Due to some problems with PCI transactions, Geode LX needs the ROM cache properties to be write-serialize + cache disabled by runtime. More details below.
Add mainboard_pre_payload() call to each mainboard as the final coreboot function before the payload is called by stage1.
Note that this patch also grows the bootblock from 16K to 20K to make room for mainboard_pre_payload().
"The problem is a transaction depth issue and bottlenecks inside the GX
and LX that go across PCI. The conditions are very complicated but it
comes down to we need write serialization for writes to PCI. If you
look in the data book you can't have write serialization and the cache
enabled on a given area. During coreboot we don't have to worry about
a write or a PCI bus master so I think we can enable caching the ROM.
After coreboot we can't be sure what will happen in the system so we
need to set it up to be safe. For example flashrom just clears the
write protect bit. If the cache were enabled (no write serialization)
and flashrom was writing the ROM we would be in a precarious position.
A PCI bus master doing a read or a write that has a hit on a tag
would cause enough bottleneck conditions that it might hit the bug. We
could change flashrom but that doesn't help other tools. We need to
leave the system in a safe state. Also, caching the ROM after it is no
longer used doesn't make much sense. So, we need a call just before
the payload runs to clean up the system."
Signed-off-by: Marc Jones <marc.jones@amd.com>
Acked-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@573 f3766cd6-281f-0410-b1cd-43a5c92072e9
turn gets cache turned on much sooner. The system boots a bit faster.
We're still far too slow, perhaps because we are not caching ROM?
Index: arch/x86/Makefile
Add ../../northbridge/amd/geodelx/geodelxinit.o object
Index: arch/x86/geodelx/geodelxinit.c
add sizeram function.
Index: arch/x86/geodelx/stage1.c
add called to northbridge_init_early()
Index: northbridge/amd/geodelx/Makefile
remove geodelxinit.o object
Index: northbridge/amd/geodelx/geodelx.c
remove call to northbridge_init_early()
remove sizeram function.
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@572 f3766cd6-281f-0410-b1cd-43a5c92072e9
emulation.
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Marc Jones <marc.jones@amd.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@570 f3766cd6-281f-0410-b1cd-43a5c92072e9
- I left LB_TAG_ intact because they are used by the payloads
- file renames are still missing. see next commit
- some lb_ renames might be missing. feel free to provide patches.
Signed-off-by: Stefan Reinauer <stepan@coresystems.de>
Acked-by: Stefan Reinauer <stepan@coresystems.de>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@564 f3766cd6-281f-0410-b1cd-43a5c92072e9
is running the VSA in the geode lx northbridge.
It builds but is not tested.
lar.h: make LAR functions SHARED
lar.c: make process_file non-static (i.e. global)
vsmsetup.c: modify to use LAR functions.
stage1.c: new function, init_archive, which is SHARED and will set up
the initial archive struct.
Note that some work remains. The use of unsigned longs and unsigned
shorts should be changed to u32/u16 as Carl-Daniel has pointed out,
Because this change requires changes elsewhere I am not including them
in this patch.
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@561 f3766cd6-281f-0410-b1cd-43a5c92072e9
processor. However, nothing in the log tells the user why the processor
is being reset.
Example log follows:
LinuxBIOS-3.0.0 Fri Jan 11 15:53:52 MST 2008 starting...
Choosing fallback boot.
[...]
Fallback failed. Try normal boot
LAR: Attempting to open 'normal/initram/segment0'.
[...]
LAR: CHECK normal/initram/segment0 @ 0xfffc49b0
start 0xfffc4a00 len 5564 reallen 5564 compression 0 entry 0x000010ca
loadaddress 0x00000000
Entry point is 0xfffc5aca
pll_reset: read msr 0x4c000014
_MSR GLCP_SYS_RSTPLL (4c000014) value is: 00000398:0000181e
Configuring PLL
LinuxBIOS-3.0.0 Fri Jan 11 15:53:52 MST 2008 starting...
Choosing fallback boot.
[...]
Print an informative message before resetting the processor.
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@553 f3766cd6-281f-0410-b1cd-43a5c92072e9
make: *** No rule to make target `LinuxBIOSv3/build/stage0-prefixed.o',
needed by `LinuxBIOSv3/build/linuxbios.initram'.
Stop.
Signed-off-by: Robert Millan <rmh@aybabtu.com>
Acked-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@552 f3766cd6-281f-0410-b1cd-43a5c92072e9
generic x86 code. For the old supported CAR sizes, the newly generated
code is equivalent, so it should be a no-brainer.
The patch is identical (except one build fix) to what was committed in
r3038 to v2.
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Marc Jones <marc.jones@amd.com>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@551 f3766cd6-281f-0410-b1cd-43a5c92072e9
numeric constants and #defines for such constants. Since grepping the
tree shouldn't be necessary to find a POST code and we already have
too many duplicated POST codes, gather almost all of them in a common
header file.
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@549 f3766cd6-281f-0410-b1cd-43a5c92072e9
return from disable_car.
- Move the cache as ram memory to 0x80000 instead of 0xc8000, as the C
range is really tricky to get right :-)
- Modify the geode disable_car to ensure the cache is flushed to ram on
the wbinvd.
With these changes, I get a payload loaded.
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@546 f3766cd6-281f-0410-b1cd-43a5c92072e9
available to C.
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@543 f3766cd6-281f-0410-b1cd-43a5c92072e9
any object already mentioned in STAGE0_OBJ.
This saves 386 bytes in qemu stage2 (~240 bytes after LZMA compression).
Build tested and runtime tested in Qemu.
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Peter Stuge <peter@stuge.se>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@540 f3766cd6-281f-0410-b1cd-43a5c92072e9
_MAINOBJECT #defined. Calls from all other files ended up in nirvana
because the compiler was not able to calculate the address of the
wrapper for the absolute call. The linker tried, but failed miserably.
Use the -combine flag and compile all of initram at once. This enables
GCC to calculate the address of the abscall wrapper, resulting in
working code.
Segher Boessenkool thinks the patched code works only by accident
because GCC has no way to specify generation of XIP code. According to
him, future GCC versions or other circumstances may break the code.
While this patch makes code work for now, it does NOT check whether the
generated code tries to write to memory outside the stack (general
writable data). That will of course fail, but I hope porters are smart
enough to avoid that.
Corey Osgood writes:
Great work tracking this down! This is okay for now, but we need to look
for a better solution in the future. Counting on porters who may or may
not remember this discussion to avoid something isn't good
future-proofing.
Checking the ELF sections for read-write data and stopping the build
with an error could make this future-proof.
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Corey Osgood <corey.osgood@gmail.com>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@537 f3766cd6-281f-0410-b1cd-43a5c92072e9
into lib/northbridgelib.c.
Signed-off-by: Uwe Hermann <uwe@hermann-uwe.de>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@534 f3766cd6-281f-0410-b1cd-43a5c92072e9
lib/lar.c:load_file_segments() and greatly simplify arch/x86/stage1.c
code as a result. While I'm at it, improve the LAR segmentation
abstraction. Stage 1 code should not have to care about internal
representation of segments, that knowledge belongs into lib/lar.c.
Constification of most function parameters in lib/lar.c led to a few
other now obvious code removals.
Build tested and runtime tested in Qemu.
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@530 f3766cd6-281f-0410-b1cd-43a5c92072e9
the x86 arch makefile. This makes porting to a new mainboard less
confusing.
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@529 f3766cd6-281f-0410-b1cd-43a5c92072e9
created by parsing the ELF file linuxbios.stage2.o. That way, we avoid
manual specification of the entry point for stage2 code. A few LAR
functions are now unused and can be removed in a following cleanup
session.
Another suggested cleanup would be factoring out a load_all_segments
function, resulting in less code duplication.
Ron says:
> This is great. We killed the blob. In the movie, it was impossible.
> FYI, the new approach works on real hardware, I am getting nice printk
> messages from stage1!
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@524 f3766cd6-281f-0410-b1cd-43a5c92072e9
and create a valid LAR header with an entry value. This will fix the
problems Ron is having on the ALIX 1C and probably allow it to boot.
It also sets the text base of the .o to 0, instead of whatever wacky
value gld is choosing, so all platforms will have the same value.
Get away from worrying about potential gld bugs, now and in the future.
As added benefit, we obsolete a lot of code without introducing
new code.
Qemu target entry point debugging has been added to be able to spot
problems with entry points in the future.
This patch is a joint work of Ron and Carl-Daniel.
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@523 f3766cd6-281f-0410-b1cd-43a5c92072e9
Signed-off-by: Uwe Hermann <uwe@hermann-uwe.de>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@519 f3766cd6-281f-0410-b1cd-43a5c92072e9
to have code in header files.
Signed-off-by: Uwe Hermann <uwe@hermann-uwe.de>
Acked-by: Stefan Reinauer <stepan@coresystems.de>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@518 f3766cd6-281f-0410-b1cd-43a5c92072e9
post_code.c (thus drop it from console.h).
Instead, have console.h #include post_code.h and drop other
explicit #includes of post_code.h in the code; console.h is already
included in pretty much every file.
Signed-off-by: Uwe Hermann <uwe@hermann-uwe.de>
Acked-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@513 f3766cd6-281f-0410-b1cd-43a5c92072e9
dereference or worse) in case the archive length is exactly
sizeof(struct lar_header). Such an archive is invalid because the
filename directly after the LAR header is always dereferenced and has to
be at least 1 byte in the "empty filename" case (only terminating \0).
Improve LAR code documentation and reorder variables in one assignment
to make the code more obvious and readable. This will help people
understand what the code does when they look at it half a year from now.
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Uwe Hermann <uwe@hermann-uwe.de>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@506 f3766cd6-281f-0410-b1cd-43a5c92072e9
This fixes for example printk calls from initram code.
Signed-off-by: Stefan Reinauer <stepan@coresystems.de>
Acked-by: Jordan Crouse <jordan.crouse@amd.com>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@505 f3766cd6-281f-0410-b1cd-43a5c92072e9
(I think) when the PAYLOAD_NONE change was put in.
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Stefan Reinauer <stepan@coresystems.de>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@503 f3766cd6-281f-0410-b1cd-43a5c92072e9
example to access the device list from Open Firmware.
Signed-off-by: Jens Freimann <jens@freimann.org>
Acked-by: Uwe Hermann <uwe@hermann-uwe.de>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@500 f3766cd6-281f-0410-b1cd-43a5c92072e9
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Stefan Reinauer <stepan@coresystems.de>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@494 f3766cd6-281f-0410-b1cd-43a5c92072e9
Remove a broken proto from a c file.
Fix protos etc. per uwe's request for const where it made
sense.
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Peter Stuge <peter@stuge.se>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@485 f3766cd6-281f-0410-b1cd-43a5c92072e9
replace with LAR" series; this set is for arch/x86.
note that the payload is now payload/segment0, payload/segment1, etc. I've extended
linuxbios to look for these. Note that you can now see all the things
that get loaded ;they're no longer hidden in an ELF header somewhere.
Elf failures are gone!
Note that I've left legacy elf support in, for now, but recommend we
get rid of it as soon as possible.
patch attached. This is a first pass. lar.c needs some refactoring but
I want to get the cmdline going. You can now have a linux payload and
it will uncompress with no problems.
This has been tested with filo and BOCHS.
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@482 f3766cd6-281f-0410-b1cd-43a5c92072e9
conformant, working) %llx in printk/printf.
While I'm at it, affected lines with 271 characters can be broken into
smaller chunks.
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Uwe Hermann <uwe@hermann-uwe.de>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@479 f3766cd6-281f-0410-b1cd-43a5c92072e9
x86 config). This was tested with gcc 4.1.0 on x86 and gcc 4.2.1 on x86_64.
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Stefan Reinauer <stepan@coresystems.de>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@466 f3766cd6-281f-0410-b1cd-43a5c92072e9
- Simplify lots of code, especially msr-related code.
- Move struct msrinit declaration into msr.h as we use it quite often
and there's no use to duplicate it again and again in each file.
- Remove unrequired variable usage (e.g. numEnabled, msrnum, val, port).
- Remove useless comments.
Signed-off-by: Uwe Hermann <uwe@hermann-uwe.de>
Acked-by: Marc Jones <marc.jones@amd.com>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@464 f3766cd6-281f-0410-b1cd-43a5c92072e9
* statically link linuxbios.initram for now (depends on printk, post_code
and die)
* greatly simplify all makefiles by creating a global Rules.make
* use $(dir $@) in Makefiles instead of absolute paths for mkdir..
* clean up Makefiles by calling components' code locally instead of in the
Makefile. (Remember: one day, no code per mainboard)
* unconditionally create .xcompile in case it changed
* add NM to xcompile
* create $(obj)/linuxbios.map with all symbols of all stages
Signed-off-by: Stefan Reinauer <stepan@coresystems.de>
Acked-by: Uwe Hermann <uwe@hermann-uwe.de>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@463 f3766cd6-281f-0410-b1cd-43a5c92072e9
- Use 'static' for functions and structs which are not meant to be public.
- Use 'const' for variables which are not meant to be modified.
- Move some prototypes into legacy.h where they belong.
- Drop prototypes for non-existing functions.
Signed-off-by: Uwe Hermann <uwe@hermann-uwe.de>
Acked-by: Marc Jones <marc.jones@amd.com>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@460 f3766cd6-281f-0410-b1cd-43a5c92072e9
Signed-off-by: Marc Jones <marc.jones@amd.com>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@459 f3766cd6-281f-0410-b1cd-43a5c92072e9
Signed-off-by: Marc Jones <marc.jones@amd.com>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@457 f3766cd6-281f-0410-b1cd-43a5c92072e9
- support decompression in the lar loader
- support compressed payloads in stage1
(copies payloads to 0x60000 unconditionally)
- hook up decompressors in Makefile
- disable compression for option_table and initram
Signed-off-by: Patrick Georgi <patrick@georgi-clan.de>
Acked-by: Stefan Reinauer <stepan@coresystems.de>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@452 f3766cd6-281f-0410-b1cd-43a5c92072e9
