I updated the v2.6.18 Linux kernel on amigaone-linux.sourceforge.net. It contains a workaround for network lockups (for the built in ethernet only), where the machine locks up after the first dozen of MBs have been transfered.
Unfortunately I couldn't fix the network data corruption (my fix didn't work as expected, although it improves ethernet operation). It looks like I need to find a way to debug this problem (I'm open for suggestions! )

BTW: already installed v2.6.18 MOL modules should still work.

_________________
A1SE: G3@600MHz, 2GB, 1GBit network card
A1XE: G4@933MHz, 2GB, refitted AC'97 codec
microA1: G3@800MHz, 1GB

- A1 Linux support -

@Geri

Thanks mate. Will install it this week :)

@Geri

I tested this 2.6.18 and it stops booting after message: "hda RDSK hda1 ... hda6". It's like any other non-2.6.8 kernel I've tested. 2.6.8 works OK. My Linux partition is hdd9. Those hda1-hda6 are my FFS/SFS partitions, 40 GB total.

_________________
Rock lobster bit me. My Workbench has always preferences. X1000 + AmigaOS4.1 FE
"Anyone can build a fast CPU. The trick is to build a fast system." -Seymour Cray

@Geri

How do we now update modules. I get the message that update-modules has been depreciated. My machine booted fine but net transfers caused a crash again but the modules may not be updated properly.

Cheers
ace

@acefnq

Quote:

acefnq wrote: @Geri How do we now update modules. I get the message that update-modules has been depreciated. My machine booted fine but net transfers caused a crash again but the modules may not be updated properly.

Update procedure for the modules:

Quote:

1. become root 2. cd / 3. rm -rf /lib/modules/2.6.18 4. tar -xzvf modules-2.6.18.tar.gz

It's strange that ethernet caused a crash on your machine. I transfered several GBs here without crash. Was it a hard lockup or is there any information in the syslog?

_________________
A1SE: G3@600MHz, 2GB, 1GBit network card
A1XE: G4@933MHz, 2GB, refitted AC'97 codec
microA1: G3@800MHz, 1GB

- A1 Linux support -

@Geri
Ok so the modules are just recognised when placed in libs as normal. I am playing around with ADRipper just now and will try again later and let you know. Thanks for working on this I know you have been putting a lot of effort in to try and fix these problems.

ace

@Geri

Thanks for doing that!

It took six tries to boot, but here I am, with 2.6.18! (I have had to boot as many as seven times to get 2.6.8 up.)

Nice not to get that red "Failed" about "udev." But I don't think I should remove hotplug until I'm really sure I won't want to use 2.6.8 any more.

For one thing, I'm back to non-recognition of affs partitions. I've already forgotten how I fixed that last time (probably not much more than a month ago); something to do with modprobe maybe.

I said:
Quote:

It took six tries to boot, but here I am, with 2.6.18! (I have had to boot as many as seven times to get 2.6.8 up.)

That needs updating. I spent several hours thereafter in OS4, then came back to fix the affs module thing.

2.6.18 booted straight up. Five minutes later, Debian crashed, and I rebooted 2.6.18 again with no trouble. Debian again failed, and again I rebooted without problem (and this time Debian has stayed up long enough for me to modprobe affs, read Usenet for a while, and type this message. 2.6.8 *never* booted successfully three times in a row!

I think this kernel is a winner.

@Tomppeli

I still have to compile some older kernels (2.6.10, 2.6.12, etc.) so that we can track down this problem. It's on my TODO list, but my spare time is too limited for all this work currently.

_________________
A1SE: G3@600MHz, 2GB, 1GBit network card
A1XE: G4@933MHz, 2GB, refitted AC'97 codec
microA1: G3@800MHz, 1GB

- A1 Linux support -

@kilaueabart

Quote:

kilaueabart wrote: 2.6.18 booted straight up. Five minutes later, Debian crashed, and I rebooted 2.6.18 again with no trouble. Debian again failed, and again I rebooted without problem (and this time Debian has stayed up long enough for me to modprobe affs, read Usenet for a while, and type this message. 2.6.8 *never* booted successfully three times in a row! I think this kernel is a winner.

That still sounds "very unstable" to me. Linux is really driving me crazy on the A1!

_________________
A1SE: G3@600MHz, 2GB, 1GBit network card
A1XE: G4@933MHz, 2GB, refitted AC'97 codec
microA1: G3@800MHz, 1GB

- A1 Linux support -

@Geri Linux is really driving me crazy on the A1

Use OS4 and OWB then :D

@Geri

Just install 2.6.18 here and all seems okay. No lockups or crashes. Been browsing arround and listening to lastfm for twenty minutes or so.

No sign of any data corruption, how does that normal manifest?

AmigaOne XE G4 800Mhz 256Mb ram

_________________
BroadBlues On Blues BroadBlues On Amiga Walker Broad

@Geri
I think you were responding to what I said about Debian (Etch) crashing after successful booting with 2.6.18 when you wrote
Quote:

That still sounds "very unstable" to me. Linux is really driving me crazy on the A1!

Having just booted 2.6.18 with no problems again today, I think it is very solid. 2.6.8 almost *never* got as far as the logon screen on the first try in the month or so I used it.

The three crashes I had yesterday were Etch crashes (or maybe my machine--anyway, none since then) rather than the kernel, as I see it.

@Kicko

Quote:

Kicko wrote: @Geri Linux is really driving me crazy on the A1 Use OS4 and OWB then :D

That doesn't make Linux run better on the A1.

@broadblues

Quote:

No sign of any data corruption, how does that normal manifest?

Data corruption happens with large data transfers over network (e.g. copy a 3GB file from another machine to the A1 and do a md5sum). I'm not sure however, if the problem is caused in combination with another PCI device or if it is a Linux cache coherency problem (the network driver handles DMA transfers a little bit different than other drivers; for performance reasons).

@kilaueabart

Quote:

Having just booted 2.6.18 with no problems again today, I think it is very solid. 2.6.8 almost *never* got as far as the logon screen on the first try in the month or so I used it. The three crashes I had yesterday were Etch crashes (or maybe my machine--anyway, none since then) rather than the kernel, as I see it.

How did the machine crash? Hard lockup or with a kernel oops on the console?

Last edited by Geri on 02-Jul-2008 at 09:20 AM.

_________________
A1SE: G3@600MHz, 2GB, 1GBit network card
A1XE: G4@933MHz, 2GB, refitted AC'97 codec
microA1: G3@800MHz, 1GB

- A1 Linux support -

@Geri

Quote:

Quote: Having just booted 2.6.18 with no problems again today, I think it is very solid. 2.6.8 almost *never* got as far as the logon screen on the first try in the month or so I used it. The three crashes I had yesterday were Etch crashes (or maybe my machine--anyway, none since then) rather than the kernel, as I see it. How did the machine crash? Hard lockup or with a kernel oops on the console?

If you mean my Etch crashes, I remember one was a lockup when I was starting Icedove. I don't remember the other two for sure, but they required reset button pushing.

The first tries with 2.6.18 were lockups after "(W)aiting for /dev to be fully populated._" (the cursor stopped blinking right there). The next one was right after "(S)tarting Xprint servers: Xprt"; this was the most popular place for 2.6.8 to die. The next hang was like the first, except after a new line feed. Fourth, "(S)etting kernel variables...done." plus line feed. The fifth and last was a repeat of the "(W)aiting for /dev to be fully poplulated," without cursor or line feed.

The next try made it all the way, and the kernel has never hung on me since. *MUCH* better on my machine than 2.6.8, which usually took three to four tries! (I admit I haven't tried any 3G file transfers--I don't have a network to transfer over.)

I still find the generic 2.6.n loss of first character on tty annoying, but that seems to be a quirk of my machine/monitor, since no one else has mentioned it (and on the tty that comes up on shut-down, the lines are moved to the right, so every line starts with a space and all the words printed in full!)

@Geri

Just now getting around to DL'ng the updated v2.6.18. Thanks for posting the patch files, too.

Just curious, would the previous v2.6.18 modules work with the updated kernel?

@DL

Quote:

DL wrote: @Geri Just now getting around to DL'ng the updated v2.6.18. Thanks for posting the patch files, too. Just curious, would the previous v2.6.18 modules work with the updated kernel?

Likely, but I didn't try it out. On the other side the code base has changed (latest Debian Etch kernel source), thus not all old modules may work as expected.

_________________
A1SE: G3@600MHz, 2GB, 1GBit network card
A1XE: G4@933MHz, 2GB, refitted AC'97 codec
microA1: G3@800MHz, 1GB

- A1 Linux support -

@Geri

Tested a couple of times. Once it freezed while KDE was up. Other times it worked good. Thansk for update.

@Geri

It worked on my SE!!!

_________________
And may the AmigaGuide you!

AmigaOne G3-SE, OS4 Final (July 2007 update), Debian Sarge, 512 MB RAM, 20 + 80 GB hard disks, NEC 3540 DVD writer, LG DVD reader, Radeon 9250, SB Live, Intracom Netfaster router, PCI USB card (NEC - OHCI/EHCI)

@mwoof

May I ask your harddrives and CD/DVDROM drives which manufacturers are they from ? I have Samsung and Seagate HD's and LiteOn DVDROM. Do you have Etch or Sarge ?

Last edited by Tomppeli on 12-Jul-2008 at 12:34 AM.

_________________
Rock lobster bit me. My Workbench has always preferences. X1000 + AmigaOS4.1 FE
"Anyone can build a fast CPU. The trick is to build a fast system." -Seymour Cray