Latest working kernel version:2.6.28 Earliest failing kernel version:2.6.29-rc1 Distribution: Debian Hardware Environment: Dell Latitude D410 Software Environment: Problem Description: Suspend-to-memory appears to work, entering sleep state with slowly blinking power LED, but hangs with blank screen on resume. Suspend-to-disk hangs before power-off. Behaviour is similar in -rc1 and -rc3. Steps to reproduce: Boot kernel (no X); attempt suspend-resume.
Created attachment 20048 [details] Patch to test Please check if this patch makes any difference.
Created attachment 20050 [details] dmesg output after suspend-to-memory and resume Better! With the patch, suspend-to-memory and suspend-to-disk both work, but resume seems slow with a number of (fleeting) error messages of the form: ohci_hcd 0000:03:00.0: can't restart, -1 I have attached the output from dmesg after suspend-to-memory and resume in the hope that it might give some clues. Thanks -- Ross
what if you unload ohci driver before suspend and load it after resume?
ping Ross. :)
Sorry for delay replying -- I'm travelling at present (London today, Seoul tomorrow) and to test the "ohci unload" strategy I had to recompile the kernel with ohci as a module. The answer is that unloading ohci makes suspend-resume snappy. If I then reload ohci, the only hardware I have with me that requires it (a Merlin XU870 HSPD modem) is unrecognised and doesn't work. Tentative conclusion: the remaining problem seems to lie in ohci_hcd
re-assign to the usb category, :)
The next step: Build 2.6.29-rc4 with CONFIG_USB_DEBUG enabled and attach the dmesg log showing what happens during the suspend/resume and when ohci-hcd is reloaded and the modem is plugged in.
The fix from comment #1 has been merged, so there's only the remaining ohci-hcd issue. Ross, can you check the patch from http://marc.info/?l=linux-kernel&m=123456499020575&w=4 and see if it changes anything?
I have not followed Alan's proposed "next step" [Comment #7] because applying the patch from marc.info suggested by Rafael [Comment #8] has made the problem go away. With the fixes from comment #1 and comment #8 both applied, suspend-resume works well whether ohci_hcd is loaded or not. Thanks everyone for your good work. From my viewpoint this bug is now completely squashed.
*** This bug has been marked as a duplicate of bug 12659 ***
Your last comment means that this is a duplicate of bug #12659. Closing.