Re: [PATCH] I2C: ISP1301_OMAP: New-style i2c driver updates, part 2

[David Brownell <david-b-yBeKhBN/0LDR7s880joybQ@public.gmane.org>]

On Sunday 16 March 2008, Felipe Balbi wrote:
> +       set_irq_type(OMAP_GPIO_IRQ(14), IRQF_SAMPLE_RANDOM);

You should provide *only* IRQF_TRIGGER_* bits when
you call set_irq_type() ... while the docs are weak
here (too), notice how kernel/irq/manage.c masks
out all except the four bits in IRQF_TRIGGER_MASK.

No irq_chip should pay attention to other bits in
that call.  Other flags, like IRQF_{DISABLED,SHARED},
are driver-specific.

Plus, the whole point is to specify the trigger mode,
so calling this the equivalent of "use the default"
(IRQF_TRIGGER_NONE == 0) is pointless.  :)

Also, I'm sure someone will be tempted to remove the
IRQF_SAMPLE_RANDOM here, and it's something of a fair
argument for externally controlled signals.  It's not
random when an external agent can control delivery of
signals at the millisecond level... hence it's not a
crypto-strong source of randomness.  Luckily most
OMAPs include a hardware RNG (except the 35xx series,
for some odd reason?) so that's not a real issue.

- Dave



_______________________________________________
i2c mailing list
i2c-GZX6beZjE8VD60Wz+7aTrA@public.gmane.org
http://lists.lm-sensors.org/mailman/listinfo/i2c