nvme: Fix cntlid validation when not using NVMEoF
authorGuilherme G. Piccoli <gpiccoli@canonical.com>
Wed, 14 Aug 2019 14:26:10 +0000 (11:26 -0300)
committerJens Axboe <axboe@kernel.dk>
Tue, 20 Aug 2019 17:02:10 +0000 (11:02 -0600)
commita89fcca8185633993018dc081d6b021d005e6d0b
tree613431c0907a7577475ebf919459a86a115b9285
parent504db087aaccdb32af61539916409f7dca31ceb5
nvme: Fix cntlid validation when not using NVMEoF

Commit 1b1031ca63b2 ("nvme: validate cntlid during controller initialisation")
introduced a validation for controllers with duplicate cntlid that runs
on nvme_init_subsystem(). The problem is that the validation relies on
ctrl->cntlid, and this value is assigned (from id_ctrl value) after the
call for nvme_init_subsystem() in nvme_init_identify() for non-fabrics
scenario. That leads to ctrl->cntlid always being 0 in case we have a
physical set of controllers in the same subsystem.

This patch fixes that by loading the discovered cntlid id_ctrl value into
ctrl->cntlid before the subsystem initialization, only for the non-fabrics
case. The patch was tested with emulated nvme devices (qemu) having two
controllers in a single subsystem. Without the patch, we couldn't make
it work failing in the duplicate check; when running with the patch, we
could see the subsystem holding both controllers.

For the fabrics case we see ctrl->cntlid has a more intricate relation
with the admin connect, so we didn't change that.

Fixes: 1b1031ca63b2 ("nvme: validate cntlid during controller initialisation")
Signed-off-by: Guilherme G. Piccoli <gpiccoli@canonical.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
drivers/nvme/host/core.c