Add support for Texas Instruments LP5569 LED driver.
Texas Instruments LP5569 is 9 channels chip with programmable engines.
It almost a copy of LP5523 with fundamental changes to regs order and
regs content.
Has difference in how the clock is handled and doesn't support detecting
clock time automatically, different handling for selftest and different
scheme for the status regs.
LED chip supports ENGINE and MUX to group LED and run precompiled code
with magic values to run patterns. This is loaded via the sysfs entry
and it's passed as a string of ASCII HEX char.
One some devices using this LED Controller (a NBG7815 Router) it was
found loading big precompiled pattern with up to 96 bytes of code. To
have support for this "extended" scenario, hardcode each engine to
support 4 pages of precompiled pattern (128 bytes of code) and 1 page
for each MUX. This gives plenty of space for any kind precompiled
pattern keeping simple logic for page handling of each engine and mux.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Link: https://lore.kernel.org/r/20240626160027.19703-21-ansuelsmth@gmail.com
Signed-off-by: Lee Jones <lee@kernel.org>
Some LED chip supports up to 16 pages and with some magic they can be
divided in 4 page for each ENGINE + 1 for each MUX. Following this we
can support bigger programs up to 128 bytes.
Rework the update_program_memory function to support program of multiple
pages instead of hardcoding it to one page per programs.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Link: https://lore.kernel.org/r/20240626160027.19703-19-ansuelsmth@gmail.com
Signed-off-by: Lee Jones <lee@kernel.org>
Generalize sysfs master_fader since their implementation is the same across
some lp55xx based LED driver.
Suggested-by: Lee Jones <lee@kernel.org>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Link: https://lore.kernel.org/r/20240626160027.19703-18-ansuelsmth@gmail.com
Signed-off-by: Lee Jones <lee@kernel.org>
Generalize sysfs engine_leds since their implementation is the same across
some lp55xx based LED driver.
While at it simplify the implementation for show_engine_leds.
Suggested-by: Lee Jones <lee@kernel.org>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Link: https://lore.kernel.org/r/20240626160027.19703-17-ansuelsmth@gmail.com
Signed-off-by: Lee Jones <lee@kernel.org>
Generalize sysfs engine_load and engine_mode since their implementation
is the same acrosso some lp55xx based LED driver.
Suggested-by: Lee Jones <lee@kernel.org>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Link: https://lore.kernel.org/r/20240626160027.19703-16-ansuelsmth@gmail.com
Signed-off-by: Lee Jones <lee@kernel.org>
Generalize stop_engine function as the implementation is the same for
most of the lp55xx based LED driver.
Suggested-by: Lee Jones <lee@kernel.org>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Link: https://lore.kernel.org/r/20240626160027.19703-15-ansuelsmth@gmail.com
Signed-off-by: Lee Jones <lee@kernel.org>
Generalize turn_off_channels function as the implementation is the same for
most of the lp55xx based LED driver.
Suggested-by: Lee Jones <lee@kernel.org>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Link: https://lore.kernel.org/r/20240626160027.19703-14-ansuelsmth@gmail.com
Signed-off-by: Lee Jones <lee@kernel.org>
Generalize set_led_current function as the implementation is the same for
most of the lp55xx based LED driver.
Suggested-by: Lee Jones <lee@kernel.org>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Link: https://lore.kernel.org/r/20240626160027.19703-13-ansuelsmth@gmail.com
Signed-off-by: Lee Jones <lee@kernel.org>
Generalize multicolor_brightness function as the implementation is the
same for most of the lp55xx based LED driver.
Suggested-by: Lee Jones <lee@kernel.org>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Link: https://lore.kernel.org/r/20240626160027.19703-12-ansuelsmth@gmail.com
Signed-off-by: Lee Jones <lee@kernel.org>
Generalize led_brightness function as the implementation is the same for
most of the lp55xx based LED driver.
Introduce a new option in device_config, reg_led_pwm_base since the reg
value is not the same for every LED chip.
Suggested-by: Lee Jones <lee@kernel.org>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Link: https://lore.kernel.org/r/20240626160027.19703-11-ansuelsmth@gmail.com
Signed-off-by: Lee Jones <lee@kernel.org>
Generalize firmware_loaded function as lp55xx based LED driver all share
the same logic.
Suggested-by: Lee Jones <lee@kernel.org>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Link: https://lore.kernel.org/r/20240626160027.19703-10-ansuelsmth@gmail.com
Signed-off-by: Lee Jones <lee@kernel.org>
LED Driver based on lp55xx all use the same logic to write memory in
SMEM. The only difference is that legacy chip doesn't support pages and
have the engine regs one after another.
To handle this apply the same logic used for load_engine also for
update_program_memory.
Introduce a new config in device_config, base_prog. For LED chip
that doesn't support pages, offset this values of 32 for each engine.
Update all lp55xx based LED driver to use this new function and define
all the required bits.
Suggested-by: Lee Jones <lee@kernel.org>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Link: https://lore.kernel.org/r/20240626160027.19703-9-ansuelsmth@gmail.com
Signed-off-by: Lee Jones <lee@kernel.org>
Generalize run_engine function for lp55xx based LED driver. The logic is
similar to every LED driver, rework it with more macro magic and account
for LED model that might have OP MODE and EXEC at base offset in the
reg.
Update any lp55xx based LED driver to use this generalized function and
declare required bits.
Suggested-by: Lee Jones <lee@kernel.org>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Link: https://lore.kernel.org/r/20240626160027.19703-8-ansuelsmth@gmail.com
Signed-off-by: Lee Jones <lee@kernel.org>
Generalize load_engine_and_select_page by reworking the implementation
and making it part of the generic load_engine function.
Add a new option in device_config, pages_per_engine used to define pages
assigned to each engine. With this option set, it's assumed LED chip
supports pages and load_engine will correctly setup the write page.
An equal amount of pages is assigned to each engine and they are
assigned from page 0.
Update any lp55xx based LED driver to define the option and use the new
function.
Suggested-by: Lee Jones <lee@kernel.org>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Link: https://lore.kernel.org/r/20240626160027.19703-7-ansuelsmth@gmail.com
Signed-off-by: Lee Jones <lee@kernel.org>
LED driver based on lp55xx have all a very similar implementation for
load_engine function. Move the function to lp55xx-common and rework the
define to be more dynamic instead of having to declare a temp array for
them.
Engine mask are the same for every LED based on lp55xx.
Suggested-by: Lee Jones <lee@kernel.org>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Link: https://lore.kernel.org/r/20240626160027.19703-6-ansuelsmth@gmail.com
Signed-off-by: Lee Jones <lee@kernel.org>
Now that stop_all_engine is generalized, probe and remove function are
the same across every lp55xx based LED driver and can be generalized.
To permit to use a common probe, make use of the OF match_data and i2c
driver_data value to store the device_config struct specific for the
LED.
Also drop the now unused exported symbol in lp55xx-common and make them
static.
Update any lp55xx based LED driver to use the new generic probe/remove.
Suggested-by: Lee Jones <lee@kernel.org>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Link: https://lore.kernel.org/r/20240626160027.19703-5-ansuelsmth@gmail.com
Signed-off-by: Lee Jones <lee@kernel.org>
In all the lp55xx based driver, we have a similar implementation of the
stop_all_engine function with the only difference of the required sleep
for the OP MODE change.
The main difference is legacy LEDs require a min of 152 us while new one
use a generic 1-2ms. The new one use a 1-2ms sleep as suggested in the
datasheet IN ALTERNATIVE to a much more robust approach by using the
newly introduced ENGINE_BUSY bit in the STATUS reg.
To better handle sleep after OP MODE change, add support for polling the
ENGINE_BUSY bit and use the legacy sleep for old LEDs.
With this change, stop_all_engine can be generalized and moved to
lp55xx-common.
To make more clear the double usage of lp55xx_reg, define a union for
additional scope of mask and shift.
Update all lp55xx based driver to use the new generalized function and
define the required bits in the device_config struct.
Suggested-by: Lee Jones <lee@kernel.org>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Link: https://lore.kernel.org/r/20240626160027.19703-4-ansuelsmth@gmail.com
Signed-off-by: Lee Jones <lee@kernel.org>
Add new ti,lp5569 compatible, this is similar to national,lp5523 with
slight change to reg order and reg type and advanced way for LED fault.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Acked-by: Conor Dooley <conor.dooley@microchip.com>
Link: https://lore.kernel.org/r/20240626160027.19703-3-ansuelsmth@gmail.com
Signed-off-by: Lee Jones <lee@kernel.org>
pwr-sel property is specific to ti,lp8501, make it conditional of the
related compatible.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Acked-by: Rob Herring (Arm) <robh@kernel.org>
Link: https://lore.kernel.org/r/20240626160027.19703-2-ansuelsmth@gmail.com
Signed-off-by: Lee Jones <lee@kernel.org>
The SY7802 is a current-regulated charge pump which can regulate two
current levels for Flash and Torch modes.
It is a high-current synchronous boost converter with 2-channel high
side current sources. Each channel is able to deliver 900mA current.
Acked-by: Lee Jones <lee@kernel.org>
Signed-off-by: André Apitzsch <git@apitzsch.eu>
Link: https://lore.kernel.org/r/20240624-sy7802-v5-2-7abc9d96bfa6@apitzsch.eu
Signed-off-by: Lee Jones <lee@kernel.org>
Default blinking period is set to 2s. This is too long to be handled by
the hardware (maximum is 1.69s).
Set the default blinking period to 1s to match what is done in the
other LED drivers.
Signed-off-by: Bastien Curutchet <bastien.curutchet@bootlin.com>
Link: https://lore.kernel.org/r/20240617143910.154546-5-bastien.curutchet@bootlin.com
Signed-off-by: Lee Jones <lee@kernel.org>
When a LED is used to drive a beeper, it uses PWM1. This can cause
conflicts if an other LED want to use PWM1 for blinking.
Disable use of hardware for blinking when one or more LEDs are used to
drive beepers.
Signed-off-by: Bastien Curutchet <bastien.curutchet@bootlin.com>
Link: https://lore.kernel.org/r/20240617143910.154546-4-bastien.curutchet@bootlin.com
Signed-off-by: Lee Jones <lee@kernel.org>
PSC0/PWM0 are used by all LEDs for brightness and blinking. This causes
conflicts when you set a brightness of a new LED while an other one is
already using PWM0 for blinking.
Use PSC1/PWM1 for blinking.
Check that no other LED is already blinking with a different PSC1/PWM1
configuration to avoid conflict.
Signed-off-by: Bastien Curutchet <bastien.curutchet@bootlin.com>
Link: https://lore.kernel.org/r/20240617143910.154546-3-bastien.curutchet@bootlin.com
Signed-off-by: Lee Jones <lee@kernel.org>
Two tables are used to configure the PWM and PSC registers of the two
PWM available in the pca9532. Magic numbers are used to access this table
instead of defines.
Add defines PCA9532_PWM_ID_0 and PCA9532_PWM_ID_1 and use them in place of
these magic numbers.
Signed-off-by: Bastien Curutchet <bastien.curutchet@bootlin.com>
Link: https://lore.kernel.org/r/20240617143910.154546-2-bastien.curutchet@bootlin.com
Signed-off-by: Lee Jones <lee@kernel.org>
Use __free for device_node values, and thus drop calls to
of_node_put.
The variable attribute __free adds a scope based cleanup to
the device node. The goal is to reduce memory management issues
in the kernel code.
The of_node_put calls were removed, and the
for_each_available_child_of_node was replaced to the equivalent
for_each_available_child_of_node_scoped which use the __free.
Suggested-by: Julia Lawall <julia.lawall@inria.fr>
Signed-off-by: MarileneGarcia <marilene.agarcia@gmail.com>
Link: https://lore.kernel.org/r/20240601031713.1307859-1-marilene.agarcia@gmail.com
Signed-off-by: Lee Jones <lee@kernel.org>
Pavel's repo hasn't been used actively for quite some time.
Let's make it official.
Cc: Pavel Machek <pavel@ucw.cz>
Signed-off-by: Lee Jones <lee@kernel.org>
The fwnode_for_each_child_node() loop requires manual intervention to
decrement the child refcount in case of an early return.
Add the missing calls to fwnode_handle_put(child) to avoid memory leaks
in the error paths.
Cc: stable@vger.kernel.org
Fixes: 679f865206 ("leds: Add mt6360 driver")
Signed-off-by: Javier Carrasco <javier.carrasco.cruz@gmail.com>
Acked-by: Pavel Machek <pavel@ucw.cz>
Link: https://lore.kernel.org/r/20240611-leds-mt6360-memleak-v1-1-93642eb5011e@gmail.com
Signed-off-by: Lee Jones <lee@kernel.org>
Use __free() for device_node values, and thus drop calls to
of_node_put().
The variable attribute __free() adds a scope based cleanup to
the device node. The goal is to reduce memory management issues
in the kernel code.
The for_each_available_child_of_node() was replaced to the equivalent
for_each_available_child_of_node_scoped() which uses the __free().
Suggested-by: Julia Lawall <julia.lawall@inria.fr>
Signed-off-by: Marilene A Garcia <marilene.agarcia@gmail.com>
Link: https://lore.kernel.org/r/20240611001740.10490-1-marilene.agarcia@gmail.com
Signed-off-by: Lee Jones <lee@kernel.org>
Currently, all LED LPG devices will call lpg_{set,clear}_pbs_trigger()
when setting brightness regardless of if they support PPG and have PBS
triggers. Check if device supports PPG before setting/clearing PBS
triggers.
Fixes: 6ab1f766a8 ("leds: rgb: leds-qcom-lpg: Add support for PPG through single SDAM")
Fixes: 5e9ff62686 ("leds: rgb: leds-qcom-lpg: Include support for PPG with dedicated LUT SDAM")
Signed-off-by: Anjelique Melendez <quic_amelende@quicinc.com>
Reviewed-by: Bjorn Andersson <andersson@kernel.org>
Link: https://lore.kernel.org/r/20240607005250.4047135-1-quic_amelende@quicinc.com
Signed-off-by: Lee Jones <lee@kernel.org>
The race fixed in timer_trig_activate() between a blocking
set_brightness() call and trigger->activate() can affect any trigger.
So move the call to flush_work() into led_trigger_set() where it can
avoid the race for all triggers.
Fixes: 0db37915d9 ("leds: avoid races with workqueue")
Fixes: 8c0f693c6e ("leds: avoid flush_work in atomic context")
Cc: stable@vger.kernel.org
Tested-by: Dustin L. Howett <dustin@howett.net>
Signed-off-by: Thomas Weißschuh <linux@weissschuh.net>
Link: https://lore.kernel.org/r/20240613-led-trigger-flush-v2-1-f4f970799d77@weissschuh.net
Signed-off-by: Lee Jones <lee@kernel.org>
There is no code that uses ID table directly, except the
struct device_driver at the end of the file. Hence, move
table closer to its user. It's always possible to access
them via a pointer.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Link: https://lore.kernel.org/r/20240606173037.3091598-7-andriy.shevchenko@linux.intel.com
Signed-off-by: Lee Jones <lee@kernel.org>
In this driver LEDs are registered using devm_led_classdev_register()
so they are automatically unregistered after module's remove() is done.
led_classdev_unregister() calls module's led_set_brightness() to turn off
the LEDs and that callback uses mutex which was destroyed already
in module's remove() so use devm API instead.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Link: https://lore.kernel.org/r/20240606173037.3091598-6-andriy.shevchenko@linux.intel.com
Signed-off-by: Lee Jones <lee@kernel.org>
Add a label so that a bit of exception handling can be better reused
at the end of this function implementation.
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Link: https://lore.kernel.org/r/5faec5de-fc36-4b38-abcb-c61954a824cd@web.de
Signed-off-by: Lee Jones <lee@kernel.org>
If both set_brightness functions return -ENOTSUPP, then the LED doesn't
support setting a fixed brightness value, and the error message isn't
helpful. This can be the case e.g. for LEDs supporting a specific hw
trigger only.
Pinched the subject line and commit message from Heiner:
Link: https://lore.kernel.org/all/44177e37-9512-4044-8991-bb23b184bf37@gmail.com/
Reworked the function to provide Heiner's required semantics whilst
simultaneously increasing readability and flow.
Cc: Pavel Machek <pavel@ucw.cz>
Cc: linux-leds@vger.kernel.org
Suggested-by: Heiner Kallweit <hkallweit1@gmail.com>
Reviewed-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: Lee Jones <lee@kernel.org>
The for_each_child_of_node() macro automatically decrements the child
refcount at the end of every iteration. On early exits, of_node_put()
must be used to manually decrement the refcount and avoid memory leaks.
The scoped version of the macro accounts for such early breaks, fixing
the early exits without the need for explicit calls to of_node_put().
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Julia Lawall <julia.lawall@inria.fr>
Closes: https://lore.kernel.org/r/202406172314.2vVgelaS-lkp@intel.com/
Fixes: b72755f5b5 ("backlight: Add new lm3509 backlight driver")
Signed-off-by: Javier Carrasco <javier.carrasco.cruz@gmail.com>
Link: https://lore.kernel.org/r/20240624-lm3509_bl_scoped-v1-1-ceba9df38f23@gmail.com
Signed-off-by: Lee Jones <lee@kernel.org>
For SoCs like gs101 that need a special regmap, register this with
of_syscon_register_regmap api, so it can be returned by
syscon_regmap_lookup_by_phandle() and friends.
For SoCs that don't require a custom regmap, revert back to syscon
creating the mmio regmap rather than duplicating the logic here.
exynos_get_pmu_regmap_by_phandle() api is also updated to retrieve
the regmap via syscon. The exynos_get_pmu_regmap_by_phandle() api
is kept around until fw_devlink support for syscon property is added
for the pinctrl-samsung driver that also runs at postcore_initcall
level.
All other exynos client drivers can revert back to
syscon_regmap_lookup_by_phandle().
Signed-off-by: Peter Griffin <peter.griffin@linaro.org>
Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Sam Protsenko <semen.protsenko@linaro.org>
Tested-by: Will McVicker <willmcvicker@google.com>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Link: https://lore.kernel.org/r/20240621115544.1655458-3-peter.griffin@linaro.org
Signed-off-by: Lee Jones <lee@kernel.org>
The of_syscon_register_regmap() API allows an externally created regmap
to be registered with syscon. This regmap can then be returned to client
drivers using the syscon_regmap_lookup_by_phandle() APIs.
The API is used by platforms where mmio access to the syscon registers is
not possible, and a underlying soc driver like exynos-pmu provides a SoC
specific regmap that can issue a SMC or hypervisor call to write the
register.
This approach keeps the SoC complexities out of syscon, but allows common
drivers such as syscon-poweroff, syscon-reboot and friends that are used
by many SoCs already to be re-used.
Signed-off-by: Peter Griffin <peter.griffin@linaro.org>
Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Sam Protsenko <semen.protsenko@linaro.org>
Tested-by: Will McVicker <willmcvicker@google.com>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Link: https://lore.kernel.org/r/20240621115544.1655458-2-peter.griffin@linaro.org
Signed-off-by: Lee Jones <lee@kernel.org>
The timeout when waiting for the PRE safe window is rather short, as
normally we would only need to wait a few dozen usecs for the problematic
scanline region to pass and we don't want to spin too long in case
something goes wrong. This however mixes badly with preemption, as we
can easily get scheduled away from the CPU for a longer time than our
timeout, in which case we would hit a spurious timeout and wrongly skip
the PRE update.
Instead of disabling preemption across the wait loop, potentially
impacting the overall system latency, use a wait loop with a fixed
max number of iterations, so time spent away from the CPU is not
accounted against the timeout budget.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Link: https://lore.kernel.org/r/20240517104549.3648939-3-l.stach@pengutronix.de
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20240517104549.3648939-3-l.stach@pengutronix.de
imx-drm doesn't mandate a modeset when the framebuffer modifier changes,
but currently the tile prefetch and resolve (TPR) configuration of the
PRE is only set up on the initial modeset.
As the TPR configuration is double buffered, same as all the other PRE
states, we can support dynamic reconfiguration of the buffer layout from
one frame to another. As switching between (super-)tiled and linear
prefetch needs to touch the CTRL register make sure to do the
reconfiguration inside the safe window.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Link: https://lore.kernel.org/r/20240517104549.3648939-2-l.stach@pengutronix.de
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20240517104549.3648939-2-l.stach@pengutronix.de
Björn Töpel <bjorn@kernel.org> says:
From: Björn Töpel <bjorn@rivosinc.com>
================================================================
Memory Hot(Un)Plug support (and ZONE_DEVICE) for the RISC-V port
================================================================
Introduction
============
To quote "Documentation/admin-guide/mm/memory-hotplug.rst": "Memory
hot(un)plug allows for increasing and decreasing the size of physical
memory available to a machine at runtime."
This series adds memory hot(un)plugging, and ZONE_DEVICE support for
the RISC-V Linux port.
MM configuration
================
RISC-V MM has the following configuration:
* Memory blocks are 128M, analogous to x86-64. It uses PMD
("hugepage") vmemmaps. From that follows that 2M (PMD) worth of
vmemmap spans 32768 pages á 4K which gets us 128M.
* The pageblock size is the minimum minimum virtio_mem size, and on
RISC-V it's 2M (2^9 * 4K).
Implementation
==============
The PGD table on RISC-V is shared/copied between for all processes. To
avoid doing page table synchronization, the first patch (patch 1)
pre-allocated the PGD entries for vmemmap/direct map. By doing that
the init_mm PGD will be fixed at kernel init, and synchronization can
be avoided all together.
The following two patches (patch 2-3) does some preparations, followed
by the actual MHP implementation (patch 4-5). Then, MHP and virtio-mem
are enabled (patch 6-7), and finally ZONE_DEVICE support is added
(patch 8).
MHP and locking
===============
TL;DR: The MHP does not step on any toes, except for ptdump.
Additional locking is required for ptdump.
Long version: For v2 I spent some time digging into init_mm
synchronization/update. Here are my findings, and I'd love them to be
corrected if incorrect.
It's been a gnarly path...
The `init_mm` structure is a special mm (perhaps not a "real" one).
It's a "lazy context" that tracks kernel page table resources, e.g.,
the kernel page table (swapper_pg_dir), a kernel page_table_lock (more
about the usage below), mmap_lock, and such.
`init_mm` does not track/contain any VMAs. Having the `init_mm` is
convenient, so that the regular kernel page table walk/modify
functions can be used.
Now, `init_mm` being special means that the locking for kernel page
tables are special as well.
On RISC-V the PGD (top-level page table structure), similar to x86, is
shared (copied) with user processes. If the kernel PGD is modified, it
has to be synched to user-mode processes PGDs. This is avoided by
pre-populating the PGD, so it'll be fixed from boot.
The in-kernel pgd regions are documented in
`Documentation/arch/riscv/vm-layout.rst`.
The distinct regions are:
* vmemmap
* vmalloc/ioremap space
* direct mapping of all physical memory
* kasan
* modules, BPF
* kernel
Memory hotplug is the process of adding/removing memory to/from the
kernel.
Adding is done in two phases:
1. Add the memory to the kernel
2. Online memory, making it available to the page allocator.
Step 1 is partially architecture dependent, and updates the init_mm
page table:
* Update the direct map page tables. The direct map is a linear map,
representing all physical memory: `virt = phys + PAGE_OFFSET`
* Add a `struct page` for each added page of memory. Update the
vmemmap (virtual mapping to the `struct page`, so we can easily
transform a kernel virtual address to a `struct page *` address.
From an MHP perspective, there are two regions of the PGD that are
updated:
* vmemmap
* direct mapping of all physical memory
The `struct mm_struct` has a couple of locks in play:
* `spinlock_t page_table_lock` protects the page table, and some
counters
* `struct rw_semaphore mmap_lock` protect an mm's VMAs
Note again that `init_mm` does not contain any VMAs, but still uses
the mmap_lock in some places.
The `page_table_lock` was originally used to to protect all pages
tables, but more recently a split page table lock has been introduced.
The split lock has a per-table lock for the PTE and PMD tables. If
split lock is disabled, all tables are guarded by
`mm->page_table_lock` (for user processes). Split page table locks are
not used for init_mm.
MHP operations is typically synchronized using
`DEFINE_STATIC_PERCPU_RWSEM(mem_hotplug_lock)`.
Actors
------
The following non-MHP actors in the kernel traverses (read), and/or
modifies the kernel PGD.
* `ptdump`
Walks the entire `init_mm`, via `ptdump_walk_pgd()` with the
`mmap_write_lock(init_mm)` taken.
Observation: ptdump can race with MHP, and needs additional locking
to avoid crashes/races.
* `set_direct_*` / `arch/riscv/mm/pageattr.c`
The `set_direct_*` functionality is used to "synchronize" the
direct map to other kernel mappings, e.g. modules/kernel text. The
direct map is using "as large huge table mappings as possible",
which means that the `set_direct_*` might need to split the direct
map.
The `set_direct_*` functions operates with the
`mmap_write_lock(init_mm)` taken.
Observation: `set_direct_*` uses the direct map, but will never
modify the same entry as MHP. If there is a mapping, that entry will
never race with MHP. Further, MHP acts when memory is offline.
* HVO / `mm/hugetlb_vmemmap`
HVO optimizes the backing `struct page` for hugetlb pages, which
means changing the "vmemmap" region. HVO can split (merge?) a
vmemmap pmd. However, it will never race with MHP, since HVO only
operates at online memory. HVO cannot touch memory being MHP added
or removed.
* `apply_to_page_range`
Walks a range, creates pages and applies a callback (setting
permissions) for the page.
When creating a table, it might use `int __pte_alloc_kernel(pmd_t
*pmd)` which takes the `init_mm.page_table_lock` to synchronize pmd
populate.
Used by: `mm/vmalloc.c` and `mm/kasan/shadow.c`. The KASAN callback
takes the `init_mm.page_table_lock` to synchronize pte creation.
Observations: `apply_to_page_range` applies to the "vmalloc/ioremap
space" region, and "kasan" region. *Not* affected by MHP.
* `apply_to_existing_page_range`
Walks a range, applies a callback (setting permissions) for the
page (no page creation).
Used by: `kernel/bpf/arena.c` and `mm/kasan/shadow.c`. The KASAN
callback takes the `init_mm.page_table_lock` to synchronize pte
creation. *Not* affected by MHP regions.
* `apply_to_existing_page_range` applies to the "vmalloc/ioremap
space" region, and "kasan" region. *Not* affected by MHP regions.
* `ioremap_page_range` and `vmap_page_range`
Uses the same internal function, and might create table entries at
the "vmalloc/ioremap space" region. Can call
`__pte_alloc_kernel()` which takes the `init_mm.page_table_lock`
synchronizing pmd populate in the region. *Not* affected by MHP
regions.
Summary:
* MHP add will never modify the same page table entries, as any of
the other actors.
* MHP remove is done when memory is offlined, and will not clash
with any of the actors.
* Functions that walk the entire kernel page table need
synchronization
* It's sufficient to add the MHP lock ptdump.
Testing
=======
This series adds basic DT supported hotplugging. There is a QEMU
series enabling MHP for the RISC-V "virt" machine here: [1]
ACPI/MSI support is still in the making for RISC-V, and prior proper
(ACPI) PCI MSI support lands [2] and NUMA SRAT support [3], it hard to
try it out.
I've prepared a QEMU branch with proper ACPI GED/PC-DIMM support [4],
and a this series with the required prerequisites [5] (AIA, ACPI AIA
MADT, ACPI NUMA SRAT).
To test with virtio-mem, e.g.:
| qemu-system-riscv64 \
| -machine virt,aia=aplic-imsic \
| -cpu rv64,v=true,vlen=256,elen=64,h=true,zbkb=on,zbkc=on,zbkx=on,zkr=on,zkt=on,svinval=on,svnapot=on,svpbmt=on \
| -nodefaults \
| -nographic -smp 8 -kernel rv64-u-boot.bin \
| -drive file=rootfs.img,format=raw,if=virtio \
| -device virtio-rng-pci \
| -m 16G,slots=3,maxmem=32G \
| -object memory-backend-ram,id=mem0,size=16G \
| -numa node,nodeid=0,memdev=mem0 \
| -serial chardev:char0 \
| -mon chardev=char0,mode=readline \
| -chardev stdio,mux=on,id=char0 \
| -device pci-serial,id=serial0,chardev=char0 \
| -object memory-backend-ram,id=vmem0,size=2G \
| -device virtio-mem-pci,id=vm0,memdev=vmem0,node=0
where "rv64-u-boot.bin" is U-boot with EFI/ACPI-support (use [6] if
you're lazy).
In the QEMU monitor:
| (qemu) info memory-devices
| (qemu) qom-set vm0 requested-size 1G
...to test DAX/KMEM, use the follow QEMU parameters:
| -object memory-backend-file,id=mem1,share=on,mem-path=virtio_pmem.img,size=4G \
| -device virtio-pmem-pci,memdev=mem1,id=nv1
and the regular ndctl/daxctl dance.
If you're brave to try the ACPI branch, add "acpi=on" to "-machine
virt", and test PC-DIMM MHP (in addition to virtio-{p},mem):
In the QEMU monitor:
| (qemu) object_add memory-backend-ram,id=mem1,size=1G
| (qemu) device_add pc-dimm,id=dimm1,memdev=mem1
You can also try hot-remove with some QEMU options, say:
| -object memory-backend-file,id=mem-1,size=256M,mem-path=/pagesize-2MB
| -device pc-dimm,id=mem1,memdev=mem-1
| -object memory-backend-file,id=mem-2,size=1G,mem-path=/pagesize-1GB
| -device pc-dimm,id=mem2,memdev=mem-2
| -object memory-backend-file,id=mem-3,size=256M,mem-path=/pagesize-2MB
| -device pc-dimm,id=mem3,memdev=mem-3
Remove "acpi=on" to run with DT.
Thanks to Alex, Andrew, David, and Oscar for all
comments/tests/fixups.
References
==========
[1] https://lore.kernel.org/qemu-devel/20240521105635.795211-1-bjorn@kernel.org/
[2] https://lore.kernel.org/linux-riscv/20240501121742.1215792-1-sunilvl@ventanamicro.com/
[3] https://lore.kernel.org/linux-riscv/cover.1713778236.git.haibo1.xu@intel.com/
[4] https://github.com/bjoto/qemu/commits/virtio-mem-pc-dimm-mhp-acpi-v2/
[5] https://github.com/bjoto/linux/commits/mhp-v4-acpi
[6] https://github.com/bjoto/riscv-rootfs-utils/tree/acpi
* b4-shazam-merge:
riscv: Enable DAX VMEMMAP optimization
riscv: mm: Add support for ZONE_DEVICE
virtio-mem: Enable virtio-mem for RISC-V
riscv: Enable memory hotplugging for RISC-V
riscv: mm: Take memory hotplug read-lock during kernel page table dump
riscv: mm: Add memory hotplugging support
riscv: mm: Add pfn_to_kaddr() implementation
riscv: mm: Refactor create_linear_mapping_range() for memory hot add
riscv: mm: Change attribute from __init to __meminit for page functions
riscv: mm: Pre-allocate vmemmap/direct map/kasan PGD entries
riscv: mm: Properly forward vmemmap_populate() altmap parameter
Link: https://lore.kernel.org/r/20240605114100.315918-1-bjorn@kernel.org
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
Now that DAX is usable, enable the DAX VMEMMAP optimization as well.
Signed-off-by: Björn Töpel <bjorn@rivosinc.com>
Reviewed-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Link: https://lore.kernel.org/r/20240605114100.315918-12-bjorn@kernel.org
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
