Commit 9fa68eae authored by Knut Petersen's avatar Knut Petersen Committed by Linus Torvalds

[PATCH] framebuffer: new driver for cyberblade/i1 graphics core

This is a framebuffer driver for the Cyberblade/i1 graphics core.

Currently tridenfb claims to support the cyberblade/i1 graphics core.  This
is of very limited truth.  Even vesafb is faster and provides more working
modes and a much better quality of the video signal.  There is a great
number of bugs in tridentfb ...  but most often it is impossible to decide
if these bugs are real bugs or if fixing them for the cyberblade/i1 core
would break support for one of the other supported chips.

Tridentfb seems to be unmaintained,and documentation for most of the
supported chips is not available.  So "fixing" cyberblade/i1 support inside
of tridentfb was not an option, it would have caused numerous
if(CYBERBLADEi1) else ...  cases and would have rendered the code to be
almost unmaintainable.

A first version of this driver was published on 2005-07-31.  A fix for a
bug reported by Jochen Hein was integrated as well as some changes
requested by Antonino A.  Daplas.

A message has been added to tridentfb to inform current users of tridentfb
to switch to cyblafb if the cyberblade/i1 graphics core is detected.

This patch is one logical change, but because of the included documentation
it is bigger than 70kb.  Therefore it is not sent to lkml and
linux-fbdev-devel,
Signed-off-by: default avatarKnut Petersen <Knut_Petersen@t-online.de>
Cc: Muli Ben-Yehuda <mulix@mulix.org>
Acked-by: default avatarAntonino Daplas <adaplas@pol.net>
Signed-off-by: default avatarAndrew Morton <akpm@osdl.org>
Signed-off-by: default avatarLinus Torvalds <torvalds@osdl.org>
parent 6062bfa1
Bugs
====
I currently don't know of any bug. Please do send reports to:
- linux-fbdev-devel@lists.sourceforge.net
- Knut_Petersen@t-online.de.
Untested features
=================
All LCD stuff is untested. If it worked in tridentfb, it should work in
cyblafb. Please test and report the results to Knut_Petersen@t-online.de.
Thanks to
=========
* Alan Hourihane, for writing the X trident driver
* Jani Monoses, for writing the tridentfb driver
* Antonino A. Daplas, for review of the first published
version of cyblafb and some code
* Jochen Hein, for testing and a helpfull bug report
Available Documentation
=======================
Apollo PLE 133 Chipset VT8601A North Bridge Datasheet, Rev. 1.82, October 22,
2001, available from VIA:
http://www.viavpsd.com/product/6/15/DS8601A182.pdf
The datasheet is incomplete, some registers that need to be programmed are not
explained at all and important bits are listed as "reserved". But you really
need the datasheet to understand the code. "p. xxx" comments refer to page
numbers of this document.
XFree/XOrg drivers are available and of good quality, looking at the code
there is a good idea if the datasheet does not provide enough information
or if the datasheet seems to be wrong.
#
# Sample fb.modes file
#
# Provides an incomplete list of working modes for
# the cyberblade/i1 graphics core.
#
# The value 4294967256 is used instead of -40. Of course, -40 is not
# a really reasonable value, but chip design does not always follow
# logic. Believe me, it's ok, and it's the way the BIOS does it.
#
# fbset requires 4294967256 in fb.modes and -40 as an argument to
# the -t parameter. That's also not too reasonable, and it might change
# in the future or might even be differt for your current version.
#
mode "640x480-50"
geometry 640 480 640 3756 8
timings 47619 4294967256 24 17 0 216 3
endmode
mode "640x480-60"
geometry 640 480 640 3756 8
timings 39682 4294967256 24 17 0 216 3
endmode
mode "640x480-70"
geometry 640 480 640 3756 8
timings 34013 4294967256 24 17 0 216 3
endmode
mode "640x480-72"
geometry 640 480 640 3756 8
timings 33068 4294967256 24 17 0 216 3
endmode
mode "640x480-75"
geometry 640 480 640 3756 8
timings 31746 4294967256 24 17 0 216 3
endmode
mode "640x480-80"
geometry 640 480 640 3756 8
timings 29761 4294967256 24 17 0 216 3
endmode
mode "640x480-85"
geometry 640 480 640 3756 8
timings 28011 4294967256 24 17 0 216 3
endmode
mode "800x600-50"
geometry 800 600 800 3221 8
timings 30303 96 24 14 0 136 11
endmode
mode "800x600-60"
geometry 800 600 800 3221 8
timings 25252 96 24 14 0 136 11
endmode
mode "800x600-70"
geometry 800 600 800 3221 8
timings 21645 96 24 14 0 136 11
endmode
mode "800x600-72"
geometry 800 600 800 3221 8
timings 21043 96 24 14 0 136 11
endmode
mode "800x600-75"
geometry 800 600 800 3221 8
timings 20202 96 24 14 0 136 11
endmode
mode "800x600-80"
geometry 800 600 800 3221 8
timings 18939 96 24 14 0 136 11
endmode
mode "800x600-85"
geometry 800 600 800 3221 8
timings 17825 96 24 14 0 136 11
endmode
mode "1024x768-50"
geometry 1024 768 1024 2815 8
timings 19054 144 24 29 0 120 3
endmode
mode "1024x768-60"
geometry 1024 768 1024 2815 8
timings 15880 144 24 29 0 120 3
endmode
mode "1024x768-70"
geometry 1024 768 1024 2815 8
timings 13610 144 24 29 0 120 3
endmode
mode "1024x768-72"
geometry 1024 768 1024 2815 8
timings 13232 144 24 29 0 120 3
endmode
mode "1024x768-75"
geometry 1024 768 1024 2815 8
timings 12703 144 24 29 0 120 3
endmode
mode "1024x768-80"
geometry 1024 768 1024 2815 8
timings 11910 144 24 29 0 120 3
endmode
mode "1024x768-85"
geometry 1024 768 1024 2815 8
timings 11209 144 24 29 0 120 3
endmode
mode "1280x1024-50"
geometry 1280 1024 1280 2662 8
timings 11114 232 16 39 0 160 3
endmode
mode "1280x1024-60"
geometry 1280 1024 1280 2662 8
timings 9262 232 16 39 0 160 3
endmode
mode "1280x1024-70"
geometry 1280 1024 1280 2662 8
timings 7939 232 16 39 0 160 3
endmode
mode "1280x1024-72"
geometry 1280 1024 1280 2662 8
timings 7719 232 16 39 0 160 3
endmode
mode "1280x1024-75"
geometry 1280 1024 1280 2662 8
timings 7410 232 16 39 0 160 3
endmode
mode "1280x1024-80"
geometry 1280 1024 1280 2662 8
timings 6946 232 16 39 0 160 3
endmode
mode "1280x1024-85"
geometry 1280 1024 1280 2662 8
timings 6538 232 16 39 0 160 3
endmode
Speed
=====
CyBlaFB is much faster than tridentfb and vesafb. Compare the performance data
for mode 1280x1024-[8,16,32]@61 Hz.
Test 1: Cat a file with 2000 lines of 0 characters.
Test 2: Cat a file with 2000 lines of 80 characters.
Test 3: Cat a file with 2000 lines of 160 characters.
All values show system time use in seconds, kernel 2.6.12 was used for
the measurements. 2.6.13 is a bit slower, 2.6.14 hopefully will include a
patch that speeds up kernel bitblitting a lot ( > 20%).
+-----------+-----------------------------------------------------+
| | not accelerated |
| TRIDENTFB +-----------------+-----------------+-----------------+
| of 2.6.12 | 8 bpp | 16 bpp | 32 bpp |
| | noypan | ypan | noypan | ypan | noypan | ypan |
+-----------+--------+--------+--------+--------+--------+--------+
| Test 1 | 4.31 | 4.33 | 6.05 | 12.81 | ---- | ---- |
| Test 2 | 67.94 | 5.44 | 123.16 | 14.79 | ---- | ---- |
| Test 3 | 131.36 | 6.55 | 240.12 | 16.76 | ---- | ---- |
+-----------+--------+--------+--------+--------+--------+--------+
| Comments | | | completely bro- |
| | | | ken, monitor |
| | | | switches off |
+-----------+-----------------+-----------------+-----------------+
+-----------+-----------------------------------------------------+
| | accelerated |
| TRIDENTFB +-----------------+-----------------+-----------------+
| of 2.6.12 | 8 bpp | 16 bpp | 32 bpp |
| | noypan | ypan | noypan | ypan | noypan | ypan |
+-----------+--------+--------+--------+--------+--------+--------+
| Test 1 | ---- | ---- | 20.62 | 1.22 | ---- | ---- |
| Test 2 | ---- | ---- | 22.61 | 3.19 | ---- | ---- |
| Test 3 | ---- | ---- | 24.59 | 5.16 | ---- | ---- |
+-----------+--------+--------+--------+--------+--------+--------+
| Comments | broken, writing | broken, ok only | completely bro- |
| | to wrong places | if bgcolor is | ken, monitor |
| | on screen + bug | black, bug in | switches off |
| | in fillrect() | fillrect() | |
+-----------+-----------------+-----------------+-----------------+
+-----------+-----------------------------------------------------+
| | not accelerated |
| VESAFB +-----------------+-----------------+-----------------+
| of 2.6.12 | 8 bpp | 16 bpp | 32 bpp |
| | noypan | ypan | noypan | ypan | noypan | ypan |
+-----------+--------+--------+--------+--------+--------+--------+
| Test 1 | 4.26 | 3.76 | 5.99 | 7.23 | ---- | ---- |
| Test 2 | 65.65 | 4.89 | 120.88 | 9.08 | ---- | ---- |
| Test 3 | 126.91 | 5.94 | 235.77 | 11.03 | ---- | ---- |
+-----------+--------+--------+--------+--------+--------+--------+
| Comments | vga=0x307 | vga=0x31a | vga=0x31b not |
| | fh=80kHz | fh=80kHz | supported by |
| | fv=75kHz | fv=75kHz | video BIOS and |
| | | | hardware |
+-----------+-----------------+-----------------+-----------------+
+-----------+-----------------------------------------------------+
| | accelerated |
| CYBLAFB +-----------------+-----------------+-----------------+
| | 8 bpp | 16 bpp | 32 bpp |
| | noypan | ypan | noypan | ypan | noypan | ypan |
+-----------+--------+--------+--------+--------+--------+--------+
| Test 1 | 8.02 | 0.23 | 19.04 | 0.61 | 57.12 | 2.74 |
| Test 2 | 8.38 | 0.55 | 19.39 | 0.92 | 57.54 | 3.13 |
| Test 3 | 8.73 | 0.86 | 19.74 | 1.24 | 57.95 | 3.51 |
+-----------+--------+--------+--------+--------+--------+--------+
| Comments | | | |
| | | | |
| | | | |
| | | | |
+-----------+-----------------+-----------------+-----------------+
TODO / Missing features
=======================
Verify LCD stuff "stretch" and "center" options are
completely untested ... this code needs to be
verified. As I don't have access to such
hardware, please contact me if you are
willing run some tests.
Interlaced video modes The reason that interleaved
modes are disabled is that I do not know
the meaning of the vertical interlace
parameter. Also the datasheet mentions a
bit d8 of a horizontal interlace parameter,
but nowhere the lower 8 bits. Please help
if you can.
low-res double scan modes Who needs it?
accelerated color blitting Who needs it? The console driver does use color
blitting for nothing but drawing the penguine,
everything else is done using color expanding
blitting of 1bpp character bitmaps.
xpanning Who needs it?
ioctls Who needs it?
TV-out Will be done later
??? Feel free to contact me if you have any
feature requests
CyBlaFB is a framebuffer driver for the Cyberblade/i1 graphics core integrated
into the VIA Apollo PLE133 (aka vt8601) south bridge. It is developed and
tested using a VIA EPIA 5000 board.
Cyblafb - compiled into the kernel or as a module?
==================================================
You might compile cyblafb either as a module or compile it permanently into the
kernel.
Unless you have a real reason to do so you should not compile both vesafb and
cyblafb permanently into the kernel. It's possible and it helps during the
developement cycle, but it's useless and will at least block some otherwise
usefull memory for ordinary users.
Selecting Modes
===============
Startup Mode
============
First of all, you might use the "vga=???" boot parameter as it is
documented in vesafb.txt and svga.txt. Cyblafb will detect the video
mode selected and will use the geometry and timings found by
inspecting the hardware registers.
video=cyblafb vga=0x317
Alternatively you might use a combination of the mode, ref and bpp
parameters. If you compiled the driver into the kernel, add something
like this to the kernel command line:
video=cyblafb:1280x1024,bpp=16,ref=50 ...
If you compiled the driver as a module, the same mode would be
selected by the following command:
modprobe cyblafb mode=1280x1024 bpp=16 ref=50 ...
None of the modes possible to select as startup modes are affected by
the problems described at the end of the next subsection.
Mode changes using fbset
========================
You might use fbset to change the video mode, see "man fbset". Cyblafb
generally does assume that you know what you are doing. But it does
some checks, especially those that are needed to prevent you from
damaging your hardware.
- only 8, 16, 24 and 32 bpp video modes are accepted
- interlaced video modes are not accepted
- double scan video modes are not accepted
- if a flat panel is found, cyblafb does not allow you
to program a resolution higher than the physical
resolution of the flat panel monitor
- cyblafb does not allow xres to differ from xres_virtual
- cyblafb does not allow vclk to exceed 230 MHz. As 32 bpp
and (currently) 24 bit modes use a doubled vclk internally,
the dotclock limit as seen by fbset is 115 MHz for those
modes and 230 MHz for 8 and 16 bpp modes.
Any request that violates the rules given above will be ignored and
fbset will return an error.
If you program a virtual y resolution higher than the hardware limit,
cyblafb will silently decrease that value to the highest possible
value.
Attempts to disable acceleration are ignored.
Some video modes that should work do not work as expected. If you use
the standard fb.modes, fbset 640x480-60 will program that mode, but
you will see a vertical area, about two characters wide, with only
much darker characters than the other characters on the screen.
Cyblafb does allow that mode to be set, as it does not violate the
official specifications. It would need a lot of code to reliably sort
out all invalid modes, playing around with the margin values will
give a valid mode quickly. And if cyblafb would detect such an invalid
mode, should it silently alter the requested values or should it
report an error? Both options have some pros and cons. As stated
above, none of the startup modes are affected, and if you set
verbosity to 1 or higher, cyblafb will print the fbset command that
would be needed to program that mode using fbset.
Other Parameters
================
crt don't autodetect, assume monitor connected to
standard VGA connector
fp don't autodetect, assume flat panel display
connected to flat panel monitor interface
nativex inform driver about native x resolution of
flat panel monitor connected to special
interface (should be autodetected)
stretch stretch image to adapt low resolution modes to
higer resolutions of flat panel monitors
connected to special interface
center center image to adapt low resolution modes to
higer resolutions of flat panel monitors
connected to special interface
memsize use if autodetected memsize is wrong ...
should never be necessary
nopcirr disable PCI read retry
nopciwr disable PCI write retry
nopcirb disable PCI read bursts
nopciwb disable PCI write bursts
bpp bpp for specified modes
valid values: 8 || 16 || 24 || 32
ref refresh rate for specified mode
valid values: 50 <= ref <= 85
mode 640x480 or 800x600 or 1024x768 or 1280x1024
if not specified, the startup mode will be detected
and used, so you might also use the vga=??? parameter
described in vesafb.txt. If you do not specify a mode,
bpp and ref parameters are ignored.
verbosity 0 is the default, increase to at least 2 for every
bug report!
vesafb allows cyblafb to be loaded after vesafb has been
loaded. See sections "Module unloading ...".
Development hints
=================
It's much faster do compile a module and to load the new version after
unloading the old module than to compile a new kernel and to reboot. So if you
try to work on cyblafb, it might be a good idea to use cyblafb as a module.
In real life, fast often means dangerous, and that's also the case here. If
you introduce a serious bug when cyblafb is compiled into the kernel, the
kernel will lock or oops with a high probability before the file system is
mounted, and the danger for your data is low. If you load a broken own version
of cyblafb on a running system, the danger for the integrity of the file
system is much higher as you might need a hard reset afterwards. Decide
yourself.
Module unloading, the vfb method
================================
If you want to unload/reload cyblafb using the virtual framebuffer, you need
to enable vfb support in the kernel first. After that, load the modules as
shown below:
modprobe vfb vfb_enable=1
modprobe fbcon
modprobe cyblafb
fbset -fb /dev/fb1 1280x1024-60 -vyres 2662
con2fb /dev/fb1 /dev/tty1
...
If you now made some changes to cyblafb and want to reload it, you might do it
as show below:
con2fb /dev/fb0 /dev/tty1
...
rmmod cyblafb
modprobe cyblafb
con2fb /dev/fb1 /dev/tty1
...
Of course, you might choose another mode, and most certainly you also want to
map some other /dev/tty* to the real framebuffer device. You might also choose
to compile fbcon as a kernel module or place it permanently in the kernel.
I do not know of any way to unload fbcon, and fbcon will prevent the
framebuffer device loaded first from unloading. [If there is a way, then
please add a description here!]
Module unloading, the vesafb method
===================================
Configure the kernel:
<*> Support for frame buffer devices
[*] VESA VGA graphics support
<M> Cyberblade/i1 support
Add e.g. "video=vesafb:ypan vga=0x307" to the kernel parameters. The ypan
parameter is important, choose any vga parameter you like as long as it is
a graphics mode.
After booting, load cyblafb without any mode and bpp parameter and assign
cyblafb to individual ttys using con2fb, e.g.:
modprobe cyblafb vesafb=1
con2fb /dev/fb1 /dev/tty1
Unloading cyblafb works without problems after you assign vesafb to all
ttys again, e.g.:
con2fb /dev/fb0 /dev/tty1
rmmod cyblafb
I tried the following framebuffer drivers:
- TRIDENTFB is full of bugs. Acceleration is broken for Blade3D
graphics cores like the cyberblade/i1. It claims to support a great
number of devices, but documentation for most of these devices is
unfortunately not available. There is _no_ reason to use tridentfb
for cyberblade/i1 + CRT users. VESAFB is faster, and the one
advantage, mode switching, is broken in tridentfb.
- VESAFB is used by many distributions as a standard. Vesafb does
not support mode switching. VESAFB is a bit faster than the working
configurations of TRIDENTFB, but it is still too slow, even if you
use ypan.
- EPIAFB (you'll find it on sourceforge) supports the Cyberblade/i1
graphics core, but it still has serious bugs and developement seems
to have stopped. This is the one driver with TV-out support. If you
do need this feature, try epiafb.
None of these drivers was a real option for me.
I believe that is unreasonable to change code that announces to support 20
devices if I only have more or less sufficient documentation for exactly one
of these. The risk of breaking device foo while fixing device bar is too high.
So I decided to start CyBlaFB as a stripped down tridentfb.
All code specific to other Trident chips has been removed. After that there
were a lot of cosmetic changes to increase the readability of the code. All
register names were changed to those mnemonics used in the datasheet. Function
and macro names were changed if they hindered easy understanding of the code.
After that I debugged the code and implemented some new features. I'll try to
give a little summary of the main changes:
- calculation of vertical and horizontal timings was fixed
- video signal quality has been improved dramatically
- acceleration:
- fillrect and copyarea were fixed and reenabled
- color expanding imageblit was newly implemented, color
imageblit (only used to draw the penguine) still uses the
generic code.
- init of the acceleration engine was improved and moved to a
place where it really works ...
- sync function has a timeout now and tries to reset and
reinit the accel engine if necessary
- fewer slow copyarea calls when doing ypan scrolling by using
undocumented bit d21 of screen start address stored in
CR2B[5]. BIOS does use it also, so this should be safe.
- cyblafb rejects any attempt to set modes that would cause vclk
values above reasonable 230 MHz. 32bit modes use a clock
multiplicator of 2, so fbset does show the correct values for
pixclock but not for vclk in this case. The fbset limit is 115 MHz
for 32 bpp modes.
- cyblafb rejects modes known to be broken or unimplemented (all
interlaced modes, all doublescan modes for now)
- cyblafb now works independant of the video mode in effect at startup
time (tridentfb does not init all needed registers to reasonable
values)
- switching between video modes does work reliably now
- the first video mode now is the one selected on startup using the
vga=???? mechanism or any of
- 640x480, 800x600, 1024x768, 1280x1024
- 8, 16, 24 or 32 bpp
- refresh between 50 Hz and 85 Hz, 1 Hz steps (1280x1024-32
is limited to 63Hz)
- pci retry and pci burst mode are settable (try to disable if you
experience latency problems)
- built as a module cyblafb might be unloaded and reloaded using
the vfb module and con2vt or might be used together with vesafb
......@@ -626,6 +626,12 @@ M: rmk@arm.linux.org.uk
W: http://www.arm.linux.org.uk/
S: Maintained
CYBLAFB FRAMEBUFFER DRIVER
P: Knut Petersen
M: Knut_Petersen@t-online.de
L: linux-fbdev-devel@lists.sourceforge.net
S: Maintained
CYCLADES 2X SYNC CARD DRIVER
P: Arnaldo Carvalho de Melo
M: acme@conectiva.com.br
......
......@@ -1177,6 +1177,32 @@ config FB_VOODOO1
Please read the <file:Documentation/fb/README-sstfb.txt> for supported
options and other important info support.
config FB_CYBLA
tristate "Cyberblade/i1 support"
depends on FB && PCI
select FB_CFB_IMAGEBLIT
select FB_SOFT_CURSOR
select VIDEO_SELECT
---help---
This driver is supposed to support the Trident Cyberblade/i1
graphics core integrated in the VIA VT8601A North Bridge,
also known as VIA Apollo PLE133.
Status:
- Developed, tested and working on EPIA 5000 and EPIA 800.
- Does work reliable on all systems with CRT/LCD connected to
normal VGA ports.
- Should work on systems that do use the internal LCD port, but
this is absolutely not tested.
Character imageblit, copyarea and rectangle fill are hw accelerated,
ypan scrolling is used by default.
Please do read <file:Documentation/fb/cyblafb/*>.
To compile this driver as a module, choose M here: the
module will be called cyblafb.
config FB_TRIDENT
tristate "Trident support"
depends on FB && PCI
......@@ -1190,8 +1216,12 @@ config FB_TRIDENT
but also on some motherboards. For more information, read
<file:Documentation/fb/tridentfb.txt>
Cyberblade/i1 support will be removed soon, use the cyblafb driver
instead.
Say Y if you have such a graphics board.
To compile this driver as a module, choose M here: the
module will be called tridentfb.
......@@ -1202,7 +1232,6 @@ config FB_TRIDENT_ACCEL
This will compile the Trident frame buffer device with
acceleration functions.
config FB_PM3
tristate "Permedia3 support"