Apply commit 283ba78415 to f12 (literally, plus one adaptation).
Change-Id: Ied7891806e269320caf968cae3de3dc792c5f8fd
Signed-off-by: Patrick Georgi <pgeorgi@google.com>
Reviewed-on: http://review.coreboot.org/7312
Tested-by: build bot (Jenkins)
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Bruce Griffith <Bruce.Griffith@se-eng.com>
The static library builder for the stub that interfaces to the
AGESA binary does not include config.h and kconfig.h, so any
header file changes that depend on Kconfig variables fail. Force
these two system headers to be included in the build of any AGESA
stub files.
Signed-off-by: Bruce Griffith <Bruce.Griffith@se-eng.com>
Change-Id: I2e8d38fa5aa21cc31b995ee3abe68ab3c3c55a68
Reviewed-on: http://review.coreboot.org/6979
Reviewed-by: Martin Roth <gaumless@gmail.com>
Tested-by: build bot (Jenkins)
Add all of the PI source that will remain part of coreboot to
build with a binary AGESA PI BLOB. This includes the gcc
makefiles, some Kconfig, and the AGESA standard library
functions.
Change vendorcode Makefile and Kconfig so that they can compile
AMD library files and use headers from outside the coreboot/src
tree.
The AGESA dispatcher is built using its own rules rather than
generic library generation rules in coreboot/Makefile and
coreboot/Makefile.inc. The AGESA source files are initially
copied from whereever they live into coreboot/build/agesa.
They are compiled from there. The binary PI directory has a
mandatory structure that places the AGESA BLOB into the same
directory as the support headers. These will nominally be
placed in the 3rdparty directory in coreboot.org.
The copy commands that were added to the the vendorcode
Makefile.inc ensure that only one thread will operate on each
source file by using a macro to generate the copy targets.
After the change, each copy target will operate on exactly one
source file.
Due to API issues, coreboot has no way to control the IMC to set up
fan control. Set a Kconfig flag that removes the ability to install
an IMC BLOB into CBFS.
Change-Id: I050b72a19086aaeba6cb65ce165297b10e3cfc45
Signed-off-by: Bruce Griffith <Bruce.Griffith@se-eng.com>
Reviewed-on: http://review.coreboot.org/6595
Tested-by: build bot (Jenkins)
Reviewed-by: WANG Siyuan <wangsiyuanbuaa@gmail.com>
Reviewed-by: Zheng Bao <zheng.bao@amd.com>
We already have these macros define in 'stdlib.h'. Make good use of them
here to avoid redefinition conflicts of the pre-processor depending on
header inclusion ordering. This has the nice side-effect of syncing up
AGESA families in this particular regard.
Change-Id: Icf911629a4a1a82b01062fe16af4c8f812b05717
Signed-off-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-on: http://review.coreboot.org/6199
Tested-by: build bot (Jenkins)
Reviewed-by: Bruce Griffith <Bruce.Griffith@se-eng.com>
Implement the fix for the erratum #712. - Processor May Hang During Graphics Memory Controller
Sequencing
The processor may hang during a graphics memory controller (GMC) sleep state transitioning. The failure may
be processor specific and may be sensitive to temperature.
Potential Effect on System:
System hang.
Suggested Workaround:
BIOS should set D18F2x408_dct[1:0] bit 31 = 1b.
See Publication # 48931 Revision: 3.08
Change-Id: I4346fd4ef3cf554ffdaaad5ab6fc84e73532e885
Signed-off-by: Rudolf Marek <r.marek@assembler.cz>
Reviewed-on: http://review.coreboot.org/6216
Tested-by: build bot (Jenkins)
Reviewed-by: Martin Roth <gaumless@gmail.com>
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
The 'm' (a memory reference) constraint makes little sense here since we
are talking about a fs relative read, rather 'ir' (immediate or
register) constraint is more sensible.
N.B. The 'p' constraint allows anything which fits the form of an address
calculation where the 'ir' constraint is just a register /xor/
immediate. Hence would produce better code here however, unfortunately,
clang does not currently support it properly.
The %b and %w constraints are also redundant and only hide errors.
The functions writefsword() and writefsdword() should use ir instead of
iq. iq is unnecessarily restrictive (it is only required for writing
bytes).
The cld in stosb is redundant (and the constraints are unnecessarily
complicated). Note that The ABI guarantees that the direction flag is
cleared. i.e. eax, ecx, edx are caller-saved, returned value in eax,
eax+edx, st0, yaddayadda, direction flag cleared. In fact bad things can
happen if you set it in some asm and do not clear it until the end of
the asm.
Line wrap these extraneously long lines found with these particular functions.
Many thanks to Christoph Mallon <christoph.mallon@gmx.de> from #llvm for
helping me with this.
Change-Id: Iaf3ad65791640e1060a2029e7ebb043f57b338a9
Signed-off-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-on: http://review.coreboot.org/5758
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Reviewed-by: Marc Jones <marc.jones@se-eng.com>
The semantics of this loop relies on an integer overflow in Index >=0
that implies a return value of (UINT8)-1 which around wraps to 0xFF, or
VOLT_UNSUPPORTED.
Change-Id: I44d68973d0a80093350b2a8a4d3b46bfbb57917a
Signed-off-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-on: http://review.coreboot.org/5801
Tested-by: build bot (Jenkins)
Reviewed-by: Marc Jones <marc.jones@se-eng.com>
The typedef'ed BIT_FIELD_NAME enum is type unsigned. The parameter
'FieldName' is decleared with type BIT_FIELD_NAME and thus the redudant
comparison of unsigned enum expression >= 0 is always true.
BIT_FIELD_NAME is declared in vendorcode/amd/agesa/f14/Proc/Mem/mm.h
Change-Id: Id2f03596c44b68e861e939f3528256d4b08c45ce
Signed-off-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-on: http://review.coreboot.org/5757
Tested-by: build bot (Jenkins)
Reviewed-by: Marc Jones <marc.jones@se-eng.com>
The conditional comparison in the for-loop construct with the constant
300000 has an index incrementor of type 'UINT16' (aka 'unsigned short')
which is always true.
Change-Id: I932c168742163be4038728fb40833231a447fa78
Signed-off-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-on: http://review.coreboot.org/5799
Tested-by: build bot (Jenkins)
Reviewed-by: Marc Jones <marc.jones@se-eng.com>
The function GfxPowerPlayLocateTdp() sets MinDeltaSclk to a maximum
sentinel value and checks DeltaSclk in a loop to minimize MinDeltaSclk.
However, MinDeltaSclk incorrectly self-assigns.
Change-Id: Id01c792057681516bba411adec268769a3549aa8
Signed-off-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-on: http://review.coreboot.org/5752
Tested-by: build bot (Jenkins)
Reviewed-by: Dave Frodin <dave.frodin@se-eng.com>
Change-Id: I31632948ce69b2d1ff63b6c920016ed6fdf9e2f8
Signed-off-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-on: http://review.coreboot.org/5760
Tested-by: build bot (Jenkins)
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Marc Jones <marc.jones@se-eng.com>
Change-Id: I1d801b9d8387e267feeb95563e55910b30ebbc34
Signed-off-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-on: http://review.coreboot.org/5790
Tested-by: build bot (Jenkins)
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Marc Jones <marc.jones@se-eng.com>
They don't exactly add clarity, but increase the risk
they're used at some obscure place.
Change-Id: Ic74f72dae3f9b7eb2343cb5c51bc44c888e1276c
Signed-off-by: Patrick Georgi <patrick@georgi-clan.de>
Reviewed-on: http://review.coreboot.org/5787
Tested-by: build bot (Jenkins)
Reviewed-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
They're _not_ part of the compiler binary, so they have
no place in $(CC_*)
Change-Id: I1e1c3c0be6f75629450a824ea834e1614d48ed9b
Signed-off-by: Patrick Georgi <patrick@georgi-clan.de>
Reviewed-on: http://review.coreboot.org/5785
Tested-by: build bot (Jenkins)
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
With the upcoming CC/CFLAGS/CPPFLAGS split,
romcc gets more CPPFLAGS, and it's picky about
directories actually existing.
Change-Id: Ib9c525296e5be0c8ace935ab8096bc98206cbcc1
Signed-off-by: Patrick Georgi <patrick@georgi-clan.de>
Reviewed-on: http://review.coreboot.org/5784
Tested-by: build bot (Jenkins)
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Make all three coreboot stages (bootblock, romstage and ramstage) aware of the
architecture specific to that stage i.e. we will have CONFIG_ARCH variables for
each of the three stages. This allows us to have an SOC with any combination of
architectures and thus every stage can be made to run on a completely different
architecture independent of others. Thus, bootblock can have an x86 arch whereas
romstage and ramstage can have arm32 and arm64 arch respectively. These stage
specific CONFIG_ARCH_ variables enable us to select the proper set of toolchain
and compiler flags for every stage.
These options can be considered as either arch or modes eg: x86 running in
different modes or ARM having different arch types (v4, v7, v8). We have got rid
of the original CONFIG_ARCH option completely as every stage can have any
architecture of its own. Thus, almost all the components of coreboot are
identified as being part of one of the three stages (bootblock, romstage or
ramstage). The components which cannot be classified as such e.g. smm, rmodules
can have their own compiler toolset which is for now set to *_i386. Hence, all
special classes are treated in a similar way and the compiler toolset is defined
using create_class_compiler defined in Makefile.
In order to meet these requirements, changes have been made to CC, LD, OBJCOPY
and family to add CC_bootblock, CC_romstage, CC_ramstage and similarly others.
Additionally, CC_x86_32 and CC_armv7 handle all the special classes. All the
toolsets are defined using create_class_compiler.
Few additional macros have been introduced to identify the class to be used at
various points, e.g.: CC_$(class) derives the $(class) part from the name of
the stage being compiled.
We have also got rid of COREBOOT_COMPILER, COREBOOT_ASSEMBLER and COREBOOT_LINKER
as they do not make any sense for coreboot as a whole. All these attributes are
associated with each of the stages.
Change-Id: I923f3d4fb097d21071030b104c372cc138c68c7b
Signed-off-by: Furquan Shaikh <furquan@google.com>
Reviewed-on: http://review.coreboot.org/5577
Tested-by: build bot (Jenkins)
Reviewed-by: Aaron Durbin <adurbin@gmail.com>
Following the same reasoning as commit
ee905a8 vendorcode/amd/agesa/fam15tn: Build as a static library
Since AGESA is stage-independent, we can build it just once, and use
the resulting static library in both rom and ram stages.
Change-Id: I8b78c462f4963fbb3a40d739196529fffedccb4c
Signed-off-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-on: http://review.coreboot.org/5441
Tested-by: build bot (Jenkins)
Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
Up until now, we were building AGESA by specifying each AGESA source
file and adding it to the list of romstage and ramstage source files.
As a result, we were compiling each AGESA source twice, despite the
fact that it does not depend on the stage we're in.
Since AGESA is stage-independent, we can build it just once, and use
the resulting static library in both rom and ram stages.
We still keep the practice of specifying every single AGESA directory
as an include dir and adding the AGESA CFLAGS to our global CFLAGS;
this is needed due to the way AGESA builds.
Change-Id: I9b23264129d1c08cb67cabc31d15a68d43ed7624
Signed-off-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
Reviewed-on: http://review.coreboot.org/5430
Tested-by: build bot (Jenkins)
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Idwer Vollering <vidwer@gmail.com>
The ROM address range is set up in the LPC PCI device, register 0x6c.
Coreboot already sets that up correctly in the bootblock, however
AGESA overrides that to 0xffffff00, which will always map the ROM from
0xff000000. This may conflict with other devices which are assigned
address space in that range.
If a device is assigned a range between 0xff000000 and the real ROM
base, accesses to that device will be diverted to the system ROM,
regardless of how other BARs are set up. Since we already need to set
up the ROM address range in the bootblock, before calling AGESA, just
remove the override from AGESA.
Note that not all AGESA versions override this mapping.
Change-Id: I592e5d087ed830c9604a04a356912c7654ce56d2
Signed-off-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
Reviewed-on: http://review.coreboot.org/5467
Tested-by: build bot (Jenkins)
Reviewed-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-by: Aaron Durbin <adurbin@google.com>
A comparison with a two's complement in gcccar.inc has dubious
GAS/AT&T notation. Clang miss-parses 0x-1 as an invalid hexadecimal
number.
Change-Id: I88baa5c2513f062ff309df05916a3832b9bd9bb1
Signed-off-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-on: http://review.coreboot.org/5277
Tested-by: build bot (Jenkins)
Reviewed-by: Ronald G. Minnich <rminnich@gmail.com>
It is inappropriate for chipset code to be redefining
types -- especially NULL to a non-pointer type. There's
only one non-straight forward change. A condition
being checked was '!ptr_type == NULL' (0 as int). That
check is actually 'ptr_type != NULL'.
Change-Id: Iab5733e5a573baba6fec94e0c955ba4fad72c836
Signed-off-by: Aaron Durbin <adurbin@chromium.org>
Reviewed-on: http://review.coreboot.org/5088
Tested-by: build bot (Jenkins)
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Patrick Georgi <patrick@georgi-clan.de>
The AGESA resumes the GPP ports in the romstage using FchInitResetGpp(),
which does FchGppPortInitS3Phase() for S3 resume. The PreInitGppLink()
looks into CMOS to figure out what ports to just force to Gen1 or
Gen2 PCIe. Then boot continues and in the ramstage the rest of GPP
init is executed. There is a problem that nobody sets properly the
PortDetected flags in the S3 path. As the consequence FchGppDynamicPowerSaving()
thinks the GPP port is not enabled and shut downs it.
The best fix would be also to remove the CMOS dependency which
might be some left over, because AGESA does not use CMOS much for
anything else. There could be also some way how to pass the GPP state
structure from romstage to ramstage possibly via hudson/resume.c
but I don't know how to do that. Similar problem is that the "late"
stage of init again "forgets" the PortDetected state.
This fix fixes the resume issue on Asus F2A85-M. With this patch applied
both GPP ports (used as PCIe x1 and internal ethernet) are working again
after resume.
Change-Id: Idaf609043abb09441c6790504d66d23e0637588f
Signed-off-by: Rudolf Marek <r.marek@assembler.cz>
Reviewed-on: http://review.coreboot.org/4671
Tested-by: build bot (Jenkins)
Reviewed-by: Idwer Vollering <vidwer@gmail.com>
Reviewed-by: Ronald G. Minnich <rminnich@gmail.com>
The bug is hard to find. We were adding the feature of fan control. We
met some strange things which could not be explained. Like, sometimes
adding printk let the error disappear. Then we traced the code by hardware
debug tool (HDT). It turned out the data in stack was overwritten.
The values of AccessWidthxx are
{ AccessWidth8 = 1,
AccessWidth16,
AccessWidth32,}
For the case of AccessWidth8, we only need to access the index/data
once. But ReadECmsg and WriteECmsg did the loop twice, 1 more time
than they are supposed to do. The data in stack next to "Value" would
be overwritten.
For all the cases, the code should be
OpFlag = OpFlag & 0x7f;
switch (OpFlag) {
case 1: /* AccessWidth8 */
OpFlag = 0;break;
case 2: /* AccessWidth16 */
OpFlag = 1;break;
case 3: /* AccessWidth32 */
OpFlag = 3;break;
case 4: /* AccessWidth64 */
OpFlag = 7;break;
default:
error;
}
Actually, the caller only takes AccessWidth8 as the parameter. We can ignore other
cases for now.
That is an AGESA bug. AMD's AGESA team own this code. They have given the
response that they are going to update this in next release. I presume let them
decide the proper way to fix that. Before that, I change the code as little
as possible to make it run without crash.
Change-Id: I566f74c242ce93f4569eedf69ca07d2fb7fb368d
Signed-off-by: Zheng Bao <zheng.bao@amd.com>
Signed-off-by: Zheng Bao <fishbaozi@gmail.com>
Reviewed-on: http://review.coreboot.org/4297
Reviewed-by: Bruce Griffith <Bruce.Griffith@se-eng.com>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Tested-by: build bot (Jenkins)
The platform initialization (PI) code v1.0.0.7 for Kabini has some
enhancements like ECC DIMM support, new CPU microcode rev 0700010B, FCH
bug fix (RTC) and so on.
Use the name Kabini instead of Kerala everywhere.
Note, the former PI code was indeed version v1.0.0.0 instead of v0.0.1.0
as used in `AGESA_VERSION_STRING`.
Change-Id: I186de1aef222cd35ea69efa93967a3ffb8da7248
Signed-off-by: WANG Siyuan <SiYuan.Wang@amd.com>
Signed-off-by: WANG Siyuan <wangsiyuanbuaa@gmail.com>
Reviewed-on: http://review.coreboot.org/3935
Tested-by: build bot (Jenkins)
Reviewed-by: Bruce Griffith <Bruce.Griffith@se-eng.com>
Northbridge code includes these headers, so they all need to
have the same name to allow different combinations of northbridge
and southbridge. This changes the sb900 names to match sb700 &
sb800, and points agesa/family12 and amd/torpedo to the new file
names.
Change-Id: I7a654ce9ae591a636a56177f64fb8cb953b4b04f
Signed-off-by: Corey Osgood <corey.osgood@gmail.com>
Reviewed-on: http://review.coreboot.org/3825
Tested-by: build bot (Jenkins)
Reviewed-by: Patrick Georgi <patrick@georgi-clan.de>
src/southbridge/amd/agesa/hudson/Kconfig config default value,
mainboard Kconfig config value for specific mainboard.
bit 1,0 - pin 0
bit 3,2 - pin 1
bit 5,4 - pin 2
bit 7,6 - pin 3
Change-Id: I54a87cf734685515a3e1850838ca7d94387172ce
Signed-off-by: WANG Siyuan <SiYuan.Wang@amd.com>
Signed-off-by: WANG Siyuan <wangsiyuanbuaa@gmail.com>
Reviewed-on: http://review.coreboot.org/3879
Tested-by: build bot (Jenkins)
Reviewed-by: Bruce Griffith <Bruce.Griffith@se-eng.com>
Reviewed-by: Dave Frodin <dave.frodin@se-eng.com>
Copied from a similar commit for Family 10h AGESA [1]
Remove the fourth argument in the comments. Luckily the compiler,
at least gcc, warns about a wrong number of arguments, and therefore
no incorrect code resulted from the wrong documentation.
[1] 07e0f1b AMD AGESA: Fix argument list for `PCIE_DDI_DATA_INITIALIZER` in comments
[2] fc47bfa Revert "AMD f14 vendorcode: Fix warning"
Change-Id: I3806e368a823e4a40d22e99b91bf3598d9ed2f15
Signed-off-by: Bruce Griffith <bruce.griffith@se-eng.com>
Reviewed-on: http://review.coreboot.org/3840
Tested-by: build bot (Jenkins)
Reviewed-by: Martin Roth <martin@se-eng.com>
This is the same patch as an earlier one applied to family 15 [1].
Static analysis often flags case statements that do not include
a terminating "break;" statement. Eclipse's CODAN is an example
of this. This changelist modifies amdlib.c to terminate
case statements with "break;".
[1] e44a89f amd/agesa/f15/Lib/amdlib.c: Add missing breaks ...
Change-Id: Ibd1ae6f2b52fde07de3d978d174975f4d93647ab
Signed-off-by: Bruce Griffith <Bruce.Griffith@se-eng.com>
Reviewed-on: http://review.coreboot.org/3839
Tested-by: build bot (Jenkins)
Reviewed-by: Martin Roth <martin@se-eng.com>
Add CONST modifiers to read-only pass-by-reference function
parameters in AGESA. This allows the use of "const" modifiers
on the declaration of lookup tables that are pass-by-reference.
These will be used to identify tables that are copied onto the
HEAP but don't need to be.
This same change was made for AMD Trinity APUs (Family15tn) [1].
[1] 283ba78 AGESA: Add "const" modifier to function parameters
Change-Id: I2bdd9fc5e027e938de9df0f923b95da934bb48dc
Signed-off-by: Bruce Griffith <Bruce.Griffith@se-eng.com>
Reviewed-by: Dave Frodin <dave.frodin@se-eng.com>
Tested-by: Bruce Griffith <bruce.griffith@se-eng.com>
Reviewed-on: http://review.coreboot.org/3837
Tested-by: build bot (Jenkins)
Reviewed-by: Marc Jones <marc.jones@se-eng.com>
This changelist was cherry-picked from merged community code
for Parmer [1] and the paths modified so that the Parmer
modification is applied against Olive Hill.
[1] 0086162 AMD SATA: Correct _them implement_ ... in comments
Change-Id: I9849e9a75dacfde15331c4200d72343a59036f14
Signed-off-by: Bruce Griffith <bruce.griffith@se-eng.com>
Reviewed-on: http://review.coreboot.org/3841
Tested-by: build bot (Jenkins)
Reviewed-by: Ronald G. Minnich <rminnich@gmail.com>
Static analysis often flags case statements that do not include
a terminating "break;" statement. Eclipse's CODAN is an example
of this. This changelist modifies amdlib.c to terminate
case statements with "break;".
Change-Id: I3d43acaf64e2e2d9717421cb547fec35e582cf8b
Signed-off-by: Bruce Griffith <Bruce.Griffith@se-eng.com>
Reviewed-on: http://review.coreboot.org/3539
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
