1 |
Karol Krizka posted <200603121736.29988.kkrizka@×××××.com>, excerpted |
2 |
below, on Sun, 12 Mar 2006 17:36:26 -0800: |
3 |
|
4 |
> Hello, |
5 |
> I am trying to disable toe agpgart support in the kernel so I can use the one |
6 |
> that commes with the binary nvidia drivers. The only problem I have is that I |
7 |
> can't seem to do it. I located the option under Device Drivers -> Characters |
8 |
> Devices but there is no checkbx button beside "/dev/agpgart (AGP Support)" |
9 |
> and it is not disabled because when I grepped though the .config I found: |
10 |
> CONFIG_AGP=y |
11 |
> CONFIG_AGP_AMD64=y |
12 |
> # CONFIG_AGP_INTEL is not set |
13 |
> |
14 |
> Is there something I need to do before I am able to disable it? This is |
15 |
> suspend2-sources-2.6.15-r8 if that matters. |
16 |
|
17 |
If there's no options beside it, it means you have something else |
18 |
configured that depends on it, so you have to disable whatever that is in |
19 |
ordered to have the option to turn what you are looking at off. |
20 |
|
21 |
The depends (what must be on to have the choice), conflicts (what can't be |
22 |
on if it's on), and selected-bys (what depends on it so automatically |
23 |
turns it on) are normally listed in the help text for an item. This one |
24 |
is no exception, so check the help text and I'll bet you find one of the |
25 |
selected-by options on, thereby forcing this one on. |
26 |
|
27 |
In many cases, simply modularizing the selected-by option will allow the |
28 |
target option to be modularized as well. That's all you'd need. |
29 |
|
30 |
Most likely, your issue is GART_IOMMU, which is the I/O Memory Management |
31 |
Unit found in the AMD64 chips (Intel's version doesn't have a hardware |
32 |
IOMMU, but a software emulation of it is enabled by the same option). |
33 |
You'll likely want that enabled if you have >3 gig of memory. If not, it |
34 |
doesn't matter and you should be safe turning it off. Whether NVidia's |
35 |
GART driver provides the same IOMMU functionality, I don't know. It |
36 |
should be documented one way or the other, however, if you look in their |
37 |
README. If they don't, and you have >3 gig of memory, you'll likely have |
38 |
to choose between access to all your memory with the kernel's version, or |
39 |
running NVidia's version. I don't know how serious the issue is without |
40 |
an IOMMU, as I'm just now upgrading to that sort of memory (I have a gig |
41 |
now but ordered 8 gig coming in next week), but from the help text for the |
42 |
IOMMU option, it only affects certain 32-bit PCI devices, so it's |
43 |
/possible/ you could do without an IOMMU. Note that GART_IOMMU is not |
44 |
configurable as a module, so you either have it enabled and built-in or |
45 |
not. If you do, then the kernel GART must be built-in and non-modularized |
46 |
as well, so unless the NVidia version supports the same thing, you don't |
47 |
have the choice of running the Nvidia GART with IOMMU support, period, |
48 |
you'll have to choose one or the other. Of course, if the Nvidia GART |
49 |
supports it or if you have <=3G memmory, there's no issue, simply disable |
50 |
the kernel IOMMU and you'll be able to disable the kernel GART as well |
51 |
(assuming that's the config option that's killing the kernel GART option |
52 |
presently). |
53 |
|
54 |
-- |
55 |
Duncan - List replies preferred. No HTML msgs. |
56 |
"Every nonfree program has a lord, a master -- |
57 |
and if you use the program, he is your master." Richard Stallman in |
58 |
http://www.linuxdevcenter.com/pub/a/linux/2004/12/22/rms_interview.html |
59 |
|
60 |
|
61 |
-- |
62 |
gentoo-amd64@g.o mailing list |