The patch titled Subject: aio-use-cancellation-list-lazily-fix has been added to the -mm tree. Its filename is aio-use-cancellation-list-lazily-fix.patch Before you just go and hit "reply", please: a) Consider who else should be cc'ed b) Prefer to cc a suitable mailing list as well c) Ideally: find the original patch on the mailing list and do a reply-to-all to that, adding suitable additional cc's *** Remember to use Documentation/SubmitChecklist when testing your code *** The -mm tree is included into linux-next and is updated there every 3-4 working days ------------------------------------------------------ From: Kent Overstreet <koverstreet@xxxxxxxxxx> Subject: aio-use-cancellation-list-lazily-fix The cancellation changes were fubar - we can't cancel a kiocb if it doesn't actually have a cancellation callback. The use of xchg() in aio_complete() was right - there we're marking the kiocb as completed - but we need to use cmpxchg() in kiocb_cancel() - a lock isn't sufficient since we're synchronizing with aio_complete() which isn't taking any locks. Signed-off-by: Kent Overstreet <koverstreet@xxxxxxxxxx> Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> --- fs/aio.c | 32 ++++++++++++++++++++++---------- include/linux/aio.h | 11 +++++++++++ 2 files changed, 33 insertions(+), 10 deletions(-) diff -puN fs/aio.c~aio-use-cancellation-list-lazily-fix fs/aio.c --- a/fs/aio.c~aio-use-cancellation-list-lazily-fix +++ a/fs/aio.c @@ -224,28 +224,40 @@ static int aio_setup_ring(struct kioctx void kiocb_set_cancel_fn(struct kiocb *req, kiocb_cancel_fn *cancel) { - if (!req->ki_list.next) { - struct kioctx *ctx = req->ki_ctx; - unsigned long flags; + struct kioctx *ctx = req->ki_ctx; + unsigned long flags; - spin_lock_irqsave(&ctx->ctx_lock, flags); + spin_lock_irqsave(&ctx->ctx_lock, flags); + + if (!req->ki_list.next) list_add(&req->ki_list, &ctx->active_reqs); - spin_unlock_irqrestore(&ctx->ctx_lock, flags); - } req->ki_cancel = cancel; + + spin_unlock_irqrestore(&ctx->ctx_lock, flags); } EXPORT_SYMBOL(kiocb_set_cancel_fn); static int kiocb_cancel(struct kioctx *ctx, struct kiocb *kiocb, struct io_event *res) { - kiocb_cancel_fn *cancel; + kiocb_cancel_fn *old, *cancel; int ret = -EINVAL; - cancel = xchg(&kiocb->ki_cancel, KIOCB_CANCELLED); - if (!cancel || cancel == KIOCB_CANCELLED) - return ret; + /* + * Don't want to set kiocb->ki_cancel = KIOCB_CANCELLED unless it + * actually has a cancel function, hence the cmpxchg() + */ + + cancel = ACCESS_ONCE(kiocb->ki_cancel); + do { + if (!cancel || cancel == KIOCB_CANCELLED) + return ret; + + BUG(); + old = cancel; + cancel = cmpxchg(&kiocb->ki_cancel, old, KIOCB_CANCELLED); + } while (cancel != old); atomic_inc(&kiocb->ki_users); spin_unlock_irq(&ctx->ctx_lock); diff -puN include/linux/aio.h~aio-use-cancellation-list-lazily-fix include/linux/aio.h --- a/include/linux/aio.h~aio-use-cancellation-list-lazily-fix +++ a/include/linux/aio.h @@ -14,6 +14,17 @@ struct kiocb; #define KIOCB_SYNC_KEY (~0U) +/* + * We use ki_cancel == KIOCB_CANCELLED to indicate that a kiocb has been either + * cancelled or completed (this makes a certain amount of sense because + * successful cancellation - io_cancel() - does deliver the completion to + * userspace). + * + * And since most things don't implement kiocb cancellation and we'd really like + * kiocb completion to be lockless when possible, we use ki_cancel to + * synchronize cancellation and completion - we only set it to KIOCB_CANCELLED + * with xchg() or cmpxchg(), see batch_complete_aio() and kiocb_cancel(). + */ #define KIOCB_CANCELLED ((void *) (~0ULL)) typedef int (kiocb_cancel_fn)(struct kiocb *, struct io_event *); _ Patches currently in -mm which might be from koverstreet@xxxxxxxxxx are mm-remove-old-aio-use_mm-comment.patch aio-remove-dead-code-from-aioh.patch gadget-remove-only-user-of-aio-retry.patch aio-remove-retry-based-aio.patch char-add-aio_readwrite-to-dev-nullzero.patch aio-kill-return-value-of-aio_complete.patch aio-kiocb_cancel.patch aio-kiocb_cancel-fix.patch aio-move-private-stuff-out-of-aioh.patch aio-dprintk-pr_debug.patch aio-do-fget-after-aio_get_req.patch aio-make-aio_put_req-lockless.patch aio-refcounting-cleanup.patch wait-add-wait_event_hrtimeout.patch wait-add-wait_event_hrtimeout-fix.patch aio-make-aio_read_evt-more-efficient-convert-to-hrtimers.patch aio-use-flush_dcache_page.patch aio-use-cancellation-list-lazily.patch aio-use-cancellation-list-lazily-fix.patch aio-change-reqs_active-to-include-unreaped-completions.patch aio-kill-batch-allocation.patch aio-kill-struct-aio_ring_info.patch aio-give-shared-kioctx-fields-their-own-cachelines.patch aio-give-shared-kioctx-fields-their-own-cachelines-fix.patch aio-reqs_active-reqs_available.patch aio-percpu-reqs_available.patch generic-dynamic-per-cpu-refcounting.patch generic-dynamic-per-cpu-refcounting-fix.patch generic-dynamic-per-cpu-refcounting-sparse-fixes.patch generic-dynamic-per-cpu-refcounting-sparse-fixes-fix.patch generic-dynamic-per-cpu-refcounting-doc.patch generic-dynamic-per-cpu-refcounting-doc-fix.patch aio-percpu-ioctx-refcount.patch aio-use-xchg-instead-of-completion_lock.patch aio-dont-include-aioh-in-schedh.patch aio-dont-include-aioh-in-schedh-fix.patch aio-dont-include-aioh-in-schedh-fix-fix.patch aio-dont-include-aioh-in-schedh-fix-3.patch aio-kill-ki_key.patch aio-kill-ki_retry.patch aio-kill-ki_retry-fix.patch aio-kill-ki_retry-fix-fix.patch block-aio-batch-completion-for-bios-kiocbs.patch block-aio-batch-completion-for-bios-kiocbs-fix.patch block-aio-batch-completion-for-bios-kiocbs-fix-fix.patch block-aio-batch-completion-for-bios-kiocbs-fix-fix-fix.patch block-aio-batch-completion-for-bios-kiocbs-fix-fix-fix-fix.patch block-aio-batch-completion-for-bios-kiocbs-fix-fix-fix-fix-fix.patch block-aio-batch-completion-for-bios-kiocbs-fix-fix-fix-fix-fix-fix.patch block-aio-batch-completion-for-bios-kiocbs-fix-fix-fix-fix-fix-fix-fix.patch virtio-blk-convert-to-batch-completion.patch mtip32xx-convert-to-batch-completion.patch aio-fix-aio_read_events_ring-types.patch aio-document-clarify-aio_read_events-and-shadow_tail.patch -- To unsubscribe from this list: send the line "unsubscribe mm-commits" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html