* Observations on CVS kernel
@ 1998-07-26 1:51 Bj|rn Augustsson
1998-07-26 1:56 ` David S. Miller
1998-07-26 19:20 ` Steve Dunham
0 siblings, 2 replies; 3+ messages in thread
From: Bj|rn Augustsson @ 1998-07-26 1:51 UTC (permalink / raw
To: ultralinux
Hi!
I installed ultrapenguin 1.0.9, and then got the latest kernel from CVS,
to test the new console system. Especially the multihead part.
The kernel is 1.2.109, from about 1700 26-Jul GMT.
The machine is a U1140E with two disks, a floppy, a CD, a Creator and a
CGsix.
I had a few problems with getting modules to work, so I decided to build
a monolithic kernel. I had problems getting the nfsd support to compile
because of __my_nfsfh_is_too_big() being unresolved in /fs/nfsd/nfsfh.c
so I configured without it. Looking at the code, I could probably just
have commented the (single) usage of it out, but I don't need the machine
to serve NFS anyway.
Mostly everything seems to work fine, but I have a few questions/comments:
My first try at a monolithic kernel failed to boot because it was to big.
(Can't remember the error message, but something along the lines of
"kernel image to big to fit in destination".)
I stripped some things from it and rebuilt without a problem.
The first one was about 1MB compressed, and the new one is about 810K
compressed. (2.2MB uncompressed)
What's the limit?
If I do L1A and then "go", the machine hangs. Under Solaris this generally
works. (I have never seen it hang, a friend says he saw a panic once.)
If I start X (BTW, very neat startup picture!), and then switch VCs, the
parts of the screen where the "border" (in the textmode) is isn't cleared.
(I still see sort of like a frame around the text area, where the old X
graphics are left.)
(X being the X I got when I installed UP 1.0.9. I haven't installed any
special fbcon X-server. I'm a bit surprised this server worked at all.)
How would I go about to set up a dual head? (X or textmode)
Both framebuffers get detected properly when booting, and the second one
(the CGsix) does _something_ (the screen clicks), but after that, no
activity ever on the second screen.
Thanks,
/August.
--
Wrong on most accounts. const Foo *foo; and Foo const *foo; mean the same: foo
being a pointer to const Foo. const Foo const *foo; would mean the same but is
illegal (double const). You are confusing this with Foo * const foo; and const
Foo * const foo; respectively. -David Kastrup, comp.os.linux.development.system
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: Observations on CVS kernel
1998-07-26 1:51 Observations on CVS kernel Bj|rn Augustsson
@ 1998-07-26 1:56 ` David S. Miller
1998-07-26 19:20 ` Steve Dunham
1 sibling, 0 replies; 3+ messages in thread
From: David S. Miller @ 1998-07-26 1:56 UTC (permalink / raw
To: ultralinux
Date: Sun, 26 Jul 1998 03:51:00 +0200
From: Bj|rn Augustsson <d3august@dtek.chalmers.se>
I had a few problems with getting modules to work, so I decided to
build a monolithic kernel. I had problems getting the nfsd support
to compile because of __my_nfsfh_is_too_big() being unresolved in
/fs/nfsd/nfsfh.c so I configured without it. Looking at the code, I
could probably just have commented the (single) usage of it out,
but I don't need the machine to serve NFS anyway.
This is a run time check, the symbol is unresolved on purpose. If the
structure in question is small enough the test will be optimized away
by the compiler. If you try to "fix" this by removing the test you
will have an nfsd which will not operate correctly at all and probably
crash the machine when used. People are actively working on a fix for
this.
What's the limit?
About 2.6MB
If I do L1A and then "go", the machine hangs. Under Solaris this
generally works. (I have never seen it hang, a friend says he saw a
panic once.)
It doesn't work because the support for it was never added. Depending
upon how much of the higher priority work gets done on the sparc64
kernel, this may or may not ever get implemented.
If I start X (BTW, very neat startup picture!), and then switch
VCs, the parts of the screen where the "border" (in the textmode)
is isn't cleared. (I still see sort of like a frame around the
text area, where the old X graphics are left.)
(X being the X I got when I installed UP 1.0.9. I haven't installed
any special fbcon X-server. I'm a bit surprised this server worked
at all.)
We worked to keep the existing X server interface to using the
graphics card work as it did before. As for the VC switching problem,
I'll let Jakub handle it.
How would I go about to set up a dual head? (X or textmode) Both
framebuffers get detected properly when booting, and the second one
(the CGsix) does _something_ (the screen clicks), but after that,
no activity ever on the second screen.
It might work if you create specific device nodes, or you run the X
server with explicit device arguments on the command line.
For dual cgsix for example I just do:
mknod /dev/cgsix0 c 29 0
mknod /dev/cgsix1 c 29 1
or add -dev arguments to the X command line.
Later,
David S. Miller
davem@dm.cobaltmicro.com
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: Observations on CVS kernel
1998-07-26 1:51 Observations on CVS kernel Bj|rn Augustsson
1998-07-26 1:56 ` David S. Miller
@ 1998-07-26 19:20 ` Steve Dunham
1 sibling, 0 replies; 3+ messages in thread
From: Steve Dunham @ 1998-07-26 19:20 UTC (permalink / raw
To: ultralinux
"David S. Miller" <davem@dm.cobaltmicro.com> writes:
> Date: Sun, 26 Jul 1998 03:51:00 +0200
> From: Bj|rn Augustsson <d3august@dtek.chalmers.se>
> If I start X (BTW, very neat startup picture!), and then switch
> VCs, the parts of the screen where the "border" (in the textmode)
> is isn't cleared. (I still see sort of like a frame around the
> text area, where the old X graphics are left.)
>
> (X being the X I got when I installed UP 1.0.9. I haven't installed
> any special fbcon X-server. I'm a bit surprised this server worked
> at all.)
>
> We worked to keep the existing X server interface to using the
> graphics card work as it did before. As for the VC switching problem,
> I'll let Jakub handle it.
>
> How would I go about to set up a dual head? (X or textmode) Both
> framebuffers get detected properly when booting, and the second one
> (the CGsix) does _something_ (the screen clicks), but after that,
> no activity ever on the second screen.
> It might work if you create specific device nodes, or you run the X
> server with explicit device arguments on the command line.
> For dual cgsix for example I just do:
> mknod /dev/cgsix0 c 29 0
> mknod /dev/cgsix1 c 29 1
> or add -dev arguments to the X command line.
I use -dev. I have xdm running, so I edit "/etc/X11/xdm/Xservers" to
read:
:0 local /usr/X11R6/bin/X -dev /dev/fb1:/dev/fb0 -fp tcp/localhost:7100 -dpi 75
(The graphics cards are a cgsix and a cgfourteen BTW. You'll want to
leave out the "-fp" bit unless you use xfs.)
Steve
dunham@cps.msu.edu
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~1998-07-26 19:20 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
1998-07-26 1:51 Observations on CVS kernel Bj|rn Augustsson
1998-07-26 1:56 ` David S. Miller
1998-07-26 19:20 ` Steve Dunham
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.