[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Fri, 6 Jun 2008 21:23:38 +0200
From: "Leon Woestenberg" <leon.woestenberg@...il.com>
To: "David Brownell" <david-b@...bell.net>
Cc: LAK <linux-arm-kernel@...ts.arm.linux.org.uk>,
"Linux Kernel list" <linux-kernel@...r.kernel.org>
Subject: Re: Locking in the (now generic) GPIO infrastructure?
Hello David, all,
On Fri, Jun 6, 2008 at 2:53 PM, David Brownell <david-b@...bell.net> wrote:
> On Wednesday 04 June 2008, Leon Woestenberg wrote:
>> include/asm-arm/arch-ixp4xx/platform.h:
>> static inline void gpio_line_set(u8 line, int value)
>> {
>> if (value == IXP4XX_GPIO_HIGH)
>> *IXP4XX_GPIO_GPOUTR |= (1 << line);
>> else if (value == IXP4XX_GPIO_LOW)
>> *IXP4XX_GPIO_GPOUTR &= ~(1 << line);
>> }
>>
>> Under a Linux kernel where multiple drivers are accessing GPIO, the
>> latter does not seem safe against preemption (assuming the memory
>> read-modify-write is not atomic).
>>
>> Shouldn't GPIO access be protected against concurrent access here?
>
> Well, against an IRQ in the middle of those read/modify/write
> sequences hidden by the "|=" and "&=" syntax. Last I knew,
> no XScale CPUs support on-chip SMP.
>
Indeed, however, I used a kernel with -rt patch (and using PREEMPT RT)
as mentioned in my original e-mail. For completeness I should have
stated this:
The interrupt handlers become kernel threads.
As such they become preemptable (to reduce latencies for any higher
priority threads, such as those from other interrupts or even RT user
tasks).
>> Documentation/gpio.txt does not really mention the locking mechanism
>> assumed to modify GPIO lines.
>
> That function isn't part of the GPIO interface, despite
> its gpio_* prefix, so that's not relevant.
>
> It resembles gpio_set_value() though. That can use at
>
In fact, on the IXP4xx, gpio_set_value() is just gpio_line_set(), so I
think it is valid to understand where the locking should occur (lowest
level, higher level?)
> most spinlocks to establish its atomicity guarantee; it's
> described as "spinlock-safe", and in distinction to the
> gpio_set_value_cansleep() call which could use a mutex or
> other sleeping synch primitive.
>
So, the solution (for the upstream work on -rt) would be to add
spinlock protection to gpio_line_set(), mutex protection for
_cansleep() variants?
Regards,
--
Leon
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists