1 |
what i could tell you is that xen is a little behind with the sources, but |
2 |
still if configured in the right way there should be no problem in running a |
3 |
box. you just have to remember that with xen you'll need a "base" system, |
4 |
then another system to run on this base with reduced priviledges and |
5 |
assigned stuff. so you'll just need a base gentoo (stage3 should be fine), |
6 |
then you should configure xen to have another stage3 with installed apps. |
7 |
the xen problem is that you'll need 2 linuxes to run one linux. but you'll |
8 |
be allowed to assign processors fragments or ram portions in the way you |
9 |
like. if you don't really need xen (you would run at least 2 different |
10 |
systems like using one linux i686 one amd64 or one linux and one windows - |
11 |
which should run well since you have vt) i think that using it would be a |
12 |
waste of time. also there are tools that let you install windows in boxes, |
13 |
like vmware and such (they're very good and fast and don't put you in the |
14 |
situation to use xen when xen is way back with kernel support and a lot of |
15 |
the new stuff in new pcs is supported only from the 2.6.21 or later kernel |
16 |
like the jmicron controller used in asus mobos). so my advice is avoid using |
17 |
xen unless it is really, really and i mean really necessary (you'll need 2 |
18 |
systems and a lot of configuration to run a single linux box that is slower |
19 |
than a native one). |
20 |
|
21 |
for the radeon driver, always use the ~amd64/~x86. always, since the driver |
22 |
is updated every month and has new things. for example the latest portage |
23 |
8.433 has aiglx support support for the latest xorg and xorg-server (even if |
24 |
compiz won't work with xorg-server-1.4) and it has a great performance |
25 |
boost. also the new radeon has a hidden option to be added into the device |
26 |
section for enabling a render: |
27 |
Option "TexturedXrender" "true" |
28 |
it seems that this has a performance boost with rendering. for ati, you'll |
29 |
also have to control that you have the following files in /usr/lib/ |
30 |
libGL.so.1, libGL.so.1.2. if one of them is missing create a symlink to the |
31 |
other libGL.so.x. this is due to a soname bug present in the latest fglrx |
32 |
releases. also be aware that there's a memory leak bug and apps like |
33 |
glxgears, amarok, firefox or others that use opengl would continue to suck |
34 |
out system memory. stay tuned to see if this happens and restart xorg when |
35 |
the system ram has reached a critical limit like 85-90%. |
36 |
or you may want to try the xf86 ati drivers for r500 (your chipset) because |
37 |
airlied, who is the main dev for the r500/600 radeon driver now works for |
38 |
red hat, has added some cool features like radeon modesetting and other |
39 |
stuff in the latest releases and you should be able to get a good working |
40 |
driver. you'll just have to add radeon support to xorg and use the radeon |
41 |
driver from ~amd64/~x86 branch. with ati these days is almost necessary to |
42 |
use this branch since the new releases add a lot of new features and fixes |
43 |
due to the recent amd/ati releasing specs and helping devs. i also would say |
44 |
you to remove the fbdev and all framebuffer ati modules or so from kernel |
45 |
since they're buggy and won't work correctly. also fb is useless if you use |
46 |
vesa or other stuff. |
47 |
is you use the fglrx then you'll also need to add the following stuff to the |
48 |
device section of your xorg.conf: |
49 |
Option "RenderAccel" "true" <--- needed for rendering stuff: forces |
50 |
sw rendering where hw doesn't work. |
51 |
Option "AllowGLXWithComposite" "true" <--- needed for compiz/beryl |
52 |
Option "HWCursor" "On" <---- needed to fix a bug with the mouse |
53 |
cursor |
54 |
Option "OpenGLOverlay" "off" <---- disables opengl overlay and |
55 |
doesn't work with Xv enabled |
56 |
Option "VideoOverlay" "on" <--- enables Xv overlay doesn't work |
57 |
with opengloverlay enabled |
58 |
Option "no_dri" "no" <---- use dri |
59 |
Option "UseInternalAGPGART" "no" <---- this is needed for some |
60 |
agpgart issues |
61 |
Option "FSAAEnable" "no" <---- disables fsaa since it is buggy on |
62 |
linux |
63 |
Option "BlockSignalsOnLock" "on" <----- needed to a have a stable |
64 |
driver |
65 |
Option "Capabilities" "0x00000000" <---- set capabilities to default |
66 |
since some cards are buggy with non default capabilities |
67 |
Option "mtrr" "off" <---- disables mtrr |
68 |
Option "XAANoOffscreenPixmaps" "true" <---- needed to fix the |
69 |
watermark issue of the latest fglrx releases |
70 |
Option "UseFastTLS" "0" <---- if you use wine don't set this to |
71 |
higher than 1. |
72 |
|
73 |
also for a working compiz you'll need this section: |
74 |
|
75 |
Section "Extensions" |
76 |
Option "DAMAGE" "true" |
77 |
Option "RENDER" "true" |
78 |
Option "Composite" "Enable" |
79 |
EndSection |
80 |
|
81 |
this is the working configuration for a fglrx running with compiz. if you |
82 |
experience issue with compiz then set the compiz setting from |
83 |
"texturefrompixmap" rendering to "force copy". |
84 |
|
85 |
2007/12/8, Florent Ouchet <ouchet.florent@×××××××.net>: |
86 |
> |
87 |
> Florent Ouchet a écrit : |
88 |
> > I experienced some problems with the following combination of software: |
89 |
> > - xen-sources 2.6.20-r6 |
90 |
> > - xen-tools 3.1.2 |
91 |
> > - ati-drivers 8.40.4 |
92 |
> > - linux 2.6.20 |
93 |
> > - xorg-server 1.3.0.0-r2 (build with support for fbdev, vesa and fglrx) |
94 |
> > |
95 |
> > The system is a Dom0 of a paravirtualized system: |
96 |
> > - Core 2 Duo 2.4GHz (vT enabled) |
97 |
> > - 2GB RAM |
98 |
> > - ATI Radeon x1300 video card |
99 |
> > |
100 |
> > Is anyone running such a configuration with success? I have to stay |
101 |
> > with the vesa driver without hardware acceleration. |
102 |
> |
103 |
> No one managed to get such a config running with optimal performances? :( |
104 |
> |
105 |
> |
106 |
> -- |
107 |
> gentoo-amd64@g.o mailing list |
108 |
> |
109 |
> |
110 |
|
111 |
|
112 |
-- |
113 |
dott. ing. beso |