main index Please update your bookmarks - per favore aggiorna i tuoi bookmark:

English version:

Italian version:

Last update of the above pages was on 2003-Sep-04.

Other notebook-related pages on my site:
- IBM Thinkpad A21m page.
- Apple Powerbook pages.

Yes, this is the "Asus A1370D and Linux" page.

WARNING: the page below contains information and news from March 2001 to November 2003. If you want latest news please go to the updated Asus+Linux main page

Pope John Paul II using a notebook computer Quick index - the news were updated on 2003-Sep-04

  1. some notes before starting;
  2. what's inside the box;
  3. technical specifications of the notebook;
  4. how to install Linux on it!!!
  5. my current BIOS setup;
  6. how to do an incremental network backup on another Linux machine;
  7. testing some nice PCMCIA cards...
  8. how to get hard disk maximum performance using hdparm;
  9. some Linux-related files you can download;
  10. how to read "internet keys" from your software;
  11. things and drivers still missing;
  12. three reasons why ReiserFS is better than EXT2;
  13. wonderful: get to the internet using the IrDA connect with Nokia 7110's internal modem!!
  14. useful: send and receive SMS messages using the IrDA connect with Nokia 7110 phone!
  15. nice: connect to a Windows network drive with Samba!
  16. again, nice: connect to an AppleTalk network with DDP!
  17. interesting: using SunDisk compactflashes and PCMCIA ATA/IDE drives.

  18. the latest news are always added/updated here!!!!


NEWS! NEWS! NEWS! - from newest to oldest - latest update: 2003-Sep-04

Some of the not yet solved problems with kernel 2.6.0-test7 and 2.6.0-test8 (this means that Asus A1xx notebook users cannot consider them as "production" kernels):

Also, the 2.6.0-test8 kernel agpgart section does actually put the AGP 2× devices in "4× speed" mode!

Uh... I told you about "30G" hard-disk. Well, it's 30 billions of bytes and few bytes more. After formatting in reiserfs mode, I got 29,124,948 kilobytes of useful space.

YES, if you compile the 2.6.0-test kernel activating the preemptible option, everything is faster! I really mean FASTER because the recompilation of this site, requiring from 1'20" to 2'00" with old hard disk and 2.4-series kernel (intensive open/read/close of files), now is performed in less than 0'40"!! This is surely due not only to new ATA/IDE features and the Hitachi disk: the ksysguardd reported a 100% use of CPU time (the program doesn't have to wait for completion of every I/O operation: the system calls were "preempted"! YAY!!!)

Well, since some days I'm using the Linux 2.6.0-test7 kernel. It has a number of interesting news. It seems to achieve a better general speed because of the "preemptible" option and the new ATA/IDE "taskfile/packed/etc" methods. But the great news is the improved ACPI support (even with Asus ACPI features... but this requires a recent Asus BIOS; my current BIOS is dated 2001, and I won't change it even if it's buggy: I don't want to experiment a new BIOS only to get some more little features). Current 2.6.0-test7 kernel version allows things like:
cat /proc/acpi/battery/BAT0/state
cat /proc/acpi/thermal_zone/THRM/temperature
cat /proc/acpi/button/lid/LID/state

The first command tells you about battery status (something like: present: yes, present rate: 907 mA, remaining capacity: 1894 mAh). The second command tells you current processor temperature (like: temperature: 33 C), and the last command is a good method, in your programs, to know if the notebook LCD display is open or closed (this is a very nice thing: your program can "poll" that /proc system file and know how much time you are out for a tea...!). Another pretty ACPI feature: if you insert the battery while the computer is on and running, you can get soon after the battery status (up to 2.4.21 kernel, the "battery inserted" check was done only at boot).

I finally got a new hard disk. I installed again SuSE 8.2 on it, and in next days I will refine all things (SuSE 9.0 will be available in next days).

It's a 30Gb Hitachi IC25N030ATMR04-0. I ordered a 40Gb hard disk on September 3rd, they claimed "not immediately available, wait some days, we will call you when the disk is here". Weeks and weeks passed. I went to them to ask at least three times, but they always said "unavailable".

They proposed me a 60Gb disk (at a price 100euro higher!), and I was even willing to buy it, but "I don't want to wait anymore!". No news.

After almost seven weeks, I went to them asking to cancel the order and getting back money. "No problem", the guy said, "but we could have immediately the 30Gb one in two days". Well let's get that. Two "work" days passed, and they told me "yes, it will be here today, but we will be closed, come tomorrow".

And, finally, I got it. From Sep'3rd to Oct'15th. I already knew it, but fell into the trap again. Don't -please- DON'T EVER trust "coming soon" offers!

This Hitachi disk (CHS:3648×255×63) is a nice 9.5mm 2.5" ATA-100 disk, manufactured by Hitachi in Aug'2003 (sadly, our A13xx notebooks will use it only in "UDMA/33" mode, look in the hardware features in this page).

SADLY, it's a 4200rpm disk requiring 5V 1A (ouch!), while the old 20Gb disk required only 0.55A: this means that it will be quite hot (up to now, it doesn't seem to do) and will drain out battery power at high speed (GRRRR!) because it requires almost DOUBLE power! This is a very bad issue, but I don't want to go back just now. It simply runs. I formatted it with a 160Mb swap partition, and the almost 30Gb remaining space in ReiserFS filesystem.



One of my friends told me he got a reply from Motorola saying "we don't provide customer support but our new Linux driver will be available for RedHat 9.0 in a couple of weeks. (this means that they don't "discontinue" anymore Linux support!!! More news soon...!)

Thomas Winischhofer, author of X/free SiS630 accelerated driver, wrote me:

1) AGP support is provided by the AGPgart kernel module, not by any of my drivers.

2) AGP is only good for 3D, not for anything else. The video and 2D accelerators are not designed to use AGP on the 630.

The above just means two important things:

  1. good news: Winischhofer's driver was already working 100% since years!!!

  2. bad news: until a decent fix for SiS/AGPgart kernel support, we cannot expect decent 3D-acceleration on our Asus A13xx notebooks. Don't bother Thomas anymore (except for sending some donation: everyone can see that his driver is 1000 times better than the official SiS driver for Windoze).

Other stuff: sometimes, while fiddling with PCMCIA ATA hard disk and local ethernet, I found the Asus USB mouse "dies". That is, it changes address (usb disconnect/reconnect) but X/free server doesn't know. Below, the dmesg output explaining this (you can halt and reboot the machine, or move the mouse connector to the other USB port and press Ctrl-Alt-BackSpace to restart X/free): hub.c: already running port 1 disabled by hub (EMI?), re-enabling...
usb.c: USB disconnect on device 00:01.2-1 address 2
hub.c: new USB device 00:01.2-1, assigned address 3
input0: USB HID v1.00 Mouse [0603:7134] on usb1:3.0

The Linux 2.6 final test phase has started. Currently, there are patches for the 2.6.0-test3 kernel (the 2.6.0 patches must be applied onto the 2.5.75 kernel). Well, I'm a little surprised; I didn't expect a new Linux version in such a short (er...) time; the 2.3 kernel series had more than one hundred patches. Currently I have no disk space to compile a full kernel; I will continue to get latest patches, hoping to get 200-250 megs free on the current hard disk to try it.

The Asus battery now doesn't charge more than 2077 mAh. Don't forget that a battery capable of even a single minute of work is one of the best remedies for short black-outs (quite frequent in some zones near here).

Aieeee! the LCD is alive again!

It was just a cabling problem. Or, maybe, it just needed ten days to return perfectly dry. I don't actually know what I did. I only remember I was hardly stretching the two Asus LCD cables (the left one with 16 useful pins, and the right one with 12 pins), and - after reboot - the screen became 1024×768 again. After some trial-and-error in which the 800×600 jerky mode randomly appeared I was able to close in all stuff and remount everything. (if the notebook boots with the full LCD, then there is no fear to get the "jerky mode"!)

Some of extra LCD electronics below the LCD display In the shot below, you can see the internal controlling section of the LCD (the things going under the "Asus" logo on the display; under the five rubber stoppers there are the LCD screws). You can also see the hard disk slot (lower-left), the three Internet keys and their connector (this is just in the center), the Chrontel 7005 TV-signal generator (on the immediate left of the Internet keys flat cable).

Now, some important considerations:

yes, water does damage any computer. But a waterflooded notebook is surprisingly a candidate to get running again: don't panic, just wait a couple of weeks to get it absolutely dry. My giant mistake was battery-charging it when it still had some water into. Water vapourized (by charger heat), and then became again water (when heat got low), and when I turned on the notebook it short-circuited something on the hard disk electronics. This is why I am now without my pretty 20Gb 2.5" hard disk...!

* for some actually weird project issues, the Asus LCD controlling stuff is an 800×600-capable thing with one or two pin (at maximum) used to recognize and drive a 1024×768 LCD display. I tried with all the strangest ModeLines but from 808×606 upto 1024×768, no resolution was accepted at all! It always gave out an unzoomed and buggy 800×600 mode (see shot in the "Catastrophe" section below) at 47KHz/78Hz but no other higher resolution within that range of frequencies. Stretching cables was the solution to get again the 60KHz/78Hz 1024×768 LCD mode...!

Note: an LCD display doesn't actually use frequency timings (they are the equivalent of a CRT classic monitor).

Er... the ethernet device is running again without big problems. I only got a warning message never seen before:
NETDEV WATCHDOG: eth0: transmit timed out
eth0: Transmit timeout, status 00000004 00000000

Oh, no!

Catastrophe! While the notebook was charging, some water flooded it. When I noticed that the power-supply LED was flashing in a weird manner, I disconnected all and wait it to be dry. But it was already done for.Broken Asus A1370D...

The 20Gb hard disk fails. The ethernet port does run from time to time only. The LCD-controller does not "zoom" and - worst thing of all! - does not do 1024×768 anymore! (see the foto below for this "replicated" 800×600 display).

SNIFF, SIGH, SOB! I tried with xfine and xvidtune programs, tried with standard 1024×768 "ModeLine" settings, but I got no results. Even the "vga=791" parameter in lilo.conf (that uses the cleanest BIOS VESA features) fails, and startx gives an extra unwanted "Virtual height" weird problem:
(EE) SIS(0): Virtual height (0) is too small for the hardware (min 128)
(EE) SIS(0): xf86ValidateModes() error
(EE) Screen(s) found, but none have a usable configuration.

Hard disk: Ermanno lent me an old IBM 1.3Gb 2.5" hard disk; I was able to install a SuSE 8.2 with KDE (getting rid of OpenOffice and other big things) and got even 100Mb free on the disk at the end of the installation. But until the "jerky 800×600 LCD mode" is there, it isn't nice to use it...!

Well, after 28 months maybe it's time to buy a new notebook or -maybe- a palmtop.

This page will be updated only in case of very great news. The information supplied below is still valid. SNIFF, SIGH, SOB!

Er... I updated to latest Winischhofer's SiS driver (X11 driver, new XF86Config options, and sisfb source in the kernel): the great news is the sisctrl program (by default was disabled) allowing you to switch resolutions and TV modes (and even TV parameters like contrast, screen position, etc) on the fly!
Sadly, I got no other news: gears still gives about 160fps and TuxRacer still plays 20 to 35fps when running on snow and ice. This means that AGP is still disabled (but this is not Winischhofer's fault: click here).

About DVD playing, video, etc...

MPlayer 0.90 official release is out since a number of weeks. Great software! I must point out some things that people often asks to me.

Some reasons why I like it:

Some reasons why I hate emails asking how to install MPlayer:

  1. currently MPlayer is distributed in source code. You have to recompile it by yourself. And this may be harder than recompiling a Linux kernel...!
  2. some kernel distributions include a full binary package, but not only it's slower than a locally compiled version (because of the CPU-detection routines)... remember that because of legal restrictions, a binary-distributed version will surely have a feature-missing block (the decss part, at least!);
  3. the MPlayer package has everything, but maybe you need other things like lame (for libmp3lame), xvid, etc;
  4. if you happen to get a compile error, it's not immediate to understand what's missing. Some common errors I got under my SuSE 8.2:
    • an header file is missing: the #include complains... Maybe you have the aa package installed but not the aa-devel that contains include files and linkable libraries (did you know about the difference between .so and .la library files? eh, eh...)
    • it says you don't have arts sound driver, but KDE's arts is actually there. Maybe you don't have /opt/kde3/bin in your PATH and the configure script didn't find the artsc-config executable...
    • you get an assembler error somewhere (a strange label was not found, and you find that grep doesn't find anything similar in the mplayer directory tree... eeeaaaah!) and you find that your gcc is a nasty version (anything after gcc 2.95.3 and before gcc 3.0.5 is surely stinking! I use gcc 3.3 but MPlayer guys still say that the best is gcc 2.95.3);
    • you get lots of warning messages and you are scared about them. Well, a "signed/unsigned comparison" warning may or may not be fatal, but hundreds (and hundreds and hundreds) of diverse warnings...!
  5. MPlayer documentation is still quite sparse. Sometimes I had to get useful information from source files, Makefile comments, incomplete documentation texts, etc. The first thing one has to try are the examples. Start from examples... after reading everything!
  6. the encoding/decoding stuff is always a mess. Do you know the true difference between H263 and H263Plus? What does compress better, an interleaved divx5 or an uninterleaved xvid stream? how many minutes you would save using a quality 4 instead of 5? (eh, eh...!); is it better a -framedrop or a -bps option to get synced again audio and video? or was it a braindamaged .AVI file? (don't forget the Windoze ultra-dirty encoders, and the great lamers fooling around...)

YES, YES, YEAAAH! This evening I tried the TV output. A friend told me that on the TV "SVHS" connector in the rear side of the notebook, the third pin of the lower row is a composite signal (whose ground is the circle size jack). Getting these two lines to a "video in" videorecorder port, I was able to get the notebook screen on TV.

Some notes. I used a videorecorder because my old 28" TV set doesn't have a "video in" jack. Any hardware coming with a composite video input jack is good (monitors, projectors, etc).

I did not need to change BIOS setup: I just left there the "Auto" mode for LCD/TV mode. So, after plugging the cable, when rebooting, I got the text mode on the LCD, but the graphics mode (the Xfree graphics mode with the stunning Winischhofer's driver) was shown on TV only (with LCD turned off). Xfree started up only at 800×600 resolution, because it seems that the SiS630 hardware doesn't support 1024×768 resolution on TV.

The notebook's TV signal is quite stable and I saw nice colors. I have only to change a bit lightning and screen-oversizing parameters in the XF86Config file (the external border occupied a quarter of the available TV CRT space).

After that, I wasn't able to switch to/from LCD (and even getting the "mixed" mode, that is LCD and TV in the same moment), as I did some time ago using Fn-F8 and Fn-F9 keys. Maybe I have an quite old version of Winischhofer driver, it's at least two months I don't update it... (will do soon!).

I was even able to record some minutes of VHS tape of my testing: the mouse arrow going here and there, opening and closing menus, windows, popups, the KDE 3.1 running great, some MPEG and AVI file played using mplayer... To record VHS video with audio, I just need another cable, to get the earphone notebook output to the "audio in" videorecorder jack.

There is still one unsolved problem the X/free stuff, at least for this class of notebooks: the AGP support. Currently the video acceleration is used without AGP feature, and this limits the performance to an almost decent speed. (this is not Winischhofer's fault: click here)

Oh... this time the latest SmartLink driver says the internal modem is a "SmartRiser" card... :-) And it still doesn't work on Asus A1370D :-(

slmdm: version 2.7.14 Apr  6 2003 16:09:06 (Smart Link Ltd.).
slmdm: country set is 0x59 (ITALY).
Smart Link AMRMO modem.
amrmo: probe 1039:7013 Silicon Integrated Systems [SiS] 56k Winmodem (Smart Link HAMR5600 compatible) : SiS630/SmartRiser card...
PCI: Found IRQ 5 for device 00:01.6
PCI: Sharing IRQ 5 with 00:01.4
This driver is not compatible with the installed modem codec.

Ouch... Linux 2.4.21 is out since a few days. There are no big news in the kernel (at least for our Asus A1000-series notebooks).

I was able to build it using gcc 3.3-prerelease (it came with SuSE 8.2). There was a large bunch of "comparison between signed and unsigned" warning messages: maybe some kernel variable was changed and... ZOT! Will be fixed in the next release :-) (well, they were just annoying compilation messages; the kernel is running without problems. LINUX IS GREAT!!! Only Linux makes you pronounce always the term "...without problems..."!!).

YES, you can read the "internet keys" of the Asus keyboard from your software! They have non-standard scancodes (0x76, 0x7f, 0x7c at the console level): one just opens the keyboard in "raw" mode (it requires root privileges) and reads them (the Linux kernel driver correctly reports press and release codes). Try with the program "showkey -s". They don't have any repetition. There is only one weird thing: the "press" and "release" statuses are transmitted to the keyboard hardware only when releasing the button (this doesn't seem a software inconsistency but a hardware issue).
While in X/free, you don't need root privileges, because the X/free event engine supports reporting of scancodes while reading for keyboard events. The scancodes are slightly different from the console ones (0x7e, 0x87, 0x84); they also happen to be transmitted (always without repetition) only when the key is released. Let's see a code snippet of getting a KeyPress event from the X/free server.
  Display *dpy;
  Window *dpw;
  XEvent e;
  char buf[10];
  int k;
  if(XCheckTypedWindowEvent(dpy, dpw, KeyPress, &e))
    XLookupString((XKeyEvent*)&e, buf, sizeof(buf), &k, NULL);
    {                         /* not a normal key: check scancode */
        case 0x7e:     /* left (email) special button was pressed */
        case 0x87:     /* center (browser) special button pressed */
        case 0x84:     /* right (quickstart) special button       */

The above code was tested under X/free 4.3 using different keyboard layouts (US, IT). The three "internet keys" lay on a strange circuitry to allow the user to power on the system pressing one of those keys (maybe they discharge a capacitor, or something like that, and a keyboard interrupt happens, and a power-on command if the system was off).

SGRUNT! KWord 1.2.1 of KDE 3.1.1 crashed during its "replaceSelectionCommand" function of "libkotext" library (there was also a weird "KoTextCursor::setIndex: 810 out of range" warning message on the screen), while editing a large document and I don't like crashes in wordprocessing applications. I will get back again using LyX (the only wordprocessing software that NEVER crashed in the last eleven years here)! ;-(

Also, it complains about the size of the footer (I used a little font but "WARNING: frame height is so small no text will fit, adjusting (was: 9.704 is: 11)") and about the page size ("WARNING: The used page layout (CUSTOM) is not supported by KPrinter. Printing in A4."). Ouch... let's wait for another KDE/KOffice release...

I will be out for some days, and I have no time to check this: have a look at the ACPI for Asus SourceForge stuff!

Uh-oh. This page is getting quite big (200k of html). Please wait some weeks, I will split it in an "history" page and a "current status" page. Sigh!

YES, an OCR (optical character recognition) solution for Linux!

The SuSE 8.2 comes with a modified version of the KDE's Kooka OCR program which includes the Kadmos OCR/ICR engine (by reRecognition AG).

I tried it with some images taken by my digital camera (!!!) because I don't have a scanner, and it ran quite well (and quite fast) on machine-printed text and even on hand-written capital letters text!

Kooka doesn't actually need a scanner (if you have one, check its Sane interface support): it just searches for readable images in your local .kde/share/apps/ScanImages/ directory. I used truecolor JPEG images and PNG images (1-bit monochrome and gray-scale) without problems.

Pros: a full-functional KDE-integrated OCR solution for Linux, running at a good speed, requiring minimum operator work.
Cons: OCR scanning does not show anything, not even a progress bar (this seems a Kadmos missing feature), and doesn't remember settings like "handmade/machine text" (this is a missing Kooka feature).

Other OCR engines: gocr is still under development, and clara is not that easy. Kooka+Kadmos is better because requires little operator work. My little experience suggests to use at least 400dpi b/w (1-bit) scanner resolution for a normal book page OCR text scanning; the use of a digital camera was just for quick testing purposes.

Motorola SM56 internal modem: it seems that Rajesh Kizhuveetil patched the Motorola module to get it to work even with gcc 3 under Linux 2.4: I have no time (for now) to check everything, but everyone should have a look to his site!!

3D acceleration running again, under SuSE 8.2!

All the files mentioned above can be downloaded on the world unique giant Winischhofer's Linux SiS630 driver page!!!

I installed SuSE 8.2 Linux distribution, more news soon...

GREAT news!

Yes, the SiS900 is a 10/100 Ethernet card. But for some STRANGE reason, using the SAME inverted RJ45 cable, it does connect to my home desktop PC (running Linux 2.4 with a gorgeous RealTek RTL8029 10/100 card) only in "half-duplex 10Mbit" mode, while connecting at 100Mbps with other 10/100 ethercards, like the one embedded in an IBM Thinkpad A21E, and the one in a Toshiba Satellite 404, and even the one in an Acer Aspire 1400 (er... please wait some days, I'm not sure of the model numbers of these three notebooks; I will correct them as soon as I can).

On the Satellite I transferred 7300 Mb in less than 30 minutes (instead of more than two hours): think that the latter was a WinXP based machine, and this means slowness, and that was while playing audio/video on the two notebooks: and still we got four times faster than a full-speed 10 Mbit connection...!

When a CD is "burned" with multisession feature, the CD-ROM head of our Asus notebook moves three times before trying the CD speed. I got one of those CDs, burnt at 16× speed, and I copied from it 425 Mb onto my hard-disk in 2'55" (2.4 Mb/sec, that's exactly 16×).


I had to get out the cpu fan: sometimes it starts, sometimes it doesn't. It's a DC 5V 0.5W Sunon GC054006VH-8 fan with 11 (eleven!) 4mm-high blades, and its base has space for three screws. I hope I will find a new one...

Yes, I saw the VGA-to-TV chip inside, it's a Chrontel 7005C-T, and it's surface-mounted just near its TV connector.

Other sad news: the battery life is still shortening. Now it seems that the charging stops at about 2230-2340 mAh (instead of 2816 mAh, that was the "official" last capacity) and then stops and restarts trying to charge every few seconds... I will have to be present every time I will have to charge it. I think that in two years I'll surely charged more than 300 times, almost always from fully discharged to fully charged.

Uh-oh. I tried Linux 2.5.66, and saw some interesting news.

Sad news: the SmartLink driver was reported to work from another user using a SIL38-based modem. But mine still doesn't work. The smlink driver v2.7.10 still reports:

slmdm: version 2.7.10 Feb 4 2003 15:18:20 (Smart Link Ltd.).
slmdm: country set is 0x59 (ITALY).
Smart Link AMRMO modem.
amrmo: probe 1039:7013 Silicon Integrated Systems [SiS] 56k Winmodem (Smart Link HAMR5600 compatible) : SiS630 card...
PCI: Found IRQ 5 for device 00:01.6
PCI: Sharing IRQ 5 with 00:01.4
This driver is not compatible with the installed modem codec.
Its modules still had to be loaded BEFORE loading the sound modules, and still reset the master volume to highest value (so I had to issue an aumix -L soon after loading sound modules in the boot scripts). Yes, this is the ugliest part of the story: at least the Asus A1370 family uses an uncommon (generally it means "cheaper") version of the Motorola chipset. It was seen as "SIL38/unknown" by the AC97 driver, but this is still "not compatible" with current 2.7.10 driver...! SNIFF, SOB!

Nice stuff: we worked hard on a local network, transferring gigabytes of stuff between two Linux machines and two W98 machines. While a Linux box was connected via "ssh -X" to the other, its Samba server worked right on the same ethernet link. Also, while transferring lots of data via the ssh, the other Linux box was connected to the Internet letting the other computers navigate and fetch mail. Not to mention printing and sharing anything other...! LINUX is STUNNING!!

AIEEEEE!!!! I printed (text and graphics) on an HP Deskjet 350 portable printer using the infrared port supported since lots of Linux kernel versions! (the one I'm using now is 2.4.20)!
I printed a Postscript file (generated by LyX) converted to HP/PCL printer language using the gs command (ghostscript) and its driver "hpdj340" (gs 6.53 doesn't have direct support for HP DeskJet 350). One just has to send data to the /dev/irlpt0 device, with cat or similar command (YES, in these weeks I'm not using the CUPS things, because I am experimenting some weird things... if you have CUPS, you don't need this strange stuff!). To get printed the Postscript file on the printer, I used one single command (ghostscript, no interactive, 300dpi resolution, batch printing):
gs -dNOPAUSE -r300 -dBATCH -sOutputFile=/dev/irlpt0 -sDEVICE=hpdj340
LINUX is great!
Also, I found connected my Nokia 7110 phone on the same port (and it could print without problems): look at the output of the irdadump command (I cut out some useless stuff):

s=1 DeskJet 350 hint=0800 [ Printer ] (27)
s=3 Nokia 7110 hint=b125 [ PnP Modem Fax Telephony IrCOMM IrOBEX ] (27)
s=* notebook hint=c404 [ Computer LAN Access IrCOMM ] (22)

OUCH!!! I had to open the keyboard cover to get access The cpu cooler fan to the cpu cooler fan: it didn't start anymore. Pushing it on front of its cover, I got more noise (the fan "rocks and rolls" and sometimes touches its cover) but -at least- it runs again. SIGH! I have to find out some other fan (a 5V high-performance littlest-size fan!) if I don't want to get all people looking at me when I power on the notebook!! :-) (er... a young nun asked me "huh? what's that FARM TRACTOR you turned on?"!!!!) :-)

The battery is slowly degradating. Now I got auto-power-off's already at 350 mAh.

Uh-oh. I found my battery charge "101% full". The reported capacity was 2816 mAh, but the charge level was 2848 mAh. But the bad news is that when the battery charge goes under 10% (that is, under 282 mAh), it simply shuts down immediately. Well, the battery is getting old... but, at least, I can still work safely for two and a half hours (!) using a not-so-resource-hungry wordprocessor.

Sniff, sob! The SmartLink SM56 softmodem driver release 2.7.7, even with soundmodules unloaded, still says "unknown codec"... :~( The PCtel softmodem driver release 0.9.6 doesn't even accept "AT" commands in its weird ttyS15 driver... :~(

I erased lots of non-KDE applications from my hard disk (mainly Gnome apps). I found that KDE programs load quickly because they "fork" from a KDE application (kdeinit) and not from a huge (or libraryless) task. KDE is simply better than Gnome (which, IMHO, requires more memory, more disk, more resources, and -sigh!- has more bugs).

BIG NEWS!!! I was able to play some DVD's using the GREAT mplayer software! (it requires an extra decss library to read some commercial DVD's). More news soon! Click here for... Capobanda!!

Sometimes - I still don't understand under which conditions it happens - the notebook starts up at 600 MHz instead of 750 MHz. I think there is a software command to get this done and, at boot, something triggers it. The slowdown happens even after a normal shutdown procedure (so I can even suspect a BIOS bug); rebooting solves the problem, setting again the processor to 750 MHz speed.
If you don't grep the "detected" string from dmesg (which now gave me a strange "Detected 598.190 MHz processor.") maybe you will never notice it (a TRUE great difference would be from an 80-cable hard disk UDMA5 interface; but I think this is not possible on this hardware). Oh, well: at 600 MHz the processor does not get quickly hot: I tried to run a cpu-intensive job, it consumed 10 minutes of continuous 100% cpu, and the temperature just went from 31º degrees to 66º degrees (so the cpu fan didn't start!).

New hardware tested:
hde: CD-220EA, ATAPI CD/DVD-ROM drive
ide2 at 0x100-0x107,0x10e on irq 10
hde: ATAPI 20X CD-ROM drive, 128kB Cache
ide_cs: hde: Vcc = 5.0, Vpp = 12.0
This is an Archos branded ATAPI CD drive with PCMCIA card adapter. In a few seconds it was recognized and worked without problems. I tried to transfer some data and got 182 Mb transferred in 165 sec (that is 1102 kb/sec, that's about 7× instead of the 20× claimed by the drive, but this is due to PCMCIA interface limits. Well, on an old 486SLC-based notebook I reported about 20 kb/sec of transfer rate under Win31...!).
Notice the differences in the Vcc/Vpp fields for the latest peripherals tested: CD/ROM: 5V and 12V; hard disk 340Mb: 5V only; Compactflash card: 3.3V only!

Other hardware tested: an old (VERY old) 3-button serial mouse. Yes, I forgot my Asus USB mouse in my office, and the only extra mouse I found at home was an Artec serial mouse, claiming "made in China" in 1992 (?). Well, I tried the mouse configuration utility of SuSE 8 sax2 setup program: it didn't work on any mouse type. Then I just added these lines to the /etc/XF86Config:
Section "InputDevice"
    Identifier  "SERIALmouse"
    Driver      "mouse"
    Option      "Name"             "Artec Serial Mouse"
    Option      "Vendor"           "Artec"
    Option      "InputFashion"     "Mouse"
    Option      "Protocol"         "Microsoft"
    Option      "Device"           "/dev/ttyS0"
    Option      "Emulate3Buttons"  "off"
And then added this line in the ServerLayout section:
    InputDevice "SERIALmouse"   "AlwaysCore"

This means that I can use now three pointer (mouse's): the internal touchpad, the external USB mouse and the external serial mouse, and they all run together (because of the "AlwaysCore" option of the latter: this is because I don't always connect them to the notebook).

Linux 2.4.20 has been released!

The "Local APIC disabled by BIOS -- reenabling. Could not enable APIC!" problem is still there (sigh!): this means that there weren't great changes in ACPI section (do you remember? this local-APIC problem is actually an ACPI driver problem. But, so far, it was harmless).

Again, we see:
PCI: BIOS reporting unknown device 00:50
PCI: Device 00:51 not found by BIOS

The AC97 modem warning now appears as:
ac97_codec: AC97 Modem codec, id: SIL38(Unknown)

The dmesg output doesn't report any other problem.

PHEW!! Here are the dates of the last Linux kernel releases:

OUCH! The 2.7.5 version of the Smart Link HAMR5600 driver features the same problems! :-(

OUCH! I tried the Smart Link HAMR5600 Linux driver (slmdm-2.7.4_i386) got from SmartLink official site, to use the notebook's internal modem. In the dmesg messages I got:

slmdm: version 2.7.4_i386 Oct 17 2002 12:42:29 (Smart Link Ltd.).
slmdm: country set is 0x59 (ITALY).
Smart Link AMRMO modem.
amrmo: probe 1039:7013 Silicon Integrated Systems [SiS] 56k Winmodem (Smart Link HAMR5600 compatible) : SiS630 card...
This driver is not compatible with the installed modem codec.
Please contact your modem provider for support.(6)
In fact it could not connect at any speed to my favourite providers. The terminal/command mode is OK, but that is pure "ATZ OK ..." emulation that does not use the softmodem hardware.
It can place a call and even report "NO ANSWER" (without any audio feedback...! I don't know if it's retraining or if it's waiting for the dial tone!), but it did never connect. SIGH!
Someone sent me a success report with this driver, but he had a Celeron-based A1000-series Asus notebook (does this mean that the Pentium3-based, like mine, have a different modem codec?). SNIFF, SOB!

Nice little strange thing: the CPU cooling starts running the first time at 70ºC; successive times, starts at 75ºC (when using battery it starts respectively at 80ºC and 85ºC).

SIGH!! The CPU cooler fan is very noisy (dust and other stuff in; uncentered axis; 19 months lifetime with at least 200-250 hours running); it's not that happy hearing it starting when the CPU temperature is too high. (To be honest, I often worked an entire day without getting the fan run: Linux kernel and programs actually use the minimum required CPU power; the 200-250 hours of above are only the "dirty hard works" like rendering POVray images, kernel compiling, etc).
The two-year warranty is not expired, but I do not want to send my precious notebook to people that MAYBE will change the fan for free but will keep there the notebook for MONTHS...!
So, what to do in the meantime, until I will get sufficient courage to open the notebook case and try to repair by myself? I wrote a little nice program that, given a process-id of a very-CPU-hungry program, stops it when the CPU temperature is high (before the fan starts) and restarts it when the CPU temperature is decently low. In this mode you use the full CPU power only 15%-30% of time (that is, of 100 seconds the program will have run only for 15-30 seconds) using a "natural cooling".

I just had my first problem with KMail: using the "dangerous" HTML mode for showing messages, KMail 1.4 (of KDE 3.0.0) gets a SIGSEGV when trying to show a regular spam message from one of the most famous Internet providers (famous for its unrequested ads...!). Disabling HTML mode helped to solve the problem.

Good news: KDE 3.1 is (almost) out!. Bad news: in the forthcoming SuSE 8.1 there is still KDE 3.0.3... :-(

SGRUNT! Some pages of "Linux Weekly News" site are now accessible only to registered users. Time to switch to another news source for Linux...!

Battery life is shortening... After a full bios battery refresh (because of strange charge levels, like 2789 mAh), I got 2816 mAh of charge. Yep, this is quite far from the "more than 3050 mAh" that I got one year ago, when the battery was only six months (and very far from the 3500 mAh of a new unused battery)...
Also, even if not used (while the AC adapter was always connected), every 12-24 hours the battery charge looses some milliamps. Example: without ever disconnecting the AC adapter or battery, on Sunday evening the /proc/acpi reported it was 2816 mAh; on Tuesday I got 2805; Wednesday morning 2799; evening 2794; Thursday morning 2789, evening 2783...!
The batteries like this one are guaranteed for at least 200-300 cycles of charge (sometimes 500). I am absolutely sure I didn't reach the 200 cycles (perhaps I'm near), and less than 10 times I charged it when it wasn't fully discharged (this means almost the cleanest life of the battery, avoiding any remote "memory effect" even if this is a NiMH). So I can think that the "shortening" of above is due only to its age (19 months).

Ouch! The Linux 2.5.39 kernel fails compiling the 8250 section... :-)

Ouch! I tried to compile the Linux 2.5.36 kernel, but the infrared driver is quite broken (well, at least this means that Dag Brattli is working hard on it...).

I just found that the Asus site links are all broken... bleah!!

The KDE3 MATRIX screensaver seems slow? Have a try with this (you need the cmatrix package, usually found in many Linux distributions like SuSE 8):
xterm +sb +j -bg black -fg green -T MATRIX -geometry 180x53 -e cmatrix

OUCH! the SuSE 8 doesn't have the glx module in the Xfree lib/modules directory (neither anywhere!). This means that we CANNOT use 3D-acceleration by simply installing Winischhofer's files... More news soon!

I tried to compile Linux kernel 2.5.31 but I got a bunch of errors (even some "#error" directives that were there just to recall to the author of the i2o module... to update the documentation files! sigh!). Will wait for one of next releases...!

Well, it seems that even with a "rich" distribution like SuSE, I am still using only GNU-GPL licensed software, because it's simply safer, better supported, more often updated or patched or debugged, and it surely comes at no cost and without any commercial advertisement compiled in!

Oh, I see again the "hard disk stop bug". I do not want the hard disk to sleep; but even issuing "hdparm -S0 /dev/hda", I sometimes happen to see it stopped... I don't know if this is a kernel bug or a SuSE bug. As of now, SuSE has not installed any acpid daemon.... :-(

Linux 2.4.19 patch: there were some errors patching the 2.4.18 kernel... because this time the source directory to patch, has to be called "linux-2.4.18" instead of "linux"!

We still have (OUCH!) the "AC97 Modem codec, id: 0x5349:0x4c26 (Unknown)" warning... it means that Ollie at SiS doesn't read my emails :-(

Some of the i2o modules stuff fails (the i2o_pci is "not found"; this is the first time I see this warning); also, the i2o_lan exits with this weird error text: "parameter rx_copybreak has max < min!"

I got also this warning: "Advanced speculative caching feature not present".

Sigh, the local APIC warning ("Local APIC disabled by BIOS -- reenabling. ...Could not enable APIC!") is still here! :-( Also, we still see that stinking "unknown device 00:50 (PCI BIOS), device 00:51 not found"

Another stooopid warning: the soundmodem stuff has... "no license" :-) I wrote to the author, but noticed also that some of the soundmodem source files are dated... 1997! :-)

SiS acceleration: again, we still have to wait Thomas work to be integrated in X/free 4.3 and Linux kernel.


My friend Ermanno convinced me to install the SuSE Linux 8.0. "It's great with security", he said. Oh, well. My system is already "extremely unpromiscuous", I answered. But since a lot of time I had already decided to get rid of the Mandrake and so I started reinstalling (ouch!)...

The installation phase seemed quite nice but, at the start of the installation of the second CD (second out of seven CDs), that I wasn't anymore able to use the mouse (ouch!!!). Later I will find that the SuSE's standard kernel does not like very much my Asus USB mouse...! (I figured Mandrake's guys laughing at me: since Ye Goode Anciente Mandrake 7.2 distribution, I got USB mouse running at any moment and without any problem!).

Also, I had problems even configuring a CUPS-networked printer (and this also was a good point for Mandrake). This means that I'll have to reconfigure it in some days...

Another little ugly thing: the SuSE Linux kernel is configured for APM (it's a 2.4.18 kernel and they still don't use the ACPI!), and I simply need ACPI to check CPU temperature and battery exact status...!

Well, even this time I have to restart from my "made-to-measure" full-optional no-stupid-stuff-included highly-personalized kernel: and now I can use again touchpad and USB-mouse without reconfiguring X/free, I see again the exact CPU temperature and battery status, and I got rid of the lots of strange things that one sees in every Linux distribution that has the need to have one kernel for all kinds of computers...!

Ermanno, I always told you that a "made-to-measure" kernel will always save us a lot of problems!

Well, I do a full-optional full-hillbilly everything-available installation and find some programs that weren't included in the Mandrake because of their "not-GNU" license (oh well, this does not mean that non GNU software is good...!)

I didn't find installed my nice locate program; I had to search it in the YaST2 SuSE package manager, but instead of the slocate (coming with Mandrake) I found the findutils, that has a little inconsistency: it says that there is a file there even if you had already deleted it! The slocate (which I will find and upgrade soon) is better: if a file does not exist anymore (or if you don't have enough privilege to read it), then it will not appear in locate's output. GRRR!!!

Another problem: I had to look around to find where my personal inits (about pcmcia, IrDA, etc) and other stuff had to be placed (in the Mandrake I had to place them at the end of /etc/rc.d/rc.local; now, the SuSE requires them at the end of /etc/init.d/boot.local).

Mandrake 8.2 was out about in the same days of the SuSE 8.0 (feb/mar 2002), but SuSE features the KDE 3.0; at a first glance, one does not notice great difference between the two versions of KDE (except a number of apparently small improvements), but this is a first glance...

Another note: for some weird reason, cardmgr (for pcmcia/pccard handling) complained at startup, so I had to specify on its command-line where to find kernel modules, so I now load it with:
device "ide-cs"
   class "ide" module "ide-cs"
In fact a problem arises when in some strange places the "ide_cs" module is instead referred as "ide-cs". This safe workaround (oh, we should call it "dirty patch") solves the problem: adding the two above lines, and changing from bind "ide_cs" to bind "ide-cs" in the "ATA/IDE fixed disk" section (coming after about 1700 lines of the same /etc/pcmcia/config file), I was able to use without any other problem as external pcmcia hard-disk (seen as /dev/hde) the compactflash card (on the pcmcia adapter) of the digital camera Kodak DC210-Plus.

IrDA: the irda rpm package was not installed by default, but after installation the Nokia 7110 correctly receives/sends SMS's and data (yep, it's because of the 2.4.18 kernel IrDA support compiled in; I am still working with my "made-to-measure" kernel!).

In /etc/fstab I needed also to add the noatime option on the line of the main partition (that I still have reiserfs formatted, in defiance to SuSE installer: I had more than 11 gigs of stuff there, and I didn't want to backup/restore all that).
The noatime option preserves the atime field update in the inodes of the files on that reiserfs partition, saving a number of write operations (quite useless, if you don't ever go look at a "last read access time" of a file... did you know that any serious Unix system, like the best one [Linux], does update the atime field every time a process goes reading it?), so saving a bit of battery power.

Other problems: here I found also Mozilla 0.9.8, but sadly, differently to a number of other great programs, it didn't support Italian language like Mandrake's one, so I could not use the same configuration directory because a lot of menu items and window descriptions simply disappear when you use another language setup (ouch!!). Well, I did need to reconfigure everything from scratch (shame on you, Mozilla programmers! your software is too "language dependant"... A "sent messages folder" file should be called always "Sent", even in Italian language! instead, I got a file called Sent and another file called Inviati or something like that!). GRRR!

Little note added here: well, I left out Mozilla and its slooooowness (its name, here in Italy, makes one think to something big and SLOW): is it nice for a P3/750 machine to wait entire seconds to open and close window messages? I left it to use KDE's KMail 1.4 that is blindingly fast when compared to Mozilla (and it recycled without problems the old mailbox files, and it has a number of other nice features, and it loads in TWO seconds! KDE is simply stunning!).

Other problems: some X/free fonts are missing (the 100dpi fonts were not installed from SuSE installer). One of my old X software needed one of those fonts and, for the very first time in some years, it complained! (well, why not to install the 100dpi fonts? I remember that somewhere I specified that the Asus 13.3" 1024×768 LCD screen is exactly 96dpi!)

The security paranoia checks of the SuSE reached one of the highest insanity level with the ssh that now does not support by default the X/free socket data transfers (do you remember that I could watch TV via ethernet using the xawtv -remote command, and I could start kppp of the home desktop computer to go to the Internet via local area network? now, by SuSE default, it's not possible).
Oops, problem solved: you have to specify manually the "-X" flag on the command line of ssh to get X/free remote calls passed...!

Another bug, this time not a SuSE bug: for some strange reason, I cannot get the "big cursor" in the KDE, even if I requested it from its control center (kcontrol).

Adding other TrueTupe fonts: it was not so simple like in Mandrake distribution (that featured a little utility program to do it), but in the susehelp documentation there was documented how to do it. I was able to add lots of TrueType fonts here, but it's quite strange for a such great distribution to not add in the yast2 package a simple module to add TTF fonts without using sed and mkfontdir manually!

Another problem: after installing the Italian keyboard layout, the KDE's keyboard applet (to get different keyboard mappings) becomes useless (GRRRRRRR!!!!). I need to work with three different keyboard layouts: italian, US standard and US with dead keys (my preferred layout)! Even issuing the loadkeys command while in root mode was useless.

Ouch! SuSE does not support the nice autologin feature found in the Mandrake and other distributions (that is: start KDE with a fixed username, without having to login and wait some time to get KDE ready). The autologin feature is a must for a computer used surely by only one person!
In the /etc/rc.d/ directory I found that the kdm (KDE display manager, the one that does the graphic login/password dialog) is launched from one of the rc.d scripts and not from /etc/inittab (like in Mandrake and RedHat distributions), so I changed the runlevel 4 to start an autologin with X/free and KDE (yes, I had to get an RPM package not found in the seven SuSE CDs: the autologin RPM package).

So, here it's how you can add the autologin feature to your SuSE distribution:

First of all, install the autologin RPM package (from two RedHat programmers; you may have to search something like "autologin-1.0.0.i586.rpm"). If you get the Mandrake's one, you will find it requires an initscripts package (that we mustn't install because in our SuSE we do want only SuSE initialization scripts!), so you'll have to install without dependencies check, using the command "rpm -ivh --nodeps autologin*rpm".

Now let's modify the /etc/inittab file:

The "respawn" is needed to restart X/free if the X-server was stopped (like pressing ctrl-alt-backspace). If you instead place "wait", then when the X-server is closed you will get a normal text-mode system login.

Now you need to get sure that the /etc/rc.d/rc4.d contains exactly the same things of /etc/rc.d/rc3.d (the only difference between these two runlevels in respect of runlevel 5 will be the autologin KDE startup). I used these commands:
cd /etc/rc.d
rm rc4.d/*
cp -a rc3.d/* rc4.d/

Then, the file /etc/sysconfig/autologin must be owned from root and have permissions 0600 ("rw------"); if not, the autologin will not run (it's a safety check) and won't even complain. This text file must contain three lines:

"turzolo" has to be changed with the username you use normally to log in (oh, you WON'T EVER use "root" to log in and work, because in this case all the security issues of Linux will be unsure, SuSE or not SuSE!).

Now you can check if it runs, issuing from root the "init 4" command. If you see X/free and KDE restart without problems, then you can modify in the /etc/inittab the default runlevel to 4 (it was 5, that is "standard login in X/free mode"). The line of the runlevel has to be: "id:4:initdefault:".

Another stoooooooopid problem of all the Linux distributions: you install whatever the installation program suggests, and after a number of months (years) you find you have much more software than you did use and know! this happens because you used only "clickable" programs from the KDE (or Gnome) menus. And, ugliest, in those menus you have lots of stoooooooopid programs that you ("The Normal Everyday User(TM)") will never use!
The SuSE also has this ugly "problem". You have to check often the YaST2 package manager at least to know what software is installed onto your disk...!

Another curious case: I installed also the Borland Kylix coming with the SuSE and, after issuing the "startkylix" command, I found that it requires a key to run. AUUUGH! The Mandrake distribution (and almost all other Linux distribution) come with software ready to use! So I think (after years and years of public domain, freeware and mostly GNU-licensed software) that if a program requires to fill a form with my personal data to get a key for "evaluating" it for a number of days, it really isn't worth even a sector of my extra-large hard disk!

Another software to disinstall immediately after SuSE installation is Opera: it's not really faster than other browsers (like Galeon or even KDE's Konqueror), and its "non-commercial" version sucks some of the useful bandwidth to... place commercials in the north-east section of the screen (ouch!!!). GNU-licensed software are a GREAT hit in user-friendliness!!!

I didn't find neither lynx nor links in the SuSE installation, but a w3m textmode browser that can even... place some graphics in the textmode terminal of X/free! But the first time I tried it on an HTTP page, it showed two commercial ads (ouch!) and so I decided to install again Ye Olde Goode lynx text browser, that surely doesn't have such "features"!!!

Last note (for now): the full boot (from turning on the computer to the definitively ready-to-use KDE system) needs just 1'40", that is about 40 (forty) seconds less than the Mandrake startup. YEAH!!

The only last thing I have to work on now is the graphics acceleration X/free driver (from Thomasite's site). More news soon!

After six months from the release of the Linux 2.4.18 kernel, we now have (yeeeeeeah!) a 4.5 Mb bzip2-compressed Linux 2.4.19 kernel patch! :-)

Ouch! Since five months and we still have Linux 2.4.18 kernel (the 2.4.19 is not out yet!). In the last weeks, the only nice news is that I got xawtv running on the local Ethernet: I was able to watch TV (even at a low resolution, because of the slow Asus network card: 10Mbps) using the xawtv -remote command (the desktop PC has a TV-card and video4linux support in the kernel): I connected via ssh and started xawtv and... WOW!!

OOPS! While switching on the notebook, I timed the boot stage:

That is: the entire Linux+KDE+everything requires two minutes to boot (2'20" for full boot completion); if I use a /sbin/init created from me, my system would load in only 26 seconds.
Well, times may vary on different notebook setups. If you have 128 Mb RAM or less, and you don't optimize the IDE-disk parameters, you may find it takes quite longer. If you disable a number of useless services (PostgreSQL, etc) you may get it in less time. Slow? Blame Mandrake boot scheme. I remember that someone proposed a "tree" of services loaded concurrently (and not one after another), but don't remember where did I read it.

Slow? Well, I don't care too much about it. I simply added a lilo.conf entry with: append="parport=0x378,7\ devfs=mount\ 1"
The "1" there is an argument for /sbin/init telling it to start at runlevel 1 and not at the default runlevel. The runlevel 1 is "single user mode" (only one terminal text window, and very little services loaded), so it completes the boot stage in less than a minute.

YES, someone told me he followed Thomasite instructions and got full graphics acceleration on an Asus A1370D under RedHat 7.2: one has to use the new SiS-framebuffer driver (the one presented in Thomasite site), which is part of the 2.4.18-ac4 kernels and 2.4.19-pre2 kernels (but NOT in the 2.4.18 official kernel, that contains the OLD driver featuring the LCD-bug!).

SUCCESS! I still have some problems with the AGP module, but I just noticed the gears software of the Mesa-demo package running at double speed (21 fps in full-screen mode and 161 fps in standard mode).

Using 32 Mb of video-RAM:

The problem may be in the continuous use of the DRI/DRM kernel functions without using the AGP bus (look above for details; I hope to solve this soon), so the kernel hangs under a 100%+ cpu usage... The problem wasn't solved disabling the GenericSiS/AGP option in the AGP section of the kernel.

News: he told me also that the Motorola SM56-softmodem should be compatible with the Smart Link HAMR-56k. More info soon!

UGLY KDE 2.x BUG: if a power-down happens during the blue screen boot phase (when you see "loading panel... loading desktop..." messages), KDE ALMOST SURELY hangs and refuses to load from that time on: you cannot use it, and you get VERY STRANGE ERRORS, like "could not find the application/octet-stream MIME type", "could not start the process", "could not create I/O slave file: unknown protocol «file»"...

This nasty bug made me remember The Ancient Times, when EVERY computer user was SUPERSTITIOUS and even the hardest computer expert had to CROSS ITS FINGERS while Windogs was booting. From a number of years, using Linux, I just FORGOT that a boot session was (by definition) a CRITICAL SESSION...!

The first time the above bug happened here, I just pressed Ctrl-Alt-Backspace and got the X-server and KDE 2.x running again without problems. The second time it didn't suffice, and then I needed to get a backup of my home .kde directory (I ALWAYS do a WEEKLY BACKUP: and YOU?!). Restoring the .kde directory only, I got all running OK again.

Now, that I got this ugly hang for the third time in three months, I just feel like a hungry piranha, because not even the .kde backup was sufficient. Will I have to REINSTALL again the entire Mandrake distribution because of a dirty few-bytes-file?!? I am SERIOUSLY considering NOT TO UPGRADE Mandrake, but to SWITCH BACK TO RedHat. Sadly, when you have one of this problems, it's always too far from next distribution release...!

Well, again I DID IT: you need to restart KDE at least once WITHOUT the «damaged» .kde directory. After loading is complete, you can then replace the «damaged» .kde with the original (backupped) one.

Well, I hope this will NOT be a KDE 3.0 bug...

Well, the Mandrake 8.2 update was a real pain. It seems that rpm failed to install correctly all dependencies (the urpmi does all checks), so I get, e.g., a segfault when using pnmscale, a bunch of warnings when using kppp, etc. GRRRR!

Uh-oh. I got a "Initializing CPU#0 - Detected 750.946 MHz processor." in the boot-up messages (after an year of 747.737 MHz or 747.739 MHz). Nice :-)

Yes, GALEON is THE web-browser for Linux! Fast, reliable, and even pretty (though it's Gnome)!

I did "upgrade" to Mandrake 8.2, but *aaargh!*, it was a true pain. It seems that the "LCD-bug" has disappeared with the bundled X/Free 4.2, but I cannot guarantee it's always true. X/Free 4.2 still doesn't support acceleration (even loading DRI support and enabling kernel support), so we still have to follow Thomasite instructions...! Some first impressions:

The one-year-warranty period has expired. Sigh! This is the current defect list:

The one-year-warranty period has expired. Sigh!

After another hang of the netatalk beta package, I upgraded it to version 1.5.2. I will test it in some weeks.

Yes, there is a bug in the IrDA section of Linux 2.4.18 kernel, in the input/output section of the IrTTY module. After getting again the strange initialization problem of the Nokia 7110 (which firmware has its strange little bugs), I found that the simple command "cat < /dev/ircomm0" was not able to flush internal Linux IrDA stack buffers: I got an endless sequence of OKs from the modem (every time cat tried to read, got always the same string). Sadly, this bug is quite hard to reproduce, so I cannot send a bug-report to the IrDA stack wizard... :-(

Another strange bug: sometimes the ACPI kernel support cannot initialize the thermal section. I placed this check-script in the .kde/Autostart directory:
if dmesg | grep "method failed"
  gmessage "`dmesg | grep method\ failed`"
Well, if you start KDE and see a message box containing something like "Ps_execute: method failed - \_TZ_.THRM._AC0 (cbb2cde8)" (failed initialization of a Thermal Zone related method), then you should reboot the machine.
I reported only once this problem because in normal work I don't need CPU-hungry tasks. This evening I found that after some minutes of hard ray-tracing work, the internal fan was still stopped. There was a temperature of 89ºC. When it gets to 105ºC then there happens a sort of "safe mode" (it seems that the cpu is slowed down to a few MHz).
This bug should disappear when the full ACPI support will be definitively completed and debugged.

Linux 2.4.18 is out (phew!); video, sound and IrDA updates and fixups. A few little problems is still there, but I can still mark very stable this kernel.

There is still a lot of people out of there using the EXT2 filesystem for partitions larger than 300-400 Mb... HEY!!! you need to read here to know why you should use ReiserFS instead of Ext2! The reiserfs is simply stunning:

No news is good news? We are still waiting for Linux 2.4.18 (ouch!), but so far, using 2.4.17 since two months, we are still happy... :-) Thomas Winischhofer (notice, the site address has been updated) is working hard on SiS630 accelerated support: now one can use the notebook as dual-headed, that is, one full image on LCD, and another screen on the external VGA connector (even if not the same resolution/mode!), also in "Xinerama" mode (the two screens can be "virtually one"). I didn't test it yet, because in these weeks I'm working hard on a paper using LyX ("LyX rulezzz!"), and I don't want to experiment new features...

Stooooooopid bug: if you issue a command like "eject song.mp3", the stoooooopid eject 2.0.10 program doesn't understand that the MP3 file is not a cdrom device but a stooooooopid MP3 file, and locks up the machine for a bunch of seconds, letting your face become stoooooooopid before you understand that it ain't a system bug, but a stoooooopid bug! I wrote to the author to get it fixed.

XFree 4.2 is out, but they promised the 4.3 version ready for May 2002. Its support for SiS630 is not much different than the 4.1 release, so we still have to follow the Winischhofer's instructions to get graphics acceleration running. I still haven't tried it.

Stunning. Someone worked on a hand-powered Linux webserver. Yes, without batteries and without AC adapters. Just your hand. Turn the lever, let it generate electricity (5V, while the internal uCdimm Linux board uses only 1.15W at 3.3V) and have it run! Yes, it's a muscle power supply, it's "the first webserver you can bring with you and use in a desert".

Well, the first two strange things happened. My notebook warranty expires in two months, and this is the first time that I care about it...!

This is my current BIOS setup (press F2 during hard boot to get to the BIOS setup):

Floppy disk drive: 1.44Mb






Power management:

Boot sequence:

  1. ATAPI CD/ROM drive
  2. IDE drive
  3. Legacy Floppy

An updated version of the acpid daemon is here on SourceForge. I do not need anymore to suggest you to patch the old acpid executable.

Some strange guys filed against our friend Werner Heuser because his site name "MobiliX" (Mobile devices and Linux) ...simply resembles Obelix.
Yes, it's true. And it's sad. More news on his site and on "Linux Weekly Edition" news site.

BAD NEWS from Motorola about the softmodem or our notebook:

Thank you for contacting us at Motorola SM56 Support. All development for the Motorola SM56 software modems was discontinued in July of 2001. Motorola currently has no plans on releasing the driver code to the public domain. Thank you for your inquiry.

Linux & SiS630: an Austrian wizard got SiS630 acceleration really running, with stunning speeds for Quake3, Descent3, TuxRacer 0.61 and other games. Great. Have a look there!

VCD: yes, I had some fun watching a VideoCD (huh? divx? boh? I don't know the standard naming for it... Xine calls it "VCD": a CD containing 650 megs of animation/video).
Yes, some time ago I had problems with Xine, but it was because I forgot to install the ogg-devel and vorbis-devel packages (that Mandrake doesn't install by default). With at least these packages installed, the Xine compiles also the sound support.
The film? Something about a red planet... English language, and Greek subtitles. Thanks to my friend Panagiotis (he also likes Ozzy!).

The rsync solution for local (and incremental!) backup is simply GREAT. I backup my entire notebook home directory (fourteen gigabytes of stuff) with just one command:
rsync -v -ae ssh --delete --delete-after --stats /home/alfmar alfmar@desktoppc:/home/
Yes: the ssh transport (safe-shell), the verbose file output ("-v --stats") and the superfluous file deletion after completing the backup. Any networked machine reachable via ssh is good (LAN, internet, etc).

About SiS direct rendering module: the dmesg reports SiS/DRM support:

More info coming soon!

Wow! TRUE local internetworking, using iptables feature of Linux 2.4! Well, I won't write here how I did (I just followed instructions from the great, easy and detailed iptables HOWTO), and I got this:

Linux 2.4.17: adds some interesting stuff (SiS direct rendering manager), updating also ACPI, IrDA, PCMCIA, parport sections.

DVD playback: YES, I got a decent DVD film. I used xmovie (coming with Mandrake 8.1), and:

DVD and DivX problems that I hope will disappear in next weeks:

Linux 2.5.xx: since I need a stable system, I won't use it.

Linux 2.4.16 is out, but without great changes (the patch is less than six kilobytes): it clears an horrible bug in the filesystems section (I didn't find any problem, but I updated to 2.4.16).

ARRRGH: the "Linux AppleTalk system hang" is not a netatalk's bug! I started an extremely-clean system (Linux 2.4.15 starting with /bin/bash instead of /sbin/init) and I printed ten pages without any problem! I think that xinetd (or something other of network-related software) may be really buggy about Linux AppleTalk/EtherTalk support...

Keyboard typing speed: yes, I wrote a 21,000+ (over twenty-one thousand!) characters text in almost 47 (forty-seven) minutes, including some corrections and inserting special characters. This means that the keyboard of this notebook isn't that hard and slow as someone could think. Well, this means also that I'm a reasonably fast typist.

IrDA port & Nokia 7110: I found that disabling "Fast RRs" in the kernel infrared-port options, the IrDA connection is easier (the irattach doesn't like too much that option... hm!?).

Wordprocessing: work with LyX with minimal pain! You can use wvWare 0.6.5 (found in the Mandrake 8.1 distribution) to convert a WinWord 9 .DOC file to CleanLatex, and then import this to LyX: it runs great! Well, you still have to work on it, but just to clear some "sloppy paragraphs", footnotes, "makeuppercase"-like functions, and other stupid things (colored text, inlined images, etc).

Trying Linux 2.4.15 "greased turkey":

SUCCESS using the Kodak DC210+ camera with the IrDA connection. You have just to use the same program that drives the camera via the serial port, changing /dev/ttyS0 (the "COM1" serial port) to /dev/ircomm0 (the IrDA port). LINUX RULEZZZZZ!!!!!!

The three "internet keys": if you start a script that shows the interrupt numbers (something repeating the grep intr /proc/stat command) you will see that the "internet keys" generate some keyboard interrupts when pressed (remember that the /proc/stat line of above contains the interrupt-total count and the single interrupt count starting from IRQ0). If you use showkey while in root mode, you will see that the third key generates a keycode 95 which we can use in our programs. But, the other two keys, generate something that showkey (that uses the keyboard in "raw" mode) doesn't recognize...

The "Audio-DJ" keys are only useful when the notebook is off. Sigh! It would have been very nice if those keys were available to user programs: with the notebook closed (but running), I would have some useful keys without the need to open... Remember that I use the notebook also as a tape recorder: silent, over three hours of continuous recording length, CD-quality, MP3 output!!!

Using the notebook as a recorder? Yes, I did it with a batch file that you can look below! Please note that $ARGS is rec arguments (we need to use two channels because mono recording is not fully supported by Linux sound driver for our sound-card), $AOUT is sox arguments to convert to mono, $LOGF is a logfile, $OUTF is the file name (date and time of recording, so I don't need to find every time a file name: the script is called r, so I just press "r" and then Enter to start immediately recording), $VOLU is a volume change (needed because the internal mic and my external mic are not that great with not really near sound sources), $LAME is the lame parameters (lame is an MP3 encoder - yes, I really do prefer the OGG/Vorbis format because of better quality and lower disk usage, but I share my University lessons recordings with my friends that have only W95 and W98...); notice that I used the "preset voice" parameter to get the quality cut down to minimum disk usage needed to store voice data; notice also the "nice" command, to run at lower priority the sox stereo-to-mono conversion and at the lowest priority the wav-mono-to-MP3 conversion, to get rec running at the maximum user speed.

ARGS="-c 2 -r 44100 -t .raw -w"
AOUT="-c 1 -r 44100 -t .wav -w"
OUTF=`date +%Y-%m-%d.%H-%M-%S.mp3`
VOLU="-v 4"
LAME="-h -m m --preset voice -v"

echo "0 OK, 0:1"
(rec $ARGS - | nice sox -s $ARGS - $VOLU $AOUT - | nice -20 lame $LAME - $OUTF) &> $LOGF
echo "L Break into program, 0:1"

stat -t $OUTF | awk '{ print $2,"bytes    mpg123",$1}'

Another nice "success report": I was out of battery and needed to backup 240 Mb of stuff to my friend Alberto. He gave me his "Delta Electronics ADP-60JB" AC Adapter of his Acer Travelmate 212DX notebook, and... it runs! Well, when Asus guys will know about this, they'll complain about WARRANTY...! I have still four months of warranty coverage, but I had no chance. I tried that AC adapter and discovered that it runs. I won't suggest to you to try just any AC adapter you find: this was a "19V 3.16A" with the same polarity of Asus AC adapter.

It seems that there is an incompatibility between Samba 2.2.1 and AX25 protocols (even if the ax0 interface was assigned an unique IP address and network) because Samba needs to broadcast some initialization packets. I have to switch off Samba server services to get rid of those Samba packets on the "listen" AX25 screen.

Linux 2.4.14: must be rebuilt on top of a clean kernel. I applied the patch to an already-compiled 2.4.13, called "make mrproper", "make clean", etc, but it had a strange error while compiling the loop device. I then took the 2.4.1 kernel, applied all patches up to 2.4.14, and finally got a compilable kernel.

Mandrake 8.1: some programs missing. The MP3 encoders, for example (because of the latest legal issues). Well, I don't care about it; I am using since a lot of time the OGG/Vorbis format, that is significantly better than MP3 (more quality in less space, with the same CPU usage). But I need a "lame MP3 encoder" because I have to keep (sigh!!) compatibility with my sad Micro$oft-based friends...

Wordprocessing: I happened to need to write a little document (two pages) that "must look nice because it won't be read". I tried three programs from the Mandrake 8.1 distribution.

KWord 1.1: in spite of the release number, it's not reliable. It hangs sometimes with a sad Micro$oft-like exception ("segment violation"), tipically in those buggy-buggy-buggy libstdc++ functions, and, in spite of the "automatic save" setup (I have a paranoid "save every two minutes") after a crash, eleven minutes of work (and they could be many many more!) went into the oblivion. I wasn't asked to recover anything, and didn't find anywhere a backup file. This is Not A Good Thing...!
A nice issue is the capability to use natively (and natively print) TrueType fonts. But... another infamous problem: I used in the document of above the Galliard and Garamond fonts, and GhostScript 6.51 couldn't calculate the truetype-to-curves translation made by KWord's (KDE's?) printer driver: it simply hangs there, in some strange cases (when using a sequence like "«super-hero»") using 100% of CPU time for hours...
Well, this KOffice, in spite of its release numbers, has still a number of nasty bugs that seem really to mean "not ready for office use"... Sigh! I hope they are GhostScript or Mandrake bugs. But I cannot rely on KWord after these strange things (and also some screen messages, like "QArray out of range...").

AbiWord 0.9.2: shame on you! It couldn't import an RTF-exported file from KWord 1.1 - neither RTF, nor HTML, XHTML, TeX. Sigh! It doesn't support TrueType fonts (even if natively supported from the X font server). Well, it's nice and doesn't seem to stop with SIGSEGVs, but it will not be my favourite editor if it doesn't support TTF fonts...

LyX 1.1.6fix3, got from Mandrake 8.1: it runs! LyX is a Good Thing. It doesn't SIGSEGV, it doesn't show strange things. Its interface, written in xforms instead gtk+ or (better) qt, is not that nice. But I wrote and get formatted a document of three sections with paragraphs/sections numbering, in a few minutes, less than any other wordprocessor! Simply great. Well, maybe you will be not really happy with its "fixed" styles and fonts (it does not support TrueType fonts because its engine is the LaTeX language), but it really answered great to the classical question "how much time I need to get my idea nicely printed?".
Oh, another Good Thing: after a lot of work (unsaved work!) on a document, I tried to reopen it and got a nice message "The saved backup is newer than the document: open this instead?". Well, I'm happy with this!

Long live LyX!!!

Mandrake 8.1

SIGH! At the very end of the installation, I found that the RPM database was corrupted. I had to do again the entire installation (and I didn't know that you can "rpm --rebuilddb"!!). This time I checked (arrrgh!! stooOOOoopid error!) "installation" instead of "upgrade" and had to solve another bunch of problems (reconfigure pcmcia, etc).

I found also that some RPMs in the downloaded CD images in ISO format fail in the CRC check ("MD5sum failed") and weren't installed. These files are netscape-common and kvirc RPMs, but I don't know if in the uninstalled RPMs there are other damaged files.

SHAME ON YOU, Mandrake guys! At the end of the installation it asks for... "XFree 3.3.6 (X-server) with Experimental Acceleration, or without"??! Where is XFree 4.1.xx? Why install a lot of XFree4 files and an XServer outdated (Jan-2000, almost two years ago)?! I would like to know why Mandrake guys think that recent X-servers are not better than the 3.3.6...!

Other Mandrake installation "bug": it creates a lilo.conf containing a "vga=788" line, that generates the "LCD bug" (must be changed to "vga=791", and after this you have to re-execute lilo).

The Mandrake 8.1 contains also the gcc 3.0.1 C/C++ compiler, but it is not present in the default options. In fact Mandrake installs the 2.96 version (but this seems to be a kernel need... do you remember the ugly "kgcc vs gcc" comedy?).

Screen energy-saving KDE option: if anything of it is enabled in the KDE control center, you will get the LCD-bug. The BIOS option runs without problems.

EXTREMELY COOL: networking with an Apple iMac and a true Apple LaserWriter! Click here to know how I did connect this Linux box via Linux kernel AppleTalk/EtherTalk support to exchange data and print files!

Linux 2.4.12 does not contain ACPI updates. Sigh! And it contains also a compile error in the IEEE-1284 section (identification of devices connected to the parallel port):

ieee1284_ops.c: In function `ecp_forward_to_reversè:
ieee1284_ops.c:365: `IEEE1284_PH_DIR_UNKNOWN' undeclared (first use in this function)

COOL: dual mouse under X/Free 3.3.x!!
You need only to add in the XF86Config file (already configured for the Synaptics touchpad) an "XInput" section for the USB mouse, and you will get them working together!!! (and so the procedure to check if the USB mouse is connected becomes useless!). Here's what I added (but you can click here to get the already-running XF86Config file):
Section "XInput"                  # the Asus USB wheel mouse
    SubSection "Mouse"
        DeviceName   "Asus USB mouse"
        Protocol     "IMPS/2"
        Device       "/dev/usbmouse"
        ZAxisMapping 4 5

The XFree 4.x syntax is quite different; you can find in the official Linux "second-mouse" HOWTO the appropriate instructions (for XFree 3.3.x and 4.x). Have fun!!! (You should even be able to add a third mouse...).

GREAT: I tried a true PCMCIA hard disk. This is a "Viper 340" 341 Mb hard disk, made by Integral Peripherals, using the two PCMCIA slots. On the bottom I can read also "made in Singapore; voltage +5V 0.5A, 5-55ºC; manufactured Mar'97" (a bit old, eh?!).
It needs just the same Linux internal ATA/IDE driver that I used to with the Kodak/SanDisk compactflash card. Have a look from the /var/log/messages file:
kernel: hde: Integral Peripherals 8340PA, ATA DISK drive
kernel: ide2 at 0x100-0x107,0x10e on irq 10
kernel: hde: 666144 sectors (341 MB) w/32KiB Cache, CHS=771/16/54
kernel:  hde: hde1
kernel: ide_cs: hde: Vcc = 5.0, Vpp = 0.0
Please note that it uses 5V instead of 3.3V as seen on the SanDisk (there is a true disk into! this is not a solid-state disk), and it has even 32 kB cache on board. Performances: cool - I copied 68 Mb in some hundreds files, from the disk to the Asus hard disk, in only 1'14": more than 910 kb/sec of throughput, even faster than the solid-state SanDisk. Another copy, five files for a total of 108 Mb, took (including a sync) only 1'24" (that is, more than 1.2 Mb/sec)
The only slow thing is... the first time I insert it in the slot, it freezes the operating system for twenty seconds to initialize!

Yeah: I successfully tried the USB keyboard and USB mouse of an iMac! I recompiled my 2.4.4 kernel with full USB-HID (Human Interface Devices) support; connected the Apple iMac USB keyboard to a notebook USB port, and the Apple iMac mouse to one of the Apple iMac keyboard USB ports, and then restarted Linux. Great: as of dmesg log below, Linux recognizes the external USB keyboard, the extra USB hub and ports, the iMac mouse (and even the Asus USB mouse if you connect it to any free USB port, either notebook port or Apple keyboard port).
Apple keyboard was recognized as an "Alps Electric M2452", while Apple mouse was recognized as a "Logitech M4848".
The notebook keyboard and the USB keyboard may be used concurrently; this is not true for the mouse because X reads data from the USB mouse from /dev/input/mouse0, and the second mouse will be recognized as /dev/input/mouse1, but this doesn't mean that a program cannot use two of them (and a touchpad) in the same moment... LINUX is great!!!

usb-ohci.c: USB OHCI at membase 0xccc1b000, IRQ 5
usb-ohci.c: usb-00:01.2, Silicon Integrated Systems [SiS] 7001
usb.c: new USB bus registered, assigned bus number 1
hub.c: USB hub found
hub.c: 3 ports detected
PCI: Found IRQ 5 for device 00:01.3
PCI: The same IRQ used for device 00:01.2
usb-ohci.c: USB OHCI at membase 0xccc1d000, IRQ 5
usb-ohci.c: usb-00:01.3, Silicon Integrated Systems [SiS] 7001 (#2)
usb.c: new USB bus registered, assigned bus number 2
hub.c: USB hub found
hub.c: 2 ports detected
usb.c: registered new driver hid
usb.c: registered new driver acm
usb.c: registered new driver usblp
usbserial.c: USB Serial support registered for Generic
usb.c: registered new driver serial
hub.c: USB new device connect on bus2/1, assigned device number 2
hub.c: USB hub found
hub.c: 4 ports detected
hub.c: USB new device connect on bus2/1/1, assigned device number 3
event0: Event device for input0
keybdev.c: Adding keyboard: input0
input0: USB HID v1.00 Keyboard [Alps Electric?M2452 M2452] on usb2:3.0
hub.c: USB new device connect on bus2/1/2, assigned device number 4
event1: Event device for input1
mouse0: PS/2 mouse device for input1
input1: USB HID v1.00 Mouse [Logitech M4848] on usb2:4.0
hub.c: USB new device connect on bus2/1/3, assigned device number 5
event2: Event device for input2
mouse1: PS/2 mouse device for input2
input2: USB HID v1.00 Mouse [ USB Mouse STD. ] on usb2:5.0
usb.c: USB disconnect on device 4
usb.c: USB disconnect on device 5
hub.c: USB new device connect on bus2/1/2, assigned device number 6
event1: Event device for input1
mouse0: PS/2 mouse device for input1
input1: USB HID v1.00 Mouse [ USB Mouse STD. ] on usb2:6.0

I got Linux 2.4.10 kernel, and I verified that the "SiS 630/540 accelerated framebuffer support" is still buggy (the "LCD bug"). So we have still to use the VESA/framebuffer mode. There is also another problem, this time in ACPI section: we cannot anymore compile it as a module (it has "unresolved externals acpi_ut_xxx" on any "ospm_xxx" module); if we compile it in the kernel, we get a system hang at the beginning without even an useful debug information screen; if we compile the ACPI support in the kernel and the functions as modules, we still cannot get them insmod'ed. Let's wait for the next release... sigh!

YAY! I found that the VESA BIOS of this notebook HAS a 1024×768×24bpp mode, despite what I told you in the last months! Do prepare for a great "vga=864" line in the /etc/lilo.conf!!!

How did I? I got the VBE (Vesa Bios Extensions) specifics from the VESA site, and read that any VGA manufacturer can add new "non-standard" modes: the software has to scan all possible VESA modes to find the needed resolution and color-depth.

So I manually tried to get information for VESA modes from 0x118 and up. The 0x118 (1024×768×24bpp standard VESA mode) is not supported, but there is an equivalent mode (named 0x160) which results supported "but not available now".

Since VESA modes support double buffering, I configured the BIOS to use 8 Mb VRAM instead of 4 Mb (while in Linux you need only 4 Mb VRAM to get 1024×768×24bpp and ×32bpp). Well, with 8 Mb it is usable even in DOSemu!

So, instead of the "vga=791" line that I suggest in this page, one can try a "vga=864" line to get 1024×768×24bpp even with an XF86_FBDev server. Someone tried this on an A1360 notebook but without success, but I'll try it as soon as possible - indeed, I do not need it here, because I use the XF86_SVGA server that can switch to 24bpp or 32bpp starting from 16bpp without "LCD bug".

I should also test other VESA modes (maybe there are 32bpp VESA-accessible modes, and even higher resolution modes for use on an external monitor), but I don't have much time to write a VESA-search-and-test-modes program and don't even have an external 1280x1024 monitor. Well, you have just to know that also here Linux performs always better than Windows.

Another little hint: Nick, from UK, remembered me that if you have less than 4 Mb Video-RAM, you surely won't get the XFree running! Until we have a full accelerated X-screen-server (4.1.xx?) using really lots of VRAM, we shouldn't setup the BIOS to select more than 4 Mb VRAM.

Just a little hint for playing DVD's: the best resource collection is DVD on SourceForge. Someone told me that the Xine player performs better than the OMS package that I suggested in these pages.

SOLVED the ugly problem with PCMCIA and Linux kernel 2.4.9: I asked to David Hinds (Linux-pcmcia developer) and he answered:

"Edit the /etc/pcmcia/config file and replace all instances of ide_cs with ide-cs."

I also compiled the ide-cs support in the kernel (not a module anymore), to get rid of all those ide-cs.o lying around... and now, YES, it runs great! The note below (with a bit of sadness) should need some little corrections, but I leave it there because it's history... (the kernel "Oops" and "BUG slab.c" simply disappeared).

Another nice note: to get something about ACPI, compile the ACPI bus manager as a module (so all ACPI functions get compiled as modules), and then you will find some interesting data in /proc/acpi.
The sad thing is that the acpid-07-Nov-2000 that comes with Mandrake 8, doesn not recognize the ACPI stuff... Well, as of Linux 2.4.9, after loading the modules using these commands:

modprobe ospm_busmgr
modprobe ospm_ec
modprobe ospm_processor
modprobe ospm_system
modprobe ospm_button
modprobe ospm_battery
modprobe ospm_ac_adapter
modprobe ospm_thermal
...I got these results below (the AC adapter is connected and the battery is already charged); I skipped some useless files:

/proc/acpi/info file contains:
ACPI-CA Version:         20010615
Sx States Supported:     S0 S1 S3 S4 S5

/proc/acpi/ac_adapter/0/stat says "on-line" (that's correct)

/proc/acpi/battery/0/info says correctly:
Present:                 yes
Design Capacity:         2976 mAh
Last Full Capacity:      2976 mAh
Battery Technology:      secondary (rechargeable)
Design Voltage:          10800 mV
Design Capacity Warning: 297 mAh
Design Capacity Low:     29 mAh
Capacity Granularity 1:  268 mAh
Capacity Granularity 2:  2679 mAh
Model Number:            BA-03
Serial Number:
Battery Type:            Li-I
OEM Info:                ASUSTek

/proc/acpi/battery/0/status says correctly:
Present:                 yes
State:                   charging
Present Rate:            0 mA
Remaining Capacity:      2958 mAh
Battery Voltage:         9108 mV

/proc/acpi/button directory contains three empty directories (lid, power and sleep)

/proc/acpi/processor/0/info contains only a "<TBD>" string (grepping "TBD" in ACPI kernel sources suggests that it means "to be developed!")

/proc/acpi/processor/0/status contains some stuff that I don't know what it exactly means:
Bus Mastering Activity:  ffffffff
C-State Utilization:     C1[649] C2[53215] C3[0]

/proc/acpi/thermal/0/status contains the temperature (in ºKelvin*10, so "3172" should be 43.6ºC), but it seems quite incorrect (the fan is not running, while the temperature should be quite low because I started up the notebook a few minutes ago):
Temperature:             3172 (1/10th degrees Kelvin)
State:                   ok
Cooling Mode:            active (noisy)
Polling Frequency:       n/a

Well, I'm again at home and I tested Linux kernel 2.4.9, but it's not that useful on this notebook because:

Also, Linux kernel 2.4.9 contains a curious bug about disk buffering. I found the fan running while no process was hogging the CPU, and 132Mb of buffers allocated...!

Some e-mails I got in the last weeks: someone said to have great problems using M1380 (M1000 series of the Asus notebooks), that it's based on the Intel I815 instead of the SiS630.
Some other complained about Windogs screen drivers, that have a number (large number) of strange, nasty, ugly effects. Only under Linux this notebook runs great!!!

GREAT: someone emailed me a copy of an Asus tech guy email stating to look at these pages to get some help about installing Linux: YAY!!!

Another "success report", this time by a Medion 9438 notebook user. The Medion is an A1360-based computer, with some differences - a 4 Gb hdd and a wonderful Panasonic 8×4×32 CD-R(W) drive (but without Audio-DJ keys).

Well, the modem runs. I got a "success report" from Alessandro B. from Italy: he compiled a 2.4.5 Linux kernel without sound support, then insmod'ed the sm56.o driver and then he could connect to the Internet at 56k speed!
There are two ugly problems: after some minutes of inactivity, the modem driver hangs; and after one call, you have to reboot the machine. I'll try it in the next days; maybe I can find a decent solution (sound support compiled as a module) to get all things running!

I got another "success report", this time by a Mitac 7521 notebook user. The Mitac is SiS630-based and had the same problem "LCD bug"; now it runs at 1024×768×32bpp!

The Linux kernel 2.4.7 is out - I tried it, and there isn't anymore the IrDA bug! Yeah!!! :-) I sent a "thank you!!" email to Dag Brattli. But the SanDisk bug is still there and I cannot use anymore the ide-cs.o module of the 2.4.1 kernel (because it gives a kernel "Oops" when inserting the compactflash card). And now, again, the best kernel versions are 2.4.3 and 2.4.4...!

ARRRGH: Motorola discontinued softmodem support, as of July 2001. Sigh!. I simply don't understand why.

As of July 2001, Motorola has exited the software modem business. Effective immediately, Motorola is discontinuing all soft modem products and future development.

I tried the latest Motorola SM56 driver for Linux: it doesn't work. Below, the email I sent them, complaining about their discontinued support (and suggesting to release softmodem documentation) and about the bugs of the SM56 driver.

Dear Sirs,

I just found that Motorola discontinues softmodems support.

I hope that this means that full documentation will be released to the public domain. If Motorola is not interested in softmodem support anymore, there are lots of people working in the open-source that are willing to continue that support. And this means that lots of Motorola softmodem cards will be sold.

Well, I tried the Motorola Softmodem SM56 Rel. 5.01 Build 01 driver that I just found on Motorola site.

The module doesn't run on my Linux-based Asus A1370D notebook, that has a SM56-based softmodem on its SiS630 mainboard; the "lspci -v" command reports:

00:01.6 Modem: Silicon Integrated Systems [SiS]: Unknown device 7013 (rev a0)
               (prog-if 00 [Generic])
  Subsystem: Asustek Computer, Inc.: Unknown device 1456
  Control: I/O+ Mem- BusMaster+ SpecCycle- MemWINV- VGASnoop-
           ParErr- Stepping- SERR- FastB2B-
  Status: Cap+ 66Mhz- UDF- FastB2B+ ParErr- DEVSEL=medium >TAbort- <TAbort-
          <MAbort- >SERR- <PERR-
  Latency: 32 (13000ns min, 2750ns max)
  Interrupt: pin B routed to IRQ 5
  Region 0: I/O ports at a400 [size=256]
  Region 1: I/O ports at a000 [size=128]
  Capabilities: [48] Power Management version 2
        Flags: PMEClk- DSI- D1- D2- AuxCurrent=55mA PME(D0-,D1-,D2-,D3hot+,D3cold+)
        Status: D0 PME-Enable- DSel=0 DScale=0 PME-

As stated in the notebook documentation and drivers, this is a true Motorola SM56 Softmodem product. The "Unknown" markers there, mean that Linux does know about that hardware (this is a sad story: for all other peripherals and cards in the notebook, there is full open-source support. You may have a look here to verify this).

Kernel 2.4.3 and 2.4.4:

after the installation procedure (as suggested on your site) the module loads correctly, but when I try to send a byte to it (example: sending ATZ to the modem), the computer hangs up for about 30 seconds, and then I get these error log messages:

      SM56: SS_DisplayFatalErr - 0x8102
      smdspGlobalSetup - Failed to detect codec, exit...
      MotNTSmSrlCreate - Smdspopen with a bad status of 0xffffffff

That is, the module is correctly loaded (and can be unloaded when not in use) modem is not detected. This means that you cleared the bug of your previous "85.1" driver release (I wrote you a note about it, but I never got any answer).

Kernel 2.4.7:

the module loads, but insmod gives a "Segmentation fault" error generating a kernel BUG in the error log messages:

       kernel BUG at slab.c:1062!
       invalid operand: 0000
       CPU:    0
       EIP:    0010:[<c0128c2b>]

and the module remains in "initializing" phase. And it cannot be used or unloaded anymore (I have to restart the machine). It seems that your driver can raise the BUG of above (in the "/usr/src/linux-2.4.7/mm/slabs.c" file), but I cannot guess if it's only a kernel bug or it's SM56-related stuff. I saw that "BUG at slab.c" line only when I tried the "insmod sm56" command.

After "insmod", the output of lsmod says:

       Module                  Size  Used by
       sm56                 1345984   1  (initializing)

Thank you for your help. I wish I can use the SM56 softmodem of my notebook soon.

Nokia 7110 - yes, it runs: click here to know how to send SMS messages via IrDA port and to know how to fetch stored SMS messages to hard disk.

I got a crossover cable and finally did some serious tests with the SiS900 internal network card. I got about one meg per second of throughput (Linux-to-Linux is one meg per second; with Linux-to/from-Windows I got only 380 kb/sec...! click here for more info).
One meg per second? That's quite fine for a 10Mbps transceiver... yes: it's just a 10Mbps, and it's also half-duplex, sigh!

I tried the Linux 2.4.6 kernel, but with the same result for 2.4.5: if the IrDA support is enabled, the kernel hangs while loading. I wrote an email to Dag Brattli; we'll have to wait at least for the next kernel patch... :-(

Now we can dance Samba: click here to know how to connect to a Windows drive on a Windows network...! (This means also that the internal network card runs without problems).

Internal SM56 modem... SIGH!
I sent a mail to SM56 Linux Support of Motorola site. First, the message was bounced (incorrect address). Then I sent it to the generic support email address: I got an automated reply saying "in 24 hours we will answer"... and they simply didn't answer. Then I checked again their site and I found that they changed the email address, and sent the email there (
Now I am still waiting for their answer...
In a few words, I wrote them about two bugs I found in their driver:

  1. they didn't think that an IRQ may be shared among several peripherals;
  2. the driver does a MOD_DEC_USE_COUNT even if a MOD_INC_USE_COUNT wasn't done (this bug should disappear as soon as they correct the initialization section to support IRQ-sharing).

This is the email I sent them - still without answer... :-(

to: S. Harris, K. Shao, A. Rosario <>, <>

about: Motorola SM56 SoftModem Serial Driver (sm56.o build 85 preview 1)

Dear Sirs,

I have an Asus A1370D notebook with SiS630 chipset and integrated Motorola SM56 modem. I installed and use only Linux on this computer.

The notebook is a Pentium III 750 Mhz, with 192 Mb RAM, 20 Gb hard disk, TFT LCD 1024x768 13.3" display, DVD 8x, SiS630 chipset.

I got all hardware running with standard Linux 2.4.4 kernel drivers, except the SM56 modem, for which I need your software driver.

So I downloaded your softmodem driver for Linux from Motorola site: "Motorola Softmodem: version SM56 Rel. 5.00 Build 85".

The modem is a true SM56 modem: I verified it using "cat /proc/pci", obtaining:

  Bus  0, device   1, function  6:
  Modem: PCI device 1039:7013 (Silicon Integrated Systems [SiS]) (rev 160).
  IRQ 5.  Master Capable.  Latency=32.  Min Gnt=52.Max Lat=11.
  I/O at 0xa400 [0xa4ff].
  I/O at 0xa000 [0xa07f].

and "lspci -v":

  00:01.6 Modem: Silicon Integrated Systems [SiS]: Unknown device 7013
(rev a0)
  (prog-if 00 [Generic])
     Subsystem: Asustek Computer, Inc.: Unknown device 1456
     Flags: bus master, medium devsel, latency 32, IRQ 5
     I/O ports at a400 [size=256]
     I/O ports at a000 [size=128]
     Capabilities: [48] Power Management version 2

The command "insmod sm56" loads the driver immediately, but after issuing a command like "echo ATZ > /dev/modem" (/dev/modem is correctly linked to /dev/sm56 that is a special device char/24/0), I got these kernel messages:

SM56: SS_DisplayFatalErr - 0x8102
smdspGlobalSetup - Failed to detect codec, exit...
MotNTSmSrlCreate - Smdspopen with a bad status of 0xffffffff
motomdm.c(914): mot_open - error -1 in SmSrlCreateport
Unable to handle kernel NULL pointer dereference at virtual address
 printing eip:
*pde = 00000000
Oops: 0000
CPU:    0
EIP:  0010:[<ccd1aee8>]
EFLAGS: 00010093
eax: 00000000   ebx: 00000000   ecx: 00000000   edx: ffffffff
esi: 00000282   edi: cb27b000   ebp: cb231df8   esp: cb231df0
ds: 0018   es: 0018   ss: 0018
Process bash (pid: 1890, stackpage=cb231000)
Stack: 00000000 cab42ee0 cb231e18 ccd4fbb3 00000000 00000010 00000246 00000000
       cb27b000 c4ddcc80 cb231e58 ccd4ece3 c85e72e0 ffffffff c4ddcc80 00000282
       cb27b12c cb231eb8 0000000a c028ce29 c037eb24 cb27b000 00000000 cb27b000
Call Trace: [<ccd4fbb3>] [<ccd4ece3>] [<c028ce29>] [<c01a3127>]
[<c0114c9e>] [<ccd4e04b>] [<ccddc0c0>]
       [<ccddc0af>] [<c01a363e>] [<c013a800>] [<c013b582>] [<c01314a6>]
[<c0130600>] [<c013052d>] [<c0130816>] [<c0106c37>]
Code: 8b 53 3c 85 d2 74 1b 83 ec 0c 68 48 40 e2 cc e8 98 4e 03 00
motomdm.c(2179): mot_intr - Out of band IRQ recieved.
last message repeated 495 times
last message repeated 2657 times
last message repeated 2291 times
last message repeated 2215 times
last message repeated 1251 times
motomdm.c(2179): mot_intr - Out of band IRQ recieved.
last message repeated 755 times
last message repeated 2045 times
last message repeated 2433 times

I happened to try to hear some audio files (using /dev/dsp) and got no sound output. I tried to "rmmod sm56" but got "device or resource busy" because its use count was -1 (negative values are not allowed, since this is a counter; maybe a MOD_DEC_USE_COUNT or similar function was called when the "use count" was already zero). In fact, asking for "lsmod", I got:
        Module                  Size  Used by
        sm56                 1191008  -1  (unused)

I had to reboot the machine to clean up the module and hear /dev/dsp music files again.

The extreme number of repetitions of the message "mot_intr, Out of band IRQ received" seems to mean that the IRQ of the modem is used from a bunch of other drivers. In fact, on my machine, "cat /proc/interrupts" reports:

  0:    1277111          XT-PIC  timer
  1:      13396          XT-PIC  keyboard
  2:          0          XT-PIC  cascade
  3:          2          XT-PIC  serial
  5:      26708          XT-PIC  SiS 7018 PCI Audio, Ricoh Co Ltd
RL5c476 II (#2), usb-ohci, usb-ohci, sm56
  7:          1          XT-PIC  parport0
  8:          1          XT-PIC  rtc
  9:          0          XT-PIC  Ricoh Co Ltd RL5c476 II
 14:      77108          XT-PIC  ide0
 15:      38029          XT-PIC  ide1
NMI:          0
ERR:          0

Because of a very strange SiS project issue, it happens that a bunch of interrupts is assigned to IRQ5: the SiS audio, the 2nd Ricoh PCMCIA controller, the two USB controllers, and even the SM56 modem.

So, it seems to me that your sm56 driver catches the IRQ of other peripherals and believes they are "unexpected" IRQs, "out of band".

This should be a simple problem of IRQ-sharing. Without the SM56 module loaded, I got all things running (USB mouse, SiS audio, PCMCIA cards), only the SiS audio stops output (without further errors) after loading the sm56.o module driver.


Interesting: I placed in /etc/rc.d/rc.local these commands:

# XF86Config switcher
# if I forgot to plug the asus usb wheel mouse, then use a touchpad X config

cd /etc/X11
rm -rf XF86Config
if dmesg | grep "USB Mouse"
        ln -s XF86Config.usb XF86Config
        ln -s XF86Config.syn XF86Config

# --- end ---

What happens? At boot time, before the X-server starts, it will change the X configuration file depending on the string "USB Mouse" in the boot messages. The file XF86Config.usb is a copy of the one that you can download here, and the XF86Config.syn is the same but with the Pointer definition changed to:

Section "Pointer"                 # the Synaptics touch-pad
  Protocol "PS/2"
  Device   "/dev/psaux"

Result: I can work with X even if I forget to plug in (or, have not some space on which use) the USB mouse.

Other news:

Old news...:

Here the adventure begins...

Before starting

Beware! All this aims to:

I chose to buy this nice notebook because:

What will I do with this machine? Everything I did with the old notebook:

All of that, using free software!! (GPL-licensed, freely distributable, free with or without source, etc). Obviously, everything under Linux.

So, I am not happy I paid for a Windows ME license that I'll never use. But it seems a bit hard to get - here in Italy - a refund for this ugly "Windows tax" (at least, on some models, one could choose "Win98" instead of "WinME", saving about 25 US$).

What's inside the boxClick here for index page.

Only WinME and ViaVoice were in Italian; the other stuff (Drakan, manuals, drivers, etc) was all English.

Technical characteristics


External connectors (clockwise, starting from left):

On the right there are the DVD drive and the floppy drive.

Some notes about hardware...

First of all, one needs surely the RAM expansion, because system RAM is shared with Video-RAM (this is one of the features of the SiS630 chipset). Example: with the standard 64 Mb on-board, if you configure 4 Mb VRAM you will get 64-4=60 Mb RAM; if you want full graphics acceleration and configure 32 Mb VRAM, you will get only 64-32=32 Mb system RAM...!
Instead, with full RAM expansion (192 Mb total, as in my case), even subtracting 32 Mb for VRAM, there still remain 160 Mb system RAM.

A little note: on the Asus site one can get the BIOS upgrade (dated 2001-03-30) to configure 64 Mb VRAM!

DVD drive. The engineering needed it single-sided (8.5 Gb): there is no space for a second head; they would have had to resize the entire case - or, to put out (read: external) the floppy drive. To read a "double-sided" disk (17 Gb) or DVD, you have to manually turn over the disk (eject, turn, re-insert).

Interesting: the hard disk is /dev/hda (primary-master) and the DVD is /dev/hdc (primary-slave): there are two PCI IDE interfaces (using always two drives, only one was needed), with a little gain in speed.
Click here for index page. Speed: not great, but interesting: the hdparm -t /dev/hda reports 18.74 Mb/sec (that is, the hard disk features 150 Mbit/sec, the best for an UDMA/33). Note that the disk does not start fully initialized; hdparm must be issued (look here for details).

CD/ROM: a bit slow; hdparm reports 1.74 Mb/sec for data CDs (that's about 12×, even if Linux itself sees it as a 24×; in any case, with other CDs - original and "home-made" - I got more speed, about 16×). Music ripping speed (using cdparanoia III 9.7) it's about at 3×.

The two PCI IDE interfaces (primary and secondary) present the disks as connected with "40-pin cables".

Other: the PCMCIA interface (Ricoh RL5c476 II) uses two interrupts (9 and 5); one of them (IRQ5) used also by soundcard, USB hub and other things (yeeeoowww!). I don't think this was really urgent; there's still a bunch of free IRQs... maybe it's a brain-damaged engineering issue of the SiS motherboard.

The keyboard: the Enter key is sized as just two keys (vertically), so it's not very easy to find it without looking the keyboard... The first keyrow (Esc, |\, F1-F12, Pause, Ins, Del) is "half-sized" (but the keys could be full-sized without re-engineering the notebook; I don't understand why). The three "Internet keys", on top right of the keyboad, generate new scan-codes and the "power-on" signal (one can power on the notebook pressing one of these keys instead of the standard power-on on the left).

Click here for index page. The "Audio DJ" keys, with a really nice blue led on the right, let you play music CDs without turning on the notebook. Really great!

Last news: the CPU can be upgraded (as they say) to "850+ MHz" (but the uPGA2 slot is not accessible to the user). The LCD display, as they say, is a 1024×768×16bit (but this is a Windows limit: X/Free, that's free, drives it correctly at 24 bit and even 32 bit, without flickering!).

I read somewhere that the so-called "SpeedStep(TM)" technology features a speed decrease of the CPU to consume less power (this happens when the batteries go down). So, a 750 MHz CPU will run at 500 MHz when low batteries (but, so far, I didn't notice it: Linux isn't performance-hungry like other so-called "operating systems").

The cat /proc/cpuinfo reports:
processor       : 0
vendor_id       : GenuineIntel
cpu family      : 6
model           : 8
model name      : Pentium III (Coppermine)
stepping        : 6
cpu MHz         : 747.733
cache size      : 256 KB
fdiv_bug        : no
hlt_bug         : no
f00f_bug        : no
coma_bug        : no
fpu             : yes
fpu_exception   : yes
cpuid level     : 2
wp              : yes
flags           : fpu vme de pse tsc msr pae mce cx8 sep mtrr pge mca
                                          cmov pat pse36 mmx fxsr sse
bogomips        : 1490.94

The cat /proc/fb reports "0 VESA VGA". The cat /proc/ide/sis reports:
--------------- Primary Channel ---------------- Secondary Channel -------------
Channel Status: On                               On
Operation Mode: Compatible                       Compatible
Cable Type:     40 pins                          40 pins
Prefetch Count: 512                              512
Drive 0:        Postwrite Enabled                Postwrite Enabled
                Prefetch  Enabled                Prefetch  Disabled
                UDMA Enabled                     UDMA Enabled
                UDMA Cycle Time    4 CLK         UDMA Cycle Time    4 CLK
                Data Active Time   2 PCICLK      Data Active Time   2 PCICLK
                Data Recovery Time 1 PCICLK      Data Recovery Time 1 PCICLK

Installation of Linux...?Click here for index page.

Mandrake 8 and Mandrake 8.1

Well, the bad-bad-bad-news is that so far the 2.4.17 Linux kernel (in the SiS630-framebuffer section) and the X/Free 4.1 server, simply don't know how to initialize the LCD display on this SiS630-based board.
You get a strange (and almost surely dangerous) white light on the LCD (we here call it "the white liquid LCD bug") and then you can only try to reboot the machine.

BUT, I got the beast running X at 1024×768 just doing so:

  1. Starting Mandrake 8 installation, you get a decent 800×600 graphics screen (and even USB mouse ready) while installing: this means that the standard VESA-framebuffer option is sufficient to go in graphics mode (this was the real kink).
  2. Install Mandrake 8 using "Expert mode", because you mustn't accept the "Try the X Configuration Now" temptation!!
    (if not "Expert mode", the Mandrake will try to test XFree graphics and... ZOT!!! the LCD bug is there).
  3. Once the installation is complete, you can restart the system in text-mode. Then you can substitute these files:
  4. Remember that ctrl-alt-backspace sequence kills the X-server (and should restore text-mode - if you're lucky).

Doing the above simple things, I got X running at 1024×768 in 32bpp since the last days of March 2001 (and, as described below, even under the old Mandrake 7.2 distribution).

So far, I still cannot try to set/reset text-mode (but a decent terminal window is always sufficient), and still have no graphics acceleration (games like SearchAndRescue or TuxRacer 0.61 do less than one or two frame per second...) because the 3.3.6 X-server of the Mandrake 8 doesn't have acceleration. But it's months that I'm working great with it.

If you still get the LCD-bug, please check here:

If you still get the LCD-bug, you surely have done one or more of these errors:

(the download section is still here).

Yes, some things are not for newbies. Sorry, Linux is a "Real Men Don't Need Too Much Help" thing. I cannot explain to all of you how to "get and compile the latest Linux kernel".

I think I've no more hints to add to this description. If you happen to get it running, please let me know if you did something different from what I've described above (I'll add your comments here).
Thank you!

IMPORTANT: the section below is somewhat obsolescent, but it costed me too much work to erase it just because it's old stuff... So, here it is. If you want to get the latest news, click here.

Arrrrgh, this hardware is just too recent even for a Mandrake 7.2; if you are a true computer wizard and have some good experience of Linux, then you can try...

Well, I do not make any express or implied warranty about what I'm going to describe, neither I will give extensive help because with next Mandrake release (Mandrake 8.0 is already out), based on Linux 2.4.x, a bunch of problems will simply disappear.

But, if anyone is in a hurry to get some results, read below:

Mandrake 7.2

After configuring the BIOS to get the DVD/CDROM bootable, I started Mandrake 7.2 installation. It hangs (sigh!) at the beginning of the installation process, while searching for pcmcia peripherals (the machine needed a hard reset - turn off and then turn on).

I restarted with the "Expert / Text" installation, and I read on the console, while the installer was checking pcmcia stuff, a tremendous kernel Oops - the pcmcia_core.o module couldn't get right with the Asus Ricoh controller, and the Mandrake distribution, in spite of Ye Olde RedHat 6, doesn't allow you to skip that check. Obviously, the problem is only of that extremely-patched 2.2.14 kernel supplied with Mandrake 7.2; the 2.4.x Linux kernel supports Asus PCMCIA with and without modules.

After some tries, I realize that *.o installation modules of Mandrake are compressed in the /lib/ file of the installation ramdisc (almost three Mb) that one can modify using the packdrake command (that is a nice Perl script, only 18k).

Let's go hacking Mandrake installer: I uncompress/extract the file, remove its ugly pcmcia_core.o, recreate the 260 modules left, restart Mandrake installation and, after the first screens, in the prompt-window (alt-F2) I erase the from the ramdisk and place there my Then I continue the installation: while checking pcmcia, the installer does not find the pcmcia_core.o but doesn't complain.

Well, I used these commands - only if you are a computer guru you will exactly understand every command (and find out a bunch of megabytes of space to execute them):

cp /lib/ .
mkdir tmp
packdrake --list | awk 'BEGIN{getline}{print $3}' >modules.list
for a in `grep -v pcmcia_core.o modules.list`
packdrake --extract tmp/ $a
cd tmp
rm pcmcia_core.o ../
grep -v pcmcia_core.o ../modules.list | packdrake -9 --build ../
cd ..
rm -r tmp
cp /lib

Sadly, there are other problems with Mandrake 7.2 installation: after copying all RPM files (more than 1200 if you really install everything from the two CDs), you will have to configure X/Free. And there you get a machine hang again (grrrr!) with even a TFT LCD strange behavior (I didn't let it show me for more than some seconds).

So, I compiled - on another computer - the 2.4.3 Linux kernel; I configured it with frame-buffer support (just to use VESA BIOS 2.0 to get in graphics mode at boot) and after a good number of tries, I got - I dunno exactly how I did - to configure even X/Free 3.3.6.

Well, you don't have to re-do the entire work, you can click here to download the files.

Summary for high-density brains:Click here for index page.

  1. start Mandrake 7.2 installation
  2. after the first screens, before the pcmcia test, go to the terminal window (using ALT-F2)
  3. remove the file /lib/ (don't panic, we're in the installation ramdisk);
  4. copy there the hacked file (the one that does not contain the pcmcia_core.o - if you do not like to use use packdrake, you can get the already-hacked file clicking here; beware, it's 2878k! the original is in the Mandrake/inst/lib directory of the first installation CD;
  5. continue installing (it won't complain about the missing pcmcia_core.o) and set up your preferred RPMs;
  6. when you get to the X/Free setup... good luck! If you can configure it without problems, let me hear your whistle. If you cannot and get a machine hang (like me), turn off and then turn on and et voilà!, you can use it in full text mode.

But this story is not at the end...!

A little note now: because of the hard disk size, I hardly suggest the reiserfs filesystem instead of ext2. First, because of the stunning speed of large directory scans (the biggest advantage with respect to ext2: have a try!); second, because if you crash (for example, you run out of batteries and get it turned off before you did a correct shutdown), the file-system check and update at the boot it's a matter of seconds (simply great!!). Third, because you don't have to bother with i-nodes (and i-node limits). I reformatted the reiserfs partition in the terminal window.

Another little note: you shouldn't need a swap partition. The disk is fast enough to let you use a file swap. Mandrake 7.2 does not allow you to do it, but... (well, I'll explain this another day!).

Good!!! Linux started!! Now, while in text mode:

Warning. In the SiS site there's a "Linux drivers" page where they state that... everything needed is already in Linux kernel! The only thing you can download from there is an XSiS_SVGA server (X/Free 3.3.6) that they created patching XF86_SVGA. But we don't need this, because it's a bit old (it's dated Feb'2000) and because the current XF86_SVGA driver (the one in Mandrake 7.2 is dated Oct'2000) supports well the SiS630 (at least, until we use an LCD display!).

After all this stuff, the machine will still start up in text mode (so you'll have to login and start manually startx). When you exit X/Free, you cannot get anymore text mode or graphics mode (except blindly restarting X/Free), but you can still press Ctrl-Alt-Del or, by root, give (blindly) the halt command to get a normal shutdown.

Well, now I can test some PCMCIA cards I have here... I place them in the PCMCIA slots and, using cardctl ident I get:
Socket 0:
product info: "SunDisk", "SDP", "5/3 0.6"
manfid: 0x0045, 0x0401
function: 4 (fixed disk)

Socket 1:
product info: "3Com Corporation", "3C589", "TP/BNC LAN Card Ver. 2a", "000002"
manfid: 0x0101, 0x0589
function: 6 (network)

These two do work and I used them without problems. The first uses the PCMCIA ATA/IDE kernel driver, the second needs the PCMCIA 3C589 kernel driver. Here are two other PCMCIA cards:
Socket 0:
product info: "CIRRUS LOGIC", "FAX MODEM", "MD144ECP", "0-0"
function: 2 (serial)

Socket 1:
product info: "Olicom", "GoCard Eth/Modem 336", "OC2232", "NONE"
manfid: 0x0121, 0x3222
function: 0 (multifunction)

The Olicom card is "unsupported" because it's not a serial modem. The Cirrus instead is a standard serial modem (you get a true /dev/ttyS2) but this card is just an old 14400 baud V32BIS modem).

Don't miss this! (hdparm)

I strongly recommend to put in /etc/rc.d/rc.local the commands:

hdparm -ivm 16 -c 1 -a 16 -u 1 /dev/hda
hdparm -a 64 -c 1 -u 1 /dev/hdc

the first speeds the hard disk, the second tries to do it also for the CD/DVD.Click here for index page.

I tried to get a little more speed - without great success: Linux correctly recognizes that the ATA66 modes udma3/4/5 are not available.

The available files here - read above and below to get some more info:

If you want to upgrade the BIOS, click to Asus Bios Upgrade Page.

What's still missing...

I did stress the SiS900 network card for more two years, and got no problems - even trying the Samba it ran without problems.

Internet browsing via GSM phone and Infrared connection!!!

I've a nice Nokia 7110 GSM cellular phone, with internal 14400 modem and infrared port.

Well, after I configured in the BIOS the infrared port of the Asus A1370D notebook as "com2" (that is, addr=0x2f8, irq=3; or, from Linux, /dev/ttyS1) and letting IrDA services run, I did:

irattach /dev/ttyS1
mknod /dev/ircomm0 c 161 0
mknod /dev/irlpt0 c 161 16
ln -s /dev/ircomm0 /dev/gsmodem
chmod 644 /dev/gsmodem

(the irattach is needed after every boot; the other commands are needed only once). Yeeeah! it runs, even if the findchip command does not report anything! :-)

Now one can place the Nokia cellular phone near the IrDA port of the notebook (on the left), and activate the "Infrared" option from the 7110's menu and... wow!!!

If you use minicom (or xminicom), you find on /dev/gsmodem a regular serial modem (you can issue the commands ATI, ATZ, ATDT...).

The ATI commands say:

And, obviously, you can configure kppp and browse Internet and get e-mail: YAY!!!!!!

I use an "Omnitel/Vodafone" account; this is only in Italy, but I copy here the configuration so it can be a help for you. Note: the data/fax GSM service may need an abilitation from the telephone company (I had to ask for it - for free - from Omnitel some years ago).

So, this is the kppp configuration that I use:

modem and kppp:


Omnitel required my GSM number as userid, and no password (they always know what phone is connecting...!).

Well, kppp it's easier to configure than the bunch of pppd files. After configuring as above, I was ready to browse Internet and get/send email with Netscape.

The Omnitel charges for the Internet data connection 0.11 euro for the first three seconds and 0.11 euro for every successive 105.3 seconds. The ppp handshaking is about 15 seconds. Sadly, the "2800 Omnitel" service is only 9600 baud.

But, who cares about GPRS?!? Now I can get/send email from everywhere! SLURP!!!! ;-)

Other Nokia 7110 features (but I still cannot report success about using them with the A1370D notebook): the 7110 can "print" and "send card via IR" or "print the entire phonebook via IR" (these functions need only minimum IrLPT and IrOBEX support from host computer).

I was able to print a phonebook entry (vCard) on an HP LaserJet 4 via Infrared printing! Well, it is not that funny (all the fields were printed without formatting; Nokia 7110 firmware does not support more than that...)

I can get stored SMS messages via IrDA port, so I don't need to erase them manually on the phone and I can save them to disk. The source code of the program I wrote to do it is in this page.

I can send SMS messages via IrDA port. I needed only to install the gsmlib 1.6 library, and execute this command:

gsmsendsms -d /dev/gsmodem -C +393492000300 3479999999 "Linux is STUNNING"

The gsmlib is not specific for the Nokia 7110, but sending SMS is 100% ok. Note: the first number is the SMSC (SMS message center: depends on your mobile phone network!), the second is the destination number, and finally follows the quoted string of message text (up to 160 characters). The gsmsendsms also correctly translates the accented ISO8859-1 vowels to Nokia GSM alphabet. YAY!

Compactflash and PCMCIA IDE/ATA disks

The digital camera Kodak DC210+ stores images on a solid-state disk (compactflash card) SunDisk SDCFB-8 of 8 Mb (it nicely declares to have even an 1 Kb cache).

The Linux kernel 2.4.3 does not like the IDE-cs module (it seems to me that the ide-cs code is not compiled!), so I had to get the ide-cs.o module from Linux kernel 2.4.1; click here to download it. After "inserting" it (insmod ide-cs.o), the cardmgr program will recognize correctly that SunDisk compactflash card (and every other PCMCIA ATA/IDE disk), marking it as the third IDE interface (ide2, that is hde/hdf) and you can /bin/mount normally.

So I created a /mnt/kodak directory and added in the /etc/fstab this line:

/dev/hde1  /mnt/kodak  msdos  user,noauto,nosuid,nodev  0 0

After inserting the compactflash card in its PCMCIA adapter in one of the two PCMCIA slots of the notebook, the "mount /mnt/kodak/" command makes me see it as an extra hard disk.

Here are the /var/log/messages messages:
cardmgr[9538]: starting, version is 3.1.20
cardmgr[9538]: watching 2 sockets
kernel: cs: IO port probe 0x0c00-0x0cff: clean.
kernel: cs: IO port probe 0x0800-0x08ff: clean.
kernel: cs: IO port probe 0x0100-0x04ff: excluding 0x480-0x48f 0x4d0-0x4d7
kernel: cs: IO port probe 0x0a00-0x0aff: clean.
cardmgr[9538]: initializing socket 0
kernel: cs: memory probe 0xa0000000-0xa0ffffff: clean.
cardmgr[9538]: socket 0: ATA/IDE Fixed Disk
kernel: hde: SunDisk SDCFB-8, ATA DISK drive
kernel: ide2 at 0x100-0x107,0x10e on irq 10
kernel: hde: 15680 sectors (8 MB) w/1KiB Cache, CHS=245/2/32
kernel: hde: hde1
kernel: ide_cs: hde: Vcc = 3.3, Vpp = 0.0
cardmgr[9538]: executing: './ide start hde'
kernel: VFS: Disk change detected on device ide2(33,1)
kernel: hde: hde1
kernel: VFS: Disk change detected on device ide2(33,1)
kernel: hde: hde1

WARNING: the pcmcia driver authors claim that the pcmcia-tree in the 2.4 kernel source will be "experimental" for all the 2.4.x series development. So, it's better use the pcmcia-cs package than the kernel stuff. I just compiled the 3.1.26 version, and used successfully its ide-cs.o module in the Linux 2.4.4 kernel. Look below for other details.

Let's dance Samba!

We can network mount a Windows drive using the Samba protocol.

I have an old PC-486 (Amd 486/120, 20 Mb RAM, Tseng ET4000 video card, NE2000-compatibile network card); on its 545 Mb hard disk there is a Windows 95 installed. I can mount that disk as a normal read/write directory from Linux.

The steps described below really did it, so you can hope to get them to run also for recent Windows versions (you can do it even with Windows 3.1, but you need then to install there the TCP/IP stack and the Win32s extensions); for all details, please look at the great Samba site: here I just describe how I did and do not want to put another entire Samba-documentation site...!

Before starting, let's give a name and fixed network address to the machines:

These "192.168.x.x" are free for private use and everyone can use it in his home network. If you don't have dozens of computers, you really don't need all the nameserver/domain/etc stuff: it's easier to have a detailed /etc/hosts file listing all fixed addresses.

Let's now configure the BIDONE.

Before starting, I invite you to check that also the network card of the BIDONE is correctly installed, and the network cable is correctly set up. In my PC486 the card it's an old NE2000-compatible with addr=0x300 and IRQ=11, and in next lines I'll refer to this. I have to specify to check the card because half of the "Windows network problems" for which people asked me for help, were because of incorrect configurations...!

Well, from the Control Panel get to Network and configure only:

Note: I refer to an Italian edition of Windows'95, so I tried to translate Italian menus and names in English... but I'm not really sure I succeeded. Anyways, I think that if you know what you are doing, you won't need more help...!

Warning: the TCP/IP needs a light configuration - since I am using fixed IP addresses, I really don't need to fiddle with name-server, DNS's and other services. So, complete the TCP/IP protocol configuration with these Properties:

Still in Control Panel --> Network, in the Identification section, I placed the computer name (BIDONE) with workgroup=WORKGROUP and description="Ye Olde 486" (well, connecting just two computers doesn't really need description and workgroup!).

Then, in Network --> Access control, choose sharing.

Now we can reboot BIDONE. Don't forget to go to My computer to enable hard disk sharing with the name ardisco and read/write permission without password.

In the /etc/hosts Linux file (and in the \WINDOWS\HOSTS Windows file) we can finally assign names and IP addresses, adding these two lines: barile bidone

Now in Linux, if you activated the network card (was it supported from kernel? was it configured? was it checked at boot? etc), you shouldn't need to activate it again using the command ifconfig eth0 barile up - just verify with the ping command that all configuration is OK.

That is: in Linux we can try to use ping bidone, and in a DOS prompt window of Windows you can use ping barile (or ping so you can understand if a computer "hears" the other. If you see packets exchanged in a few milliseconds, then we can finally mount the remote directory from Linux, using this command:

smbmount //bidone/ardisco "" -c "mount /mnt"

Beware the syntax! I wrote really "" (that's a fake password) and "mount /mnt" (that's a command string).

After doing this, the ardisco share of the bidone computer is mounted on /mnt directory of Linux (choose one for your needs).

Let's ask for ls -al /mnt and... ZOT!!! The ardisco is ready there!

If you activate WinPopup of Windows, you can also send winpopup messages using the command smbclient -M bidone 

What? You complain because I didn't tell you to compile Samba modules in the kernel? To import Samba shares, you don't need Samba modules in the Linux kernel!

Performances. I got about 380 kb/sec of throughput - this is because of hardware limits and Win*ows limits, but it's quite good to make backups and decent file exchange.

Well, that's all: happy Samba dancing!

Connecting this Linux box to an Apple iMac & LaserWriter via Ethernet!!!

Yeah, I did it! This applies not only to this Asus notebook, but to every Linux machine with an Ethernet network adapter.

The iMac here is a 233 MHz PowerPC G3 machine with 32 Mb RAM and Finder 8.5 installed, with an Apple LaserWriter Connect 360 printer connected to its Ethernet port.

Stage 1. Check that pam-devel, quota and lsof packages are installed (well, these shouldn't be really always needed; I read about them only in documentation and contrib directories in the netatalk package).

Have a compiled kernel with AppleTalk DDP support in it, and boot with this kernel. If you compiled DDP as a module, check with the modprobe ipddp command (while root user) and look at it with the /sbin/lsmod program.

Well, you surely have Ethernet support in the kernel (on this notebook you need only the sis900 driver).

While in root user, modify the /etc/services file adding these lines below (the tcp and udp lines should be already there, so you should add only the ddp lines):
# appletalk support

rtmp            1/ddp           # Routing Table Maintenance Protocol
nbp             2/ddp           # Name Binding Protocol
echo            4/ddp           # AppleTalk Echo Protocol
zip             6/ddp           # Zone Information Protocol

afpovertcp      548/tcp         # AFP over TCP
afpovertcp      548/udp

Get the latest netatalk package from its official Sourceforge site (click here to go to download page for; in the Mandrake 8 distribution there is no netatalk stuff). I found on SourceForge the "1.5-pre8" version; the final "1.5" should be out soon (look in the news section).

While in root user, uncompress it somewhere (I suggest /usr/src), configure it (./configure), compile (make) and install it (make install) without adding anything to the default options.

Stage 2. Let's go now in the /usr/local/etc/netatalk directory (on different distributions it may stay in some other place; I found it just there). Let's add to the atalkd.conf file (that by default has only comment lines) a tiny line containing only "eth0" (without quotes!).

We have to modify also the afpd.conf adding this line (all in one single line!!!):

- -transall -uamlist,, -nosavepassword -loginmesg "Welcome to the Linux jungle!"

We may add now to the AppleVolumes.default file (that by default contains only a definition of a generic user home directory) an useful line:
/tmp    "Temporary files"

We can now connect with a "cross-over" network cable the Ethernet ports of the Asus notebook and Apple iMac, and finally (while root user) start the /usr/local/sbin/atalkd daemon and wait a bunch of seconds that it gently places himself in background. Since my Ethertalk connection is not a permanent connection, I didn't place it also in one of the /etc/rc.d/ startup files.

Stage 3. Using /usr/local/bin/getzones -m we should know about the Apple network "zone", but it simply fails (timeout) because this iMac is not connected to other Apple computers/hubs. Well, the interesting thing is that the /usr/local/bin/nbplkup command shows us system and printer names actually ready to use in the AppleTalk local domain; here's what I got on the screen (note that you don't have to type in also the "$"... it's just the command prompt!):

$ nbplkup

BOMBOLO:AFPServer 41201.173:249
BOMBOLO:ARA - Client-Only 41201.173:2
BOMBOLO:PPCToolBox 41201.173:251
BOMBOLO: Power Macintosh 41201.173:252
BOMBOLO:Workstation 41201.173:4

There's also a ping-style command called aecho: using /usr/local/bin/aecho SystemName we can see response times on the network (below, we see times less than one millisecond):

# aecho BOMBOLO:AFPServer
14 bytes from 41201.173: aep_seq=0. time=0. ms
14 bytes from 41201.173: aep_seq=1. time=0. ms
14 bytes from 41201.173: aep_seq=2. time=0. ms
14 bytes from 41201.173: aep_seq=3. time=0. ms
14 bytes from 41201.173: aep_seq=4. time=0. ms

----41201.173 AEP Statistics----
5 packets sent, 5 packets received, 0% packet loss
round-trip (ms) min/avg/max = 0/0/0

Now we can also start the AFPS server (/usr/local/sbin/afpd) and, on the iMac, call the "Network browse" (I don't know its English exact name because I have here an Italian Finder version) of the "apple" menu. YAY! Here are the Linux directories: home and temporary, ready for browsing.

This iMac here has an Internet PPP default connection, so it tries - before connecting to Linux box - to dial to the Internet; I simply aborted this operation and immediately I got Linux directories (as of AppleVolumes.default definitions of above). Since it does not use passwords, we understand that it is a "guest"-type login (it will use the nobody Linux username, the only that can write only in the /tmp directory); I won't study and describe a password-based login because I've no more spare time...

Well, click on one of the Linux directories and drag'n'drop files from iMac desktop to Linux and from Linux to iMac: it runs!!!

Have a look at the Linux Ethernet port status: it has been set up as "EtherTalk Phase 2":

$ ifconfig eth0
eth0 Link encap:Ethernet HWaddr 0E:EE:1E:1E:8E:5E
inet addr: Bcast: Mask:
EtherTalk Phase 2 addr:65280/117
RX packets:16 errors:0 dropped:0 overruns:0 frame:0
TX packets:81 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 RX bytes:1093 (1.0 Kb) TX bytes:3850 (3.7 Kb)

And now, the printer. The one here is an Apple LaserWriter Select 360 with a Farallon EtherMac iPrint Adapter (a device to convert the old-style proprietary Apple ports to Ethernet ports: and we just need Ethernet ports).

In order to print files, we need only the atalkd daemon (the afpd daemon is needed only to let the iMac see the Linux resources; it must be loaded after the atalkd, not before).

Using nbplkup and papstatus we can see the printer on-line and idle (but only after its full initialization after the "cold boot" - of about 30 seconds). Using the pap command we can print a PostScript file:

$ nbplkup
LaserWriter Select 360:LaserWriter 65280.122:128
$ papstatus -p "LaserWriter Select 360"
status: idle
$ pap -p "LaserWriter Select 360" /tmp/

Sadly, because of a nasty ugly bug of the 1.5-pre8 version of the netatalk package, soon after the successful printing I saw a system hang and I needed to hard reset the notebook (this is because the atalkd uses root privileges, and a bug there is really able to hang the machine). I hope next versions will clear out this bug: it ain't nice to get a system hang after a successful printing...!

Other folklore: my Postscript file had LF as line terminator, while the ones generated by iMac software had CR only - but the print tests succeeded in any case (I still don't know if it was so because of an atalkd filter or it is a LaserWriter feature). The Select 360 Postscript, as generated from iMac software, is not correctly displayed by the gv program, but I think this is not more than technical curiosity...

Happy "Appling"!!!