scheduling problems in X with 2.6.0

scheduling problems in X with 2.6.0

[martin f krafft]

[-- Attachment #1: Type: text/plain, Size: 2070 bytes --]

Hi all,

Ever since I moved to 2.6.0, I have been experiencing something I'd
like to call scheduling problems while working in X. I have tried
using preemptive mode, and turning it off, but the symptoms are the
same:

whenever there is continuous disk access (e.g. tar, rsync, dd),
X will not respond for a couple of seconds every couple of seconds.
With that I mean that the mouse will freeze as well as all screen
output, and then resume after a couple of seconds.

I wonder if I am the only one with that problem. The machine in
question is a dual AMD 2400+ with 2Gb of RAM and a Maxtor DiamondMax
drive spinning at 7200 RPM. The drive is configured as follows
(hdparm):

 multcount    = 16 (on)
 IO_support   =  1 (32-bit)
 unmaskirq    =  0 (off)
 using_dma    =  1 (on)
 keepsettings =  0 (off)
 readonly     =  0 (off)
 readahead    = 256 (on)
 geometry     = 16383/255/63, sectors = 241254720, start = 0
 Model=IC35L120AVV207-0, FwRev=V24OA63A, SerialNo=VNVD06G4C1NXZL
 Config={ HardSect NotMFM HdSw>15uSec Fixed DTR>10Mbs }
 RawCHS=16383/16/63, TrkSize=0, SectSize=0, ECCbytes=52
 BuffType=DualPortCache, BuffSize=1821kB, MaxMultSect=16, MultSect=16
 CurCHS=65535/1/63, CurSects=4128705, LBA=yes, LBAsects=241254720
 IORDY=on/off, tPIO={min:240,w/IORDY:120}, tDMA={min:120,rec:120}
 PIO modes:  pio0 pio1 pio2 pio3 pio4 
 DMA modes:  mdma0 mdma1 mdma2 
 UDMA modes: udma0 udma1 udma2 udma3 udma4 *udma5 
 AdvancedPM=yes: disabled (255) WriteCache=enabled
 Drive conforms to: ATA/ATAPI-6 T13 1410D revision 3a: 

I would appreciate any pointers to the source of the problem, or
general hints!

-- 
martin;              (greetings from the heart of the sun.)
  \____ echo mailto: !#^."<*>"|tr "<*> mailto:" net@madduck
 
invalid/expired pgp subkeys? use subkeys.pgp.net as keyserver!
 
"never try to explain computers to a layman.
 it's easier to explain sex to a virgin."
                                                    -- robert heinlein
 
(note, however, that virgins tend to know a lot about computers.)

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: scheduling problems in X with 2.6.0

[Nick Piggin]

martin f krafft wrote:

>Hi all,
>
>Ever since I moved to 2.6.0, I have been experiencing something I'd
>like to call scheduling problems while working in X. I have tried
>using preemptive mode, and turning it off, but the symptoms are the
>same:
>
>whenever there is continuous disk access (e.g. tar, rsync, dd),
>X will not respond for a couple of seconds every couple of seconds.
>With that I mean that the mouse will freeze as well as all screen
>output, and then resume after a couple of seconds.
>
>I wonder if I am the only one with that problem. The machine in
>question is a dual AMD 2400+ with 2Gb of RAM and a Maxtor DiamondMax
>drive spinning at 7200 RPM. The drive is configured as follows
>(hdparm):
>

Can you post 20 or so lines of 'vmstat 1' captured while the problem is
happening? See if you can see which lines correspond to X freezing (ie.
watch the xterm), but that might be impossible if everything freezes.

Re: scheduling problems in X with 2.6.0

[martin f krafft]

[-- Attachment #1: Type: text/plain, Size: 3109 bytes --]

also sprach Nick Piggin <piggin@cyberone.com.au> [2004.01.07.1630 +0100]:
> Can you post 20 or so lines of 'vmstat 1' captured while the problem is
> happening? See if you can see which lines correspond to X freezing (ie.
> watch the xterm), but that might be impossible if everything freezes.

procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----
 r  b   swpd   free   buff  cache   si   so    bi    bo   in    cs us sy id wa
 2  2      0 1137424 235064 456972    0    0   736   712 4303  8686 15  6 17 62
 0  3      0 1135696 236116 457008    0    0  1000   688 4321  8004 11 15 14 61
 2  2      0 1132336 237724 457304    0    0  1508    40 4475 10604 25 12 19 44
 1  2      0 1129008 238852 457400    0    0  1036  1616 4476  9232 17 11 13 59
 1  2      0 1125488 240740 457552    0    0  1836  1408 4509  9179 15 10  4 72
 0  3      0 1121328 243212 457732    0    0  2340   988 4634 11376 27 14 12 47
 2  5      0 1117744 245336 457920    0    0  2040    44 4538  9344 17 17 16 51
 1  1      0 1115632 246228 458116    0    0   776   784 4339  9230 22  9  5 63
 0  3      0 1112624 247448 458256    0    0  1108  1424 4456  9141 22 10 22 46
 1  5      0 1110896 248104 458416    0    0   576   536 4294  8767 18  8  6 68
 3  1      0 1107728 250068 458492    0    0  1876   792 4626  9652 20 10 23 47
 1  2      0 1105872 251028 458620    0    0   916  1236 4399  8839 16 17 19 48
 0  2      0 1102224 253096 458864    0    0  1956   500 4583  9880 29 11 13 49
 2  1      0 1098896 255128 459008    0    0  1972   464 4628  9729 18  8 23 51
 0  2      0 1096448 256500 459132    0    0  1316   920 4476  9485 18 10  3 71
 0  3      0 1094016 257884 459244    0    0  1280  1004 4467  9896 23  9  8 59
 0  3      0 1091328 258800 459416    0    0   848   824 4316  8754 17 17 12 53
 1  2      0 1087744 260236 459476    0    0  1392    28 4447  9077 13  8 16 62
 0  2      0 1086096 261028 459704    0    0   692   788 4226  8918 24  8 12 56
 0  3      0 1085776 261284 459652    0    0   256   760 4200  7934  6  4 29 63
 4  2      0 1083632 262220 459872    0    0   796   504 4321  9680 29 10 10 51
 0  3      0 1082800 262924 459848    0    0   704   456 4227  7885  6 13 17 64
 2  2      0 1081824 263500 460020    0    0   512    28 4228  9024 19  6 13 62
 1  2      0 1080608 264360 460044    0    0   824   388 4238  9320 14  7 13 66
 2  2      0 1079648 264964 460188    0    0   544   388 4260  8962 19  7 13 60
 0  2      0 1078160 265800 460304    0    0   808   360 4264  8831 15  9 28 48

Although I can't see a clear pattern of when exactly it froze, it
definitely froze everytime right before outputting a line with
r > 1. Sometimes, it also froze when r == 1 and sometimes even with
r == 0.

Thanks!

-- 
martin;              (greetings from the heart of the sun.)
  \____ echo mailto: !#^."<*>"|tr "<*> mailto:" net@madduck
 
invalid/expired pgp subkeys? use subkeys.pgp.net as keyserver!
 
"wer ein warum hat, dem ist kein wie zu schwer."
                                                 - friedrich nietzsche

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: scheduling problems in X with 2.6.0

[Nick Piggin]

martin f krafft wrote:

>also sprach Nick Piggin <piggin@cyberone.com.au> [2004.01.07.1630 +0100]:
>
>>Can you post 20 or so lines of 'vmstat 1' captured while the problem is
>>happening? See if you can see which lines correspond to X freezing (ie.
>>watch the xterm), but that might be impossible if everything freezes.
>>
>
>procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----
> r  b   swpd   free   buff  cache   si   so    bi    bo   in    cs us sy id wa
> 2  2      0 1137424 235064 456972    0    0   736   712 4303  8686 15  6 17 62
> 0  3      0 1135696 236116 457008    0    0  1000   688 4321  8004 11 15 14 61
> 2  2      0 1132336 237724 457304    0    0  1508    40 4475 10604 25 12 19 44
> 1  2      0 1129008 238852 457400    0    0  1036  1616 4476  9232 17 11 13 59
> 1  2      0 1125488 240740 457552    0    0  1836  1408 4509  9179 15 10  4 72
> 0  3      0 1121328 243212 457732    0    0  2340   988 4634 11376 27 14 12 47
> 2  5      0 1117744 245336 457920    0    0  2040    44 4538  9344 17 17 16 51
> 1  1      0 1115632 246228 458116    0    0   776   784 4339  9230 22  9  5 63
> 0  3      0 1112624 247448 458256    0    0  1108  1424 4456  9141 22 10 22 46
> 1  5      0 1110896 248104 458416    0    0   576   536 4294  8767 18  8  6 68
> 3  1      0 1107728 250068 458492    0    0  1876   792 4626  9652 20 10 23 47
> 1  2      0 1105872 251028 458620    0    0   916  1236 4399  8839 16 17 19 48
> 0  2      0 1102224 253096 458864    0    0  1956   500 4583  9880 29 11 13 49
> 2  1      0 1098896 255128 459008    0    0  1972   464 4628  9729 18  8 23 51
> 0  2      0 1096448 256500 459132    0    0  1316   920 4476  9485 18 10  3 71
> 0  3      0 1094016 257884 459244    0    0  1280  1004 4467  9896 23  9  8 59
> 0  3      0 1091328 258800 459416    0    0   848   824 4316  8754 17 17 12 53
> 1  2      0 1087744 260236 459476    0    0  1392    28 4447  9077 13  8 16 62
> 0  2      0 1086096 261028 459704    0    0   692   788 4226  8918 24  8 12 56
> 0  3      0 1085776 261284 459652    0    0   256   760 4200  7934  6  4 29 63
> 4  2      0 1083632 262220 459872    0    0   796   504 4321  9680 29 10 10 51
> 0  3      0 1082800 262924 459848    0    0   704   456 4227  7885  6 13 17 64
> 2  2      0 1081824 263500 460020    0    0   512    28 4228  9024 19  6 13 62
> 1  2      0 1080608 264360 460044    0    0   824   388 4238  9320 14  7 13 66
> 2  2      0 1079648 264964 460188    0    0   544   388 4260  8962 19  7 13 60
> 0  2      0 1078160 265800 460304    0    0   808   360 4264  8831 15  9 28 48
>
>Although I can't see a clear pattern of when exactly it froze, it
>definitely froze everytime right before outputting a line with
>r > 1. Sometimes, it also froze when r == 1 and sometimes even with
>r == 0.
>

OK so its not VM stalls. In fact, it looks like your system is
only under a moderate load.

You could try my alternate CPU scheduler which would tell us if your
problem is the scheduler or something else. Its against the mm tree.
http://www.kerneltrap.org/~npiggin/v29p6.gz

Re: scheduling problems in X with 2.6.0

[martin f krafft]

[-- Attachment #1: Type: text/plain, Size: 970 bytes --]

also sprach Nick Piggin <piggin@cyberone.com.au> [2004.01.08.1634 +0100]:
> You could try my alternate CPU scheduler which would tell us if your
> problem is the scheduler or something else. Its against the mm tree.
> http://www.kerneltrap.org/~npiggin/v29p6.gz

From a quick glance, it mentions the Pentium 4 and Hyperthreading.
This *is* an SMP system, but it uses Athlon XPs. Will your scheduler
still work?

-- 
martin;              (greetings from the heart of the sun.)
  \____ echo mailto: !#^."<*>"|tr "<*> mailto:" net@madduck
 
invalid/expired pgp subkeys? use subkeys.pgp.net as keyserver!
 
government announcement - the government announced today that it is
changing its mascot to a condom because it more clearly reflects the
government's political stance. a condom stands up to inflation, halts
production, destroys the next generation, protects a bunch of pricks
and finally, gives you a sense of security while you're being screwed!

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: scheduling problems in X with 2.6.0

[martin f krafft]

[-- Attachment #1: Type: text/plain, Size: 630 bytes --]

also sprach Nick Piggin <piggin@cyberone.com.au> [2004.01.08.1634 +0100]:
> OK so its not VM stalls. In fact, it looks like your system is
> only under a moderate load.

I would not expect so. I find it kinda hard to load a system with 2Gb
of RAM and a dual Athlon XP 2400+ using only rsync... the
harddrive bottleneck prevents it from breaking a sweat.

-- 
martin;              (greetings from the heart of the sun.)
  \____ echo mailto: !#^."<*>"|tr "<*> mailto:" net@madduck
 
invalid/expired pgp subkeys? use subkeys.pgp.net as keyserver!
 
linux is like a wigwam.
no gates, no windoze, and an apache inside.

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: scheduling problems in X with 2.6.0

[Nick Piggin]

martin f krafft wrote:

>also sprach Nick Piggin <piggin@cyberone.com.au> [2004.01.08.1634 +0100]:
>
>>You could try my alternate CPU scheduler which would tell us if your
>>problem is the scheduler or something else. Its against the mm tree.
>>http://www.kerneltrap.org/~npiggin/v29p6.gz
>>
>
>From a quick glance, it mentions the Pentium 4 and Hyperthreading.
>This *is* an SMP system, but it uses Athlon XPs. Will your scheduler
>still work?
>

Yeah, my interactivity stuff has to be applied after my SMP stuff,
so I had to give you both ;)

You'll want to run make oldconfig after patching, and you needn't say
yes to CONFIG_SCHED_SMT, although if you did it wouldn't hurt either,
it just wouldn't do anything.

Nick