With the recent switch to SMM module loader v2, the size of the SMM for
module google/volteer increased to above 64K in size, and thus failed to
install the permanent SMM handler. Turns out, the devicetree is all
pulled into the SMM build because of elog, which calls
`pci_dev_is_wake_source`, and is the only user of `struct device` in
SMM. Changing this function to take a pci_devfn_t instead allows the
linker to remove almost the entire devicetree from SMM (only usage left
is when disabling HECI via SMM).
BUG=b:186661594
TEST=Verify loaded program size of `smm.elf` for google/volteer is
almost ~50% smaller.
Signed-off-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
Change-Id: I4c39e5188321c8711d6479b15065e5aaedad8f38
Reviewed-on: https://review.coreboot.org/c/coreboot/+/52765
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Furquan Shaikh <furquan@google.com>
On Intel 6-series PCHs, the GCAP register is R/WO (Read / Write Once),
and needs to be written to after the HD Audio controller is taken out
of reset. Add a Kconfig option to read and write back GCAP in order to
lock it down. Follow-up commits will select this option when switching
platforms to use common Azalia code, to preserve original behaviour.
Change-Id: I70bab20816fb6c0bf7bff35c3d2f5828cd96172d
Signed-off-by: Patrick Rudolph <patrick.rudolph@9elements.com>
Signed-off-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/50794
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
The default of 32 buses per hotplug bridge is rather high. Especially
for platforms that limit MMConf space to 64 buses: they run out of
numbers if there is more than a single hotplug bridge.
Lower the default to
* 8 if MMConf is limited to 64 or less buses,
* 16 if MMConf is limited to 128 or less buses.
Change-Id: I06d522dd92ceea9f4798273b26f947a5333800c3
Signed-off-by: Nico Huber <nico.h@gmx.de>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/52069
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Drop unnecessary typedefs and rename DDR4-specific definitions to avoid
name clashes, as done for DDR3 in earlier commits. This allows including
and using both DDR3 and DDR4 headers in the same compilation unit.
Change-Id: I17f1cd88f83251ec23e9783a617f4d2ed41b07f0
Signed-off-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/51898
Reviewed-by: Michael Niewöhner <foss@mniewoehner.de>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
The timeout is never reached when the codec is functioning properly.
Using a small timeout value can result in spurious errors with some
codecs, e.g. a codec that is slow to respond but operates correctly.
When a codec is non-operative, the timeout is only reached once per
verb table, thus the impact on booting time is relatively small. So,
use a reasonably long enough timeout to cover all possible cases.
Remove the unconditional 25 µs delay and increase the timeout delay.
The new value of 1 ms is the maximum of all existing implementations.
Currently, the only boards using this code are AMD reference boards:
- AMD Bilby
- AMD Mandolin
- AMD Padmelon
Change-Id: Ia5e4829d404dcecdb9e7a377e896a319cb38531a
Signed-off-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/51634
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Nico Huber <nico.h@gmx.de>
Use timer.h helpers instead of open-coding timeout handling in polling
loops. The 25-microsecond delay in `wait_for_valid` looks odd, and may
be removed in subsequent commits. For now, preserve existing behavior.
Change-Id: Id1227c6812618597c37408a7bf53bcbcae97374a
Signed-off-by: Patrick Rudolph <patrick.rudolph@9elements.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/50789
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
I was bugged by spurious "Failed to enable LTR" messages for years.
Looking at the the current algorithm, it is flawed in multiple ways:
* It looks like the author didn't know they implemented a
recursive algorithm (pciexp_enable_ltr()) inside another
recursive algorithm (pciexp_scan_bridge()). Thus, at every
tree level, everything is run again for the whole sub-
tree.
* LTR is enabled no matter if `.set_ltr_max_latencies` is
implemented or not. Leaving the endpoints' LTR settings
at 0: They are told to always report zero tolerance.
In theory, depending on the root-complex implementation,
this may result in higher power consumption than without
LTR messages.
* `.set_ltr_max_latencies` is only considered for the direct
parent of a device. Thus, even with it implemented, an
endpoint below a (non-root) bridge may suffer from the 0
settings as described above.
* Due to the double-recursive nature, LTR is enabled starting
with the endpoints, then moving up the tree, while the PCIe
spec tells us to do it in the exact opposite order.
With the current implementation of pciexp_scan_bridge(), it is
hard to hook anything in that runs for each device from top to
bottom. So the proposed solution still adds some redundancy:
First, for every device that uses pciexp_scan_bus(), we enable
LTR if possible (see below). Then, when returning from the bus-
scanning recursion, we enable LTR for every device and configure
the maximum latencies (if supported). The latter runs again on
all bridges, because it's hard to know if pciexp_scan_bus() was
used for them.
When to enable LTR:
* For all devices that implement `.set_ltr_max_latencies`.
* For all devices below a bridge that has it enabled already.
Change-Id: I2c5b8658f1fc8cec15e8b0824464c6fc9bee7e0e
Signed-off-by: Nico Huber <nico.h@gmx.de>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/51328
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-by: Tim Wawrzynczak <twawrzynczak@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
On some mainboards, codec configuration depends on settings that are
only known at runtime, which is impossible to specify using one verb
table. Add an optional `mainboard_azalia_program_runtime_verbs` hook
where mainboards can program runtime-dependent codec verbs.
Change-Id: I7efeba5c26051aeb5061cce191ace08c304a6c70
Signed-off-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/50388
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Patrick Rudolph <siro@das-labor.org>
On some boards, Azalia configuration depends on config settings that are
not known at compile-time. Expose a function to program a verb table, to
be used in subsequent commits.
Change-Id: Ie9607f6e733df66f0ca26a4bb70e0864ce1d4512
Signed-off-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/50387
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Patrick Rudolph <siro@das-labor.org>
It was commented that the need for the delay was mainly related
to external displays and only with VBIOS execution. Move the
delay such that it is done only when we actually need to execute
the VBIOS aka option rom.
A delay is currently only defined for librem/purism_bdw in
its Kconfig. As the description of the issue sounds like it
would equally happen on other platforms when VBIOS is involved,
promote the Kconfig visible option to global scope.
Change-Id: I4503158576f35057373f003586bbf76af4d59b3d
Signed-off-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/48787
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Revise the following aspects to follow coreboot's coding style:
- Drop braces for single-statement condition and loop bodies.
- Use `__func__` to print the current function's name.
- Reflow pointer dereferences to fit in a single line.
- Adjust the `*` position in pointer variable declarations.
- Drop unnecessary `else` statements.
BUG = N/A
TEST = Build Compulab Intense-PC with secure oprom enabled
Change-Id: I780251d946d5bea97658476d61d25555ec768dfc
Signed-off-by: Frans Hendriks <fhendriks@eltan.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/49963
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
This is necessary because ASL Memory32Fixed values cannot contain
operations, even if they can be evaluated to constants. Add a sanity
check in pci_mmio_cfg.h to ensure consistency with MMCONF_BUS_NUMBER.
Change-Id: I8f0b5edf166580cc12c1363d8d6b6ef0f2854be9
Signed-off-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/50033
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Systems can boot to the OS without a display. Don't kill the boot
process based on a vBIOS error, instead just display a warning.
If the issue is actually fatal for some reason, it's going to die
at some point anyway.
BUG=b:175843172
TEST=Boot morphius to OS without a display
BRANCH=Zork
Signed-off-by: Martin Roth <martinroth@chromium.org>
Change-Id: I7d261321cdbe423dd754f6a354e5f50b53563fcb
Reviewed-on: https://review.coreboot.org/c/coreboot/+/49764
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Raul Rangel <rrangel@chromium.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-by: Nico Huber <nico.h@gmx.de>
Platform can now select VGA_ROM_RUN_DEFAULT Kconfig to perform graphics
initialization for PCI-E based discrete card through VGA OpRom
(SoC or Mainboard user can't select VGA_ROM_RUN directly because
it's part of choice option).
(Note: Some payloads, like SeaBIOS, are also able to run Option ROMs,
so coreboot does not need to enable VGA_ROM_RUN Kconfig)
For payload like depthcharge, create VGA_ROM_RUN_DEFAULT Kconfig
for mainboard to select design with DGPU where OpROM is embedded
inside the DGPU card.
Allow auto selection of VGA_ROM_RUN_DEFAULT from VGA_BIOS Kconfig.
Also NO_GFX_INIT Kconfig to avoid running VGA_ROM_RUN
by default in case SeaBIOS is used.
TEST=Able to get Pre-OS splash screen with AMD Radeon RX 5700 PCI-E
DGPU when mainboard user selects VGA_ROM_RUN_DEFAULT.
Change-Id: Iecb2fcdb105af449bc20ad727759cdef17d5e376
Signed-off-by: Subrata Banik <subrata.banik@intel.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/49016
Reviewed-by: Nico Huber <nico.h@gmx.de>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Introduce a new device `gpio` that is going to be used for generic
abstraction of gpio operations in the devicetree.
The general idea behind this is that every chip can have gpios that
shall be accessible in a very generic way by any driver through the
devicetree.
The chip that implements the chip-specific gpio operations has to assign
them to the generic device operations struct, which then gets assigned
to the gpio device during device probing. See CB:48583 for how this gets
done for the SoCs using intelblocks/gpio.
The gpio device then can be added to the devicetree with an alias name
like in the following example:
chip soc/whateverlake
device gpio 0 alias soc_gpio on end
...
end
Any driver that requires access to this gpio device needs to have a
device pointer (or multiple) and an option for specifying the gpio to be
used in its chip config like this:
struct drivers_ipmi_config {
...
DEVTREE_CONST struct device *gpio_dev;
u16 post_complete_gpio;
...
};
The device `soc_gpio` can then be linked to the chip driver's `gpio_dev`
above by using the syntax `use ... as ...`, which was introduced in
commit 8e1ea52:
chip drivers/ipmi
use soc_gpio as gpio_dev
register "bmc_jumper_gpio" = "GPP_D22"
...
end
The IPMI driver can then use the generic gpio operations without any
knowlege of the chip's specifics:
unsigned int gpio_val;
const struct gpio_operations *gpio_ops;
gpio_ops = dev_get_gpio_ops(conf->gpio_dev);
gpio_val = gpio_ops->get(conf->bmc_jumper_gpio);
For a full example have a look at CB:48096 and CB:48095.
This change adds the new device type to sconfig and adds generic gpio
operations to the `device_operations` struct. Also, a helper for getting
the gpio operations from a device after checking them for NULL pointers
gets added.
Successfully tested on Supermicro X11SSM-F with CB:48097, X11SSH-TF with
CB:48711 and OCP DeltaLake with CB:48672.
Change-Id: Ic4572ad8b37bd1afd2fb213b2c67fb8aec536786
Tested-by: Johnny Lin <Johnny_Lin@wiwynn.com>
Tested-by: Michael Niewöhner <foss@mniewoehner.de>
Tested-by: Patrick Rudolph <siro@das-labor.org>
Signed-off-by: Michael Niewöhner <foss@mniewoehner.de>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/48582
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Nico Huber <nico.h@gmx.de>
Many uses of `azalia_set_bits` are used to toggle the reset bit. To
avoid having to repeat the register operations and the corresponding
comment, create two helpers with self-explanatory names. They will be
put to use in subsequent commits, with one change for each function.
Change-Id: If0594fdaf99319f08a2e272cd37958f0f216e654
Signed-off-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/48355
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Patrick Rudolph <siro@das-labor.org>
The `4` here doesn't have to do with the size of u32. Instead, it is
because the verb header contains the number of jacks, which is the
number of four-verb groups.
Change-Id: I3956ce5ec2a7abc29982504cf75b262a1c098af5
Signed-off-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/48352
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Patrick Rudolph <siro@das-labor.org>
