D945GCLF2 (Dual Core Atom) w/ Fedora 10?

[darethehair]

Hello Folks:

With the demise of my Pundit-based MythTV box, I am seriously considering an Atom-based desktop, using the D945GCLF2 motherboard -- with Fedora.

It would be great to get some 'thumbs up' for such a configuration. The vibes I am getting from this community so far are:

- Fedora 10 works well on Atom-based netbooks
- Fedora 9 has trouble on the D945GCLF (single core) motherboard

Can I extrapolate from this that Fedora 10 will work great on the D945GCLF2?

As well, the 'Fedora 10 Installation Guide' indicates that the appropriate version to use for the D945GCLF2 is the 'x86_64' one. Is this true? I am never used a 64 bit version of linux before.

[Phirsk]

I've been running Fedora 9 on my D945GCLF2 based test system fine for 5 weeks. Am now downloading Fedora 10 ISO to give that a try

Oh, and no, the 64-bit version is _not_ the right one for these boards - use the standard i386 version. I think you'll also want to read up on using (or not) the HyperThreading. If you enable this, Fedora will basically act as if you have 4 cores. This can be good if you run lots of small tasks/threads, but it sounds like you're more likely to want it disabled (i.e. 2 powerful cores, instead of 4 weak ones).

Good luck.
Phirsk.

[darethehair]

Phirsk:

Thanks for confirming success with F9! Please confirm success with F10 as well!

I am hoping that the 'TV Out' works as well? I have not (and intend not) to have a VGA monitor connected to the MythTV box. I hope that the 'TV Out' works for both the console, and GUI desktop under F10 (?).

[Phirsk]

I'm afraid that the s-video 'TV Out' port won't be on my test list, I'm only using the VGA port. In the past, though, I've found s-video to work OK, but sometimes to be black and white only.

I should have Fedora 10 on the Atom board in the next couple of days, and will try to give a rundown of it.

Phirsk.

[Phirsk]

Bad news, i'm afraid...
below are my build notes for GUI installation of Fedora 10 on the Intel dual-core board:

bios - disable hyperthreading
boot from fedora 10 dvd
skip media check (if it's damaged, we'll find out soon enough anyway)
setup localisation, partitions, root password etc
confirm packages to be installed
wait half an hour or so, depending on packages
reboot when prompted

system boots to CLI. login as root

ran 'startx', which fails with:
(EE) open /dev/fb0: No Such File or directory
(EE) intel(): no valid modes
(EE) Screen(s) found, but none have a usable configuration


looked about online, but couldn't find what was wrong.
ran out of time, so had to start CLI-only installation.

Sorry I couldn't help further, but I think if you spend some time researching, it should work.

Phirsk.

[darethehair]

Phirsk:

Ouch! This is not good news! Do you intend to keep trying until success if achieved?

Maybe I will try the 'live' version first, rather than go through all the same effort -- just to see if there is 'hope...

It is possible that the 'x86_64' version would work better? Was there a need to disable hyperthreading, as you did?

EDIT: I wonder if this is the bug that you ran into?

https://bugzilla.redhat.com/show_bug.cgi?id=458807

I notice that on my (other) F10 install, that the version of the xorg 'i810' driver is newer:

xorg-x11-drv-i810-2.5.0-3.fc10.i386

I wonder if the newer driver fixes this problem? Can you try changing your 'xorg.conf' to 'vesa', just to see if you get a screen?

(if it sounds like I know what I am talking about, it is just an illusion...)

EDIT #2: Great news! I just installed Fedora 10 onto my just-assembled D945GCLF2-based mini-PC, and the X windows/Gnome desktop looks fine!

I am not sure what I did that was so different from what Phirsk did above, but during Anaconda installation I picked the all three clickable repositories:

- Installation Repo
- Fedora 10 - i386
- Fedora 10 - i386 - Updates

So far -- in just the past few minutes -- I got sound working, including the microphone.

The default GPU fan *is* too noisy, so one of my next steps is to replace it with a 40mm 'Mini-kaze' fan.

[Phirsk]

I disabled HT as I only wanted 2 cores running, not 4, but this wouldn't affect the graphical issue I got stuck on.

I'm glad you sorted it out - with hindsight, I built it off-line, so it may not have benefited from online updates.

Phirsk

[everestx]

Fellas, I've been struggling a bit with my install on the same board. The install went fine but I keep getting a crash as kacpid hogs the CPU. Can't find anything about a bug with this exaclty, but just a heads up in case anyone sees the same problem. Here's the messy output from /var/log/messages. I tried both the kernel from the install media and the kernel from updates with same results. Tried disabling hyperthreading with no success. Updated the BIOS from intel with no success. Is this a faulty motherboard?

Dec 16 16:31:16 localhost kernel: BUG: soft lockup - CPU#0 stuck for 61s! [kacpid:105]
Dec 16 16:31:16 localhost kernel: Modules linked in: udf crc_itu_t fuse i915 drm autofs4 sunrpc ip6t_REJECT nf_conntrack_ipv6 ip6ta
ble_filter ip6_tables ipv6 dm_multipath uinput snd_ice1724 snd_rawmidi snd_ice17xx_ak4xxx snd_ac97_codec ac97_bus snd_ak4xxx_adda s
nd_ak4114 snd_seq_dummy snd_seq_oss snd_seq_midi_event snd_seq snd_seq_device snd_pcm_oss snd_mixer_oss iTCO_wdt snd_pcm iTCO_vendo
r_support snd_timer ppdev snd_page_alloc parport_pc snd_pt2258 i2c_i801 i2c_core parport snd_i2c snd soundcore r8169 mii pcspkr lir
c_mceusb2 lirc_dev serio_raw ata_generic pata_acpi [last unloaded: microcode]
Dec 16 16:31:16 localhost kernel:
Dec 16 16:31:16 localhost kernel: Pid: 105, comm: kacpid Not tainted (2.6.27.7-134.fc10.i686 #1)
Dec 16 16:31:16 localhost kernel: EIP: 0060:[<c055867f>] EFLAGS: 00000297 CPU: 0
Dec 16 16:31:16 localhost kernel: EIP is at acpi_ut_valid_acpi_name+0x14/0x3f
Dec 16 16:31:16 localhost kernel: EAX: cb290412 EBX: f7931e0f ECX: 00000003 EDX: cb290453
Dec 16 16:31:16 localhost kernel: ESI: f5444e00 EDI: f7816c48 EBP: f7931e14 ESP: f7931e0c
Dec 16 16:31:16 localhost kernel: DS: 007b ES: 007b FS: 00d8 GS: 0000 SS: 0068
Dec 16 16:31:16 localhost kernel: CR0: 8005003b CR2: a9331000 CR3: 0087b000 CR4: 000006d0
Dec 16 16:31:16 localhost kernel: DR0: 00000000 DR1: 00000000 DR2: 00000000 DR3: 00000000
Dec 16 16:31:16 localhost kernel: DR6: ffff0ff0 DR7: 00000400
Dec 16 16:31:16 localhost kernel: [<c05520b8>] acpi_ns_search_and_enter+0x4a/0x14e
Dec 16 16:31:16 localhost kernel: [<c0551c71>] acpi_ns_lookup+0x1ed/0x2ec
Dec 16 16:31:16 localhost kernel: [<c05477d8>] acpi_ds_create_operand+0xca/0x1d7
Dec 16 16:31:16 localhost kernel: [<c0547925>] acpi_ds_evaluate_name_path+0x40/0xe3
Dec 16 16:31:16 localhost kernel: [<c0546875>] acpi_ds_exec_end_op+0x88/0x373
Dec 16 16:31:16 localhost kernel: [<c0555035>] acpi_ps_parse_loop+0x5cd/0x720
Dec 16 16:31:16 localhost kernel: [<c0554468>] acpi_ps_parse_aml+0x6d/0x25b
Dec 16 16:31:16 localhost kernel: [<c05556b5>] acpi_ps_execute_method+0x11c/0x1d7
Dec 16 16:31:16 localhost kernel: [<c05529e6>] acpi_ns_evaluate+0x12e/0x188
Dec 16 16:31:16 localhost kernel: [<c054ab1d>] acpi_ev_asynch_execute_gpe_method+0xab/0xf2
Dec 16 16:31:16 localhost kernel: [<c0544709>] acpi_os_execute_deferred+0x20/0x2c
Dec 16 16:31:16 localhost kernel: [<c0439e9d>] run_workqueue+0x89/0x113
Dec 16 16:31:16 localhost kernel: [<c05446e9>] ? acpi_os_execute_deferred+0x0/0x2c
Dec 16 16:31:16 localhost kernel: [<c0439ff3>] worker_thread+0xcc/0xd8
Dec 16 16:31:16 localhost kernel: [<c043cd4e>] ? autoremove_wake_function+0x0/0x33
Dec 16 16:31:16 localhost kernel: [<c0439f27>] ? worker_thread+0x0/0xd8
Dec 16 16:31:16 localhost kernel: [<c043caab>] kthread+0x3b/0x61
Dec 16 16:31:16 localhost kernel: [<c043ca70>] ? kthread+0x0/0x61
Dec 16 16:31:16 localhost kernel: [<c04048f7>] kernel_thread_helper+0x7/0x10
Dec 16 16:31:16 localhost kernel: =======================

[darethehair]

everestx:

Sorry to read about your difficulties, and also that I don't have the 'smarts' to comment on those bizarre error messages. All I can suggest is the following:

1) Do what I did: disable hyperthreading (at least initially) and let the same three 'repos' do your upgrade stuff (no more and no less) -- or is that what you already did?

2) Install another OS or Linux -- to see if the board works OK or not with them

[everestx]

I think another OS could tell me if this is hardware related (or BIOS related), but I didn't really want to do another install if I didn't have to. I don't see anything from anyone else anywhere, so I might have to do that tonight.

I am going to post to the fedora-list too and see if I get any response. It's really killing me as I am using it as my HD frontend on MythTV and it won't play HD when kacpid takes over 100% CPU. If I hear anything I'll post back here.

[Phirsk]

Are you sure you're using the Fedora-10-i386 image ? not 64-bit or i686?

Phirsk

[everestx]

I believe I am using fedora-10-i686. I don't have the exact version right now since I am remote and 'no surprise to me' the system is unresponsive again when I try ssh. I'll have to reboot when I get home and check the kernel version.

Should it be i386 since it's the atom processor?

[everestx]

My apologies Phirsk.. I see from the top of the thread that i386 was noted as being preferred. I didn't catch that the first time. Maybe I'll try an install of that tonight. I guess mitigating this error would be a good reason to switch.

I should be able to hard code that somewhere and install the i386 version.

[Phirsk]

I think you need to download the ISO (or ISO's) again, but the i386 version. i686 and 64 bit versions won't work.
Good luck on the install, and make sure you read all of Everestx's comments, and you should be good

Phirsk.

[everestx]

There was a common bug with FC6 that caused the i586 kernel to install on i686 hardware, so there's a documented way to install a non-matching kernel ( http://fedoraproject.org/wiki/Bugs/F...97983a3822f143 ). I'll try that. Worst case, I still have to re-install, but everything else is perfect, so I would hate to do a clean wipe.