On 12/25/21 at 04:21pm, Guilherme G. Piccoli wrote: > On 23/12/2021 22:35, Dave Young wrote: > > Hi Guilherme, > > [...] > > If only the doc update, I think it is fine to be another follup-up > > patch. > > > > About your 1st option in patch log, there is crash_kexec_post_notifiers > > kernel param which can be used to switch on panic notifiers before kdump > > bootup. Another way probably you can try to move panic print to be > > panic notifier. Have this been discussed before? > > > > Hey Dave, thanks for the suggestion. I've considered that but didn't > like the idea. My reasoning was: allowing post notifiers on kdump will > highly compromise the reliability, whereas the panic_print is a solo > option, and not very invasive. > > To mix it with all panic notifiers would just increase a lot the risk of > a kdump failure. Put in other words: if I'm a kdump user and in order to > have this panic_print setting I'd also need to enable post notifiers, > certainly I'll not use the feature, 'cause I don't wanna risk kdump too > much. Hi Guilherme, yes, I have the same concern. But there could be more things like the panic_print in the future, it looks odd to have more kernel cmdline params though. > > One other option I've considered however, and I'd appreciate your > opinion here, would be a new option on crash_kexec_post_notifiers that > allows the users to select *which few notifiers* they want to enable. > Currently it's all or nothing, and this approach is too heavy/risky I > believe. Allowing customization on which post notifiers the user wants > would be much better and in this case, having a post notifier for > panic_print makes a lot of sense. What do you think? It is definitely a good idea, I'm more than glad to see this if you would like to work on this! > > Thanks! > Thanks Dave
