Sluggish performance with Slackware 12.2 2.6.27.7-smp on Gigabyte GA-8S649MF

kslen · 05-11-2009, 10:14 PM

Recompiled as suggested above. Still no luck.

Code:

Using /lib/modules/2.6.27.7-twk/kernel/fs/reiserfs/reiserfs.ko
mount: mounting /dev/sda1 on /mnt failed: No such file or directory
ERROR: No /sbin/init found on rootdev (or not mounted). Trouble ahead.
       You can try to fix it. Type 'exit' when things are done.

/bin/sh: can't access tty; job control turned off
/ $ _

Same story with /dev/hda1.

I'll check out the compositing later today. Need to catch me some Zs.

Ilgar · 05-12-2009, 03:19 AM

Well then it seems you need the old library to use the drive. I made a Google search about the model number, there were a couple of hits where it was recognized as sda. There was also another case (kernel 2.6.9) where it was the device hda and hdparm gave 517 MB/s cached read speed. So your low speed in hda mode could be caused some other thing which causes an overall slow down.

At this point I can only think of "deep" things like dynamic ticks, cpu throttling (cpufreq), smp as the possible causes. These features were quite stable in 2.6.27, but who knows. As a somewhat different thing I noticed you have the it87 module, but I'm not sure if it could cause this sort of behaviour.

Do your system logs indicate any errors or unusual things?

forum1793 · 05-12-2009, 07:27 AM

Is it possible your bios settings direct the drive to be treated as ide? Many BIOSs have that capability as some systems (OS) cannot boot from an sata drive. My system has 5 sata ports and the bios allows for all of them to be treated as sata or as ide, but it also has a second feature allowing one of the five to be treated differently, so I can have 4 satas and treat the 5th as sata or ide. This is useful for booting with OS that needs ide (hd or cdrom). Early on I had very slow performance with disk operation and I think this was the cause. But also around the same time I upgraded slackware, probably from 12 to 12.1 so that might have done it also.

kslen · 05-14-2009, 10:07 AM

Sorry for the late response. Been busy times lately.

Actually, the SATA option in BIOS was set to IDE. Changed it to RAID which were the only other option and I'm now doing a recompile with SiS ATA support as changing the BIOS option had no effect on the kernel mentioned in my previous post. This time around I'm leaving out all Generic ATA from SATA/PATA drivers "page" in the kernel setup. Perhaps that'll do the trick..?

Plugged in a SATA drive and hdparm -tT results were much slower than the IDE drive results in previous replies.

Also browsed the logfiles and they were chugging along quite contently. IDE drive running at ATA/100, chipsets detected correctly and so forth.

kslen · 05-14-2009, 10:12 AM

Oh, yea. it87 is my precious.. Needs my temp readings.

And no, I doubt it. I fired up lm_sensors late in the game and the issue was present before touching that module. I'll unload it and run hdparm just to be sure though.

Thanks for sticking with me this far. ^^

kslen · 05-14-2009, 01:16 PM

Could anyone load this config and see if there's something blatantly obvious I'm leaving out besides the generic drivers?

I'm getting the same error about not finding kernel modules and thus not /dev/hda1 as mentioned in previous post.

Download fail: config-fail

Ilgar · 05-14-2009, 02:33 PM

Your ext3 filesystem is compiled as a module. Did you make an initrd to load support for it?

Edit: Also, in case it's something about hda not being found, under ATA/ATAPI/MFM/RLL support, you might need to add

generic/default IDE chipset support
and/or
Generic PCI IDE Chipset Support (or just mark the controller on your board instead).

forum1793 · 05-14-2009, 10:11 PM

I think sata requires scsi support and will require scsi disk support. Your config shows that is off. Should be:

#
# SCSI support type (disk, tape, CD-ROM)
#
CONFIG_BLK_DEV_SD=y

Try adding scsi disk support into the kernel.

kslen · 05-21-2009, 09:33 AM

Code:

> make clean
> make menuconfig
Enabled SCSI, disabled SMP.
> make
> make modules_install

> make kernelversion                                                           
2.6.27.7

> cp arch/i386/boot/bzImage /boot/bzImage-2.6.27.7

> cp System.map /boot/System.map-2.6.27.7

> cd /boot/                                                                    
> mkinitrd -c -k 2.6.27.7 -m ext2:ext3:reiserfs -f reiserfs -r /dev/hda1       
OK: /lib/modules/2.6.27.7/kernel/fs/ext2/ext2.ko added.
OK: /lib/modules/2.6.27.7/kernel/fs/mbcache.ko added.
OK: /lib/modules/2.6.27.7/kernel/fs/jbd/jbd.ko added.
OK: /lib/modules/2.6.27.7/kernel/fs/ext3/ext3.ko added.
OK: /lib/modules/2.6.27.7/kernel/fs/reiserfs/reiserfs.ko added.
3984 blocks

/etc/lilo.conf

Code:

image = /boot/bzImage-2.6.27.7
  initrd = /boot/initrd.gz
  root = /dev/hda1
  label = Slackware-twk
  read-only

hdparm -tT /dev/sda

Code:

/dev/sda:
 Timing cached reads:   254 MB in  2.01 seconds = 126.62 MB/sec
 Timing buffered disk reads:  188 MB in  3.03 seconds =  62.04 MB/sec

hdparm -tT /dev/hda

Code:

/dev/hda:
 Timing cached reads:   252 MB in  2.01 seconds = 125.28 MB/sec
 Timing buffered disk reads:  164 MB in  3.00 seconds =  54.58 MB/sec

lspci

Code:

00:00.0 Host bridge: Silicon Integrated Systems [SiS] SiS649 Host (rev 10)
00:01.0 PCI bridge: Silicon Integrated Systems [SiS] PCI-to-PCI bridge
00:02.0 ISA bridge: Silicon Integrated Systems [SiS] SiS965 [MuTIOL Media IO] (rev 48)
00:02.5 IDE interface: Silicon Integrated Systems [SiS] 5513 [IDE] (rev 01)
00:02.7 Multimedia audio controller: Silicon Integrated Systems [SiS] AC'97 Sound Controller (rev a0)
00:03.0 USB Controller: Silicon Integrated Systems [SiS] USB 1.1 Controller (rev 0f)
00:03.1 USB Controller: Silicon Integrated Systems [SiS] USB 1.1 Controller (rev 0f)
00:03.2 USB Controller: Silicon Integrated Systems [SiS] USB 1.1 Controller (rev 0f)
00:03.3 USB Controller: Silicon Integrated Systems [SiS] USB 2.0 Controller
00:05.0 RAID bus controller: Silicon Integrated Systems [SiS] 182 SATA/RAID Controller (rev 01)
00:06.0 PCI bridge: Silicon Integrated Systems [SiS] PCI-to-PCI bridge
00:07.0 PCI bridge: Silicon Integrated Systems [SiS] PCI-to-PCI bridge
00:09.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL-8139/8139C/8139C+ (rev 10)
00:0c.0 FireWire (IEEE 1394): VIA Technologies, Inc. VT6306 Fire II IEEE 1394 OHCI Link Layer Controller (rev 80)
01:00.0 VGA compatible controller: nVidia Corporation G72 [GeForce 7300 LE] (rev a1)

lsmod

Code:

Module                  Size  Used by
it87                   19600  0
hwmon_vid               2944  1 it87
snd_seq_dummy           2692  0
snd_seq_oss            27264  0
snd_seq_midi_event      5888  1 snd_seq_oss
snd_seq                40912  5 snd_seq_dummy,snd_seq_oss,snd_seq_midi_event
snd_seq_device          6284  3 snd_seq_dummy,snd_seq_oss,snd_seq
snd_pcm_oss            34848  0
snd_mixer_oss          13568  1 snd_pcm_oss
ipv6                  222100  22
lp                      9380  0
usbhid                 39808  0
hid                    35968  1 usbhid
usblp                  10880  0
ppdev                   7172  0
evdev                   8736  4
nvidia               7221244  0
thermal                15516  0
processor              22560  1 thermal
thermal_sys            10912  2 thermal,processor
snd_intel8x0           27804  0
ohci_hcd               21264  0
parport_pc             23588  1
agpgart                27184  1 nvidia
rtc_cmos                8608  0
rtc_core               14232  1 rtc_cmos
rtc_lib                 2560  1 rtc_core
snd_ac97_codec         94500  1 snd_intel8x0
ac97_bus                1664  1 snd_ac97_codec
ohci1394               27824  0
parport                29256  3 lp,ppdev,parport_pc
snd_pcm                61448  3 snd_pcm_oss,snd_intel8x0,snd_ac97_codec
i2c_core               19472  1 nvidia
snd_timer              17796  2 snd_seq,snd_pcm
ieee1394               73540  1 ohci1394
psmouse                35216  0
snd                    44324  9 
snd_seq_oss,snd_seq,snd_seq_device,snd_pcm_oss,snd_mixer_oss,snd_intel8x0,snd_ac97_codec,snd_pcm,snd_timer
soundcore               5600  1 snd
snd_page_alloc          7304  2 snd_intel8x0,snd_pcm
button                  6160  0
hwmon                   2196  2 it87,thermal_sys
ehci_hcd               30988  0
serio_raw               4996  0

The kernel boots successfully now with support for both ATA and SATA but with horrible performance as usual.

Kernel config: Download less fail.

kslen · 05-21-2009, 03:03 PM

Oh, and..

dmesg | grep CPU

Code:

Initializing CPU#0
SLUB: Genslabs=12, HWalign=128, Order=0-3, MinObjects=0, CPUs=1, Nodes=1
CPU: Trace cache: 12K uops, L1 D cache: 16K
CPU: L2 cache: 1024K
Intel machine check reporting enabled on CPU#0.
CPU0: Intel P4/Xeon Extended MCE MSRs (24) available
CPU: Intel(R) Pentium(R) 4 CPU 3.20GHz stepping 01

I do notice a slight performance decrease all in all without HT.

By the way. Any tips as to what is completely useless in the kernel are most welcome.

kslen · 05-23-2009, 06:07 PM

Quote:

Originally Posted by kslen

> mkinitrd -c -k 2.6.27.7 -m ext2:ext3:reiserfs -f reiserfs -r /dev/hda1

Here's the bastard which have been throwing me off balance all this time after recompiles..

It should be;

Code:

mkinitrd -c -k 2.6.27.7 -m ext2:ext3:reiserfs -f reiserfs -r /dev/sda1

ofc..

The drive previously listed as /dev/hda is now listed as /dev/sda, yet there's no improvement in read speed.

I've gotta backtrack a few steps in order to verify the SiS modules performance to the generic ATA support which is currently compiled into the kernel.

Be reporting back soon.

kslen · 05-23-2009, 08:02 PM

Ahoy.

I've recompiled the kernel exactly as disturbed1 suggested earlier and the it booted gleefully. The IDE drive is listed as /dev/sda, however; also as mentioned earlier. The hdparm -tT /dev/sda results are pretty much exactly the same as the results previously posted results..

disturbed1 · 05-24-2009, 01:40 AM

My nforce2 chipset has the choice to use either module, like you, I noticed no speed increase between /dev/hd$ and /dev/sd$ for IDE drives, but... the system seemed to be more responsive using libata.

The main problem with your hdparm -tT results were the cached reads. For an IDE drive, your buffered disk reads were with in reason (40-60MB/s). All of my ATA 100 hard drives test close to what your getting, but the cached reads are much higher, in the 5,000 - 6,000MB/s range. The cached reads should be this high because it is reading from the cached memory.

Could be that my IDE drives have between 8MB and 32MB of cache.

Have you tried running a smart test on the drive just to rule anything out? You could run smartctl --all /dev/sda to get a brief summary of the drive. I tend to use the manufacturer's tools to diagnose hard drives. These come as bootable isos, and can run a complete hard drive test. Depending on the drive size, you may want to let it run over night, it can take a while.

Some other things I've seen that cause problems are more than one device on the IDE channel, where one of those devices are faulty, or slower. Like having an optical drive and hard drive share the same cable. Faulty cables, and temp issues.

Should be noted that hdparm is like glxgears, not really a benchmark

dd if=/dev/zero of=test.bin bs=512000 count=1000
The above will write a 512MB file and report the write speed. Best to perform more than once, the first one is usually not a true result.

There's been an on going discussion over a huge Kernel regression since 2.6.18 which coincides with the time CFQ became the default scheduler. Not everyone has an issue, and it's been near impossible for the smart people

to nail down. If you search this forum, there's a post about changing the default scheduler. You can try that to see if it helps.

Ilgar · 05-24-2009, 03:51 AM

Well I have similar thoughts as disturbed1, but what confuses me is that kslen says he didn't notice this performance loss on XP, so it probably isn't faulty hardware. I know one kernel regression which causes loss of system responsiveness during heavy disk I/O (which starts around 2.6.17) and yes, I also remember rare complaints about CFQ which was made the default around 2.6.22 or 23 as far as I can remember. But none of these should affect the cached read speed I think. I suffer from the first regression and I have a slow disk but cached read rate is way better (around 720 MB/s). I suspect another deeper problem could be causing this.

... I just made another Google search. I've come across people reporting bad performance with the Sis 5513 controller. Some of these people seemed to get better performance with the old drivers. If you keep your generic kernel around, could you try the boot options

Code:

ide0=ata66 ide0=autotune hda=autotune

And finally, you could also try the options

Code:

noapic acpi=off

to see if these subsystems could be part of the trouble. These likely won't solve your problem but they are the only things I can think of now.

kslen · 05-24-2009, 03:04 PM

After one more attempt at scrounging around for a BIOS update, I finally found one. And guess what..

hdparm -tT /dev/sda

Code:

Timing cached reads:   1034 MB in 2.00 seconds = 516.58 MB/sec
Timing buffered disk reads:  166 MB in 3.02 seconds = 54.88 MB/ssec

My choppy issues are no more.

If anyone should happen to be looking for a BIOS upgrade or drivers in general for the Gigabyte GA-8S649MF motherboard, look no further:
Link to support site.

Now I've just got to figure out how to boot the kernel successfully as my SATA drive seems to steal sda during boot. Should be easy, but one never knows in these parts of the woods....

If you're wondering "Why the frack is this guy booting offa an IDE drive when he's got a SATA drive in there?". Well, the SATA drive will move along with me in not a too distant future, the rest of the machine will remain at its current location and I don't really need a SATA drive for the work it will be doing, so an IDE drive will suffice. ^^

In any case, great props to you all for the help. I learned a bunch. Greatly appreciate it.

Until next time. ^^