Re: [PATCH v2 10/13] EDAC/altera: Skip the panic notifier if kdump is loaded

[Guilherme G. Piccoli]

On 17/08/2022 16:34, Borislav Petkov wrote:
> [...]
> 
> What is "the failure risk for kdump"?
> 
> Some of the notifiers which run before kdump might fail and thus prevent
> the machine from kdumping?
>

Exactly; some notifiers could break the machine and prevent a successful
kdump. The EDAC one is consider medium risk, due to invasive operations
(register readings on panic situation).


> [...] 
> My question stands: if kdump is loaded and the s10_edac_dberr_handler()
> does not read the the fatal errors and they don't get shown in dmesg
> before the machine panics, how do you intend to show that information to
> the user?
> 
> Because fatal errors are something you absolutely wanna show, at least,
> in dmesg!
> 
> I don't think you can "read" the errors from vmcore - they need to be
> read from the hw registers before the machine dies.
> 

My understanding is the same as yours, i.e., this is not possible to
collect from vmcore, it requires register reading. But again: if you
kdump your machine today, you won't collect this information, patch
changed nothing in that regard.

The one thing it changes is that you'd skip the altera register dump if
kdump is set AND you managed to also set "crash_kexec_post_notifiers".

In case you / Dinh / Tony disagrees with the patch, it's fine and we can
discard it, but then this notifier couldn't run early in the refactor we
are doing, it'd postponed to run later. This are is full of trade-offs,
we just need to choose what compromise solution is preferred by the
majority of developers =)

Cheers,


Guilherme