View previous topic :: View next topic |
Author |
Message |
CptPajamas Tux's lil' helper
Joined: 02 Apr 2004 Posts: 77 Location: Santa Fe, NM
|
Posted: Fri Nov 11, 2005 4:53 pm Post subject: Fujitsu S2110 Turion 64, Broadcom 4400, segfaults [SOLVED] |
|
|
Greetings!
I've recently acquired a new Fujitsu Lifebook S2110 for my own use through my place of work, however upon installing Gentoo on the system, I've encountered an error exclusive to the networking hardware. The rest of the machine boots appropriately and runs just fine, however upon either doing default runlevel net.eth0 or manual net.eth0 after boot (using both built-in b44.o and modulized b44.o), the machine displays the following message:
Quote: | Kernel BUG at "arch/x86_64/kernel/traps.c" :338
invalid operand: 0000 [1]
CPU 0
Modules linked in: b44 mii
Pid: 4669, comm: dhcpcd Not Tainted 2.6.13-gentoo-r5
RIP: 0010:[<fffffffff8010f320>] <fffffffff8010f320>{out_of_line_bug+0}
RSP: 0018:fff81007704bd00 EFLAGS: 000100006 |
etc, etc, etc, process dhpcd dies, stack, call trace, code, RIP /sbin/dhcpcd ${dhcpcd_IFACE} ${IFACE}
Any ideas as to narrowing this down further? The 2005.0 AMD64 universal livecd runs just fine with DHCP and acquiring an IP address through my network.
Any suggestions are appreciated.
Cheers! _________________ =[ Nate Metheny
=[ Director of Technology
=[ Santa Fe Institute
Last edited by CptPajamas on Wed Feb 22, 2006 6:29 pm; edited 1 time in total |
|
Back to top |
|
|
josef.navratil n00b
Joined: 13 Nov 2005 Posts: 2
|
Posted: Sun Nov 13, 2005 2:50 am Post subject: Re: Fujitsu S2110 Turion 64, Broadcom 4400, Driver segfaults |
|
|
CptPajamas wrote: | Greetings!
I've recently acquired a new Fujitsu Lifebook S2110 for my own use through my place of work, however upon installing Gentoo on the system, I've encountered an error exclusive to the networking hardware. The rest of the machine boots appropriately and runs just fine, however upon either doing default runlevel net.eth0 or manual net.eth0 after boot (using both built-in b44.o and modulized b44.o), the machine displays the following message:
Quote: | Kernel BUG at "arch/x86_64/kernel/traps.c" :338
invalid operand: 0000 [1]
CPU 0
Modules linked in: b44 mii
Pid: 4669, comm: dhcpcd Not Tainted 2.6.13-gentoo-r5
RIP: 0010:[<fffffffff8010f320>] <fffffffff8010f320>{out_of_line_bug+0}
RSP: 0018:fff81007704bd00 EFLAGS: 000100006 |
etc, etc, etc, process dhpcd dies, stack, call trace, code, RIP /sbin/dhcpcd ${dhcpcd_IFACE} ${IFACE}
Any ideas as to narrowing this down further? The 2005.0 AMD64 universal livecd runs just fine with DHCP and acquiring an IP address through my network.
Any suggestions are appreciated.
Cheers! |
This is not a network/driver error I think. It looks like A64/ATI Xpress200 segfault bug. Just append noapictimer option to your kernel boot parameters (in /boot/grub/grub.conf).
Like this (mine):
title=Gentoo Hardened Linux 2.6.13-r2
root (hd0,0)
kernel /kernel-2.6.13-hardened-r2 root=/dev/sda3 vga=0x318 noapictimer
All yours segfaults should gone. |
|
Back to top |
|
|
CptPajamas Tux's lil' helper
Joined: 02 Apr 2004 Posts: 77 Location: Santa Fe, NM
|
Posted: Sun Nov 13, 2005 9:25 pm Post subject: |
|
|
I'll give that a try first thing tomorrow morning. I was curious about whether APIC may be involved, considering the only way I could get the livecd to boot was using noapic.
Thanks a lot! _________________ =[ Nate Metheny
=[ Director of Technology
=[ Santa Fe Institute |
|
Back to top |
|
|
adsmith Veteran
Joined: 26 Sep 2004 Posts: 1386 Location: NC, USA
|
Posted: Sun Nov 13, 2005 10:08 pm Post subject: |
|
|
please let us know if this works. My wife will be getting this exact machine in a week or two, and I'd like it to be as painless as possible.... |
|
Back to top |
|
|
CptPajamas Tux's lil' helper
Joined: 02 Apr 2004 Posts: 77 Location: Santa Fe, NM
|
Posted: Mon Nov 14, 2005 4:28 pm Post subject: |
|
|
Hm I manually changed the boot line in grub during boot time to include the noapictimer option, however it didn't have any effect on the error. Any other ideas? Anyone? Bueller?
EDIT: I'm going to try a stage1 on the machine and see if it makes any difference. _________________ =[ Nate Metheny
=[ Director of Technology
=[ Santa Fe Institute |
|
Back to top |
|
|
CptPajamas Tux's lil' helper
Joined: 02 Apr 2004 Posts: 77 Location: Santa Fe, NM
|
Posted: Wed Nov 16, 2005 4:49 pm Post subject: |
|
|
Personal notes: The 2005.0 livecd only boots with noapic iommu=soft, everything's supported and works fine. I'll be finishing the stage1 install today and will be trying out the ethernet again. _________________ =[ Nate Metheny
=[ Director of Technology
=[ Santa Fe Institute |
|
Back to top |
|
|
adsmith Veteran
Joined: 26 Sep 2004 Posts: 1386 Location: NC, USA
|
Posted: Wed Nov 16, 2005 6:18 pm Post subject: |
|
|
Interesting. Do you have any info as to what iommu=soft actually *does*?
It's mentioned only briefly in the kernel docs. |
|
Back to top |
|
|
CptPajamas Tux's lil' helper
Joined: 02 Apr 2004 Posts: 77 Location: Santa Fe, NM
|
Posted: Wed Nov 16, 2005 6:31 pm Post subject: |
|
|
I'm pretty sure it has something to do with the difference between a hardware and software AGPGART that's present in the AMD64 chips -- using IOMMU compiled into the kernel disables AGPGART support, and vice versa. _________________ =[ Nate Metheny
=[ Director of Technology
=[ Santa Fe Institute |
|
Back to top |
|
|
CptPajamas Tux's lil' helper
Joined: 02 Apr 2004 Posts: 77 Location: Santa Fe, NM
|
|
Back to top |
|
|
adsmith Veteran
Joined: 26 Sep 2004 Posts: 1386 Location: NC, USA
|
Posted: Wed Nov 16, 2005 11:52 pm Post subject: |
|
|
Thank you!
These will be very useful when I'm installing this machine this coming weekend. |
|
Back to top |
|
|
josef.navratil n00b
Joined: 13 Nov 2005 Posts: 2
|
Posted: Thu Nov 17, 2005 11:26 am Post subject: |
|
|
Congrats
Didn't you notice something like twice faster timer? Just type 'date' few times and see the seconds... |
|
Back to top |
|
|
CptPajamas Tux's lil' helper
Joined: 02 Apr 2004 Posts: 77 Location: Santa Fe, NM
|
Posted: Thu Nov 17, 2005 8:55 pm Post subject: |
|
|
Bad news ...
DRM was built into the kernel ... I tried to take it out, rebooted, trying to emerge ati-drivers, and the dhcpcd process seg faulted with the same error again. I'm frustrated enough at this point to just give up and try it again with the next kernel revision and/or Gentoo 2005.x release.
I did however rebuild the kernel exactly the way it was previous to the change, and it still failed. Flaky, flaky, flaky. _________________ =[ Nate Metheny
=[ Director of Technology
=[ Santa Fe Institute
Last edited by CptPajamas on Thu Nov 17, 2005 11:52 pm; edited 1 time in total |
|
Back to top |
|
|
adsmith Veteran
Joined: 26 Sep 2004 Posts: 1386 Location: NC, USA
|
Posted: Thu Nov 17, 2005 11:32 pm Post subject: |
|
|
what about using plain old vesa? |
|
Back to top |
|
|
CptPajamas Tux's lil' helper
Joined: 02 Apr 2004 Posts: 77 Location: Santa Fe, NM
|
Posted: Thu Nov 17, 2005 11:50 pm Post subject: |
|
|
I was using VESA for framebuffering, but I built the radeon DRM into the kernel for xorg. I realized after the fact that in order for the ati-drivers to build, DRM has to not be in the kernel, hence the reason I changed it in the first place.
I'll monkey with it some more; I've never had to tweak so much to get Gentoo working on a laptop, 64-bit or otherwise, before. Quite odd.
It's no use to get a generic xorg running with the ati driver if the ethernet doesn't work.
EDIT: DRM out of kernel, IOMMU built in, iommu=soft appended to grub.conf, booting again with dhcpcd success. _________________ =[ Nate Metheny
=[ Director of Technology
=[ Santa Fe Institute |
|
Back to top |
|
|
adsmith Veteran
Joined: 26 Sep 2004 Posts: 1386 Location: NC, USA
|
Posted: Sat Nov 19, 2005 1:51 pm Post subject: |
|
|
So, my machine is mostly installed, and I'm not having any of these problems. I'm wondering if you have bad media, a broken toolkit, or perhaps something is funny with your kernel config.
It has certainly never failed to boot, on the liveCD or otherwise. I am currently just using "iommu=soft", and not using the noapic* options, but I tired those as well. They didn't seem to help or hurt anything, so I left them off.
The only problem I do have is that the b44 turns itself off and on repeatedly, apparently because it improperly detects the beat signal. |
|
Back to top |
|
|
adsmith Veteran
Joined: 26 Sep 2004 Posts: 1386 Location: NC, USA
|
Posted: Sat Nov 19, 2005 1:53 pm Post subject: |
|
|
Aside: which drivers you are using for the SD card reader or the touchpad? |
|
Back to top |
|
|
adsmith Veteran
Joined: 26 Sep 2004 Posts: 1386 Location: NC, USA
|
Posted: Mon Nov 21, 2005 3:16 pm Post subject: |
|
|
Updates:
My system is essentially installed, all the way up to KDE, OOffice, etc. Here is the hardware status:
- kernel option are "noapic" I turned iommu=soft on and off, but it doesn't seem to make a difference for anything. Noapic is needed to prevent CPU "lost ticks" errors. I'm not convinced that noapictimer completely solves this, so I just went for noapic.
- audio is alsa's atiixp module
- The touchpad works as a basic psmouse device. If you enable evdev as well, then the synaptics X11 driver makes the scroll buttons, etc, work nicely.
- X works fairly well with the vesa driver. Have not succeeded in getting kernel-DRM or commercial ati/ radeon drivers to work, but haven't tried since this is a "work" machine. flgrx works, too.
- Broadcom ethernet b44 driver is insane, and the commercial "bcm4401" driver is only for 32-bit kernels. the b44 driver improperly detects the network beat and keeps bringing itself up and down, spastically. Someone online also suggested using "pci=noapci" or "noapci" as a boot option, but this didn't help, and the latter of course kill all power management. However, after only limited testing a solution appears to be "ethtool -s eth0 autoneg off speed 100 duplex full" That is, turn off autonegotiation. I'm not sure if this actually solves the problem, as the connection still seems slower than it ought to be, but the error messages don't show up in the system logs. I just put that ethtool line in local.start.
- The built-in SD card reader doesn't work. I have no idea how to get this to work, and apparently few others online do. It it not on the USB bus. It may be on the PCMCIA or the PCI bus.
- The Atheros wireless is not quite stable, but the madwifi-ng snapshot drivers will work fairly well (download directly -- the ones in portage are old)
- USB works with ehci and ohci.
- PCMCIA works with normal "yenta" modules and pcmcia-tools .
- with APCI processor and k8powernow built into the kernel (not as modules), CPU frequency scaling works great. It has two modes, ~800MHz, and ~1600MHz. Also, the power button, lid switch, and ac-power detection all work great in APCI.
If the b44 driver is really stable, then all the essential parts work very well, aside from the wireless, which I am hopeful about. I have doubts that the SD card reader will work any time soon, as there is no unified way to access these devices yet.
Of course, there are also the normal amd64 issues to deal with, but that's another thing altogether, and it was relatively painless (once I gave up on crossdev/distcc)
[edit]: added audio.
[edit]: added PCMCIA and madwifi and flgrx
Last edited by adsmith on Tue Feb 21, 2006 5:52 pm; edited 3 times in total |
|
Back to top |
|
|
CptPajamas Tux's lil' helper
Joined: 02 Apr 2004 Posts: 77 Location: Santa Fe, NM
|
Posted: Mon Nov 21, 2005 4:15 pm Post subject: |
|
|
Thanks for all the great info. I've been working on getting the x200 radeon to work without success, but I feel as if this may be a refresh value issue and not a driver issue. I'll try the straight VESA driver out, and even the "ati" driver (not "fglrx") to see if these have more success.
Out of curiosity, if you're using framebuffering, what configuration are you using? (mode, font, any kernel compiled-in-fonts, etc)
I do seem to have a pretty stable machine going, compiling fast, and I do have to testify / agree that the CPU scaling works really great, as do most of the other ACPI options. PCMCIA, last time I tried, as broken (but the ethernet was causing the traps.c seg fault at that time too) ... I'll be testing this more with my orinoco card(s).
The SD reader I haven't had success with either. Touchpad wise, I'm sticking with PS/2, but now I'm curious about the scroll buttons.
Interesting that you didn't have the similar problems with the livecd ... was it a 2005.0? I'm also keeping the wireless switch in the "off" position, and for some reason am curious if this causes some issue with APIC.
Thanks again for the followup. _________________ =[ Nate Metheny
=[ Director of Technology
=[ Santa Fe Institute |
|
Back to top |
|
|
adsmith Veteran
Joined: 26 Sep 2004 Posts: 1386 Location: NC, USA
|
Posted: Mon Nov 21, 2005 4:24 pm Post subject: |
|
|
I haven't played with framebuffering at all. It boots fast enough that I don't mind the short burst of ugliness. I also hope to get suspend working, so we never see a boot screen ever again!
To get the scroll buttons working on the touchpad, just emerge synaptics and use the xorg snippet for "Alps" on the Synaptics gentoo-wiki entry: http://gentoo-wiki.com/HARDWARE_Synaptics_Touchpad
I was using the straight amd64 2005.1 universal install CD. I think I had the WiFi switch "on" during install, though it is off now without any problems. I have had "noapic" set for a long time, so I don't know if that matters or not for curing those hypothetical problems. |
|
Back to top |
|
|
adsmith Veteran
Joined: 26 Sep 2004 Posts: 1386 Location: NC, USA
|
Posted: Tue Nov 22, 2005 4:22 am Post subject: |
|
|
Update: I now have the ATI (fglrx) X driver working. very nice -- 700+ FPS in glxgears. |
|
Back to top |
|
|
adsmith Veteran
Joined: 26 Sep 2004 Posts: 1386 Location: NC, USA
|
Posted: Tue Nov 22, 2005 4:46 am Post subject: |
|
|
Also, the Atheros WiFi card is listed as supported on the MadWifi homepage, but the newest version is still masked in portage. Hence, it appears this will be available soon. |
|
Back to top |
|
|
CptPajamas Tux's lil' helper
Joined: 02 Apr 2004 Posts: 77 Location: Santa Fe, NM
|
Posted: Tue Feb 21, 2006 4:35 pm Post subject: |
|
|
Quick question for adsmith:
I recently revisited this issue and found that the 2005.1 x86 Universal CD works great with no kernel options; I'm up to configuring Xorg with video drivers, and have gotten the "vesa" driver to function, but no go with "radeon" or "fglrx."
Can you kindly post your xorg.conf section referencing the fglrx setup here?
Thanks! _________________ =[ Nate Metheny
=[ Director of Technology
=[ Santa Fe Institute |
|
Back to top |
|
|
adsmith Veteran
Joined: 26 Sep 2004 Posts: 1386 Location: NC, USA
|
Posted: Tue Feb 21, 2006 5:40 pm Post subject: |
|
|
I'll have to post it later, as my wife has the laptop at work. However, the config was simply to emerge the ati drivers, then run ATI's xorg configurator (flgrx-config or something?) Then, I added the synaptics stuff for the touchpad.
Some other notes:
1) kernel 2.6.15 claims to include a driver for those extra buttons! It isn't supported in x86_64/amd64 yet, though.
2) atheros stabiltiy is quickly improving, if you check with the atheros-ng snapshots. For now, I'm still using an old PCMCIA card. |
|
Back to top |
|
|
CptPajamas Tux's lil' helper
Joined: 02 Apr 2004 Posts: 77 Location: Santa Fe, NM
|
Posted: Tue Feb 21, 2006 6:50 pm Post subject: |
|
|
thx for the notes.
I did run the fglrxconfig in order to make the correct xorg.conf, however it gave me the fatal:
(EE) No Screens Found
with the fglrx and radeon drivers.
I'll take another look at some configs.
EDIT: Found it to be the 8.14 fglrx drivers. the 8.20x work nicely, albeit without DRI.
Thanks! _________________ =[ Nate Metheny
=[ Director of Technology
=[ Santa Fe Institute |
|
Back to top |
|
|
|
|
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
|
|