diff options
Diffstat (limited to 'fs')
| -rw-r--r-- | fs/btrfs/block-group.c | 1 | ||||
| -rw-r--r-- | fs/btrfs/ctree.h | 5 | ||||
| -rw-r--r-- | fs/btrfs/delalloc-space.c | 7 | ||||
| -rw-r--r-- | fs/btrfs/disk-io.c | 6 | ||||
| -rw-r--r-- | fs/btrfs/file.c | 43 | ||||
| -rw-r--r-- | fs/btrfs/inode-map.c | 4 | ||||
| -rw-r--r-- | fs/btrfs/inode.c | 12 | ||||
| -rw-r--r-- | fs/btrfs/ioctl.c | 6 | ||||
| -rw-r--r-- | fs/btrfs/qgroup.c | 4 | ||||
| -rw-r--r-- | fs/btrfs/relocation.c | 9 | ||||
| -rw-r--r-- | fs/dax.c | 5 | ||||
| -rw-r--r-- | fs/gfs2/ops_fstype.c | 1 | ||||
| -rw-r--r-- | fs/io_uring.c | 207 |
13 files changed, 158 insertions, 152 deletions
diff --git a/fs/btrfs/block-group.c b/fs/btrfs/block-group.c index bf7e3f23bba7..670700cb1110 100644 --- a/fs/btrfs/block-group.c +++ b/fs/btrfs/block-group.c @@ -1761,6 +1761,7 @@ int btrfs_read_block_groups(struct btrfs_fs_info *info) btrfs_err(info, "bg %llu is a mixed block group but filesystem hasn't enabled mixed block groups", cache->key.objectid); + btrfs_put_block_group(cache); ret = -EINVAL; goto error; } diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 19d669d12ca1..fe2b8765d9e6 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -734,8 +734,6 @@ struct btrfs_fs_info { struct btrfs_workqueue *fixup_workers; struct btrfs_workqueue *delayed_workers; - /* the extent workers do delayed refs on the extent allocation tree */ - struct btrfs_workqueue *extent_workers; struct task_struct *transaction_kthread; struct task_struct *cleaner_kthread; u32 thread_pool_size; @@ -2489,8 +2487,7 @@ int btrfs_subvolume_reserve_metadata(struct btrfs_root *root, int nitems, bool use_global_rsv); void btrfs_subvolume_release_metadata(struct btrfs_fs_info *fs_info, struct btrfs_block_rsv *rsv); -void btrfs_delalloc_release_extents(struct btrfs_inode *inode, u64 num_bytes, - bool qgroup_free); +void btrfs_delalloc_release_extents(struct btrfs_inode *inode, u64 num_bytes); int btrfs_delalloc_reserve_metadata(struct btrfs_inode *inode, u64 num_bytes); u64 btrfs_account_ro_block_groups_free_space(struct btrfs_space_info *sinfo); diff --git a/fs/btrfs/delalloc-space.c b/fs/btrfs/delalloc-space.c index d949d7d2abed..db9f2c58eb4a 100644 --- a/fs/btrfs/delalloc-space.c +++ b/fs/btrfs/delalloc-space.c @@ -381,7 +381,6 @@ int btrfs_delalloc_reserve_metadata(struct btrfs_inode *inode, u64 num_bytes) out_qgroup: btrfs_qgroup_free_meta_prealloc(root, qgroup_reserve); out_fail: - btrfs_inode_rsv_release(inode, true); if (delalloc_lock) mutex_unlock(&inode->delalloc_mutex); return ret; @@ -418,7 +417,6 @@ void btrfs_delalloc_release_metadata(struct btrfs_inode *inode, u64 num_bytes, * btrfs_delalloc_release_extents - release our outstanding_extents * @inode: the inode to balance the reservation for. * @num_bytes: the number of bytes we originally reserved with - * @qgroup_free: do we need to free qgroup meta reservation or convert them. * * When we reserve space we increase outstanding_extents for the extents we may * add. Once we've set the range as delalloc or created our ordered extents we @@ -426,8 +424,7 @@ void btrfs_delalloc_release_metadata(struct btrfs_inode *inode, u64 num_bytes, * temporarily tracked outstanding_extents. This _must_ be used in conjunction * with btrfs_delalloc_reserve_metadata. */ -void btrfs_delalloc_release_extents(struct btrfs_inode *inode, u64 num_bytes, - bool qgroup_free) +void btrfs_delalloc_release_extents(struct btrfs_inode *inode, u64 num_bytes) { struct btrfs_fs_info *fs_info = inode->root->fs_info; unsigned num_extents; @@ -441,7 +438,7 @@ void btrfs_delalloc_release_extents(struct btrfs_inode *inode, u64 num_bytes, if (btrfs_is_testing(fs_info)) return; - btrfs_inode_rsv_release(inode, qgroup_free); + btrfs_inode_rsv_release(inode, true); } /** diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index 044981cf6df9..402b61bf345c 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -2008,7 +2008,6 @@ static void btrfs_stop_all_workers(struct btrfs_fs_info *fs_info) btrfs_destroy_workqueue(fs_info->readahead_workers); btrfs_destroy_workqueue(fs_info->flush_workers); btrfs_destroy_workqueue(fs_info->qgroup_rescan_workers); - btrfs_destroy_workqueue(fs_info->extent_workers); /* * Now that all other work queues are destroyed, we can safely destroy * the queues used for metadata I/O, since tasks from those other work @@ -2214,10 +2213,6 @@ static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info, max_active, 2); fs_info->qgroup_rescan_workers = btrfs_alloc_workqueue(fs_info, "qgroup-rescan", flags, 1, 0); - fs_info->extent_workers = - btrfs_alloc_workqueue(fs_info, "extent-refs", flags, - min_t(u64, fs_devices->num_devices, - max_active), 8); if (!(fs_info->workers && fs_info->delalloc_workers && fs_info->submit_workers && fs_info->flush_workers && @@ -2228,7 +2223,6 @@ static int btrfs_init_workqueues(struct btrfs_fs_info *fs_info, fs_info->endio_freespace_worker && fs_info->rmw_workers && fs_info->caching_workers && fs_info->readahead_workers && fs_info->fixup_workers && fs_info->delayed_workers && - fs_info->extent_workers && fs_info->qgroup_rescan_workers)) { return -ENOMEM; } diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c index 27e5b269e729..435a502a3226 100644 --- a/fs/btrfs/file.c +++ b/fs/btrfs/file.c @@ -1692,7 +1692,7 @@ again: force_page_uptodate); if (ret) { btrfs_delalloc_release_extents(BTRFS_I(inode), - reserve_bytes, true); + reserve_bytes); break; } @@ -1704,7 +1704,7 @@ again: if (extents_locked == -EAGAIN) goto again; btrfs_delalloc_release_extents(BTRFS_I(inode), - reserve_bytes, true); + reserve_bytes); ret = extents_locked; break; } @@ -1772,8 +1772,7 @@ again: else free_extent_state(cached_state); - btrfs_delalloc_release_extents(BTRFS_I(inode), reserve_bytes, - true); + btrfs_delalloc_release_extents(BTRFS_I(inode), reserve_bytes); if (ret) { btrfs_drop_pages(pages, num_pages); break; @@ -2068,25 +2067,7 @@ int btrfs_sync_file(struct file *file, loff_t start, loff_t end, int datasync) struct btrfs_trans_handle *trans; struct btrfs_log_ctx ctx; int ret = 0, err; - u64 len; - /* - * If the inode needs a full sync, make sure we use a full range to - * avoid log tree corruption, due to hole detection racing with ordered - * extent completion for adjacent ranges, and assertion failures during - * hole detection. - */ - if (test_bit(BTRFS_INODE_NEEDS_FULL_SYNC, - &BTRFS_I(inode)->runtime_flags)) { - start = 0; - end = LLONG_MAX; - } - - /* - * The range length can be represented by u64, we have to do the typecasts - * to avoid signed overflow if it's [0, LLONG_MAX] eg. from fsync() - */ - len = (u64)end - (u64)start + 1; trace_btrfs_sync_file(file, datasync); btrfs_init_log_ctx(&ctx, inode); @@ -2113,6 +2094,19 @@ int btrfs_sync_file(struct file *file, loff_t start, loff_t end, int datasync) atomic_inc(&root->log_batch); /* + * If the inode needs a full sync, make sure we use a full range to + * avoid log tree corruption, due to hole detection racing with ordered + * extent completion for adjacent ranges, and assertion failures during + * hole detection. Do this while holding the inode lock, to avoid races + * with other tasks. + */ + if (test_bit(BTRFS_INODE_NEEDS_FULL_SYNC, + &BTRFS_I(inode)->runtime_flags)) { + start = 0; + end = LLONG_MAX; + } + + /* * Before we acquired the inode's lock, someone may have dirtied more * pages in the target range. We need to make sure that writeback for * any such pages does not start while we are logging the inode, because @@ -2139,8 +2133,11 @@ int btrfs_sync_file(struct file *file, loff_t start, loff_t end, int datasync) /* * We have to do this here to avoid the priority inversion of waiting on * IO of a lower priority task while holding a transaction open. + * + * Also, the range length can be represented by u64, we have to do the + * typecasts to avoid signed overflow if it's [0, LLONG_MAX]. */ - ret = btrfs_wait_ordered_range(inode, start, len); + ret = btrfs_wait_ordered_range(inode, start, (u64)end - (u64)start + 1); if (ret) { up_write(&BTRFS_I(inode)->dio_sem); inode_unlock(inode); diff --git a/fs/btrfs/inode-map.c b/fs/btrfs/inode-map.c index 63cad7865d75..37345fb6191d 100644 --- a/fs/btrfs/inode-map.c +++ b/fs/btrfs/inode-map.c @@ -501,13 +501,13 @@ again: ret = btrfs_prealloc_file_range_trans(inode, trans, 0, 0, prealloc, prealloc, prealloc, &alloc_hint); if (ret) { - btrfs_delalloc_release_extents(BTRFS_I(inode), prealloc, true); + btrfs_delalloc_release_extents(BTRFS_I(inode), prealloc); btrfs_delalloc_release_metadata(BTRFS_I(inode), prealloc, true); goto out_put; } ret = btrfs_write_out_ino_cache(root, trans, path, inode); - btrfs_delalloc_release_extents(BTRFS_I(inode), prealloc, false); + btrfs_delalloc_release_extents(BTRFS_I(inode), prealloc); out_put: iput(inode); out_release: diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 0f2754eaa05b..c3f386b7cc0b 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2206,7 +2206,7 @@ again: ClearPageChecked(page); set_page_dirty(page); - btrfs_delalloc_release_extents(BTRFS_I(inode), PAGE_SIZE, false); + btrfs_delalloc_release_extents(BTRFS_I(inode), PAGE_SIZE); out: unlock_extent_cached(&BTRFS_I(inode)->io_tree, page_start, page_end, &cached_state); @@ -4951,7 +4951,7 @@ again: if (!page) { btrfs_delalloc_release_space(inode, data_reserved, block_start, blocksize, true); - btrfs_delalloc_release_extents(BTRFS_I(inode), blocksize, true); + btrfs_delalloc_release_extents(BTRFS_I(inode), blocksize); ret = -ENOMEM; goto out; } @@ -5018,7 +5018,7 @@ out_unlock: if (ret) btrfs_delalloc_release_space(inode, data_reserved, block_start, blocksize, true); - btrfs_delalloc_release_extents(BTRFS_I(inode), blocksize, (ret != 0)); + btrfs_delalloc_release_extents(BTRFS_I(inode), blocksize); unlock_page(page); put_page(page); out: @@ -8709,7 +8709,7 @@ static ssize_t btrfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter) } else if (ret >= 0 && (size_t)ret < count) btrfs_delalloc_release_space(inode, data_reserved, offset, count - (size_t)ret, true); - btrfs_delalloc_release_extents(BTRFS_I(inode), count, false); + btrfs_delalloc_release_extents(BTRFS_I(inode), count); } out: if (wakeup) @@ -9059,7 +9059,7 @@ again: unlock_extent_cached(io_tree, page_start, page_end, &cached_state); if (!ret2) { - btrfs_delalloc_release_extents(BTRFS_I(inode), PAGE_SIZE, true); + btrfs_delalloc_release_extents(BTRFS_I(inode), PAGE_SIZE); sb_end_pagefault(inode->i_sb); extent_changeset_free(data_reserved); return VM_FAULT_LOCKED; @@ -9068,7 +9068,7 @@ again: out_unlock: unlock_page(page); out: - btrfs_delalloc_release_extents(BTRFS_I(inode), PAGE_SIZE, (ret != 0)); + btrfs_delalloc_release_extents(BTRFS_I(inode), PAGE_SIZE); btrfs_delalloc_release_space(inode, data_reserved, page_start, reserved_space, (ret != 0)); out_noreserve: diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c index de730e56d3f5..7c145a41decd 100644 --- a/fs/btrfs/ioctl.c +++ b/fs/btrfs/ioctl.c @@ -1360,8 +1360,7 @@ again: unlock_page(pages[i]); put_page(pages[i]); } - btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT, - false); + btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT); extent_changeset_free(data_reserved); return i_done; out: @@ -1372,8 +1371,7 @@ out: btrfs_delalloc_release_space(inode, data_reserved, start_index << PAGE_SHIFT, page_cnt << PAGE_SHIFT, true); - btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT, - true); + btrfs_delalloc_release_extents(BTRFS_I(inode), page_cnt << PAGE_SHIFT); extent_changeset_free(data_reserved); return ret; diff --git a/fs/btrfs/qgroup.c b/fs/btrfs/qgroup.c index c4bb69941c77..3ad151655eb8 100644 --- a/fs/btrfs/qgroup.c +++ b/fs/btrfs/qgroup.c @@ -3629,7 +3629,7 @@ int __btrfs_qgroup_reserve_meta(struct btrfs_root *root, int num_bytes, return 0; BUG_ON(num_bytes != round_down(num_bytes, fs_info->nodesize)); - trace_qgroup_meta_reserve(root, type, (s64)num_bytes); + trace_qgroup_meta_reserve(root, (s64)num_bytes, type); ret = qgroup_reserve(root, num_bytes, enforce, type); if (ret < 0) return ret; @@ -3676,7 +3676,7 @@ void __btrfs_qgroup_free_meta(struct btrfs_root *root, int num_bytes, */ num_bytes = sub_root_meta_rsv(root, num_bytes, type); BUG_ON(num_bytes != round_down(num_bytes, fs_info->nodesize)); - trace_qgroup_meta_reserve(root, type, -(s64)num_bytes); + trace_qgroup_meta_reserve(root, -(s64)num_bytes, type); btrfs_qgroup_free_refroot(fs_info, root->root_key.objectid, num_bytes, type); } diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c index 00504657b602..5cd42b66818c 100644 --- a/fs/btrfs/relocation.c +++ b/fs/btrfs/relocation.c @@ -3277,6 +3277,8 @@ static int relocate_file_extent_cluster(struct inode *inode, if (!page) { btrfs_delalloc_release_metadata(BTRFS_I(inode), PAGE_SIZE, true); + btrfs_delalloc_release_extents(BTRFS_I(inode), + PAGE_SIZE); ret = -ENOMEM; goto out; } @@ -3297,7 +3299,7 @@ static int relocate_file_extent_cluster(struct inode *inode, btrfs_delalloc_release_metadata(BTRFS_I(inode), PAGE_SIZE, true); btrfs_delalloc_release_extents(BTRFS_I(inode), - PAGE_SIZE, true); + PAGE_SIZE); ret = -EIO; goto out; } @@ -3326,7 +3328,7 @@ static int relocate_file_extent_cluster(struct inode *inode, btrfs_delalloc_release_metadata(BTRFS_I(inode), PAGE_SIZE, true); btrfs_delalloc_release_extents(BTRFS_I(inode), - PAGE_SIZE, true); + PAGE_SIZE); clear_extent_bits(&BTRFS_I(inode)->io_tree, page_start, page_end, @@ -3342,8 +3344,7 @@ static int relocate_file_extent_cluster(struct inode *inode, put_page(page); index++; - btrfs_delalloc_release_extents(BTRFS_I(inode), PAGE_SIZE, - false); + btrfs_delalloc_release_extents(BTRFS_I(inode), PAGE_SIZE); balance_dirty_pages_ratelimited(inode->i_mapping); btrfs_throttle(fs_info); } @@ -220,10 +220,11 @@ static void *get_unlocked_entry(struct xa_state *xas, unsigned int order) for (;;) { entry = xas_find_conflict(xas); + if (!entry || WARN_ON_ONCE(!xa_is_value(entry))) + return entry; if (dax_entry_order(entry) < order) return XA_RETRY_ENTRY; - if (!entry || WARN_ON_ONCE(!xa_is_value(entry)) || - !dax_is_locked(entry)) + if (!dax_is_locked(entry)) return entry; wq = dax_entry_waitqueue(xas, entry, &ewait.key); diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c index 681b44682b0d..dc61af2c4d5e 100644 --- a/fs/gfs2/ops_fstype.c +++ b/fs/gfs2/ops_fstype.c @@ -1600,6 +1600,7 @@ static int gfs2_meta_get_tree(struct fs_context *fc) } static const struct fs_context_operations gfs2_meta_context_ops = { + .free = gfs2_fc_free, .get_tree = gfs2_meta_get_tree, }; diff --git a/fs/io_uring.c b/fs/io_uring.c index 67dbe0201e0d..a30c4f622cb3 100644 --- a/fs/io_uring.c +++ b/fs/io_uring.c @@ -197,6 +197,7 @@ struct io_ring_ctx { unsigned sq_entries; unsigned sq_mask; unsigned sq_thread_idle; + unsigned cached_sq_dropped; struct io_uring_sqe *sq_sqes; struct list_head defer_list; @@ -212,6 +213,7 @@ struct io_ring_ctx { struct { unsigned cached_cq_tail; + atomic_t cached_cq_overflow; unsigned cq_entries; unsigned cq_mask; struct wait_queue_head cq_wait; @@ -420,7 +422,8 @@ static struct io_ring_ctx *io_ring_ctx_alloc(struct io_uring_params *p) static inline bool __io_sequence_defer(struct io_ring_ctx *ctx, struct io_kiocb *req) { - return req->sequence != ctx->cached_cq_tail + ctx->rings->sq_dropped; + return req->sequence != ctx->cached_cq_tail + ctx->cached_sq_dropped + + atomic_read(&ctx->cached_cq_overflow); } static inline bool io_sequence_defer(struct io_ring_ctx *ctx, @@ -567,9 +570,8 @@ static void io_cqring_fill_event(struct io_ring_ctx *ctx, u64 ki_user_data, WRITE_ONCE(cqe->res, res); WRITE_ONCE(cqe->flags, 0); } else { - unsigned overflow = READ_ONCE(ctx->rings->cq_overflow); - - WRITE_ONCE(ctx->rings->cq_overflow, overflow + 1); + WRITE_ONCE(ctx->rings->cq_overflow, + atomic_inc_return(&ctx->cached_cq_overflow)); } } @@ -735,6 +737,14 @@ static unsigned io_cqring_events(struct io_rings *rings) return READ_ONCE(rings->cq.tail) - READ_ONCE(rings->cq.head); } +static inline unsigned int io_sqring_entries(struct io_ring_ctx *ctx) +{ + struct io_rings *rings = ctx->rings; + + /* make sure SQ entry isn't read before tail */ + return smp_load_acquire(&rings->sq.tail) - ctx->cached_sq_head; +} + /* * Find and free completed poll iocbs */ @@ -864,19 +874,11 @@ static void io_iopoll_reap_events(struct io_ring_ctx *ctx) mutex_unlock(&ctx->uring_lock); } -static int io_iopoll_check(struct io_ring_ctx *ctx, unsigned *nr_events, - long min) +static int __io_iopoll_check(struct io_ring_ctx *ctx, unsigned *nr_events, + long min) { - int iters, ret = 0; - - /* - * We disallow the app entering submit/complete with polling, but we - * still need to lock the ring to prevent racing with polled issue - * that got punted to a workqueue. - */ - mutex_lock(&ctx->uring_lock); + int iters = 0, ret = 0; - iters = 0; do { int tmin = 0; @@ -912,6 +914,21 @@ static int io_iopoll_check(struct io_ring_ctx *ctx, unsigned *nr_events, ret = 0; } while (min && !*nr_events && !need_resched()); + return ret; +} + +static int io_iopoll_check(struct io_ring_ctx *ctx, unsigned *nr_events, + long min) +{ + int ret; + + /* + * We disallow the app entering submit/complete with polling, but we + * still need to lock the ring to prevent racing with polled issue + * that got punted to a workqueue. + */ + mutex_lock(&ctx->uring_lock); + ret = __io_iopoll_check(ctx, nr_events, min); mutex_unlock(&ctx->uring_lock); return ret; } @@ -1877,7 +1894,7 @@ static int io_poll_add(struct io_kiocb *req, const struct io_uring_sqe *sqe) static enum hrtimer_restart io_timeout_fn(struct hrtimer *timer) { struct io_ring_ctx *ctx; - struct io_kiocb *req; + struct io_kiocb *req, *prev; unsigned long flags; req = container_of(timer, struct io_kiocb, timeout.timer); @@ -1885,6 +1902,15 @@ static enum hrtimer_restart io_timeout_fn(struct hrtimer *timer) atomic_inc(&ctx->cq_timeouts); spin_lock_irqsave(&ctx->completion_lock, flags); + /* + * Adjust the reqs sequence before the current one because it + * will consume a slot in the cq_ring and the the cq_tail pointer + * will be increased, otherwise other timeout reqs may return in + * advance without waiting for enough wait_nr. + */ + prev = req; + list_for_each_entry_continue_reverse(prev, &ctx->timeout_list, list) + prev->sequence++; list_del(&req->list); io_cqring_fill_event(ctx, req->user_data, -ETIME); @@ -1903,6 +1929,7 @@ static int io_timeout(struct io_kiocb *req, const struct io_uring_sqe *sqe) struct io_ring_ctx *ctx = req->ctx; struct list_head *entry; struct timespec64 ts; + unsigned span = 0; if (unlikely(ctx->flags & IORING_SETUP_IOPOLL)) return -EINVAL; @@ -1951,9 +1978,17 @@ static int io_timeout(struct io_kiocb *req, const struct io_uring_sqe *sqe) if (ctx->cached_sq_head < nxt_sq_head) tmp += UINT_MAX; - if (tmp >= tmp_nxt) + if (tmp > tmp_nxt) break; + + /* + * Sequence of reqs after the insert one and itself should + * be adjusted because each timeout req consumes a slot. + */ + span++; + nxt->sequence++; } + req->sequence -= span; list_add(&req->list, entry); spin_unlock_irq(&ctx->completion_lock); @@ -2292,11 +2327,11 @@ static int io_req_set_file(struct io_ring_ctx *ctx, const struct sqe_submit *s, } static int __io_queue_sqe(struct io_ring_ctx *ctx, struct io_kiocb *req, - struct sqe_submit *s, bool force_nonblock) + struct sqe_submit *s) { int ret; - ret = __io_submit_sqe(ctx, req, s, force_nonblock); + ret = __io_submit_sqe(ctx, req, s, true); /* * We async punt it if the file wasn't marked NOWAIT, or if the file @@ -2343,7 +2378,7 @@ static int __io_queue_sqe(struct io_ring_ctx *ctx, struct io_kiocb *req, } static int io_queue_sqe(struct io_ring_ctx *ctx, struct io_kiocb *req, - struct sqe_submit *s, bool force_nonblock) + struct sqe_submit *s) { int ret; @@ -2356,18 +2391,17 @@ static int io_queue_sqe(struct io_ring_ctx *ctx, struct io_kiocb *req, return 0; } - return __io_queue_sqe(ctx, req, s, force_nonblock); + return __io_queue_sqe(ctx, req, s); } static int io_queue_link_head(struct io_ring_ctx *ctx, struct io_kiocb *req, - struct sqe_submit *s, struct io_kiocb *shadow, - bool force_nonblock) + struct sqe_submit *s, struct io_kiocb *shadow) { int ret; int need_submit = false; if (!shadow) - return io_queue_sqe(ctx, req, s, force_nonblock); + return io_queue_sqe(ctx, req, s); /* * Mark the first IO in link list as DRAIN, let all the following @@ -2396,7 +2430,7 @@ static int io_queue_link_head(struct io_ring_ctx *ctx, struct io_kiocb *req, spin_unlock_irq(&ctx->completion_lock); if (need_submit) - return __io_queue_sqe(ctx, req, s, force_nonblock); + return __io_queue_sqe(ctx, req, s); return 0; } @@ -2404,8 +2438,7 @@ static int io_queue_link_head(struct io_ring_ctx *ctx, struct io_kiocb *req, #define SQE_VALID_FLAGS (IOSQE_FIXED_FILE|IOSQE_IO_DRAIN|IOSQE_IO_LINK) static void io_submit_sqe(struct io_ring_ctx *ctx, struct sqe_submit *s, - struct io_submit_state *state, struct io_kiocb **link, - bool force_nonblock) + struct io_submit_state *state, struct io_kiocb **link) { struct io_uring_sqe *sqe_copy; struct io_kiocb *req; @@ -2432,6 +2465,8 @@ err: return; } + req->user_data = s->sqe->user_data; + /* * If we already have a head request, queue this one for async * submittal once the head completes. If we don't have a head but @@ -2458,7 +2493,7 @@ err: INIT_LIST_HEAD(&req->link_list); *link = req; } else { - io_queue_sqe(ctx, req, s, force_nonblock); + io_queue_sqe(ctx, req, s); } } @@ -2538,12 +2573,13 @@ static bool io_get_sqring(struct io_ring_ctx *ctx, struct sqe_submit *s) /* drop invalid entries */ ctx->cached_sq_head++; - rings->sq_dropped++; + ctx->cached_sq_dropped++; + WRITE_ONCE(rings->sq_dropped, ctx->cached_sq_dropped); return false; } -static int io_submit_sqes(struct io_ring_ctx *ctx, struct sqe_submit *sqes, - unsigned int nr, bool has_user, bool mm_fault) +static int io_submit_sqes(struct io_ring_ctx *ctx, unsigned int nr, + bool has_user, bool mm_fault) { struct io_submit_state state, *statep = NULL; struct io_kiocb *link = NULL; @@ -2557,19 +2593,23 @@ static int io_submit_sqes(struct io_ring_ctx *ctx, struct sqe_submit *sqes, } for (i = 0; i < nr; i++) { + struct sqe_submit s; + + if (!io_get_sqring(ctx, &s)) + break; + /* * If previous wasn't linked and we have a linked command, * that's the end of the chain. Submit the previous link. */ if (!prev_was_link && link) { - io_queue_link_head(ctx, link, &link->submit, shadow_req, - true); + io_queue_link_head(ctx, link, &link->submit, shadow_req); link = NULL; shadow_req = NULL; } - prev_was_link = (sqes[i].sqe->flags & IOSQE_IO_LINK) != 0; + prev_was_link = (s.sqe->flags & IOSQE_IO_LINK) != 0; - if (link && (sqes[i].sqe->flags & IOSQE_IO_DRAIN)) { + if (link && (s.sqe->flags & IOSQE_IO_DRAIN)) { if (!shadow_req) { shadow_req = io_get_req(ctx, NULL); if (unlikely(!shadow_req)) @@ -2577,24 +2617,24 @@ static int io_submit_sqes(struct io_ring_ctx *ctx, struct sqe_submit *sqes, shadow_req->flags |= (REQ_F_IO_DRAIN | REQ_F_SHADOW_DRAIN); refcount_dec(&shadow_req->refs); } - shadow_req->sequence = sqes[i].sequence; + shadow_req->sequence = s.sequence; } out: if (unlikely(mm_fault)) { - io_cqring_add_event(ctx, sqes[i].sqe->user_data, + io_cqring_add_event(ctx, s.sqe->user_data, -EFAULT); } else { - sqes[i].has_user = has_user; - sqes[i].needs_lock = true; - sqes[i].needs_fixed_file = true; - io_submit_sqe(ctx, &sqes[i], statep, &link, true); + s.has_user = has_user; + s.needs_lock = true; + s.needs_fixed_file = true; + io_submit_sqe(ctx, &s, statep, &link); submitted++; } } if (link) - io_queue_link_head(ctx, link, &link->submit, shadow_req, true); + io_queue_link_head(ctx, link, &link->submit, shadow_req); if (statep) io_submit_state_end(&state); @@ -2603,7 +2643,6 @@ out: static int io_sq_thread(void *data) { - struct sqe_submit sqes[IO_IOPOLL_BATCH]; struct io_ring_ctx *ctx = data; struct mm_struct *cur_mm = NULL; mm_segment_t old_fs; @@ -2618,14 +2657,27 @@ static int io_sq_thread(void *data) timeout = inflight = 0; while (!kthread_should_park()) { - bool all_fixed, mm_fault = false; - int i; + bool mm_fault = false; + unsigned int to_submit; if (inflight) { unsigned nr_events = 0; if (ctx->flags & IORING_SETUP_IOPOLL) { - io_iopoll_check(ctx, &nr_events, 0); + /* + * inflight is the count of the maximum possible + * entries we submitted, but it can be smaller + * if we dropped some of them. If we don't have + * poll entries available, then we know that we + * have nothing left to poll for. Reset the + * inflight count to zero in that case. + */ + mutex_lock(&ctx->uring_lock); + if (!list_empty(&ctx->poll_list)) + __io_iopoll_check(ctx, &nr_events, 0); + else + inflight = 0; + mutex_unlock(&ctx->uring_lock); } else { /* * Normal IO, just pretend everything completed. @@ -2639,7 +2691,8 @@ static int io_sq_thread(void *data) timeout = jiffies + ctx->sq_thread_idle; } - if (!io_get_sqring(ctx, &sqes[0])) { + to_submit = io_sqring_entries(ctx); + if (!to_submit) { /* * We're polling. If we're within the defined idle * period, then let us spin without work before going @@ -2670,7 +2723,8 @@ static int io_sq_thread(void *data) /* make sure to read SQ tail after writing flags */ smp_mb(); - if (!io_get_sqring(ctx, &sqes[0])) { + to_submit = io_sqring_entries(ctx); + if (!to_submit) { if (kthread_should_park()) { finish_wait(&ctx->sqo_wait, &wait); break; @@ -2688,19 +2742,8 @@ static int io_sq_thread(void *data) ctx->rings->sq_flags &= ~IORING_SQ_NEED_WAKEUP; } - i = 0; - all_fixed = true; - do { - if (all_fixed && io_sqe_needs_user(sqes[i].sqe)) - all_fixed = false; - - i++; - if (i == ARRAY_SIZE(sqes)) - break; - } while (io_get_sqring(ctx, &sqes[i])); - /* Unless all new commands are FIXED regions, grab mm */ - if (!all_fixed && !cur_mm) { + if (!cur_mm) { mm_fault = !mmget_not_zero(ctx->sqo_mm); if (!mm_fault) { use_mm(ctx->sqo_mm); @@ -2708,8 +2751,9 @@ static int io_sq_thread(void *data) } } - inflight += io_submit_sqes(ctx, sqes, i, cur_mm != NULL, - mm_fault); + to_submit = min(to_submit, ctx->sq_entries); + inflight += io_submit_sqes(ctx, to_submit, cur_mm != NULL, + mm_fault); /* Commit SQ ring head once we've consumed all SQEs */ io_commit_sqring(ctx); @@ -2726,8 +2770,7 @@ static int io_sq_thread(void *data) return 0; } -static int io_ring_submit(struct io_ring_ctx *ctx, unsigned int to_submit, - bool block_for_last) +static int io_ring_submit(struct io_ring_ctx *ctx, unsigned int to_submit) { struct io_submit_state state, *statep = NULL; struct io_kiocb *link = NULL; @@ -2741,7 +2784,6 @@ static int io_ring_submit(struct io_ring_ctx *ctx, unsigned int to_submit, } for (i = 0; i < to_submit; i++) { - bool force_nonblock = true; struct sqe_submit s; if (!io_get_sqring(ctx, &s)) @@ -2752,8 +2794,7 @@ static int io_ring_submit(struct io_ring_ctx *ctx, unsigned int to_submit, * that's the end of the chain. Submit the previous link. */ if (!prev_was_link && link) { - io_queue_link_head(ctx, link, &link->submit, shadow_req, - force_nonblock); + io_queue_link_head(ctx, link, &link->submit, shadow_req); link = NULL; shadow_req = NULL; } @@ -2775,27 +2816,16 @@ out: s.needs_lock = false; s.needs_fixed_file = false; submit++; - - /* - * The caller will block for events after submit, submit the - * last IO non-blocking. This is either the only IO it's - * submitting, or it already submitted the previous ones. This - * improves performance by avoiding an async punt that we don't - * need to do. - */ - if (block_for_last && submit == to_submit) - force_nonblock = false; - - io_submit_sqe(ctx, &s, statep, &link, force_nonblock); + io_submit_sqe(ctx, &s, statep, &link); } - io_commit_sqring(ctx); if (link) - io_queue_link_head(ctx, link, &link->submit, shadow_req, - !block_for_last); + io_queue_link_head(ctx, link, &link->submit, shadow_req); if (statep) io_submit_state_end(statep); + io_commit_sqring(ctx); + return submit; } @@ -3636,21 +3666,10 @@ SYSCALL_DEFINE6(io_uring_enter, unsigned int, fd, u32, to_submit, wake_up(&ctx->sqo_wait); submitted = to_submit; } else if (to_submit) { - bool block_for_last = false; - to_submit = min(to_submit, ctx->sq_entries); - /* - * Allow last submission to block in a series, IFF the caller - * asked to wait for events and we don't currently have - * enough. This potentially avoids an async punt. - */ - if (to_submit == min_complete && - io_cqring_events(ctx->rings) < min_complete) - block_for_last = true; - mutex_lock(&ctx->uring_lock); - submitted = io_ring_submit(ctx, to_submit, block_for_last); + submitted = io_ring_submit(ctx, to_submit); mutex_unlock(&ctx->uring_lock); } if (flags & IORING_ENTER_GETEVENTS) { |