kernel 2.6 migration steps?

[Forge]

lodger - It won't matter to your kernel.

As I've noted earlier in the thread.... linux-headers-2.4.19-r1 has NOTHING to do with your running kernel. Current and standards-compliant Linux systems keep TWO sets of headers. One is kept in /lib/modules/*version*/build/ and matches the RUNNING kernel. The other is kept in /usr/include/ and is used by glibc. This second one is expected to NOT change with the kernel, only with the stable kernel tree.

In other words: I'm running 2.6.0-test3.
The headers in /lib/modules/2.6.0-test3/build/include are from 2.6.0-test3, and that is good.

The headers in /usr/include/ are from 2.4.19-r1, and that, too, is good.

If I foolishly copied my 2.6.0 headers into /usr/include, it wouldn't help anything, and might BREAK a lot of things!

Some people seem to be assuming the headers in /usr/include need to come from 2.5.75+ in order to compile glibc with nptl, and that's NOT so! Only the headers in /lib/modules/*/build matter to the nptl check!

[jerome187]

will the nvidia-driver work with the 2.6 beta3 kernel?

[Forge]

If you use the portage version, yes, it patches for 2.6.0 automatically. If you install the Nvidia driver by hand, then no, you're out of luck or in for some patching.

[jerome187]

this is a little off topic, but when i emerge a new kernel do i have to change the linux link thing in /usr/src/ to point to the new kernel sources i want to use?

[ericsean]

I've been trying to get the development kernels working since 2.5.74, upto and including 2.6.0-test3. I can compile the kernel fine, and it boots fine, however, my network connectivity is always broken on bootup. doing ifconfig tells me that eth0 is configured correctly, with the correct ip, netmask, broadcast and gateway addresses. however, if i ping a known ip address on my network, i get no reply to my ping.

i thought that perhaps iptables now defaulted to blocking all traffic or something, but 'iptables -L' showed no rules.

i tried to check my routing table by typing 'route', but it wouldn't resolve and just waited - is this a clue? typing 'route -n' showed the whole routing table without resolving things, and it looked correct (compared against a dump of my 2.4.21 routing table).

i'm really bemused as to what the problem is, as i've seen no-one else under gentoo having the same problems. i disabled all the APIC stuff as I heard that might screw some of the network stuff up if used on UP boxes (which mine is). my network card is an intel pci etherexpress, so is using the e100 module (was eepro100 under 2.4 i think).

if anyone can help or suggest other ways to give more information about my problem, i'd be really grateful.

[TheOneLord]

Hi, there

If compiled the new linux-2.6.0-test3-bk3-Kernel. It runs just fine. But at boottime I get some errors:

Code: Select all

FATAL: Module /dev/rd/initrd not found
[...]
FATAL: Module st not found
FATAL: Modules ide_probe_mod not found
FATAL: Module ide_tape not found

The Log-File contains the following:

Code: Select all

[...]
Aug 19 16:11:58 [modprobe] FATAL: Module /dev/cpu not found._
Aug 19 16:11:59 [kernel] agpgart: Found an AGP 2.0 compliant device at 0000:00:0
0.0.
Aug 19 16:12:11 [kde(pam_unix)] session opened for user root by (uid=0)
[...]

1. Why does this kernel search for the module /dev/rd/initrd, and why does it not exist?

2. What is the Module "st"?

3. "ide_probe_mod"... ?

4. Why the hell is it searching for the ide_tape-Module? I compiled the Kernel without any Tape-support

5. How should I create /dev/cpu or which option in Kernel is it?

The new Kernel runs very fine, but I want to cutout such errors during boot... could somebody help me?

thx TOL

[plate]

There's a bug report devoted to all those "FATAL: Module blah not found" error messages - and numerous other bugs as well. In fact, bugs.gentoo.org is swarming with those... It's got nothing to do with the kernel, this is a Gentoo-only problem (cosmetic, granted, but it's still annoying) in either baselayout or module-init-tools or both, I don't know. I have trouble understanding what's going on and decided to disregard this whole topic until someone comes up with a solution straight from the portage tree...

[ericsean]

i finally got mine to work. i'm not quite sure which one it was, but in the end, i dropped the whole ACPI stuff out off the kernel and changed my network driver to donald becker's eepro driver instead of the alternate one. the noapic and pci=noacpi as kernel parameters was not enough to get it to work for me.

i was so pleased that alsa worked out of the box. happy days. hope this helps someone.

[TheOneLord]

thx @ plate

So I keep waiting for an update. While the stability doesn't really decrease by those fatal errors I'm going to switch to this new kernel permanentaly.

[nailed]

I wanted to build a new kernel for my server (running 2.5.70-mm8 at the moment) and i though about building in support for pptp vpn only to find there's no ppp_mppe support either in the kernel or available anywhere else on the internet (recent pppd does have patches for 2.2 and 2.4 series of kernel) and my particular flavor of ATM nic (lanai based but it's a chipset from TI) is still unsupported:

lanai: In lanai_dev_open()
lanai: PCI says board_id=0, board_rev=0
lanai: about to reset board
lanai: register says board_id=3, board_rev=2
lanai: Found register board-id 3 -- not a Lanai 25.6
lanai: about to reset board
lanai: lanai_start() failed, err=19
lanai: no adaptor found

lspci output for the card:

00:12.0 ATM network controller: Efficient Networks, Inc SpeedStream (LANAI)
Subsystem: Texas Instruments: Unknown device 6050
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-
Interrupt: pin A routed to IRQ 10
BIST result: 00
Region 0: Memory at ffb40000 (32-bit, non-prefetchable) [disabled] [size=256K]
Expansion ROM at ffbc0000 [disabled] [size=32K]
Capabilities: [40] Power Management version 1
Flags: PMEClk- DSI+ D1- D2- AuxCurrent=0mA PME(D0-,D1-,D2-,D3hot-,D3cold-)
Status: D0 PME-Enable- DSel=0 DScale=0 PME-

i'd be very happy to send the board out to a kernel hacker if it could result in getting some Linux support for it.

Nailed

[asterism]

this is a little off topic, but when i emerge a new kernel do i have to change the linux link thing in /usr/src/ to point to the new kernel sources i want to use?

Some of the ebuilds (like iptables) depend on the kernel sources so you probably want to make a symbolic link to the kernel sources you are using.

Also, well, it's less typing later.

I link /usr/src/linux to the 2.4 kernel and /usr/src/linux-beta to the 2.6 kernel I'm using.

[butters]

OK I am not migrating to 2.6.0, I did that on my desktop with little to no problems. I am now trying to install gentoo on my new centrino laptop with 2.6.0-test4-mm1 right from the start. Everything appears to work just fine, minus those annoying /etc/modprobe.conf problems associated with module-init-tools, expect for pcmcia-cs. When I first tried emerging pcmcia-cs, it noted that my running kernel was 2.4.x (the Knoppix CD), but my kernel source pointed to 2.6.0-test4-mm1. Not a problem, just fix that when I boot the new kernel.

But now pcmcia-cs complains that my version of the module utilities are too old for my kernel. It correctly sees version 0.9.13 of module-init-tools, but insists that I need at least 2.1.85. This would make sense if I was using modutils currently at 2.4.25, but 2.6.x uses module-init-tools. I'm sure there's a workaround for this, right? Is someone on top of all of these module-init-tools problems?

[fury]

I just got pcmcia working for 2.6.0-test1.

first I had to enable all the pcmcia stuff in the kernel then

I just had to edit /usr/src/linux/include/scsi/scsi.h
and make lines 228-30 look like:

Code: Select all

typedef struct scsi_lun {
    u_int8_t  scsi_lun[8];
} ScsiLun;

then emerge pcmcia-cs
then I just had to add "yenta_socket" to /etc/modules.autoload.d/kernel-2.6

It up and running now with my linksys wpc11 card.

hth,
sa

Nope, That fixed a few error messages but It's still crashing for me when i emerge pcmcia-cs. So sad. I was having this problem ever since early versions of 2.5, you'd think a bug this big would be fixed by now (I presume it's with pcmcia-cs) Oh well, I guess I'll keep waiting.

[Jevin Sweval]

I ahve teh same pcmcia_cs problem. We need a fix soon!

[sa]

this fix i mentioned used to work, but now they have changed the scsi.h file so you need to either change all instances of u8 to u_int8_t in the scsi.h file or I believe adding

Code: Select all

#include <asm/types.h>

to the top of the scsi.h file should fix it too.

[JeffBlair]

I get the same module not found error. But, after that, it is trying to load the system modual, and not finding it. I think this is stoping my NIC from starting up right. I will load all of the other moduals, including nvnet, but it won't start the network. It gives me the error right after it says that it is going to use /etc/modules.autoload.d/kernel-2.6
I am running the mm5 build and that is upgraded from the mm4 one. Thanks for any help.

[robdavies]

Anyone else seen this, when I boot into 2.4 kernel it doesn't happen. Am not sure why there's delay on login/out, but it might be because the UID is 500.

[robdavies]

So if you want to move back and forth with 2.4, building glibc with NPTL enabled (nptl in USE) is not a good idea.