Am 17.06.22 um 15:25 schrieb Martin Wilck:
On Fri, 2022-06-17 at 05:55 -0400, Cristian Rodríguez wrote:
On Wed, Jun 15, 2022 at 12:15 PM Manfred Schwarb <manfred99@gmx.ch> wrote:
Why should they be "not stable" or "not accurate"?
device numbers were never stable.please do not use this stuff to mount filesystems, use UUID= instead.
I don't think Manfred was talking about device numbers (aka major/minor numbers).
I was talking about the sentence "... initialized with persistent storage symlinks (those located in /dev/disk/by-*) which later were considered broken or not accurate" Which means all /dev/disk/by-* are considered legacy, i.e. even /dev/disk/by-uuid/* But probably I got it wrong. Probably the meaning is "some of them later were considered broken" On my box, I have #>ls -l /dev/disk/by-id/ |grep "nvme0n1$" lrwxrwxrwx 1 root root 13 2022-06-17 15:48 -Samsung_SSD_970_PRO_512GB_S463NF0K512719M -> ../../nvme0n1 lrwxrwxrwx 1 root root 13 2022-06-17 15:48 nvme-20025385581b03fdf -> ../../nvme0n1 lrwxrwxrwx 1 root root 13 2022-06-17 15:48 nvme-SNVMe_Samsung_SSD_970_S463NF0K512719M -> ../../nvme0n1 lrwxrwxrwx 1 root root 13 2022-06-17 15:48 nvme-Samsung_SSD_970_PRO_512GB_S463NF0K512719M -> ../../nvme0n1 lrwxrwxrwx 1 root root 13 2022-06-17 15:48 nvme-eui.0025385581b03fdf -> ../../nvme0n1 And the the script tells me: legacy: /dev/disk/by-id/-Samsung_SSD_970_PRO_512GB_S463NF0K512719M /dev/disk/by-id/nvme-20025385581b03fdf /dev/disk/by-id/nvme-SNVMe_Samsung_SSD_970_S463NF0K512719M which means probably, that /dev/disk/by-id/nvme-Samsung_SSD_970_PRO_512GB_S463NF0K512719M /dev/disk/by-id/nvme-eui.0025385581b03fdf are OK. Cheers, Manfred
Anyway, by-path links aren't stable either. PCI bus/device/function enumeration can change when hardware is added to or removed from the system, or sometimes even when the BIOS is updated.
SCSI host/bus/target/lun enumeration is also not cast in stone. Especially the host number may change depending on probing order.
Martin