[Mellanox] align platform_reboot to use "hardware reboot" #3320
Conversation
Signed-off-by: Mykola Faryma <mykolaf@mellanox.com>
|
retest this please |
| @@ -31,4 +31,6 @@ if [[ "${EXIT_CODE}" != "${EXIT_SUCCESS}" ]]; then | |||
| fi | |||
| fi | |||
|
|
|||
| exec /sbin/reboot $@ | |||
| echo 1 > /bsp/system/pwr_cycle | |||
There was a problem hiding this comment.
Is it possible to fail? If you stop the BSP service, the file will disappear, and finally reboot will not work. #Closed
There was a problem hiding this comment.
Good point, changed to use the sysfs path itself.
This option is ignored in new code. I remember our discussion concluded that 'force/hardware-reboot' is the default option. When user want to 'normal/graceful-reboot', he/she should provide an option. Refers to: device/mellanox/x86_64-mlnx_msn2700-r0/platform_reboot:13 in 9b655b1. [](commit_id = 9b655b1, deletion_comment = False) |
Signed-off-by: Mykola Faryma <mykolaf@mellanox.com>
|
@qiluo-msft We decided to align with other vendors in using hw-reboot as the default option. Introducing graceful reboot option was put into backlog. The -f option only provides ability to overcome fw-upgrade failure. |
|
retest broadcom please |
|
@jleveque @mykolaf with this change can we know the real reboot reason? User issued a software reboot command, but actually it is triggered by a power cycle behind the scenes, and the "show reboot-cause" will tell the user that the reboot cause is not a "software reboot" but a "power cycle", seems quite confusing, would like to hear from you. |
|
@keboliu: This change definitely has the potential to provide the incorrect reboot reason. There needs to be accounting performed on the platform side to prevent this (e.g., if there is a way to check whether the software triggered the hardware reset via some register). |
|
@jleveque @stephenxs I think we may need to change current logic, maybe combine check software reboot cause and hardware reboot cause? in this case, we do have a hardware reboot cause - power cycle, with current logic the software reboot cause check will be skipped. |
If most of the vendors implement reboot by using powercycle, is it better to update the logic of process-reboot-cause to checking /host/reboot-cause/reboot-cause.txt first and then checking the platform-dependent hardware reboot cause? |
I don't understand, how this option is ignored in new code? |
Signed-off-by: Mykola Faryma mykolaf@mellanox.com
- What I did
Aligned Mellanox specific platform_reboot script to perform a power cycle instead of calling /sbin/reboot. Also added verbosity to the fw upgrade script call.
Motivation: switches were reported "stuck" after a reboot call, most vendors use "hardware reboot" in their platform_reboot scripts. We now go with the same approach of a more harsh reboot.
- How I did it
- How to verify it
- Description for the changelog
Mellanox platform_reboot to use power cycle
- A picture of a cute animal (not mandatory but encouraged)