]> git.proxmox.com Git - pve-storage.git/commit
Revert "Fix #2020: use /sys to map nvmeXnY to nvmeX"
authorFiona Ebner <f.ebner@proxmox.com>
Mon, 28 Nov 2022 11:29:49 +0000 (12:29 +0100)
committerThomas Lamprecht <t.lamprecht@proxmox.com>
Wed, 30 Nov 2022 15:33:09 +0000 (16:33 +0100)
commit668c15043f933e3025d955f7e0ae9962c13a942c
tree65e27fbab64f0a16010c82bd9ef0f85a708fd997
parent4d5fb344ac663dd0abd66054c863b59e6b5d2741
Revert "Fix #2020: use /sys to map nvmeXnY to nvmeX"

This reverts commit c3442aa5546b029a524928d10c7ecabe0024c137.

Nowadays, relying on 'readlink /sys/block/nvmeXnY/device' won't always
lead to the correct device, as reported in the community forum[0],
where it results in '../../nvme-subsys0' and there's no matching entry
under '/dev/'.

Since Linux kernel 5.4, in particular commit 733e4b69d508 ("nvme:
Assign subsys instance from first ctrl"), the problematic situation
from bug #2020 shouldn't happen anymore.

Stated more clearly by the commit's author here[1]:
> Indeed, that commit will make the naming a bit more sane and will
> definitely prevent mistaken identity. It is still possible to
> observe controllers with instances that don't match their
> namespaces, but it is impossible to get a namespace instance that
> matches a non-owning controller.

The only other user of get_sysdir_info() doesn't use the 'device'
entry, so reverting that part is fine too.

[0] https://forum.proxmox.com/threads/113962/
[1] https://github.com/linux-nvme/nvme-cli/issues/510#issuecomment-552508647

Signed-off-by: Fiona Ebner <f.ebner@proxmox.com>
Tested-by: Stefan Hrdlicka <s.hrdlicka@proxmox.com>
PVE/Diskmanage.pm
test/disk_tests/nvme_smart/nvme0/model [deleted file]
test/disk_tests/nvme_smart/nvme0n1/device [deleted symlink]
test/disk_tests/nvme_smart/nvme0n1/device/model [new file with mode: 0644]