Re: [PATCH 3/3] nvme-pci: make nvme reset more reliable

[Ming Lei <ming.lei@redhat.com>]

On Mon, May 25, 2020 at 10:01:18PM -0700, Dongli Zhang wrote:
> 
> 
> On 5/20/20 4:56 AM, Ming Lei wrote:
> > During waiting for in-flight IO completion in reset handler, timeout
> 
> Does this indicate the window since nvme_start_queues() in nvme_reset_work(),
> that is, just after the queues are unquiesced again?

Right, nvme_start_queues() starts to dispatch requests again, and
nvme_wait_freeze() waits completion of all these in-flight IOs.

> 
> If v2 is required in the future, how about to mention the specific function to
> that it would be much more easier to track the issue.

Not sure it is needed, cause it is quite straightforward.

> 
> > or controller failure still may happen, then the controller is deleted
> > and all inflight IOs are failed. This way is too violent.
> > 
> > Improve the reset handling by replacing nvme_wait_freeze with query
> > & check controller. If all ns queues are frozen, the controller is reset
> > successfully, otherwise check and see if the controller has been disabled.
> > If yes, break from the current recovery and schedule a fresh new reset.
> > 
> > This way avoids to failing IO & removing controller unnecessarily.
> > 
> > Cc: Christoph Hellwig <hch@lst.de>
> > Cc: Sagi Grimberg <sagi@grimberg.me>
> > Cc: Keith Busch <kbusch@kernel.org>
> > Cc: Max Gurtovoy <maxg@mellanox.com>
> > Signed-off-by: Ming Lei <ming.lei@redhat.com>
> > ---
> >  drivers/nvme/host/pci.c | 37 ++++++++++++++++++++++++++++++-------
> >  1 file changed, 30 insertions(+), 7 deletions(-)
> > 
> > diff --git a/drivers/nvme/host/pci.c b/drivers/nvme/host/pci.c
> > index ce0d1e79467a..b5aeed33a634 100644
> > --- a/drivers/nvme/host/pci.c
> > +++ b/drivers/nvme/host/pci.c
> > @@ -24,6 +24,7 @@
> >  #include <linux/io-64-nonatomic-lo-hi.h>
> >  #include <linux/sed-opal.h>
> >  #include <linux/pci-p2pdma.h>
> > +#include <linux/delay.h>
> >  
> >  #include "trace.h"
> >  #include "nvme.h"
> > @@ -1235,9 +1236,6 @@ static enum blk_eh_timer_return nvme_timeout(struct request *req, bool reserved)
> >  	 * shutdown, so we return BLK_EH_DONE.
> >  	 */
> >  	switch (dev->ctrl.state) {
> > -	case NVME_CTRL_CONNECTING:
> > -		nvme_change_ctrl_state(&dev->ctrl, NVME_CTRL_DELETING);
> > -		/* fall through */
> >  	case NVME_CTRL_DELETING:
> >  		dev_warn_ratelimited(dev->ctrl.device,
> >  			 "I/O %d QID %d timeout, disable controller\n",
> > @@ -2393,7 +2391,8 @@ static void nvme_dev_disable(struct nvme_dev *dev, bool shutdown)
> >  		u32 csts = readl(dev->bar + NVME_REG_CSTS);
> >  
> >  		if (dev->ctrl.state == NVME_CTRL_LIVE ||
> > -		    dev->ctrl.state == NVME_CTRL_RESETTING) {
> > +		    dev->ctrl.state == NVME_CTRL_RESETTING ||
> > +		    dev->ctrl.state == NVME_CTRL_CONNECTING) {
> >  			freeze = true;
> >  			nvme_start_freeze(&dev->ctrl);
> >  		}
> > @@ -2504,12 +2503,29 @@ static void nvme_remove_dead_ctrl(struct nvme_dev *dev)
> >  		nvme_put_ctrl(&dev->ctrl);
> >  }
> >  
> > +static bool nvme_wait_freeze_and_check(struct nvme_dev *dev)
> > +{
> > +	bool frozen;
> > +
> > +	while (true) {
> > +		frozen = nvme_frozen(&dev->ctrl);
> > +		if (frozen)
> > +			break;
> 
> ... and how about to comment that the below is because of nvme timeout handler
> as explained in another email (if v2 would be sent) so that it is not required
> to query for "online_queues" with cscope :)
> 
> > +		if (!dev->online_queues)
> > +			break;
> > +		msleep(5);

Fine.


Thanks,
Ming