suspend/resume [was: Devsummit pictures]

[Warner Losh]

On Aug 31, 2012, at 8:14 AM, Alexey Dokuchaev wrote:

> On Fri, Aug 31, 2012 at 07:58:36AM -0600, Warner Losh wrote:
>> On Aug 31, 2012, at 4:40 AM, Alexey Dokuchaev wrote:
>>> On Fri, Aug 31, 2012 at 12:11:42PM +0200, Hans Petter Selasky wrote:
>>>>> and then new USB stack had broken it again for me (NEC Versa S950, YMMV).
>>>> 
>>>> Should be fixed in all -stable with regard to USB and suspend and resume.
>>> 
>>> To be fair to Hans Petter, maybe his USB stack itself does not have a bug,
>>> but instead uncovers a bug somewhere else.  OTOH, my laptop does not care.
>> 
>> You can determine if it is the usb stack by removing it and trying again.
> 
> Actually I did.  Resume was still broken.  But backing out r229370 (only
> USB-related changes) fixes the hang.  I am puzzled...  And while keyboard
> does not work in general, tty switching does.

That's very odd results.  Perhaps is suggests needing to run many tests?

keyboard not working means the process isn't reading the data.  Since tty switching is working, does breaking in ddb with control escape work (or is that alt-escape)?

>> If possible, a serial console will tell you for sure...
> 
> Serial console is practically useless if you have to work via USB-RS232
> adapter.  "call doadump" also hangs.  So it looks like I have to find
> firewire machine, revive attilio@'s netdump [1] code, or implement phk@'s
> soundcard tricks

Yes, you need a real serial port, not a usb one to do this.  However, if the USB-RS232 is on the other end, it works well (I was using just such a setup with gdb the other day).

Hmmm, call doadump fails, but that means you can get into ddb?

Firewire works well, and is very cheap.  I don't know if it works with a plug-in card though.

> [1] http://lists.freebsd.org/pipermail/freebsd-current/2010-October/020405.html

Now that's cool...

Warner