The patch below does not apply to the 5.9-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to stable@vger.kernel.org.
thanks,
greg k-h
------------------ original commit in Linus's tree ------------------
From e297822b20e7fe683e107aea46e6402adcf99c70 Mon Sep 17 00:00:00 2001
From: Pavel Begunkov asml.silence@gmail.com Date: Wed, 18 Nov 2020 14:56:26 +0000 Subject: [PATCH] io_uring: order refnode recycling
Don't recycle a refnode until we're done with all requests of nodes ejected before.
Signed-off-by: Pavel Begunkov asml.silence@gmail.com Cc: stable@vger.kernel.org # v5.7+ Signed-off-by: Jens Axboe axboe@kernel.dk
diff --git a/fs/io_uring.c b/fs/io_uring.c index 5cb194ca4fce..7d4b755ab451 100644 --- a/fs/io_uring.c +++ b/fs/io_uring.c @@ -205,6 +205,7 @@ struct fixed_file_ref_node { struct list_head file_list; struct fixed_file_data *file_data; struct llist_node llist; + bool done; };
struct fixed_file_data { @@ -7323,10 +7324,6 @@ static void __io_file_put_work(struct fixed_file_ref_node *ref_node) kfree(pfile); }
- spin_lock(&file_data->lock); - list_del(&ref_node->node); - spin_unlock(&file_data->lock); - percpu_ref_exit(&ref_node->refs); kfree(ref_node); percpu_ref_put(&file_data->refs); @@ -7353,17 +7350,32 @@ static void io_file_put_work(struct work_struct *work) static void io_file_data_ref_zero(struct percpu_ref *ref) { struct fixed_file_ref_node *ref_node; + struct fixed_file_data *data; struct io_ring_ctx *ctx; - bool first_add; + bool first_add = false; int delay = HZ;
ref_node = container_of(ref, struct fixed_file_ref_node, refs); - ctx = ref_node->file_data->ctx; + data = ref_node->file_data; + ctx = data->ctx; + + spin_lock(&data->lock); + ref_node->done = true; + + while (!list_empty(&data->ref_list)) { + ref_node = list_first_entry(&data->ref_list, + struct fixed_file_ref_node, node); + /* recycle ref nodes in order */ + if (!ref_node->done) + break; + list_del(&ref_node->node); + first_add |= llist_add(&ref_node->llist, &ctx->file_put_llist); + } + spin_unlock(&data->lock);
- if (percpu_ref_is_dying(&ctx->file_data->refs)) + if (percpu_ref_is_dying(&data->refs)) delay = 0;
- first_add = llist_add(&ref_node->llist, &ctx->file_put_llist); if (!delay) mod_delayed_work(system_wq, &ctx->file_put_work, 0); else if (first_add) @@ -7387,6 +7399,7 @@ static struct fixed_file_ref_node *alloc_fixed_file_ref_node( INIT_LIST_HEAD(&ref_node->node); INIT_LIST_HEAD(&ref_node->file_list); ref_node->file_data = ctx->file_data; + ref_node->done = false; return ref_node; }
@@ -7482,7 +7495,7 @@ static int io_sqe_files_register(struct io_ring_ctx *ctx, void __user *arg,
file_data->node = ref_node; spin_lock(&file_data->lock); - list_add(&ref_node->node, &file_data->ref_list); + list_add_tail(&ref_node->node, &file_data->ref_list); spin_unlock(&file_data->lock); percpu_ref_get(&file_data->refs); return ret; @@ -7641,7 +7654,7 @@ static int __io_sqe_files_update(struct io_ring_ctx *ctx, if (needs_switch) { percpu_ref_kill(&data->node->refs); spin_lock(&data->lock); - list_add(&ref_node->node, &data->ref_list); + list_add_tail(&ref_node->node, &data->ref_list); data->node = ref_node; spin_unlock(&data->lock); percpu_ref_get(&ctx->file_data->refs);