XX @@ int qcow2_decompress_cluster(BlockDriverState *bs, uint64_t cluster_offset) 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_read(bs->file, coffset >> 9, s->cluster_data, - nb_csectors); + 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 + sector_offset, csize) < 0) { + s->cluster_data, csize) < 0) { return -EIO; } s->cluster_cache_offset = coffset; -- 2.19.2 From: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> 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 <vsementsov@virtuozzo.com> Reviewed-by: Alberto Garcia <berto@igalia.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/qcow2.h | 4 --- block/qcow2.c | 96 ++++++++++++++++++++++++++------------------------- 2 files changed, 49 insertions(+), 51 deletions(-) diff --git a/block/qcow2.h b/block/qcow2.h index XXXXXXX..XXXXXXX 100644 --- a/block/qcow2.h +++ b/block/qcow2.h @@ -XXX,XX +XXX,XX @@ 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; @@ -XXX,XX +XXX,XX @@ 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); diff --git a/block/qcow2.c b/block/qcow2.c index XXXXXXX..XXXXXXX 100644 --- a/block/qcow2.c +++ b/block/qcow2.c @@ -XXX,XX +XXX,XX @@ 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; @@ -XXX,XX +XXX,XX @@ 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); @@ -XXX,XX +XXX,XX @@ 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: @@ -XXX,XX +XXX,XX @@ 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) { @@ -XXX,XX +XXX,XX @@ 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); } @@ -XXX,XX +XXX,XX @@ 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); @@ -XXX,XX +XXX,XX @@ 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); - } + buf = g_try_malloc(csize); + if (!buf) { + return -ENOMEM; + } + iov.iov_base = buf; + iov.iov_len = csize; + qemu_iovec_init_external(&local_qiov, &iov, 1); - iov.iov_base = s->cluster_data; - 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) { - return ret; - } - if (qcow2_decompress(s->cluster_cache, s->cluster_size, - s->cluster_data, csize) < 0) { - return -EIO; - } - s->cluster_cache_offset = coffset; + BLKDBG_EVENT(bs->file, BLKDBG_READ_COMPRESSED); + ret = bdrv_co_preadv(bs->file, coffset, csize, &local_qiov, 0); + if (ret < 0) { + goto fail; } - return 0; + + 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) -- 2.19.2 From: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> Do decompression in threads, like it is already done for compression. This improves asynchronous compressed reads performance. Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> Reviewed-by: Alberto Garcia <berto@igalia.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/qcow2.c | 34 +++++++++++++++++++++++++++------- 1 file changed, 27 insertions(+), 7 deletions(-) diff --git a/block/qcow2.c b/block/qcow2.c index XXXXXXX..XXXXXXX 100644 --- a/block/qcow2.c +++ b/block/qcow2.c @@ -XXX,XX +XXX,XX @@ static ssize_t qcow2_decompress(void *dest, size_t dest_size, #define MAX_COMPRESS_THREADS 4 +typedef ssize_t (*Qcow2CompressFunc)(void *dest, size_t dest_size, + const void *src, size_t src_size); typedef struct Qcow2CompressData { void *dest; size_t dest_size; const void *src; size_t src_size; ssize_t ret; + + Qcow2CompressFunc func; } Qcow2CompressData; static int qcow2_compress_pool_func(void *opaque) { Qcow2CompressData *data = opaque; - data->ret = qcow2_compress(data->dest, data->dest_size, - data->src, data->src_size); + data->ret = data->func(data->dest, data->dest_size, + data->src, data->src_size); return 0; } @@ -XXX,XX +XXX,XX @@ static void qcow2_compress_complete(void *opaque, int ret) qemu_coroutine_enter(opaque); } -/* See qcow2_compress definition for parameters description */ -static ssize_t qcow2_co_compress(BlockDriverState *bs, - void *dest, size_t dest_size, - const void *src, size_t src_size) +static ssize_t coroutine_fn +qcow2_co_do_compress(BlockDriverState *bs, void *dest, size_t dest_size, + const void *src, size_t src_size, Qcow2CompressFunc func) { BDRVQcow2State *s = bs->opaque; BlockAIOCB *acb; @@ -XXX,XX +XXX,XX @@ static ssize_t qcow2_co_compress(BlockDriverState *bs, .dest_size = dest_size, .src = src, .src_size = src_size, + .func = func, }; while (s->nb_compress_threads >= MAX_COMPRESS_THREADS) { @@ -XXX,XX +XXX,XX @@ static ssize_t qcow2_co_compress(BlockDriverState *bs, return arg.ret; } +static ssize_t coroutine_fn +qcow2_co_compress(BlockDriverState *bs, void *dest, size_t dest_size, + const void *src, size_t src_size) +{ + return qcow2_co_do_compress(bs, dest, dest_size, src, src_size, + qcow2_compress); +} + +static ssize_t coroutine_fn +qcow2_co_decompress(BlockDriverState *bs, void *dest, size_t dest_size, + const void *src, size_t src_size) +{ + return qcow2_co_do_compress(bs, dest, dest_size, src, src_size, + qcow2_decompress); +} + /* XXX: put compressed sectors first, then all the cluster aligned tables to avoid losing bytes in alignment */ static coroutine_fn int @@ -XXX,XX +XXX,XX @@ qcow2_co_preadv_compressed(BlockDriverState *bs, goto fail; } - if (qcow2_decompress(out_buf, s->cluster_size, buf, csize) < 0) { + if (qcow2_co_decompress(bs, out_buf, s->cluster_size, buf, csize) < 0) { ret = -EIO; goto fail; } -- 2.19.2 RawPosixAIOData contains a lot of fields for several separate operations that are to be processed in a worker thread and that need different parameters. The struct is currently rather unorganised, with unions that cover some, but not all operations, and even one #define for field names instead of a union. Clean this up to have some common fields and a single union. As a side effect, on x86_64 the struct shrinks from 72 to 48 bytes. Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/file-posix.c | 89 +++++++++++++++++++++++++--------------------- 1 file changed, 49 insertions(+), 40 deletions(-) diff --git a/block/file-posix.c b/block/file-posix.c index XXXXXXX..XXXXXXX 100644 --- a/block/file-posix.c +++ b/block/file-posix.c @@ -XXX,XX +XXX,XX @@ static int64_t raw_getlength(BlockDriverState *bs); typedef struct RawPosixAIOData { BlockDriverState *bs; + int aio_type; int aio_fildes; - union { - struct iovec *aio_iov; - void *aio_ioctl_buf; - }; - int aio_niov; - uint64_t aio_nbytes; -#define aio_ioctl_cmd aio_nbytes /* for QEMU_AIO_IOCTL */ + off_t aio_offset; - int aio_type; + uint64_t aio_nbytes; + union { + struct { + struct iovec *iov; + int niov; + } io; + struct { + uint64_t cmd; + void *buf; + } ioctl; struct { int aio_fd2; off_t aio_offset2; - }; + } copy_range; struct { PreallocMode prealloc; Error **errp; - }; + } truncate; }; } RawPosixAIOData; @@ -XXX,XX +XXX,XX @@ static ssize_t handle_aiocb_ioctl(RawPosixAIOData *aiocb) { int ret; - ret = ioctl(aiocb->aio_fildes, aiocb->aio_ioctl_cmd, aiocb->aio_ioctl_buf); + ret = ioctl(aiocb->aio_fildes, aiocb->ioctl.cmd, aiocb->ioctl.buf); if (ret == -1) { return -errno; } @@ -XXX,XX +XXX,XX @@ static ssize_t handle_aiocb_rw_vector(RawPosixAIOData *aiocb) do { if (aiocb->aio_type & QEMU_AIO_WRITE) len = qemu_pwritev(aiocb->aio_fildes, - aiocb->aio_iov, - aiocb->aio_niov, + aiocb->io.iov, + aiocb->io.niov, aiocb->aio_offset); else len = qemu_preadv(aiocb->aio_fildes, - aiocb->aio_iov, - aiocb->aio_niov, + aiocb->io.iov, + aiocb->io.niov, aiocb->aio_offset); } while (len == -1 && errno == EINTR); @@ -XXX,XX +XXX,XX @@ static ssize_t handle_aiocb_rw(RawPosixAIOData *aiocb) * If there is just a single buffer, and it is properly aligned * we can just use plain pread/pwrite without any problems. */ - if (aiocb->aio_niov == 1) { - return handle_aiocb_rw_linear(aiocb, aiocb->aio_iov->iov_base); + if (aiocb->io.niov == 1) { + return handle_aiocb_rw_linear(aiocb, aiocb->io.iov->iov_base); } /* * We have more than one iovec, and all are properly aligned. @@ -XXX,XX +XXX,XX @@ static ssize_t handle_aiocb_rw(RawPosixAIOData *aiocb) char *p = buf; int i; - for (i = 0; i < aiocb->aio_niov; ++i) { - memcpy(p, aiocb->aio_iov[i].iov_base, aiocb->aio_iov[i].iov_len); - p += aiocb->aio_iov[i].iov_len; + for (i = 0; i < aiocb->io.niov; ++i) { + memcpy(p, aiocb->io.iov[i].iov_base, aiocb->io.iov[i].iov_len); + p += aiocb->io.iov[i].iov_len; } assert(p - buf == aiocb->aio_nbytes); } @@ -XXX,XX +XXX,XX @@ static ssize_t handle_aiocb_rw(RawPosixAIOData *aiocb) size_t count = aiocb->aio_nbytes, copy; int i; - for (i = 0; i < aiocb->aio_niov && count; ++i) { + for (i = 0; i < aiocb->io.niov && count; ++i) { copy = count; - if (copy > aiocb->aio_iov[i].iov_len) { - copy = aiocb->aio_iov[i].iov_len; + if (copy > aiocb->io.iov[i].iov_len) { + copy = aiocb->io.iov[i].iov_len; } - memcpy(aiocb->aio_iov[i].iov_base, p, copy); + memcpy(aiocb->io.iov[i].iov_base, p, copy); assert(count >= copy); p += copy; count -= copy; @@ -XXX,XX +XXX,XX @@ static ssize_t handle_aiocb_copy_range(RawPosixAIOData *aiocb) { uint64_t bytes = aiocb->aio_nbytes; off_t in_off = aiocb->aio_offset; - off_t out_off = aiocb->aio_offset2; + off_t out_off = aiocb->copy_range.aio_offset2; while (bytes) { ssize_t ret = copy_file_range(aiocb->aio_fildes, &in_off, - aiocb->aio_fd2, &out_off, + aiocb->copy_range.aio_fd2, &out_off, bytes, 0); trace_file_copy_file_range(aiocb->bs, aiocb->aio_fildes, in_off, - aiocb->aio_fd2, out_off, bytes, 0, ret); + aiocb->copy_range.aio_fd2, out_off, bytes, + 0, ret); if (ret == 0) { /* No progress (e.g. when beyond EOF), let the caller fall back to * buffer I/O. */ @@ -XXX,XX +XXX,XX @@ static int handle_aiocb_truncate(RawPosixAIOData *aiocb) struct stat st; int fd = aiocb->aio_fildes; int64_t offset = aiocb->aio_offset; - Error **errp = aiocb->errp; + PreallocMode prealloc = aiocb->truncate.prealloc; + Error **errp = aiocb->truncate.errp; if (fstat(fd, &st) < 0) { result = -errno; @@ -XXX,XX +XXX,XX @@ static int handle_aiocb_truncate(RawPosixAIOData *aiocb) } current_length = st.st_size; - if (current_length > offset && aiocb->prealloc != PREALLOC_MODE_OFF) { + if (current_length > offset && prealloc != PREALLOC_MODE_OFF) { error_setg(errp, "Cannot use preallocation for shrinking files"); return -ENOTSUP; } - switch (aiocb->prealloc) { + switch (prealloc) { #ifdef CONFIG_POSIX_FALLOCATE case PREALLOC_MODE_FALLOC: /* @@ -XXX,XX +XXX,XX @@ static int handle_aiocb_truncate(RawPosixAIOData *aiocb) default: result = -ENOTSUP; error_setg(errp, "Unsupported preallocation mode: %s", - PreallocMode_str(aiocb->prealloc)); + PreallocMode_str(prealloc)); return result; } @@ -XXX,XX +XXX,XX @@ static int aio_worker(void *arg) case QEMU_AIO_READ: ret = handle_aiocb_rw(aiocb); if (ret >= 0 && ret < aiocb->aio_nbytes) { - iov_memset(aiocb->aio_iov, aiocb->aio_niov, ret, + iov_memset(aiocb->io.iov, aiocb->io.niov, ret, 0, aiocb->aio_nbytes - ret); ret = aiocb->aio_nbytes; @@ -XXX,XX +XXX,XX @@ static int paio_submit_co_full(BlockDriverState *bs, int fd, acb->bs = bs; acb->aio_type = type; acb->aio_fildes = fd; - acb->aio_fd2 = fd2; - acb->aio_offset2 = offset2; acb->aio_nbytes = bytes; acb->aio_offset = offset; if (qiov) { - acb->aio_iov = qiov->iov; - acb->aio_niov = qiov->niov; + 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); @@ -XXX,XX +XXX,XX @@ raw_regular_truncate(BlockDriverState *bs, int fd, int64_t offset, .aio_fildes = fd, .aio_type = QEMU_AIO_TRUNCATE, .aio_offset = offset, - .prealloc = prealloc, - .errp = errp, + .truncate = { + .prealloc = prealloc, + .errp = errp, + }, }; /* @bs can be NULL, bdrv_get_aio_context() returns the main context then */ @@ -XXX,XX +XXX,XX @@ static BlockAIOCB *hdev_aio_ioctl(BlockDriverState *bs, acb->aio_type = QEMU_AIO_IOCTL; acb->aio_fildes = s->fd; acb->aio_offset = 0; - acb->aio_ioctl_buf = buf; - acb->aio_ioctl_cmd = req; + 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); } -- 2.19.2 Getting the thread pool of the AioContext of a block node and scheduling some work in it is an operation that is already done twice, and we'll get more instances. Factor it out into a separate function. Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/file-posix.c | 17 ++++++++++------- 1 file changed, 10 insertions(+), 7 deletions(-) diff --git a/block/file-posix.c b/block/file-posix.c index XXXXXXX..XXXXXXX 100644 --- a/block/file-posix.c +++ b/block/file-posix.c @@ -XXX,XX +XXX,XX @@ static int aio_worker(void *arg) return ret; } +static int coroutine_fn raw_thread_pool_submit(BlockDriverState *bs, + ThreadPoolFunc func, void *arg) +{ + /* @bs can be NULL, bdrv_get_aio_context() returns the main context then */ + ThreadPool *pool = aio_get_thread_pool(bdrv_get_aio_context(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); - ThreadPool *pool; acb->bs = bs; acb->aio_type = type; @@ -XXX,XX +XXX,XX @@ static int paio_submit_co_full(BlockDriverState *bs, int fd, } trace_file_paio_submit_co(offset, bytes, type); - pool = aio_get_thread_pool(bdrv_get_aio_context(bs)); - return thread_pool_submit_co(pool, aio_worker, acb); + return raw_thread_pool_submit(bs, aio_worker, acb); } static inline int paio_submit_co(BlockDriverState *bs, int fd, @@ -XXX,XX +XXX,XX @@ raw_regular_truncate(BlockDriverState *bs, int fd, int64_t offset, PreallocMode prealloc, Error **errp) { RawPosixAIOData *acb = g_new(RawPosixAIOData, 1); - ThreadPool *pool; *acb = (RawPosixAIOData) { .bs = bs, @@ -XXX,XX +XXX,XX @@ raw_regular_truncate(BlockDriverState *bs, int fd, int64_t offset, }, }; - /* @bs can be NULL, bdrv_get_aio_context() returns the main context then */ - pool = aio_get_thread_pool(bdrv_get_aio_context(bs)); - return thread_pool_submit_co(pool, aio_worker, acb); + return raw_thread_pool_submit(bs, aio_worker, acb); } static int coroutine_fn raw_co_truncate(BlockDriverState *bs, int64_t offset, -- 2.19.2 aio_worker() doesn't add anything interesting, it's only a useless indirection. Call the handler function directly instead. As we know that this handler function is only called from coroutine context and the coroutine stays around until the worker thread finishes, we can keep RawPosixAIOData on the stack. Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/file-posix.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/block/file-posix.c b/block/file-posix.c index XXXXXXX..XXXXXXX 100644 --- a/block/file-posix.c +++ b/block/file-posix.c @@ -XXX,XX +XXX,XX @@ 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; @@ -XXX,XX +XXX,XX @@ static int aio_worker(void *arg) ret = handle_aiocb_copy_range(aiocb); break; case QEMU_AIO_TRUNCATE: - ret = handle_aiocb_truncate(aiocb); - break; + g_assert_not_reached(); default: error_report("invalid aio request (0x%x)", aiocb->aio_type); ret = -EINVAL; @@ -XXX,XX +XXX,XX @@ 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, @@ -XXX,XX +XXX,XX @@ 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, -- 2.19.2 aio_worker() doesn't add anything interesting, it's only a useless indirection. Call the handler function directly instead. As we know that this handler function is only called from coroutine context and the coroutine stays around until the worker thread finishes, we can keep RawPosixAIOData on the stack. Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/file-posix.c | 22 +++++++++++++++++----- 1 file changed, 17 insertions(+), 5 deletions(-) diff --git a/block/file-posix.c b/block/file-posix.c index XXXXXXX..XXXXXXX 100644 --- a/block/file-posix.c +++ b/block/file-posix.c @@ -XXX,XX +XXX,XX @@ 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; @@ -XXX,XX +XXX,XX @@ static int aio_worker(void *arg) ret = handle_aiocb_write_zeroes_unmap(aiocb); break; case QEMU_AIO_COPY_RANGE: - ret = handle_aiocb_copy_range(aiocb); - break; case QEMU_AIO_TRUNCATE: g_assert_not_reached(); default: @@ -XXX,XX +XXX,XX @@ 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; @@ -XXX,XX +XXX,XX @@ 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 = { -- 2.19.2 aio_worker() doesn't add anything interesting, it's only a useless indirection. Call the handler function directly instead. As we know that this handler function is only called from coroutine context and the coroutine stays around until the worker thread finishes, we can keep RawPosixAIOData on the stack. Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/file-posix.c | 53 +++++++++++++++++++++++++++++----------------- 1 file changed, 34 insertions(+), 19 deletions(-) diff --git a/block/file-posix.c b/block/file-posix.c index XXXXXXX..XXXXXXX 100644 --- a/block/file-posix.c +++ b/block/file-posix.c @@ -XXX,XX +XXX,XX @@ 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 @@ -XXX,XX +XXX,XX @@ 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; @@ -XXX,XX +XXX,XX @@ static int aio_worker(void *arg) 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: case QEMU_AIO_TRUNCATE: g_assert_not_reached(); @@ -XXX,XX +XXX,XX @@ raw_co_pdiscard(BlockDriverState *bs, int64_t offset, int bytes) return paio_submit_co(bs, s->fd, offset, NULL, bytes, QEMU_AIO_DISCARD); } -static int coroutine_fn raw_co_pwrite_zeroes( - BlockDriverState *bs, int64_t offset, - int bytes, BdrvRequestFlags flags) +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) @@ -XXX,XX +XXX,XX @@ hdev_co_pdiscard(BlockDriverState *bs, int64_t offset, int bytes) 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); @@ -XXX,XX +XXX,XX @@ 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, -- 2.19.2 aio_worker() doesn't add anything interesting, it's only a useless indirection. Call the handler function directly instead. As we know that this handler function is only called from coroutine context and the coroutine stays around until the worker thread finishes, we can keep RawPosixAIOData on the stack. Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/file-posix.c | 32 ++++++++++++++++++++++++-------- 1 file changed, 24 insertions(+), 8 deletions(-) diff --git a/block/file-posix.c b/block/file-posix.c index XXXXXXX..XXXXXXX 100644 --- a/block/file-posix.c +++ b/block/file-posix.c @@ -XXX,XX +XXX,XX @@ static int handle_aiocb_copy_range(void *opaque) 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; @@ -XXX,XX +XXX,XX @@ static int aio_worker(void *arg) ret = handle_aiocb_ioctl(aiocb); break; case QEMU_AIO_DISCARD: - ret = handle_aiocb_discard(aiocb); - break; case QEMU_AIO_WRITE_ZEROES: case QEMU_AIO_WRITE_ZEROES | QEMU_AIO_DISCARD: case QEMU_AIO_COPY_RANGE: @@ -XXX,XX +XXX,XX @@ 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, + }; - return paio_submit_co(bs, s->fd, offset, NULL, bytes, QEMU_AIO_DISCARD); + if (blkdev) { + acb.aio_type |= QEMU_AIO_BLKDEV; + } + + return raw_thread_pool_submit(bs, handle_aiocb_discard, &acb); +} + +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 @@ -XXX,XX +XXX,XX @@ 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, -- 2.19.2 aio_worker() doesn't add anything interesting, it's only a useless indirection. Call the handler function directly instead. As we know that this handler function is only called from coroutine context and the coroutine stays around until the worker thread finishes, we can keep RawPosixAIOData on the stack. Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/file-posix.c | 16 +++++++++++----- 1 file changed, 11 insertions(+), 5 deletions(-) diff --git a/block/file-posix.c b/block/file-posix.c index XXXXXXX..XXXXXXX 100644 --- a/block/file-posix.c +++ b/block/file-posix.c @@ -XXX,XX +XXX,XX @@ static ssize_t handle_aiocb_ioctl(RawPosixAIOData *aiocb) return 0; } -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; @@ -XXX,XX +XXX,XX @@ static int aio_worker(void *arg) 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_FLUSH: case QEMU_AIO_DISCARD: case QEMU_AIO_WRITE_ZEROES: case QEMU_AIO_WRITE_ZEROES | QEMU_AIO_DISCARD: @@ -XXX,XX +XXX,XX @@ 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); @@ -XXX,XX +XXX,XX @@ 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, -- 2.19.2 aio_worker() for reads and writes isn't boring enough yet. It still does some postprocessing for handling short reads and turning the result into the right return value. However, there is no reason why handle_aiocb_rw() couldn't do the same, and even without duplicating code between the read and write path. So move the code there. Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/file-posix.c | 40 ++++++++++++++++++++-------------------- 1 file changed, 20 insertions(+), 20 deletions(-) diff --git a/block/file-posix.c b/block/file-posix.c index XXXXXXX..XXXXXXX 100644 --- a/block/file-posix.c +++ b/block/file-posix.c @@ -XXX,XX +XXX,XX @@ 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. @@ -XXX,XX +XXX,XX @@ 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; } @@ -XXX,XX +XXX,XX @@ 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) { @@ -XXX,XX +XXX,XX @@ 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 @@ -XXX,XX +XXX,XX @@ static int aio_worker(void *arg) 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_IOCTL: ret = handle_aiocb_ioctl(aiocb); -- 2.19.2 aio_worker() doesn't add anything interesting, it's only a useless indirection. Call the handler function directly instead. As we know that this handler function is only called from coroutine context and the coroutine stays around until the worker thread finishes, we can keep RawPosixAIOData on the stack. Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/file-posix.c | 27 +++++++++++++++++++-------- 1 file changed, 19 insertions(+), 8 deletions(-) diff --git a/block/file-posix.c b/block/file-posix.c index XXXXXXX..XXXXXXX 100644 --- a/block/file-posix.c +++ b/block/file-posix.c @@ -XXX,XX +XXX,XX @@ 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; @@ -XXX,XX +XXX,XX @@ static int aio_worker(void *arg) ssize_t ret = 0; switch (aiocb->aio_type & QEMU_AIO_TYPE_MASK) { - case QEMU_AIO_READ: - ret = handle_aiocb_rw(aiocb); - break; - case QEMU_AIO_WRITE: - ret = handle_aiocb_rw(aiocb); - break; case QEMU_AIO_IOCTL: ret = handle_aiocb_ioctl(aiocb); break; + case QEMU_AIO_READ: + case QEMU_AIO_WRITE: case QEMU_AIO_FLUSH: case QEMU_AIO_DISCARD: case QEMU_AIO_WRITE_ZEROES: @@ -XXX,XX +XXX,XX @@ 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; @@ -XXX,XX +XXX,XX @@ 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, -- 2.19.2 The function is not used any more, remove it. Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/file-posix.c | 34 ---------------------------------- 1 file changed, 34 deletions(-) diff --git a/block/file-posix.c b/block/file-posix.c index XXXXXXX..XXXXXXX 100644 --- a/block/file-posix.c +++ b/block/file-posix.c @@ -XXX,XX +XXX,XX @@ 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) { -- 2.19.2 No real reason to keep using the callback based mechanism here when the rest of the file-posix driver is coroutine based. Changing it brings ioctls more in line with how other request types work. Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- include/scsi/pr-manager.h | 8 +++----- block/file-posix.c | 21 +++++++++++---------- scsi/pr-manager.c | 21 +++++++++------------ scsi/trace-events | 2 +- 4 files changed, 24 insertions(+), 28 deletions(-) diff --git a/include/scsi/pr-manager.h b/include/scsi/pr-manager.h index XXXXXXX..XXXXXXX 100644 --- a/include/scsi/pr-manager.h +++ b/include/scsi/pr-manager.h @@ -XXX,XX +XXX,XX @@ #include "qapi/visitor.h" #include "qom/object_interfaces.h" #include "block/aio.h" +#include "qemu/coroutine.h" #define TYPE_PR_MANAGER "pr-manager" @@ -XXX,XX +XXX,XX @@ typedef struct PRManagerClass { } PRManagerClass; bool pr_manager_is_connected(PRManager *pr_mgr); -BlockAIOCB *pr_manager_execute(PRManager *pr_mgr, - AioContext *ctx, int fd, - struct sg_io_hdr *hdr, - BlockCompletionFunc *complete, - void *opaque); +int coroutine_fn pr_manager_execute(PRManager *pr_mgr, AioContext *ctx, int fd, + struct sg_io_hdr *hdr); PRManager *pr_manager_lookup(const char *id, Error **errp); diff --git a/block/file-posix.c b/block/file-posix.c index XXXXXXX..XXXXXXX 100644 --- a/block/file-posix.c +++ b/block/file-posix.c @@ -XXX,XX +XXX,XX @@ 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; + 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); } } @@ -XXX,XX +XXX,XX @@ static BlockAIOCB *hdev_aio_ioctl(BlockDriverState *bs, 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); + return thread_pool_submit_co(pool, aio_worker, acb); } #endif /* linux */ @@ -XXX,XX +XXX,XX @@ static BlockDriver bdrv_host_device = { /* generic scsi device */ #ifdef __linux__ - .bdrv_aio_ioctl = hdev_aio_ioctl, + .bdrv_co_ioctl = hdev_co_ioctl, #endif }; @@ -XXX,XX +XXX,XX @@ 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__ */ diff --git a/scsi/pr-manager.c b/scsi/pr-manager.c index XXXXXXX..XXXXXXX 100644 --- a/scsi/pr-manager.c +++ b/scsi/pr-manager.c @@ -XXX,XX +XXX,XX @@ static int pr_manager_worker(void *opaque) } -BlockAIOCB *pr_manager_execute(PRManager *pr_mgr, - AioContext *ctx, int fd, - struct sg_io_hdr *hdr, - BlockCompletionFunc *complete, - void *opaque) +int coroutine_fn pr_manager_execute(PRManager *pr_mgr, AioContext *ctx, int fd, + struct sg_io_hdr *hdr) { - PRManagerData *data = g_new(PRManagerData, 1); ThreadPool *pool = aio_get_thread_pool(ctx); + PRManagerData data = { + .pr_mgr = pr_mgr, + .fd = fd, + .hdr = hdr, + }; - trace_pr_manager_execute(fd, hdr->cmdp[0], hdr->cmdp[1], opaque); - data->pr_mgr = pr_mgr; - data->fd = fd; - data->hdr = hdr; + trace_pr_manager_execute(fd, hdr->cmdp[0], hdr->cmdp[1]); /* The matching object_unref is in pr_manager_worker. */ object_ref(OBJECT(pr_mgr)); - return thread_pool_submit_aio(pool, pr_manager_worker, - data, complete, opaque); + return thread_pool_submit_co(pool, pr_manager_worker, &data); } bool pr_manager_is_connected(PRManager *pr_mgr) diff --git a/scsi/trace-events b/scsi/trace-events index XXXXXXX..XXXXXXX 100644 --- a/scsi/trace-events +++ b/scsi/trace-events @@ -XXX,XX +XXX,XX @@ # scsi/pr-manager.c -pr_manager_execute(int fd, int cmd, int sa, void *opaque) "fd=%d cmd=0x%02x service action=0x%02x opaque=%p" +pr_manager_execute(int fd, int cmd, int sa) "fd=%d cmd=0x%02x service action=0x%02x" pr_manager_run(int fd, int cmd, int sa) "fd=%d cmd=0x%02x service action=0x%02x" -- 2.19.2 aio_worker() doesn't add anything interesting, it's only a useless indirection. Call the handler function directly instead. As we know that this handler function is only called from coroutine context and the coroutine stays around until the worker thread finishes, we can keep RawPosixAIOData on the stack. This was the last user of aio_worker(), so the function goes away now. Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/file-posix.c | 55 +++++++++++++--------------------------------- 1 file changed, 15 insertions(+), 40 deletions(-) diff --git a/block/file-posix.c b/block/file-posix.c index XXXXXXX..XXXXXXX 100644 --- a/block/file-posix.c +++ b/block/file-posix.c @@ -XXX,XX +XXX,XX @@ static int hdev_probe_geometry(BlockDriverState *bs, HDGeometry *geo) } #endif -static ssize_t handle_aiocb_ioctl(RawPosixAIOData *aiocb) +static int handle_aiocb_ioctl(void *opaque) { + RawPosixAIOData *aiocb = opaque; int ret; ret = ioctl(aiocb->aio_fildes, aiocb->ioctl.cmd, aiocb->ioctl.buf); @@ -XXX,XX +XXX,XX @@ 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_IOCTL: - ret = handle_aiocb_ioctl(aiocb); - break; - case QEMU_AIO_READ: - case QEMU_AIO_WRITE: - case QEMU_AIO_FLUSH: - case QEMU_AIO_DISCARD: - case QEMU_AIO_WRITE_ZEROES: - case QEMU_AIO_WRITE_ZEROES | QEMU_AIO_DISCARD: - case QEMU_AIO_COPY_RANGE: - case QEMU_AIO_TRUNCATE: - g_assert_not_reached(); - 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) { @@ -XXX,XX +XXX,XX @@ 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; ret = fd_open(bs); @@ -XXX,XX +XXX,XX @@ hdev_co_ioctl(BlockDriverState *bs, unsigned long int req, void *buf) } } - 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_co(pool, aio_worker, acb); + 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 */ -- 2.19.2 From: Alberto Garcia <berto@igalia.com> Most callers of bdrv_reopen() only use it to switch a BlockDriverState between read-only and read-write, so this patch adds a new function that does just that. We also want to get rid of the flags parameter in the bdrv_reopen() API, so this function sets the "read-only" option and passes the original flags (which will then be updated in bdrv_reopen_prepare()). Signed-off-by: Alberto Garcia <berto@igalia.com> Reviewed-by: Max Reitz <mreitz@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- include/block/block.h | 2 ++ block.c | 17 +++++++++++++++++ 2 files changed, 19 insertions(+) diff --git a/include/block/block.h b/include/block/block.h index XXXXXXX..XXXXXXX 100644 --- a/include/block/block.h +++ b/include/block/block.h @@ -XXX,XX +XXX,XX @@ BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue, QDict *options, int flags); int bdrv_reopen_multiple(AioContext *ctx, BlockReopenQueue *bs_queue, Error **errp); int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp); +int bdrv_reopen_set_read_only(BlockDriverState *bs, bool read_only, + Error **errp); int bdrv_reopen_prepare(BDRVReopenState *reopen_state, BlockReopenQueue *queue, Error **errp); void bdrv_reopen_commit(BDRVReopenState *reopen_state); diff --git a/block.c b/block.c index XXXXXXX..XXXXXXX 100644 --- a/block.c +++ b/block.c @@ -XXX,XX +XXX,XX @@ int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp) return ret; } +int bdrv_reopen_set_read_only(BlockDriverState *bs, bool read_only, + Error **errp) +{ + int ret; + BlockReopenQueue *queue; + QDict *opts = qdict_new(); + + qdict_put_bool(opts, BDRV_OPT_READ_ONLY, read_only); + + bdrv_subtree_drained_begin(bs); + queue = bdrv_reopen_queue(NULL, bs, opts, bdrv_get_flags(bs)); + ret = bdrv_reopen_multiple(bdrv_get_aio_context(bs), queue, errp); + bdrv_subtree_drained_end(bs); + + return ret; +} + static BlockReopenQueueEntry *find_parent_in_reopen_queue(BlockReopenQueue *q, BdrvChild *c) { -- 2.19.2 From: Alberto Garcia <berto@igalia.com> This patch replaces the bdrv_reopen() calls that set and remove the BDRV_O_RDWR flag with the new bdrv_reopen_set_read_only() function. Signed-off-by: Alberto Garcia <berto@igalia.com> Reviewed-by: Max Reitz <mreitz@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/block.c b/block.c index XXXXXXX..XXXXXXX 100644 --- a/block.c +++ b/block.c @@ -XXX,XX +XXX,XX @@ static int bdrv_backing_update_filename(BdrvChild *c, BlockDriverState *base, const char *filename, Error **errp) { BlockDriverState *parent = c->opaque; - int orig_flags = bdrv_get_flags(parent); + bool read_only = bdrv_is_read_only(parent); int ret; - if (!(orig_flags & BDRV_O_RDWR)) { - ret = bdrv_reopen(parent, orig_flags | BDRV_O_RDWR, errp); + if (read_only) { + ret = bdrv_reopen_set_read_only(parent, false, errp); if (ret < 0) { return ret; } @@ -XXX,XX +XXX,XX @@ static int bdrv_backing_update_filename(BdrvChild *c, BlockDriverState *base, error_setg_errno(errp, -ret, "Could not update backing file link"); } - if (!(orig_flags & BDRV_O_RDWR)) { - bdrv_reopen(parent, orig_flags, NULL); + if (read_only) { + bdrv_reopen_set_read_only(parent, true, NULL); } return ret; -- 2.19.2 From: Alberto Garcia <berto@igalia.com> This patch replaces the bdrv_reopen() calls that set and remove the BDRV_O_RDWR flag with the new bdrv_reopen_set_read_only() function. Signed-off-by: Alberto Garcia <berto@igalia.com> Reviewed-by: Max Reitz <mreitz@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/commit.c | 16 ++++++---------- 1 file changed, 6 insertions(+), 10 deletions(-) diff --git a/block/commit.c b/block/commit.c index XXXXXXX..XXXXXXX 100644 --- a/block/commit.c +++ b/block/commit.c @@ -XXX,XX +XXX,XX @@ typedef struct CommitBlockJob { BlockBackend *base; BlockDriverState *base_bs; BlockdevOnError on_error; - int base_flags; + bool base_read_only; char *backing_file_str; } CommitBlockJob; @@ -XXX,XX +XXX,XX @@ static void commit_clean(Job *job) /* restore base open flags here if appropriate (e.g., change the base back * to r/o). These reopens do not need to be atomic, since we won't abort * even on failure here */ - if (s->base_flags != bdrv_get_flags(s->base_bs)) { - bdrv_reopen(s->base_bs, s->base_flags, NULL); + if (s->base_read_only) { + bdrv_reopen_set_read_only(s->base_bs, true, NULL); } g_free(s->backing_file_str); @@ -XXX,XX +XXX,XX @@ void commit_start(const char *job_id, BlockDriverState *bs, const char *filter_node_name, Error **errp) { CommitBlockJob *s; - int orig_base_flags; BlockDriverState *iter; BlockDriverState *commit_top_bs = NULL; Error *local_err = NULL; @@ -XXX,XX +XXX,XX @@ void commit_start(const char *job_id, BlockDriverState *bs, } /* convert base to r/w, if necessary */ - orig_base_flags = bdrv_get_flags(base); - if (!(orig_base_flags & BDRV_O_RDWR)) { - bdrv_reopen(base, orig_base_flags | BDRV_O_RDWR, &local_err); - if (local_err != NULL) { - error_propagate(errp, local_err); + s->base_read_only = bdrv_is_read_only(base); + if (s->base_read_only) { + if (bdrv_reopen_set_read_only(base, false, errp) != 0) { goto fail; } } @@ -XXX,XX +XXX,XX @@ void commit_start(const char *job_id, BlockDriverState *bs, goto fail; } - s->base_flags = orig_base_flags; s->backing_file_str = g_strdup(backing_file_str); s->on_error = on_error; -- 2.19.2 From: Alberto Garcia <berto@igalia.com> This patch replaces the bdrv_reopen() calls that set and remove the BDRV_O_RDWR flag with the new bdrv_reopen_set_read_only() function. Signed-off-by: Alberto Garcia <berto@igalia.com> Reviewed-by: Max Reitz <mreitz@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/commit.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) diff --git a/block/commit.c b/block/commit.c index XXXXXXX..XXXXXXX 100644 --- a/block/commit.c +++ b/block/commit.c @@ -XXX,XX +XXX,XX @@ int bdrv_commit(BlockDriverState *bs) BlockDriverState *commit_top_bs = NULL; BlockDriver *drv = bs->drv; int64_t offset, length, backing_length; - int ro, open_flags; + int ro; int64_t n; int ret = 0; uint8_t *buf = NULL; @@ -XXX,XX +XXX,XX @@ int bdrv_commit(BlockDriverState *bs) } ro = bs->backing->bs->read_only; - open_flags = bs->backing->bs->open_flags; if (ro) { - if (bdrv_reopen(bs->backing->bs, open_flags | BDRV_O_RDWR, NULL)) { + if (bdrv_reopen_set_read_only(bs->backing->bs, false, NULL)) { return -EACCES; } } @@ -XXX,XX +XXX,XX @@ ro_cleanup: if (ro) { /* ignoring error return here */ - bdrv_reopen(bs->backing->bs, open_flags & ~BDRV_O_RDWR, NULL); + bdrv_reopen_set_read_only(bs->backing->bs, true, NULL); } return ret; -- 2.19.2 From: Alberto Garcia <berto@igalia.com> This patch replaces the bdrv_reopen() calls that set and remove the BDRV_O_RDWR flag with the new bdrv_reopen_set_read_only() function. Signed-off-by: Alberto Garcia <berto@igalia.com> Reviewed-by: Max Reitz <mreitz@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/stream.c | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) diff --git a/block/stream.c b/block/stream.c index XXXXXXX..XXXXXXX 100644 --- a/block/stream.c +++ b/block/stream.c @@ -XXX,XX +XXX,XX @@ typedef struct StreamBlockJob { BlockDriverState *base; BlockdevOnError on_error; char *backing_file_str; - int bs_flags; + bool bs_read_only; } StreamBlockJob; static int coroutine_fn stream_populate(BlockBackend *blk, @@ -XXX,XX +XXX,XX @@ static void stream_clean(Job *job) BlockDriverState *bs = blk_bs(bjob->blk); /* Reopen the image back in read-only mode if necessary */ - if (s->bs_flags != bdrv_get_flags(bs)) { + if (s->bs_read_only) { /* Give up write permissions before making it read-only */ blk_set_perm(bjob->blk, 0, BLK_PERM_ALL, &error_abort); - bdrv_reopen(bs, s->bs_flags, NULL); + bdrv_reopen_set_read_only(bs, true, NULL); } g_free(s->backing_file_str); @@ -XXX,XX +XXX,XX @@ void stream_start(const char *job_id, BlockDriverState *bs, { StreamBlockJob *s; BlockDriverState *iter; - int orig_bs_flags; + bool bs_read_only; /* Make sure that the image is opened in read-write mode */ - orig_bs_flags = bdrv_get_flags(bs); - if (!(orig_bs_flags & BDRV_O_RDWR)) { - if (bdrv_reopen(bs, orig_bs_flags | BDRV_O_RDWR, errp) != 0) { + bs_read_only = bdrv_is_read_only(bs); + if (bs_read_only) { + if (bdrv_reopen_set_read_only(bs, false, errp) != 0) { return; } } @@ -XXX,XX +XXX,XX @@ void stream_start(const char *job_id, BlockDriverState *bs, s->base = base; s->backing_file_str = g_strdup(backing_file_str); - s->bs_flags = orig_bs_flags; + s->bs_read_only = bs_read_only; s->on_error = on_error; trace_stream_start(bs, base, s); @@ -XXX,XX +XXX,XX @@ void stream_start(const char *job_id, BlockDriverState *bs, return; fail: - if (orig_bs_flags != bdrv_get_flags(bs)) { - bdrv_reopen(bs, orig_bs_flags, NULL); + if (bs_read_only) { + bdrv_reopen_set_read_only(bs, true, NULL); } } -- 2.19.2 From: Alberto Garcia <berto@igalia.com> This patch replaces the bdrv_reopen() calls that set and remove the BDRV_O_RDWR flag with the new bdrv_reopen_set_read_only() function. Signed-off-by: Alberto Garcia <berto@igalia.com> Reviewed-by: Max Reitz <mreitz@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- blockdev.c | 8 ++------ 1 file changed, 2 insertions(+), 6 deletions(-) diff --git a/blockdev.c b/blockdev.c index XXXXXXX..XXXXXXX 100644 --- a/blockdev.c +++ b/blockdev.c @@ -XXX,XX +XXX,XX @@ void qmp_change_backing_file(const char *device, BlockDriverState *image_bs = NULL; Error *local_err = NULL; bool ro; - int open_flags; int ret; bs = qmp_get_root_bs(device, errp); @@ -XXX,XX +XXX,XX @@ void qmp_change_backing_file(const char *device, } /* if not r/w, reopen to make r/w */ - open_flags = image_bs->open_flags; ro = bdrv_is_read_only(image_bs); if (ro) { - bdrv_reopen(image_bs, open_flags | BDRV_O_RDWR, &local_err); - if (local_err) { - error_propagate(errp, local_err); + if (bdrv_reopen_set_read_only(image_bs, false, errp) != 0) { goto out; } } @@ -XXX,XX +XXX,XX @@ void qmp_change_backing_file(const char *device, } if (ro) { - bdrv_reopen(image_bs, open_flags, &local_err); + bdrv_reopen_set_read_only(image_bs, true, &local_err); error_propagate(errp, local_err); } -- 2.19.2 From: Alberto Garcia <berto@igalia.com> This patch replaces the bdrv_reopen() call that set and remove the BDRV_O_RDWR flag with the new bdrv_reopen_set_read_only() function. Signed-off-by: Alberto Garcia <berto@igalia.com> Reviewed-by: Max Reitz <mreitz@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- blockdev.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/blockdev.c b/blockdev.c index XXXXXXX..XXXXXXX 100644 --- a/blockdev.c +++ b/blockdev.c @@ -XXX,XX +XXX,XX @@ static void external_snapshot_commit(BlkActionState *common) * bdrv_reopen_multiple() across all the entries at once, because we * don't want to abort all of them if one of them fails the reopen */ if (!atomic_read(&state->old_bs->copy_on_read)) { - bdrv_reopen(state->old_bs, state->old_bs->open_flags & ~BDRV_O_RDWR, - NULL); + bdrv_reopen_set_read_only(state->old_bs, true, NULL); } aio_context_release(aio_context); -- 2.19.2 From: Alberto Garcia <berto@igalia.com> The 'block-commit' QMP command is implemented internally using two different drivers. If the source image is the active layer then the mirror driver is used (commit_active_start()), otherwise the commit driver is used (commit_start()). In both cases the destination image must be put temporarily in read-write mode. This is done correctly in the latter case, but what commit_active_start() does is copy all flags instead. This patch replaces the bdrv_reopen() calls in that function with bdrv_reopen_set_read_only() so that only the read-only status is changed. A similar change is made in mirror_exit(), which is also used by the 'drive-mirror' and 'blockdev-mirror' commands. Signed-off-by: Alberto Garcia <berto@igalia.com> Reviewed-by: Max Reitz <mreitz@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/mirror.c | 19 ++++++++++++------- 1 file changed, 12 insertions(+), 7 deletions(-) diff --git a/block/mirror.c b/block/mirror.c index XXXXXXX..XXXXXXX 100644 --- a/block/mirror.c +++ b/block/mirror.c @@ -XXX,XX +XXX,XX @@ static int mirror_exit_common(Job *job) if (s->should_complete && !abort) { BlockDriverState *to_replace = s->to_replace ?: src; + bool ro = bdrv_is_read_only(to_replace); - if (bdrv_get_flags(target_bs) != bdrv_get_flags(to_replace)) { - bdrv_reopen(target_bs, bdrv_get_flags(to_replace), NULL); + if (ro != bdrv_is_read_only(target_bs)) { + bdrv_reopen_set_read_only(target_bs, ro, NULL); } /* The mirror job has no requests in flight any more, but we need to @@ -XXX,XX +XXX,XX @@ void commit_active_start(const char *job_id, BlockDriverState *bs, BlockCompletionFunc *cb, void *opaque, bool auto_complete, Error **errp) { - int orig_base_flags; + bool base_read_only; Error *local_err = NULL; - orig_base_flags = bdrv_get_flags(base); + base_read_only = bdrv_is_read_only(base); - if (bdrv_reopen(base, bs->open_flags, errp)) { - return; + if (base_read_only) { + if (bdrv_reopen_set_read_only(base, false, errp) < 0) { + return; + } } mirror_start_job(job_id, bs, creation_flags, base, NULL, speed, 0, 0, @@ -XXX,XX +XXX,XX @@ void commit_active_start(const char *job_id, BlockDriverState *bs, error_restore_flags: /* ignore error and errp for bdrv_reopen, because we want to propagate * the original error */ - bdrv_reopen(base, orig_base_flags, NULL); + if (base_read_only) { + bdrv_reopen_set_read_only(base, true, NULL); + } return; } -- 2.19.2 From: Alberto Garcia <berto@igalia.com> No one is using this function anymore, so we can safely remove it. Signed-off-by: Alberto Garcia <berto@igalia.com> Reviewed-by: Max Reitz <mreitz@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- include/block/block.h | 1 - block.c | 21 --------------------- 2 files changed, 22 deletions(-) diff --git a/include/block/block.h b/include/block/block.h index XXXXXXX..XXXXXXX 100644 --- a/include/block/block.h +++ b/include/block/block.h @@ -XXX,XX +XXX,XX @@ BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue, BlockDriverState *bs, QDict *options, int flags); int bdrv_reopen_multiple(AioContext *ctx, BlockReopenQueue *bs_queue, Error **errp); -int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp); int bdrv_reopen_set_read_only(BlockDriverState *bs, bool read_only, Error **errp); int bdrv_reopen_prepare(BDRVReopenState *reopen_state, diff --git a/block.c b/block.c index XXXXXXX..XXXXXXX 100644 --- a/block.c +++ b/block.c @@ -XXX,XX +XXX,XX @@ cleanup: return ret; } - -/* Reopen a single BlockDriverState with the specified flags. */ -int bdrv_reopen(BlockDriverState *bs, int bdrv_flags, Error **errp) -{ - int ret = -1; - Error *local_err = NULL; - BlockReopenQueue *queue; - - bdrv_subtree_drained_begin(bs); - - queue = bdrv_reopen_queue(NULL, bs, NULL, bdrv_flags); - ret = bdrv_reopen_multiple(bdrv_get_aio_context(bs), queue, &local_err); - if (local_err != NULL) { - error_propagate(errp, local_err); - } - - bdrv_subtree_drained_end(bs); - - return ret; -} - int bdrv_reopen_set_read_only(BlockDriverState *bs, bool read_only, Error **errp) { -- 2.19.2 From: Alberto Garcia <berto@igalia.com> When reopen_f() puts a block device in the reopen queue, some of the new options are passed using a QDict, but others ("read-only" and the cache options) are passed as flags. This patch puts those flags in the QDict. This way the flags parameter becomes redundant and we'll be able to get rid of it in a subsequent patch. Signed-off-by: Alberto Garcia <berto@igalia.com> Reviewed-by: Max Reitz <mreitz@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- qemu-io-cmds.c | 27 ++++++++++++++++++++++++++- tests/qemu-iotests/133 | 9 +++++++++ tests/qemu-iotests/133.out | 8 ++++++++ 3 files changed, 43 insertions(+), 1 deletion(-) diff --git a/qemu-io-cmds.c b/qemu-io-cmds.c index XXXXXXX..XXXXXXX 100644 --- a/qemu-io-cmds.c +++ b/qemu-io-cmds.c @@ -XXX,XX +XXX,XX @@ #include "qemu/osdep.h" #include "qapi/error.h" +#include "qapi/qmp/qdict.h" #include "qemu-io.h" #include "sysemu/block-backend.h" #include "block/block.h" @@ -XXX,XX +XXX,XX @@ static int reopen_f(BlockBackend *blk, int argc, char **argv) int flags = bs->open_flags; bool writethrough = !blk_enable_write_cache(blk); bool has_rw_option = false; + bool has_cache_option = false; BlockReopenQueue *brq; Error *local_err = NULL; @@ -XXX,XX +XXX,XX @@ static int reopen_f(BlockBackend *blk, int argc, char **argv) error_report("Invalid cache option: %s", optarg); return -EINVAL; } + has_cache_option = true; break; case 'o': if (!qemu_opts_parse_noisily(&reopen_opts, optarg, 0)) { @@ -XXX,XX +XXX,XX @@ static int reopen_f(BlockBackend *blk, int argc, char **argv) } qopts = qemu_opts_find(&reopen_opts, NULL); - opts = qopts ? qemu_opts_to_qdict(qopts, NULL) : NULL; + opts = qopts ? qemu_opts_to_qdict(qopts, NULL) : qdict_new(); qemu_opts_reset(&reopen_opts); + if (qdict_haskey(opts, BDRV_OPT_READ_ONLY)) { + if (has_rw_option) { + error_report("Cannot set both -r/-w and '" BDRV_OPT_READ_ONLY "'"); + qobject_unref(opts); + return -EINVAL; + } + } else { + qdict_put_bool(opts, BDRV_OPT_READ_ONLY, !(flags & BDRV_O_RDWR)); + } + + if (qdict_haskey(opts, BDRV_OPT_CACHE_DIRECT) || + qdict_haskey(opts, BDRV_OPT_CACHE_NO_FLUSH)) { + if (has_cache_option) { + error_report("Cannot set both -c and the cache options"); + qobject_unref(opts); + return -EINVAL; + } + } else { + qdict_put_bool(opts, BDRV_OPT_CACHE_DIRECT, flags & BDRV_O_NOCACHE); + qdict_put_bool(opts, BDRV_OPT_CACHE_NO_FLUSH, flags & BDRV_O_NO_FLUSH); + } + bdrv_subtree_drained_begin(bs); brq = bdrv_reopen_queue(NULL, bs, opts, flags); bdrv_reopen_multiple(bdrv_get_aio_context(bs), brq, &local_err); diff --git a/tests/qemu-iotests/133 b/tests/qemu-iotests/133 index XXXXXXX..XXXXXXX 100755 --- a/tests/qemu-iotests/133 +++ b/tests/qemu-iotests/133 @@ -XXX,XX +XXX,XX @@ echo IMGOPTSSYNTAX=false $QEMU_IO -f null-co -c 'reopen' -c 'info' \ "json:{'driver': 'null-co', 'size': 65536}" +echo +echo "=== Check that mixing -c/-r/-w and their corresponding options is forbidden ===" +echo + +$QEMU_IO -c 'reopen -r -o read-only=on' $TEST_IMG +$QEMU_IO -c 'reopen -w -o read-only=on' $TEST_IMG +$QEMU_IO -c 'reopen -c none -o cache.direct=on' $TEST_IMG +$QEMU_IO -c 'reopen -c writeback -o cache.direct=on' $TEST_IMG +$QEMU_IO -c 'reopen -c directsync -o cache.no-flush=on' $TEST_IMG # success, all done echo "*** done" rm -f $seq.full diff --git a/tests/qemu-iotests/133.out b/tests/qemu-iotests/133.out index XXXXXXX..XXXXXXX 100644 --- a/tests/qemu-iotests/133.out +++ b/tests/qemu-iotests/133.out @@ -XXX,XX +XXX,XX @@ Cannot change the option 'driver' format name: null-co format name: null-co + +=== Check that mixing -c/-r/-w and their corresponding options is forbidden === + +Cannot set both -r/-w and 'read-only' +Cannot set both -r/-w and 'read-only' +Cannot set both -c and the cache options +Cannot set both -c and the cache options +Cannot set both -c and the cache options *** done -- 2.19.2 From: Alberto Garcia <berto@igalia.com> This function is used to put the hidden and secondary disks in read-write mode before launching the backup job, and back in read-only mode afterwards. This patch does the following changes: - Use an options QDict with the "read-only" option instead of passing the changes as flags only. - Simplify the code (it was unnecessarily complicated and verbose). - Fix a bug due to which the secondary disk was not being put back in read-only mode when writable=false (because in this case orig_secondary_flags always had the BDRV_O_RDWR flag set). - Stop clearing the BDRV_O_INACTIVE flag. The flags parameter to bdrv_reopen_queue() becomes redundant and we'll be able to get rid of it in a subsequent patch. Signed-off-by: Alberto Garcia <berto@igalia.com> Reviewed-by: Max Reitz <mreitz@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/replication.c | 45 +++++++++++++++++++++------------------------ 1 file changed, 21 insertions(+), 24 deletions(-) diff --git a/block/replication.c b/block/replication.c index XXXXXXX..XXXXXXX 100644 --- a/block/replication.c +++ b/block/replication.c @@ -XXX,XX +XXX,XX @@ #include "block/block_backup.h" #include "sysemu/block-backend.h" #include "qapi/error.h" +#include "qapi/qmp/qdict.h" #include "replication.h" typedef enum { @@ -XXX,XX +XXX,XX @@ typedef struct BDRVReplicationState { char *top_id; ReplicationState *rs; Error *blocker; - int orig_hidden_flags; - int orig_secondary_flags; + bool orig_hidden_read_only; + bool orig_secondary_read_only; int error; } BDRVReplicationState; @@ -XXX,XX +XXX,XX @@ static void secondary_do_checkpoint(BDRVReplicationState *s, Error **errp) } } +/* This function is supposed to be called twice: + * first with writable = true, then with writable = false. + * The first call puts s->hidden_disk and s->secondary_disk in + * r/w mode, and the second puts them back in their original state. + */ static void reopen_backing_file(BlockDriverState *bs, bool writable, Error **errp) { BDRVReplicationState *s = bs->opaque; BlockReopenQueue *reopen_queue = NULL; - int orig_hidden_flags, orig_secondary_flags; - int new_hidden_flags, new_secondary_flags; Error *local_err = NULL; if (writable) { - orig_hidden_flags = s->orig_hidden_flags = - bdrv_get_flags(s->hidden_disk->bs); - new_hidden_flags = (orig_hidden_flags | BDRV_O_RDWR) & - ~BDRV_O_INACTIVE; - orig_secondary_flags = s->orig_secondary_flags = - bdrv_get_flags(s->secondary_disk->bs); - new_secondary_flags = (orig_secondary_flags | BDRV_O_RDWR) & - ~BDRV_O_INACTIVE; - } else { - orig_hidden_flags = (s->orig_hidden_flags | BDRV_O_RDWR) & - ~BDRV_O_INACTIVE; - new_hidden_flags = s->orig_hidden_flags; - orig_secondary_flags = (s->orig_secondary_flags | BDRV_O_RDWR) & - ~BDRV_O_INACTIVE; - new_secondary_flags = s->orig_secondary_flags; + s->orig_hidden_read_only = bdrv_is_read_only(s->hidden_disk->bs); + s->orig_secondary_read_only = bdrv_is_read_only(s->secondary_disk->bs); } bdrv_subtree_drained_begin(s->hidden_disk->bs); bdrv_subtree_drained_begin(s->secondary_disk->bs); - if (orig_hidden_flags != new_hidden_flags) { - reopen_queue = bdrv_reopen_queue(reopen_queue, s->hidden_disk->bs, NULL, - new_hidden_flags); + if (s->orig_hidden_read_only) { + int flags = bdrv_get_flags(s->hidden_disk->bs); + QDict *opts = qdict_new(); + qdict_put_bool(opts, BDRV_OPT_READ_ONLY, !writable); + reopen_queue = bdrv_reopen_queue(reopen_queue, s->hidden_disk->bs, + opts, flags); } - if (!(orig_secondary_flags & BDRV_O_RDWR)) { + if (s->orig_secondary_read_only) { + int flags = bdrv_get_flags(s->secondary_disk->bs); + QDict *opts = qdict_new(); + qdict_put_bool(opts, BDRV_OPT_READ_ONLY, !writable); reopen_queue = bdrv_reopen_queue(reopen_queue, s->secondary_disk->bs, - NULL, new_secondary_flags); + opts, flags); } if (reopen_queue) { -- 2.19.2 From: Alberto Garcia <berto@igalia.com> Now that all callers are passing all flag changes as QDict options, the flags parameter is no longer necessary, so we can get rid of it. Signed-off-by: Alberto Garcia <berto@igalia.com> Reviewed-by: Max Reitz <mreitz@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- include/block/block.h | 3 +-- block.c | 5 +++-- block/replication.c | 6 ++---- qemu-io-cmds.c | 2 +- 4 files changed, 7 insertions(+), 9 deletions(-) diff --git a/include/block/block.h b/include/block/block.h index XXXXXXX..XXXXXXX 100644 --- a/include/block/block.h +++ b/include/block/block.h @@ -XXX,XX +XXX,XX @@ BlockDriverState *bdrv_open(const char *filename, const char *reference, BlockDriverState *bdrv_new_open_driver(BlockDriver *drv, const char *node_name, int flags, Error **errp); BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue, - BlockDriverState *bs, - QDict *options, int flags); + BlockDriverState *bs, QDict *options); int bdrv_reopen_multiple(AioContext *ctx, BlockReopenQueue *bs_queue, Error **errp); int bdrv_reopen_set_read_only(BlockDriverState *bs, bool read_only, Error **errp); diff --git a/block.c b/block.c index XXXXXXX..XXXXXXX 100644 --- a/block.c +++ b/block.c @@ -XXX,XX +XXX,XX @@ static BlockReopenQueue *bdrv_reopen_queue_child(BlockReopenQueue *bs_queue, BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue, BlockDriverState *bs, - QDict *options, int flags) + QDict *options) { + int flags = bdrv_get_flags(bs); return bdrv_reopen_queue_child(bs_queue, bs, options, flags, NULL, NULL, 0); } @@ -XXX,XX +XXX,XX @@ int bdrv_reopen_set_read_only(BlockDriverState *bs, bool read_only, qdict_put_bool(opts, BDRV_OPT_READ_ONLY, read_only); bdrv_subtree_drained_begin(bs); - queue = bdrv_reopen_queue(NULL, bs, opts, bdrv_get_flags(bs)); + queue = bdrv_reopen_queue(NULL, bs, opts); ret = bdrv_reopen_multiple(bdrv_get_aio_context(bs), queue, errp); bdrv_subtree_drained_end(bs); diff --git a/block/replication.c b/block/replication.c index XXXXXXX..XXXXXXX 100644 --- a/block/replication.c +++ b/block/replication.c @@ -XXX,XX +XXX,XX @@ static void reopen_backing_file(BlockDriverState *bs, bool writable, bdrv_subtree_drained_begin(s->secondary_disk->bs); if (s->orig_hidden_read_only) { - int flags = bdrv_get_flags(s->hidden_disk->bs); QDict *opts = qdict_new(); qdict_put_bool(opts, BDRV_OPT_READ_ONLY, !writable); reopen_queue = bdrv_reopen_queue(reopen_queue, s->hidden_disk->bs, - opts, flags); + opts); } if (s->orig_secondary_read_only) { - int flags = bdrv_get_flags(s->secondary_disk->bs); QDict *opts = qdict_new(); qdict_put_bool(opts, BDRV_OPT_READ_ONLY, !writable); reopen_queue = bdrv_reopen_queue(reopen_queue, s->secondary_disk->bs, - opts, flags); + opts); } if (reopen_queue) { diff --git a/qemu-io-cmds.c b/qemu-io-cmds.c index XXXXXXX..XXXXXXX 100644 --- a/qemu-io-cmds.c +++ b/qemu-io-cmds.c @@ -XXX,XX +XXX,XX @@ static int reopen_f(BlockBackend *blk, int argc, char **argv) } bdrv_subtree_drained_begin(bs); - brq = bdrv_reopen_queue(NULL, bs, opts, flags); + brq = bdrv_reopen_queue(NULL, bs, opts); bdrv_reopen_multiple(bdrv_get_aio_context(bs), brq, &local_err); bdrv_subtree_drained_end(bs); -- 2.19.2 From: Alberto Garcia <berto@igalia.com> Now that all callers are passing the new options using the QDict we no longer need the 'flags' parameter. This patch makes the following changes: 1) The update_options_from_flags() call is no longer necessary so it can be removed. 2) The update_flags_from_options() call is now used in all cases, and is moved down a few lines so it happens after the options QDict contains the final set of values. 3) The flags parameter is removed. Now the flags are initialized using the current value (for the top-level node) or the parent flags (after inherit_options()). In both cases the initial values are updated to reflect the new options in the QDict. This happens in bdrv_reopen_queue_child() (as explained above) and in bdrv_reopen_prepare(). Signed-off-by: Alberto Garcia <berto@igalia.com> Reviewed-by: Max Reitz <mreitz@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block.c | 48 +++++++++++++++++++----------------------------- 1 file changed, 19 insertions(+), 29 deletions(-) diff --git a/block.c b/block.c index XXXXXXX..XXXXXXX 100644 --- a/block.c +++ b/block.c @@ -XXX,XX +XXX,XX @@ BlockDriverState *bdrv_open(const char *filename, const char *reference, static BlockReopenQueue *bdrv_reopen_queue_child(BlockReopenQueue *bs_queue, BlockDriverState *bs, QDict *options, - int flags, const BdrvChildRole *role, QDict *parent_options, int parent_flags) @@ -XXX,XX +XXX,XX @@ static BlockReopenQueue *bdrv_reopen_queue_child(BlockReopenQueue *bs_queue, BlockReopenQueueEntry *bs_entry; BdrvChild *child; - QDict *old_options, *explicit_options; + QDict *old_options, *explicit_options, *options_copy; + int flags; + QemuOpts *opts; /* Make sure that the caller remembered to use a drained section. This is * important to avoid graph changes between the recursive queuing here and @@ -XXX,XX +XXX,XX @@ static BlockReopenQueue *bdrv_reopen_queue_child(BlockReopenQueue *bs_queue, /* * Precedence of options: * 1. Explicitly passed in options (highest) - * 2. Set in flags (only for top level) - * 3. Retained from explicitly set options of bs - * 4. Inherited from parent node - * 5. Retained from effective options of bs + * 2. Retained from explicitly set options of bs + * 3. Inherited from parent node + * 4. Retained from effective options of bs */ - if (!parent_options) { - /* - * Any setting represented by flags is always updated. If the - * corresponding QDict option is set, it takes precedence. Otherwise - * the flag is translated into a QDict option. The old setting of bs is - * not considered. - */ - update_options_from_flags(options, flags); - } - /* Old explicitly set values (don't overwrite by inherited value) */ if (bs_entry) { old_options = qdict_clone_shallow(bs_entry->state.explicit_options); @@ -XXX,XX +XXX,XX @@ static BlockReopenQueue *bdrv_reopen_queue_child(BlockReopenQueue *bs_queue, /* Inherit from parent node */ if (parent_options) { - QemuOpts *opts; - QDict *options_copy; - assert(!flags); + flags = 0; role->inherit_options(&flags, options, parent_flags, parent_options); - options_copy = qdict_clone_shallow(options); - opts = qemu_opts_create(&bdrv_runtime_opts, NULL, 0, &error_abort); - qemu_opts_absorb_qdict(opts, options_copy, NULL); - update_flags_from_options(&flags, opts); - qemu_opts_del(opts); - qobject_unref(options_copy); + } else { + flags = bdrv_get_flags(bs); } /* Old values are used for options that aren't set yet */ @@ -XXX,XX +XXX,XX @@ static BlockReopenQueue *bdrv_reopen_queue_child(BlockReopenQueue *bs_queue, bdrv_join_options(bs, options, old_options); qobject_unref(old_options); + /* We have the final set of options so let's update the flags */ + options_copy = qdict_clone_shallow(options); + opts = qemu_opts_create(&bdrv_runtime_opts, NULL, 0, &error_abort); + qemu_opts_absorb_qdict(opts, options_copy, NULL); + update_flags_from_options(&flags, opts); + qemu_opts_del(opts); + qobject_unref(options_copy); + /* bdrv_open_inherit() sets and clears some additional flags internally */ flags &= ~BDRV_O_PROTOCOL; if (flags & BDRV_O_RDWR) { @@ -XXX,XX +XXX,XX @@ static BlockReopenQueue *bdrv_reopen_queue_child(BlockReopenQueue *bs_queue, qdict_extract_subqdict(options, &new_child_options, child_key_dot); g_free(child_key_dot); - bdrv_reopen_queue_child(bs_queue, child->bs, new_child_options, 0, + bdrv_reopen_queue_child(bs_queue, child->bs, new_child_options, child->role, options, flags); } @@ -XXX,XX +XXX,XX @@ BlockReopenQueue *bdrv_reopen_queue(BlockReopenQueue *bs_queue, BlockDriverState *bs, QDict *options) { - int flags = bdrv_get_flags(bs); - return bdrv_reopen_queue_child(bs_queue, bs, options, flags, - NULL, NULL, 0); + return bdrv_reopen_queue_child(bs_queue, bs, options, NULL, NULL, 0); } /* -- 2.19.2 From: Alberto Garcia <berto@igalia.com> This function takes four options (cache.direct, cache.no-flush, read-only and auto-read-only) from a QemuOpts object and updates the flags accordingly. If any of those options is not set (because it was missing from the original QDict or because it had an invalid value) then the function aborts with a failed assertion: $ qemu-io -c 'reopen -o read-only=foo' hd.qcow2 block.c:1126: update_flags_from_options: Assertion `qemu_opt_find(opts, BDRV_OPT_CACHE_DIRECT)' failed. Aborted This assertion is unnecessary, and it forces any caller of bdrv_reopen() to pass all the aforementioned four options. This may have made sense in order to remove ambiguity when bdrv_reopen() was taking both flags and options, but that's not the case anymore. It's also unnecessary if we want to validate the option values, because bdrv_reopen_prepare() already takes care of that, as we can see if we remove the assertions: $ qemu-io -c 'reopen -o read-only=foo' hd.qcow2 Parameter 'read-only' expects 'on' or 'off' Signed-off-by: Alberto Garcia <berto@igalia.com> Reviewed-by: Max Reitz <mreitz@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block.c | 4 ---- tests/qemu-iotests/133 | 9 +++++++++ tests/qemu-iotests/133.out | 7 +++++++ 3 files changed, 16 insertions(+), 4 deletions(-) diff --git a/block.c b/block.c index XXXXXXX..XXXXXXX 100644 --- a/block.c +++ b/block.c @@ -XXX,XX +XXX,XX @@ static void update_flags_from_options(int *flags, QemuOpts *opts) { *flags &= ~(BDRV_O_CACHE_MASK | BDRV_O_RDWR | BDRV_O_AUTO_RDONLY); - assert(qemu_opt_find(opts, BDRV_OPT_CACHE_NO_FLUSH)); if (qemu_opt_get_bool_del(opts, BDRV_OPT_CACHE_NO_FLUSH, false)) { *flags |= BDRV_O_NO_FLUSH; } - assert(qemu_opt_find(opts, BDRV_OPT_CACHE_DIRECT)); if (qemu_opt_get_bool_del(opts, BDRV_OPT_CACHE_DIRECT, false)) { *flags |= BDRV_O_NOCACHE; } - assert(qemu_opt_find(opts, BDRV_OPT_READ_ONLY)); if (!qemu_opt_get_bool_del(opts, BDRV_OPT_READ_ONLY, false)) { *flags |= BDRV_O_RDWR; } - assert(qemu_opt_find(opts, BDRV_OPT_AUTO_READ_ONLY)); if (qemu_opt_get_bool_del(opts, BDRV_OPT_AUTO_READ_ONLY, false)) { *flags |= BDRV_O_AUTO_RDONLY; } diff --git a/tests/qemu-iotests/133 b/tests/qemu-iotests/133 index XXXXXXX..XXXXXXX 100755 --- a/tests/qemu-iotests/133 +++ b/tests/qemu-iotests/133 @@ -XXX,XX +XXX,XX @@ $QEMU_IO -c 'reopen -w -o read-only=on' $TEST_IMG $QEMU_IO -c 'reopen -c none -o cache.direct=on' $TEST_IMG $QEMU_IO -c 'reopen -c writeback -o cache.direct=on' $TEST_IMG $QEMU_IO -c 'reopen -c directsync -o cache.no-flush=on' $TEST_IMG + +echo +echo "=== Check that invalid options are handled correctly ===" +echo + +$QEMU_IO -c 'reopen -o read-only=foo' $TEST_IMG +$QEMU_IO -c 'reopen -o cache.no-flush=bar' $TEST_IMG +$QEMU_IO -c 'reopen -o cache.direct=baz' $TEST_IMG +$QEMU_IO -c 'reopen -o auto-read-only=qux' $TEST_IMG # success, all done echo "*** done" rm -f $seq.full diff --git a/tests/qemu-iotests/133.out b/tests/qemu-iotests/133.out index XXXXXXX..XXXXXXX 100644 --- a/tests/qemu-iotests/133.out +++ b/tests/qemu-iotests/133.out @@ -XXX,XX +XXX,XX @@ Cannot set both -r/-w and 'read-only' Cannot set both -c and the cache options Cannot set both -c and the cache options Cannot set both -c and the cache options + +=== Check that invalid options are handled correctly === + +Parameter 'read-only' expects 'on' or 'off' +Parameter 'cache.no-flush' expects 'on' or 'off' +Parameter 'cache.direct' expects 'on' or 'off' +Parameter 'auto-read-only' expects 'on' or 'off' *** done -- 2.19.2 From: Alberto Garcia <berto@igalia.com> Towards the end of bdrv_reopen_queue_child(), before starting to process the children, the update_flags_from_options() function is called in order to have BDRVReopenState.flags in sync with the options from the QDict. This is necessary because during the reopen process flags must be updated for all nodes in the queue so bdrv_is_writable_after_reopen() and the permission checks work correctly. Because of that, calling update_flags_from_options() again in bdrv_reopen_prepare() doesn't really change the flags (they are already up-to-date). But we need to call it in order to remove those options from QemuOpts and that way indicate that they have been processed. Signed-off-by: Alberto Garcia <berto@igalia.com> Reviewed-by: Max Reitz <mreitz@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/block.c b/block.c index XXXXXXX..XXXXXXX 100644 --- a/block.c +++ b/block.c @@ -XXX,XX +XXX,XX @@ int bdrv_reopen_prepare(BDRVReopenState *reopen_state, BlockReopenQueue *queue, Error **errp) { int ret = -1; + int old_flags; Error *local_err = NULL; BlockDriver *drv; QemuOpts *opts; @@ -XXX,XX +XXX,XX @@ int bdrv_reopen_prepare(BDRVReopenState *reopen_state, BlockReopenQueue *queue, goto error; } + /* This was already called in bdrv_reopen_queue_child() so the flags + * are up-to-date. This time we simply want to remove the options from + * QemuOpts in order to indicate that they have been processed. */ + old_flags = reopen_state->flags; update_flags_from_options(&reopen_state->flags, opts); + assert(old_flags == reopen_state->flags); discard = qemu_opt_get_del(opts, BDRV_OPT_DISCARD); if (discard != NULL) { -- 2.19.2 From: Christian Borntraeger <borntraeger@de.ibm.com> "-machine pc" will not work all architectures. Lets fall back to the default machine by not specifying it. In addition we also need to specify -no-shutdown on s390 as qemu will exit otherwise. Cc: qemu-stable@nongnu.org Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com> Reviewed-by: Eric Blake <eblake@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- tests/qemu-iotests/235 | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/tests/qemu-iotests/235 b/tests/qemu-iotests/235 index XXXXXXX..XXXXXXX 100755 --- a/tests/qemu-iotests/235 +++ b/tests/qemu-iotests/235 @@ -XXX,XX +XXX,XX @@ qemu_img_create('-f', iotests.imgfmt, '-o', 'preallocation=metadata', disk, str(size)) vm = QEMUMachine(iotests.qemu_prog) -vm.add_args('-machine', 'pc,accel=kvm') +vm.add_args('-machine', 'accel=kvm') +if iotests.qemu_default_machine == 's390-ccw-virtio': + vm.add_args('-no-shutdown') vm.add_args('-drive', 'id=src,file=' + disk) vm.launch() -- 2.19.2 The following changes since commit b2896c1b09878fd1c4b485b3662f8beecbe0fef4: Merge tag 'vfio-updates-20230509.0' of https://gitlab.com/alex.williamson/qemu into staging (2023-05-10 11:20:35 +0100) are available in the Git repository at: https://repo.or.cz/qemu/kevin.git tags/for-upstream for you to fetch changes up to 58a2e3f5c37be02dac3086b81bdda9414b931edf: block: compile out assert_bdrv_graph_readable() by default (2023-05-10 14:16:54 +0200) ---------------------------------------------------------------- Block layer patches - Graph locking, part 3 (more block drivers) - Compile out assert_bdrv_graph_readable() by default - Add configure options for vmdk, vhdx and vpc - Fix use after free in blockdev_mark_auto_del() - migration: Attempt disk reactivation in more failure scenarios - Coroutine correctness fixes ---------------------------------------------------------------- Emanuele Giuseppe Esposito (5): nbd: Mark nbd_co_do_establish_connection() and callers GRAPH_RDLOCK block: Mark bdrv_co_get_allocated_file_size() and callers GRAPH_RDLOCK block: Mark bdrv_co_get_info() and callers GRAPH_RDLOCK block: Mark bdrv_co_debug_event() GRAPH_RDLOCK block: Mark BlockDriver callbacks for amend job GRAPH_RDLOCK Eric Blake (1): migration: Attempt disk reactivation in more failure scenarios Kevin Wolf (18): block: Fix use after free in blockdev_mark_auto_del() iotests/nbd-reconnect-on-open: Fix NBD socket path qcow2: Don't call bdrv_getlength() in coroutine_fns block: Consistently call bdrv_activate() outside coroutine block: bdrv/blk_co_unref() for calls in coroutine context block: Don't call no_coroutine_fns in qmp_block_resize() iotests: Test resizing image attached to an iothread test-bdrv-drain: Don't modify the graph in coroutines graph-lock: Add GRAPH_UNLOCKED(_PTR) graph-lock: Fix GRAPH_RDLOCK_GUARD*() to be reader lock block: .bdrv_open is non-coroutine and unlocked nbd: Remove nbd_co_flush() wrapper function vhdx: Require GRAPH_RDLOCK for accessing a node's parent list mirror: Require GRAPH_RDLOCK for accessing a node's parent list block: Mark bdrv_query_bds_stats() and callers GRAPH_RDLOCK block: Mark bdrv_query_block_graph_info() and callers GRAPH_RDLOCK block: Mark bdrv_recurse_can_replace() and callers GRAPH_RDLOCK block: Mark bdrv_refresh_limits() and callers GRAPH_RDLOCK Paolo Bonzini (1): block: add missing coroutine_fn annotations Stefan Hajnoczi (2): aio-wait: avoid AioContext lock in aio_wait_bh_oneshot() block: compile out assert_bdrv_graph_readable() by default Vladimir Sementsov-Ogievskiy (1): block: add configure options for excluding vmdk, vhdx and vpc meson_options.txt | 8 ++ configure | 1 + block/coroutines.h | 5 +- block/qcow2.h | 4 +- include/block/aio-wait.h | 2 +- include/block/block-global-state.h | 19 +++- include/block/block-io.h | 23 +++-- include/block/block_int-common.h | 37 +++---- include/block/block_int-global-state.h | 4 +- include/block/graph-lock.h | 20 ++-- include/block/qapi.h | 7 +- include/sysemu/block-backend-global-state.h | 5 +- block.c | 25 ++++- block/amend.c | 8 +- block/blkverify.c | 5 +- block/block-backend.c | 10 +- block/crypto.c | 8 +- block/graph-lock.c | 3 + block/io.c | 12 +-- block/mirror.c | 18 +++- block/nbd.c | 50 +++++---- block/parallels.c | 6 +- block/qapi.c | 6 +- block/qcow.c | 6 +- block/qcow2-refcount.c | 2 +- block/qcow2.c | 48 ++++----- block/qed.c | 24 ++--- block/quorum.c | 4 +- block/raw-format.c | 2 +- block/vdi.c | 6 +- block/vhdx.c | 15 +-- block/vmdk.c | 20 ++-- block/vpc.c | 6 +- blockdev.c | 25 +++-- hw/block/dataplane/virtio-blk.c | 3 +- hw/scsi/virtio-scsi-dataplane.c | 2 - migration/migration.c | 24 +++-- qemu-img.c | 2 + tests/unit/test-bdrv-drain.c | 112 ++++++++++++++------- util/aio-wait.c | 2 +- block/meson.build | 18 +++- meson.build | 5 + scripts/meson-buildoptions.sh | 13 +++ tests/qemu-iotests/tests/iothreads-resize | 71 +++++++++++++ tests/qemu-iotests/tests/iothreads-resize.out | 11 ++ tests/qemu-iotests/tests/nbd-reconnect-on-open | 3 +- tests/qemu-iotests/tests/nbd-reconnect-on-open.out | 4 +- 47 files changed, 474 insertions(+), 240 deletions(-) create mode 100755 tests/qemu-iotests/tests/iothreads-resize create mode 100644 tests/qemu-iotests/tests/iothreads-resize.out From: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru> Let's add --enable / --disable configure options for these formats, so that those who don't need them may not build them. Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru> Message-Id: <20230421092758.814122-1-vsementsov@yandex-team.ru> Reviewed-by: Kevin Wolf <kwolf@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- meson_options.txt | 6 ++++++ block/meson.build | 18 +++++++++++++----- meson.build | 3 +++ scripts/meson-buildoptions.sh | 9 +++++++++ 4 files changed, 31 insertions(+), 5 deletions(-) diff --git a/meson_options.txt b/meson_options.txt index XXXXXXX..XXXXXXX 100644 --- a/meson_options.txt +++ b/meson_options.txt @@ -XXX,XX +XXX,XX @@ option('qcow1', type: 'feature', value: 'auto', description: 'qcow1 image format support') option('vdi', type: 'feature', value: 'auto', description: 'vdi image format support') +option('vhdx', type: 'feature', value: 'auto', + description: 'vhdx image format support') +option('vmdk', type: 'feature', value: 'auto', + description: 'vmdk image format support') +option('vpc', type: 'feature', value: 'auto', + description: 'vpc image format support') option('vvfat', type: 'feature', value: 'auto', description: 'vvfat image format support') option('qed', type: 'feature', value: 'auto', diff --git a/block/meson.build b/block/meson.build index XXXXXXX..XXXXXXX 100644 --- a/block/meson.build +++ b/block/meson.build @@ -XXX,XX +XXX,XX @@ block_ss.add(files( 'snapshot-access.c', 'throttle-groups.c', 'throttle.c', - 'vhdx-endian.c', - 'vhdx-log.c', - 'vhdx.c', - 'vmdk.c', - 'vpc.c', 'write-threshold.c', ), zstd, zlib, gnutls) @@ -XXX,XX +XXX,XX @@ endif if get_option('vdi').allowed() block_ss.add(files('vdi.c')) endif +if get_option('vhdx').allowed() + block_ss.add(files( + 'vhdx-endian.c', + 'vhdx-log.c', + 'vhdx.c' + )) +endif +if get_option('vmdk').allowed() + block_ss.add(files('vmdk.c')) +endif +if get_option('vpc').allowed() + block_ss.add(files('vpc.c')) +endif if get_option('cloop').allowed() block_ss.add(files('cloop.c')) endif diff --git a/meson.build b/meson.build index XXXXXXX..XXXXXXX 100644 --- a/meson.build +++ b/meson.build @@ -XXX,XX +XXX,XX @@ if have_block summary_info += {'dmg support': get_option('dmg').allowed()} summary_info += {'qcow v1 support': get_option('qcow1').allowed()} summary_info += {'vdi support': get_option('vdi').allowed()} + summary_info += {'vhdx support': get_option('vhdx').allowed()} + summary_info += {'vmdk support': get_option('vmdk').allowed()} + summary_info += {'vpc support': get_option('vpc').allowed()} summary_info += {'vvfat support': get_option('vvfat').allowed()} summary_info += {'qed support': get_option('qed').allowed()} summary_info += {'parallels support': get_option('parallels').allowed()} diff --git a/scripts/meson-buildoptions.sh b/scripts/meson-buildoptions.sh index XXXXXXX..XXXXXXX 100644 --- a/scripts/meson-buildoptions.sh +++ b/scripts/meson-buildoptions.sh @@ -XXX,XX +XXX,XX @@ meson_options_help() { printf "%s\n" ' VDUSE block export support' printf "%s\n" ' vfio-user-server' printf "%s\n" ' vfio-user server support' + printf "%s\n" ' vhdx vhdx image format support' printf "%s\n" ' vhost-crypto vhost-user crypto backend support' printf "%s\n" ' vhost-kernel vhost kernel backend support' printf "%s\n" ' vhost-net vhost-net kernel acceleration support' @@ -XXX,XX +XXX,XX @@ meson_options_help() { printf "%s\n" ' vhost-vdpa vhost-vdpa kernel backend support' printf "%s\n" ' virglrenderer virgl rendering support' printf "%s\n" ' virtfs virtio-9p support' + printf "%s\n" ' vmdk vmdk image format support' printf "%s\n" ' vmnet vmnet.framework network backend support' printf "%s\n" ' vnc VNC server' printf "%s\n" ' vnc-jpeg JPEG lossy compression for VNC server' printf "%s\n" ' vnc-sasl SASL authentication for VNC server' + printf "%s\n" ' vpc vpc image format support' printf "%s\n" ' vte vte support for the gtk UI' printf "%s\n" ' vvfat vvfat image format support' printf "%s\n" ' whpx WHPX acceleration support' @@ -XXX,XX +XXX,XX @@ _meson_option_parse() { --disable-vduse-blk-export) printf "%s" -Dvduse_blk_export=disabled ;; --enable-vfio-user-server) printf "%s" -Dvfio_user_server=enabled ;; --disable-vfio-user-server) printf "%s" -Dvfio_user_server=disabled ;; + --enable-vhdx) printf "%s" -Dvhdx=enabled ;; + --disable-vhdx) printf "%s" -Dvhdx=disabled ;; --enable-vhost-crypto) printf "%s" -Dvhost_crypto=enabled ;; --disable-vhost-crypto) printf "%s" -Dvhost_crypto=disabled ;; --enable-vhost-kernel) printf "%s" -Dvhost_kernel=enabled ;; @@ -XXX,XX +XXX,XX @@ _meson_option_parse() { --disable-virglrenderer) printf "%s" -Dvirglrenderer=disabled ;; --enable-virtfs) printf "%s" -Dvirtfs=enabled ;; --disable-virtfs) printf "%s" -Dvirtfs=disabled ;; + --enable-vmdk) printf "%s" -Dvmdk=enabled ;; + --disable-vmdk) printf "%s" -Dvmdk=disabled ;; --enable-vmnet) printf "%s" -Dvmnet=enabled ;; --disable-vmnet) printf "%s" -Dvmnet=disabled ;; --enable-vnc) printf "%s" -Dvnc=enabled ;; @@ -XXX,XX +XXX,XX @@ _meson_option_parse() { --disable-vnc-jpeg) printf "%s" -Dvnc_jpeg=disabled ;; --enable-vnc-sasl) printf "%s" -Dvnc_sasl=enabled ;; --disable-vnc-sasl) printf "%s" -Dvnc_sasl=disabled ;; + --enable-vpc) printf "%s" -Dvpc=enabled ;; + --disable-vpc) printf "%s" -Dvpc=disabled ;; --enable-vte) printf "%s" -Dvte=enabled ;; --disable-vte) printf "%s" -Dvte=disabled ;; --enable-vvfat) printf "%s" -Dvvfat=enabled ;; -- 2.40.1 From: Paolo Bonzini <pbonzini@redhat.com> After the recent introduction of many new coroutine callbacks, a couple calls from non-coroutine_fn to coroutine_fn have sneaked in; fix them. Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Message-Id: <20230406101752.242125-1-pbonzini@redhat.com> Reviewed-by: Kevin Wolf <kwolf@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- include/block/graph-lock.h | 4 ++-- block/mirror.c | 4 ++-- 2 files changed, 4 insertions(+), 4 deletions(-) diff --git a/include/block/graph-lock.h b/include/block/graph-lock.h index XXXXXXX..XXXXXXX 100644 --- a/include/block/graph-lock.h +++ b/include/block/graph-lock.h @@ -XXX,XX +XXX,XX @@ typedef struct GraphLockable { } GraphLockable; * unlocked. TSA_ASSERT() makes sure that the following calls know that we * hold the lock while unlocking is left unchecked. */ -static inline GraphLockable * TSA_ASSERT(graph_lock) TSA_NO_TSA +static inline GraphLockable * TSA_ASSERT(graph_lock) TSA_NO_TSA coroutine_fn graph_lockable_auto_lock(GraphLockable *x) { bdrv_graph_co_rdlock(); return x; } -static inline void TSA_NO_TSA +static inline void TSA_NO_TSA coroutine_fn graph_lockable_auto_unlock(GraphLockable *x) { bdrv_graph_co_rdunlock(); diff --git a/block/mirror.c b/block/mirror.c index XXXXXXX..XXXXXXX 100644 --- a/block/mirror.c +++ b/block/mirror.c @@ -XXX,XX +XXX,XX @@ static inline int64_t mirror_clip_bytes(MirrorBlockJob *s, /* Round offset and/or bytes to target cluster if COW is needed, and * return the offset of the adjusted tail against original. */ -static int mirror_cow_align(MirrorBlockJob *s, int64_t *offset, - uint64_t *bytes) +static int coroutine_fn mirror_cow_align(MirrorBlockJob *s, int64_t *offset, + uint64_t *bytes) { bool need_cow; int ret = 0; -- 2.40.1 From: Stefan Hajnoczi <stefanha@redhat.com> There is no need for the AioContext lock in aio_wait_bh_oneshot(). It's easy to remove the lock from existing callers and then switch from AIO_WAIT_WHILE() to AIO_WAIT_WHILE_UNLOCKED() in aio_wait_bh_oneshot(). Document that the AioContext lock should not be held across aio_wait_bh_oneshot(). Holding a lock across aio_poll() can cause deadlock so we don't want callers to do that. This is a step towards getting rid of the AioContext lock. Cc: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> Message-Id: <20230404153307.458883-1-stefanha@redhat.com> Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> Reviewed-by: Emanuele Giuseppe Esposito <eesposit@redhat.com> Reviewed-by: Kevin Wolf <kwolf@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- include/block/aio-wait.h | 2 +- hw/block/dataplane/virtio-blk.c | 3 ++- hw/scsi/virtio-scsi-dataplane.c | 2 -- util/aio-wait.c | 2 +- 4 files changed, 4 insertions(+), 5 deletions(-) diff --git a/include/block/aio-wait.h b/include/block/aio-wait.h index XXXXXXX..XXXXXXX 100644 --- a/include/block/aio-wait.h +++ b/include/block/aio-wait.h @@ -XXX,XX +XXX,XX @@ void aio_wait_kick(void); * * Run a BH in @ctx and wait for it to complete. * - * Must be called from the main loop thread with @ctx acquired exactly once. + * Must be called from the main loop thread without @ctx acquired. * Note that main loop event processing may occur. */ void aio_wait_bh_oneshot(AioContext *ctx, QEMUBHFunc *cb, void *opaque); diff --git a/hw/block/dataplane/virtio-blk.c b/hw/block/dataplane/virtio-blk.c index XXXXXXX..XXXXXXX 100644 --- a/hw/block/dataplane/virtio-blk.c +++ b/hw/block/dataplane/virtio-blk.c @@ -XXX,XX +XXX,XX @@ void virtio_blk_data_plane_stop(VirtIODevice *vdev) s->stopping = true; trace_virtio_blk_data_plane_stop(s); - aio_context_acquire(s->ctx); aio_wait_bh_oneshot(s->ctx, virtio_blk_data_plane_stop_bh, s); + aio_context_acquire(s->ctx); + /* Wait for virtio_blk_dma_restart_bh() and in flight I/O to complete */ blk_drain(s->conf->conf.blk); diff --git a/hw/scsi/virtio-scsi-dataplane.c b/hw/scsi/virtio-scsi-dataplane.c index XXXXXXX..XXXXXXX 100644 --- a/hw/scsi/virtio-scsi-dataplane.c +++ b/hw/scsi/virtio-scsi-dataplane.c @@ -XXX,XX +XXX,XX @@ void virtio_scsi_dataplane_stop(VirtIODevice *vdev) } s->dataplane_stopping = true; - aio_context_acquire(s->ctx); aio_wait_bh_oneshot(s->ctx, virtio_scsi_dataplane_stop_bh, s); - aio_context_release(s->ctx); blk_drain_all(); /* ensure there are no in-flight requests */ diff --git a/util/aio-wait.c b/util/aio-wait.c index XXXXXXX..XXXXXXX 100644 --- a/util/aio-wait.c +++ b/util/aio-wait.c @@ -XXX,XX +XXX,XX @@ void aio_wait_bh_oneshot(AioContext *ctx, QEMUBHFunc *cb, void *opaque) assert(qemu_get_current_aio_context() == qemu_get_aio_context()); aio_bh_schedule_oneshot(ctx, aio_wait_bh, &data); - AIO_WAIT_WHILE(ctx, !data.done); + AIO_WAIT_WHILE_UNLOCKED(NULL, !data.done); } -- 2.40.1 job_cancel_locked() drops the job list lock temporarily and it may call aio_poll(). We must assume that the list has changed after this call. Also, with unlucky timing, it can end up freeing the job during job_completed_txn_abort_locked(), making the job pointer invalid, too. For both reasons, we can't just continue at block_job_next_locked(job). Instead, start at the head of the list again after job_cancel_locked() and skip those jobs that we already cancelled (or that are completing anyway). Cc: qemu-stable@nongnu.org Signed-off-by: Kevin Wolf <kwolf@redhat.com> Message-Id: <20230503140142.474404-1-kwolf@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- blockdev.c | 18 ++++++++++++++---- 1 file changed, 14 insertions(+), 4 deletions(-) diff --git a/blockdev.c b/blockdev.c index XXXXXXX..XXXXXXX 100644 --- a/blockdev.c +++ b/blockdev.c @@ -XXX,XX +XXX,XX @@ void blockdev_mark_auto_del(BlockBackend *blk) JOB_LOCK_GUARD(); - for (job = block_job_next_locked(NULL); job; - job = block_job_next_locked(job)) { - if (block_job_has_bdrv(job, blk_bs(blk))) { + do { + job = block_job_next_locked(NULL); + while (job && (job->job.cancelled || + job->job.deferred_to_main_loop || + !block_job_has_bdrv(job, blk_bs(blk)))) + { + job = block_job_next_locked(job); + } + if (job) { + /* + * This drops the job lock temporarily and polls, so we need to + * restart processing the list from the start after this. + */ job_cancel_locked(&job->job, false); } - } + } while (job); dinfo->auto_del = 1; } -- 2.40.1 Socket paths need to be short to avoid failures. This is why there is a iotests.sock_dir (defaulting to /tmp) separate from the disk image base directory. Make use of it to fix failures in too deeply nested test directories. Fixes: ab7f7e67a7e7b49964109501dfcde4ec29bae60e Signed-off-by: Kevin Wolf <kwolf@redhat.com> Message-Id: <20230503165019.8867-1-kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- tests/qemu-iotests/tests/nbd-reconnect-on-open | 3 ++- tests/qemu-iotests/tests/nbd-reconnect-on-open.out | 4 ++-- 2 files changed, 4 insertions(+), 3 deletions(-) diff --git a/tests/qemu-iotests/tests/nbd-reconnect-on-open b/tests/qemu-iotests/tests/nbd-reconnect-on-open index XXXXXXX..XXXXXXX 100755 --- a/tests/qemu-iotests/tests/nbd-reconnect-on-open +++ b/tests/qemu-iotests/tests/nbd-reconnect-on-open @@ -XXX,XX +XXX,XX @@ from iotests import qemu_img_create, file_path, qemu_io_popen, qemu_nbd, \ iotests.script_initialize(supported_fmts=['qcow2']) -disk, nbd_sock = file_path('disk', 'nbd-sock') +disk = file_path('disk') +nbd_sock = file_path('nbd-sock', base_dir=iotests.sock_dir) def create_args(open_timeout): diff --git a/tests/qemu-iotests/tests/nbd-reconnect-on-open.out b/tests/qemu-iotests/tests/nbd-reconnect-on-open.out index XXXXXXX..XXXXXXX 100644 --- a/tests/qemu-iotests/tests/nbd-reconnect-on-open.out +++ b/tests/qemu-iotests/tests/nbd-reconnect-on-open.out @@ -XXX,XX +XXX,XX @@ read 1048576/1048576 bytes at offset 0 1 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) Check fail to connect with 0 seconds of timeout -qemu-io: can't open: Failed to connect to 'TEST_DIR/PID-nbd-sock': No such file or directory +qemu-io: can't open: Failed to connect to 'SOCK_DIR/PID-nbd-sock': No such file or directory qemu_io finished in 0..0.2 seconds, OK Check fail to connect with 1 seconds of timeout -qemu-io: can't open: Failed to connect to 'TEST_DIR/PID-nbd-sock': No such file or directory +qemu-io: can't open: Failed to connect to 'SOCK_DIR/PID-nbd-sock': No such file or directory qemu_io finished in 1..1.2 seconds, OK -- 2.40.1 From: Eric Blake <eblake@redhat.com> Commit fe904ea824 added a fail_inactivate label, which tries to reactivate disks on the source after a failure while s->state == MIGRATION_STATUS_ACTIVE, but didn't actually use the label if qemu_savevm_state_complete_precopy() failed. This failure to reactivate is also present in commit 6039dd5b1c (also covering the new s->state == MIGRATION_STATUS_DEVICE state) and 403d18ae (ensuring s->block_inactive is set more reliably). Consolidate the two labels back into one - no matter HOW migration is failed, if there is any chance we can reach vm_start() after having attempted inactivation, it is essential that we have tried to restart disks before then. This also makes the cleanup more like migrate_fd_cancel(). Suggested-by: Kevin Wolf <kwolf@redhat.com> Signed-off-by: Eric Blake <eblake@redhat.com> Message-Id: <20230502205212.134680-1-eblake@redhat.com> Acked-by: Peter Xu <peterx@redhat.com> Reviewed-by: Juan Quintela <quintela@redhat.com> Reviewed-by: Kevin Wolf <kwolf@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- migration/migration.c | 24 ++++++++++++++---------- 1 file changed, 14 insertions(+), 10 deletions(-) diff --git a/migration/migration.c b/migration/migration.c index XXXXXXX..XXXXXXX 100644 --- a/migration/migration.c +++ b/migration/migration.c @@ -XXX,XX +XXX,XX @@ static void migration_completion(MigrationState *s) MIGRATION_STATUS_DEVICE); } if (ret >= 0) { + /* + * Inactivate disks except in COLO, and track that we + * have done so in order to remember to reactivate + * them if migration fails or is cancelled. + */ s->block_inactive = !migrate_colo(); qemu_file_set_rate_limit(s->to_dst_file, INT64_MAX); ret = qemu_savevm_state_complete_precopy(s->to_dst_file, false, @@ -XXX,XX +XXX,XX @@ static void migration_completion(MigrationState *s) rp_error = await_return_path_close_on_source(s); trace_migration_return_path_end_after(rp_error); if (rp_error) { - goto fail_invalidate; + goto fail; } } if (qemu_file_get_error(s->to_dst_file)) { trace_migration_completion_file_err(); - goto fail_invalidate; + goto fail; } if (migrate_colo() && s->state == MIGRATION_STATUS_ACTIVE) { @@ -XXX,XX +XXX,XX @@ static void migration_completion(MigrationState *s) return; -fail_invalidate: - /* If not doing postcopy, vm_start() will be called: let's regain - * control on images. - */ - if (s->state == MIGRATION_STATUS_ACTIVE || - s->state == MIGRATION_STATUS_DEVICE) { +fail: + if (s->block_inactive && (s->state == MIGRATION_STATUS_ACTIVE || + s->state == MIGRATION_STATUS_DEVICE)) { + /* + * If not doing postcopy, vm_start() will be called: let's + * regain control on images. + */ Error *local_err = NULL; qemu_mutex_lock_iothread(); bdrv_activate_all(&local_err); if (local_err) { error_report_err(local_err); - s->block_inactive = true; } else { s->block_inactive = false; } qemu_mutex_unlock_iothread(); } -fail: migrate_set_state(&s->state, current_active_state, MIGRATION_STATUS_FAILED); } -- 2.40.1 There is a bdrv_co_getlength() now, which should be used in coroutine context. This requires adding GRAPH_RDLOCK to some functions so that this still compiles with TSA because bdrv_co_getlength() is GRAPH_RDLOCK. Signed-off-by: Kevin Wolf <kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Message-Id: <20230504115750.54437-2-kwolf@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/qcow2.h | 4 +++- block/qcow2-refcount.c | 2 +- block/qcow2.c | 19 +++++++++---------- 3 files changed, 13 insertions(+), 12 deletions(-) diff --git a/block/qcow2.h b/block/qcow2.h index XXXXXXX..XXXXXXX 100644 --- a/block/qcow2.h +++ b/block/qcow2.h @@ -XXX,XX +XXX,XX @@ int qcow2_change_refcount_order(BlockDriverState *bs, int refcount_order, void *cb_opaque, Error **errp); int coroutine_fn GRAPH_RDLOCK qcow2_shrink_reftable(BlockDriverState *bs); int64_t coroutine_fn qcow2_get_last_cluster(BlockDriverState *bs, int64_t size); -int coroutine_fn qcow2_detect_metadata_preallocation(BlockDriverState *bs); + +int coroutine_fn GRAPH_RDLOCK +qcow2_detect_metadata_preallocation(BlockDriverState *bs); /* qcow2-cluster.c functions */ int qcow2_grow_l1_table(BlockDriverState *bs, uint64_t min_size, diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c index XXXXXXX..XXXXXXX 100644 --- a/block/qcow2-refcount.c +++ b/block/qcow2-refcount.c @@ -XXX,XX +XXX,XX @@ int coroutine_fn qcow2_detect_metadata_preallocation(BlockDriverState *bs) qemu_co_mutex_assert_locked(&s->lock); - file_length = bdrv_getlength(bs->file->bs); + file_length = bdrv_co_getlength(bs->file->bs); if (file_length < 0) { return file_length; } diff --git a/block/qcow2.c b/block/qcow2.c index XXXXXXX..XXXXXXX 100644 --- a/block/qcow2.c +++ b/block/qcow2.c @@ -XXX,XX +XXX,XX @@ static void qcow2_join_options(QDict *options, QDict *old_options) } } -static int coroutine_fn qcow2_co_block_status(BlockDriverState *bs, - bool want_zero, - int64_t offset, int64_t count, - int64_t *pnum, int64_t *map, - BlockDriverState **file) +static int coroutine_fn GRAPH_RDLOCK +qcow2_co_block_status(BlockDriverState *bs, bool want_zero, int64_t offset, + int64_t count, int64_t *pnum, int64_t *map, + BlockDriverState **file) { BDRVQcow2State *s = bs->opaque; uint64_t host_offset; @@ -XXX,XX +XXX,XX @@ preallocate_co(BlockDriverState *bs, uint64_t offset, uint64_t new_length, * all of the allocated clusters (otherwise we get failing reads after * EOF). Extend the image to the last allocated sector. */ - file_length = bdrv_getlength(s->data_file->bs); + file_length = bdrv_co_getlength(s->data_file->bs); if (file_length < 0) { error_setg_errno(errp, -file_length, "Could not get file size"); ret = file_length; @@ -XXX,XX +XXX,XX @@ qcow2_co_copy_range_from(BlockDriverState *bs, case QCOW2_SUBCLUSTER_UNALLOCATED_PLAIN: case QCOW2_SUBCLUSTER_UNALLOCATED_ALLOC: if (bs->backing && bs->backing->bs) { - int64_t backing_length = bdrv_getlength(bs->backing->bs); + int64_t backing_length = bdrv_co_getlength(bs->backing->bs); if (src_offset >= backing_length) { cur_write_flags |= BDRV_REQ_ZERO_WRITE; } else { @@ -XXX,XX +XXX,XX @@ qcow2_co_truncate(BlockDriverState *bs, int64_t offset, bool exact, goto fail; } - old_file_size = bdrv_getlength(bs->file->bs); + old_file_size = bdrv_co_getlength(bs->file->bs); if (old_file_size < 0) { error_setg_errno(errp, -old_file_size, "Failed to inquire current file length"); @@ -XXX,XX +XXX,XX @@ qcow2_co_truncate(BlockDriverState *bs, int64_t offset, bool exact, break; } - old_file_size = bdrv_getlength(bs->file->bs); + old_file_size = bdrv_co_getlength(bs->file->bs); if (old_file_size < 0) { error_setg_errno(errp, -old_file_size, "Failed to inquire current file length"); @@ -XXX,XX +XXX,XX @@ qcow2_co_pwritev_compressed_part(BlockDriverState *bs, * align end of file to a sector boundary to ease reading with * sector based I/Os */ - int64_t len = bdrv_getlength(bs->file->bs); + int64_t len = bdrv_co_getlength(bs->file->bs); if (len < 0) { return len; } -- 2.40.1 Migration code can call bdrv_activate() in coroutine context, whereas other callers call it outside of coroutines. As it calls other code that is not supposed to run in coroutines, standardise on running outside of coroutines. This adds a no_co_wrapper to switch to the main loop before calling bdrv_activate(). Cc: qemu-stable@nongnu.org Signed-off-by: Kevin Wolf <kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Message-Id: <20230504115750.54437-3-kwolf@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- include/block/block-global-state.h | 6 +++++- block/block-backend.c | 10 +++++++++- 2 files changed, 14 insertions(+), 2 deletions(-) diff --git a/include/block/block-global-state.h b/include/block/block-global-state.h index XXXXXXX..XXXXXXX 100644 --- a/include/block/block-global-state.h +++ b/include/block/block-global-state.h @@ -XXX,XX +XXX,XX @@ int bdrv_amend_options(BlockDriverState *bs_new, QemuOpts *opts, BlockDriverState *check_to_replace_node(BlockDriverState *parent_bs, const char *node_name, Error **errp); -int bdrv_activate(BlockDriverState *bs, Error **errp); +int no_coroutine_fn bdrv_activate(BlockDriverState *bs, Error **errp); + +int coroutine_fn no_co_wrapper +bdrv_co_activate(BlockDriverState *bs, Error **errp); + void bdrv_activate_all(Error **errp); int bdrv_inactivate_all(void); diff --git a/block/block-backend.c b/block/block-backend.c index XXXXXXX..XXXXXXX 100644 --- a/block/block-backend.c +++ b/block/block-backend.c @@ -XXX,XX +XXX,XX @@ void blk_activate(BlockBackend *blk, Error **errp) return; } - bdrv_activate(bs, errp); + /* + * Migration code can call this function in coroutine context, so leave + * coroutine context if necessary. + */ + if (qemu_in_coroutine()) { + bdrv_co_activate(bs, errp); + } else { + bdrv_activate(bs, errp); + } } bool coroutine_fn blk_co_is_inserted(BlockBackend *blk) -- 2.40.1 These functions must not be called in coroutine context, because they need write access to the graph. Cc: qemu-stable@nongnu.org Signed-off-by: Kevin Wolf <kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Message-Id: <20230504115750.54437-4-kwolf@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- include/block/block-global-state.h | 3 ++- include/sysemu/block-backend-global-state.h | 5 ++++- block.c | 2 +- block/crypto.c | 6 +++--- block/parallels.c | 6 +++--- block/qcow.c | 6 +++--- block/qcow2.c | 14 +++++++------- block/qed.c | 6 +++--- block/vdi.c | 6 +++--- block/vhdx.c | 6 +++--- block/vmdk.c | 18 +++++++++--------- block/vpc.c | 6 +++--- 12 files changed, 44 insertions(+), 40 deletions(-) diff --git a/include/block/block-global-state.h b/include/block/block-global-state.h index XXXXXXX..XXXXXXX 100644 --- a/include/block/block-global-state.h +++ b/include/block/block-global-state.h @@ -XXX,XX +XXX,XX @@ void bdrv_img_create(const char *filename, const char *fmt, bool quiet, Error **errp); void bdrv_ref(BlockDriverState *bs); -void bdrv_unref(BlockDriverState *bs); +void no_coroutine_fn bdrv_unref(BlockDriverState *bs); +void coroutine_fn no_co_wrapper bdrv_co_unref(BlockDriverState *bs); void bdrv_unref_child(BlockDriverState *parent, BdrvChild *child); BdrvChild *bdrv_attach_child(BlockDriverState *parent_bs, BlockDriverState *child_bs, diff --git a/include/sysemu/block-backend-global-state.h b/include/sysemu/block-backend-global-state.h index XXXXXXX..XXXXXXX 100644 --- a/include/sysemu/block-backend-global-state.h +++ b/include/sysemu/block-backend-global-state.h @@ -XXX,XX +XXX,XX @@ blk_co_new_open(const char *filename, const char *reference, QDict *options, int blk_get_refcnt(BlockBackend *blk); void blk_ref(BlockBackend *blk); -void blk_unref(BlockBackend *blk); + +void no_coroutine_fn blk_unref(BlockBackend *blk); +void coroutine_fn no_co_wrapper blk_co_unref(BlockBackend *blk); + void blk_remove_all_bs(void); BlockBackend *blk_by_name(const char *name); BlockBackend *blk_next(BlockBackend *blk); diff --git a/block.c b/block.c index XXXXXXX..XXXXXXX 100644 --- a/block.c +++ b/block.c @@ -XXX,XX +XXX,XX @@ int coroutine_fn bdrv_co_create_opts_simple(BlockDriver *drv, ret = 0; out: - blk_unref(blk); + blk_co_unref(blk); return ret; } diff --git a/block/crypto.c b/block/crypto.c index XXXXXXX..XXXXXXX 100644 --- a/block/crypto.c +++ b/block/crypto.c @@ -XXX,XX +XXX,XX @@ block_crypto_co_create_generic(BlockDriverState *bs, int64_t size, ret = 0; cleanup: qcrypto_block_free(crypto); - blk_unref(blk); + blk_co_unref(blk); return ret; } @@ -XXX,XX +XXX,XX @@ block_crypto_co_create_luks(BlockdevCreateOptions *create_options, Error **errp) ret = 0; fail: - bdrv_unref(bs); + bdrv_co_unref(bs); return ret; } @@ -XXX,XX +XXX,XX @@ fail: bdrv_co_delete_file_noerr(bs); } - bdrv_unref(bs); + bdrv_co_unref(bs); qapi_free_QCryptoBlockCreateOptions(create_opts); qobject_unref(cryptoopts); return ret; diff --git a/block/parallels.c b/block/parallels.c index XXXXXXX..XXXXXXX 100644 --- a/block/parallels.c +++ b/block/parallels.c @@ -XXX,XX +XXX,XX @@ static int coroutine_fn parallels_co_create(BlockdevCreateOptions* opts, ret = 0; out: - blk_unref(blk); - bdrv_unref(bs); + blk_co_unref(blk); + bdrv_co_unref(bs); return ret; exit: @@ -XXX,XX +XXX,XX @@ parallels_co_create_opts(BlockDriver *drv, const char *filename, done: qobject_unref(qdict); - bdrv_unref(bs); + bdrv_co_unref(bs); qapi_free_BlockdevCreateOptions(create_options); return ret; } diff --git a/block/qcow.c b/block/qcow.c index XXXXXXX..XXXXXXX 100644 --- a/block/qcow.c +++ b/block/qcow.c @@ -XXX,XX +XXX,XX @@ static int coroutine_fn qcow_co_create(BlockdevCreateOptions *opts, g_free(tmp); ret = 0; exit: - blk_unref(qcow_blk); - bdrv_unref(bs); + blk_co_unref(qcow_blk); + bdrv_co_unref(bs); qcrypto_block_free(crypto); return ret; } @@ -XXX,XX +XXX,XX @@ qcow_co_create_opts(BlockDriver *drv, const char *filename, fail: g_free(backing_fmt); qobject_unref(qdict); - bdrv_unref(bs); + bdrv_co_unref(bs); qapi_free_BlockdevCreateOptions(create_options); return ret; } diff --git a/block/qcow2.c b/block/qcow2.c index XXXXXXX..XXXXXXX 100644 --- a/block/qcow2.c +++ b/block/qcow2.c @@ -XXX,XX +XXX,XX @@ qcow2_co_create(BlockdevCreateOptions *create_options, Error **errp) goto out; } - blk_unref(blk); + blk_co_unref(blk); blk = NULL; /* @@ -XXX,XX +XXX,XX @@ qcow2_co_create(BlockdevCreateOptions *create_options, Error **errp) } } - blk_unref(blk); + blk_co_unref(blk); blk = NULL; /* Reopen the image without BDRV_O_NO_FLUSH to flush it before returning. @@ -XXX,XX +XXX,XX @@ qcow2_co_create(BlockdevCreateOptions *create_options, Error **errp) ret = 0; out: - blk_unref(blk); - bdrv_unref(bs); - bdrv_unref(data_bs); + blk_co_unref(blk); + bdrv_co_unref(bs); + bdrv_co_unref(data_bs); return ret; } @@ -XXX,XX +XXX,XX @@ finish: } qobject_unref(qdict); - bdrv_unref(bs); - bdrv_unref(data_bs); + bdrv_co_unref(bs); + bdrv_co_unref(data_bs); qapi_free_BlockdevCreateOptions(create_options); return ret; } diff --git a/block/qed.c b/block/qed.c index XXXXXXX..XXXXXXX 100644 --- a/block/qed.c +++ b/block/qed.c @@ -XXX,XX +XXX,XX @@ static int coroutine_fn bdrv_qed_co_create(BlockdevCreateOptions *opts, ret = 0; /* success */ out: g_free(l1_table); - blk_unref(blk); - bdrv_unref(bs); + blk_co_unref(blk); + bdrv_co_unref(bs); return ret; } @@ -XXX,XX +XXX,XX @@ bdrv_qed_co_create_opts(BlockDriver *drv, const char *filename, fail: qobject_unref(qdict); - bdrv_unref(bs); + bdrv_co_unref(bs); qapi_free_BlockdevCreateOptions(create_options); return ret; } diff --git a/block/vdi.c b/block/vdi.c index XXXXXXX..XXXXXXX 100644 --- a/block/vdi.c +++ b/block/vdi.c @@ -XXX,XX +XXX,XX @@ static int coroutine_fn vdi_co_do_create(BlockdevCreateOptions *create_options, ret = 0; exit: - blk_unref(blk); - bdrv_unref(bs_file); + blk_co_unref(blk); + bdrv_co_unref(bs_file); g_free(bmap); return ret; } @@ -XXX,XX +XXX,XX @@ vdi_co_create_opts(BlockDriver *drv, const char *filename, done: qobject_unref(qdict); qapi_free_BlockdevCreateOptions(create_options); - bdrv_unref(bs_file); + bdrv_co_unref(bs_file); return ret; } diff --git a/block/vhdx.c b/block/vhdx.c index XXXXXXX..XXXXXXX 100644 --- a/block/vhdx.c +++ b/block/vhdx.c @@ -XXX,XX +XXX,XX @@ static int coroutine_fn vhdx_co_create(BlockdevCreateOptions *opts, ret = 0; delete_and_exit: - blk_unref(blk); - bdrv_unref(bs); + blk_co_unref(blk); + bdrv_co_unref(bs); g_free(creator); return ret; } @@ -XXX,XX +XXX,XX @@ vhdx_co_create_opts(BlockDriver *drv, const char *filename, fail: qobject_unref(qdict); - bdrv_unref(bs); + bdrv_co_unref(bs); qapi_free_BlockdevCreateOptions(create_options); return ret; } diff --git a/block/vmdk.c b/block/vmdk.c index XXXXXXX..XXXXXXX 100644 --- a/block/vmdk.c +++ b/block/vmdk.c @@ -XXX,XX +XXX,XX @@ exit: if (pbb) { *pbb = blk; } else { - blk_unref(blk); + blk_co_unref(blk); blk = NULL; } } @@ -XXX,XX +XXX,XX @@ vmdk_co_do_create(int64_t size, if (strcmp(blk_bs(backing)->drv->format_name, "vmdk")) { error_setg(errp, "Invalid backing file format: %s. Must be vmdk", blk_bs(backing)->drv->format_name); - blk_unref(backing); + blk_co_unref(backing); ret = -EINVAL; goto exit; } ret = vmdk_read_cid(blk_bs(backing), 0, &parent_cid); - blk_unref(backing); + blk_co_unref(backing); if (ret) { error_setg(errp, "Failed to read parent CID"); goto exit; @@ -XXX,XX +XXX,XX @@ vmdk_co_do_create(int64_t size, blk_bs(extent_blk)->filename); created_size += cur_size; extent_idx++; - blk_unref(extent_blk); + blk_co_unref(extent_blk); } /* Check whether we got excess extents */ extent_blk = extent_fn(-1, extent_idx, flat, split, compress, zeroed_grain, opaque, NULL); if (extent_blk) { - blk_unref(extent_blk); + blk_co_unref(extent_blk); error_setg(errp, "List of extents contains unused extents"); ret = -EINVAL; goto exit; @@ -XXX,XX +XXX,XX @@ vmdk_co_do_create(int64_t size, ret = 0; exit: if (blk) { - blk_unref(blk); + blk_co_unref(blk); } g_free(desc); g_free(parent_desc_line); @@ -XXX,XX +XXX,XX @@ vmdk_co_create_opts_cb(int64_t size, int idx, bool flat, bool split, errp)) { goto exit; } - bdrv_unref(bs); + bdrv_co_unref(bs); exit: g_free(ext_filename); return blk; @@ -XXX,XX +XXX,XX @@ static BlockBackend * coroutine_fn vmdk_co_create_cb(int64_t size, int idx, return NULL; } blk_set_allow_write_beyond_eof(blk, true); - bdrv_unref(bs); + bdrv_co_unref(bs); if (size != -1) { ret = vmdk_init_extent(blk, size, flat, compress, zeroed_grain, errp); if (ret) { - blk_unref(blk); + blk_co_unref(blk); blk = NULL; } } diff --git a/block/vpc.c b/block/vpc.c index XXXXXXX..XXXXXXX 100644 --- a/block/vpc.c +++ b/block/vpc.c @@ -XXX,XX +XXX,XX @@ static int coroutine_fn vpc_co_create(BlockdevCreateOptions *opts, } out: - blk_unref(blk); - bdrv_unref(bs); + blk_co_unref(blk); + bdrv_co_unref(bs); return ret; } @@ -XXX,XX +XXX,XX @@ vpc_co_create_opts(BlockDriver *drv, const char *filename, fail: qobject_unref(qdict); - bdrv_unref(bs); + bdrv_co_unref(bs); qapi_free_BlockdevCreateOptions(create_options); return ret; } -- 2.40.1 This QMP handler runs in a coroutine, so it must use the corresponding no_co_wrappers instead. Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=2185688 Cc: qemu-stable@nongnu.org Signed-off-by: Kevin Wolf <kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Message-Id: <20230504115750.54437-5-kwolf@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- blockdev.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/blockdev.c b/blockdev.c index XXXXXXX..XXXXXXX 100644 --- a/blockdev.c +++ b/blockdev.c @@ -XXX,XX +XXX,XX @@ void coroutine_fn qmp_block_resize(const char *device, const char *node_name, return; } - blk = blk_new_with_bs(bs, BLK_PERM_RESIZE, BLK_PERM_ALL, errp); + blk = blk_co_new_with_bs(bs, BLK_PERM_RESIZE, BLK_PERM_ALL, errp); if (!blk) { return; } @@ -XXX,XX +XXX,XX @@ void coroutine_fn qmp_block_resize(const char *device, const char *node_name, bdrv_co_lock(bs); bdrv_drained_end(bs); - blk_unref(blk); + blk_co_unref(blk); bdrv_co_unlock(bs); } -- 2.40.1 This tests that trying to resize an image with QMP block_resize doesn't hang or otherwise fail when the image is attached to a device running in an iothread. This is a regression test for the recent fix that changed qmp_block_resize, which is a coroutine based QMP handler, to avoid calling no_coroutine_fns directly. Signed-off-by: Kevin Wolf <kwolf@redhat.com> Message-Id: <20230509134133.373408-1-kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- tests/qemu-iotests/tests/iothreads-resize | 71 +++++++++++++++++++ tests/qemu-iotests/tests/iothreads-resize.out | 11 +++ 2 files changed, 82 insertions(+) create mode 100755 tests/qemu-iotests/tests/iothreads-resize create mode 100644 tests/qemu-iotests/tests/iothreads-resize.out diff --git a/tests/qemu-iotests/tests/iothreads-resize b/tests/qemu-iotests/tests/iothreads-resize new file mode 100755 index XXXXXXX..XXXXXXX --- /dev/null +++ b/tests/qemu-iotests/tests/iothreads-resize @@ -XXX,XX +XXX,XX @@ +#!/usr/bin/env bash +# group: rw auto quick +# +# Test resizing an image that is attached to a separate iothread +# +# Copyright (C) 2023 Red Hat, Inc. +# +# This program is free software; you can redistribute it and/or modify +# it under the terms of the GNU General Public License as published by +# the Free Software Foundation; either version 2 of the License, or +# (at your option) any later version. +# +# This program is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +# GNU General Public License for more details. +# +# You should have received a copy of the GNU General Public License +# along with this program. If not, see <http://www.gnu.org/licenses/>. +# + +# creator +owner=kwolf@redhat.com + +seq=`basename $0` +echo "QA output created by $seq" + +status=1 # failure is the default! + +_cleanup() +{ + _cleanup_test_img +} +trap "_cleanup; exit \$status" 0 1 2 3 15 + +# get standard environment, filters and checks +cd .. +. ./common.rc +. ./common.filter + +# Resizing images is only supported by a few block drivers +_supported_fmt raw qcow2 qed +_supported_proto file +_require_devices virtio-scsi-pci + +size=64M +_make_test_img $size + +qmp() { +cat <<EOF +{"execute":"qmp_capabilities"} +{'execute': 'block_resize', + 'arguments': {'node-name': 'img', 'size': 134217728}} +{"execute":"quit"} +EOF +} + +qmp | $QEMU -S -display none \ + -drive if=none,format=$IMGFMT,file="$TEST_IMG",node-name=img \ + -object iothread,id=t0 \ + -device virtio-scsi-pci,iothread=t0 \ + -device scsi-hd,drive=none0 \ + -qmp stdio \ + | _filter_qmp + +_img_info | _filter_img_info + +# success, all done +echo "*** done" +rm -f $seq.full +status=0 diff --git a/tests/qemu-iotests/tests/iothreads-resize.out b/tests/qemu-iotests/tests/iothreads-resize.out new file mode 100644 index XXXXXXX..XXXXXXX --- /dev/null +++ b/tests/qemu-iotests/tests/iothreads-resize.out @@ -XXX,XX +XXX,XX @@ +QA output created by iothreads-resize +Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=67108864 +QMP_VERSION +{"return": {}} +{"return": {}} +{"return": {}} +{"timestamp": {"seconds": TIMESTAMP, "microseconds": TIMESTAMP}, "event": "SHUTDOWN", "data": {"guest": false, "reason": "host-qmp-quit"}} +image: TEST_DIR/t.IMGFMT +file format: IMGFMT +virtual size: 128 MiB (134217728 bytes) +*** done -- 2.40.1 test-bdrv-drain contains a few test cases that are run both in coroutine and non-coroutine context. Running the entire code including the setup and shutdown in coroutines is incorrect because graph modifications can generally not happen in coroutines. Change the test so that creating and destroying the test nodes and BlockBackends always happens outside of coroutine context. Signed-off-by: Kevin Wolf <kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Message-Id: <20230504115750.54437-6-kwolf@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- tests/unit/test-bdrv-drain.c | 112 +++++++++++++++++++++++------------ 1 file changed, 75 insertions(+), 37 deletions(-) diff --git a/tests/unit/test-bdrv-drain.c b/tests/unit/test-bdrv-drain.c index XXXXXXX..XXXXXXX 100644 --- a/tests/unit/test-bdrv-drain.c +++ b/tests/unit/test-bdrv-drain.c @@ -XXX,XX +XXX,XX @@ static void do_drain_begin_unlocked(enum drain_type drain_type, BlockDriverState } } +static BlockBackend * no_coroutine_fn test_setup(void) +{ + BlockBackend *blk; + BlockDriverState *bs, *backing; + + blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL); + bs = bdrv_new_open_driver(&bdrv_test, "test-node", BDRV_O_RDWR, + &error_abort); + blk_insert_bs(blk, bs, &error_abort); + + backing = bdrv_new_open_driver(&bdrv_test, "backing", 0, &error_abort); + bdrv_set_backing_hd(bs, backing, &error_abort); + + bdrv_unref(backing); + bdrv_unref(bs); + + return blk; +} + static void do_drain_end_unlocked(enum drain_type drain_type, BlockDriverState *bs) { if (drain_type != BDRV_DRAIN_ALL) { @@ -XXX,XX +XXX,XX @@ static void do_drain_end_unlocked(enum drain_type drain_type, BlockDriverState * } } -static void test_drv_cb_common(enum drain_type drain_type, bool recursive) +static void test_drv_cb_common(BlockBackend *blk, enum drain_type drain_type, + bool recursive) { - BlockBackend *blk; - BlockDriverState *bs, *backing; + BlockDriverState *bs = blk_bs(blk); + BlockDriverState *backing = bs->backing->bs; BDRVTestState *s, *backing_s; BlockAIOCB *acb; int aio_ret; QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, NULL, 0); - blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL); - bs = bdrv_new_open_driver(&bdrv_test, "test-node", BDRV_O_RDWR, - &error_abort); s = bs->opaque; - blk_insert_bs(blk, bs, &error_abort); - - backing = bdrv_new_open_driver(&bdrv_test, "backing", 0, &error_abort); backing_s = backing->opaque; - bdrv_set_backing_hd(bs, backing, &error_abort); /* Simple bdrv_drain_all_begin/end pair, check that CBs are called */ g_assert_cmpint(s->drain_count, ==, 0); @@ -XXX,XX +XXX,XX @@ static void test_drv_cb_common(enum drain_type drain_type, bool recursive) g_assert_cmpint(s->drain_count, ==, 0); g_assert_cmpint(backing_s->drain_count, ==, 0); - - bdrv_unref(backing); - bdrv_unref(bs); - blk_unref(blk); } static void test_drv_cb_drain_all(void) { - test_drv_cb_common(BDRV_DRAIN_ALL, true); + BlockBackend *blk = test_setup(); + test_drv_cb_common(blk, BDRV_DRAIN_ALL, true); + blk_unref(blk); } static void test_drv_cb_drain(void) { - test_drv_cb_common(BDRV_DRAIN, false); + BlockBackend *blk = test_setup(); + test_drv_cb_common(blk, BDRV_DRAIN, false); + blk_unref(blk); +} + +static void coroutine_fn test_drv_cb_co_drain_all_entry(void) +{ + BlockBackend *blk = blk_all_next(NULL); + test_drv_cb_common(blk, BDRV_DRAIN_ALL, true); } static void test_drv_cb_co_drain_all(void) { - call_in_coroutine(test_drv_cb_drain_all); + BlockBackend *blk = test_setup(); + call_in_coroutine(test_drv_cb_co_drain_all_entry); + blk_unref(blk); } -static void test_drv_cb_co_drain(void) +static void coroutine_fn test_drv_cb_co_drain_entry(void) { - call_in_coroutine(test_drv_cb_drain); + BlockBackend *blk = blk_all_next(NULL); + test_drv_cb_common(blk, BDRV_DRAIN, false); } -static void test_quiesce_common(enum drain_type drain_type, bool recursive) +static void test_drv_cb_co_drain(void) { - BlockBackend *blk; - BlockDriverState *bs, *backing; - - blk = blk_new(qemu_get_aio_context(), BLK_PERM_ALL, BLK_PERM_ALL); - bs = bdrv_new_open_driver(&bdrv_test, "test-node", BDRV_O_RDWR, - &error_abort); - blk_insert_bs(blk, bs, &error_abort); + BlockBackend *blk = test_setup(); + call_in_coroutine(test_drv_cb_co_drain_entry); + blk_unref(blk); +} - backing = bdrv_new_open_driver(&bdrv_test, "backing", 0, &error_abort); - bdrv_set_backing_hd(bs, backing, &error_abort); +static void test_quiesce_common(BlockBackend *blk, enum drain_type drain_type, + bool recursive) +{ + BlockDriverState *bs = blk_bs(blk); + BlockDriverState *backing = bs->backing->bs; g_assert_cmpint(bs->quiesce_counter, ==, 0); g_assert_cmpint(backing->quiesce_counter, ==, 0); @@ -XXX,XX +XXX,XX @@ static void test_quiesce_common(enum drain_type drain_type, bool recursive) g_assert_cmpint(bs->quiesce_counter, ==, 0); g_assert_cmpint(backing->quiesce_counter, ==, 0); - - bdrv_unref(backing); - bdrv_unref(bs); - blk_unref(blk); } static void test_quiesce_drain_all(void) { - test_quiesce_common(BDRV_DRAIN_ALL, true); + BlockBackend *blk = test_setup(); + test_quiesce_common(blk, BDRV_DRAIN_ALL, true); + blk_unref(blk); } static void test_quiesce_drain(void) { - test_quiesce_common(BDRV_DRAIN, false); + BlockBackend *blk = test_setup(); + test_quiesce_common(blk, BDRV_DRAIN, false); + blk_unref(blk); +} + +static void coroutine_fn test_quiesce_co_drain_all_entry(void) +{ + BlockBackend *blk = blk_all_next(NULL); + test_quiesce_common(blk, BDRV_DRAIN_ALL, true); } static void test_quiesce_co_drain_all(void) { - call_in_coroutine(test_quiesce_drain_all); + BlockBackend *blk = test_setup(); + call_in_coroutine(test_quiesce_co_drain_all_entry); + blk_unref(blk); +} + +static void coroutine_fn test_quiesce_co_drain_entry(void) +{ + BlockBackend *blk = blk_all_next(NULL); + test_quiesce_common(blk, BDRV_DRAIN, false); } static void test_quiesce_co_drain(void) { - call_in_coroutine(test_quiesce_drain); + BlockBackend *blk = test_setup(); + call_in_coroutine(test_quiesce_co_drain_entry); + blk_unref(blk); } static void test_nested(void) -- 2.40.1 For some functions, it is part of their interface to be called without holding the graph lock. Add a new macro to document this. The macro expands to TSA_EXCLUDES(), which is a relatively weak check because it passes in cases where the compiler just doesn't know if the lock is held. Function pointers can't be checked at all. Therefore, its primary purpose is documentation. Signed-off-by: Kevin Wolf <kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Message-Id: <20230504115750.54437-7-kwolf@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- include/block/graph-lock.h | 2 ++ 1 file changed, 2 insertions(+) diff --git a/include/block/graph-lock.h b/include/block/graph-lock.h index XXXXXXX..XXXXXXX 100644 --- a/include/block/graph-lock.h +++ b/include/block/graph-lock.h @@ -XXX,XX +XXX,XX @@ extern BdrvGraphLock graph_lock; */ #define GRAPH_WRLOCK TSA_REQUIRES(graph_lock) #define GRAPH_RDLOCK TSA_REQUIRES_SHARED(graph_lock) +#define GRAPH_UNLOCKED TSA_EXCLUDES(graph_lock) /* * TSA annotations are not part of function types, so checks are defeated when @@ -XXX,XX +XXX,XX @@ extern BdrvGraphLock graph_lock; */ #define GRAPH_RDLOCK_PTR TSA_GUARDED_BY(graph_lock) #define GRAPH_WRLOCK_PTR TSA_GUARDED_BY(graph_lock) +#define GRAPH_UNLOCKED_PTR /* * register_aiocontext: -- 2.40.1 GRAPH_RDLOCK_GUARD() and GRAPH_RDLOCK_GUARD_MAINLOOP() only take a reader lock for the graph, so the correct annotation for them to use is TSA_ASSERT_SHARED rather than TSA_ASSERT. Signed-off-by: Kevin Wolf <kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Message-Id: <20230504115750.54437-8-kwolf@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- include/block/graph-lock.h | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) diff --git a/include/block/graph-lock.h b/include/block/graph-lock.h index XXXXXXX..XXXXXXX 100644 --- a/include/block/graph-lock.h +++ b/include/block/graph-lock.h @@ -XXX,XX +XXX,XX @@ typedef struct GraphLockable { } GraphLockable; #define GML_OBJ_() (&(GraphLockable) { }) /* - * This is not marked as TSA_ACQUIRE() because TSA doesn't understand the + * This is not marked as TSA_ACQUIRE_SHARED() because TSA doesn't understand the * cleanup attribute and would therefore complain that the graph is never - * unlocked. TSA_ASSERT() makes sure that the following calls know that we - * hold the lock while unlocking is left unchecked. + * unlocked. TSA_ASSERT_SHARED() makes sure that the following calls know that + * we hold the lock while unlocking is left unchecked. */ -static inline GraphLockable * TSA_ASSERT(graph_lock) TSA_NO_TSA coroutine_fn +static inline GraphLockable * TSA_ASSERT_SHARED(graph_lock) TSA_NO_TSA coroutine_fn graph_lockable_auto_lock(GraphLockable *x) { bdrv_graph_co_rdlock(); @@ -XXX,XX +XXX,XX @@ typedef struct GraphLockableMainloop { } GraphLockableMainloop; #define GMLML_OBJ_() (&(GraphLockableMainloop) { }) /* - * This is not marked as TSA_ACQUIRE() because TSA doesn't understand the + * This is not marked as TSA_ACQUIRE_SHARED() because TSA doesn't understand the * cleanup attribute and would therefore complain that the graph is never - * unlocked. TSA_ASSERT() makes sure that the following calls know that we - * hold the lock while unlocking is left unchecked. + * unlocked. TSA_ASSERT_SHARED() makes sure that the following calls know that + * we hold the lock while unlocking is left unchecked. */ -static inline GraphLockableMainloop * TSA_ASSERT(graph_lock) TSA_NO_TSA +static inline GraphLockableMainloop * TSA_ASSERT_SHARED(graph_lock) TSA_NO_TSA graph_lockable_auto_lock_mainloop(GraphLockableMainloop *x) { bdrv_graph_rdlock_main_loop(); -- 2.40.1 Drivers were a bit confused about whether .bdrv_open can run in a coroutine and whether or not it holds a graph lock. It cannot keep a graph lock from the caller across the whole function because it both changes the graph (requires a writer lock) and does I/O (requires a reader lock). Therefore, it should take these locks internally as needed. The functions used to be called in coroutine context during image creation. This was buggy for other reasons, and as of commit 32192301, all block drivers go through no_co_wrappers. So it is not called in coroutine context any more. Fix qcow2 and qed to work with the correct assumptions: The graph lock needs to be taken internally instead of just assuming it's already there, and the coroutine path is dead code that can be removed. Signed-off-by: Kevin Wolf <kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Message-Id: <20230504115750.54437-9-kwolf@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- include/block/block_int-common.h | 8 ++++---- block.c | 6 +++--- block/qcow2.c | 15 ++++++--------- block/qed.c | 18 ++++++++---------- 4 files changed, 21 insertions(+), 26 deletions(-) diff --git a/include/block/block_int-common.h b/include/block/block_int-common.h index XXXXXXX..XXXXXXX 100644 --- a/include/block/block_int-common.h +++ b/include/block/block_int-common.h @@ -XXX,XX +XXX,XX @@ struct BlockDriver { void (*bdrv_reopen_abort)(BDRVReopenState *reopen_state); void (*bdrv_join_options)(QDict *options, QDict *old_options); - int (*bdrv_open)(BlockDriverState *bs, QDict *options, int flags, - Error **errp); + int GRAPH_UNLOCKED_PTR (*bdrv_open)( + BlockDriverState *bs, QDict *options, int flags, Error **errp); /* Protocol drivers should implement this instead of bdrv_open */ - int (*bdrv_file_open)(BlockDriverState *bs, QDict *options, int flags, - Error **errp); + int GRAPH_UNLOCKED_PTR (*bdrv_file_open)( + BlockDriverState *bs, QDict *options, int flags, Error **errp); void (*bdrv_close)(BlockDriverState *bs); int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_create)( diff --git a/block.c b/block.c index XXXXXXX..XXXXXXX 100644 --- a/block.c +++ b/block.c @@ -XXX,XX +XXX,XX @@ out: * bdrv_refresh_total_sectors() which polls when called from non-coroutine * context. */ -static int bdrv_open_driver(BlockDriverState *bs, BlockDriver *drv, - const char *node_name, QDict *options, - int open_flags, Error **errp) +static int no_coroutine_fn GRAPH_UNLOCKED +bdrv_open_driver(BlockDriverState *bs, BlockDriver *drv, const char *node_name, + QDict *options, int open_flags, Error **errp) { Error *local_err = NULL; int i, ret; diff --git a/block/qcow2.c b/block/qcow2.c index XXXXXXX..XXXXXXX 100644 --- a/block/qcow2.c +++ b/block/qcow2.c @@ -XXX,XX +XXX,XX @@ static void coroutine_fn qcow2_open_entry(void *opaque) QCow2OpenCo *qoc = opaque; BDRVQcow2State *s = qoc->bs->opaque; - assume_graph_lock(); /* FIXME */ + GRAPH_RDLOCK_GUARD(); qemu_co_mutex_lock(&s->lock); qoc->ret = qcow2_do_open(qoc->bs, qoc->options, qoc->flags, true, @@ -XXX,XX +XXX,XX @@ static int qcow2_open(BlockDriverState *bs, QDict *options, int flags, /* Initialise locks */ qemu_co_mutex_init(&s->lock); - if (qemu_in_coroutine()) { - /* From bdrv_co_create. */ - qcow2_open_entry(&qoc); - } else { - assert(qemu_get_current_aio_context() == qemu_get_aio_context()); - qemu_coroutine_enter(qemu_coroutine_create(qcow2_open_entry, &qoc)); - BDRV_POLL_WHILE(bs, qoc.ret == -EINPROGRESS); - } + assert(!qemu_in_coroutine()); + assert(qemu_get_current_aio_context() == qemu_get_aio_context()); + qemu_coroutine_enter(qemu_coroutine_create(qcow2_open_entry, &qoc)); + BDRV_POLL_WHILE(bs, qoc.ret == -EINPROGRESS); + return qoc.ret; } diff --git a/block/qed.c b/block/qed.c index XXXXXXX..XXXXXXX 100644 --- a/block/qed.c +++ b/block/qed.c @@ -XXX,XX +XXX,XX @@ typedef struct QEDOpenCo { int ret; } QEDOpenCo; -static void coroutine_fn GRAPH_RDLOCK bdrv_qed_open_entry(void *opaque) +static void coroutine_fn bdrv_qed_open_entry(void *opaque) { QEDOpenCo *qoc = opaque; BDRVQEDState *s = qoc->bs->opaque; + GRAPH_RDLOCK_GUARD(); + qemu_co_mutex_lock(&s->table_lock); qoc->ret = bdrv_qed_do_open(qoc->bs, qoc->options, qoc->flags, qoc->errp); qemu_co_mutex_unlock(&s->table_lock); @@ -XXX,XX +XXX,XX @@ static int bdrv_qed_open(BlockDriverState *bs, QDict *options, int flags, }; int ret; - assume_graph_lock(); /* FIXME */ - ret = bdrv_open_file_child(NULL, options, "file", bs, errp); if (ret < 0) { return ret; } bdrv_qed_init_state(bs); - if (qemu_in_coroutine()) { - bdrv_qed_open_entry(&qoc); - } else { - assert(qemu_get_current_aio_context() == qemu_get_aio_context()); - qemu_coroutine_enter(qemu_coroutine_create(bdrv_qed_open_entry, &qoc)); - BDRV_POLL_WHILE(bs, qoc.ret == -EINPROGRESS); - } + assert(!qemu_in_coroutine()); + assert(qemu_get_current_aio_context() == qemu_get_aio_context()); + qemu_coroutine_enter(qemu_coroutine_create(bdrv_qed_open_entry, &qoc)); + BDRV_POLL_WHILE(bs, qoc.ret == -EINPROGRESS); + return qoc.ret; } -- 2.40.1 The only thing nbd_co_flush() does is call nbd_client_co_flush(). Just use that function directly in the BlockDriver definitions and remove the wrapper. Signed-off-by: Kevin Wolf <kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Message-Id: <20230504115750.54437-10-kwolf@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/nbd.c | 11 +++-------- 1 file changed, 3 insertions(+), 8 deletions(-) diff --git a/block/nbd.c b/block/nbd.c index XXXXXXX..XXXXXXX 100644 --- a/block/nbd.c +++ b/block/nbd.c @@ -XXX,XX +XXX,XX @@ fail: return ret; } -static int coroutine_fn nbd_co_flush(BlockDriverState *bs) -{ - return nbd_client_co_flush(bs); -} - static void nbd_refresh_limits(BlockDriverState *bs, Error **errp) { BDRVNBDState *s = (BDRVNBDState *)bs->opaque; @@ -XXX,XX +XXX,XX @@ static BlockDriver bdrv_nbd = { .bdrv_co_pwritev = nbd_client_co_pwritev, .bdrv_co_pwrite_zeroes = nbd_client_co_pwrite_zeroes, .bdrv_close = nbd_close, - .bdrv_co_flush_to_os = nbd_co_flush, + .bdrv_co_flush_to_os = nbd_client_co_flush, .bdrv_co_pdiscard = nbd_client_co_pdiscard, .bdrv_refresh_limits = nbd_refresh_limits, .bdrv_co_truncate = nbd_co_truncate, @@ -XXX,XX +XXX,XX @@ static BlockDriver bdrv_nbd_tcp = { .bdrv_co_pwritev = nbd_client_co_pwritev, .bdrv_co_pwrite_zeroes = nbd_client_co_pwrite_zeroes, .bdrv_close = nbd_close, - .bdrv_co_flush_to_os = nbd_co_flush, + .bdrv_co_flush_to_os = nbd_client_co_flush, .bdrv_co_pdiscard = nbd_client_co_pdiscard, .bdrv_refresh_limits = nbd_refresh_limits, .bdrv_co_truncate = nbd_co_truncate, @@ -XXX,XX +XXX,XX @@ static BlockDriver bdrv_nbd_unix = { .bdrv_co_pwritev = nbd_client_co_pwritev, .bdrv_co_pwrite_zeroes = nbd_client_co_pwrite_zeroes, .bdrv_close = nbd_close, - .bdrv_co_flush_to_os = nbd_co_flush, + .bdrv_co_flush_to_os = nbd_client_co_flush, .bdrv_co_pdiscard = nbd_client_co_pdiscard, .bdrv_refresh_limits = nbd_refresh_limits, .bdrv_co_truncate = nbd_co_truncate, -- 2.40.1 From: Emanuele Giuseppe Esposito <eesposit@redhat.com> This adds GRAPH_RDLOCK annotations to declare that callers of nbd_co_do_establish_connection() need to hold a reader lock for the graph. Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Message-Id: <20230504115750.54437-11-kwolf@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/coroutines.h | 5 +++-- block/nbd.c | 39 +++++++++++++++++++++------------------ 2 files changed, 24 insertions(+), 20 deletions(-) diff --git a/block/coroutines.h b/block/coroutines.h index XXXXXXX..XXXXXXX 100644 --- a/block/coroutines.h +++ b/block/coroutines.h @@ -XXX,XX +XXX,XX @@ bdrv_co_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos); int coroutine_fn GRAPH_RDLOCK bdrv_co_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos); -int coroutine_fn +int coroutine_fn GRAPH_RDLOCK nbd_co_do_establish_connection(BlockDriverState *bs, bool blocking, Error **errp); @@ -XXX,XX +XXX,XX @@ bdrv_common_block_status_above(BlockDriverState *bs, int64_t *map, BlockDriverState **file, int *depth); -int co_wrapper_mixed + +int co_wrapper_mixed_bdrv_rdlock nbd_do_establish_connection(BlockDriverState *bs, bool blocking, Error **errp); #endif /* BLOCK_COROUTINES_H */ diff --git a/block/nbd.c b/block/nbd.c index XXXXXXX..XXXXXXX 100644 --- a/block/nbd.c +++ b/block/nbd.c @@ -XXX,XX +XXX,XX @@ int coroutine_fn nbd_co_do_establish_connection(BlockDriverState *bs, int ret; IO_CODE(); + assert_bdrv_graph_readable(); assert(!s->ioc); s->ioc = nbd_co_establish_connection(s->conn, &s->info, blocking, errp); @@ -XXX,XX +XXX,XX @@ static bool nbd_client_connecting(BDRVNBDState *s) } /* Called with s->requests_lock taken. */ -static coroutine_fn void nbd_reconnect_attempt(BDRVNBDState *s) +static void coroutine_fn GRAPH_RDLOCK nbd_reconnect_attempt(BDRVNBDState *s) { int ret; bool blocking = s->state == NBD_CLIENT_CONNECTING_WAIT; @@ -XXX,XX +XXX,XX @@ static coroutine_fn int nbd_receive_replies(BDRVNBDState *s, uint64_t handle) } } -static int coroutine_fn nbd_co_send_request(BlockDriverState *bs, - NBDRequest *request, - QEMUIOVector *qiov) +static int coroutine_fn GRAPH_RDLOCK +nbd_co_send_request(BlockDriverState *bs, NBDRequest *request, + QEMUIOVector *qiov) { BDRVNBDState *s = (BDRVNBDState *)bs->opaque; int rc, i = -1; @@ -XXX,XX +XXX,XX @@ static int coroutine_fn nbd_co_receive_blockstatus_reply(BDRVNBDState *s, return iter.ret; } -static int coroutine_fn nbd_co_request(BlockDriverState *bs, NBDRequest *request, - QEMUIOVector *write_qiov) +static int coroutine_fn GRAPH_RDLOCK +nbd_co_request(BlockDriverState *bs, NBDRequest *request, + QEMUIOVector *write_qiov) { int ret, request_ret; Error *local_err = NULL; @@ -XXX,XX +XXX,XX @@ static int coroutine_fn nbd_co_request(BlockDriverState *bs, NBDRequest *request return ret ? ret : request_ret; } -static int coroutine_fn nbd_client_co_preadv(BlockDriverState *bs, int64_t offset, - int64_t bytes, QEMUIOVector *qiov, - BdrvRequestFlags flags) +static int coroutine_fn GRAPH_RDLOCK +nbd_client_co_preadv(BlockDriverState *bs, int64_t offset, int64_t bytes, + QEMUIOVector *qiov, BdrvRequestFlags flags) { int ret, request_ret; Error *local_err = NULL; @@ -XXX,XX +XXX,XX @@ static int coroutine_fn nbd_client_co_preadv(BlockDriverState *bs, int64_t offse return ret ? ret : request_ret; } -static int coroutine_fn nbd_client_co_pwritev(BlockDriverState *bs, int64_t offset, - int64_t bytes, QEMUIOVector *qiov, - BdrvRequestFlags flags) +static int coroutine_fn GRAPH_RDLOCK +nbd_client_co_pwritev(BlockDriverState *bs, int64_t offset, int64_t bytes, + QEMUIOVector *qiov, BdrvRequestFlags flags) { BDRVNBDState *s = (BDRVNBDState *)bs->opaque; NBDRequest request = { @@ -XXX,XX +XXX,XX @@ static int coroutine_fn nbd_client_co_pwritev(BlockDriverState *bs, int64_t offs return nbd_co_request(bs, &request, qiov); } -static int coroutine_fn nbd_client_co_pwrite_zeroes(BlockDriverState *bs, int64_t offset, - int64_t bytes, BdrvRequestFlags flags) +static int coroutine_fn GRAPH_RDLOCK +nbd_client_co_pwrite_zeroes(BlockDriverState *bs, int64_t offset, int64_t bytes, + BdrvRequestFlags flags) { BDRVNBDState *s = (BDRVNBDState *)bs->opaque; NBDRequest request = { @@ -XXX,XX +XXX,XX @@ static int coroutine_fn nbd_client_co_pwrite_zeroes(BlockDriverState *bs, int64_ return nbd_co_request(bs, &request, NULL); } -static int coroutine_fn nbd_client_co_flush(BlockDriverState *bs) +static int coroutine_fn GRAPH_RDLOCK nbd_client_co_flush(BlockDriverState *bs) { BDRVNBDState *s = (BDRVNBDState *)bs->opaque; NBDRequest request = { .type = NBD_CMD_FLUSH }; @@ -XXX,XX +XXX,XX @@ static int coroutine_fn nbd_client_co_flush(BlockDriverState *bs) return nbd_co_request(bs, &request, NULL); } -static int coroutine_fn nbd_client_co_pdiscard(BlockDriverState *bs, int64_t offset, - int64_t bytes) +static int coroutine_fn GRAPH_RDLOCK +nbd_client_co_pdiscard(BlockDriverState *bs, int64_t offset, int64_t bytes) { BDRVNBDState *s = (BDRVNBDState *)bs->opaque; NBDRequest request = { @@ -XXX,XX +XXX,XX @@ static int coroutine_fn nbd_client_co_pdiscard(BlockDriverState *bs, int64_t off return nbd_co_request(bs, &request, NULL); } -static int coroutine_fn nbd_client_co_block_status( +static int coroutine_fn GRAPH_RDLOCK nbd_client_co_block_status( BlockDriverState *bs, bool want_zero, int64_t offset, int64_t bytes, int64_t *pnum, int64_t *map, BlockDriverState **file) { -- 2.40.1 This adds GRAPH_RDLOCK annotations to declare that functions accessing the parent list of a node need to hold a reader lock for the graph. As it happens, they already do. Signed-off-by: Kevin Wolf <kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Message-Id: <20230504115750.54437-12-kwolf@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/vhdx.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/block/vhdx.c b/block/vhdx.c index XXXXXXX..XXXXXXX 100644 --- a/block/vhdx.c +++ b/block/vhdx.c @@ -XXX,XX +XXX,XX @@ exit: * There are 2 headers, and the highest sequence number will represent * the active header */ -static int vhdx_create_new_headers(BlockBackend *blk, uint64_t image_size, - uint32_t log_size) +static int coroutine_fn GRAPH_RDLOCK +vhdx_create_new_headers(BlockBackend *blk, uint64_t image_size, + uint32_t log_size) { BlockDriverState *bs = blk_bs(blk); BdrvChild *child; @@ -XXX,XX +XXX,XX @@ exit: * .---- ~ ----------- ~ ------------ ~ ---------------- ~ -----------. * 1MB */ -static int coroutine_fn vhdx_co_create(BlockdevCreateOptions *opts, - Error **errp) +static int coroutine_fn GRAPH_RDLOCK +vhdx_co_create(BlockdevCreateOptions *opts, Error **errp) { BlockdevCreateOptionsVhdx *vhdx_opts; BlockBackend *blk = NULL; -- 2.40.1 This adds GRAPH_RDLOCK annotations to declare that functions accessing the parent list of a node need to hold a reader lock for the graph. As it happens, they already do. Signed-off-by: Kevin Wolf <kwolf@redhat.com> Message-Id: <20230504115750.54437-13-kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/mirror.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/block/mirror.c b/block/mirror.c index XXXXXXX..XXXXXXX 100644 --- a/block/mirror.c +++ b/block/mirror.c @@ -XXX,XX +XXX,XX @@ static MirrorOp *coroutine_fn active_write_prepare(MirrorBlockJob *s, return op; } -static void coroutine_fn active_write_settle(MirrorOp *op) +static void coroutine_fn GRAPH_RDLOCK active_write_settle(MirrorOp *op) { uint64_t start_chunk = op->offset / op->s->granularity; uint64_t end_chunk = DIV_ROUND_UP(op->offset + op->bytes, -- 2.40.1 From: Emanuele Giuseppe Esposito <eesposit@redhat.com> This adds GRAPH_RDLOCK annotations to declare that callers of bdrv_co_get_allocated_file_size() need to hold a reader lock for the graph. Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Message-Id: <20230504115750.54437-14-kwolf@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- include/block/block-io.h | 7 +++++-- include/block/block_int-common.h | 2 +- block.c | 4 +++- block/vmdk.c | 2 +- 4 files changed, 10 insertions(+), 5 deletions(-) diff --git a/include/block/block-io.h b/include/block/block-io.h index XXXXXXX..XXXXXXX 100644 --- a/include/block/block-io.h +++ b/include/block/block-io.h @@ -XXX,XX +XXX,XX @@ int64_t coroutine_mixed_fn bdrv_nb_sectors(BlockDriverState *bs); int64_t coroutine_fn GRAPH_RDLOCK bdrv_co_getlength(BlockDriverState *bs); int64_t co_wrapper_mixed_bdrv_rdlock bdrv_getlength(BlockDriverState *bs); -int64_t coroutine_fn bdrv_co_get_allocated_file_size(BlockDriverState *bs); -int64_t co_wrapper bdrv_get_allocated_file_size(BlockDriverState *bs); +int64_t coroutine_fn GRAPH_RDLOCK +bdrv_co_get_allocated_file_size(BlockDriverState *bs); + +int64_t co_wrapper_bdrv_rdlock +bdrv_get_allocated_file_size(BlockDriverState *bs); BlockMeasureInfo *bdrv_measure(BlockDriver *drv, QemuOpts *opts, BlockDriverState *in_bs, Error **errp); diff --git a/include/block/block_int-common.h b/include/block/block_int-common.h index XXXXXXX..XXXXXXX 100644 --- a/include/block/block_int-common.h +++ b/include/block/block_int-common.h @@ -XXX,XX +XXX,XX @@ struct BlockDriver { int64_t coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_getlength)( BlockDriverState *bs); - int64_t coroutine_fn (*bdrv_co_get_allocated_file_size)( + int64_t coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_get_allocated_file_size)( BlockDriverState *bs); BlockMeasureInfo *(*bdrv_measure)(QemuOpts *opts, BlockDriverState *in_bs, diff --git a/block.c b/block.c index XXXXXXX..XXXXXXX 100644 --- a/block.c +++ b/block.c @@ -XXX,XX +XXX,XX @@ exit: * sums the size of all data-bearing children. (This excludes backing * children.) */ -static int64_t coroutine_fn bdrv_sum_allocated_file_size(BlockDriverState *bs) +static int64_t coroutine_fn GRAPH_RDLOCK +bdrv_sum_allocated_file_size(BlockDriverState *bs) { BdrvChild *child; int64_t child_size, sum = 0; @@ -XXX,XX +XXX,XX @@ int64_t coroutine_fn bdrv_co_get_allocated_file_size(BlockDriverState *bs) { BlockDriver *drv = bs->drv; IO_CODE(); + assert_bdrv_graph_readable(); if (!drv) { return -ENOMEDIUM; diff --git a/block/vmdk.c b/block/vmdk.c index XXXXXXX..XXXXXXX 100644 --- a/block/vmdk.c +++ b/block/vmdk.c @@ -XXX,XX +XXX,XX @@ static void vmdk_close(BlockDriverState *bs) error_free(s->migration_blocker); } -static int64_t coroutine_fn +static int64_t coroutine_fn GRAPH_RDLOCK vmdk_co_get_allocated_file_size(BlockDriverState *bs) { int i; -- 2.40.1 From: Emanuele Giuseppe Esposito <eesposit@redhat.com> This adds GRAPH_RDLOCK annotations to declare that callers of bdrv_co_get_info() need to hold a reader lock for the graph. Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Message-Id: <20230504115750.54437-15-kwolf@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- include/block/block-io.h | 7 +++++-- include/block/block_int-common.h | 4 ++-- block.c | 2 ++ block/crypto.c | 2 +- block/io.c | 11 +++++------ block/mirror.c | 8 ++++++-- block/raw-format.c | 2 +- 7 files changed, 22 insertions(+), 14 deletions(-) diff --git a/include/block/block-io.h b/include/block/block-io.h index XXXXXXX..XXXXXXX 100644 --- a/include/block/block-io.h +++ b/include/block/block-io.h @@ -XXX,XX +XXX,XX @@ const char *bdrv_get_node_name(const BlockDriverState *bs); const char *bdrv_get_device_name(const BlockDriverState *bs); const char *bdrv_get_device_or_node_name(const BlockDriverState *bs); -int coroutine_fn bdrv_co_get_info(BlockDriverState *bs, BlockDriverInfo *bdi); -int co_wrapper_mixed bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi); +int coroutine_fn GRAPH_RDLOCK +bdrv_co_get_info(BlockDriverState *bs, BlockDriverInfo *bdi); + +int co_wrapper_mixed_bdrv_rdlock +bdrv_get_info(BlockDriverState *bs, BlockDriverInfo *bdi); ImageInfoSpecific *bdrv_get_specific_info(BlockDriverState *bs, Error **errp); diff --git a/include/block/block_int-common.h b/include/block/block_int-common.h index XXXXXXX..XXXXXXX 100644 --- a/include/block/block_int-common.h +++ b/include/block/block_int-common.h @@ -XXX,XX +XXX,XX @@ struct BlockDriver { BlockDriverState *bs, int64_t offset, int64_t bytes, QEMUIOVector *qiov, size_t qiov_offset); - int coroutine_fn (*bdrv_co_get_info)(BlockDriverState *bs, - BlockDriverInfo *bdi); + int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_get_info)( + BlockDriverState *bs, BlockDriverInfo *bdi); ImageInfoSpecific *(*bdrv_get_specific_info)(BlockDriverState *bs, Error **errp); diff --git a/block.c b/block.c index XXXXXXX..XXXXXXX 100644 --- a/block.c +++ b/block.c @@ -XXX,XX +XXX,XX @@ int coroutine_fn bdrv_co_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) int ret; BlockDriver *drv = bs->drv; IO_CODE(); + assert_bdrv_graph_readable(); + /* if bs->drv == NULL, bs is closed, so there's nothing to do here */ if (!drv) { return -ENOMEDIUM; diff --git a/block/crypto.c b/block/crypto.c index XXXXXXX..XXXXXXX 100644 --- a/block/crypto.c +++ b/block/crypto.c @@ -XXX,XX +XXX,XX @@ fail: return ret; } -static int coroutine_fn +static int coroutine_fn GRAPH_RDLOCK block_crypto_co_get_info_luks(BlockDriverState *bs, BlockDriverInfo *bdi) { BlockDriverInfo subbdi; diff --git a/block/io.c b/block/io.c index XXXXXXX..XXXXXXX 100644 --- a/block/io.c +++ b/block/io.c @@ -XXX,XX +XXX,XX @@ BdrvTrackedRequest *coroutine_fn bdrv_co_get_self_request(BlockDriverState *bs) /** * Round a region to cluster boundaries */ -void coroutine_fn bdrv_round_to_clusters(BlockDriverState *bs, - int64_t offset, int64_t bytes, - int64_t *cluster_offset, - int64_t *cluster_bytes) +void coroutine_fn GRAPH_RDLOCK +bdrv_round_to_clusters(BlockDriverState *bs, int64_t offset, int64_t bytes, + int64_t *cluster_offset, int64_t *cluster_bytes) { BlockDriverInfo bdi; IO_CODE(); @@ -XXX,XX +XXX,XX @@ void coroutine_fn bdrv_round_to_clusters(BlockDriverState *bs, } } -static coroutine_fn int bdrv_get_cluster_size(BlockDriverState *bs) +static int coroutine_fn GRAPH_RDLOCK bdrv_get_cluster_size(BlockDriverState *bs) { BlockDriverInfo bdi; int ret; @@ -XXX,XX +XXX,XX @@ fail: return ret; } -static inline int coroutine_fn +static inline int coroutine_fn GRAPH_RDLOCK bdrv_co_write_req_prepare(BdrvChild *child, int64_t offset, int64_t bytes, BdrvTrackedRequest *req, int flags) { diff --git a/block/mirror.c b/block/mirror.c index XXXXXXX..XXXXXXX 100644 --- a/block/mirror.c +++ b/block/mirror.c @@ -XXX,XX +XXX,XX @@ static uint64_t coroutine_fn mirror_iteration(MirrorBlockJob *s) } else if (ret >= 0 && !(ret & BDRV_BLOCK_DATA)) { int64_t target_offset; int64_t target_bytes; - bdrv_round_to_clusters(blk_bs(s->target), offset, io_bytes, - &target_offset, &target_bytes); + WITH_GRAPH_RDLOCK_GUARD() { + bdrv_round_to_clusters(blk_bs(s->target), offset, io_bytes, + &target_offset, &target_bytes); + } if (target_offset == offset && target_bytes == io_bytes) { mirror_method = ret & BDRV_BLOCK_ZERO ? @@ -XXX,XX +XXX,XX @@ static int coroutine_fn mirror_run(Job *job, Error **errp) */ bdrv_get_backing_filename(target_bs, backing_filename, sizeof(backing_filename)); + bdrv_graph_co_rdlock(); if (!bdrv_co_get_info(target_bs, &bdi) && bdi.cluster_size) { s->target_cluster_size = bdi.cluster_size; } else { s->target_cluster_size = BDRV_SECTOR_SIZE; } + bdrv_graph_co_rdunlock(); if (backing_filename[0] && !bdrv_backing_chain_next(target_bs) && s->granularity < s->target_cluster_size) { s->buf_size = MAX(s->buf_size, s->target_cluster_size); diff --git a/block/raw-format.c b/block/raw-format.c index XXXXXXX..XXXXXXX 100644 --- a/block/raw-format.c +++ b/block/raw-format.c @@ -XXX,XX +XXX,XX @@ static BlockMeasureInfo *raw_measure(QemuOpts *opts, BlockDriverState *in_bs, return info; } -static int coroutine_fn +static int coroutine_fn GRAPH_RDLOCK raw_co_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) { return bdrv_co_get_info(bs->file->bs, bdi); -- 2.40.1 From: Emanuele Giuseppe Esposito <eesposit@redhat.com> This adds GRAPH_RDLOCK annotations to declare that callers of bdrv_co_debug_event() need to hold a reader lock for the graph. Unfortunately we cannot use a co_wrapper_bdrv_rdlock (i.e. make the coroutine wrapper a no_coroutine_fn), because the function is called (using the BLKDBG_EVENT macro) by mixed functions that run both in coroutine and non-coroutine context (for example many of the functions in qcow2-cluster.c and qcow2-refcount.c). Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Message-Id: <20230504115750.54437-16-kwolf@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- include/block/block-io.h | 9 +++++---- include/block/block_int-common.h | 4 ++-- block.c | 2 ++ 3 files changed, 9 insertions(+), 6 deletions(-) diff --git a/include/block/block-io.h b/include/block/block-io.h index XXXXXXX..XXXXXXX 100644 --- a/include/block/block-io.h +++ b/include/block/block-io.h @@ -XXX,XX +XXX,XX @@ void *qemu_try_blockalign0(BlockDriverState *bs, size_t size); void bdrv_enable_copy_on_read(BlockDriverState *bs); void bdrv_disable_copy_on_read(BlockDriverState *bs); -void coroutine_fn bdrv_co_debug_event(BlockDriverState *bs, - BlkdebugEvent event); -void co_wrapper_mixed bdrv_debug_event(BlockDriverState *bs, - BlkdebugEvent event); +void coroutine_fn GRAPH_RDLOCK +bdrv_co_debug_event(BlockDriverState *bs, BlkdebugEvent event); + +void co_wrapper_mixed_bdrv_rdlock +bdrv_debug_event(BlockDriverState *bs, BlkdebugEvent event); #define BLKDBG_EVENT(child, evt) \ do { \ diff --git a/include/block/block_int-common.h b/include/block/block_int-common.h index XXXXXXX..XXXXXXX 100644 --- a/include/block/block_int-common.h +++ b/include/block/block_int-common.h @@ -XXX,XX +XXX,XX @@ struct BlockDriver { int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_check)( BlockDriverState *bs, BdrvCheckResult *result, BdrvCheckMode fix); - void coroutine_fn (*bdrv_co_debug_event)(BlockDriverState *bs, - BlkdebugEvent event); + void coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_debug_event)( + BlockDriverState *bs, BlkdebugEvent event); /* io queue for linux-aio */ void coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_io_plug)(BlockDriverState *bs); diff --git a/block.c b/block.c index XXXXXXX..XXXXXXX 100644 --- a/block.c +++ b/block.c @@ -XXX,XX +XXX,XX @@ BlockStatsSpecific *bdrv_get_specific_stats(BlockDriverState *bs) void coroutine_fn bdrv_co_debug_event(BlockDriverState *bs, BlkdebugEvent event) { IO_CODE(); + assert_bdrv_graph_readable(); + if (!bs || !bs->drv || !bs->drv->bdrv_co_debug_event) { return; } -- 2.40.1 From: Emanuele Giuseppe Esposito <eesposit@redhat.com> This adds GRAPH_RDLOCK annotations to declare that callers of amend callbacks in BlockDriver need to hold a reader lock for the graph. Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Message-Id: <20230504115750.54437-17-kwolf@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- include/block/block_int-common.h | 12 ++++++------ block/amend.c | 8 +++++++- 2 files changed, 13 insertions(+), 7 deletions(-) diff --git a/include/block/block_int-common.h b/include/block/block_int-common.h index XXXXXXX..XXXXXXX 100644 --- a/include/block/block_int-common.h +++ b/include/block/block_int-common.h @@ -XXX,XX +XXX,XX @@ struct BlockDriver { * to allow driver-specific initialization code that requires * the BQL, like setting up specific permission flags. */ - int (*bdrv_amend_pre_run)(BlockDriverState *bs, Error **errp); + int GRAPH_RDLOCK_PTR (*bdrv_amend_pre_run)( + BlockDriverState *bs, Error **errp); /* * This function is invoked under BQL after .bdrv_co_amend() * to allow cleaning up what was done in .bdrv_amend_pre_run(). */ - void (*bdrv_amend_clean)(BlockDriverState *bs); + void GRAPH_RDLOCK_PTR (*bdrv_amend_clean)(BlockDriverState *bs); /* * Return true if @to_replace can be replaced by a BDS with the @@ -XXX,XX +XXX,XX @@ struct BlockDriver { int (*bdrv_probe)(const uint8_t *buf, int buf_size, const char *filename); - int coroutine_fn (*bdrv_co_amend)(BlockDriverState *bs, - BlockdevAmendOptions *opts, - bool force, - Error **errp); + int coroutine_fn GRAPH_RDLOCK_PTR (*bdrv_co_amend)( + BlockDriverState *bs, BlockdevAmendOptions *opts, bool force, + Error **errp); /* aio */ BlockAIOCB * GRAPH_RDLOCK_PTR (*bdrv_aio_preadv)(BlockDriverState *bs, diff --git a/block/amend.c b/block/amend.c index XXXXXXX..XXXXXXX 100644 --- a/block/amend.c +++ b/block/amend.c @@ -XXX,XX +XXX,XX @@ static int coroutine_fn blockdev_amend_run(Job *job, Error **errp) { BlockdevAmendJob *s = container_of(job, BlockdevAmendJob, common); int ret; + GRAPH_RDLOCK_GUARD(); job_progress_set_remaining(&s->common, 1); ret = s->bs->drv->bdrv_co_amend(s->bs, s->opts, s->force, errp); @@ -XXX,XX +XXX,XX @@ static int coroutine_fn blockdev_amend_run(Job *job, Error **errp) return ret; } -static int blockdev_amend_pre_run(BlockdevAmendJob *s, Error **errp) +static int GRAPH_RDLOCK +blockdev_amend_pre_run(BlockdevAmendJob *s, Error **errp) { if (s->bs->drv->bdrv_amend_pre_run) { return s->bs->drv->bdrv_amend_pre_run(s->bs, errp); @@ -XXX,XX +XXX,XX @@ static void blockdev_amend_free(Job *job) { BlockdevAmendJob *s = container_of(job, BlockdevAmendJob, common); + bdrv_graph_rdlock_main_loop(); if (s->bs->drv->bdrv_amend_clean) { s->bs->drv->bdrv_amend_clean(s->bs); } + bdrv_graph_rdunlock_main_loop(); bdrv_unref(s->bs); } @@ -XXX,XX +XXX,XX @@ void qmp_x_blockdev_amend(const char *job_id, BlockDriver *drv = bdrv_find_format(fmt); BlockDriverState *bs; + GRAPH_RDLOCK_GUARD_MAINLOOP(); + bs = bdrv_lookup_bs(NULL, node_name, errp); if (!bs) { return; -- 2.40.1 This adds GRAPH_RDLOCK annotations to declare that callers of bdrv_query_bds_stats() need to hold a reader lock for the graph because it accesses the children list of a node. Signed-off-by: Kevin Wolf <kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Message-Id: <20230504115750.54437-18-kwolf@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- block/qapi.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/block/qapi.c b/block/qapi.c index XXXXXXX..XXXXXXX 100644 --- a/block/qapi.c +++ b/block/qapi.c @@ -XXX,XX +XXX,XX @@ static void bdrv_query_blk_stats(BlockDeviceStats *ds, BlockBackend *blk) = bdrv_latency_histogram_stats(&hgram[BLOCK_ACCT_FLUSH]); } -static BlockStats *bdrv_query_bds_stats(BlockDriverState *bs, - bool blk_level) +static BlockStats * GRAPH_RDLOCK +bdrv_query_bds_stats(BlockDriverState *bs, bool blk_level) { BdrvChild *parent_child; BlockDriverState *filter_or_cow_bs; @@ -XXX,XX +XXX,XX @@ BlockStatsList *qmp_query_blockstats(bool has_query_nodes, BlockBackend *blk; BlockDriverState *bs; + GRAPH_RDLOCK_GUARD_MAINLOOP(); + /* Just to be safe if query_nodes is not always initialized */ if (has_query_nodes && query_nodes) { for (bs = bdrv_next_node(NULL); bs; bs = bdrv_next_node(bs)) { -- 2.40.1 This adds GRAPH_RDLOCK annotations to declare that callers of bdrv_query_block_graph_info() need to hold a reader lock for the graph because it accesses the children list of a node. Signed-off-by: Kevin Wolf <kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Message-Id: <20230504115750.54437-19-kwolf@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- include/block/qapi.h | 7 ++++--- qemu-img.c | 2 ++ 2 files changed, 6 insertions(+), 3 deletions(-) diff --git a/include/block/qapi.h b/include/block/qapi.h index XXXXXXX..XXXXXXX 100644 --- a/include/block/qapi.h +++ b/include/block/qapi.h @@ -XXX,XX +XXX,XX @@ #ifndef BLOCK_QAPI_H #define BLOCK_QAPI_H +#include "block/graph-lock.h" #include "block/snapshot.h" #include "qapi/qapi-types-block-core.h" @@ -XXX,XX +XXX,XX @@ void bdrv_query_image_info(BlockDriverState *bs, bool flat, bool skip_implicit_filters, Error **errp); -void bdrv_query_block_graph_info(BlockDriverState *bs, - BlockGraphInfo **p_info, - Error **errp); +void GRAPH_RDLOCK +bdrv_query_block_graph_info(BlockDriverState *bs, BlockGraphInfo **p_info, + Error **errp); void bdrv_snapshot_dump(QEMUSnapshotInfo *sn); void bdrv_image_info_specific_dump(ImageInfoSpecific *info_spec, diff --git a/qemu-img.c b/qemu-img.c index XXXXXXX..XXXXXXX 100644 --- a/qemu-img.c +++ b/qemu-img.c @@ -XXX,XX +XXX,XX @@ static BlockGraphInfoList *collect_image_info_list(bool image_opts, } bs = blk_bs(blk); + GRAPH_RDLOCK_GUARD_MAINLOOP(); + /* * Note that the returned BlockGraphInfo object will not have * information about this image's backing node, because we have opened -- 2.40.1 This adds GRAPH_RDLOCK annotations to declare that callers of bdrv_recurse_can_replace() need to hold a reader lock for the graph because it accesses the children list of a node. Signed-off-by: Kevin Wolf <kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Message-Id: <20230504115750.54437-20-kwolf@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- include/block/block-global-state.h | 5 +++-- include/block/block_int-common.h | 4 ++-- include/block/block_int-global-state.h | 4 ++-- block/blkverify.c | 5 +++-- block/mirror.c | 4 ++++ block/quorum.c | 4 ++-- blockdev.c | 3 +++ 7 files changed, 19 insertions(+), 10 deletions(-) diff --git a/include/block/block-global-state.h b/include/block/block-global-state.h index XXXXXXX..XXXXXXX 100644 --- a/include/block/block-global-state.h +++ b/include/block/block-global-state.h @@ -XXX,XX +XXX,XX @@ int bdrv_amend_options(BlockDriverState *bs_new, QemuOpts *opts, Error **errp); /* check if a named node can be replaced when doing drive-mirror */ -BlockDriverState *check_to_replace_node(BlockDriverState *parent_bs, - const char *node_name, Error **errp); +BlockDriverState * GRAPH_RDLOCK +check_to_replace_node(BlockDriverState *parent_bs, const char *node_name, + Error **errp); int no_coroutine_fn bdrv_activate(BlockDriverState *bs, Error **errp); diff --git a/include/block/block_int-common.h b/include/block/block_int-common.h index XXXXXXX..XXXXXXX 100644 --- a/include/block/block_int-common.h +++ b/include/block/block_int-common.h @@ -XXX,XX +XXX,XX @@ struct BlockDriver { * same data as @bs without it affecting @bs's behavior (that is, * without it being visible to @bs's parents). */ - bool (*bdrv_recurse_can_replace)(BlockDriverState *bs, - BlockDriverState *to_replace); + bool GRAPH_RDLOCK_PTR (*bdrv_recurse_can_replace)( + BlockDriverState *bs, BlockDriverState *to_replace); int (*bdrv_probe_device)(const char *filename); diff --git a/include/block/block_int-global-state.h b/include/block/block_int-global-state.h index XXXXXXX..XXXXXXX 100644 --- a/include/block/block_int-global-state.h +++ b/include/block/block_int-global-state.h @@ -XXX,XX +XXX,XX @@ int bdrv_child_try_set_perm(BdrvChild *c, uint64_t perm, uint64_t shared, */ int bdrv_child_refresh_perms(BlockDriverState *bs, BdrvChild *c, Error **errp); -bool bdrv_recurse_can_replace(BlockDriverState *bs, - BlockDriverState *to_replace); +bool GRAPH_RDLOCK bdrv_recurse_can_replace(BlockDriverState *bs, + BlockDriverState *to_replace); /* * Default implementation for BlockDriver.bdrv_child_perm() that can diff --git a/block/blkverify.c b/block/blkverify.c index XXXXXXX..XXXXXXX 100644 --- a/block/blkverify.c +++ b/block/blkverify.c @@ -XXX,XX +XXX,XX @@ static int coroutine_fn GRAPH_RDLOCK blkverify_co_flush(BlockDriverState *bs) return bdrv_co_flush(s->test_file->bs); } -static bool blkverify_recurse_can_replace(BlockDriverState *bs, - BlockDriverState *to_replace) +static bool GRAPH_RDLOCK +blkverify_recurse_can_replace(BlockDriverState *bs, + BlockDriverState *to_replace) { BDRVBlkverifyState *s = bs->opaque; diff --git a/block/mirror.c b/block/mirror.c index XXXXXXX..XXXXXXX 100644 --- a/block/mirror.c +++ b/block/mirror.c @@ -XXX,XX +XXX,XX @@ static int mirror_exit_common(Job *job) * Cannot use check_to_replace_node() here, because that would * check for an op blocker on @to_replace, and we have our own * there. + * + * TODO Pull out the writer lock from bdrv_replace_node() to here */ + bdrv_graph_rdlock_main_loop(); if (bdrv_recurse_can_replace(src, to_replace)) { bdrv_replace_node(to_replace, target_bs, &local_err); } else { @@ -XXX,XX +XXX,XX @@ static int mirror_exit_common(Job *job) "would not lead to an abrupt change of visible data", to_replace->node_name, target_bs->node_name); } + bdrv_graph_rdunlock_main_loop(); bdrv_drained_end(target_bs); if (local_err) { error_report_err(local_err); diff --git a/block/quorum.c b/block/quorum.c index XXXXXXX..XXXXXXX 100644 --- a/block/quorum.c +++ b/block/quorum.c @@ -XXX,XX +XXX,XX @@ static coroutine_fn GRAPH_RDLOCK int quorum_co_flush(BlockDriverState *bs) return result; } -static bool quorum_recurse_can_replace(BlockDriverState *bs, - BlockDriverState *to_replace) +static bool GRAPH_RDLOCK +quorum_recurse_can_replace(BlockDriverState *bs, BlockDriverState *to_replace) { BDRVQuorumState *s = bs->opaque; int i; diff --git a/blockdev.c b/blockdev.c index XXXXXXX..XXXXXXX 100644 --- a/blockdev.c +++ b/blockdev.c @@ -XXX,XX +XXX,XX @@ static void blockdev_mirror_common(const char *job_id, BlockDriverState *bs, BlockDriverState *unfiltered_bs; int job_flags = JOB_DEFAULT; + GLOBAL_STATE_CODE(); + GRAPH_RDLOCK_GUARD_MAINLOOP(); + if (!has_speed) { speed = 0; } -- 2.40.1 This adds GRAPH_RDLOCK annotations to declare that callers of bdrv_refresh_limits() need to hold a reader lock for the graph because it accesses the children list of a node. Signed-off-by: Kevin Wolf <kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Message-Id: <20230504115750.54437-21-kwolf@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- include/block/block-global-state.h | 5 ++++- include/block/block_int-common.h | 3 ++- block.c | 9 +++++++++ block/io.c | 1 - 4 files changed, 15 insertions(+), 3 deletions(-) diff --git a/include/block/block-global-state.h b/include/block/block-global-state.h index XXXXXXX..XXXXXXX 100644 --- a/include/block/block-global-state.h +++ b/include/block/block-global-state.h @@ -XXX,XX +XXX,XX @@ int bdrv_reopen_set_read_only(BlockDriverState *bs, bool read_only, BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs, const char *backing_file); void bdrv_refresh_filename(BlockDriverState *bs); -void bdrv_refresh_limits(BlockDriverState *bs, Transaction *tran, Error **errp); + +void GRAPH_RDLOCK +bdrv_refresh_limits(BlockDriverState *bs, Transaction *tran, Error **errp); + int bdrv_commit(BlockDriverState *bs); int bdrv_make_empty(BdrvChild *c, Error **errp); int bdrv_change_backing_file(BlockDriverState *bs, const char *backing_file, diff --git a/include/block/block_int-common.h b/include/block/block_int-common.h index XXXXXXX..XXXXXXX 100644 --- a/include/block/block_int-common.h +++ b/include/block/block_int-common.h @@ -XXX,XX +XXX,XX @@ struct BlockDriver { int (*bdrv_debug_resume)(BlockDriverState *bs, const char *tag); bool (*bdrv_debug_is_suspended)(BlockDriverState *bs, const char *tag); - void (*bdrv_refresh_limits)(BlockDriverState *bs, Error **errp); + void GRAPH_RDLOCK_PTR (*bdrv_refresh_limits)( + BlockDriverState *bs, Error **errp); /* * Returns 1 if newly created images are guaranteed to contain only diff --git a/block.c b/block.c index XXXXXXX..XXXXXXX 100644 --- a/block.c +++ b/block.c @@ -XXX,XX +XXX,XX @@ bdrv_open_driver(BlockDriverState *bs, BlockDriver *drv, const char *node_name, return ret; } + bdrv_graph_rdlock_main_loop(); bdrv_refresh_limits(bs, NULL, &local_err); + bdrv_graph_rdunlock_main_loop(); + if (local_err) { error_propagate(errp, local_err); return -EINVAL; @@ -XXX,XX +XXX,XX @@ static int bdrv_set_file_or_backing_noperm(BlockDriverState *parent_bs, } out: + bdrv_graph_rdlock_main_loop(); bdrv_refresh_limits(parent_bs, tran, NULL); + bdrv_graph_rdunlock_main_loop(); return 0; } @@ -XXX,XX +XXX,XX @@ static void bdrv_reopen_commit(BDRVReopenState *reopen_state) qdict_del(bs->explicit_options, "backing"); qdict_del(bs->options, "backing"); + bdrv_graph_rdlock_main_loop(); bdrv_refresh_limits(bs, NULL, NULL); + bdrv_graph_rdunlock_main_loop(); bdrv_refresh_total_sectors(bs, bs->total_sectors); } @@ -XXX,XX +XXX,XX @@ int bdrv_append(BlockDriverState *bs_new, BlockDriverState *bs_top, out: tran_finalize(tran, ret); + bdrv_graph_rdlock_main_loop(); bdrv_refresh_limits(bs_top, NULL, NULL); + bdrv_graph_rdunlock_main_loop(); if (new_context && old_context != new_context) { aio_context_release(new_context); diff --git a/block/io.c b/block/io.c index XXXXXXX..XXXXXXX 100644 --- a/block/io.c +++ b/block/io.c @@ -XXX,XX +XXX,XX @@ void bdrv_refresh_limits(BlockDriverState *bs, Transaction *tran, Error **errp) bool have_limits; GLOBAL_STATE_CODE(); - assume_graph_lock(); /* FIXME */ if (tran) { BdrvRefreshLimitsState *s = g_new(BdrvRefreshLimitsState, 1); -- 2.40.1 From: Stefan Hajnoczi <stefanha@redhat.com> reader_count() is a performance bottleneck because the global aio_context_list_lock mutex causes thread contention. Put this debugging assertion behind a new ./configure --enable-debug-graph-lock option and disable it by default. The --enable-debug-graph-lock option is also enabled by the more general --enable-debug option. Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> Message-Id: <20230501173443.153062-1-stefanha@redhat.com> Reviewed-by: Kevin Wolf <kwolf@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com> --- meson_options.txt | 2 ++ configure | 1 + block/graph-lock.c | 3 +++ meson.build | 2 ++ scripts/meson-buildoptions.sh | 4 ++++ 5 files changed, 12 insertions(+) diff --git a/meson_options.txt b/meson_options.txt index XXXXXXX..XXXXXXX 100644 --- a/meson_options.txt +++ b/meson_options.txt @@ -XXX,XX +XXX,XX @@ option('rng_none', type: 'boolean', value: false, description: 'dummy RNG, avoid using /dev/(u)random and getrandom()') option('coroutine_pool', type: 'boolean', value: true, description: 'coroutine freelist (better performance)') +option('debug_graph_lock', type: 'boolean', value: false, + description: 'graph lock debugging support') option('debug_mutex', type: 'boolean', value: false, description: 'mutex debugging support') option('debug_stack_usage', type: 'boolean', value: false, diff --git a/configure b/configure index XXXXXXX..XXXXXXX 100755 --- a/configure +++ b/configure @@ -XXX,XX +XXX,XX @@ for opt do --enable-debug) # Enable debugging options that aren't excessively noisy debug_tcg="yes" + meson_option_parse --enable-debug-graph-lock "" meson_option_parse --enable-debug-mutex "" meson_option_add -Doptimization=0 fortify_source="no" diff --git a/block/graph-lock.c b/block/graph-lock.c index XXXXXXX..XXXXXXX 100644 --- a/block/graph-lock.c +++ b/block/graph-lock.c @@ -XXX,XX +XXX,XX @@ void bdrv_graph_rdunlock_main_loop(void) void assert_bdrv_graph_readable(void) { + /* reader_count() is slow due to aio_context_list_lock lock contention */ +#ifdef CONFIG_DEBUG_GRAPH_LOCK assert(qemu_in_main_thread() || reader_count()); +#endif } void assert_bdrv_graph_writable(void) diff --git a/meson.build b/meson.build index XXXXXXX..XXXXXXX 100644 --- a/meson.build +++ b/meson.build @@ -XXX,XX +XXX,XX @@ if get_option('debug_stack_usage') and have_coroutine_pool have_coroutine_pool = false endif config_host_data.set10('CONFIG_COROUTINE_POOL', have_coroutine_pool) +config_host_data.set('CONFIG_DEBUG_GRAPH_LOCK', get_option('debug_graph_lock')) config_host_data.set('CONFIG_DEBUG_MUTEX', get_option('debug_mutex')) config_host_data.set('CONFIG_DEBUG_STACK_USAGE', get_option('debug_stack_usage')) config_host_data.set('CONFIG_GPROF', get_option('gprof')) @@ -XXX,XX +XXX,XX @@ summary_info += {'PIE': get_option('b_pie')} summary_info += {'static build': config_host.has_key('CONFIG_STATIC')} summary_info += {'malloc trim support': has_malloc_trim} summary_info += {'membarrier': have_membarrier} +summary_info += {'debug graph lock': get_option('debug_graph_lock')} summary_info += {'debug stack usage': get_option('debug_stack_usage')} summary_info += {'mutex debugging': get_option('debug_mutex')} summary_info += {'memory allocator': get_option('malloc')} diff --git a/scripts/meson-buildoptions.sh b/scripts/meson-buildoptions.sh index XXXXXXX..XXXXXXX 100644 --- a/scripts/meson-buildoptions.sh +++ b/scripts/meson-buildoptions.sh @@ -XXX,XX +XXX,XX @@ meson_options_help() { printf "%s\n" ' QEMU' printf "%s\n" ' --enable-cfi Control-Flow Integrity (CFI)' printf "%s\n" ' --enable-cfi-debug Verbose errors in case of CFI violation' + printf "%s\n" ' --enable-debug-graph-lock' + printf "%s\n" ' graph lock debugging support' printf "%s\n" ' --enable-debug-mutex mutex debugging support' printf "%s\n" ' --enable-debug-stack-usage' printf "%s\n" ' measure coroutine stack usage' @@ -XXX,XX +XXX,XX @@ _meson_option_parse() { --datadir=*) quote_sh "-Ddatadir=$2" ;; --enable-dbus-display) printf "%s" -Ddbus_display=enabled ;; --disable-dbus-display) printf "%s" -Ddbus_display=disabled ;; + --enable-debug-graph-lock) printf "%s" -Ddebug_graph_lock=true ;; + --disable-debug-graph-lock) printf "%s" -Ddebug_graph_lock=false ;; --enable-debug-mutex) printf "%s" -Ddebug_mutex=true ;; --disable-debug-mutex) printf "%s" -Ddebug_mutex=false ;; --enable-debug-stack-usage) printf "%s" -Ddebug_stack_usage=true ;; -- 2.40.1