Xfree 4.2.0 on Sony Picturebook C1VPK - shifted display

[WildBill]

OK, some background:

Machine: Sony Picturebook C1VPK.
Graphics Chipset: ATI Rage Mobility
LCD resolution: 1024x480 (yes, that's correct - it's a half-height screen on an ultraportable)

I've got a 30GB drive in the machine, with the Win2K build that came with it (never use it), Mandrake 8.1, and now Gentoo 1.1a.

I have been using Mandrake 8.1 for some time, and everything works just fine. I decided to move to Gentoo to get a custom-compiled environment in order to get all the speed I could out of the Transmeta Crusoe that powers this machine.

I installed Gentoo in another partition, and got the base OS up. I ran an "emerge xfree" and wet to bed. Woke up to a completed install of X.

I mounted the Mandrake partition, and copied XF86Config-4 to the /etc/X11/ directory on my Gentoo install, then ran "startx". The GUI came up, but it's got a strange white horizontal bar across the screen, about 40 pixels or so from the bottom. This bar is about 20 pixels wide, and below the bar is what appears to be the top of the screen. The mouse works, and when I move them mouse all the way up, the pointer goes off the (physical) top of the screen and appears at the bottom, until it hits the bottom of the white bar.

I can't seem to get a working XF86Config-4 for this configuration. Again, this same file works perfectly under Mandrake 8.1 (which uses Xfree 4.1). I've tried xvidtune and xf86cfg, and I can get the effect to minimize (by using xf86cfg, which splits the bar into two small bars across the top and bottom of the screen) but it's a very noticeable effect still.

Here's the Modeline in question:

[fafhrd]

I have the C1VP (I think the "K" just means it came with Win2k and not Me).

In recent kernels, 2.4.18, and 2.4.19-pre* and 2.4.19-pre*-ac*, the vaio fullscreen console patch seems to be poison to X. I was getting the same exact problems. I cleanly untarred a fresh kernel source, did not add that patch, and recompiled.

Also, although it may just be my laptop, the full screen console patch made the usually inaudible LCD hum very audible (a high pitched wail, actually).

It'd be nice to have full screen console working, but X is much more important to me.

I hope this helps.

ADDENUM:
Forgot to say, I have the exact some modeline, and it works for me (with the above configuration).

[WildBill]

Thanks for replying! I tracked it down this weekend to the full-screen patch - when I didn't apply that, it worked. You're right there...

However, I am wondering if it's more related to Xfree86 4.2 than the kernel. I have Mandrake 8.1 on the same drive (in another partition) running kernel 2.4.18 with the full-screen patch, and it works fine. Mandrake 8.1 uses X 4.1. I tried to emerge X 4.1 under Gentoo, but I get a compilation error everytime I try. I've given up on full-screen console (for now), as X is definitely more important.

FYI, since you're using a Crusoe-powered machine, too - what CFLAG optimizations are you using? I tried compiling with "march=i686" and it made the system so slow it was unbearable. Changing to "mcpu=i686" resulted in much greater speed - it's now much faster than the Mandrake 8.1 install on the same drive.

[fafhrd]

I also noticed pretty early on that -march=i686 just wasn't right for the crusoe chip that is in my c1vp. I ended up going with "-mcpu=i686 -march=i586 -O3 -pipe" as my full, always-on optimizations. I find only using pentium mmx (i.e. i586) instructions or lower, and yet optimizing for i686 code placement works quite well with this chip.

Okay, so now I got one for you: if you use the sonypi driver to allow access to the buttons, jog dial, etc, and meye driver for the camera, have you gotten any power management (sleep?) stuff working? I have not, and from the Documentation dir in the kernel source, it doesn't seem these want to play together. I suppose we'll have to wait until the ACPI stuff is "all done".

[WildBill]

Thanks for the code optimizations... I'll try those.

And no, I haven't got sonypi and ACPI playing nice, at least sleep-wise. I do have sonypi and ACPI working to the point where ACPI reports accurate battery life, so I guess that's something. I wrote the author of the ACPI modules, and he said that sleep/suspend was still very much alpha-code, and we shouldn't expect it to work at all. I can get my machine to sleep by issuing an 'echo "3" > /proc/acpi/sleep', but it won't wake up. The author seemed pleased that it went to sleep at all.

I'll have to try those optimizations. Any idea on how I can do that without b0rking the system? Can I just do bootstrap and emerge system again?