Re: [PATCH 9/9] x86/mwait-idle: Add C-states validation

[Roger Pau Monné <roger.pau@xxxxxxxxxx>]

On Mon, May 04, 2026 at 11:34:40AM +0200, Jan Beulich wrote:
> On 24.04.2026 21:15, Roger Pau Monné wrote:
> > On Thu, Mar 12, 2026 at 05:58:21PM +0100, Jan Beulich wrote:
> >> @@ -1589,6 +1594,41 @@ static char *__init get_cmdline_field(ch
> >>  }
> >>  
> >>  /**
> >> + * validate_cmdline_cstate - Validate a C-state from cmdline.
> >> + * @state: The C-state to validate.
> >> + * @prev_state: The previous C-state in the table or NULL.
> >> + *
> >> + * Return: 0 if the C-state is valid or -EINVAL otherwise.
> > 
> > Hm, I know we picked this up from upstream, but this function would
> > better return a boolean, rather than 0 or -EINVAL.
> 
> I agree, but I didn't want to deviate from their code purely for cosmetic
> reasons.
> 
> >> +static int __init validate_cmdline_cstate(struct cpuidle_state *state,
> >> +                                    struct cpuidle_state *prev_state)
> >> +{
> >> +  if (state->exit_latency == 0)
> >> +          /* Exit latency 0 can only be used for the POLL state */
> >> +          return -EINVAL;
> >> +
> >> +  if (state->exit_latency > MAX_CMDLINE_LATENCY_US)
> >> +          return -EINVAL;
> >> +
> >> +  if (state->target_residency > MAX_CMDLINE_RESIDENCY_US)
> >> +          return -EINVAL;
> >> +
> >> +  if (state->target_residency < state->exit_latency)
> >> +          return -EINVAL;
> >> +
> >> +  if (!prev_state)
> >> +          return 0;
> >> +
> >> +  if (state->exit_latency <= prev_state->exit_latency)
> >> +          return -EINVAL;
> >> +
> >> +  if (state->target_residency <= prev_state->target_residency)
> >> +          return -EINVAL;
> > 
> > I'm not an expert on C-states, but isn't this checking against the
> > previous value kind of defeating part of the purpose of the command
> > line?
> 
> I don't know. The question would need raising to the author.
> 
> > Also, it might help to also write down those limits in the command
> > line documentation.
> 
> What do you mean there? Some of the values are universal, but some
> checks are against model-specific values. I don't think you mean to
> enumerate them all?

Maybe it's indeed not very useful.  What I referring to was something
along the lines of: "the command line provided residency and latency
values must be smaller than the default ones".  As noted above it
seems weird to me than higher than current values cannot be set,
albeit I have no idea what's the expected usage of this interface.

Thanks, Roger.