On Mon, Oct 24, 2011 at 03:26, Linus Walleij wrote:
On Sat, Oct 22, 2011 at 7:44 PM, Mike Frysinger wrote:
On Tue, Oct 4, 2011 at 16:35, Grant Likely wrote:
On Sat, Oct 01, 2011 at 12:39:21PM +0200, Linus Walleij wrote:
2011/9/30 Grant Likely:
I'm not convinced that the sysfs approach is actually the right interface here (I'm certainly not a fan of the gpio sysfs i/f), and I'd rather not be putting in unneeded stuff until the userspace i/f is hammered out.
Actually, thinking about it I cannot see what would be wrong with /dev/gpio0 & friends in the first place.
Using sysfs as swiss army knife for custom I/O does not seem like it would be long-term viable so thanks for this observation, and I think we need /dev/gpio* put on some mental roadmap somewhere.
Agreed. I don't want to be in the situation we are now with GPIO, where every time I look at the sysfs interface I shudder.
the problem with that is it doesn't scale. if i have a device with over 150 GPIOs on the SoC itself (obviously GPIO expanders can make that much bigger), i don't want to see 150+ device nodes in /dev/. that's a pretty big waste. sysfs only allocates/frees resources when userspace actually wants to utilize a GPIO.
I was more thinking along the lines of one device per GPIO controller, then you ioctl() to ask /dev/gpio0 how many pins it has or so.
that brings its own set of trade offs. this might be OK from a debugging point of view, but it means security wise we have to grant access on a per-gpiochip basis instead of a per-gpio basis. i think the sysfs interface has this granularity of support already as the root user can chmod/chown the files after exporting them.
Grant suggested we extend UIO to export GPIOs. this would be a good trade off i think -- sysfs is a good on-the-fly debugging/scripting interface, but UIO gets us the performance. sysfs overhead can be mitigated by using pwrite/pread, but without pwritev/preadv, we're stuck with 1-transition-per-syscall. -mike