From: Alistair Francis <alistair.francis@wdc.com>

The NVMe Base Specification 2.1 states that:

"""
A host requests an explicit persistent connection ... by specifying a
non-zero Keep Alive Timer value in the Connect command.
"""

As such if we are starting a persistent connection to a discovery
controller and the KATO is currently 0 we need to update KATO to a non
zero value to avoid continuous timeouts on the target.

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
---
 drivers/nvme/host/core.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
index 812c1565114f..bb9685b67338 100644
--- a/drivers/nvme/host/core.c
+++ b/drivers/nvme/host/core.c
@@ -4986,8 +4986,14 @@ void nvme_start_ctrl(struct nvme_ctrl *ctrl)
 	 * checking that they started once before, hence are reconnecting back.
 	 */
 	if (test_bit(NVME_CTRL_STARTED_ONCE, &ctrl->flags) &&
-	    nvme_discovery_ctrl(ctrl))
+	    nvme_discovery_ctrl(ctrl)) {
+		if (!ctrl->kato) {
+			nvme_stop_keep_alive(ctrl);
+			ctrl->kato = NVME_DEFAULT_KATO;
+			nvme_start_keep_alive(ctrl);
+		}
 		nvme_change_uevent(ctrl, "NVME_EVENT=rediscover");
+	}
 
 	if (ctrl->queue_count > 1) {
 		nvme_queue_scan(ctrl);
-- 
2.50.1

On Tue, Sep 02, 2025 at 01:52:11PM +1000, alistair23@gmail.com wrote:
> From: Alistair Francis <alistair.francis@wdc.com>
> 
> The NVMe Base Specification 2.1 states that:
> 
> """
> A host requests an explicit persistent connection ... by specifying a
> non-zero Keep Alive Timer value in the Connect command.
> """
> 
> As such if we are starting a persistent connection to a discovery
> controller and the KATO is currently 0 we need to update KATO to a non
> zero value to avoid continuous timeouts on the target.

Thanks, applied to nvme-6.18.

Looks good:

Reviewed-by: Christoph Hellwig <hch@lst.de>

On Tue, 2025-09-02 at 13:52 +1000, alistair23@gmail.com wrote:
> From: Alistair Francis <alistair.francis@wdc.com>
> 
> The NVMe Base Specification 2.1 states that:
> 
> """
> A host requests an explicit persistent connection ... by specifying a
> non-zero Keep Alive Timer value in the Connect command.
> """
> 
> As such if we are starting a persistent connection to a discovery
> controller and the KATO is currently 0 we need to update KATO to a
> non
> zero value to avoid continuous timeouts on the target.
> 
> 

When would this ever happen? Won't nvme-cli & nvme/host/fabrics.c in
the kernel ensure a PDC (persistent discovery controller) would always
have the KATO either default set to NVMF_DEF_DISC_TMO (i.e. 30s) or any
positive int value & not zero?

Do you have a test log for the above scenario where the KATO ends up
being zero for a PDC?

-Martin

On Tue, Sep 2, 2025 at 8:35 PM Martin George <martinus.gpy@gmail.com> wrote:
>
> On Tue, 2025-09-02 at 13:52 +1000, alistair23@gmail.com wrote:
> > From: Alistair Francis <alistair.francis@wdc.com>
> >
> > The NVMe Base Specification 2.1 states that:
> >
> > """
> > A host requests an explicit persistent connection ... by specifying a
> > non-zero Keep Alive Timer value in the Connect command.
> > """
> >
> > As such if we are starting a persistent connection to a discovery
> > controller and the KATO is currently 0 we need to update KATO to a
> > non
> > zero value to avoid continuous timeouts on the target.
> >
> >
>
> When would this ever happen? Won't nvme-cli & nvme/host/fabrics.c in

It occurs if you perform a `nvme connect` to the discovery nqn of a
Linux target.

> the kernel ensure a PDC (persistent discovery controller) would always
> have the KATO either default set to NVMF_DEF_DISC_TMO (i.e. 30s) or any
> positive int value & not zero?

The kernel doesn't set a default for discovery controllers (hence this patch).

nvme-cli will only set NVMF_DEF_DISC_TMO if the `--persistent`
connection option is supplied to `nvme discover`. But it doesn't set a
KATO for `nvme connect`, even though it's a persistent connection.

Note, that I think that is a bug in nvme-cli and it should be setting
a non zero KATO. I plan on patching that. At the same time if the
kernel knows it's a persistent discovery connection it should also be
setting a default KATO. The kernel is already using a default non-zero
KATO for non discovery nqns (see nvmf_parse_options()). This just
extends the default to apply to persistent discovery controllers.

>
> Do you have a test log for the above scenario where the KATO ends up
> being zero for a PDC?

I do, it's just a lot of keep-alive timeout prints on the target

Alistair

>
> -Martin

On 9/2/25 05:52, alistair23@gmail.com wrote:
> From: Alistair Francis <alistair.francis@wdc.com>
> 
> The NVMe Base Specification 2.1 states that:
> 
> """
> A host requests an explicit persistent connection ... by specifying a
> non-zero Keep Alive Timer value in the Connect command.
> """
> 
> As such if we are starting a persistent connection to a discovery
> controller and the KATO is currently 0 we need to update KATO to a non
> zero value to avoid continuous timeouts on the target.
> 
> Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
> ---
>   drivers/nvme/host/core.c | 8 +++++++-
>   1 file changed, 7 insertions(+), 1 deletion(-)
> 
Reviewed-by: Hannes Reinecke <hare@suse.de>

Cheers,

Hannes
-- 
Dr. Hannes Reinecke                  Kernel Storage Architect
hare@suse.de                                +49 911 74053 688
SUSE Software Solutions GmbH, Frankenstr. 146, 90461 Nürnberg
HRB 36809 (AG Nürnberg), GF: I. Totev, A. McDonald, W. Knoblich