2.6.19-rc4-mm1: noidlehz problems
Hi!
First, it would be nice if we had someone listed as a maintainer of noidlehz stuff...
Then... I'm getting strange messages from noidlehz each time I unplug/replug AC power (perhaps due to interrupt latency?).
Disabling NO_HZ and high resolution timers due to timer broadcasting (C3 stops local apic)
Adding 987988k swap on /dev/sda1. Priority:-1 extents:1 across:987988k
Disabling NO_HZ and high resolution timers due to timer broadcasting (C3 stops local apic)
EXT2-fs warning (device sda2): ext2_fill_super: mounting ext3 filesystem as ext2
pcmcia: Detected deprecated PCMCIA ioctl usage from process: cardmgr.
pcmcia: This interface will soon be removed from the kernel; please expect breakage unless you upgrade to new tools.
pcmcia: see
http://www.kernel.org/pub/linux/utils/kernel/pcmcia/pcmcia.html for details.
cs: IO port probe 0x310-0x380: clean.
cs: IO port probe 0xa00-0xaff: clean.
Disabling NO_HZ and high resolution timers due to timer broadcasting (C3 stops local apic)
Disabling NO_HZ and high resolution timers due to timer broadcasting (C3 stops local apic)
e1000: eth0: e1000_watchdog: NIC Link is Up 100 Mbps Full Duplex
e1000: eth0: e1000_watchdog: 10/100 speed: disabling TSOcoda_read_super: Bad mount datacoda_read_super: device index: 0coda_read_super: rootfid is (01234567.ffffffff.080519b0.00000000)coda_upcall: Venus dead on (op,un) (7.2) flags 10
Failure of coda_cnode_make for root: error -19acpiphp_glue: cannot get bridge info
Disabling NO_HZ and high resolution timers due to timer broadcasting (C3 stops local apic)
Disabling NO_HZ and high resolution timers due to timer broadcasting (C3 stops local apic)
Disabling NO_HZ and high resolution timers due to timer broadcasting (C3 stops local apic)
Disabling NO_HZ and high resolution timers due to timer broadcasting (C3 stops local apic)
...I'd expect one such message, not many of them. Something seems seriously wrong there...
Plus, suspend to RAM and disk is broken in -rc4-mm1. Suspend to RAM dies with screaming speaker, suspend to disk returns but machine is mostly toast (and screaming, looks like timer problem, beeps never end). I'll disable NO_HZ and try again.
Pavel
Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek (cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
Hi!
Quoted 14 lines
First, it would be nice if we had someone listed as a maintainer of noidlehz stuff...
Then... I'm getting strange messages from noidlehz each time I unplug/replug AC power (perhaps due to interrupt latency?).
Disabling
NO_HZand high resolution timers due to timer broadcasting (C3 stops local apic)
Adding 987988k swap on/dev/sda1. Priority:-1 extents:1 across:987988k
DisablingNO_HZand high resolution timers due to timer broadcasting (C3 stops local apic)
EXT2-fs warning (device sda2):ext2_fill_super: mounting ext3 filesystem as ext2
...
Quoted 5 lines
Disabling
NO_HZand high resolution timers due to timer broadcasting (C3 stops local apic)
DisablingNO_HZand high resolution timers due to timer broadcasting (C3 stops local apic)
Quoted 7 lines
...I'd expect one such message, not many of them. Something seems seriously wrong there...
Plus, suspend to RAM and disk is broken in -rc4-mm1. Suspend to RAM dies with screaming speaker, suspend to disk returns but machine is mostly toast (and screaming, looks like timer problem, beeps never end). I'll disable
NO_HZand try again.
Disabling NO_HZ did not help, but I disabled high resolution timers, and s2ram now works. s2disk also started working.
Pavel
Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek (cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
Quoted 7 lines · Pavel Machek
Hi!
First, it would be nice if we had someone listed as a maintainer of noidlehz stuff...
Then... I'm getting strange messages from noidlehz each time I unplug/replug AC power (perhaps due to interrupt latency?).
there probably is a different story going on.
When you unplug/replug AC power, several bioses change the meaning of the software C-states in your system.
(there is a mapping between software visible C states and the hardware C-states)
In some (hardware) C-states, the local apic timer stops (as does the TSC), while in others it keeps running. If you change from AC to battery, the bios can change the meaning of a software C-state from one where local apic timer keeps going to one where it stops. This obviously upsets the hrtimers/tickless code since that uses local apic timer for event generation....
Signature · 2 lines
if you want to mail me at work (you don't), use arjan (at) linux.intel.com Test the interaction between Linux and your BIOS via http://www.linuxfirmwarekit.org
Quoted 6 lines · Arjan van de Ven
Quoted 5 lines · Pavel Machek
Hi!
First, it would be nice if we had someone listed as a maintainer of noidlehz stuff...
Quoted 8 lines
Quoted 2 lines
Then... I'm getting strange messages from noidlehz each time I unplug/replug AC power (perhaps due to interrupt latency?).
there probably is a different story going on.
When you unplug/replug AC power, several bioses change the meaning of the software C-states in your system.
(there is a mapping between software visible C states and the hardware C-states)
Yep, my notebook is one of such machines, so your theory looks very plausible.
Pavel
Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek (cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
Quoted 6 lines · Arjan van de Ven
In some (hardware) C-states, the local apic timer stops (as does the TSC), while in others it keeps running. If you change from AC to battery, the bios can change the meaning of a software C-state from one where local apic timer keeps going to one where it stops. This obviously upsets the hrtimers/tickless code since that uses local apic timer for event generation....
Is there any hope of working around this? I'd have expected that the most useful case for the tickless code was also the case where we want to be using C3/C4...
Signature · 1 line
Matthew Garrett | [email protected]
Quoted 9 lines · Matthew Garrett
Quoted 6 lines · Arjan van de Ven
In some (hardware) C-states, the local apic timer stops (as does the TSC), while in others it keeps running. If you change from AC to battery, the bios can change the meaning of a software C-state from one where local apic timer keeps going to one where it stops. This obviously upsets the hrtimers/tickless code since that uses local apic timer for event generation....
Quoted 3 lines
Is there any hope of working around this? I'd have expected that the most useful case for the tickless code was also the case where we want to be using C3/C4...
Can you try the following patches please ?
http://tglx.de/projects/hrtimers/2.6.19-rc4-mm1/patch-2.6.19-rc4-mm1-hrt-dyntick1.patch
This addresses the C34/C4 issue
tglx