On 08/15/2014 11:07 PM, Tony Lindgren wrote:

Nice, now it mostly works for me with off-idle too :) That is as long
as I have the DMA channels commented out in the .dts file.

And I'm still seeing an occasional hang with pstore console just
showing:

[  289.076538] In-band Error seen by MPU  at address 0
[  289.076538] ------------[ cut here ]------------
[  289.076568] WARNING: CPU: 0 PID: 99 at drivers/bus/omap_l3_smx.c:162 omap3_l3_app_irq+0xdc/0x134()
[  289.076599] Modules linked in:
[  289.076599] CPU: 0 PID: 99 Comm: test-idle-off-8 Tainted: G        W      3.16.0+ #510
[  289.076629] [<c0016c44>] (unwind_backtrace) from [<c00129c8>] (show_stack+0x20/0x24)
[  289.076660] [<c00129c8>] (show_stack) from [<c0714cd4>] (dump_stack+0x88/0xa4)

Okay. So this backtrace does not show more like from where the access
is happening?

Which most likely means there's still some glitch with the
runtime PM somewhere and registers are being accessed when
not clocked. I _think_ I did not see it when I did not have
console=ttyS2,115200 in my cmdline but was using just pstore
console.

quoted

The device name is ttyS based instead of ttyO. If a ttyO based node name
is required please ask udev for it. If both driver are activated (this
and omap-serial) then this serial driver will take control over the
device due to the link order

That's still not going to help with the existing kernel cmdlines
and existing installs.. I wonder if we can just do a minimal
dummy serial-omap.c that just proxies all the ttyO read/write
access to ttyS?

Hmm. So you are not a friend of the udev solution?. For now the driver
is "default n" and you have to explicit enable it and _then_ you should
be able to update your command line, etc.
If I introduce a kernel proxy for compatibility I assume that people
will wake up once that compatibility piece is gone.

However, if you insist? I tried to make a symlink but nobody does this
in kernel. The "rtc -> rtc0" and friends seem to come from udev _or_
distro. So I sff could a second device node with the same major/minor.
That would work for userland but not for the kernel console? So we need
a proxy-console for this.

Before I spent time on this proxy-console I would like to hear from Greg
that he is okay with this.

Regards,

Tony

Sebastian