nvme: mask CSE effects for security receive

The nvme driver will freeze the IO queues in response to an admin
command with CSE bits set. These bits notify the host that the command
that's about to be executed needs to be done exclusively, hence the
freeze.

The Security Receive command is often reported by multiple vendors with
CSE bits set. The reason for this is that the result depends on the
previous Security Send. This has nothing to do with IO queues, though,
so the driver is taking an overly cautious response to seeing this
passthrough command, while unable to fufill the intended admin queue
action.

Rather than freeze IO during this harmless command, mask off the
effects. This freezing is observed to cause IO latency spikes when host
software periodically validates the security state of the drives.

Signed-off-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>

1 files changed, 17 insertions, 0 deletions

diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
index df929ba9bcc2..d1c9402389f9 100644
--- a/drivers/nvme/host/core.c
+++ b/drivers/nvme/host/core.c
@@ -3091,6 +3091,23 @@ static void nvme_init_known_nvm_effects(struct nvme_ctrl *ctrl)
 	log->acs[nvme_admin_sanitize_nvm] |= cpu_to_le32(NVME_CMD_EFFECTS_LBCC |
 						NVME_CMD_EFFECTS_CSE_MASK);
 
+	/*
+	 * The spec says the result of a security receive command depends on
+	 * the previous security send command. As such, many vendors log this
+	 * command as one to submitted only when no other commands to the same
+	 * namespace are outstanding. The intention is to tell the host to
+	 * prevent mixing security send and receive.
+	 *
+	 * This driver can only enforce such exclusive access against IO
+	 * queues, though. We are not readily able to enforce such a rule for
+	 * two commands to the admin queue, which is the only queue that
+	 * matters for this command.
+	 *
+	 * Rather than blindly freezing the IO queues for this effect that
+	 * doesn't even apply to IO, mask it off.
+	 */
+	log->acs[nvme_admin_security_recv] &= ~NVME_CMD_EFFECTS_CSE_MASK;
+
 	log->iocs[nvme_cmd_write] |= cpu_to_le32(NVME_CMD_EFFECTS_LBCC);
 	log->iocs[nvme_cmd_write_zeroes] |= cpu_to_le32(NVME_CMD_EFFECTS_LBCC);
 	log->iocs[nvme_cmd_write_uncor] |= cpu_to_le32(NVME_CMD_EFFECTS_LBCC);