[opensuse] 10.3 Kernel problem with P5A-B motherboard.
I should be most grateful for advice on how to investigate this problem further. I have installed opensuse 10.3 on a P5A-B motherboard (Aladdin V chipset) with 512MB of SDRAM, Cyrix MII-300 processor. This runs slowly, with somewhat garbled graphics (ATI Radeon 7000) if I use X windows. Without X windows, in run level 3, the load average after booting (and after it has done various database things and settled) is about 0.50. Even if I start it with so-called safe mode and remove all loadable modules apart from those needed to drive the IDE interface the load average never falls below about 0.20. In the first case the CPU usage is about 60% idle and about 30% "si" (is this context switching?). In the second, CPU is about 95% idle, but the load average varies from 0.16 to 0.22. I assume some driver or high priority kernel process too fundamental to show up by "ps" is in a loop occupying processor time. Certainly the machine is even more sluggish than the hardware should warrant (and I am using older hardware in a router running SuSE 10.1 with load averages about .01 to .02). This is clearly too vague for a bug report, can anyone suggest how I can pin this down more, for instance find out where the CPU is spending most of its time? Many thanks, -- Roger Hayter -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
In message <VHvTtBE+xZYHFAlX@kalahari.uninhabited.net>, Roger Hayter <roger@hayter.org> writes
I should be most grateful for advice on how to investigate this problem further. I have installed opensuse 10.3 on a P5A-B motherboard (Aladdin V chipset) with 512MB of SDRAM, Cyrix MII-300 processor. This runs slowly, with somewhat garbled graphics (ATI Radeon 7000) if I use X windows. Without X windows, in run level 3, the load average after booting (and after it has done various database things and settled) is about 0.50. Even if I start it with so-called safe mode and remove all loadable modules apart from those needed to drive the IDE interface the load average never falls below about 0.20. In the first case the CPU usage is about 60% idle and about 30% "si" (is this context switching?). In the second, CPU is about 95% idle, but the load average varies from 0.16 to 0.22.
I assume some driver or high priority kernel process too fundamental to show up by "ps" is in a loop occupying processor time. Certainly the machine is even more sluggish than the hardware should warrant (and I am using older hardware in a router running SuSE 10.1 with load averages about .01 to .02).
This is clearly too vague for a bug report, can anyone suggest how I can pin this down more, for instance find out where the CPU is spending most of its time?
Can anyone advise me where to look in order to trace this problem further? I can find a lot of information on how to trace the behaviour of a C program, running it in a controlled environment: I can't find any information on how to begin finding out what is going wrong in a whole operating system. Any suggestions gratefully received. -- Roger Hayter -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Saturday 15 December 2007 05:04:20 pm Roger Hayter wrote:
Can anyone advise me where to look in order to trace this problem further? I can find a lot of information on how to trace the behaviour of a C program, running it in a controlled environment: I can't find any information on how to begin finding out what is going wrong in a whole operating system. Any suggestions gratefully received.
For the begin try using 'top' to see what is using CPU. Except kernel there is number of system processes that are running all the time like hald, d-bus etc. On a slower machine they would need more time to accomplish tasks and load will appear almost steady. -- Regards, Rajko -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
In message <200712151809.43861.rmatov101@charter.net>, Rajko M. <rmatov101@charter.net> writes
On Saturday 15 December 2007 05:04:20 pm Roger Hayter wrote:
Can anyone advise me where to look in order to trace this problem further? I can find a lot of information on how to trace the behaviour of a C program, running it in a controlled environment: I can't find any information on how to begin finding out what is going wrong in a whole operating system. Any suggestions gratefully received.
For the begin try using 'top' to see what is using CPU.
Except kernel there is number of system processes that are running all the time like hald, d-bus etc. On a slower machine they would need more time to accomplish tasks and load will appear almost steady.
Thanks. I agree with this and the original figures in my post were from "top". But 97% of the time the CPU is doing nothing useful although some process(es) not visible in "top" are raising the load average. -- Roger Hayter -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Roger Hayter wrote: [snipped] Socket 7 systems are so old that there ought to be a K6/2 @ 500MHz or more or a K6-III+ @400 or more to be had near you for between $0 and $25 or so. My P2A-B has a K6/2-500. Any K6* of 400HMz or more should add enough performance to get around several problems. IIRC, that Cyrix CPU only supports a FSB speed of up to 75 MHz, while those K6 chips will all do 100 on the P5A-B. Next, Socket 7 systems were originally designed to depend on motherboard cache for RAM. I don't remember if the Ali chipset does better than most, but I doubt it supports cache for all 512M. I do remember my K6/2-550 was considerably slower on benchmarks with 384M than with 256M on a Tyan S1590 Trinity @ 100 MHz FSB (Via MVP3 chipset). The cache on a K6-III+ chip gets around any shortage of motherboard cache, and can usually be run at 50-150 MHZ above its official rating. So, the OS might be slow and have problems that need a solution, but a cheap or free CPU upgrade should go a long way to alleviate some pain. Have you run a RAM checker like memtest86? Does the Linux actually find all 512M? I thought Socket 7 chips were limited to 384M. -- " Our Constitution was made only for a moral and religious people. It is wholly inadequate to the government of any other." John Adams Team OS/2 ** Reg. Linux User #211409 Felix Miata *** http://mrmazda.no-ip.com/ -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
In message <47647EC4.9090501@ij.net>, Felix Miata <mrmazda@ij.net> writes
Roger Hayter wrote: [snipped]
Socket 7 systems are so old that there ought to be a K6/2 @ 500MHz or more or a K6-III+ @400 or more to be had near you for between $0 and $25 or so. My P2A-B has a K6/2-500. Any K6* of 400HMz or more should add enough performance to get around several problems. IIRC, that Cyrix CPU only supports a FSB speed of up to 75 MHz, while those K6 chips will all do 100 on the P5A-B. Next, Socket 7 systems were originally designed to depend on motherboard cache for RAM. I don't remember if the Ali chipset does better than most, but I doubt it supports cache for all 512M. I do remember my K6/2-550 was considerably slower on benchmarks with 384M than with 256M on a Tyan S1590 Trinity @ 100 MHz FSB (Via MVP3 chipset). The cache on a K6-III+ chip gets around any shortage of motherboard cache, and can usually be run at 50-150 MHZ above its official rating.
So, the OS might be slow and have problems that need a solution, but a cheap or free CPU upgrade should go a long way to alleviate some pain.
Have you run a RAM checker like memtest86? Does the Linux actually find all 512M? I thought Socket 7 chips were limited to 384M.
Yes, sorry I was reading the swap size I had set, it is using 256MB ram! I take your point I could make the system faster, but I am fairly sure I shouldn't need to. The CPU is 97% idle at rest with unnecessary processes stopped but the load average is still 0.2+ - something is wrong. -- Roger Hayter -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Sat, Dec 15, 2007 at 11:04:20PM +0000, Roger Hayter wrote:
In message <VHvTtBE+xZYHFAlX@kalahari.uninhabited.net>, Roger Hayter <roger@hayter.org> writes
I should be most grateful for advice on how to investigate this problem further. I have installed opensuse 10.3 on a P5A-B motherboard (Aladdin V chipset) with 512MB of SDRAM, Cyrix MII-300 processor. This runs slowly, with somewhat garbled graphics (ATI Radeon 7000) if I use X windows. Without X windows, in run level 3, the load average after booting (and after it has done various database things and settled) is about 0.50. Even if I start it with so-called safe mode and remove all loadable modules apart from those needed to drive the IDE interface the load average never falls below about 0.20. In the first case the CPU usage is about 60% idle and about 30% "si" (is this context switching?). In the second, CPU is about 95% idle, but the load average varies from 0.16 to 0.22.
I assume some driver or high priority kernel process too fundamental to show up by "ps" is in a loop occupying processor time. Certainly the machine is even more sluggish than the hardware should warrant (and I am using older hardware in a router running SuSE 10.1 with load averages about .01 to .02).
This is clearly too vague for a bug report, can anyone suggest how I can pin this down more, for instance find out where the CPU is spending most of its time?
Can anyone advise me where to look in order to trace this problem further?
The load idle seems a bit high for just sitting there and doing nothing. Can you run 'powertop' while in text mode (no X running) and see what is showing up as being the thing that is waking the processor up the most? Hopefully that tool will work, but I don't know, as that is a very old processor. thanks, greg k-h -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
In message <20071216050549.GA7706@suse.de>, Greg KH <gregkh@suse.de> writes
On Sat, Dec 15, 2007 at 11:04:20PM +0000, Roger Hayter wrote:
Can anyone advise me where to look in order to trace this problem further?
The load idle seems a bit high for just sitting there and doing nothing.
Can you run 'powertop' while in text mode (no X running) and see what is showing up as being the thing that is waking the processor up the most?
Hopefully that tool will work, but I don't know, as that is a very old processor.
Many thanks for your suggestion. Powertop shows that there are about 250 interrupt events per second, 95%+ being "extra timer interrupts". The answer is about the same whether or not I enable ACPI in the kernel, except for some not wholly credible figures are added (C3 state 159% for example) if I enable ACPI. I can't find any explicit definition of "extra timer interrupts" on the powertop web site, do they mean anything to you? -- Roger Hayter -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Sun, Dec 16, 2007 at 09:41:06PM +0000, Roger Hayter wrote:
In message <20071216050549.GA7706@suse.de>, Greg KH <gregkh@suse.de> writes
On Sat, Dec 15, 2007 at 11:04:20PM +0000, Roger Hayter wrote:
Can anyone advise me where to look in order to trace this problem further?
The load idle seems a bit high for just sitting there and doing nothing.
Can you run 'powertop' while in text mode (no X running) and see what is showing up as being the thing that is waking the processor up the most?
Hopefully that tool will work, but I don't know, as that is a very old processor.
Many thanks for your suggestion. Powertop shows that there are about 250 interrupt events per second, 95%+ being "extra timer interrupts". The answer is about the same whether or not I enable ACPI in the kernel, except for some not wholly credible figures are added (C3 state 159% for example) if I enable ACPI.
Hm, it sounds like powertop will just not work for your older processor, as it is not exporting the information that powertop needs in order to try to figure these things out :( I really don't know what else to suggest at this time, sorry. greg k-h -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
In message <VHvTtBE+xZYHFAlX@kalahari.uninhabited.net>, Roger Hayter <roger@hayter.org> writes
I should be most grateful for advice on how to investigate this problem further. I have installed opensuse 10.3 on a P5A-B motherboard (Aladdin V chipset) with 512MB of SDRAM, Cyrix MII-300 processor. This runs slowly, with somewhat garbled graphics (ATI Radeon 7000) if I use X windows. Without X windows, in run level 3, the load average after booting (and after it has done various database things and settled) is about 0.50. Even if I start it with so-called safe mode and remove all loadable modules apart from those needed to drive the IDE interface the load average never falls below about 0.20. In the first case the CPU usage is about 60% idle and about 30% "si" (is this context switching?). In the second, CPU is about 95% idle, but the load average varies from 0.16 to 0.22.
I assume some driver or high priority kernel process too fundamental to show up by "ps" is in a loop occupying processor time. Certainly the machine is even more sluggish than the hardware should warrant (and I am using older hardware in a router running SuSE 10.1 with load averages about .01 to .02).
This is clearly too vague for a bug report, can anyone suggest how I can pin this down more, for instance find out where the CPU is spending most of its time?
As a matter of interest this is almost completely explained by the resources used by "top". This little programme increases the load average from about .02 to about .25 with significant values for "wait" and "context switching". I was testing with it running in another console, in the erroneous belief it would make little difference. I am sure it didn't used to be such a great consumer of resources! -- Roger Hayter -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Roger Hayter wrote:
In message <VHvTtBE+xZYHFAlX@kalahari.uninhabited.net>, Roger Hayter <roger@hayter.org> writes
I should be most grateful for advice on how to investigate this problem further. I have installed opensuse 10.3 on a P5A-B motherboard (Aladdin V chipset) with 512MB of SDRAM, Cyrix MII-300 processor. This runs slowly, with somewhat garbled graphics (ATI Radeon 7000) if I use X windows. Without X windows, in run level 3, the load average after booting (and after it has done various database things and settled) is about 0.50. Even if I start it with so-called safe mode and remove all loadable modules apart from those needed to drive the IDE interface the load average never falls below about 0.20. In the first case the CPU usage is about 60% idle and about 30% "si" (is this context switching?). In the second, CPU is about 95% idle, but the load average varies from 0.16 to 0.22.
I assume some driver or high priority kernel process too fundamental to show up by "ps" is in a loop occupying processor time. Certainly the machine is even more sluggish than the hardware should warrant (and I am using older hardware in a router running SuSE 10.1 with load averages about .01 to .02).
This is clearly too vague for a bug report, can anyone suggest how I can pin this down more, for instance find out where the CPU is spending most of its time?
As a matter of interest this is almost completely explained by the resources used by "top". This little programme increases the load average from about .02 to about .25 with significant values for "wait" and "context switching". I was testing with it running in another console, in the erroneous belief it would make little difference. I am sure it didn't used to be such a great consumer of resources!
In top, si is NOT context switching. si = servicing software interrupt hi = servicing hardware interrupt -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
In message <47B3C26A.9070605@hotpop.com>, Aaron Kulkis <akulkis00@hotpop.com> writes
Roger Hayter wrote:
In message <VHvTtBE+xZYHFAlX@kalahari.uninhabited.net>, Roger Hayter <roger@hayter.org> writes
I should be most grateful for advice on how to investigate this problem further. I have installed opensuse 10.3 on a P5A-B motherboard (Aladdin V chipset) with 512MB of SDRAM, Cyrix MII-300 processor. This runs slowly, with somewhat garbled graphics (ATI Radeon 7000) if I use X windows. Without X windows, in run level 3, the load average after booting (and after it has done various database things and settled) is about 0.50. Even if I start it with so-called safe mode and remove all loadable modules apart from those needed to drive the IDE interface the load average never falls below about 0.20. In the first case the CPU usage is about 60% idle and about 30% "si" (is this context switching?). In the second, CPU is about 95% idle, but the load average varies from 0.16 to 0.22.
I assume some driver or high priority kernel process too fundamental to show up by "ps" is in a loop occupying processor time. Certainly the machine is even more sluggish than the hardware should warrant (and I am using older hardware in a router running SuSE 10.1 with load averages about .01 to .02).
This is clearly too vague for a bug report, can anyone suggest how I can pin this down more, for instance find out where the CPU is spending most of its time?
As a matter of interest this is almost completely explained by the resources used by "top". This little programme increases the load average from about .02 to about .25 with significant values for "wait" and "context switching". I was testing with it running in another console, in the erroneous belief it would make little difference. I am sure it didn't used to be such a great consumer of resources!
In top, si is NOT context switching.
si = servicing software interrupt hi = servicing hardware interrupt
Thanks, that is helpful, or will be if I get around to investigating what is actually going on. I sort of assumed the abbreviation was as used elsewhere. -- Roger Hayter -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
participants (5)
-
Aaron Kulkis -
Felix Miata -
Greg KH -
Rajko M. -
Roger Hayter