Re: [RFC] devlink: health: add remediation type

On Sat, 6 Mar 2021 15:48:11 +0100 Andrew Lunn wrote:

+/**

quoted

+ * enum devlink_health_reporter_remedy - severity of remediation procedure
+ * @DLH_REMEDY_NONE: transient error, no remediation required
+ * @DLH_REMEDY_COMP_RESET: associated device component (e.g. device queue)
+ *			will be reset
+ * @DLH_REMEDY_RESET: full device reset, will result in temporary unavailability
+ *			of the device, device configuration should not be lost
+ * @DLH_REMEDY_REINIT: device will be reinitialized and configuration lost
+ * @DLH_REMEDY_POWER_CYCLE: device requires a power cycle to recover
+ * @DLH_REMEDY_REIMAGE: device needs to be reflashed
+ * @DLH_REMEDY_BAD_PART: indication of failing hardware, device needs to be
+ *			replaced
+ *
+ * Used in %DEVLINK_ATTR_HEALTH_REPORTER_REMEDY, categorizes the health reporter
+ * by the severity of the required remediation, and indicates the remediation
+ * type to the user if it can't be applied automatically (e.g. "reimage").
+ */
+enum devlink_health_reporter_remedy {
+	DLH_REMEDY_NONE = 1,
+	DLH_REMEDY_COMP_RESET,
+	DLH_REMEDY_RESET,
+	DLH_REMEDY_REINIT,
+	DLH_REMEDY_POWER_CYCLE,
+	DLH_REMEDY_REIMAGE,
+	DLH_REMEDY_BAD_PART,
+};  

Hi Jakub

Are there any cases where the host is the problem, not the device? The
host driver needs to be unloaded and reloaded? The host needs a
reboot?

I was thinking of REINIT addressing that case. Maybe RELOAD would 
be a better name since that's what the devlink command is called?
But also to answer your direct question, I haven't seen such case 
in practice, I don't think, just trying to cover all the bases.