https://bugzilla.novell.com/show_bug.cgi?id=305162#c1 Holger Macht changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |olh@novell.com, seife@novell.com AssignedTo|behlert@novell.com |pavel@novell.com --- Comment #1 from Holger Macht 2007-08-28 02:40:50 MST ---

From the log:

--- Di 28. Aug 09:31:02 CEST 2007: done running suspend hooks. + /usr/sbin/s2ram Allocated buffer at 0x2010 (base is 0x0) ES: 0x0201 EBX: 0x0000 Function not supported? Switching from vt7 to vt1 Calling save_state fbcon fb0 state 1 Calling restore_state_from fbcon fb0 state 0 switching back to vt7 + RET=0 + set +x Di 28. Aug 09:33:38 CEST 2007: running resume hooks. --- Pavel, Stefan, Olaf, any idea? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=305162#c3 --- Comment #3 from Felix Möller 2007-08-29 14:23:37 MST --- Did todays factory upgrade: # rpm -q kernel-default suspend kernel-default-2.6.22.5-2 suspend-0.50.20070731-12 but still: Wed Aug 29 22:13:04 CEST 2007: done running suspend hooks. + /usr/sbin/s2ram Allocated buffer at 0x2010 (base is 0x0) ES: 0x0201 EBX: 0x0000 Function not supported? Switching from vt7 to vt1 Calling save_state fbcon fb0 state 1 Calling restore_state_from fbcon fb0 state 0 switching back to vt7 + RET=0 + set +x Wed Aug 29 22:15:26 CEST 2007: running resume hooks. Any idea what to do? As far as I can research my system did suspend fine with factory of 2007-08-10 (~ Beta1). :-( The time difference does not seem to be constant now its 2:22 in my original report it was 2:36 ... -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=305162#c5 --- Comment #5 from Felix Möller 2007-09-04 09:31:09 MST --- (In reply to comment #4 from Pavel Machek)

Does system still work properly after resume? Does "time sleep 1" produce something reasonable? # time sleep 1

real 0m1.003s user 0m0.000s sys 0m0.004s The system works perfectly after resume. Even the WLAN is reconnecting. ;) But booting takes 1 minute and resuming 3 so it is unusable. :-(

Where is the time spent? While suspending? The system suspends within 2 seconds or so. The time is spent resuming.

Stefan, how widespread is this problem? Can we reproduce it on machine where VGA works during resume, or where we have serial/netconsole?

-- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=305162#c7 --- Comment #7 from Pavel Machek 2007-09-04 15:31:17 MST --- Unfortunately, /var/log/messages shows timestamps from syslog (not from kernel) that are useless. If you increase console loglevel (sysrq-9), can you see the messages during resume? Or perhaps enable timestamps for printk (it is in kernel config somewhere, IIRC). -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=305162 Pavel Machek changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |ASSIGNED -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=305162#c10 --- Comment #10 from Pavel Machek 2007-09-05 02:34:52 MST --- Ok, so it is fast with minimal modules, and slow with the full set? Can you selectively load/unload modules to find out which one causes the slowdown? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=305162#c12 Felix Möller changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |trenn@novell.com --- Comment #12 from Felix Möller 2007-09-08 07:24:41 MST --- After several hours of rebooting I found the culprit: acpi_cpufreq I have to add an "install acpi_cpufreq /bin/true" to /etc/modprobe.d/blacklist to solve the problem. Reading through the changelog there are two commits which look related, therefore adding Thomas. * Mo Aug 13 2007 - trenn@suse.de - patches.fixes/cpufreq_check_negative_load.patch: ondemand: add a check to avoid negative load calculation (https://bugzilla.novell.com/show_bug.cgi?id=297355). - patches.fixes/cpufreq_fix_tickless_acounting.patch: ondemand: fix tickless accounting and software coordination bug (https://bugzilla.novell.com/show_bug.cgi?id=297355). * Mo Aug 06 2007 - trenn@suse.de - patches.arch/acpi_autoloading_acpica.patch: - patches.arch/acpi_autoloading_interface.patch: - patches.arch/acpi_autoloading_modaliases.patch: - patches.fixes/cpufreq_move_policy_init.patch: ondemand as default governor. - patches.fixes/cpufreq_ondemand_as_default.patch: - Update config files. These are ACPI autoloading (via udev) and use ondemand as default governor when loading low-level cpufreq driver patches. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=305162#c14 Felix Möller changed: What |Removed |Added ---------------------------------------------------------------------------- Platform|Other |i586 --- Comment #14 from Felix Möller 2007-09-09 10:01:47 MST --- You were right with the config option. I am compiling the kernel right now. The system is i586 with a Core Duo. It is an MacBook of early 2006. Will report back later today. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=305162#c16 --- Comment #16 from Thomas Renninger 2007-09-09 11:20:22 MST ---

Thomas what else do you need? Time :) Just kidding...

Maybe Venkatesh can have a look at this, it looks very Intel specific. Please ping me at end of the week or beginning of next week if Venkatesh also has no time or nobody else had a look at this. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=305162#c18 --- Comment #18 from Felix Möller 2007-09-09 12:59:38 MST --- nmi_watchdog=0 watchdog does not help. But just adding SUSPEND_MODULES="acpi_cpufreq" to /etc/pm/config.d/config helps. This makes frequency-scaling working for me again and I do not have to blacklist acpi_cpufreq any longer. Sadly I do not have the old kernels anymore. But as far as I can research my system did suspend fine with factory of 2007-08-10 (~ Beta1). :-( -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=305162#c20 --- Comment #20 from Venkatesh Pallipadi 2007-09-11 12:16:12 MST --- If I understand correctly, there are two recent changes that can be potentially causing this. * Mo Aug 13 2007 - trenn@suse.de - patches.fixes/cpufreq_check_negative_load.patch: ondemand: add a check to avoid negative load calculation (https://bugzilla.novell.com/show_bug.cgi?id=297355). - patches.fixes/cpufreq_fix_tickless_acounting.patch: ondemand: fix tickless accounting and software coordination bug (https://bugzilla.novell.com/show_bug.cgi?id=297355). * Mo Aug 06 2007 - trenn@suse.de - patches.arch/acpi_autoloading_acpica.patch: : : Can you rollback the two patches that went on Aug 13 and see whether the problem is still there? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=305162#c22 --- Comment #22 from Thomas Renninger 2007-09-11 16:04:30 MST --- Created an attachment (id=163374) --> (https://bugzilla.novell.com/attachment.cgi?id=163374) cpufreq_fix_tickless_acounting.patch -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=305162#c24 Thomas Renninger changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |NEEDINFO Info Provider| |gregkh@novell.com --- Comment #24 from Thomas Renninger 2007-09-13 13:14:51 MST --- Greg, do you know whether something changed here? Andi said, we could blacklist the device, but (possibly?) only if there is a driver for it? Here the relevant information collected: Since arround beta1 my system (Intel MacBook) takes 3 minutes to resume. During the last alphas the system took 3 seconds to resume. So something is broken here. :-( For me it looks interesting around line 800: [ 130.548000] PM: Writing back config space on device 0000:00:07.0 at offset 4 (was 0, writing 90444000) [ 130.548000] PM: Writing back config space on device 0000:00:07.0 at offset 1 (was b00000, writing b00002) [ 270.900000] PM: Writing back config space on device 0000:00:1b.0 at offset f (was 100, writing 10b) [ 270.900000] PM: Writing back config space on device 0000:00:1b.0 at offset 4 (was 4, writing 90440004) # lspci -s 00:07.0 -nn 00:07.0 Performance counters [Class 1101]: Intel Corporation Unknown device [8086:27a3] (rev 03) -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=305162#c26 --- Comment #26 from Rafael Wysocki 2007-09-21 07:57:46 MST --- There are some patches from Thomas Gleixner that fix similar issues in the current Linus' tree. AFAICS, they are these two patches plus-minus some minor stuff: http://www.sisk.pl/kernel/hibernation_and_suspend/2.6.23-rc6/patches/32-time... http://www.sisk.pl/kernel/hibernation_and_suspend/2.6.23-rc6/patches/38-cpu-... -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=305162#c28 --- Comment #28 from Stephan Kulow 2007-09-21 08:52:09 MST --- I tend to disagree actually. I would rather see some more testing than risking breaking the final with somethign we could have avoided -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=305162#c30 Stephan Kulow changed: What |Removed |Added ---------------------------------------------------------------------------- Flag| |SHIP_STOPPER- --- Comment #30 from Stephan Kulow 2007-09-21 09:08:13 MST --- yes. But 10.3 won't actively waiting for this fix -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=305162#c32 --- Comment #32 from Felix Möller 2007-09-24 01:16:05 MST --- Thomas I think you might want to have a look at http://kerneltrap.org/mailarchive/linux-kernel/2007/9/22/272671 . -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=305162#c34 --- Comment #34 from Felix Möller 2007-09-24 13:26:28 MST --- Created an attachment (id=174436) --> (https://bugzilla.novell.com/attachment.cgi?id=174436) patch from current kernel-git I applied the attached patch, which is a copy from the kernel-git-repository to my kernel-source-2.6.22.5-29 and now the system resume and suspends fine as far as i can see. Maybe this is something for the 2.6.22.x releases? -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=305162#c36 --- Comment #36 from Rafael Wysocki 2007-09-25 03:53:55 MST --- (In reply to comment #34 from Felix Möller)

Created an attachment (id=174436) --> (https://bugzilla.novell.com/attachment.cgi?id=174436) [details] patch from current kernel-git

I applied the attached patch, which is a copy from the kernel-git-repository to my kernel-source-2.6.22.5-29 and now the system resume and suspends fine as far as i can see.

Maybe this is something for the 2.6.22.x releases?

I think this patch is -stable material. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=305162#c38 Thomas Renninger changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |RESOLVED Resolution| |FIXED --- Comment #38 from Thomas Renninger 2007-09-25 07:43:21 MST --- You mean with: echo 1 >/sys/module/processor/max_cstate at suspend and echo 3 >/sys/module/processor/max_cstate at resume time? Better not, the patch exactly does the same where it should be done. Patch from comment #34 added -> closing. Rafael, could you add me to CC if you ask for stable inclusion for this one or tell me whether it will hit stable kernel or not (if you are CC'ed by requester...). Thanks Felix and Rafael for finally resolving this one. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=305162#c40 --- Comment #40 from Thomas Renninger 2007-09-25 08:38:19 MST --- Rafael, FYI: I know these Vaios have a broken lapic timer in deeper C-states. I expect the IPI timer broadcast setup which is a workaround for those in 10.2 is messed up in timer suspend functions by Thomas Gleissner's nohz/clocksource/whatever patches that came in some time ago. If I understood that right, they are still thinking about a right fix, which might be a modification where IPI broadcast timers break, I will post to lkml as soon as I find the time. The workaround for those Vaios to use IPI broadcast timer interrupts is located here(formerly AMDs where blacklisted to use IPI, I added Pentium Ms to 10.2 some time ago, but this fell out due to CLOCKSOURCE patches and if I read the code right all machines are now using IPI interrupts): 2.6.22 (drivers/acpi/processor_idle.c): #ifdef CONFIG_GENERIC_CLOCKEVENTS unsigned long reason; reason = pr->power.timer_broadcast_on_state < INT_MAX ? CLOCK_EVT_NOTIFY_BROADCAST_ON : CLOCK_EVT_NOTIFY_BROADCAST_OFF; clockevents_notify(reason, &pr->id); #else This is just a guess, but I could imagine the clocksoure (or whatever) timer suspend functions are messing up the IPI broadcast timer setup somehow. Therefore the Sony Vaios take that long at suspend time (hitting keys should help to be quicker if this is the problem). With 10.2, without blacklisting those machines even booted very slow (nolapic or pressing keys (producing interrupts) helped). I haven't followed up the lkml thread, don't know whether this info is still needed. I will search for the thread if I find the time, but time for me is precious at the moment, maybe this helps a bit fixing up the root cause... Of course this is just a guess, it may be something else :) -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=305162#c42 --- Comment #42 from Rafael Wysocki 2007-09-25 14:24:19 MST --- Referring to Comment #40: This certainly is valuable information, as it may help understand some strange behavior that we observe on these boxes from time to time (especially on the Andrew's one). OTOH, I'm not sure if it's still relevant as far as suspend is concerned. Thanks for it anyway. :-) -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=305162 Thomas Renninger changed: What |Removed |Added ---------------------------------------------------------------------------- Status|RESOLVED |VERIFIED -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.