[workaround] Re: resume slow on Thinkpad T42 FreeBSD 8-STABLE

[Ian Smith]

Thanks to a recent post by Bengt Ahlgren in mobile@, the 60 second stall 
or freeze on resume from S3 suspend - an issue that's persisted since 
8.0-R through present 8-STABLE and that has only been reported to affect 
various models of Thinkpad (T23, T30, T42 and X40 to date) - has been 
confirmed as being due to uhci.  Bengt discovered that a successful 
workaround is to simply not have uhci loaded during suspend/resume.

On my T23, currently at 8.2-RELEASE i386, building a kernel from only

 include GENERIC
 ident NO_UHCI
 nodevice uhci
 nodevice ohci	# not used
 nodevice ehci	# not used on T23 (USB 1 only)

and adding

 uhci_load="YES" to /boot/loader.conf,
 kldunload uhci  to /etc/rc.suspend and
 kldload uhci    to /etc/rc.resume

entirely eliminated my 60 second stall, and I once again feel confident 
to build a working, travelling 8-STABLE system.  Thanks again Bengt, and 
I hope that this will work also for you Ted, and for Vitaly and others.

So far I've seen no ill effects from having USB devices attached when 
suspending, except that filesystems mounted via USB are inaccessible on 
resume.  Even then there are no panics as used to occur pre-8.x in such 
circumstances, umount/mount working to remount the fs .. though I've so 
far only tested that with a PC-BSD memstick image mounted r/o.

I know nothing about the USB code, but would be happy to test patches; 
maybe this problem with uhci can be identified and fixed before 9.0?

cheers, Ian


On Wed, 22 Sep 2010, Ian Smith wrote:
 > On Thu, 25 Feb 2010 17:35:51 -0800, Ted Faber wrote:
 >  > On Thu, Feb 25, 2010 at 05:18:22PM +1100, Ian Smith wrote:
 >  > > On Wed, 24 Feb 2010, Ted Faber wrote:
 >  > > 
 >  > >  > I'm running a recent (yesterday) FreeBSD 8-STABLE on my T42.  When I
 >  > >  > resume from suspend the fan starts screen returns immediately (to text
 >  > >  > mode) and then the system idles for about a minute until it shakes
 >  > >  > itself awake.  The keyboard LEDs cycle, the disk runs and X returns.
 >  > > 
 >  > > Good to hear, Ted .. I thought it was just me :)  Your symptoms appear 
 >  > > entirely identical to mine as posted on Dec 13th to mobile@ and acpi@, 
 >  > > see thread 'Thinkpad T23 60 second stall on resuming 8.0-RELEASE/i386' 
 >  > > http://lists.freebsd.org/pipermail/freebsd-acpi/2009-December/006192.html
 >  > > 
 >  > > Nate Lawson thought it an ATA rather than an ACPI problem and I assume 
 >  > > he's likely right, but I've had too much $work (and too little solar 
 >  > > power during an exceptionally wet summer) to follow it up.
 >  > 
 >  > Yeah, that sounds like the problem.  I'll poke at ata if I get the
 >  > chance, but that seems unlikely for a while.
 >  > 
 >  > > In case relevant, my ad0 is a 120GB Fujitsu MHV2120AH, but your TOSHIBA 
 >  > > MK4026GAX looks more likely to be the original disk?
 >  > 
 >  > Likely, but I can't be sure.  I bought it used.
 >  > 
 >  > > 
 >  > > I've no time to spend on hunting this now, and know nothing about ATA 
 >  > > anyway, assuming that's where the problem lies, but I'll be quick to 
 >  > > test any suggested solution/s!
 >  > 
 >  > Like I say, assuming I ever get free time, I'll have a look.
 > 
 > Brought my Thinkpad T23 (actually a replacement, similar model, same HD) 
 > to '8.1-STABLE #0: Sun Sep 5 15:53:00 EST 2010' but this issue persists, 
 > as it does if booted from 8.1-i386-memstick, HD present but not mounted; 
 > this one has the CDRW rather than the original's DVD-ROM, if relevant.
 > 
 > Before excruciating detail of the issue in mobile@ and acpi@ again in 
 > hope of debugging hints, have you (Ted or anyone) any further news on 
 > Thinkpad T-series (or other) 60 second stalls/freezes on S3 resume?
 > 
 > [Please read the abovementioned thread before asking any questions]
 > 
 > cheers, Ian