Misc todo for epia

From CourvilleWiki

Contents 1 Misc todo for epia: things I haven't managed to do yet and worth doing 1.1 issue with pvr250 and bttv when using irda: 1.2 improve speed of hda 1.3 mouse goes gooffy with belkin kvm when switching from linux boxes under xfree 1.4 increase glxgears score 1.5 update to latest BIOS 1.6 fix mtrr setup 1.7 enabling DRI 1.8 tv out options for xfree 1.9 resolve grounding issue with morex box 1.10 compilation flags to test for enhanced speed: 1.11 migration to 2.6 1.12 mythtv polish 1.13 lm_sensors configuration 1.14 Test directfb support with mpeg 1.15 5.1 sound with alsa

Misc todo for epia: things I haven't managed to do yet and worth doing

issue with pvr250 and bttv when using irda:

Solution: comment out in drivers/lirc_i2c/lirc_i2c.c

/* request_module("bttv"); */
/* request_module("rivatv"); */

improve speed of hda

• use udma5

hdparm -c1 -d1 -X69 /dev/hda

mouse goes gooffy with belkin kvm when switching from linux boxes under xfree

Suggestions that I collected on the web from http://web.belkin.com/support/faq_qa.asp?pid=190&cid=#887

• disable gpm (!?)
• autoprobe mouse protocol (instead of IMPS/2) in /etc/X11/XF86Config-4 by specifying

Option "Protocol" "auto"

increase glxgears score

Surprisingly this can be achieved by tweaking the BIOS: reset to "optimal defaults" and correct boot devices.

update to latest BIOS

Get it from [1]. _IMPORTANT_: do not forget afterwards to clear NVRAM with the special switch on the board. Note that flashing from an usb key _*doesn't work*_ because awdflash utility deactivates USB support _before_ reading the new BIOS in thus disabling the device on which you boot!!! Solution is to create a custom cd for floppyless configurations. The flash syntax is: awdflash.exe BIOS.bin /py/sn/cd/cp/tiny. ISO of the latest BIOS is kindly provided by totte: [2] Proper configuration of the BIOS for reaching good XFree performance is to set the AGP apperture to 128MB and video memory (framebuffer) to 32MB.

fix mtrr setup

Videoram should be less than 65536kb: 32678ko is enough. To fix it just issue TWICE

echo "disable=3" > /proc/mtrr
echo "disable=3" > /proc/mtrr

this is required to have only reg01 reg02 reg04 left in /proc/mtrr. In latest XFree this issue is fixed. This way you switch from a situation where:

#cat /proc/mtrr
reg00: base=0x00000000 ( 0MB), size= 128MB: write-back, count=1
reg01: base=0x08000000 ( 128MB), size= 64MB: write-back, count=1
reg02: base=0x0c000000 ( 192MB), size= 32MB: write-back, count=1
reg03: base=0xd0000000 (3328MB), size= 128MB: write-combining, count=2
reg04: base=0xd8000000 (3456MB), size= 64MB: write-combining, count=2
#echo "disable=3" > /proc/mtrr
#echo "disable=3" > /proc/mtrr
#cat /proc/mtrr
reg00: base=0x00000000 ( 0MB), size= 128MB: write-back, count=1
reg01: base=0x08000000 ( 128MB), size= 64MB: write-back, count=1
reg02: base=0x0c000000 ( 192MB), size= 32MB: write-back, count=1
reg04: base=0xd8000000 (3456MB), size= 64MB: write-combining, count=1

According to _totte_, 2 separate regions are seen because that both memories in the AGP aperture and in the frame buffer are set to write-combining: actually both the via X driver and the via dri driver sets up the same write-combining regions.

enabling DRI

• it requires in /etc/X11/XF86Config-4 the presence of

Section "DRI"
Mode 0666
EndSection

• module via_v4l_drv need to be loaded AFTER agpgart
• reduce via_dri size: strip --strip-debug via_dri.so

tv out options for xfree

Section "Device"

Identifier "CLE266"
Driver "via"
VideoRam 32768

Option "NoDDCValue"
Option "ActiveDevice" "TV"
Option "TVDotCrawl" "True"
Option "PAL"
Option "TVVScan" "over"
Option "TVOutput" "S-Video"

EndSection

resolve grounding issue with morex box

Just ground better the motherboard by linking with a simple wire the closest screw to the PS2 plug to one of the screws used for the cdrom.

compilation flags to test for enhanced speed:

Rroet advises:

HOST="i686-pc-linux-gnu"
CFLAGS="-Os -march=i686 -msse -mmmx"

Actually the epia board doesn't seem to have any onboard cache so -O2 and -O3 optimizations can be skipped, since they are not to be used anyway. Thus just put in -Os for targetting faster execution. Totte reports that -O2 performs better than -O3 speed wise.

migration to 2.6

• add in /etc/modutils/epiam

install via /sbin/modprobe via_agp; /sbin/modprobe --ignore-install via; /bin/true

• must do a modprobe via_agp before starting via_v4l_drv or X in order to get dri usable as a non-root user
• viafb pre patch available at [3] and use for mapping the console to the framebuffer the tool called "con2fb" [4]
• former viafb parameters were viafb TVtype=1 TVoverscan0=0 TVon=1

tar jxf linux-2.6.3.tar.bz2
cd linux-2.6.3/
zcat ../via-v4l-1.4a-drm.patch.gz | patch -p1

mythtv polish

• auto mythtv login on mediabox: su - mythtv --login -c /usr/X11R6/bin/startx & in place of kdm or other

lm_sensors configuration

Check if following is the same than at home, modules to be loaded are: i2c-viapro, i2c-isa, eeprom and vt1211

chip "vt1211-*" "vt8231-*"
    ignore temp1
    ignore in6
    set config 12
    ignore in0
    ignore in1
    ignore temp2
    ignore temp4
    ignore temp5
    ignore temp6
    ignore temp7
    ignore fan1
    ignore fan2
    label in2 "VCore1"
    label in3 "+5V"
    label in4 "+12V"
    label in5 "+3.3V"

    label temp3 "Proc Temp"
    compute in2 ((@ * 100) - 3) / (0.5952 * 95.8), (@ * 0.5952 * 0.958) + .03
    compute in3 ((@ * 100) - 3) / (0.4167 * 95.8), (@ * 0.4167 * 0.958) + .03
    compute in4 ((@ * 100) - 3) / (0.1754 * 95.8), (@ * 0.1754 * 0.958) + .03
    compute in5 ((@ * 100) - 3) / (0.6296 * 95.8), (@ * 0.6296 * 0.958) + .03

    set vrm 9.1
    set in2_min vid * 0.97
    set in2_max vid * 1.03
    set in3_min 5.0 * 0.95
    set in3_max 5.0 * 1.05
    set in4_min 12.0 * 0.90
    set in4_max 12.0 * 1.10
    set in5_min 3.3 * 0.95
    set in5_max 3.3 * 1.05

    compute temp3  (@ - 65) / 0.9686,  (@ * 0.9686) + 65

    set temp2_hyst 55
    set temp2_over 60
    set temp3_hyst 60
    set temp3_over 65
    set temp4_hyst 60
    set temp4_over 65

Test directfb support with mpeg

• reboot the machine with the following kernel option vga=791 video=vesa:wrap,mtrr,vram:32 (use 789 for 800x600)
• install DirectFB-0.9.20 and run ldconfig
• load cle266vgaio module obtained from either from the 2.4.2x-epia1 kernel, or from DirectFB CVS
• compile Andreas mtest: [5] and run it on a video file
• according to cozzap (quoting): "mtest doesn't do any demultiplexing, therefore you'll need the raw video stream - the 'elementary stream' or ES. An mpeg file typically contains different streams - usually one video and at least one audio. I won't go into the details here, but typically the video stream has an ID of e0-ef".
• the extraction of the video itself from the mpeg stream can be achieved through ts2es from dvb-mpegtools package

5.1 sound with alsa

• requires recent version of alsa
• in alsamixer control speakers that way: front with master, rear with headphones, master with pcm and subwoofer/center with LFE