Linux on a Toshiba Satellite 1800-712 (23/7/2002) (update) (update2)

I succeded in installing Red Hat 7.3 on a Toshiba Satellite 1800-712 . My choice fell on RH just because it was the first distribution I had at hand, no deeper reason. All in all, the installation was pretty straightforward and most of the hardware I tested was recognized by the installer so, just out of the box. I'm writing this notes mainly to report that it worked and to note down the few additional post-installation steps which were necessary for me. I didn't yet try out extensively the hardware, but I'd say that the bottom line is, save the winmodem, enough of the machine works well under linux.

Hardware information:

I bought the computer in a chain store in Italy on 10/7/2002. I found no information on the net about this particular -712 submodel, however its configuration is pretty similar to that of other 1800- series Satellites. There is plenty of information about them, and this is why I had a rather easy game.

Main sources of information:

• LINUX on Laptops : in particular the links to pages about linux on specifical Satellite 1800s with the same chipsets.

• Toshiba Service and support database: though it doesn't list this -712 submodel, it's a very good source for a) configuration information of related variants, b) quirks about installing linux (or BSD, or Beos, or Solaris) on this production line. Toshiba officially supports only another OS, but the fact that they published this site mends the bad impression. 

The configuration of the machine, as I got it,  is the following:

(untested means only that I didn't test it. Check at the above mentioned links for further information)

Linux installation notes:

The Red Hat installer recognizes most of the hardware. I had to choose manually only in a couple of occurrencies:

• as pointing device, I chose ALPS pointing device instead of PS/2 mouse (though, that might work too), 3 button emulation.

• as video card, the installer suggests VESA. The installer knows about Trident Aladdin CyberBlade Xp, though, and this is what I chose in my first instance. The video driver can be replaced with a better one, see below . 

• as monitor, choose generic 1024x768 Laptop panel.

• As for partitioning the hard disk: I did it, given my usual messy style, in several steps and second and third thoughts, and ended up with the partition table below. I understand that by using standard partition resizing tools, such as FIPS provided on the RH disk, the process should be straightforward. Note: it is always, always good to keep a linux boot diskette at hand, just in case some re-installer writes on the MBR: then you just boot from the diskette, mount /boot , and rerun grub-install (supposing you choose grub as bootloader) to restore the MBR.

partition table:# /sbin/fdisk -l /dev/hda

Disco /dev/hda: 255 testine, 63 settori, 2432 cilindri
Unità = cilindri di 16065 * 512 byte

Dispositivo Avvio    Inizio    Fine   Blocchi   Id  Sistema
/dev/hda1   *         1       800   6425968+   b  Win95 FAT32
/dev/hda2           801       813    104422+  83  Linux
/dev/hda3           814      2432  13004617+   f  Win95 Ext'd (LBA)
/dev/hda5          1801      2350   4417843+   b  Win95 FAT32
/dev/hda6           814      1800   7928014+  83  Linux
/dev/hda7          2351      2432    658633+  82  Linux swap

Partition table entries are not in disk order

The rationale for making /dev/hda5 a Win95 Fat32 partition was to have a common exchange area between two OSs. I'm quite skeptical about this necessity, however, and I might change this scheme later.

• Getting rid of that other crapped OS: as said above. Originally I got a demo machine, and messed up a little in the process of resizing the partition. There must be two places where the relevant partition information is written onto the disk, as at one stage I found myself with fdisk telling me that /dev/hda1 was 6.2 Gb and df telling that it was 18.9 Gb, and similarly under the other OS. So I decided to use the Toshiba Software Restore CD to reformat /dev/hda1 to the desired size and to fill that with the original software, and yes, my Restore CD lets me do it without pretending to reformat the whole disk. At the moment I kept the partition without accepting that f***ed licence, I'll end up reformatting it someday. For the moment I keep it for when  I'll be ostinate enough to try the process described in http://www.freego.it/articles.php?show=4 . Note, however, that that system boots that OS up only if in the BIOS you set the hard disk in " Enhanced IDE mode".

Post install fine tuning:

Accelerated trident driver:

A direct replacement for the video driver is available at http://www.xfree86.org/~alanh/ . As the kernel provided with RH 7.3 is 2.4.18-3, i.e. advanced enough, it is sufficient to replace /usr/X11R6/lib/modules/drivers/trident_drv.o and to restart the X server. I didn't perform specifical speed tests, as a rule of thumb I'd say that dragging a solid window across the desktop is twice as responsive as before. However, the replacement driver has problems with the out-of-the-box libmpeg. With this driver, noatun / kaboodle / plaympeg drive the card to an unusable mode whenever I try to display mpegs, and I have to reboot to recover.

Post-install changes

• chmod +s /usr/bin/amd  - amd works both putting the machine in standby and in suspend. A third low consumption mode is achieved when closing the cover; this only turns off the backlighting of the screen, and it is due to the BIOS. I understand that "suspend" means putting the RAM in a standby mode and stopping the CPU; there seems to be no such think like a dump of the RAM to a hybernation partition on the hard disk (I'm not observing any heavy disk activity before suspension). If it is not like this, I'll discover it once my filesystem will be eventually trashed, as I didn't allocate any hybernation partition on the disk. 

• chmod +s /usr/bin/cdrdao /usr/bin/cdwrite 

Software added out of the distribution:

•  LyX 1.2.0

Software of the distribution which didn't work for me:

• rpmfind, gnorpm: they are not able to connect to the rpm server, though they work perfectly as standalone applications.

• xine crashes (even upgrading to 0.9.12)

Remaining hardware/software issues:

Modem:

The soft modem, based on a Lucent SCORPIO chip, is marked as non-supported even by www.linmodems.org.

Mouse and sound after suspend

apm works and is able to put the computer in suspend mode, etiher calling apm -s from the shell, or choosing from a graphical menu like that of gdm or klaptop . However, after resuming the sound server of kde is either jerky or mute, and some of the mouse actions are crippled. In particular, it seems that tap-touchpad is not anymore processed as a mouse click, it is not possible to point and drag, and so on. I didn't understand to which extent the BIOS is to blame and to which the upper layers. However, adding these lines to /etc/sysconfig/apmd , the sound behaves better (in the next window manager session):

RESTORESOUND="yes"
RESTORESOUNDPROGS="yes"
SOUNDMODULES="sound sr_mod trident ac97_codec soundcore"

The mouse behaves normally after a kde logout/login, which hints at a problem solvable within kde.

Reboot

Sometimes a reboot from linux shuts correctly the system down, but then freezes the machine with a black screen and the fan at full power, while some other time the process works perfectly. I understand that it could depend from either the BIOS settings or amd, but didn't yet isolate the reason and its cure. I observed that reboot succeeds if the machine has been suspended and revived. I've seen the problem reported for another Satellite in dejanews. When the machine freezes, the only choice is to press for 1 second the power button; then the machine peeps for the next 15 seconds, and finally powers off. A further press of the power button turns the machine normally on again.

External mouse

Hot plugging an external mouse is ok, but hot unplugging leaves the trackpoint unseen until the next X server restart.

Hard disk mode

• Speed (/sbin/hdparm -Tt /dev/hda) is dependent on the BIOS setting "Battery Save Mode" (which also affects the speed of the CPU btw). In "Low power ", hdparm reports ~15Mb/sec for buffer-cached read speed, while this increases to ~80Mb/sec in "High Power" mode. I have seen the Toshiba software restore disk reporting up to 120Mb/sec during the software restore process . According to /proc/ide/ali, the bus is working at 33 Mhz. Passing idebus=66 to the distribution kernel doesn't help, it lowers the buffer-cached throughput to 57Mb/sec.

• According to hdparm, the disk should support mode udma5. However, /sbin/hdparm -d1X69 /dev/hda seem to have no effect, and the highest mode in which I succeed to put the disk is udma2.

CDRW/DVD

I have been able to read and to burn CDs (used koncd). I didn't find yet a software suitable fror playing DVDs.

Extended keys

F-keys and Fn-F1-F5, Fn-F10-12 work just out of the box.
Note that Fn-F2 switches between the "Battery Save Modes" , thus affecting the performance of the hard disk .  The current mode is hinted by the brightness of the screen, which becomes a little dimmer in  "Low power".
Fn-F5 blacks out and revives the screen, I suppose it is switching between internal and external video.
I didn't experiment with the exotic "internet key" or with the "cd player keys" out of the keyboard; it is quite possible that they can be made usable for some real purpose.

Update - 8/9/2003 (perhaps will be expanded someday)

Since then I have succeeded upgrading to RedHat9, but I didn't have had the patience to sort out once more all the miscellaneous problems. I.e., sound is slightly worse, I tried but not succeeded to recompile a new kernel with UDMA5 support, there are problems of random sleep, didn't try out the network, and so on. I read on the net that things in reality improved, it's just that I didn't invest time in fine tuning. Say, the softmodem should now be operable. I have been able to play DVDs, trying out various different versions of xine and ogle, but alltogether the result was quite jumpy (up to ~12fps, which allows to see something but falls behind). I'm told however that the right combination of X driver, kernel, library does.
Useful links:

• once more the Toshiba Service and support database, which contains a lot of new information,

• Tomas Pospisek page

• jxh's page

Update - september/october 2008

By now the laptop's batteries are almost gone [they are reported as 100% full, but last only 10 min at most, likely sign of one element with reduced capacity, unless the ACPI went nut], and the key 'g' triggers almost spontaneously. I wanted to give the last chance to the old beast, installed Xubuntu 8.04 off the single cd i386, and kept populating the software as soon as the ADSL network was up. That practically resurrected the laptop. I was initially particularly happy of the lightweightedness of Xfce (I could still run decently off 256Mb with half RAM free), but started to have doubts after populating a little the system software and the xfce panel applets. However, practically everything I've tried so far worked out of the box, which is remarkable.

Only one thing required a minimal post-install tweaking - the video card is not recognised automatically, and the display gets stubbornly into 800x600 vesa mode. The cure is amply described in the ubuntu forums (e.g., in this thread and this thread). For me, I simply needed to know that a) the configuration program is called displayconfig-gtk; b) you have to save the resulting xorg.conf somewhere, and then to copy it in /etc/X11.

Video output would then come shifted by 4 pixels to the right, leaving a tiny black bar at the left of the display. What fixed it was setting the initial text mode to a higher resolution, like appending vga=791 or vga=773 to the kernel options in /boot/grub/menu.lst (which takes the place of /boot/grub/grub.conf in other distros). I had no success in fixing this by playing with either Modeline or Option XvHsync in xorg.conf, which I initially tried, and anyway the higher text resolution is just nicer.

As for other kernel options suggested on jxh's page, which I was using in the past:

• idebus=66 is, according to what I read on net posts, apparently useless: the bus speed is 33MHz, and the option is only meant to convince the kernel of something different, in case of odd speed not automatically recognized; in any event, it does not make the IDE interface faster.

• agp_try_unsupported=1: no idea of whether it has any use here

• tosh_fn=0x62 possibly helps in having some of Fn-Fx working
  
  
  

As for other peripherals:

• I quickly managed to setup my ISP adsl modem (just copy the 3 relevant firmware files in /lib/firmware/ and plug in the usb, literally, even though it took me some wrong attempts just to get there, more older ramblings here and here), and

• my cheap webcam (a Vgear talkcam something) was surprisingly immediately recognized by cheese (it never happened me with other installations).

• Afterwards I bought a Ralink PCMCIA wireless card, and that too just worked out of the box. Well, I got several kernel panics while trying to download package updates, one afternoon, but otherwise net-browsing seems to work smoothly...

Well' I heard well of Ubuntu...

Things I'm not yet happy about

Video driver related

I remember that I made several attempts at the time, swapping various trident kernel models I got with different distributions and on the net. With the new installation I'm running again in familiar problems, but I don't remember/didn't really document how I solved them at the time.

• video-intensive apps (xine, totem) could at time crash the screen as well, completely locking the system, with vivid striped/plaid patterns on the screen. Also this seen described on the net. Only the emergency ten-second press of the power button had any effect in that situation. This used to happen now and then, but seems much more frequent in this installation.

  • Trident module specific options which don't work, for the record:

  Option "PciRetry" “on” # ? no effect seen yet

  Option "ShadowFB" "on" # crashes the video driver and hangs the laptop

  Option "AccelMethod" “EXA” # puts all consoles in an unusable video mode, but at least doesn't hang the machine

  ACPI (Sleep modes, reboot)

The idea is to have them working as handled by gnome-power-manager, since gnome-power-manager can be used as toolbar applet, which is convenient, the effect of events like closing the lid or pressing the power button can be configured, and all is neatly integrated in the gnome session. Out of the box, these are the problems:

• Suspend apparently works, but when resuming the screen is unusable. Too bad given the present state of my battery. Namely, the laptop wakes up with fan running at full speed, screen showing a blurred, bright color shade, with some moire, changing tonality and brightness during the first several seconds. Very impressing because it really looks as if the screen is burning out. I remember having seen the same effect described on the net. Ctrl-Alt-Backspace doesn't reset X to any usable mode. However, the laptop is not dead, as I can Ctrl-Alt-F1, log in as blindly as root and shutdown (or just Ctrl-Alt-Del to reboot, which hangs anyway, but shutdowns cleanly). I remember that the problem was completely solved for me up to the last RH9 installation, but now how I did. Different kernel? Kernel recompiled with some Toshiba specific flag? Most likely, just a lucky special trident.o. Anyway, the solution seems at hand, since running pm-suspend as root, without any optional quirk, just works fine. I have to check what is really going on with hal, freedesktop, etc., and possibly send a patched fdi. On the net I've found about this bug. The fists stop about pm-suspend quirks is obviously this one.

• Hibernate in fact causes a complete shutdown; when the laptop is powered again, the system logs in as the last user and restores the last session. Not exactly as efficient as I would have expected, but I understand that this is what it should.

• Reboot causes the system to shutdown, and then hangs with a black screen and fan running at full speed; only the 10 sec press of the power button turns completely off the machine. It was like that in my previous installations, and I've also read on the net of it as a known problem, probably unsolved because of lack of proprietary details about the ACPI. My observation was that, when suspend used to work, if the machine was suspended at least once since the linux system was up, then reboot was working as expected. It is not anymore the case now, even suspending with pm-suspend and resuming cleanly, tghat is the laptop would sdubsequently hang at reboot.

  Sound

  Essentially works (e.g. if I play sound files with totem, if totem doesn't crash the screen – I should try with some other audio player, the basic Xubuntu installation doesn't seem to provide one), but I have no session manager sounds, and moreover I have to open the mixer each time to rise the zeroed PCM volume. Some sound problem existed in my previous installations as well; for instance with RH9 one channel was always muted at the beginning of every session, and only touching the volume control unmuted it.

  Performance

  Performance, as I already knew, but forgot, is enormously influenced by the "battery saving mode". Had to remember to set Maximum performance in the bios. Expecially hard disk speed is highly influenced (see above), and since with only 256MB the machine is often swapping, that is slow. As for the hard disk speed, I'm still puzzled. According to jhx, it should be possible to achieve a buffered read throughput of the order of ~100Mb/s, and udma mode 4 or 5. No matter what I do, I can't get beyond ~50Mb/s and udma2. Support for the Ali 15x3 is enabled in my kernel, though I didn't find the “enable DMA as default” option mentioned in the module description. Another observation is that

# hdparm -i /dev/hda

/dev/hda:

Model=TOSHIBA MK2018GAP, FwRev=M1.42 A, SerialNo=42C75599T

Config={ Fixed }

RawCHS=16383/16/63, TrkSize=0, SectSize=0, ECCbytes=46

BuffType=unknown, BuffSize=0kB, MaxMultSect=16, MultSect=off

CurCHS=16383/16/63, CurSects=16514064, LBA=yes, LBAsects=39070080

IORDY=on/off, tPIO={min:120,w/IORDY:120}, tDMA={min:120,rec:120}

PIO modes: pio0 pio1 pio2 pio3 pio4

DMA modes: sdma0 sdma1 sdma2 mdma0 mdma1 mdma2

UDMA modes: udma0 udma1 *udma2 udma3 udma4 udma5

AdvancedPM=yes: unknown setting WriteCache=enabled

Drive conforms to: Unspecified: ATA/ATAPI-1,2,3,4,5

What Tomas Pospisek reports differs in that Config is { HardSect NotMFM HdSw>15uSec Fixed DTR>10Mbs }, BuffType is DualPortCache, BuffSize is 1768kB, tPIO min: is 240. The last two lines of the hdparm report are also somewhat different, bu that might be due to different formatting of hdparm across versions. Either his laptop had a better disk than mine, or there is some parameter I don't know how to set (more likely, as I noted above that I remember having seen a report about 120Mb/s once in Windows). I can set MultSect=16 with hdparm -m, but that alone doesn't help.

cpu clock scaling is apparently not supported. Maybe it would be a matter of recompiling the kernel with the proper options, but I didn't find them yet. According to my present options (Celeron coppermine), it should, but the scaling applet CPUFreq complains.

• Recompiling the kernel, I managed to have some sensors supported. Despite hdparm being apparently able to read the hard disk temperature, the sensor applet complains about.

  I got toshutils, but I am not able to change all settings, like fan speed or display brightness. One observation is that the power mode has to be set as “user” in the bios, for some of them to be settable at all.

  Xorg cpu usage is significant for the tiniest X action, like moving the mouse around; panel applets causing frequent refresh, like xfce4-systemload are taking constantly some 10% cpu each. What is going on? Maybe I've got an unaccelerated trident driver?

• IrDA support might be working with this distro, I haven't yet tested. The boot message reports about Superlogic chip found, IrDA not present, IRPcc (?) working.

  Stupid question, but how do I set location dependent proxies for all applications? At work I sit behind one firewall if I use the ethernet cable, and behind another if I use wireless; at home I have no firewall if I use the ADSL connection to my ISP. I know about firefox proxy setting and /root/.synaptics/synaptic.conf, but how to make the process automatic? What about ping, telnet, ftp,...

Serial port not working

I've just discovered this (11/1/2009). The problem is documented (this question and this bug entry), and apparently due to the kernel confusing the infrared port with the serial port. I've tried a few kernels (2.6.24-3, 2.6.24-19, 2.6.24-23) during the process of updating them, but none works.

The symptom is:

echo blahblah > /dev/ttyS0
/dev/ttyS0: No such device.

Additionally, /sys/bus/platform/devices/serial8250/tty/ shows ttyS1, ttyS2, ttyS3, but no ttyS0. The serial resource indeed appears under /sys/bus/pnp/devices/00:0a (not 02 as reported for other laptops).

One of the proposed workarounds for other laptops is to disable IR related stuff in the BIOS, but I don't find any related entry here.

Another proposed workaround is to “bounce” the port by writing 2 and then 0 to /sys/bus/pnp/devices/00:0a/power/state, but this virtual file does not exist in my installation.

Passing the boot option pnpacpi=off, I do get a dmesg line

ttyS0: LSR safety check engaged!

whenever I attempt to access the port, and some setserial help message at boot (as if setserial was called with wrong parameters, have to check what is really in the init.d script) which I haven't yet been able to read.

Of course, if anybody has further suggestions, or just wants to say hi there, I am reachable at this address (remove the nospam).