The code in hw/xfree86/os-support/bus/sparcPci.c:simbaCheckBus()
is trying to mimmick VGA routing by disabling I/O space responses
behind the Simba PCI-PCI controller.
Unfortunately, doing this also happens to disable access to the
IDE controller I/O space registers, thus crashing the system. The
granularity of the I/O disabling in the Simba controller is not
fine enough to disable VGA without also disabling the IDE controller
registers.
Formerly we sized an array with a compile time constant, then initialized
its size to the same constant, but the Linux PCI init code would increase
that "constant". So if you happened to have more than 128 PCI devices,
you'd happily scribble into whatever variables happened to be in .bss
after that array.
Only really fixed for Linux atm. Other OSes will simply (still) fail to
work on video devices above the 128th PCI device.
/sys/devices reflects the bus topology, and we don't care that much.
Easier (and more reliable) to just look in /sys/bus/pci/devices, which
is a flat view.
Currently, the call to linuxPciOpenFile() is always made for read
only access which causes the subsequent mmap call to fail when the
memory is mapped read/write.
Xorg #9692
If we're mapping something in the "legacy range" (0-1Mb), we shouldn't
expand the requested range to the entire 0-1Mb range. Typically this
is for mapping the VGA frame buffer, and some platforms support mmap of
the frame buffer but not the entire 0-1Mb range.
For example, HP sx1000 and sx2000 ia64 platforms can have memory from
0-0x9ffff, VGA frame buffer from 0xa0000-0xbffff, and memory from
0xc0000-0xfffff. On these platforms, we can't map the entire 0-1Mb
range with the same attribute because the memory only supports WB,
while the frame buffer supports only UC. But an mmap of just the
frame buffer should work fine.
Mach64 driver bails out on ia64 because it cannot map device
memory. It turns out that some bogus and unneeded code attempts
to find the root bridge of the device and fails to do so proberly
as there this host-to-pci bridge is not existant. This code has
been around for years although it completely unclear what it had
been intended for. Fixing this by eliminating the bogus code.
broken for any 32 bit X server running on a 64 bit kernel) so #ifdef
them out for now. the PCI rework tree will make all this crap go away,
so I think we can tolerate the extra #ifdef for the next release.
Get rid of almost all uses of these definitions. They're still defined for
delinquent out-of-tree drivers, and also for the Mesa build. As well as
for miinitext.c. But largely gone.
function pointers to implement a level of flexability that was never
used. The code also had unused support for extracting a single image
type from a larger expansion ROM.
Fix the spelling of PCI_BIOS_OPEN_FIRMWARE.
Fix a couple errors in #ifdef debug code.
These changes have been tested on x86 and x86-64 Linux.
Class information is not, and never has been, stored there. Therefore,
this is just a bunch of elaborate code to read 0x00000000.
This has received testing on x86 and x86-64 Linux.
Attempt to build xeglmodule.c, not xglxmodule.c.
Add xf86Sbus.h to EXTRA_DIST, as _HEADERS doesn't appear to get the same
treatment as _SOURCES in terms of automatically DISTing.
Jonathan Lim
Adam Jackson
David Nusinow
Matthias Hopf
Jay Estabrook
Eric Anholt
Bjorn Helgaas
Keith Packard
Egbert Eich
Donnie Berkholz
Jesse Barnes
Adam Jackson
Daniel Stone
Ian Romanick
Adam Jackson
Dave Airlie
Ian Romanick
Kevin E Martin
Eric Anholt
Kristian Høgsberg
Alan Coopersmith