On Wed, Jan 23, 2013 at 10:05:41PM +0200, Aaro Koskinen wrote:
> On Wed, Jan 23, 2013 at 12:23:02PM +0200, Felipe Balbi wrote:
> > Hi folks,
> >
> > it's now rebased on today's i2c-embedded/for-next. I boot tested with my
> > Blaze board.
> >
> > Aaro, if you could verify it doesn't break N900, I'd be glad.
>
> It works, but it still introduces a ~4 second delay during the boot:
>
> [ 0.180847] usbcore: registered new device driver usb
> [ 0.181518] musb-omap2430 musb-omap2430: invalid resource
> [ 4.289093] twl 1-0048: PIH (irq 23) chaining IRQs 338..346
> [ 4.289276] twl 1-0048: power (irq 343) chaining IRQs 346..353
> [ 4.321228] twl4030_gpio twl4030_gpio: gpio (irq 338) chaining IRQs
> 354..371
> [ 4.446655] VUSB1V5: 1500 mV normal standby
This delay is coming from omap_i2c_wait_for_bb(), with ARDY interrupt
disabled we need to poll for BB bit longer, it seems for every transfer.
Replacing msleep(1) with usleep_range(800, 1200) drops it half, and with
just cpu_relax() the delay is gone - but obviously may increase the CPU
time usage.
BTW, with TWL RTC on OMAP boards you can quickly sanity check I2C
throughput:
while /bin/true; do cat /sys/class/rtc/rtc0/time ; done | uniq -c
For these patches I get (ignoring the first line of the output):
Before:
88 00:13:07
89 00:13:08
89 00:13:09
88 00:13:10
89 00:13:11
After:
26 00:00:30
26 00:00:31
26 00:00:32
26 00:00:33
26 00:00:34
After + omap_i2c_wait_for_bb() msleep() replaced with cpu_relax():
91 00:04:59
90 00:05:00
91 00:05:01
90 00:05:02
91 00:05:03
A.
--
To unsubscribe from this list: send the line "unsubscribe linux-i2c" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
