In nvmf_dev_write we did check if the /dev/nvme-fabrics device node's private data is already set and then create a controller data structure afterwards. The private data is protected by the nvmf_dev_mutex, but there is no need to hold it while calling nvmf_create_ctrl().
This also reduces the number of lockdep complaints in the 'nvme connect' with fcloop scenario. [ 9.703333] ============================= [ 9.704797] WARNING: suspicious RCU usage [ 9.706250] 4.17.0-rc5+ #883 Not tainted [ 9.708146] ----------------------------- [ 9.708868] ./include/linux/rcupdate.h:304 Illegal context switch in RCU read-side critical section! [ 9.710511] [ 9.710511] other info that might help us debug this: [ 9.710511] [ 9.711959] [ 9.711959] rcu_scheduler_active = 2, debug_locks = 1 [ 9.713142] 3 locks held by nvme/1420: [ 9.713800] #0: (ptrval) (nvmf_dev_mutex){+.+.}, at: nvmf_dev_write+0x6a/0xb7d [nvme_fabrics] [ 9.717279] #1: (ptrval) (rcu_read_lock){....}, at: hctx_lock+0x56/0xd0 [ 9.718636] [ 9.718636] stack backtrace: [ 9.720266] CPU: 2 PID: 1420 Comm: nvme Not tainted 4.17.0-rc5+ #883 [ 9.721446] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.0.0-prebuilt.qemu-project.org 04/01/2014 [ 9.723003] Call Trace: [ 9.723453] dump_stack+0x78/0xb3 [ 9.724059] ___might_sleep+0xde/0x250 [ 9.724749] kmem_cache_alloc_trace+0x1ae/0x270 [ 9.725565] fcloop_fcp_req+0x32/0x1a0 [nvme_fcloop] [ 9.726428] nvme_fc_start_fcp_op.part.39+0x193/0x4c0 [nvme_fc] [ 9.727425] blk_mq_dispatch_rq_list+0x7f/0x4a0 [ 9.728219] ? blk_mq_flush_busy_ctxs+0xa8/0xf0 [ 9.729035] blk_mq_sched_dispatch_requests+0x16e/0x170 [ 9.729984] __blk_mq_run_hw_queue+0x79/0xd0 [ 9.730737] __blk_mq_delay_run_hw_queue+0x11c/0x160 [ 9.731647] blk_mq_run_hw_queue+0x63/0xc0 [ 9.732357] blk_mq_sched_insert_request+0xb2/0x140 [ 9.733204] blk_execute_rq+0x64/0xc0 [ 9.733840] __nvme_submit_sync_cmd+0x63/0xd0 [nvme_core] [ 9.734772] nvmf_connect_admin_queue+0x11e/0x190 [nvme_fabrics] [ 9.735815] ? mark_held_locks+0x6b/0x90 [ 9.736504] nvme_fc_create_association+0x35b/0x970 [nvme_fc] [ 9.737489] nvme_fc_create_ctrl+0x5d2/0x830 [nvme_fc] [ 9.738379] nvmf_dev_write+0x92d/0xb7d [nvme_fabrics] [ 9.739253] __vfs_write+0x21/0x130 [ 9.739895] ? selinux_file_permission+0xe9/0x140 [ 9.740698] ? security_file_permission+0x2f/0xb0 [ 9.741526] vfs_write+0xbd/0x1c0 [ 9.742095] ksys_write+0x40/0xa0 [ 9.742681] ? do_syscall_64+0xd/0x190 [ 9.743335] do_syscall_64+0x51/0x190 [ 9.744415] entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Johannes Thumshirn <jthumsh...@suse.de> --- drivers/nvme/host/fabrics.c | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-) diff --git a/drivers/nvme/host/fabrics.c b/drivers/nvme/host/fabrics.c index febf82639b40..757a49b9c5a8 100644 --- a/drivers/nvme/host/fabrics.c +++ b/drivers/nvme/host/fabrics.c @@ -1022,18 +1022,17 @@ static ssize_t nvmf_dev_write(struct file *file, const char __user *ubuf, if (IS_ERR(buf)) return PTR_ERR(buf); + ctrl = nvmf_create_ctrl(nvmf_device, buf, count); + if (IS_ERR(ctrl)) + return PTR_ERR(ctrl); + mutex_lock(&nvmf_dev_mutex); if (seq_file->private) { + nvme_delete_ctrl_sync(ctrl); ret = -EINVAL; goto out_unlock; } - ctrl = nvmf_create_ctrl(nvmf_device, buf, count); - if (IS_ERR(ctrl)) { - ret = PTR_ERR(ctrl); - goto out_unlock; - } - seq_file->private = ctrl; out_unlock: -- 2.16.3