"ksoftirqd/0" always use 3-8% CPU. about linux-sunxi HOT 11 CLOSED

isn't this kind of too long?

from linux-sunxi.

I'm not tested <15 sec.
But 15sec is no problem to use hdmi. After connect cable we should wait <15sec.

from linux-sunxi.

isnt the original delay set at 200ms?

hdmi_delay_ms(200); @ drv_hdmi.c

from linux-sunxi.

Yes, you right. Then better to increase hdmi_delay_ms. And make suspend for it.

from linux-sunxi.

i have tesed 1sec with good response. but I believe 2 sec is reasonable for cutting down on the polling and getting a good detection.

from linux-sunxi.

thanks @Quarx2k for the fix. This is a long issue of Allwinner. I looked at this today. Another way to fix is to change in drivers/video/sun4i/hdmi/aw/hdmi_core.c function __s32 main_Hpd_Check(void); change the hdmi_delay_ms(1) to hdmi_delay_ms(10). i can not explain it with my current understanding of the kernel schedule. can anyone explain this?

from linux-sunxi.

I think that we could simply replace hdmi_delay_ms() with msleep(). Please correct me if I am wrong.

from linux-sunxi.

On Tue, Sep 11, 2012 at 6:19 PM, Aliaksei Katovich <[email protected]

wrote:

I think that we could simply replace hdmi_delay_ms() with msleep(). Please
correct me if I am wrong.

yes, that also works. can anyone explain this?

—
Reply to this email directly or view it on GitHubhttps://github.com/amery/linux-allwinner/issues/43#issuecomment-8455122.

Keep simple, stay foolish.

from linux-sunxi.

does msleep(1) produce the same effect as hdmi_delay_ms(1)?

from linux-sunxi.

On Tue, Sep 11, 2012 at 6:46 PM, Aliaksei Katovich <[email protected]

wrote:

does msleep(1) produce the same effect as hdmi_delay_ms(1)?

no. hdmi_delay_ms uses set_current_state(TASK_INTERRUPTIBLE), but msleep
uses schedule_timeout_uninterruptible(timeout). Interesting, i need to
google the difference between TASK_INTERRUPTIBLE and TASK_UNINTERRUPTIBLE.
but from hardware point of view, they are the same, wait for some time, and
check again.

—
Reply to this email directly or view it on GitHubhttps://github.com/amery/linux-allwinner/issues/43#issuecomment-8455695.

Keep simple, stay foolish.

from linux-sunxi.

On Tue, Sep 11, 2012 at 6:46 PM, Aliaksei Katovich
<[email protected]

wrote:
does msleep(1) produce the same effect as hdmi_delay_ms(1)?

no. hdmi_delay_ms uses set_current_state(TASK_INTERRUPTIBLE), but
msleep
uses schedule_timeout_uninterruptible(timeout). Interesting, i need to
google the difference between TASK_INTERRUPTIBLE and
TASK_UNINTERRUPTIBLE.
but from hardware point of view, they are the same, wait for some time,
and
check again.

I would anyway recommend to use plain msleep() here, regardless of your
findings. E.g. jiffies calculations like this (see. hdmi_delay_ms)

    __u32 timeout = t*HZ/1000

look suspicious to me.

â
Reply to this email directly or view it on
GitHub<https://github.com/amery/linux-allwinner/issues/43#issuecomment-
8455695>.

--
Keep simple, stay foolish.

—
Reply to this email directly or [1]view it on GitHub.
[J6T91GIPIyhU-8ti4GCGP71oDyae5wcKbWrmQLuV4jMv0PuTIo6eLnz7eoGYz3O_.gif]

References

1. https://github.com/amery/linux-allwinner/issues/43#issuecomment-8456078

from linux-sunxi.