Per AMD, the Integrated Micro Controller is not a supported feature of
the Stoney Ridge APU. Systems are expected to implement an external EC
for desired features. Remove all stoney IMC files and functions from
src/soc/amd/stoneyridge.
There are 2 "IMC bits" left (and used) that are not truly IMC. New BKDG
describe these bits, so a new patch will be released later to fix the
names and comment.
BUG=b:111780177
TEST=Build grunt and gardenia
Change-Id: I6a24e4c3f03d04713a030b884c611d9c64c4cb3a
Signed-off-by: Richard Spiegel <richard.spiegel@silverbackltd.com>
Reviewed-on: https://review.coreboot.org/27651
Reviewed-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
The careena board needs different video settings to pass eye diagram test,
which does not affect negatively the grunt board. In preparation for new
VBIOS, create code that allows changing eDP training parameter.
BUG=b:111673328
TEST=Tested in child patch.
Change-Id: Ic0452618bfc5e05b9ef8280bb8ba398ec7b4ce95
Signed-off-by: Richard Spiegel <richard.spiegel@silverbackltd.com>
Reviewed-on: https://review.coreboot.org/27625
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martinroth@google.com>
Old return value was not used, and function body
has die() in case of errors in allocation.
Change-Id: I89b0e9c927d395ac6d27201e0b3a8658e9585187
Signed-off-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-on: https://review.coreboot.org/27261
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Garrett Kirkendall <garrett.kirkendall@amd.corp-partner.google.com>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
When AllocationMethod == ByHost, buffer has to be
provided by caller.
Improve code symmetry, the named parameter is now
always pointer to the struct.
Change-Id: I2085f7d5d63ef96f4bd9d5194af099634c402820
Signed-off-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-on: https://review.coreboot.org/27112
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Garrett Kirkendall <garrett.kirkendall@amd.corp-partner.google.com>
Reviewed-by: Marshall Dawson <marshalldawson3rd@gmail.com>
The stoneyridge bootblock no longer makes AGESA calls. Remove the
support files from the bootblock build.
TEST=boot Grunt
Change-Id: I14d2336d5fb766a1acf5e812337ae0ab3ca4a6c1
Signed-off-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Reviewed-on: https://review.coreboot.org/27255
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-by: Martin Roth <martinroth@google.com>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Remove residual code that allowed successful building of the heapmanager
code. Now that stoneyridge no longer makes AGESA calls in bootblock, it
is safe to elimate the workaround.
BUG=b:74518368
TEST=boot Grunt
Change-Id: Ie169a691a177bcd8283c31c8188ce28bcbce82af
Signed-off-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Reviewed-on: https://review.coreboot.org/27254
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-by: Martin Roth <martinroth@google.com>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
The entire StdHeader field is really supposed to
be forked from a template for each entry into the
AGESA API. Current code assumes only Callout would
be relevant, which is not quite the case.
Change-Id: I0cc66d01d62fa8dc6bb7c9f9fab6fa4753827554
Signed-off-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-on: https://review.coreboot.org/27111
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Garrett Kirkendall <garrett.kirkendall@amd.corp-partner.google.com>
Reviewed-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Implement an optional callout for AgesaHeapRebase which allows AGESA
to override any internal hardcoded heap addresses.
Designate a region in CAR that may be used for pre-mem heap and return
that address before DRAM is configured. After DRAM is up, the address
in cbmem is returned.
TEST=Boot grunt with patchstack and experimental blob
BUG=b:74518368
Change-Id: Ieda202a6064302b21707bd7ddfabc132cd85ed45
Signed-off-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Reviewed-on: https://review.coreboot.org/25458
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martinroth@google.com>
Implement a new AGESA callout that may be used to find the correct
temporary location in DRAM to store heap data.
Near the end of AmdInitPost, AGESA migrates its heap from a CAR-based
location to a temporary region. Once cbmem has been established, the
heap will be relocated again in AmdInitEnv from the temp location to
the final one.
This patch does not materially affect the behavior of AGESA's heap
management. It only puts coreboot in control of the location. Future
work may refactor the copying.
TEST=Boot grunt with patchstack and experimental blob
BUG=b:74518368
Change-Id: Ibc5cc988e3e80d78f50cf0195e952b657141e570
Signed-off-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Reviewed-on: https://review.coreboot.org/26146
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martinroth@google.com>
This won't actually get called yet since the GPIO pin has not been
configured as SMI.
BUG=b:80295434
TEST=grunt: Made sure events could be processed.
Change-Id: I189e26196e4543b3e34bff5d9df8566eff07d585
Signed-off-by: Raul E Rangel <rrangel@chromium.org>
Reviewed-on: https://review.coreboot.org/26546
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Use of device_t has been abandoned in ramstage.
Change-Id: I82089475eb43d58303d1091f35aee06f1f04b4a4
Signed-off-by: Elyes HAOUAS <ehaouas@noos.fr>
Reviewed-on: https://review.coreboot.org/26459
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Marc Jones <marc@marcjonesconsulting.com>
This patch ensures that user can select a specific AP to run
a function.
BUG=b:74436746
BRANCH=none
TEST=Able to run functions over APs with argument.
Change-Id: Iff2f34900ce2a96ef6ff0779b651f25ebfc739ad
Signed-off-by: Subrata Banik <subrata.banik@intel.com>
Reviewed-on: https://review.coreboot.org/26034
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
This patch ensures that user can pass a function with given argument
list to execute over APs.
BUG=b:74436746
BRANCH=none
TEST=Able to run functions over APs with argument.
Change-Id: I668b36752f6b21cb99cd1416c385d53e96117213
Signed-off-by: Subrata Banik <subrata.banik@intel.com>
Reviewed-on: https://review.coreboot.org/25725
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Nico Huber <nico.h@gmx.de>
It was decided to not add the buffers definitions, so the todo message
is obsolete. Replace it with minimum instructions about when a new buffer
will be needed.
It was also noticed a typo in one command. MBOX_BIOS_CMD_C3_DATA_INFO is
about S3 transition, so it should be called MBOX_BIOS_CMD_S3_DATA_INFO.
BUG=b:77940747
TEST=None.
Change-Id: I6143d7e85476061395962b95ad8864ac32a1d4a3
Signed-off-by: Richard Spiegel <richard.spiegel@silverbackltd.com>
Reviewed-on: https://review.coreboot.org/25740
Reviewed-by: Martin Roth <martinroth@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Found-by: Coverity (CID 1387031: Memory - illegal accesses
(BUFFER_SIZE_WARNING)). Calling strncpy with a maximum size argument of
19 bytes on destination array "dimm->module_part_number" of size 19 bytes
might leave the destination string unterminated. Fix the size parameter.
BUG=b:76202696
TEST=Build and boot kahlee, using special debug code to see the output
strings, which was later removed.
Change-Id: I18fa5e9c73401575441b6810f1db80d11666368c
Signed-off-by: Richard Spiegel <richard.spiegel@silverbackltd.com>
Reviewed-on: https://review.coreboot.org/25419
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martinroth@google.com>
There are sections of code that are almost identical and they can be
converted to auxiliary procedures. For allocating heap, 3 sizes (the
buffer size of currently being examined node, the buffer size of the
current best fit node and the minimum size for a buffer that will need
to be split if selected as the best fit) are used often so they could
be stored in temporary variables. These 2 changes will make code shorter,
with less indentation problems and overall easier to read. The actual
logic of the code is not changed.
BUG=b:77940747
TEST=Build and boot grunt.
Change-Id: Ib4c69981eab7452228ccae9ed9bc288c8baceffe
Signed-off-by: Richard Spiegel <richard.spiegel@silverbackltd.com>
Reviewed-on: https://review.coreboot.org/25703
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
dimm_info.serial had a strange contract. The SPD spec defines a 4 byte
serial number. dimm_info.serial required a 4 character ascii string with
a null terminator.
This change makes the serial field so it matches the SPD spec.
smbios.c will then translate the byte array into hex and set it on the
smbios table.
There were only two callers that set the serial number:
* haswell/raminit.c: already does a memcpy(serial, spd->serial, 4), so
it already matches the new contract.
* amd_late_init.c: Previously copied the last 4 characters. Requires
decoding the serial number into a byte array.
google/cyan/spd/spd.c: This could be updated to pass the serial number,
but it uses a hard coded spd.bin.
Testing this on grunt, dmidecode now shows the full serial number:
Serial Number: 00000000
BUG=b:65403853
TEST=tested on grunt
Change-Id: Ifc58ad9ea4cdd2abe06a170a39b1f32680e7b299
Signed-off-by: Raul E Rangel <rrangel@chromium.org>
Reviewed-on: https://review.coreboot.org/25343
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
The converter was setting SMBIOS values when dimm_info expects SPD
values.
dmidecode now shows the following:
Memory Device
Array Handle: 0x0000
Error Information Handle: Not Provided
Total Width: 64 bits
Data Width: 64 bits
Size: 8192 MB
Form Factor: SODIMM
Set: None
Locator: Channel-0-DIMM-0
Bank Locator: BANK 0
Type: DDR4
Type Detail: Synchronous
Speed: 933 MT/s
Manufacturer: Hynix/Hyundai
Serial Number: 00000000
Asset Tag: Not Specified
Part Number: HMAA51S6AMR6N-UH
Rank: 1
Configured Clock Speed: 933 MT/s
Minimum Voltage: Unknown
Maximum Voltage: Unknown
Configured Voltage: Unknown
Example debug output:
AGESA TYPE 17 DMI INFO:
Handle: 1
TotalWidth: 64
DataWidth: 64
MemorySize: 8192
DeviceSet: 0
Speed: 1200
ManufacturerIdCode: 44416
Attributes: 1
ExtSize: 0
ConfigSpeed: 933
MemoryType: 0x1a
FormFactor: 0xd
DeviceLocator: DIMM 0
BankLocator: CHANNEL A
SerialNumber(8): ' 00000000'
PartNumber(20): 'HMAA51S6AMR6N-UH '
CBMEM_ID_MEMINFO:
dimm_size: 8192
ddr_type: 0x1a
ddr_frequency: 933
rank_per_dimm: 1
channel_num: 0
dimm_num: 0
bank_locator: 0
mod_id: 44416
mod_type: 0x4
bus_width: 3
serial: 0x00000000
module_part_number(18): 'HMAA51S6AMR6N-UH '
The serial number we get from AGESA (at least on my
board) is always 00000000. I'm assuming this is because the SPD info is
compiled in.
`mosys memory spd print all` is still failing though. I will look into
that next.
BUG=b:65403853
BRANCH=dimm-info
TEST=Test output above
Change-Id: I076bc3f965f81a9374c8976da48c7fdce014dc0c
Signed-off-by: Raul E Rangel <rrangel@chromium.org>
Reviewed-on: https://review.coreboot.org/25304
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
As part of moving AGESA calls from bootblock to romstage, callback function
AGESA_HALT_THIS_AP must be available at romstage.
BUG=b:74236170
TEST=Build and boot grunt, actual test will be performed at a later patch.
Change-Id: I0992b2de5856881c19191ec4f637168727686524
Signed-off-by: Richard Spiegel <richard.spiegel@silverbackltd.com>
Reviewed-on: https://review.coreboot.org/25527
Reviewed-by: Martin Roth <martinroth@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Example output:
AGESA TYPE 17 DMI INFO:
Handle: 1
TotalWidth: 64
DataWidth: 64
MemorySize: 8192
DeviceSet: 0
Speed: 1200
ManufacturerIdCode: 44416
Attributes: 1
ExtSize: 0
ConfigSpeed: 933
MemoryType: 0x1a
FormFactor: 0xd
DeviceLocator: DIMM 0
BankLocator: CHANNEL A
SerialNumber(8): 00000000
PartNumber(20): HMAA51S6AMR6N-UH
CBMEM_ID_MEMINFO:
dimm_size: 0
ddr_type: 0x1a
ddr_frequency: 1200
rank_per_dimm: 1
channel_num: 0
dimm_num: 0
bank_locator: 0
mod_id: 44416
mod_type: 0x1a
bus_width: 64
serial(4): 0000
module_part_number(23): HMAA51S6AMR6N-UH ��@
dimm_size, mod_type, bus_width need to be updated so they return the
correct values. module_part_number is missing a null terminator due to
the AGESA part number being larger than the dimm_info buffer.
Example dmidecode output:
Memory Device
Array Handle: 0x0000
Error Information Handle: Not Provided
Total Width: 8 bits
Data Width: 8 bits
Size: No Module Installed
Form Factor: Unknown
Set: None
Locator: Channel-0-DIMM-0
Bank Locator: BANK 0
Type: DDR4
Type Detail: Synchronous
Speed: 1200 MT/s
Manufacturer: Hynix/Hyundai
Serial Number: 0000
Asset Tag: Not Specified
Part Number: HMAA51S6AMR6N-UH
Rank: 1
Configured Clock Speed: 1200 MT/s
Minimum Voltage: Unknown
Maximum Voltage: Unknown
Configured Voltage: Unknown
To enable the output set CONFIG_DEBUG_RAM_SETUP.
The Kconfig change is required in order to enable
CONFIG_DEBUG_RAM_SETUP, otherwise it's not a valid option.
BUG=b:65403853
TEST=Test output shown above
Signed-off-by: Raul E Rangel <rrangel@chromium.org>
Change-Id: I5eac00b9400056357915761287770a400b3f9f8b
Reviewed-on: https://review.coreboot.org/25303
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Richard Spiegel <richard.spiegel@silverbackltd.com>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
It was required for all cores use the same CAR teardown function
(exit_car.S and gcccar.inc). AGESA has already been modified to do the
AP to do the call out. Create assembly code to call chipset_teardown_car
and then enter an endless loop with halt instruction. Then create the
call out that will call this new assembly code.
BUG=b:70338633
AGESA COMMIT=3313d277
TEST=Created a debug version of AGESA that would print the returned
status of HALT_THIS_AP. Build code without the fix, see the return.
Build code with the fix, see that there's no return.
Change-Id: I05ee405812211d93dfdbdc5ee7d9978c2eb585e1
Signed-off-by: Richard Spiegel <richard.spiegel@silverbackltd.com>
Reviewed-on: https://review.coreboot.org/24999
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martinroth@google.com>
Save the UMA base and size settings returned by AGESA
in amdinitpost();
Change-Id: Id96cc65582118ad41d397b1a600cab1615676a55
Signed-off-by: Marc Jones <marcj303@gmail.com>
Reviewed-on: https://review.coreboot.org/23818
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
Reviewed-by: Martin Roth <martinroth@google.com>
Kahlee DIMM have invalid string when it comes to part number
(bytes 0x149-0x15c). We currently force a NA string, but grunt has the
proper strings. Just let the string go through, and a second commit
within smbios.c will be responsible for testing the string and taking
proper action.
BUG=b:73122207
TEST=Build, boot and record serial output for kahlee while injecting
different strings to dmi17->PartNumber. Remove string injection before
committing.
Change-Id: I427262873f9ec80f459245e5f509e28a68de3074
Signed-off-by: Richard Spiegel <richard.spiegel@silverbackltd.com>
Reviewed-on: https://review.coreboot.org/23825
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martinroth@google.com>
SMBIOS structure type 17 is not being generated because memory info is
not being stored to cbmem. This has to happen after AGESA AmdInitLate
has run, but before SMBIOS is generated. There's a need to convert
format between AGESA generated info, and what is required in cbmem.
Create a procedure that transfers information between AGESA and cbmem,
and call it from agesawrapper_post_device() after AmdLateInit is called.
BUG=b:65403853
TEST=build and run kahlee, verify if SMBIOS structure type 17 is being
generated, and if associated strings are what should be expected.
Change-Id: I151a8f1348c9bafceb38bab1f79d3002c5f6b31b
Signed-off-by: Richard Spiegel <richard.spiegel@silverbackltd.com>
Reviewed-on: https://review.coreboot.org/23644
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
AMD support in coreboot has typically not used the AmdS3FinalRestore()
Entry Point. Add a call to it immediately prior to resuming to the OS.
BUG=b:69614064
TEST=Check console log for execution
Change-Id: Iadc4438d8cda9766002f6edade3c7b00b23b98b4
Signed-off-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Reviewed-on: https://review.coreboot.org/23443
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martinroth@google.com>
Add new functions that can execute InitRtb, InitResume, LateResume,
and FinalResume.
Note that the name AmdInitRtb supersedes the deprecated AmdS3Save.
TEST=Suspend/Resume Kahlee with complete S3 patch stack
BUG=b:69614064
Change-Id: I5c6a9c1a679a1c4d3f7d1d3b41a32efd0a2c2c01
Signed-off-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Reviewed-on: https://review.coreboot.org/22730
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martinroth@google.com>
The AGESA spec states that "Failure to release a structure can cause
undesired outcomes."
Uncomment the one in AmdInitLate(). The function only dealocates the
structure used for the AGESA entry point, and not the internal data
used by coreboot.
Release the structure in AmdInitEnv(). This appears to have been an
omission years ago when duplicating agesawrapper.c for every mainboard
was still common.
BUG=b:70671742
TEST=Build and boot Kahlee, inspect console log
Change-Id: Ib1ff94ec2acdc845c5e4b4ed7088061cfc0c55f3
Signed-off-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Reviewed-on: https://review.coreboot.org/22888
Reviewed-by: Martin Roth <martinroth@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Add functions that the wrapper will call to get and save the S3 data.
The wrapper requires two types of data saved:
* Non-volatile: Information that is the minimum required for bringing
the DRAM controller back online. This change uses the common
mrc_cache driver to manage the storage
* Volatile: May be stored in DRAM; information required to complete
the system restoration process.
TEST=Suspend/Resume Kahlee with complete S3 patch stack
BUG=b:69614064
Change-Id: Ie60162ea10f053393bc84e927dbd80c9279e6b63
Signed-off-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Reviewed-on: https://review.coreboot.org/22727
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martinroth@google.com>
Improve the file with:
* C99 inializations for structures
* reorder include files for aesthetics
* remove extraneous whitespace
* remove a stale comment
* make variable naming consistent
* make function arguments consistent
This change clears up all remaining checkpatch issues with the wrapper
with the exception of errors created with AMD definitions, e.g.
ERROR: need consistent spacing around '*' (ctx:WxV)
#32: FILE: src/soc/amd/common/block/pi/agesawrapper.c:32:
void __attribute__((weak)) SetFchMidParams(FCH_INTERFACE *params) {}
BUG=b:62240746
TEST=Build and boot Kahlee
Change-Id: I40985e0cf50df6aa4d830937e7f6b6e7908f72fe
Signed-off-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Reviewed-on: https://review.coreboot.org/22889
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
Revert most of 4f3f47b "amd/common: Define regions in cbmem". This
puts the management of the heap space back to its traditional
methodology. Subsequent patches that were to have used these
subregions have been reworked.
BUG=b:69614064
Change-Id: Ib3d40bcf61c50dbc481b60e7b5286f65a529b912
Signed-off-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Reviewed-on: https://review.coreboot.org/23516
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martinroth@google.com>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
The cbmem location holding the heap will be used to store additional
information in subsequent patches. Remove the static designation from
agesa_heap_base.
Change-Id: Ic607432fd6500ef69b5d47793896cf12a699d8b7
Signed-off-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Reviewed-on: https://review.coreboot.org/22721
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
A subsequent patch will use GetHeapBase() in more files than
heapmanager.c. Convert it to a format more similar to existing
coreboot source.
Change-Id: I8362af849fc9d7cb1b8a93113e8d78dcac51c20a
Signed-off-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Reviewed-on: https://review.coreboot.org/22903
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
In 6c747068 "amd/stoneyridge: Put AGESA heap into cbmem" the AGESA
heap was moved completely into cbmem. This was a departure from the
"late cbmem init" method of adding it late in post, then storing the
S3 volatile data to the region. Remove the hardcoded base address
that was missed in that commit.
To prepare for S3 support, split the region into subregions for
heap, AGESA's S3 volatile storage, and an MTRR save area.
BUG=b:69614064
Change-Id: I06c137f56516f3a04091d1191cd657a0aa07320b
Signed-off-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Reviewed-on: https://review.coreboot.org/22720
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martinroth@google.com>
Remove the original spi.c file that writes S3 NV data to flash in a
proprietary format. The s3 folder is retained to facilitate new
development.
Change-Id: I1b5fe8e854c3d2dd71506c2acd6ff73e4b86d7d4
Signed-off-by: Marshall Dawson <marshalldawson3rd@gmail.com>
Reviewed-on: https://review.coreboot.org/23305
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Richard Spiegel <richard.spiegel@silverbackltd.com>
Reviewed-by: Martin Roth <martinroth@google.com>
The deallocation was always subtracting the header, even when it
shouldn't. This caused problems for the allocator where buffer
sizes were incorrect and freed and used buffers could collide.
Fix the deallocation size.
Clear deallocated concatinated buffer header memory.
Fix the initial calculation of the total buffer size
available to be allocated.
BUG=b:71764350
TEST= Boot grunt.
BRANCH=none
Change-Id: I2789ddf72d662f24709dc5d9873741169cc4ef36
Signed-off-by: Marc Jones <marcj303@gmail.com>
Reviewed-on: https://review.coreboot.org/23284
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
This reverts commit 0f5651584ebb8e2ccfa151275bfd2f70e74bae9b.
This is not the correct fix for the heap allocator.
It looks like the root cause is in the buffer size of the
deallocate function.
Change-Id: I33c479a30d89a665677d3e4914194ae8136504af
Signed-off-by: Marc Jones <marcj303@gmail.com>
Reviewed-on: https://review.coreboot.org/23245
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Chris Ching <chingcodes@chromium.org>
Reviewed-by: Martin Roth <martinroth@google.com>
The heap allocator would try to split a buffer node that
was too small for another node. In the failing case, the buffer
node was 0x140 bytes and the requested size was 0x133 bytes.
The logic would check that there was room for the header and
buffer and try to split the buffer node. The buffer node header
is 0xC bytes, so 0x13F bytes are need. The problem is that it didn't
leave room for another node header and a little space for a buffer.
BUG=b:71764350
TEST= Boot grunt.
BRANCH=none
Change-Id: Iece5e12d5787415a335bb953985331a5dc312152
Signed-off-by: Marc Jones <marcj303@gmail.com>
Reviewed-on: https://review.coreboot.org/23211
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martinroth@google.com>
Reviewed-by: Daniel Kurtz <djkurtz@google.com>
Creating initial common acpi and implement halt.h
BUG=b:71575631
BRANCH=none
TEST=put poweroff() call in Kahlee's mainboard_final and board turns off
correctly
Change-Id: Ie7dd9851dcb240c53f2487b4f4b8a3e51d6b98d6
Signed-off-by: Chris Ching <chingcodes@chromium.org>
Reviewed-on: https://review.coreboot.org/23074
Reviewed-by: Marc Jones <marc@marcjonesconsulting.com>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martinroth@google.com>
CONFIG_AGESA_SPLIT_MEMORY_FILES controls whether AGESA is split into
pre- and post-memory binaries when it is built. Building AGESA this way
is required when doing the new "load post-memory AGESA binary into ram"
feature.
Thus, condition this new path on the CONFIG option being enabled.
BUG=b:71641792
TEST=build and boot kahlee with CONFIG_AGESA_SPLIT_MEMORY_FILES disabled
Change-Id: Ibec9db67437c57092e0f7acf0e3185865dc02688
Signed-off-by: Daniel Kurtz <djkurtz@chromium.org>
Reviewed-on: https://review.coreboot.org/23141
Reviewed-by: Justin TerAvest <teravest@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martinroth@google.com>
