Re: h4350 dmesg

Hello Daniel,

Thursday, February 22, 2007, 7:18:51 AM, you wrote:

> Hi Paul,

> I've compiled from CVS HEAD, and included the output of dmesg at the end
> of this email.

  Thanks! I'm sorry that it takes me more time to get back to you to
do more tests. Preliminary results are good:

> h4000_machine_is_h4300: read a700 from spi, counter: 131

I consistently get 0 on my h4100. Well, let's do more tests to check
consistency of results on h4300. So, please apply attached patch, and
build kernel with CONFIG_IPAQ_H4300_KBD=n (or they will conflict).
With it, line above will be printed to screen every few seconds. So,
what we're looking for:

1. Please check if value (a700) will be consistent. If not, will there
be any pattern values? Will it ever be 0?
2. Press and hold a key on keyboard and wait for that message. It is
expected that low byte of value will be != 0. But what about high
byte?
3. Reboot few times, and see, if consistency is preserved/no.
4. In h4300_detect.c, find line:

   asic3_write_register(a3, ASIC3_SPI_TXDATA, 0xaa);

change 0xaa to 0, then to 0xff. Will there be any changes to value?

  So, as you see, there're some few tests to perform ;-). But with those
results, we'll have full picture of the situation, and know if this
idea of detection is reliable enough, or we need to think about
something else.

  This is also a good sample of what kind of work is needed to get WiFi
and flash working ;-).

> A couple of other notes. Now that power management is
> working well I am using my ipaq as my primary PDA. Here is what I have
> observed so far (all this is with the old kernel from 2-3 weeks ago.

  Very good news, thanks for this testing!

> I'll continue to test with the new ones):

> First the good news:
> 1) I've been lightly using it for about two weeks now, and have not
> charged it once in the past week. The battery has VERY nice longevity.
> 2) Consistantly works well even with repetitive sleep/wake cycles. No
> problems at all.

  Very good!

> 3) Sleep wakup works very nicely. Much nicer than on my Sharp Zaurus
> 5500 which was always slow and sluggish.

  Glad you treat it as such. I personally has intuitive suspicion that
with each new mainline kernel version, resume has increasing latency
;-). I guess we'll need to profile it later yet. More important will
be to make sure that BT and WiFi work across suspend cycle well
(that's yet quite a bit of work).

> 4) Seems very stable so far. Again more stable than on my Zaurus which
> would crash often enough to be annoying.

> Now the stuff that still seems to be troublesome:
> 1) when hci_uart is loaded the power is drained VERY quickly even when
> the device is in sleep mode. (this needs a more formal test, but this
> is my preliminary observation)

  Hm, that's a bit strange. I ensured that BT chip is powered off when
it is not used, and during suspend. Maybe something else related is
activated still (like some clock). Indeed, needs formal testing.

> 2) the powerbutton works to turn the device on, but not to turn it off.
> This is true both under opie and gpe so I suspect that some keycode is
> not right. (is there some command I can use to detect the message sent
> by that button?)

  What distro do you use? This works for both Angstrom and my unofficial
build of Familiar 0.8.4.

> Please let me know if there are any specific long-term (or otherwise)
> tests that you would like me to perform.

  Let's have h4100 vs h4300 issue hammered out, so we didn't have
community split here.

> Thanks,
> Daniel Steen

[]

-- 
Best regards,
 Paul                            mailto:pmiscml_at_gmail.com