You are not logged in.

#326 2013-01-27 01:19:46

KaiSforza
Member
Registered: 2012-04-22
Posts: 133
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Hspasta wrote:

Serious thanks for such quick builds for the rc kernels.

You're welcome (and sorry for being 5 or 6 hours late).

Let's hope this fixes my GPU hang problem I've had issued with the last few 3.8rc's. At least the power regression seems to be fine for me.

Edit: Power regression still fixed, but GPU hang still there sad

What's in /sys/kernel/debug/dri/0/i915_error_state ?


Thinkpad T420 | Intel 3000 | systemd {,--user}
PKGBUILDs I use | pywer AUR helper

Offline

#327 2013-01-27 01:34:04

Hspasta
Member
Registered: 2011-12-24
Posts: 187
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

error:

Time: 1359247451 s 583326 us
PCI ID: 0x0126
EIR: 0x00000000
IER: 0x8c649081
PGTBL_ER: 0x00000000
FORCEWAKE: 0x00000001
DERRMR: 0x0000fffe
CCID: 0x0007010d
  fence[0] = bda03300611001
  fence[1] = a3d10070a392001
  fence[2] = a6590070a61a001
  fence[3] = a6d90070a69a001
  fence[4] = a7190070a6da001
  fence[5] = a7590070a71a001
  fence[6] = 9fd000709f91001
  fence[7] = a6990070a65a001
  fence[8] = a01000709fd1001
  fence[9] = a0500070a011001
  fence[10] = a0d00070a091001
  fence[11] = a0900070a051001
  fence[12] = a2d10070a292001
  fence[13] = a3110070a2d2001
  fence[14] = a3510070a312001
  fence[15] = a3910070a352001
  INSTDONE_0: 0xffffffff
  INSTDONE_1: 0xbfffffff
  INSTDONE_2: 0x00000000
  INSTDONE_3: 0x00000000
ERROR: 0x00000000
DONE_REG: 0x00000000
render command stream:
  HEAD: 0x0001daf0
  TAIL: 0x0001e148
  CTL: 0x0001f801
  ACTHD: 0x005fc01c
  IPEIR: 0x00000000
  IPEHR: 0x01800001
  INSTDONE: 0xffffffff
  BBADDR: 0x5fc01d80000200
  INSTPS: 0x80000200
  INSTPM: 0x00000080
  FADDR: 0x005fc200
  RC PSMI: 0x00000010
  FAULT_REG: 0x00000000
  SYNC_0: 0x00000000 [last synced 0x00000000]
  SYNC_1: 0x0000088d [last synced 0x0000088f]
  seqno: 0x00000888
  waiting: yes
  ring->head: 0x00000000
  ring->tail: 0x0001e148
bsd command stream:
  HEAD: 0x00000000
  TAIL: 0x00000000
  CTL: 0x0001f001
  ACTHD: 0x00000000
  IPEIR: 0x00000000
  IPEHR: 0x00000000
  INSTDONE: 0xffeffffe
  INSTPS: 0x00000000
  INSTPM: 0x00000000
  FADDR: 0x00023000
  RC PSMI: 0x00000018
  FAULT_REG: 0x00000000
  SYNC_0: 0x0000088d [last synced 0x00000000]
  SYNC_1: 0x00000888 [last synced 0x00000000]
  seqno: 0x00000000
  waiting: no
  ring->head: 0x00000000
  ring->tail: 0x00000000
blt command stream:
  HEAD: 0x0001e010
  TAIL: 0x0001e110
  CTL: 0x0001f401
  ACTHD: 0x0001e01c
  IPEIR: 0x00000000
  IPEHR: 0x0b140001
  INSTDONE: 0xfffffff7
  INSTPS: 0x00000000
  INSTPM: 0x00000000
  FADDR: 0x00062038
  RC PSMI: 0x00000010
  FAULT_REG: 0x00000000
  SYNC_0: 0x00000888 [last synced 0x00000895]
  SYNC_1: 0x00000000 [last synced 0x00000000]
  seqno: 0x0000088d
  waiting: no
  ring->head: 0x00000000
  ring->tail: 0x0001e110
Active [90]:
  08040000     8192 0010 0000 889 0 dirty render snooped (LLC)
  005fc000     4096 0018 0000 889 0 p dirty render snooped (LLC)
  00070000     8192 0050 0000 88a 0 P dirty render snooped (LLC)
  0975c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  0971c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  096dc000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  0969c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  0965c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  0961c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  095dc000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  0959c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  0955c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  0951c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  094dc000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  0949c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  0945c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  0941c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  093dc000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  0939c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  0935c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  0931c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  092dc000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  0929c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  0925c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  0921c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  091dc000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  0919c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  08dd4000    32768 0004 0000 88a 0 X dirty render snooped (LLC)
  07f31000  1048576 0004 0000 88a 0 X dirty render snooped (LLC)
  0915c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  0911c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  090dc000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  0909c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  0905c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  0901c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  08fdc000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  08f9c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  08f5c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  08f1c000   262144 0004 0000 88a 0 X dirty render snooped (LLC)
  097a2000    32768 001e 0000 88a 0 dirty render snooped (LLC)
  08042000     4096 0010 0010 88b 88b dirty render snooped (LLC)
  097aa000     8192 0010 0000 88b 0 dirty render snooped (LLC)
  08edc000   262144 0004 0000 88b 0 X dirty render snooped (LLC)
  08043000  6291456 0002 0002 88b 88b Y dirty render snooped (LLC)
  08643000  6291456 0002 0002 88b 88b Y dirty render snooped (LLC)
  08c53000  1572864 0002 0002 88b 88b dirty render snooped (LLC)
  0979d000     4096 0020 0000 88b 0 dirty render snooped (LLC)
  0979e000     4096 0020 0000 88b 0 dirty render snooped (LLC)
  08e9c000   262144 0004 0000 88b 0 X dirty render snooped (LLC)
  08e5c000   262144 0004 0000 88b 0 X dirty render snooped (LLC)
  08e1c000   262144 0004 0000 88b 0 X dirty render snooped (LLC)
  08ddc000   262144 0004 0000 88b 0 X dirty render snooped (LLC)
  097ac000    57344 0004 0000 88b 0 render snooped (LLC)
  097ba000    20480 0004 0000 88b 0 render snooped (LLC)
  0a3d2000    32768 001e 0000 88b 0 dirty render snooped (LLC)
  08c50000     8192 0010 0000 88e 0 dirty render snooped (LLC)
  005f9000     4096 0018 0000 88e 0 p dirty render snooped (LLC)
  005f8000     4096 0018 0000 88f 0 p dirty render snooped (LLC)
  005f7000     4096 0008 0000 890 0 p dirty blt snooped (LLC)
  03c3c000  1835008 0002 0002 891 891 X dirty render snooped (LLC)
  07f06000    16384 0004 0000 891 0 dirty render snooped (LLC)
  03520000    12288 0004 0000 891 0 render snooped (LLC)
  0060d000    16384 0018 0000 891 0 p dirty render snooped (LLC)
  074c6000 10747904 0004 0000 892 88b X dirty render snooped (LLC) (name: 3)
  005f6000     4096 0018 0000 892 0 p dirty render snooped (LLC)
  005f5000     4096 0008 0000 893 0 p dirty blt snooped (LLC)
  0749b000   176128 0002 0002 894 894 X dirty render snooped (LLC)
  03441000    12288 0004 0000 894 0 dirty render snooped (LLC)
  03432000    12288 0004 0000 894 0 dirty render snooped (LLC)
  0a75a000     8192 0004 0000 894 0 render snooped (LLC)
  0a291000     4096 0004 0000 894 0 dirty render snooped (LLC)
  018bc000     4096 0004 0000 894 0 dirty render snooped (LLC)
  03438000    12288 0004 0000 894 0 dirty render snooped (LLC)
  0342f000    12288 0004 0000 894 0 dirty render snooped (LLC)
  00609000    16384 0018 0000 894 0 p dirty render snooped (LLC)
  005f4000     4096 0008 0000 895 0 p dirty blt snooped (LLC)
  01773000    36864 0010 0010 896 896 dirty render snooped (LLC)
  0a75c000   180224 0002 0002 896 896 X dirty render snooped (LLC)
  06a59000     8192 0004 0000 896 0 dirty render snooped (LLC)
  033a4000   262144 0004 0000 896 0 dirty render snooped (LLC)
  03424000     4096 0004 0000 896 0 dirty render snooped (LLC)
  02fa2000  4194304 0004 0000 896 0 Y dirty render snooped (LLC)
  05de5000    24576 0002 0002 896 896 X dirty render snooped (LLC)
  03bdb000    61440 0004 0000 896 0 render snooped (LLC)
  03b1d000   122880 0004 0000 896 0 dirty render snooped (LLC)
  0352f000    12288 0004 0000 896 0 render snooped (LLC)
  0a788000     8192 0004 0000 896 0 render snooped (LLC)
  00611000  6070272 0002 0002 896 896 P X dirty render uncached (name: 4) (fence: 0)
  05feb000   262144 0020 0000 896 0 render snooped (LLC)
  005f3000     4096 0018 0000 896 0 p dirty render snooped (LLC)
Pinned [31]:
  00000000     4096 0001 0001 0 0 P snooped (LLC)
  00001000   131072 0040 0040 0 0 P dirty snooped (LLC)
  00021000     4096 0001 0001 0 0 P snooped (LLC)
  00022000     4096 0001 0001 0 0 P snooped (LLC)
  00023000   131072 0040 0040 0 0 P dirty snooped (LLC)
  00043000     4096 0001 0001 0 0 P snooped (LLC)
  00044000   131072 0040 0040 0 0 P dirty snooped (LLC)
  00070000     8192 0050 0000 88a 0 P dirty render snooped (LLC)
  00072000  5763072 0000 0000 0 0 P uncached (name: 1)
  005f1000     4096 0018 0000 0 0 p dirty snooped (LLC)
  005f2000     4096 0008 0000 0 0 p dirty snooped (LLC)
  005f3000     4096 0018 0000 896 0 p dirty render snooped (LLC)
  005f4000     4096 0008 0000 895 0 p dirty blt snooped (LLC)
  005f5000     4096 0008 0000 893 0 p dirty blt snooped (LLC)
  005f6000     4096 0018 0000 892 0 p dirty render snooped (LLC)
  005f7000     4096 0008 0000 890 0 p dirty blt snooped (LLC)
  005f8000     4096 0018 0000 88f 0 p dirty render snooped (LLC)
  005f9000     4096 0018 0000 88e 0 p dirty render snooped (LLC)
  005fa000     4096 0008 0000 0 0 p dirty snooped (LLC)
  005fb000     4096 0008 0000 0 0 p dirty snooped (LLC)
  005fc000     4096 0018 0000 889 0 p dirty render snooped (LLC)
  005fd000     4096 0018 0000 0 0 p dirty snooped (LLC)
  005fe000     4096 0018 0000 0 0 p dirty snooped (LLC)
  005ff000     4096 0018 0000 0 0 p dirty snooped (LLC)
  00600000     4096 0018 0000 0 0 p dirty snooped (LLC)
  00601000    16384 0018 0000 0 0 p dirty snooped (LLC)
  00605000    16384 0018 0000 0 0 p dirty snooped (LLC)
  00609000    16384 0018 0000 894 0 p dirty render snooped (LLC)
  0060d000    16384 0018 0000 891 0 p dirty render snooped (LLC)
  00611000  6070272 0002 0002 896 896 P X dirty render uncached (name: 4) (fence: 0)
  00bdb000    16384 0040 0040 0 0 P dirty uncached
render ring --- gtt_offset = 0x005fc000
00000000 :  11000001
....

I don't know how much use this info is.

Offline

#328 2013-01-27 03:55:41

the sad clown
Member
From: 192.168.0.X
Registered: 2011-03-20
Posts: 833

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

KaiSforza wrote:

Ah. looks like a config issue. I'll fix that config and build it soon.

EDIT: Fixed. 3.8.rc5-g949db15-3 is up.

Thanks that fixed everything.  Also, I'm glad to report much better GPU stats.  No more 100% GPU hang ups, and I've tried by suspending and other activity I suspected of causing it.  Active now goes down to 25% and battery reports a discharge rate of 8.73 W (which is good for me as far as I can tell)

Last edited by the sad clown (2013-01-27 03:59:33)


I laugh, yet the joke is on me

Offline

#329 2013-01-27 20:45:14

KaiSforza
Member
Registered: 2012-04-22
Posts: 133
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Hspasta wrote:

error:

<lots of things that I have no idea about>

I don't know how much use this info is.

I'd say start a new thread about this. Maybe others are having this issue. I'm not getting anything like this at all, oddly enough. Are you using hybrid graphics or anything?


Thinkpad T420 | Intel 3000 | systemd {,--user}
PKGBUILDs I use | pywer AUR helper

Offline

#330 2013-01-29 07:56:01

estan
Member
Registered: 2011-03-11
Posts: 13

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Also suffering from this (i7 X220). Has anyone seen an improvement with 3.7.4?

Offline

#331 2013-01-29 08:40:02

hendry
Member
From: Singapore
Registered: 2009-09-08
Posts: 115
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

My X220 is running comparatively cool ~46C on 3.7.4
temp.png

Offline

#332 2013-01-29 09:25:51

the sad clown
Member
From: 192.168.0.X
Registered: 2011-03-20
Posts: 833

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

reporting back.  I'm going back to 3.7.4.  while the 3.8 does indeed fix the gpu hangup, it's a huge regression for video playback.  I was using mplayer and video output driver xv actually reported insufficient resources to play a .flv I pulled from youtube.  Switching to vaapi worked, but only by pushing more of the demand onto the gpu.  My cpu still reported something like 50% useage.  Under 3.7.4, xv functions fine (8% cpu and 50 ops/s gpu) and vaapi almost completely eliminated cpu usage (0-1% and 120 ops/s gpu).  Power demands also went way up between 3.7 and 3.8.

In short, 3.8 "fixes" the gpu hang up problem, but only at the cost of some serious video related inefficiencies.


I laugh, yet the joke is on me

Offline

#333 2013-01-29 18:23:08

the sad clown
Member
From: 192.168.0.X
Registered: 2011-03-20
Posts: 833

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

I still don't know what is triggering the bug in 3.8, but it is a hung gpu.  Here is the pertinent lines from journalctl:

Jan 29 10:58:04 isaac-laptop kernel: [drm:i915_hangcheck_hung] *ERROR* Hangcheck timer elapsed... GPU hung
Jan 29 10:58:04 isaac-laptop kernel: [drm] capturing error event; look for more information in /debug/dri/0/i915_error_state
Jan 29 10:58:05 isaac-laptop kernel: [drm:kick_ring] *ERROR* Kicking stuck wait on render ring

I looked over the i915_error_state, but it wasn't helpful for someone of my level.  Here it is if someone else thinks they could make use of it: https://docs.google.com/file/d/0B33qmZD … R4bGM/edit

According to powertop, prior to hanging (and when booting 3.7), vaapi will utilize ~100% of the gpu with ~120 ops/s.  After hanging, mplayer -vo vaapi still uses ~100% of gpu, but only manages ~70 ops/s.  The rest of computational requirements seem to be pushed onto the cpu.  Does this mean the gpu has only partially hung?  Not even sure what that means.

Here are the benchmarks I pulled from mplayer on a 4.9M 34 sec .mp4 video:
mplayer -nosound -benchmark -vo ***

booting 3.8

VAAPI
BENCHMARKs: VC:   0.310s VO:  19.495s A:   0.000s Sys:   0.067s =   19.872s
BENCHMARK%: VC:  1.5580% VO: 98.1027% A:  0.0000% Sys:  0.3393% = 100.0000%

x11
BENCHMARKs: VC:   2.722s VO:   2.025s A:   0.000s Sys:   0.110s =    4.857s
BENCHMARK%: VC: 56.0402% VO: 41.6910% A:  0.0000% Sys:  2.2688% = 100.0000%

gl
BENCHMARKs: VC:   5.733s VO:  21.942s A:   0.000s Sys:   0.068s =   27.744s
BENCHMARK%: VC: 20.6647% VO: 79.0885% A:  0.0000% Sys:  0.2468% = 100.0000%

gl2
BENCHMARKs: VC:   5.810s VO:  23.416s A:   0.000s Sys:   0.072s =   29.298s
BENCHMARK%: VC: 19.8295% VO: 79.9235% A:  0.0000% Sys:  0.2471% = 100.0000%

gl_nosw
BENCHMARKs: VC:   5.679s VO:  21.754s A:   0.000s Sys:   0.068s =   27.502s
BENCHMARK%: VC: 20.6508% VO: 79.1007% A:  0.0000% Sys:  0.2485% = 100.0000%

booting 3.7.4

xv
BENCHMARKs: VC:   2.264s VO:   0.128s A:   0.000s Sys:   0.052s =    2.444s
BENCHMARK%: VC: 92.6409% VO:  5.2299% A:  0.0000% Sys:  2.1292% = 100.0000%

x11
BENCHMARKs: VC:   3.223s VO:   2.074s A:   0.000s Sys:   0.101s =    5.398s
BENCHMARK%: VC: 59.7080% VO: 38.4205% A:  0.0000% Sys:  1.8715% = 100.0000%

VAAPI
BENCHMARKs: VC:   0.239s VO:  12.260s A:   0.000s Sys:   0.125s =   12.625s
BENCHMARK%: VC:  1.8966% VO: 97.1136% A:  0.0000% Sys:  0.9898% = 100.0000%

gl
BENCHMARKs: VC:   5.085s VO:   7.451s A:   0.000s Sys:   0.114s =   12.650s
BENCHMARK%: VC: 40.1961% VO: 58.9013% A:  0.0000% Sys:  0.9026% = 100.0000%

gl2
BENCHMARKs: VC:   5.156s VO:   7.356s A:   0.000s Sys:   0.113s =   12.625s
BENCHMARK%: VC: 40.8394% VO: 58.2658% A:  0.0000% Sys:  0.8948% = 100.0000%

gl_nosw
BENCHMARKs: VC:   5.125s VO:   7.397s A:   0.000s Sys:   0.116s =   12.637s
BENCHMARK%: VC: 40.5506% VO: 58.5324% A:  0.0000% Sys:  0.9169% = 100.0000%

As you can see, processing the video codec (VC) remains roughly the same, it's the video output that suffers.  It's like video hardware acceleration still works, but with horrible inefficiencies.

edit: I wonder if this is a separate bug, maybe with i915?  I'll test uxa vs sna.

edit-2: Yeah, it's a problem with intel sna.  Switching back to uxa completely solves the problem.  Sorry for bothering you about it.

Last edited by the sad clown (2013-01-29 18:46:33)


I laugh, yet the joke is on me

Offline

#334 2013-02-01 03:46:23

Hspasta
Member
Registered: 2011-12-24
Posts: 187
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

@the sad clown
That's a nice find! I was having trouble with the same thing. I never considered uxa vs sna. Finally I can have a system without any power regressions or gpu issues...at least for the time being.

Offline

#335 2013-02-01 06:08:52

KaiSforza
Member
Registered: 2012-04-22
Posts: 133
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Just finished building 3.8-rc6. Grab it from the usual place. (See my sig for details)


Thinkpad T420 | Intel 3000 | systemd {,--user}
PKGBUILDs I use | pywer AUR helper

Offline

#336 2013-02-01 12:20:07

sva_h4cky0
Member
From: Surabaya, Indonesia
Registered: 2009-03-25
Posts: 109
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

i just buy cheap laptop with intel gpu/igp, intel pentium b960. after installing Arch on it, reboot, and i get blank screen. LOL ^_^. quick look with systemctl, it appears that X failed to start.
after several reboot finally i can login into KDE. so basically it's happened at random, but last night my system crash once when playing movies.

Linux-ck 3.7.5-2 KDE 4.9.98, using "sna" with boot parameters "ro quiet splash elevator=bfq i915.i915_enable_rc6=1 i915.i915_enable_fbc=1 i915.lvds_downclock=1 i915.semaphores=1"
Lenovo G480


"Am I not good enough for you?"

Offline

#337 2013-02-01 17:11:23

the sad clown
Member
From: 192.168.0.X
Registered: 2011-03-20
Posts: 833

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

sva_h4cky0 wrote:

i just buy cheap laptop with intel gpu/igp, intel pentium b960. after installing Arch on it, reboot, and i get blank screen. LOL ^_^. quick look with systemctl, it appears that X failed to start.
after several reboot finally i can login into KDE. so basically it's happened at random, but last night my system crash once when playing movies.

Linux-ck 3.7.5-2 KDE 4.9.98, using "sna" with boot parameters "ro quiet splash elevator=bfq i915.i915_enable_rc6=1 i915.i915_enable_fbc=1 i915.lvds_downclock=1 i915.semaphores=1"
Lenovo G480

You need to start a new thread if you are having graphical issues.  Try to include things like journalctl logs and what you are using to launch X (I'm assuming KDM?).


I laugh, yet the joke is on me

Offline

#338 2013-02-01 17:52:23

KaiSforza
Member
Registered: 2012-04-22
Posts: 133
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

I've still not noticed any issues on 3.8-rc6. I've put it through quite a few suspend-resume cycles and reboots, and still no spiking of the gpu/cpu.


Thinkpad T420 | Intel 3000 | systemd {,--user}
PKGBUILDs I use | pywer AUR helper

Offline

#339 2013-02-01 18:14:17

sva_h4cky0
Member
From: Surabaya, Indonesia
Registered: 2009-03-25
Posts: 109
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

the sad clown wrote:

You need to start a new thread if you are having graphical issues.  Try to include things like journalctl logs and what you are using to launch X (I'm assuming KDM?).

yeah thank's, sorry if i bother others. i tough my issue might be related with intel "sna" and power issue. i'll keep monitoring this thread.

im using kdm to start X. when blank screen occurs, im able to switch to tty1 and "systemctl status kdm-plymouth.service" give me output something like "x failed to start" (sorry forgot the exact words, im not native english speaker smile ). since its rarely happen at booting, i'll ignore at moment till i get some free time to looking at logs.

i'll try 3.8-rc later, thank's


"Am I not good enough for you?"

Offline

#340 2013-02-01 19:37:12

the sad clown
Member
From: 192.168.0.X
Registered: 2011-03-20
Posts: 833

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

KaiSforza wrote:

I've still not noticed any issues on 3.8-rc6. I've put it through quite a few suspend-resume cycles and reboots, and still no spiking of the gpu/cpu.

Thanks for the new build.  I've tested it out, and agree that I haven't had any problems with it.  I was even able to use sna again without a hitch.  I did want to bring this to your attention though, not because it is causing me any problems, but because it might cause confusion:

resolving dependencies...
looking for inter-conflicts...

Targets (1): linux-mainline-3.8.rc6.0.g88b62b9-1

Total Installed Size:   62.44 MiB

Proceed with installation? [Y/n] y
(1/1) checking package integrity                                                                                                        [###################################################################################] 100%
(1/1) loading package files                                                                                                             [###################################################################################] 100%
(1/1) checking for file conflicts                                                                                                       [###################################################################################] 100%
(1/1) checking available disk space                                                                                                     [###################################################################################] 100%
(1/1) installing linux-mainline                                                                                                         [###################################################################################] 100%
>>> Updating module dependencies. Please wait ...
>>> Generating initial ramdisk, using mkinitcpio.  Please wait...
==> Building image from preset: 'default'
  -> -k /boot/vmlinuz-linux-mainline -c /etc/mkinitcpio.conf -g /boot/initramfs-linux-mainline.img
==> Starting build: 3.8.0-1-rc5-mainline-dirty
  -> Running build hook: [base]
  -> Running build hook: [udev]
  -> Running build hook: [autodetect]
  -> Running build hook: [modconf]
  -> Running build hook: [block]
  -> Running build hook: [filesystems]
  -> Running build hook: [keyboard]
  -> Running build hook: [fsck]
==> Generating module dependencies
==> Creating gzip initcpio image: /boot/initramfs-linux-mainline.img
==> Image generation successful

Notice that while the target is "linux-mainline-3.8.rc6.0.g88b62b9-1" the build name remains "3.8.0-1-rc5-mainline-dirty", which might cause some to think that they have either downloaded the wrong one, or if when they see that upon reboot, that somehow GRUB is picking up on the old rc5 and not the new kernel.


I laugh, yet the joke is on me

Offline

#341 2013-02-01 19:48:17

KaiSforza
Member
Registered: 2012-04-22
Posts: 133
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

the sad clown wrote:
==> Building image from preset: 'default'
  -> -k /boot/vmlinuz-linux-mainline -c /etc/mkinitcpio.conf -g /boot/initramfs-linux-mainline.img
==> Starting build: 3.8.0-1-rc5-mainline-dirty
  -> Running build hook: [base]
  -> Running build hook: [udev]
  -> Running build hook: [autodetect]
  -> Running build hook: [modconf]
  -> Running build hook: [block]
  -> Running build hook: [filesystems]
  -> Running build hook: [keyboard]
  -> Running build hook: [fsck]
==> Generating module dependencies
==> Creating gzip initcpio image: /boot/initramfs-linux-mainline.img
==> Image generation successful

Notice that while the target is "linux-mainline-3.8.rc6.0.g88b62b9-1" the build name remains "3.8.0-1-rc5-mainline-dirty", which might cause some to think that they have either downloaded the wrong one, or if when they see that upon reboot, that somehow GRUB is picking up on the old rc5 and not the new kernel.

Wait...what...let me fix that... ;_;

EDIT: I'll be fixing that. It'll require a rebuild, but for anyone that's seeing this, it still builds the 3.8-rc6 initrd.

EDIT2: Yeah. Actually fixed and put back into the repo.

Last edited by KaiSforza (2013-02-01 21:09:13)


Thinkpad T420 | Intel 3000 | systemd {,--user}
PKGBUILDs I use | pywer AUR helper

Offline

#342 2013-02-05 05:20:36

The Compiler
Member
From: Winterthur, Switzerland
Registered: 2011-05-01
Posts: 132
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Just had it with 3.8 RC6. Not after a suspend but during normal usage where my GPU was used a bit more (viewing google maps).


Currently running a crowdfunding campaign to work on qutebrowser full-time for a few weeks:

http://igg.me/at/qutebrowser

Offline

#343 2013-02-05 13:30:45

Fedes
Member
From: Buenos Aires, Argentina
Registered: 2012-10-23
Posts: 36

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Hey guys,

Found this on the Kernel Changelog for 3.7.6, probably related and fixed on this version:

commit ec8a7ca08363c563f77ea9d38727822de46a3887
Author: Jani Nikula <jani.nikula@intel.com>
Date:   Thu Jan 17 10:24:09 2013 +0200

    drm/i915: fix FORCEWAKE posting reads
   
    commit b514407547890686572606c9dfa4b7f832db9958 upstream.
   
    We stopped reading FORCEWAKE for posting reads in
   
    commit 8dee3eea3ccd3b6c00a8d3a08dd715d6adf737dd
    Author: Ben Widawsky <ben@bwidawsk.net>
    Date:   Sat Sep 1 22:59:50 2012 -0700
   
        drm/i915: Never read FORCEWAKE
   
    and started using something from the same cacheline instead. On the
    bug reporter's machine this broke entering rc6 states after a
    suspend/resume cycle. It turns out reading ECOBUS as posting read
    worked fine, while GTFIFODBG did not, preventing RC6 states after
    suspend/resume per the bug report referenced below. It's not entirely
    clear why, but clearly GTFIFODBG was nowhere near the same cacheline
    or address range as FORCEWAKE.
   
    Trying out various registers for posting reads showed that all tested
    registers for which NEEDS_FORCE_WAKE() (in i915_drv.c) returns true
    work. Conversely, most (but not quite all) registers for which
    NEEDS_FORCE_WAKE() returns false do not work. Details in the referenced
    bug.
   
    Based on the above, add posting reads on ECOBUS where GTFIFODBG was
    previously relied on.
   
    In true cargo cult spirit, add posting reads for FORCEWAKE_VLV writes as
    well, but instead of ECOBUS, use FORCEWAKE_ACK_VLV which is in the same
    address range as FORCEWAKE_VLV.
   
    v2: Add more details to the commit message. No functional changes.
   
    Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=52411
    Reported-and-tested-by: Alexander Bersenev <bay@hackerdom.ru>
    CC: Ben Widawsky <ben@bwidawsk.net>
    Signed-off-by: Jani Nikula <jani.nikula@intel.com>
    Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
    Cc: stable@vger.kernel.org
    [danvet: add cc: stable and make the commit message a bit clearer that
    this is a regression fix and what exactly broke.]
    Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
    Signed-off-by: CAI Qian <caiqian@redhat.com>
    Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Offline

#344 2013-02-06 18:34:32

cullum
Member
Registered: 2013-02-06
Posts: 3

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

On the latest mainline kernel, I was able to reduce power consumption on my Thinkpad T530 from 22W  to 8W by disabling optimus in the BIOS. Apparently the nvidia card was running all the time. Took awhile to figure out, because PowerTOP wasnt saying anything about it. I couldn't say for sure what caused that to happen. I've been running optimus on this laptop for months with no problems until a few weeks ago.

It's nice to see 10+ hours of battery life remaining again.

Offline

#345 2013-02-06 19:15:09

donniezazen
Member
From: Salt Lake City
Registered: 2011-06-24
Posts: 671
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

@cullum You could also try Bumblebee.

Offline

#346 2013-02-09 01:34:42

KaiSforza
Member
Registered: 2012-04-22
Posts: 133
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

Another week, another rc kernel.
Linux-mainline 3.8-rc7 is up in my repository if you want it.


Thinkpad T420 | Intel 3000 | systemd {,--user}
PKGBUILDs I use | pywer AUR helper

Offline

#347 2013-02-09 19:05:58

KaiSforza
Member
Registered: 2012-04-22
Posts: 133
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

So, I ran my computer through the ringer yesterday. I think I rebooted ~30 times for testing other things, but decided that I'd keep an eye out for power spikes too. Didn't see any. No suspend/resume cycles, but I can do that later today.


Thinkpad T420 | Intel 3000 | systemd {,--user}
PKGBUILDs I use | pywer AUR helper

Offline

#348 2013-02-09 22:04:46

Anthony25
Member
Registered: 2012-10-24
Posts: 43

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

The kernel 3.7.6 version improves a lot the energy management of the CPU/intel GPU, and it was a big problem for me because with Optimus (and bumblebee), when I wanted to play at Trine 2 and switch on the Nvidia graphic card, the CPU was so hot that it was underclocking itself to not crash...

Now it seems to be good, even if the bug after the suspend is not fixed yet, you still have to switch on the powersave governor before the suspend to keep the RC6 active on the intel card (and switch on the governor you want after the wake up).

Offline

#349 2013-02-10 10:42:26

hendry
Member
From: Singapore
Registered: 2009-09-08
Posts: 115
Website

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

I don't seem to be going into the coveted RC6 state as much in 3.7.6 astonishingly. Am I losing my mind? 3.7.5 was better for me with my X220.

Offline

#350 2013-02-11 09:50:18

degeberg
Member
From: Denmark
Registered: 2011-02-27
Posts: 6

Re: Kernel 3.6.2 Power Regressions (Sandy Bridge)

hendry wrote:

I don't seem to be going into the coveted RC6 state as much in 3.7.6 astonishingly. Am I losing my mind? 3.7.5 was better for me with my X220.

No, I have the same problem.

Offline

Board footer

Powered by FluxBB