After several kernels being compiled and tried, here are the conclusions of 4 days work. I own a 5600D Clevo/Sager/Promedion, specs as follow: P4 2.4, 512 DDR 1 slot, ati 9000 Mobility 64 DDR, 1400x1050 TFT, 40 gigs HDD, DVD/CDRW Combo, floppy.
This is a true exploration throught hardware. The first thing i would like to bring is the design: My machine is built around a 82845 845 (Brookdale) Chipset labelled i845E by Intel connected to a ICH3M Southbridge. The ICH3M southbridge is a implementation for ata100 on ultra mobiles systems, and is adapted to mobile pentium III processor.
Next is the fact that the i845E chipset is supposed to be connected to a ICH4 chipset, as it is shown in the manufacturer design here http://www.intel.com/design/chipsets...hip+prod_845e&
Is it a special implementation on our laptops ? If yes where is it documented ? These questions have made me consider a hardware flaw, but i am not sure.
I have compiled and run successfully 2.4.18, 2.4.20, 2.4.20 with preempt, ide, acpi,acpi-swsp patches in the stable category, and 2.5.54, 2.5.62 in the devel category.
I'm currently running a 2.4.21-pre4 kernel, with the latest acpi patches. This is running of my choice distro, Debian/GNU Linux Sid Testing/Unstable.
With this i am glad to say that:
1) USB 2.0 works :
ehci-hcd 02:0a.2: VIA Technologies, Inc. USB 2.0
ehci-hcd 02:0a.2: irq 5, pci mem e091fc00
usb.c: new USB bus registered, assigned bus number 3
ehci-hcd 02:0a.2: USB 2.0 enabled, EHCI 0.95, driver 2002-Dec-20
2) Firewire modules are loading
Module Size Used by Tainted:
ohci1394 17512 0 (unused)
ieee1394 34060 0 [ohci1394]
ohci1394: $Rev: 693 $ Ben Collins <bcollins@debian.org>
ohci1394_0: OHCI-1394 1.1 (PCI): IRQ=[11] MMIO=[e0204000-e02047ff] Max Packet=[2048]
ieee1394: SelfID completion called outside of bus reset!
ieee1394: Host added: Node[00:1023] GUID[0000000000140816] [Linux OHCI-1394]
Unfortunately no firewire device, so maybe someone could try it.
3) Sound : I use Alsa 0.9rc7 compiled from source with this options:
./configure --with-cards=intel8x0 --with-sequencer=yes --with-oss=yes --with-isapnp=auto --with-kernel=/home/laclasse/linux-2.4.21-pre4
I have compiled the OSS intel kernel drivers as modules, and i would recommend anybody to install Alsa. There is a huge performance boost on all sound apps, games as well. Doc to compile and download of source packages can be found at http://www.alsa-project.org
4) DMA transfers on the ICH3M Southbridge are a problems to me. I bet they are a problem to anybody cause i suspect a hardware flaw as commented earlier
First on normal kernels (2.4.18 to 2.4.20 unpatched):
ICH3M: IDE controller at PCI slot 00:1f.1
PCI: Disabling device 00:1f.1 due to ressources collisions
ICH3M: not 100% native mode: will probe irqs later
ide0: BM-DMA at 0x1860-0x1867, BIOS settings: hda: : pio, hdb: : pio
ide1: BM-DMA at 0x1868-0x186f, BIOS settings: hdc: : pio, hdd: : pio
So this is of interest, as the we can see that the IDE controller is in pio mode and not DMA.
at this point i run # hdparm -t /dev/hda
Timing buffered disk reads: 64 MB in 3.49 seconds = 18.33 MB/sec
Now according to http://www.intel.com/support/graphics/linux/udma.htm , this is a burst pio mode. Not DMA. I challenge ANYBODY to go in the bios of their Sager/Eurocom/Alienware/Promedion/Clevo 5600D and manage to CHANGE the default [pio/dma2] to UDAM5, as the chipset should support, and as advertised.
Then i run a 2.4.21-pre4 pre patched with the latest hedrix ide-patches, ide-2.4.20-pre5-ac4 and here is the difference:
ICH3M: IDE controller at PCI slot 00:1f.1
PCI: Enabling device 00:1f.1 (0005 -> 0007)
ICH3M: chipset revision 2
ICH3M: not 100% native mode: will probe irqs later
ide0: BM-DMA at 0x1860-0x1867, BIOS settings: hda: : pio, hdb: : pio
ide1: BM-DMA at 0x1868-0x186f, BIOS settings: hdc: : pio, hdd: : pio
Halas still no DMA enabled. Off course the hardisk reports
hda: host protected area => 1
hda: 78140160 sectors (40008 MB) w/2048KiB Cache, CHS=4864/255/63, UDMA(100)
I then made a script to run at boot :
#!/bin/bash
hdparm -c1d1X69 /dev/hda ;
hdparm -u1 -k1 /dev/hda ;
echo "Forcing DMA and 32bit transfert ......................done" ;
echo "Enabling ATA 100 Mode ................................done" ;
As well i have append="idebus=66 ide0=autotune ide1=autotune" in /etc/lilo.conf. So i get
Kernel command line: BOOT_IMAGE=Linux-2.4.21pre ro root=301 idebus=66 ide0=autotune ide1=autotune
ide: Assuming 66MHz system bus speed for PIO modes
now a hdparm -t /dev/hda reads :
Timing buffered disk reads: 64 MB in 3.43 seconds = 18.66 MB/sec
Well wasn't worth the patches. Note that i am using a 25 % throttle on the p4 at all times, causes playing RtCW in linux for 2 hours has brought the lappy to 89 Celcius, and i will not recommend it. The acpi patches to whatever kernel you are essential.
get them here http://sourceforge.net/projects/acpi
and you can try the software suspend here
http://sourceforge.net/projects/swsusp
After the ACPI is successful, you can control the processor throttle, limit temperature, as well as checking the temperature. There is a gkrellm patch to monitor the temperature at http://gkacpi.sourceforge.net
You can see the throttling states by issuing
#cat /proc/acpi/processor/CPU0/throttlingstate count: 8
active state: T2
states:
T0: 00%
T1: 12%
*T2: 25%
T3: 37%
T4: 50%
T5: 62%
T6: 75%
T7: 87%
As you can see i use a 25 % throttle. you can change that by doing :
# echo -n 0 > /proc/acpi/processor/CPU0/throttling
#
this will set your processor at 100 %, no throttle.
Note that i didn't comment much on the deval kernels, cause apart from the benefits of having Alsa integrated, and the acpi, preempt and ide patches in, it didn't offer much improvement in the IDE transfer mode point.
I did email the Sager US support service about that issue to no answers as of today. Everybody seems to ignore the fact that this machine should have 3 X superior disk transfers.
Apart that every other parts of the machine is really satisfactory. You can get the Front panel Audio DJ button and the 3 function buttons working with lineakd as posted here somewhere.
Seems nothing, but realise that if the IDE is on pio, the kernel and CPU overhead is not negligeable at all when running lot of I/O related tasks.
Please feel free to comment as i love to stand corrected.
This is a true exploration throught hardware. The first thing i would like to bring is the design: My machine is built around a 82845 845 (Brookdale) Chipset labelled i845E by Intel connected to a ICH3M Southbridge. The ICH3M southbridge is a implementation for ata100 on ultra mobiles systems, and is adapted to mobile pentium III processor.
Next is the fact that the i845E chipset is supposed to be connected to a ICH4 chipset, as it is shown in the manufacturer design here http://www.intel.com/design/chipsets...hip+prod_845e&
Is it a special implementation on our laptops ? If yes where is it documented ? These questions have made me consider a hardware flaw, but i am not sure.
I have compiled and run successfully 2.4.18, 2.4.20, 2.4.20 with preempt, ide, acpi,acpi-swsp patches in the stable category, and 2.5.54, 2.5.62 in the devel category.
I'm currently running a 2.4.21-pre4 kernel, with the latest acpi patches. This is running of my choice distro, Debian/GNU Linux Sid Testing/Unstable.
With this i am glad to say that:
1) USB 2.0 works :
ehci-hcd 02:0a.2: VIA Technologies, Inc. USB 2.0
ehci-hcd 02:0a.2: irq 5, pci mem e091fc00
usb.c: new USB bus registered, assigned bus number 3
ehci-hcd 02:0a.2: USB 2.0 enabled, EHCI 0.95, driver 2002-Dec-20
2) Firewire modules are loading
Module Size Used by Tainted:
ohci1394 17512 0 (unused)
ieee1394 34060 0 [ohci1394]
ohci1394: $Rev: 693 $ Ben Collins <bcollins@debian.org>
ohci1394_0: OHCI-1394 1.1 (PCI): IRQ=[11] MMIO=[e0204000-e02047ff] Max Packet=[2048]
ieee1394: SelfID completion called outside of bus reset!
ieee1394: Host added: Node[00:1023] GUID[0000000000140816] [Linux OHCI-1394]
Unfortunately no firewire device, so maybe someone could try it.
3) Sound : I use Alsa 0.9rc7 compiled from source with this options:
./configure --with-cards=intel8x0 --with-sequencer=yes --with-oss=yes --with-isapnp=auto --with-kernel=/home/laclasse/linux-2.4.21-pre4
I have compiled the OSS intel kernel drivers as modules, and i would recommend anybody to install Alsa. There is a huge performance boost on all sound apps, games as well. Doc to compile and download of source packages can be found at http://www.alsa-project.org
4) DMA transfers on the ICH3M Southbridge are a problems to me. I bet they are a problem to anybody cause i suspect a hardware flaw as commented earlier
First on normal kernels (2.4.18 to 2.4.20 unpatched):
ICH3M: IDE controller at PCI slot 00:1f.1
PCI: Disabling device 00:1f.1 due to ressources collisions
ICH3M: not 100% native mode: will probe irqs later
ide0: BM-DMA at 0x1860-0x1867, BIOS settings: hda: : pio, hdb: : pio
ide1: BM-DMA at 0x1868-0x186f, BIOS settings: hdc: : pio, hdd: : pio
So this is of interest, as the we can see that the IDE controller is in pio mode and not DMA.
at this point i run # hdparm -t /dev/hda
Timing buffered disk reads: 64 MB in 3.49 seconds = 18.33 MB/sec
Now according to http://www.intel.com/support/graphics/linux/udma.htm , this is a burst pio mode. Not DMA. I challenge ANYBODY to go in the bios of their Sager/Eurocom/Alienware/Promedion/Clevo 5600D and manage to CHANGE the default [pio/dma2] to UDAM5, as the chipset should support, and as advertised.
Then i run a 2.4.21-pre4 pre patched with the latest hedrix ide-patches, ide-2.4.20-pre5-ac4 and here is the difference:
ICH3M: IDE controller at PCI slot 00:1f.1
PCI: Enabling device 00:1f.1 (0005 -> 0007)
ICH3M: chipset revision 2
ICH3M: not 100% native mode: will probe irqs later
ide0: BM-DMA at 0x1860-0x1867, BIOS settings: hda: : pio, hdb: : pio
ide1: BM-DMA at 0x1868-0x186f, BIOS settings: hdc: : pio, hdd: : pio
Halas still no DMA enabled. Off course the hardisk reports
hda: host protected area => 1
hda: 78140160 sectors (40008 MB) w/2048KiB Cache, CHS=4864/255/63, UDMA(100)
I then made a script to run at boot :
#!/bin/bash
hdparm -c1d1X69 /dev/hda ;
hdparm -u1 -k1 /dev/hda ;
echo "Forcing DMA and 32bit transfert ......................done" ;
echo "Enabling ATA 100 Mode ................................done" ;
As well i have append="idebus=66 ide0=autotune ide1=autotune" in /etc/lilo.conf. So i get
Kernel command line: BOOT_IMAGE=Linux-2.4.21pre ro root=301 idebus=66 ide0=autotune ide1=autotune
ide: Assuming 66MHz system bus speed for PIO modes
now a hdparm -t /dev/hda reads :
Timing buffered disk reads: 64 MB in 3.43 seconds = 18.66 MB/sec
Well wasn't worth the patches. Note that i am using a 25 % throttle on the p4 at all times, causes playing RtCW in linux for 2 hours has brought the lappy to 89 Celcius, and i will not recommend it. The acpi patches to whatever kernel you are essential.
get them here http://sourceforge.net/projects/acpi
and you can try the software suspend here
http://sourceforge.net/projects/swsusp
After the ACPI is successful, you can control the processor throttle, limit temperature, as well as checking the temperature. There is a gkrellm patch to monitor the temperature at http://gkacpi.sourceforge.net
You can see the throttling states by issuing
#cat /proc/acpi/processor/CPU0/throttlingstate count: 8
active state: T2
states:
T0: 00%
T1: 12%
*T2: 25%
T3: 37%
T4: 50%
T5: 62%
T6: 75%
T7: 87%
As you can see i use a 25 % throttle. you can change that by doing :
# echo -n 0 > /proc/acpi/processor/CPU0/throttling
#
this will set your processor at 100 %, no throttle.
Note that i didn't comment much on the deval kernels, cause apart from the benefits of having Alsa integrated, and the acpi, preempt and ide patches in, it didn't offer much improvement in the IDE transfer mode point.
I did email the Sager US support service about that issue to no answers as of today. Everybody seems to ignore the fact that this machine should have 3 X superior disk transfers.
Apart that every other parts of the machine is really satisfactory. You can get the Front panel Audio DJ button and the 3 function buttons working with lineakd as posted here somewhere.
Seems nothing, but realise that if the IDE is on pio, the kernel and CPU overhead is not negligeable at all when running lot of I/O related tasks.
Please feel free to comment as i love to stand corrected.
