1 |
Sorry to resort to this, but bugzilla buffalo's me. |
2 |
|
3 |
The release 1 CD is incompatable with the PCChips 841 LMR motherboard. |
4 |
Actually the problem is with all 2.4 kernels after 2.4.10. The install |
5 |
fails during PCI setup with the message "PCI: System does not support |
6 |
PCI". The relevant code from pci-pc.c in the 2.4.18 kernel is |
7 |
|
8 |
1184 if ((pci_probe & (PCI_PROBE_CONF1 | PCI_PROBE_CONF2)) |
9 |
1185 && (pci_root_ops = pci_check_direct())) { |
10 |
|
11 |
pci_check_direct returns a nul when no direct access method is detected |
12 |
and the side effect of making the check is to wipe out the pointer to |
13 |
the bios access method functions previously assigned to pci_root_ops. |
14 |
|
15 |
A work around for kernels 2.4.10 - 2.4.18 is to use the kernel boot |
16 |
parameter pci=bios. For the 2.4.19 kernel using this parameter causes a |
17 |
failure and hex dump in pci_init_sis5513 start (the sis5513 is the ide |
18 |
controller) earlier in the boot process. |
19 |
|
20 |
Since scanpci -1 works correctly there appears to be a problem with the |
21 |
conf1 detection in pci-pc.c. A further comment about that below. |
22 |
|
23 |
On both the rc and the 1.0 CDs the init.d/pnp script terminates with the |
24 |
error: |
25 |
lspci: -f: Invalid function number. |
26 |
and doesn't load the modules. The particular script command that fails |
27 |
is lspci -s 00:02.7. I downloaded/installed pciutils-2.1.10 and it runs |
28 |
that command correctly so the script is not the problem. Apparently |
29 |
only an update of the utilities is needed. |
30 |
|
31 |
Since the pci=bios parameter causes a problem with setting up the |
32 |
sis5513 when using the 2.4.19 kernel, I am trying to determine why |
33 |
pci_pc.c doesn't detect the conf1 direct method while scanpci has no |
34 |
trouble using it. I don't have the answer yet but have found that |
35 |
scanpci appears to find all the pci devices using either the bios or |
36 |
direct conf1 method whereas lspci only finds the function 0 devices for |
37 |
each slot using the direct method. Thus, if code from lspci has found |
38 |
it's way into the kernel ( lspci and pci_pc.c appear to have the same |
39 |
maintainer) this could be source of the problem. I'll just have to |
40 |
stick with the 2.4.18 kernel for now. |
41 |
|
42 |
I may be the only person attempting to run linux on a 841LMR motherboard |
43 |
and thus the only person to have these problems. But, I thought the |
44 |
info might be useful to you. |
45 |
|
46 |
-- john |