Will be necessary for DBE61 automatic memory size selection.
Signed-off-by: Mart Raudsepp <mart.raudsepp@artecdesign.ee>
Acked-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@1014 f3766cd6-281f-0410-b1cd-43a5c92072e9
This is a complete rewrite of my earlier stack checker proposal.
It works for CAR and RAM, has better abstraction and actually gives us
nice results.
The stack checker is default off due to its rather measurable impact on
boot speed.
Diagnostic messages are printed on first initialization, directly after
RAM init and directly before passing control to the payload. Sample qemu
log is attached. Extract from that log follows:
coreboot-3.0.986 Fri Nov 7 04:04:37 CET 2008 starting...
(console_loglevel=8)
Initial lowest stack is 0x0008fe98
Choosing fallback boot.
[...]
Done RAM init code
After RAM init, lowest stack is 0x0008fe30
Done printk() buffer move
[...]
LAR: load_file_segments: Failed for normal/payload
Before handoff to payload, lowest stack is 0x0008bf50
FATAL: No usable payload found.
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/coreboot-v3@1012 f3766cd6-281f-0410-b1cd-43a5c92072e9
cpu setup is nonexistent. No car either. Work remains ...
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@1000 f3766cd6-281f-0410-b1cd-43a5c92072e9
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@998 f3766cd6-281f-0410-b1cd-43a5c92072e9
much of this builds.
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@995 f3766cd6-281f-0410-b1cd-43a5c92072e9
The largest changes are to get_pci_bridge_ops, and related changes to make it
compile and use correct declarations.
While I was doing that I moved the checks for CONFIG_<BUS>_PLUGIN_SUPPORT to
the Makefile.
The only functional difference is a possible NULL dereference in a debug
statement.
I also added a few more consts, now that my other patch is in.
Signed-off-by: Myles Watson <mylesgw@gmail.com>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@983 f3766cd6-281f-0410-b1cd-43a5c92072e9
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@972 f3766cd6-281f-0410-b1cd-43a5c92072e9
Does not yet build
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
Signed-off-by: Corey Osgood <corey.osgood@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@967 f3766cd6-281f-0410-b1cd-43a5c92072e9
Most substantive change is getting rid of 'initialized', which was only
ever needed in v2 due to an implementation mistake.
With Uwe's comments taken into account,
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Myles Watson <mylesgw@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@961 f3766cd6-281f-0410-b1cd-43a5c92072e9
cleanups/fixes.
Fixup device tree code. Add/change methods as needed.
This should help serengeti.
Signed-off-by: Ronald G. Minnich<rminnich@gmail.com>
Acked-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Corey Osgood <corey.osgood@gmail.com>
Acked-by: Marc Jones <marc.jones@amd.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@954 f3766cd6-281f-0410-b1cd-43a5c92072e9
CAR disabling operation. Untested, but it should work.
Please note that the code is incomplete, but that should at least not
affect stage2.
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/coreboot-v3@945 f3766cd6-281f-0410-b1cd-43a5c92072e9
I am totally convinced these are right. I am going on travel for a week
and want these in your hands.
Carl-Daniel as acked these, but for lack of time to get firefox going
right now,
Current serengeti output
DRAM(40)01000000-00ffffff, ->(1), R, W, 2 nodes, 1
DRAM(48)01000000-00ffffff, ->(1), R, W, 2 nodes, 1
DRAM(50)01000000-00ffffff, ->(1), R, W, 2 nodes, 1
DRAM(58)01000000-00ffffff, ->(1), R, W, 2 nodes, 1
DRAM(60)00000000-00ffffff, ->(4), , , No interleave, 0
DRAM(68)00000000-00ffffff, ->(0), R, W, 8 nodes, 0
DRAM(70)00000000-00ffffff, ->(0), , , No interleave, 0
DRAM(78)00000000-00ffffff, ->(0), , , No interleave, 0
MMIO(80)01a00000-1100ffff, ->(0,2), , , CPU disable 0, Lock 0, Non
posted 0
MMIO(88)75060000-0000ffff, ->(2,0), , , CPU disable 0, Lock 0, Non
posted 0
MMIO(90)51040000-3f00ffff, ->(0,0), , , CPU disable 1, Lock 0, Non
posted 0
MMIO(98)00000000-0000ffff, ->(0,0), R, W, CPU disable 0, Lock 0, Non
posted 0
MMIO(a0)01c00000-1100ffff, ->(0,1), , , CPU disable 0, Lock 0, Non
posted 1
MMIO(a8)75000000-0000ffff, ->(2,0), , , CPU disable 0, Lock 0, Non
posted 0
MMIO(b0)51040000-0000ffff, ->(0,0), , , CPU disable 1, Lock 0, Non
posted 0
MMIO(b8)00000000-0000ffff, ->(0,0), , , CPU disable 0, Lock 0, Non
posted 0
PCIIO(c0)00001010-00003110, ->(0,1), , ,VGA 0 ISA 0
PCIIO(c8)00000750-00000000, ->(2,0), , ,VGA 0 ISA 1
PCIIO(d0)00002510-00000000, ->(0,0), , ,VGA 1 ISA 0
PCIIO(d8)00000000-00000000, ->(0,0), , ,VGA 0 ISA 0
CONFIG(e0)00000000-00000000 ->(0,0), CE 0
CONFIG(e4)00000000-00000000 ->(0,0), CE 0
CONFIG(e8)00000000-00000000 ->(0,0), CE 0
CONFIG(ec)00000000-00000000 ->(0,0), CE 0
Either the DRAM output is wrong or there is a real problem with our
DRAM programming.
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@941 f3766cd6-281f-0410-b1cd-43a5c92072e9
expansion ROMs are active.
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@940 f3766cd6-281f-0410-b1cd-43a5c92072e9
arch_write_tables(), before a number of routines that write/reserve
stuff are executed (in my test this only affects the 0x0-0x500 region
but I notice there's other stuff too).
Attached patch moves it down, solving the problem. Because stage1 can no
longer assume the MBI is at 0xf0000, I had to add a return path for stage2
to give it a pointer, using its exit status value.
Signed-off-by: Robert Millan <rmh@aybabtu.com>
Acked-by: Jordan Crouse <jordan.crouse@amd.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@936 f3766cd6-281f-0410-b1cd-43a5c92072e9
at any given moment. The code is generic enough to handle this in a
processor-specific way behind the scenes if needed.
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/coreboot-v3@933 f3766cd6-281f-0410-b1cd-43a5c92072e9
have a CAR area outside the usual RAM area. For those processors, we
have to implement a stack copying and switching mechanism. Since gcc
can't be told that the stack just moved, split stage1_main() into
stage1_phase1() and stage1_phase2() and stage1_phase3().
stage1_phase1() is the new entry point in stage1 and will handle
everything up to the point where we want to disable CAR.
Switching the stack, disabling CAR and handling other tasks related to
the stack switch (printk buffer move) is all wrapped in the
stage1_phase2() function.
stage1_phase2() calls disable_car() which then calls stage1_phase3().
stage1_phase3() is the former second half of stage1_main().
Notes about this patch:
- Code flow is almost unchanged for Qemu, K8 and Geode. No extensive new
testing required.
- We can support stack-keeping and stack-relocating architectures at the
same time, so C7 is definitely supportable
- The comment in stage1_phase2 says "some of this is not yet done". That
refers to the nonexisting code for stack switching on C7.
- "Minimal changes, maximum benefit".
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/coreboot-v3@932 f3766cd6-281f-0410-b1cd-43a5c92072e9
Build-tested with the AMD dbm690t board.
Signed-off-by: Uwe Hermann <uwe@hermann-uwe.de>
Acked-by: Marc Jones <marc.jones@amd.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@922 f3766cd6-281f-0410-b1cd-43a5c92072e9
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@896 f3766cd6-281f-0410-b1cd-43a5c92072e9
Signed-off-by: Ronald G. Minnich <rminnch@gmail.com>
Acked-by: Ronald G. Minnich <rminnch@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@880 f3766cd6-281f-0410-b1cd-43a5c92072e9
mainboard that we learned with the serengeti that we needed. New
function in pnp that is for reading. new prototype in pnp.h. New
constants for ite8716f.
This board does not build yet; we are exercising code in k8 north that
the serengeti did not enable. More tomorrow.
Now that we have two boards under way we can hopefully see our way to
getting more put in. The 690 is the obvious next choice.
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@876 f3766cd6-281f-0410-b1cd-43a5c92072e9
moved to stage1 ROM code. Make the struct for nodeid/coreid generic to
x86. Create the functions for existing architectures are a model for
future architectures (VIA coming soon we hope).
Move includes so that things build correctly now.
This is actually a small patch that impacted a number of files due to
include order changes. This is build and boot tested on simnow and
build tested on geode.
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Stefan Reinauer <stepan@coresystems.de>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@872 f3766cd6-281f-0410-b1cd-43a5c92072e9
AMP TinyGX still builds, this won't affect other platforms.
clean up 8111 stage1 code; add function to smbus,
memreset_setup_amd8111, for the 8111 specific parts of memreset.
include k8 .h to reduce warnings. Turn some things into functions (romcc
legacy cleanup) and put them in .c files.
simnow actually successfully gets through a reset cycle now.
Next is to fix the fidvid code.
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@868 f3766cd6-281f-0410-b1cd-43a5c92072e9
smbus.
Set up a global var variable called spd_inited. It is set when spd is
inited.
For simple cases, nothing is visible to initram main. For complex cases,
initram main can do the work and set this variable.
This compiles and runs on dbe62, which is actually meaningless since
dbe62
has not smbus, but hey ...
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@863 f3766cd6-281f-0410-b1cd-43a5c92072e9
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@853 f3766cd6-281f-0410-b1cd-43a5c92072e9
This code has been tested on dbe62, and builds for qemu as well.
the next step is testing on simnow.
k8.h: add more prototypes and some required inline functions.
cpu.h: same
serengeti: expand defines in mainboard.h, though we need a better
mechanism; continue to fix initram.c, add new support files to Makefile
lib/console.c: include globalvars.h
lib/lar.c: Provide more informative print as the lar is scanned.
k8 north: needed reset_test.c from v2, fixes to raminit.c
arch/x86
Kconfig: new CONFIG variable CBMEMK, meaning coreboot mem k, memory
used for coreboot.
init_cpus.c: functions to start up CPUs
stage1_mtrr.c: bring over early mtrr support from v2.
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@847 f3766cd6-281f-0410-b1cd-43a5c92072e9
what you get are warnings.
There is lots of room for improvement as we move to all CAR
code, but that will take time.
I hope to get this to really compile over the weekend.
At the same time, if anybody wants to take a crack at it, your efforts
are welcome.
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@843 f3766cd6-281f-0410-b1cd-43a5c92072e9
processor BIST is nonzero. Checking it in initram makes no sense. Having
it as global variable is unnecessary as well. Link BIST is an entirely
different animal.
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/coreboot-v3@834 f3766cd6-281f-0410-b1cd-43a5c92072e9
We need the sys_info struct in the global variables struct for
cache as ram on k8. The sys_info struct is generally very useful
so it makes sense to start accomodating it.
This patch adds an (empty for now) sys_info struct for geode.
It add sys_info to the global variables struct.
It removes global variables from console.h to a new file,
globalvars.h. Very little code needs to include this file.
This patch is tested on the dbe62 and simnow with no problems.
k8 compilation is now broken but I'm working on it. I'm going through
the eyeballs-bleed code on k8 startup to document it and with any luck
we'll have more functionality by the end of today. But it's hard ...
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@828 f3766cd6-281f-0410-b1cd-43a5c92072e9
current compilation unit to be marked as externally_visible. We have
EXPORT_SYMBOL exactly for that purpose.
This applies to the following symbols used by x86emu and/or vm86:
- pci_read_config8
- pci_read_config16
- pci_read_config32
- pci_write_config8
- pci_write_config16
- pci_write_config32
- dev_find_pci_device
- dev_find_slot
It also applies to the main entry point of stage2:
- stage2
With this patch, I can use -fwhole-program for stage2 without any
problems. For standard compilation, this is a noop.
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/coreboot-v3@827 f3766cd6-281f-0410-b1cd-43a5c92072e9
This means that basic resource maps are working, initial hypertransport
setup is working, the amd8111 ISA device is working, config space is
working for all the parts, we can grow the FLASH part address space to
more than 64k, and in general we're having a good time.
Here is the output:
coreboot-3.0.824 Tue Aug 26 22:18:21 PDT 2008 starting...
(console_loglevel=8)
Choosing fallback boot.
LAR: Attempting to open 'fallback/initram/segment0'.
LAR: Start 0xfff80000 len 0x80000
LAR: normal/option_table@0xfff80000, size 1776
LAR: normal/initram/segment0@0xfff80740, size 24404
LAR: normal/stage2/segment0@0xfff866f0, size 1
LAR: normal/stage2/segment1@0xfff86750, size 18542
LAR: normal/stage2/segment2@0xfff8b010, size 559
LAR: normal/payload/segment0@0xfff8b290, size 18142
LAR: bootblock@0xffff7fc0, size 32768
LAR: File not found!
LAR: Run file fallback/initram/segment0 failed: No such file.
Fallback failed. Try normal boot
LAR: Attempting to open 'normal/initram/segment0'.
LAR: Start 0xfff80000 len 0x80000
LAR: normal/option_table@0xfff80000, size 1776
LAR: normal/initram/segment0@0xfff80740, size 24404
LAR: CHECK normal/initram/segment0 @ 0xfff80740
start 0xfff80790 len 24404 reallen 24404 compression 0 entry 0x00000004
loadaddress 0x00000000
Entry point is 0xfff80794
Hi there from stage1
stage1 returns
run_file returns with 0
Goal for tomorrow is to get initram done.
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@826 f3766cd6-281f-0410-b1cd-43a5c92072e9
image, and fails:
LAR build/coreboot.rom
Bootblock coreboot.bootblock does not appear to be a bootblock.
Error adding the bootblock to the LAR.
make: *** [/home/rminnich/src/bios/coreboot-v3/build/coreboot.rom] Error
1
Next step is to get rid of all warnings that are not #warning.
Then it is on to simnow.
Anyone who wants to work on the warnings is most welcome to.
DBE62 still builds with no problems.
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@808 f3766cd6-281f-0410-b1cd-43a5c92072e9
Geode still builds fine.
include/lib.h includes a new function, cycles(), which is a u64 and
architecture-defined. (Thanks, Plan 9, for a sensible idea).
All rdtsc removed in favor of cycles()
All other changes are k8 specific. None of these changes adversely
impact existing platforms AFAICT.
Goal is that by 31/8/8, we're testing on simnow.
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@807 f3766cd6-281f-0410-b1cd-43a5c92072e9
Continue to upgrade northbridge for k8.
Add a new standard include (which is optional on some chipsets),
mainboard.h, which will define important mainboard constants that
1. do not belong in dts
2. do not belong in Kconfig
3. are so tightly tied down to the mainboard they should probably not be
visible, i.e. the value of the variable is defined by artwork on the
mainboard, such as the socket type.
This file resolves the long-standing question of where certain
mainboard-dependent, compile-time control variables belong.
We've not resolved this issue in two years so here's how
we're going to do it. The first use of this is in the definition of
CPU_SOCKET_TYPE, needed by the northbridge code.
These changes do not affect existing Geode builds (tested on DBE62).
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@792 f3766cd6-281f-0410-b1cd-43a5c92072e9
It's all REV_F now and we're wasting time pretending otherwise. This
change simplifies the code and will continue to simplify the code.
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@788 f3766cd6-281f-0410-b1cd-43a5c92072e9
existing code does not work due to the characteristics of stage1. This
has been broken since r729.
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/coreboot-v3@786 f3766cd6-281f-0410-b1cd-43a5c92072e9
code now unconditionally pushes an empty pointer to the stack which is a
placeholder for the pointer to global variable storage. That pointer and
the global variable storage are initialized in global_vars_init().
Creating global variables is now a piece of cake. You don't even have to
touch any asm code, just add them to struct global_vars.
Build tested on all targets, boot tested on Qemu.
NOTES:
- The code is not yet MP safe, but that's due to v3 not being MP safe in
general (and the comments contradict the code regarding MP features).
- K8 code now works by accident.
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Segher Boessenkool <segher@kernel.crashing.org>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@785 f3766cd6-281f-0410-b1cd-43a5c92072e9
Added a stepping enum to k8.h. This will allow us to do things like this:
if (cpu_stepping(node) < E0)
and so on instead of is_cpu_pre_e0_in_bsp or whatever it is.
Added and fixed Kconfig variables.
Broke out northbridge by function, so we can see what goes with what.
This tree still builds a working DBE62 coreboot that boots a kernel; no harm done to existing ports.
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@781 f3766cd6-281f-0410-b1cd-43a5c92072e9
not that readable anyway, so kill them and use standard definitions
instead.
Introduce EXPORT_SYMBOL for shared symbols. EXPORT_SYMBOL tells the
compiler to use the standard calling conventions for a given symbol and
not to optimize it away.
Benefits:
- We can later use gcc -combine -fwhole-program without problems.
- It's a correctness fix for some optimizations.
- We could check for duplicated exported functions at link time.
- We could check whether exported functions are linked into initram or
stage2 by accident.
- We could generate usage statistics and possibly optimize away unused
shared functions.
- Through the above points, significant side reductions of 10-40%
Build and boot tested on qemu.
Build tested on all targets.
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/coreboot-v3@780 f3766cd6-281f-0410-b1cd-43a5c92072e9
Calling non-PIC code from PIC code needs an ABI wrapper and we don't
provide one. Our trick with function pointers is exceedingly fragile:
- it depends on gcc not being clever enough
- it forces us to compile all initram source files in one go
- parallelizing initram compilation breaks the code
- compiling one initram source file at a time breaks the code
- enabling higher optimizations breaks the code
- enabling -fwhole-program breaks the code
- declaring the function pointers const breaks the code
- it's an undocumented side effect of gcc which will go away
- we need excessively ugly shared function wrappers
- the shared function wrappers had more than their fair share of bugs
- almost nobody understands the wrappers completely
- Segher warns against them: "So why do you think this should work?
You're telling it to link PIC to non-PIC. Did you read the manual? It's
just not allowed. It cannot ever work."
Kill the SHARED wrappers and use a real ABI wrapper.
The wrapper code is autogenerated on demand.
Any function compiled into stage0 is now shared by default, yet the size
and code generation of stage0/1/2 code are unchanged. Initram code size
does decrease quite a bit and the difficulty of creating shared
functions is now zero.
The patch includes extensive documentation about the inner workings of
the new wrappers and the reasons why they look like this.
Build and boot tested on qemu.
Build tested on all targets.
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Ron says:
Wow. we've need this fix for a long time.
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@775 f3766cd6-281f-0410-b1cd-43a5c92072e9
printk buffer management to it.
Build tested and boot tested and result tested on Qemu.
Adding a new global variable is not as easy as it looks, but the
comments in the code should be good enough to tell you how.
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/coreboot-v3@760 f3766cd6-281f-0410-b1cd-43a5c92072e9
all stages, blighting everything with the same code, compiled different
ways. In this change, we see that:
- basic conf ops are compiled into stage0, where they are used.
- they are called directly from initram
- they are used to initialize the pci_cf8_conf1 structure in stage 2,
but the call still goes to stage0!
one copy of the code.
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@752 f3766cd6-281f-0410-b1cd-43a5c92072e9
