The filename chip.h has a special purpose with the generation
of static devicetree, where the configuration structure name matches
the path to the chip.h file. For example, soc/intel/skylake/chip.h
defines struct soc_intel_skylake_config.
The renamed file did not follow this convention and the structure it
defines would conflict with one defined soc/intel/common/chip.h if such
is ever added.
Change-Id: Id3d56bf092c6111d2293136865b053b095e92d6b
Signed-off-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/35657
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
Reviewed-by: Furquan Shaikh <furquan@google.com>
CONFIG_CBFS_SIZE should only be used as a parameter to generate the
default FMAP.
This also swaps around FMDFILE and CBFS_SIZE to avoid that the
CBFS_SIZE entry disappears when filling in the FMDFILE entry below it.
One advantage is that if code references CONFIG_CBFS_SIZE the jenkins
buildtest will most likely fail as many boards provide an FMD file.
Change-Id: Ic7926e1638d7fb49ba61af28d682315786c3c39e
Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/35597
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Nico Huber <nico.h@gmx.de>
Reviewed-by: Werner Zeh <werner.zeh@siemens.com>
CB:35377 changed the behavior of find_fmap_directory() to return
pointer to CBMEM_ID_FMAP if fmap is cached in
cbmem. lb_boot_media_params() calls find_fmap_directory to add offset
of fmap in flash to coreboot table. However, because of the change in
behavior of find_fmap_directory(), it ended up adding 0 as the offset.
This change adds a new function get_fmap_flash_offset() which returns
the offset of fmap in flash. Ideally, all payloads should move to
using the FMAP from CBMEM. However, in order to maintain compatibility
with payloads which are not updated, ensure that fmap_offset is
updated correctly.
Since find_fmap_directory() is no longer used outside fmap.c, this
change also removes it from fmap.h and limits scope to fmap.c.
In a follow up patch, we need to push a change to libpayload to expose
the fmap cache pointer to lib_sysinfo.
BUG=b:141723751
Change-Id: I7ff6e8199143d1a992a83d7de1e3b44813b733f4
Signed-off-by: Furquan Shaikh <furquan@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/35639
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Patrick Rudolph <siro@das-labor.org>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
Reviewed-by: Shelley Chen <shchen@google.com>
dev_find_slot() can sometimes fail to return the desired device object
prior to full PCI enumeration. Comment the declaration and
implementation accordingly to help the user understand the problem and
avoid its usage.
Change-Id: I3fe1f24ff015d3e4f272323947f057e4c910186c
Signed-off-by: Aaron Durbin <adurbin@chromium.org>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/35632
Reviewed-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
The vendor id option set here is useless as most SSVID registers get
filled with 0x8086 (their VID) by default, anyway.
Besides that the Kconfig option isn't meant for retrofit ports, cf.
commit 7e1c83e31b (Add Kconfig options to override Subsystem Vendor and
Device ID). The right place would be the devicetree.
Signed-off-by: Michael Niewöhner <foss@mniewoehner.de>
Change-Id: If67c679bb342f63096902535734106e4f1651118
Reviewed-on: https://review.coreboot.org/c/coreboot/+/35524
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Patrick Rudolph <siro@das-labor.org>
All solid state devices have vendor id defined by JEDEC specification JEP106,
which originally allocated only 7 bits for it plus parity. When number of
vendors exploded beyond 126, a banking proposition came maintaining
compatibility with older vendors while allowing for 4 extra bits (16 banks)
through the introduction of the concept "Continuation code", denoted by the
byte value of 0x7f.
Examples:
0xfe, 0x60, 0x18, 0x00, 0x00 => vendor 0xfe of bank o
0x7f, 0x7f, 0xfe, 0x60, 0x18 => vendor 0xfe of bank 2
BUG=b:141535133
TEST=Build and boot grunt.
Change-Id: I16c5df70b8ba65017d1a45c79e90a76d1f78550c
Signed-off-by: Richard Spiegel <richard.spiegel@silverbackltd.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/35589
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Edward O'Callaghan <quasisec@chromium.org>
Most of the X11 boards with socket LGA1151 are basically the same boards
with just some minor differences like different NICs (1 GbE, 10 GbE),
number of NICs / PCIe ports etc.
There are about 20 boards that can be added, if there is a community for
testing.
To be able to add more x11 boards easily like x11ssm (see CB:35427) this
restructures the x11ssh tree to represent a "X11 LGA1151 series". There
were multiple suggestions for the structure like grouping by series
(x10, x11, x...), grouping by chipset or by cpu family.
It turned out that there are some "X11 series" boards that are
completely different. Grouping by chipset or cpu family suffers from the
same problem. This is why finally we agreed on grouping by series and
socket ("X11 LGA1151 series").
The structure uses the common baseboard scheme, while there is no "real"
baseboard we know of. By checking images, comparing logs etc. we came to
the conclusion that Supermicro does have some base layout which is only
modified a bit for the different boards.
X11SSH-TF was moved to the variants/ folder with it's gpio.h. As we
expect the other boards to have mostly the same device tree, there is a
common devicetree that gets overridden by each variant's overridetree.
Besides that some very minor modifications happened (formatting, fixing
comments, ...) but not much.
Documentation is reworked in CB:35547
Change-Id: I8dc4240ae042760a845e890b923ad40478bb8e29
Signed-off-by: Michael Niewöhner <foss@mniewoehner.de>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/35426
Reviewed-by: Patrick Rudolph <siro@das-labor.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
It is only called in ramstage. Even if it was called in
romstage, execution flow is such that BSP and AP CPUs
should not be able to enter update_microcode() routine
concurrently.
Also the Kconfig guarding the spin_lock() calls are not
selected nor are the lock variables declared for these
platforms.
Change-Id: I1c2e106f10e8420e942b3ed082c677c0db95557c
Signed-off-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/35586
Reviewed-by: Michał Żygowski <michal.zygowski@3mdeb.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
CB:29633 switched platform to use sb/common spi implementation,
which worked until CAR_GLOBAL was removed in CB:30506.
Revert the changes back to usage of CAR_GLOBAL in the common spi
driver so that flashconsole will work again in romsatge for
fsp_broadwell_de.
Test: verify flashconsole functional on out-of-tree Broadwell-DE board
Change-Id: I72e5db1583199b5ca4b6ec54661282544d326f0f
Signed-off-by: Matt DeVillier <matt.devillier@gmail.com>
Signed-off-by: Werner Zeh <werner.zeh@siemens.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/32880
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Fix regression from commit ecea916
cpu/intel/common: Extend FSB detection to cover TSC
MSR_EBC_FREQUENCY_ID (0x2c) was not defined for affected
CPU models and rdmsr() caused reset loops. Implementations
deviate from public documentation.
Change to IA32_PERF_STATUS (0x198) already used in i945/udelay.c
to detect FSB to TSC multiplier.
Change-Id: I7a91da221920a7e7bfccb98d76115b5c89e3b52e
Signed-off-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/35548
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
Reviewed-by: HAOUAS Elyes <ehaouas@noos.fr>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
The Razer Blade Stealth H2U is a KabyLake System using:
- Intel KBL 7500U
- ITE8528E SuperIO
- Intel 600P Series NVMe SSD
- Either four MT52L1G32D4PG (16GB) or MT52L512MB32D4PG (8GB)
of soldered memory in dualchannel mode
- (Optional) Touchscreen
- HDMI 2.0a via DP-1: Paradetech PS175
- AlpineRidge Thunderbolt 3 controller
- TPS65982 USB-PD power switch / multiplexer
Even though it has a 16MB chip equipped (W25Q128.V) only the first 8MB
are used and mapped via IFD. The rest is left empty (0xFF). The flash is
not secured in any way and can be read via flashrom. It should be the
source for this port's IFD and ME blobs.
Working:
- USB-A Ports left and right
- Speakers
- Touchscreen (USB)
- Onboard Keyboard in Linux
- NVMe SSD
- SeaBIOS, Tianocore and Grub Payloads
- Webcam
- Powersaving Modes
- Battery state and LID switch, sometimes slow to update.
- Touchpad (I2C-HID)
- Headphones
Not part of this commit:
- Thunderbolt / USB-C (Requires advanced EC signaling)
- Full HDMI support (Currently requires plugged connection at boot)
Change-Id: I7ede881d631e1863f07f5130f84bc3b8ca61a350
Signed-off-by: Johanna Schander <coreboot@mimoja.de>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/34475
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Jonathan Neuschäfer <j.neuschaefer@gmx.net>
When accessing register with multiple bit fields, the common approach is
to use clrsetbits_le32, for example:
clrsetbits(®, (1 << 0) | (0x3 << 1) | (0x7 << 10),
(1 << 0) | (0x1 << 1) | (0x5 << 10));
This hard to maintain because we have to calculate the mask values
manually, make sure the duplicated shift (offset) was set correctly.
And it may be even worse if the value to set will be based on some
runtime values (that many developers will do a if-block with two very
similar argument list), and leaving lots of magic numbers.
We want to encourage developers always giving field names, and have a
better way of setting fields. The proposed utility macros are:
DEFINE_BITFIELD(name, high_bit, low_bit)
EXTRACT_BITFIELD(value, name)
WRITE32_BITFIELDS(addr, name, value, [name2, value2, ...])
READ32_BITFIELD(addr, name)
Where a developer can easily convert from data sheet like
BITS NAME
26:24 SEC_VIO
Into a declaration
DEFINE_BITFIELD(SEC_VIO, 26, 24)
Then, a simple call can set the field as:
WRITE32_BITFIELDS(®, SEC_VIO, 2);
That is much easier to understand than
clrsetbits_le32(®, 0x7 << 24, 0x2 << 24);
And to extract the value:
READ32_BITFIELD(®, SEC_VIO)
That is equivalent to:
(read32(®) & 0x3) >> 24
Change-Id: I8a1b17142f7a7dc6c441b0b1ee67d60d73ec8cc8
Signed-off-by: Hung-Te Lin <hungte@chromium.org>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/35463
Reviewed-by: Julius Werner <jwerner@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
- configure DMA fw_cfg
- add support to read using fw_cfg_dma
- provide fw config version id info in logs
BUG=N/A
TEST=Build and boot using qemu-i440fx.
Change-Id: I0be5355b124af40aba62c0840790d46ed0fe80a2
Signed-off-by: Himanshu Sahdev <himanshusah@hcl.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/35365
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Patrick Rudolph <siro@das-labor.org>
Devices using eMCP may run at a high DRAM frequency (e.g., 3600Mbps)
while those with discrete DRAM can only run at 3200Mbps. This patch
enables 3600Mbps for eMCP DDR for better system performance.
BUG=b:80501386
BRANCH=none
TEST=Boots correctly and stress test passes on Kukui
Change-Id: Iab6a9c2c390feeb9497b051a255b29566909e656
Signed-off-by: Huayang Duan <huayang.duan@mediatek.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/34990
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Hung-Te Lin <hungte@chromium.org>
