[PATCH V5 0/9] nvme: pci: fix & improve timeout handling

[ming.lei@redhat.com (Ming Lei)]

Hi Keith,

On Fri, May 11, 2018@02:50:28PM -0600, Keith Busch wrote:
> On Fri, May 11, 2018@08:29:24PM +0800, Ming Lei wrote:
> > Hi,
> > 
> > The 1st patch introduces blk_quiesce_timeout() and blk_unquiesce_timeout()
> > for NVMe, meantime fixes blk_sync_queue().
> > 
> > The 2nd patch covers timeout for admin commands for recovering controller
> > for avoiding possible deadlock.
> > 
> > The 3rd and 4th patches avoid to wait_freeze on queues which aren't frozen.
> > 
> > The last 5 patches fixes several races wrt. NVMe timeout handler, and
> > finally can make blktests block/011 passed. Meantime the NVMe PCI timeout
> > mecanism become much more rebost than before.
> > 
> > gitweb:
> > 	https://github.com/ming1/linux/commits/v4.17-rc-nvme-timeout.V5
> 
> Hi Ming,
> 
> First test with simulated broken links is unsuccessful. I'm getting
> stuck here:
> 
> [<0>] blk_mq_freeze_queue_wait+0x46/0xb0
> [<0>] blk_cleanup_queue+0x78/0x170
> [<0>] nvme_ns_remove+0x137/0x1a0 [nvme_core]
> [<0>] nvme_remove_namespaces+0x86/0xc0 [nvme_core]
> [<0>] nvme_remove+0x6b/0x130 [nvme]
> [<0>] pci_device_remove+0x36/0xb0
> [<0>] device_release_driver_internal+0x157/0x220
> [<0>] nvme_remove_dead_ctrl_work+0x29/0x40 [nvme]
> [<0>] process_one_work+0x170/0x350
> [<0>] worker_thread+0x2e/0x380
> [<0>] kthread+0x111/0x130
> [<0>] ret_from_fork+0x1f/0x30
> 
> 
> Here's the last parts of the kernel logs capturing the failure:
> 
> [  760.679105] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679116] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679120] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679124] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679127] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679131] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679135] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679138] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679141] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679144] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679148] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679151] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679155] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679158] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679161] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679164] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679169] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679172] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679176] nvme nvme1: EH 0: before shutdown
> [  760.679177] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679181] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679185] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679189] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679192] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679196] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679199] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679202] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679240] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679243] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.679246] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> 
> ( above repeats a few more hundred times )
> 
> [  760.679960] nvme nvme1: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff
> [  760.701468] nvme nvme1: EH 0: after shutdown, top eh: 1
> [  760.727099] pci_raw_set_power_state: 62 callbacks suppressed
> [  760.727103] nvme 0000:86:00.0: Refused to change power state, currently in D3

EH may not cover this kind of failure, so it fails in the 1st try.

> [  760.727483] nvme nvme1: EH 0: state 4, eh_done -19, top eh 1
> [  760.727485] nvme nvme1: EH 0: after recovery -19
> [  760.727488] nvme nvme1: EH: fail controller

The above issue(hang in nvme_remove()) is still an old issue, which
is because queues are kept as quiesce during remove, so could you
please test the following change?

diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
index 1dec353388be..c78e5a0cde06 100644
--- a/drivers/nvme/host/core.c
+++ b/drivers/nvme/host/core.c
@@ -3254,6 +3254,11 @@ void nvme_remove_namespaces(struct nvme_ctrl *ctrl)
         */
        if (ctrl->state == NVME_CTRL_DEAD)
                nvme_kill_queues(ctrl);
+       else {
+               if (ctrl->admin_q)
+                       blk_mq_unquiesce_queue(ctrl->admin_q);
+               nvme_start_queues(ctrl);
+       }

        down_write(&ctrl->namespaces_rwsem);
        list_splice_init(&ctrl->namespaces, &ns_list);

BTW, in my environment, it is hard to trigger this failure, so not see
this issue, but I did verify the nested EH which can recover from error
in reset.

Thanks,
Ming