Pull irqchip updates from Marc Zyngier:
- Add support for the STM32MP13 variant
- Move parent device away from struct irq_chip
- Remove all instances of non-const strings assigned to
struct irq_chip::name, enabling a nice cleanup for VIC and GIC)
- Simplify the Qualcomm PDC driver
- A bunch of SiFive PLIC cleanups
- Add support for a new variant of the Meson GPIO block
- Add support for the irqchip side of the Apple M1 PMU
- Add support for the Apple M1 Pro/Max AICv2 irqchip
- Add support for the Qualcomm MPM wakeup gadget
- Move the Xilinx driver over to the generic irqdomain handling
- Tiny speedup for IPIs on GICv3 systems
- The usual odd cleanups
Link: https://lore.kernel.org/all/20220313105142.704579-1-maz@kernel.org
Qualcomm SoCs based on the RPM architecture have a MSM Power Manager (MPM)
in always-on domain. In addition to managing resources during sleep, the
hardware also has an interrupt controller that monitors the interrupts
when the system is asleep, wakes up the APSS when one of these interrupts
occur and replays it to GIC after it becomes operational.
It adds an irqchip driver for this interrupt controller, and here are
some notes about it.
- For given SoC, a fixed number of MPM pins are supported, e.g. 96 pins
on QCM2290. Each of these MPM pins can be either a MPM_GIC pin or
a MPM_GPIO pin. The mapping between MPM_GIC pin and GIC interrupt
is defined by SoC, as well as the mapping between MPM_GPIO pin and
GPIO number. The former mapping is retrieved from device tree, while
the latter is defined in TLMM pinctrl driver.
- The power domain (PD) .power_off hook is used to notify RPM that APSS
is about to power collapse. This requires MPM PD be the parent PD of
CPU cluster.
- When SoC gets awake from sleep mode, the driver will receive an
interrupt from RPM, so that it can replay interrupt for particular
polarity.
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20220308080534.3384532-3-shawn.guo@linaro.org
* irq/aic-v2:
: .
: Add support for the interrupt controller found is the latest
: incarnation of Apple M1 systems, courtesy of Hector Martin.
: .
irqchip/apple-aic: Add support for AICv2
irqchip/apple-aic: Support multiple dies
irqchip/apple-aic: Dynamically compute register offsets
irqchip/apple-aic: Switch to irq_domain_create_tree and sparse hwirqs
irqchip/apple-aic: Add Fast IPI support
dt-bindings: interrupt-controller: apple,aic2: New binding for AICv2
PCI: apple: Change MSI handling to handle 4-cell AIC fwspec form
Signed-off-by: Marc Zyngier <maz@kernel.org>
Introduce support for the new AICv2 hardware block in t6000/t6001 SoCs.
It seems these blocks are missing the information required to compute
the event register offset in the capability registers, so we specify
that in the DT as a second reg entry.
Signed-off-by: Hector Martin <marcan@marcan.st>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20220309192123.152028-8-marcan@marcan.st
This allows us to directly use the hardware event number as the hwirq
number. Since IRQ events have bit 16 set (type=1), FIQs now move to
starting at hwirq number 0.
This will become more important once multi-die support is introduced in
a later commit.
Signed-off-by: Hector Martin <marcan@marcan.st>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20220309192123.152028-5-marcan@marcan.st
The newer AICv2 present in t600x SoCs does not have legacy IPI support
at all. Since t8103 also supports Fast IPIs, implement support for this
first. The legacy IPI code is left as a fallback, so it can be
potentially used by older SoCs in the future.
The vIPI code is shared; only the IPI firing/acking bits change for Fast
IPIs.
Signed-off-by: Hector Martin <marcan@marcan.st>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20220309192123.152028-4-marcan@marcan.st
This new incompatible revision of the AIC peripheral introduces
multi-die support. This binding is based on apple,aic, but
changes interrupt-cells to add a new die argument.
Also adds a second reg entry to specify the offset of the event
register. Inexplicably, the capability registers allow us to compute
other register offsets, but not this one. This allows us to keep
forward-compatibility with future SoCs that will likely implement
different die counts, thus shifting the event register. Apple also
specify the offset explicitly in their device tree...
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Hector Martin <marcan@marcan.st>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20220309192123.152028-3-marcan@marcan.st
* irq/meson-gpio:
: .
: Expand meson-gpio support to deal with the new Meson-S4 SoC
: .
irqchip/meson-gpio: Add support for meson s4 SoCs
irqchip/meson-gpio: add select trigger type callback
irqchip/meson-gpio: support more than 8 channels gpio irq
dt-bindings: interrupt-controller: New binding for Meson-S4 SoCs
Signed-off-by: Marc Zyngier <maz@kernel.org>
The meson s4 SoCs support 12 gpio irq lines compared with previous
serial chips and have something different, details are as below.
IRQ Number:
- 80:68 13 pins on bank Z
- 67:48 20 pins on bank X
- 47:36 12 pins on bank H
- 35:24 12 pins on bank D
- 23:22 2 pins on bank E
- 21:14 8 pins on bank C
- 13:0 13 pins on bank B
Signed-off-by: Qianggui Song <qianggui.song@amlogic.com>
[maz: fixed some W=1 build warnings]
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20220225055207.1048-5-qianggui.song@amlogic.com
* irq/misc-5.18:
: .
: Misc irq chip changes for 5.18
:
: - GICv3: Relax ordering of previous stores to only include the ISH domain
:
: - nvic: Unmap MMIo region on probe failure
:
: - xilinx: Switch to GENERIC_IRQ_MULTI_HANDLER when used on microblaze
: .
irqchip/xilinx: Switch to GENERIC_IRQ_MULTI_HANDLER
irqchip/nvic: Release nvic_base upon failure
irqchip/gic-v3: Use dsb(ishst) to order writes with ICC_SGI1R_EL1 accesses
Signed-off-by: Marc Zyngier <maz@kernel.org>
* irq/plic-cleanups:
: .
: SiFive PLIC cleanups from Niklas Cassel:
:
: - Clarify some of the namings in the driver
:
: - Make sure S-mode interrupts are disabled when running in M-mode
: .
irqchip/sifive-plic: Disable S-mode IRQs if running in M-mode
irqchip/sifive-plic: Improve naming scheme for per context offsets
Signed-off-by: Marc Zyngier <maz@kernel.org>
generic_handle_irq() is invoked from a regular interrupt service
routine. This handler will become a forced-threaded handler on
PREEMPT_RT and will be invoked with enabled interrupts. The
generic_handle_irq() must be invoked with disabled interrupts in order
to avoid deadlocks.
Instead of manually disabling interrupts before invoking use
generic_handle_irq_safe() which can be invoked with enabled and disabled
interrupts.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Mauro Carvalho Chehab <mchehab@kernel.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Link: https://lore.kernel.org/r/20220211181500.1856198-5-bigeasy@linutronix.de
When detecting a context for a privilege mode different from the current
running privilege mode, we simply skip to the next context register.
This means that we never clear the S-mode enable bits when running in
M-mode.
On canaan k210, a bunch of S-mode interrupts are enabled by the bootrom.
These S-mode specific interrupts should never trigger, since we never set
the mie.SEIE bit in the parent interrupt controller (riscv-intc).
However, we will be able to see the mip.SEIE bit set as pending.
This isn't a good default when CONFIG_RISCV_M_MODE is set, since in that
case we will never enter a lower privilege mode (e.g. S-mode).
Let's clear the S-mode enable bits when running the kernel in M-mode, such
that we won't have a interrupt pending bit set, which we will never clear.
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Reviewed-by: Anup Patel <anup@brainfault.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20220302131544.3166154-3-Niklas.Cassel@wdc.com
The PLIC supports a fixed number of contexts (15872).
Each context has fixed register offsets in PLIC.
The number of contexts that we need to initialize depends on the privilege
modes supported by each hart. Therefore, this mapping between PLIC context
registers to hart privilege modes is platform specific, and is currently
supplied via device tree.
For example, canaan,k210 has the following mapping:
Context0: hart0 M-mode
Context1: hart0 S-mode
Context2: hart1 M-mode
Context3: hart1 S-mode
While sifive,fu540 has the following mapping:
Context0: hart0 M-mode
Context1: hart1 M-mode
Context2: hart1 S-mode
Because the number of contexts per hart is not fixed, the names
ENABLE_PER_HART and CONTEXT_PER_HART for the register offsets are quite
confusing and might mislead the reader to think that these are fixed
register offsets per hart.
Rename the offsets to more clearly highlight that these are per PLIC
context and not per hart.
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Reviewed-by: Anup Patel <anup@brainfault.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20220302131544.3166154-2-Niklas.Cassel@wdc.com
* irq/qcom-pdc-cleanup:
: .
: Spring cleanup for the Qualcomm PDC driver, simplifying its
: use of irq domains, replacing open-coded functionnalities with
: the core code equivalent, and fixing the dodgy locking.
: .
irqchip/qcom-pdc: Drop open coded version of __assign_bit()
irqchip/qcom-pdc: Fix broken locking
irqchip/qcom-pdc: Kill qcom_pdc_translate helper
irqchip/qcom-pdc: Kill non-wakeup irqdomain
irqchip/qcom-pdc: Kill PDC_NO_PARENT_IRQ
Signed-off-by: Marc Zyngier <maz@kernel.org>
pdc_enable_intr() serves as a primitive to qcom_pdc_gic_{en,dis}able,
and has a raw spinlock for mutual exclusion, which is uses with
interruptible primitives.
This means that this critical section can itself be interrupted.
Should the interrupt also be a PDC interrupt, and the endpoint driver
perform an irq_disable() on that interrupt, we end-up in a deadlock.
Fix this by using the irqsave/irqrestore variants of the locking
primitives.
Signed-off-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Maulik Shah <quic_mkshah@quicinc.com>
Link: https://lore.kernel.org/r/20220224101226.88373-5-maz@kernel.org
A careful look at the way the PDC driver works shows that:
- all interrupts are in the same space
- all interrupts are treated the same
And yet the driver creates two domains based on whether
the interrupt gets mapped directly or from the pinctrl code,
which is obviously a waste of resources.
Kill the non-wakeup domain and unify all the interrupt handling.
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20220224101226.88373-3-maz@kernel.org
PDC_NO_PARENT_IRQ is pretty pointless, as all it indicates is
that the PDC terminates the interrupt hierarchy. Which is
exactly the same as not having a mapping in the GIC space.
This is also bad practice to treat the absence of a hwirq
as a hwirq itself.
Just explicitly use the region mapping pointer, and drop
the definition.
Signed-off-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Maulik Shah <quic_mkshah@quicinc.com>
Link: https://lore.kernel.org/r/20220224101226.88373-2-maz@kernel.org
A dsb(ishst) barrier should be enough to order previous writes with
the system register generating the SGI, as we only need to guarantee
the visibility of data to other CPUs in the inner shareable domain
before we send the SGI.
A micro-benchmark is written to verify the performance impact on
kunpeng920 machine with 2 sockets, each socket has 2 dies, and
each die has 24 CPUs, so totally the system has 2 * 2 * 24 = 96
CPUs. ~2% performance improvement can be seen by this benchmark.
The code of benchmark module:
#include <linux/module.h>
#include <linux/timekeeping.h>
volatile int data0 ____cacheline_aligned;
volatile int data1 ____cacheline_aligned;
volatile int data2 ____cacheline_aligned;
volatile int data3 ____cacheline_aligned;
volatile int data4 ____cacheline_aligned;
volatile int data5 ____cacheline_aligned;
volatile int data6 ____cacheline_aligned;
static void ipi_latency_func(void *val)
{
}
static int __init ipi_latency_init(void)
{
ktime_t stime, etime, delta;
int cpu, i;
int start = smp_processor_id();
stime = ktime_get();
for ( i = 0; i < 1000; i++)
for (cpu = 0; cpu < 96; cpu++) {
data0 = data1 = data2 = data3 = data4 = data5 = data6 = cpu;
smp_call_function_single(cpu, ipi_latency_func, NULL, 1);
}
etime = ktime_get();
delta = ktime_sub(etime, stime);
printk("%s ipi from cpu%d to cpu0-95 delta of 1000times:%lld\n",
__func__, start, delta);
return 0;
}
module_init(ipi_latency_init);
static void ipi_latency_exit(void)
{
}
module_exit(ipi_latency_exit);
MODULE_DESCRIPTION("IPI benchmark");
MODULE_LICENSE("GPL");
run the below commands 10 times on both Vanilla and the kernel with this
patch:
# taskset -c 0 insmod test.ko
# rmmod test
The result on vanilla:
ipi_latency_init ipi from cpu0 to cpu0-95 delta of 1000times:126757449
ipi_latency_init ipi from cpu0 to cpu0-95 delta of 1000times:126784249
ipi_latency_init ipi from cpu0 to cpu0-95 delta of 1000times:126177703
ipi_latency_init ipi from cpu0 to cpu0-95 delta of 1000times:127022281
ipi_latency_init ipi from cpu0 to cpu0-95 delta of 1000times:126184883
ipi_latency_init ipi from cpu0 to cpu0-95 delta of 1000times:127374585
ipi_latency_init ipi from cpu0 to cpu0-95 delta of 1000times:125778089
ipi_latency_init ipi from cpu0 to cpu0-95 delta of 1000times:126974441
ipi_latency_init ipi from cpu0 to cpu0-95 delta of 1000times:127357625
ipi_latency_init ipi from cpu0 to cpu0-95 delta of 1000times:126228184
The result on the kernel with this patch:
ipi_latency_init ipi from cpu0 to cpu0-95 delta of 1000times:124467401
ipi_latency_init ipi from cpu0 to cpu0-95 delta of 1000times:123474209
ipi_latency_init ipi from cpu0 to cpu0-95 delta of 1000times:123558497
ipi_latency_init ipi from cpu0 to cpu0-95 delta of 1000times:122993951
ipi_latency_init ipi from cpu0 to cpu0-95 delta of 1000times:122984223
ipi_latency_init ipi from cpu0 to cpu0-95 delta of 1000times:123323609
ipi_latency_init ipi from cpu0 to cpu0-95 delta of 1000times:124507583
ipi_latency_init ipi from cpu0 to cpu0-95 delta of 1000times:123386963
ipi_latency_init ipi from cpu0 to cpu0-95 delta of 1000times:123340664
ipi_latency_init ipi from cpu0 to cpu0-95 delta of 1000times:123285324
Signed-off-by: Barry Song <song.bao.hua@hisilicon.com>
[maz: tidied up commit message]
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20220220061910.6155-1-21cnbao@gmail.com
Pull locking fix from Borislav Petkov:
"Fix a NULL ptr dereference when dumping lockdep chains through
/proc/lockdep_chains"
* tag 'locking_urgent_for_v5.17_rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
lockdep: Correct lock_classes index mapping
Pull x86 fixes from Borislav Petkov:
- Fix the ptrace regset xfpregs_set() callback to behave according to
the ABI
- Handle poisoned pages properly in the SGX reclaimer code
* tag 'x86_urgent_for_v5.17_rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/ptrace: Fix xfpregs_set()'s incorrect xmm clearing
x86/sgx: Fix missing poison handling in reclaimer
Pull scheduler fix from Borislav Petkov:
"Fix task exposure order when forking tasks"
* tag 'sched_urgent_for_v5.17_rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
sched: Fix yet more sched_fork() races
Pull EDAC fix from Borislav Petkov:
"Fix a long-standing struct alignment bug in the EDAC struct allocation
code"
* tag 'edac_urgent_for_v5.17_rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/ras/ras:
EDAC: Fix calculation of returned address and next offset in edac_align_ptr()
Pull SCSI fixes from James Bottomley:
"Three fixes, all in drivers.
The ufs and qedi fixes are minor; the lpfc one is a bit bigger because
it involves adding a heuristic to detect and deal with common but not
standards compliant behaviour"
* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
scsi: ufs: core: Fix divide by zero in ufshcd_map_queues()
scsi: lpfc: Fix pt2pt NVMe PRLI reject LOGO loop
scsi: qedi: Fix ABBA deadlock in qedi_process_tmf_resp() and qedi_process_cmd_cleanup_resp()
Pull dmaengine fixes from Vinod Koul:
"A bunch of driver fixes for:
- ptdma error handling in init
- lock fix in at_hdmac
- error path and error num fix for sh dma
- pm balance fix for stm32"
* tag 'dmaengine-fix-5.17' of git://git.kernel.org/pub/scm/linux/kernel/git/vkoul/dmaengine:
dmaengine: shdma: Fix runtime PM imbalance on error
dmaengine: sh: rcar-dmac: Check for error num after dma_set_max_seg_size
dmaengine: stm32-dmamux: Fix PM disable depth imbalance in stm32_dmamux_probe
dmaengine: sh: rcar-dmac: Check for error num after setting mask
dmaengine: at_xdmac: Fix missing unlock in at_xdmac_tasklet()
dmaengine: ptdma: Fix the error handling path in pt_core_init()
Pull i2c fixes from Wolfram Sang:
"Some driver updates, a MAINTAINERS fix, and additions to COMPILE_TEST
(so we won't miss build problems again)"
* 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
MAINTAINERS: remove duplicate entry for i2c-qcom-geni
i2c: brcmstb: fix support for DSL and CM variants
i2c: qup: allow COMPILE_TEST
i2c: imx: allow COMPILE_TEST
i2c: cadence: allow COMPILE_TEST
i2c: qcom-cci: don't put a device tree node before i2c_add_adapter()
i2c: qcom-cci: don't delete an unregistered adapter
i2c: bcm2835: Avoid clock stretching timeouts
Pull input fixes from Dmitry Torokhov:
- a fix for Synaptics touchpads in RMI4 mode failing to suspend/resume
properly because I2C client devices are now being suspended and
resumed asynchronously which changed the ordering
- a change to make sure we do not set right and middle buttons
capabilities on touchpads that are "buttonpads" (i.e. do not have
separate physical buttons)
- a change to zinitix touchscreen driver adding more compatible
strings/IDs
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
Input: psmouse - set up dependency between PS/2 and SMBus companions
Input: zinitix - add new compatible strings
Input: clear BTN_RIGHT/MIDDLE on buttonpads
Pull power supply fixes from Sebastian Reichel:
"Three regression fixes for the 5.17 cycle:
- build warning fix for power-supply documentation
- pointer size fix in cw2015 battery driver
- OOM handling in bq256xx charger driver"
* tag 'for-v5.17-rc' of git://git.kernel.org/pub/scm/linux/kernel/git/sre/linux-power-supply:
power: supply: bq256xx: Handle OOM correctly
power: supply: core: fix application of sizeof to pointer
power: supply: fix table problem in sysfs-class-power
Pull mount_setattr test/doc fixes from Christian Brauner:
"This contains a fix for one of the selftests for the mount_setattr
syscall to create idmapped mounts, an entry for idmapped mounts for
maintainers, and missing kernel documentation for the helper we split
out some time ago to get and yield write access to a mount when
changing mount properties"
* tag 'fs.mount_setattr.v5.17-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/brauner/linux:
fs: add kernel doc for mnt_{hold,unhold}_writers()
MAINTAINERS: add entry for idmapped mounts
tests: fix idmapped mount_setattr test
Pull pidfd fix from Christian Brauner:
"This fixes a problem reported by lockdep when installing a pidfd via
fd_install() with siglock and the tasklisk write lock held in
copy_process() when calling clone()/clone3() with CLONE_PIDFD.
Originally a pidfd was created prior to holding any of these locks but
this required a call to ksys_close(). So quite some time ago in
6fd2fe494b ("copy_process(): don't use ksys_close() on cleanups") we
switched to a get_unused_fd_flags() + fd_install() model.
As part of that we moved fd_install() as late as possible. This was
done for two main reasons. First, because we needed to ensure that we
call fd_install() past the point of no return as once that's called
the fd is live in the task's file table. Second, because we tried to
ensure that the fd is visible in /proc/<pid>/fd/<pidfd> right when the
task is visible.
This fix moves the fd_install() to an even later point which means
that a task will be visible in proc while the pidfd isn't yet under
/proc/<pid>/fd/<pidfd>.
While this is a user visible change it's very unlikely that this will
have any impact. Nobody should be relying on that and if they do we
need to come up with something better but again, it's doubtful this is
relevant"
* tag 'pidfd.v5.17-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/brauner/linux:
copy_process(): Move fd_install() out of sighand->siglock critical section
