Re: [PATCH v2 1/2] i8k: Autodetect maximal fan speed and fan RPM multiplier

["Pali Rohár" <pali.rohar@gmail.com>]

[-- Attachment #1: Type: Text/Plain, Size: 2429 bytes --]

On Friday 19 December 2014 20:28:08 Guenter Roeck wrote:
> On Fri, Dec 19, 2014 at 07:51:25PM +0100, Pali Rohár wrote:
> > On Friday 19 December 2014 19:32:37 Guenter Roeck wrote:
> > > > -static int i8k_fan_mult;
> > > > -static int i8k_pwm_mult;
> > > > -static int i8k_fan_max = I8K_FAN_HIGH;
> > > > +static int i8k_fan_mult[2];
> > > > +static int i8k_pwm_mult[2];
> > > > +static int i8k_fan_max[2];
> > > 
> > > The rationale for this change is not explained in the
> > > commit log.
> > > 
> > > Do you have any indication that those values would ever be
> > > different for the two fans, ie that you actually need
> > > arrays here ?
> > 
> > I do not know... But if we decide to use only single value
> > for multiplier and max value which fan to use for
> > autodetection?
> 
> That does not answer my question. That you can not decide
> which fan to use for auto-detection does not mean that the
> result of that auto-detection would be different for
> different fans.
> 

Really I do not know if some dell products which have more fans 
(some Precision models have 2) and each fan is using different 
multiplier or has different max speed value.

> > > > @@ -271,8 +274,25 @@ static int i8k_get_fan_speed(int
> > > > fan)
> > > > 
> > > >  {
> > > >  
> > > >  	struct smm_regs regs = { .eax = I8K_SMM_GET_SPEED, };
> > > > 
> > > > +	if (fan < 0 || fan >= ARRAY_SIZE(i8k_fan_mult))
> > > > +		return -EINVAL;
> > > > +
> > > 
> > > This range check (and probably others) is still
> > > unnecessary.
> > > 
> > > Guenter
> > 
> > No, it is necessary. Function i8k_get_fan_speed is called
> > from ioctl callback with value which comes from userspace.
> > If userspace specify fan out of that array we can get
> > kernel panic.
> 
> Yes, but just because you introduced an array for various
> variables, and you still have the unnecessary check for other
> callers.
> 
> If you want to return -EINVAL for bad ioctl parameters, add a
> range check check there. But that would be a separate patch.
> 
> Guenter
> 

So you want to move that checks into ioctl code and do not do 
checks in smm functions?

> > > >  	regs.ebx = fan & 0xff;
> > > > 
> > > > -	return i8k_smm(&regs) ? : (regs.eax & 0xffff) *
> > > > i8k_fan_mult; +	return i8k_smm(&regs) ? : (regs.eax &
> > > > 0xffff) * i8k_fan_mult[fan]; +}

-- 
Pali Rohár
pali.rohar@gmail.com

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 198 bytes --]