my installation is tired

[gapan]

Maybe it would be a good idea to run sync with a cronjob, perhaps hourly, maybe even more often.

[mimosa]

I just ran sync and now the sound works again! Thanks gapan!

However, the basic problem remains, but without being able to boot a rescue disk, there's not much I can do. That is a BIOS problem and nothing to do with Salix, so I will have to look elsewhere for solutions, or find a kind person to lend me their external CD drive

Meanwhile,thanks for the general lesson that I need to be much more concerned about filesystem integrity

[Shador]

You can try shutdown -Fr now which should reboot your system enforcing a file system recheck. Furthermore I guess you might be able remount your root read only in runlevel 1.
An installation disk is enough by the way, too, if you want to do a simple fsck.

Weird that sync fixed your sound problem. It's supposed to write back all data to every attached filesystem.

Regarding your problems booting the LiveCD: You certainly need an initrd and some other parameters for the kernel. Here's what's in the default grub2 menu:

Code: Select all

  linux /boot/vmlinuz root=/dev/ram0 rw quiet lang=${locale} keyb=${kb} ${runlevel} ${toram} ${numlock} ${synaptics} ${env} changes=slxsave.xfs 
  initrd /boot/initrd.gz

Just strip all variables, replacing locale and kb as appropriate and runlevel with 4. I'd be interested in which way grub2 fails on your system. I had grub2 fail on one machine too when booting from the CD, but didn't find the time to investigate the issue with the grub devs.

[mimosa]

So I should type that in at the grub prompt just as it is with the substitutions you mention? How does grub know it is supposed to be looking in the USB? Or presumably if it were grub legacy (which I know a lot better than grub 2, not that that's saying much):

Code: Select all

root (hd1,0)

kernel /boot/vmlinuz root=/dev/sdb1 rw quiet toram lang=en setkb=br ...
[in fact I'm a bit lost here already ... keyboard and locale aren't crucial anyway because I can set them later ... ]

initrd /boot/initrd.gz

boot

I had grub legacy on this machine, but installed grub2 just now to see if I could reproduce the effect I got before where the USB, incomprehensibly, booted on typing enter (or was it ESC?) at the grub prompt. No dice, so I will probably put grub legacy back, but I'd be glad to try the experiment you mention with grub2 first if you give me exact instructions, or confirm the existing ones are idiot-proof. Then, if it works, I suppose I'd better actually keep grub2 and learn how to use that too. Incidentally, I have Slackware-Salix native lilo on my other identical machine, and the initial two-tone screen is very beautiful, with the intuitive options "Linux" and Windows" and no endless failsafes and memtests. My wife, who is not curious about computers, loves it.

As for the magical effect of sync, it may have been serendipity (or an intermittent hardware fault; I live in a very humid subtropical climate with hungry insects with Catholic tastes, too) rather than your traditional causation, but if in doubt I'll give the credit to gapan.

[Shador]

It's important that you leave the root parameter at /dev/ram0! I don't know how exactly the the linux-live scripts determine the device it's being booted from but I guess it's some information the bootloader passes to the kernel.

[mimosa]

Ok, but isn't there an equivalent of

Code: Select all

root (hdX,Y)

in grub legacy, where you tell it where to look in the first place? Otherwise, I could surely be booting any of the OS's on the hard drive, too. Just to confirm, I am typing at the grub prompt, which I reach by typing "C" at the menu inviting me to choose an OS from the hard disk. But I want to boot SalixLive on a USB, which unfortunately will not work automatically under this BIOS.

Then hopefully it will fail in an interesting way.

[Shador]

You just change environment variables like with bash. Just root= should work but "set root=(hd0,1)" definitely does. But keep in mind that partitions are now enumerated from 0 with Grub2. (i.e. 1st partition is 0)

[mimosa]

I didn`t manage to get Grub2 back - the configuration program appeared to work, though unlike the one for grub legacy it didn`t ask me any questions about other OS`s, just whether I wanted Grub in the MBR or elsewhere; but when I rebooted, grub legacy (which I had got back earlier) was still there. Since the immediate problem was booting from USB, I left it at that. I then managed to get a Ubuntu stick going and used fsck, which gave my Salix partition a clean bill of health. but couldn't cope with the Windows one which certainly seems to be implicated in the problem. Just to recap:

Code: Select all

FATAL ERROR: Bad primary partition 1: Partition ends in the final partial cylinder
                                                                                                                     Press any key to exit cfdisk

I booted Salix live like this:

Code: Select all

root (hd1,0)
kernel /boot/vmlinuz root=/dev/ram0 rw
initrd /boot/initrd.gz
boot

which was bumpy but worked. It didn't like the parameter toram, though, as has happened to me before with I think Puppy, complaining of lack of room, though this machine has 1GB. I mention it because I would be quite interested in knowing what the correct thing is in grub legacy language, as this could then be included in my menu.lst as a boot-Salix-USB option.

fsck under Salix gave a similar problem (mount says the fs on sda1 is "fuseblk"), and I couldn't find anything that looked relevant with gslapt:

Code: Select all

root[one]# fsck /dev/sda1
fsck 1.41.8 (11-July-2009)
fsck: fsck.ntfs-3g: not found

fsck: Error 2 while executing fsck.ntfs-3g for /dev/sda1
root[one]# 

Apart from the odd unpredictable glitch, everything seems to work more or less ok, including the apparently totally compromised Windows partition - not that I use that very much - but it's obviously not a desirable situation in the medium term. Meanwhile, I have backed up, and could in principle repartition and reinstall everything, but there is a lot of data involved and anyway, this shouldn't have happened in the first place. My research tells me that fsck shouldn't even be needed for Reiser and other "journaling" fs's, which I imagine means they are robust enough to take the odd powercut (had one just now and lost this post, please excuse the telegraphic style). Is that right?

[gapan]

No, every file system needs fsck. Some do not get corrupted that easy, but it's possible that they do get corrupted at some point.

Concerning your ntfs partition, I've experienced something similar at a friends PC. Unless I repaired the ntfs partition from inside windows, linux couldn't mount it. The problem was also caused by hard shutdown too.

[Shador]

A filesystem can only be nearly completely protected at the cost of (severe) performance drawbacks. As I already mentioned you can change the mount options of your fs driver (at least for ext3/4; others probably too) to behave that way or at least more that way.