When an interrupt is to be serviced, the convention is to mask the
interrupt at the chip and unmask after servicing the interrupt. Enabling
and disabling the interrupt at the PDC irqchip causes an interrupt storm
due to the way dual edge interrupts are handled in hardware.
Skip configuring the PDC when the IRQ is masked and unmasked, instead
use the irq_enable/irq_disable callbacks to toggle the IRQ_ENABLE
register at the PDC. The PDC's IRQ_ENABLE register is only used during
the monitoring mode when the system is asleep and is not needed for
active mode detection.
Signed-off-by: Lina Iyer <ilina@codeaurora.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Stephen Boyd <swboyd@chromium.org>
Link: https://lore.kernel.org/r/1573855915-9841-4-git-send-email-ilina@codeaurora.org
Add process for the situation that more than one irq is coming to
a single chip at the same time. The original code will only respond
to the lowest setted bit in JZ_REG_INTC_PENDING, and then exit the
interrupt dispatch function. After exiting the interrupt dispatch
function, since the second interrupt has not yet responded, the
interrupt dispatch function is again entered to process the second
interrupt. This creates additional unnecessary overhead, and the
more interrupts that occur at the same time, the more overhead is
added. The improved method in this patch is to check whether there
are still unresponsive interrupts after processing the lowest
setted bit interrupt. If there are any, the processing will be
processed according to the bit in JZ_REG_INTC_PENDING, and the
interrupt dispatch function will be exited until all processing
is completed.
Signed-off-by: Zhou Yanjie <zhouyanjie@zoho.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Paul Cercueil <paul@crapouillou.net>
Link: https://lore.kernel.org/r/1570015525-27018-6-git-send-email-zhouyanjie@zoho.com
We have so far always injected/cleared VLPIs using either
INT+SYNC or CLEAR+SYNC sequences, but that's pretty wrong
for two reasons:
- SYNC only synchronises physical LPIs
- The collection ID that for the associated LPI doesn't match
the redistributor the vPE is associated with
Instead, send an {INT,CLEAR}+VSYNC for forwarded LPIs, ensuring
that the ITS synchronises against the virtual pending table.
Reported-by: Zenghui Yu <yuzenghui@huawei.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20191108165805.3071-10-maz@kernel.org
We have so far alwways invalidated VLPIs usinc an INV+SYNC
sequence, but that's pretty wrong for two reasons:
- SYNC only synchronises physical LPIs
- The collection ID that for the associated LPI doesn't match
the redistributor the vPE is associated with
Instead, send an INV+VSYNC for forwarded LPIs, ensuring that
the ITS can properly synchronise the invalidation of VLPIs.
Fixes: 015ec0386a ("irqchip/gic-v3-its: Add VLPI configuration handling")
Reported-by: Zenghui Yu <yuzenghui@huawei.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20191108165805.3071-9-maz@kernel.org
We allocate the collection mapping on device creation, but somehow
free it on the irqdomain free path, which is pretty inconsistent
and has led to bugs in the past.
Move it to the point where we teardown the device, making the
alloc/free symetric.
Signed-off-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Zenghui Yu <yuzenghui@huawei.com>
Link: https://lore.kernel.org/r/20191108165805.3071-2-maz@kernel.org
The LS1021A allows inverting the polarity of six interrupt lines
IRQ[0:5] via the scfg_intpcr register, effectively allowing
IRQ_TYPE_LEVEL_LOW and IRQ_TYPE_EDGE_FALLING for those. We just need to
check the type, set the relevant bit in INTPCR accordingly, and fixup
the type argument before calling the GIC's irq_set_type.
In fact, the power-on-reset value of the INTPCR register on the LS1021A
is so that all six lines have their polarity inverted. Hence any
hardware connected to those lines is unusable without this: If the line
is indeed active low, the generic GIC code will reject an irq spec with
IRQ_TYPE_LEVEL_LOW, while if the line is active high, we must obviously
disable the polarity inversion (writing 0 to the relevant bit) before
unmasking the interrupt.
Some other Layerscape SOCs (LS1043A, LS1046A) have a similar feature,
just with a different number of external interrupt lines (and a
different POR value for the INTPCR register). This driver should be
prepared for supporting those by properly filling out the device tree
node. I have the reference manuals for all three boards, but I've only
tested the driver on an LS1021A.
Unfortunately, the Kconfig symbol ARCH_LAYERSCAPE only exists on
arm64, so do as is done for irq-ls-scfg-msi.c: introduce a new symbol
which is set when either ARCH_LAYERSCAPE or SOC_LS1021A is set.
Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20191107122115.6244-3-linux@rasmusvillemoes.dk
On some specific chips like 7211 we need to leave some interrupts
untouched/forwarded to the VPU which is another agent in the system
making use of that interrupt controller hardware (goes to both ARM GIC
and VPU L1 interrupt controller). Make that possible by using the
existing brcm,int-fwd-mask property and take necessary actions to avoid
masking that interrupt as well as not allowing Linux to map them.
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20191024201415.23454-6-f.fainelli@gmail.com
Many of the privileged CSRs exist in a supervisor and machine version
that are used very similarly. Provide versions of the CSR names and
fields that map to either the S-mode or M-mode variant depending on
a new CONFIG_RISCV_M_MODE kconfig symbol.
Contains contributions from Damien Le Moal <Damien.LeMoal@wdc.com>
and Paul Walmsley <paul.walmsley@sifive.com>.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Thomas Gleixner <tglx@linutronix.de> # for drivers/clocksource, drivers/irqchip
[paul.walmsley@sifive.com: updated to apply]
Signed-off-by: Paul Walmsley <paul.walmsley@sifive.com>
When the VHE code was reworked, a lot of the vgic stuff was moved around,
but the GICv4 residency code did stay untouched, meaning that we come
in and out of residency on each flush/sync, which is obviously suboptimal.
To address this, let's move things around a bit:
- Residency entry (flush) moves to vcpu_load
- Residency exit (sync) moves to vcpu_put
- On blocking (entry to WFI), we "put"
- On unblocking (exit from WFI), we "load"
Because these can nest (load/block/put/load/unblock/put, for example),
we now have per-VPE tracking of the residency state.
Additionally, vgic_v4_put gains a "need doorbell" parameter, which only
gets set to true when blocking because of a WFI. This allows a finer
control of the doorbell, which now also gets disabled as soon as
it gets signaled.
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20191027144234.8395-2-maz@kernel.org
Pull the second lot of irqchip updates for 5.4 from Marc Zyngier:
- Sifive PLIC: force driver to skip non-relevant contexts
- GICv4: Don't send VMOVP commands to ITSs that don't have
this vPE mapped
Modify plic_init() to skip .dts interrupt contexts other
than supervisor external interrupt.
The .dts entry for plic may specify multiple interrupt contexts.
For example, it may assign two entries IRQ_M_EXT and IRQ_S_EXT,
in that order, to the same interrupt controller. This patch
modifies plic_init() to skip the IRQ_M_EXT context since
IRQ_S_EXT is currently the only supported context.
If IRQ_M_EXT is not skipped, plic_init() will report "handler
already present for context" when it comes across the IRQ_S_EXT
context in the next iteration of its loop.
Without this patch, .dts would have to be edited to replace the
value of IRQ_M_EXT with -1 for it to be skipped.
Signed-off-by: Alan Mikhak <alan.mikhak@sifive.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Paul Walmsley <paul.walmsley@sifive.com> # arch/riscv
Link: https://lkml.kernel.org/r/1571933503-21504-1-git-send-email-alan.mikhak@sifive.com
On a system without Single VMOVP support (say GITS_TYPER.VMOVP == 0),
we will map vPEs only on ITSs that will actually control interrupts
for the given VM. And when moving a vPE, the VMOVP command will be
issued only for those ITSs.
But when issuing VMOVPs we seemed fail to present the exact ITSList
to ITSs who are actually included in the synchronization operation.
The its_list_map we're currently using includes all ITSs in the system,
even though some of them don't have the corresponding vPE mapping at all.
Introduce get_its_list() to get the per-VM its_list_map, to indicate
which ITSs have vPE mappings for the given VM, and use this map as
the expected ITSList when building VMOVP. This is hopefully a performance
gain not to do some synchronization with those unsuspecting ITSs.
And initialize the whole command descriptor to zero at beginning, since
the seq_num and its_list should be RES0 when GITS_TYPER.VMOVP == 1.
Signed-off-by: Zenghui Yu <yuzenghui@huawei.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/1571802386-2680-1-git-send-email-yuzenghui@huawei.com
The GICv3 architecture specification is incredibly misleading when it
comes to PMR and the requirement for a DSB. It turns out that this DSB
is only required if the CPU interface sends an Upstream Control
message to the redistributor in order to update the RD's view of PMR.
This message is only sent when ICC_CTLR_EL1.PMHE is set, which isn't
the case in Linux. It can still be set from EL3, so some special care
is required. But the upshot is that in the (hopefuly large) majority
of the cases, we can drop the DSB altogether.
This relies on a new static key being set if the boot CPU has PMHE
set. The drawback is that this static key has to be exported to
modules.
Cc: Will Deacon <will@kernel.org>
Cc: James Morse <james.morse@arm.com>
Cc: Julien Thierry <julien.thierry.kdev@gmail.com>
Cc: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Pull irqchip fixes from Marc Zyngier:
- Add retrigger support to Amazon's al-fic driver
- Add SAM9X60 support to Atmel's AIC5 irqchip
- Fix GICv3 maximum interrupt calculation
- Convert SiFive's PLIC to the fasteoi IRQ flow
Pull MIPS updates from Paul Burton:
"Main MIPS changes:
- boot_mem_map is removed, providing a nice cleanup made possible by
the recent removal of bootmem.
- Some fixes to atomics, in general providing compiler barriers for
smp_mb__{before,after}_atomic plus fixes specific to Loongson CPUs
or MIPS32 systems using cmpxchg64().
- Conversion to the new generic VDSO infrastructure courtesy of
Vincenzo Frascino.
- Removal of undefined behavior in set_io_port_base(), fixing the
behavior of some MIPS kernel configurations when built with recent
clang versions.
- Initial MIPS32 huge page support, functional on at least Ingenic
SoCs.
- pte_special() is now supported for some configurations, allowing
among other things generic fast GUP to be used.
- Miscellaneous fixes & cleanups.
And platform specific changes:
- Major improvements to Ingenic SoC support from Paul Cercueil,
mostly enabled by the inclusion of the new TCU (timer-counter unit)
drivers he's spent a very patient year or so working on. Plus some
fixes for X1000 SoCs from Zhou Yanjie.
- Netgear R6200 v1 systems are now supported by the bcm47xx platform.
- DT updates for BMIPS, Lantiq & Microsemi Ocelot systems"
* tag 'mips_5.4' of git://git.kernel.org/pub/scm/linux/kernel/git/mips/linux: (89 commits)
MIPS: Detect bad _PFN_SHIFT values
MIPS: Disable pte_special() for MIPS32 with RiXi
MIPS: ralink: deactivate PCI support for SOC_MT7621
mips: compat: vdso: Use legacy syscalls as fallback
MIPS: Drop Loongson _CACHE_* definitions
MIPS: tlbex: Remove cpu_has_local_ebase
MIPS: tlbex: Simplify r3k check
MIPS: Select R3k-style TLB in Kconfig
MIPS: PCI: refactor ioc3 special handling
mips: remove ioremap_cachable
mips/atomic: Fix smp_mb__{before,after}_atomic()
mips/atomic: Fix loongson_llsc_mb() wreckage
mips/atomic: Fix cmpxchg64 barriers
MIPS: Octeon: remove duplicated include from dma-octeon.c
firmware: bcm47xx_nvram: Allow COMPILE_TEST
firmware: bcm47xx_nvram: Correct size_t printf format
MIPS: Treat Loongson Extensions as ASEs
MIPS: Remove dev_err() usage after platform_get_irq()
MIPS: dts: mscc: describe the PTP ready interrupt
MIPS: dts: mscc: describe the PTP register range
...
As per GIC spec, ITLinesNumber indicates the maximum SPI INTID that
the GIC implementation supports. And the maximum SPI INTID an
implementation might support is 1019 (field value 11111).
max(GICD_TYPER_SPIS(...), 1020) is not what we actually want for
GIC_LINE_NR. Fix it to min(GICD_TYPER_SPIS(...), 1020).
Signed-off-by: Zenghui Yu <yuzenghui@huawei.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/1568789850-14080-1-git-send-email-yuzenghui@huawei.com
Pull core irq updates from Thomas Gleixner:
"Updates from the irq departement:
- Update the interrupt spreading code so it handles numa node with
different CPU counts properly.
- A large overhaul of the ARM GiCv3 driver to support new PPI and SPI
ranges.
- Conversion of all alloc_fwnode() users to use physical addresses
instead of virtual addresses so the virtual addresses are not
leaked. The physical address is sufficient to identify the
associated interrupt chip.
- Add support for Marvel MMP3, Amlogic Meson SM1 interrupt chips.
- Enforce interrupt threading at compile time if RT is enabled.
- Small updates and improvements all over the place"
* 'irq-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (37 commits)
irqchip/gic-v3-its: Fix LPI release for Multi-MSI devices
irqchip/uniphier-aidet: Use devm_platform_ioremap_resource()
irqdomain: Add the missing assignment of domain->fwnode for named fwnode
irqchip/mmp: Coexist with GIC root IRQ controller
irqchip/mmp: Mask off interrupts from other cores
irqchip/mmp: Add missing chained_irq_{enter,exit}()
irqchip/mmp: Do not use of_address_to_resource() to get mux regs
irqchip/meson-gpio: Add support for meson sm1 SoCs
dt-bindings: interrupt-controller: New binding for the meson sm1 SoCs
genirq/affinity: Remove const qualifier from node_to_cpumask argument
genirq/affinity: Spread vectors on node according to nr_cpu ratio
genirq/affinity: Improve __irq_build_affinity_masks()
irqchip: Remove dev_err() usage after platform_get_irq()
irqchip: Add include guard to irq-partition-percpu.h
irqchip/mmp: Do not call irq_set_default_host() on DT platforms
irqchip/gic-v3-its: Remove the redundant set_bit for lpi_map
irqchip/gic-v3: Add quirks for HIP06/07 invalid GICD_TYPER erratum 161010803
irqchip/gic: Skip DT quirks when evaluating IIDR-based quirks
irqchip/gic-v3: Warn about inconsistent implementations of extended ranges
irqchip/gic-v3: Add EPPI range support
...
When allocating a range of LPIs for a Multi-MSI capable device,
this allocation extended to the closest power of 2.
But on the release path, the interrupts are released one by
one. This results in not releasing the "extra" range, leaking
the its_device. Trying to reprobe the device will then fail.
Fix it by releasing the LPIs the same way we allocate them.
Fixes: 8208d1708b ("irqchip/gic-v3-its: Align PCI Multi-MSI allocation on their size")
Reported-by: Jiaxing Luo <luojiaxing@huawei.com>
Tested-by: John Garry <john.garry@huawei.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/f5e948aa-e32f-3f74-ae30-31fee06c2a74@huawei.com
When running in M-mode, the S-mode plic handlers are still listed in the
device tree. Ignore them by setting the maximum threshold.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Marc Zyngier <maz@kernel.org>
Signed-off-by: Paul Walmsley <paul.walmsley@sifive.com>
On mmp3, there's an extra set of ICU registers (ICU2) that handle
interrupts on the extra cores. When masking off interrupts on MP1,
these should be masked as well.
We add a new interrupt controller via device tree to identify when we're
looking at an mmp3 machine via compatible field of "marvell,mmp3-intc".
[lkundrak@v3.sk: Changed "mrvl,mmp3-intc" compatible strings to
"marvell,mmp3-intc". Tidied up the subject line a bit.]
Signed-off-by: Andres Salomon <dilinger@queued.net>
Signed-off-by: Lubomir Rintel <lkundrak@v3.sk>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20190822092643.593488-9-lkundrak@v3.sk
--
Changes since v1:
- Moved mmp3-specific mmp_icu2_base initialization from mmp_init_bases() to
mmp3_of_init() so that we don't have to check for marvell,mmp3-intc
compatibility twice.
- Drop an superfluous call to irq_set_default_host()
arch/arm/mach-mmp/regs-icu.h | 3 +++
drivers/irqchip/irq-mmp.c | 48 ++++++++++++++++++++++++++++++++++++
2 files changed, 51 insertions(+)
Signed-off-by: Andres Salomon <dilinger@queued.net>
Signed-off-by: Lubomir Rintel <lkundrak@v3.sk>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20190822092643.593488-9-lkundrak@v3.sk
The "regs" property of the "mrvl,mmp2-mux-intc" devices are silly. They
are offsets from intc's base, not addresses on the parent bus. At this
point it probably can't be fixed.
On an OLPC XO-1.75 machine, the muxes are children of the intc, not the
axi bus, and thus of_address_to_resource() won't work. We should treat
the values as mere integers as opposed to bus addresses.
Signed-off-by: Lubomir Rintel <lkundrak@v3.sk>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Acked-by: Pavel Machek <pavel@ucw.cz>
Link: https://lore.kernel.org/r/20190822092643.593488-7-lkundrak@v3.sk
The meson sm1 SoCs uses the same type of GPIO interrupt controller IP
block as the other meson SoCs, A total of 100 pins can be spied on:
- 223:100 undefined (no interrupt)
- 99:97 3 pins on bank GPIOE
- 96:77 20 pins on bank GPIOX
- 76:61 16 pins on bank GPIOA
- 60:53 8 pins on bank GPIOC
- 52:37 16 pins on bank BOOT
- 36:28 9 pins on bank GPIOH
- 27:12 16 pins on bank GPIOZ
- 11:0 12 pins in the AO domain
Mapping is the same as the g12a family but the sm1 controller
allows to trig an irq on both edges of the input signal. This was
not possible with the previous SoCs families
Signed-off-by: Jerome Brunet <jbrunet@baylibre.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Tested-by: Kevin Hilman <khilman@baylibre.com>
Reviewed-by: Kevin Hilman <khilman@baylibre.com>
Link: https://lore.kernel.org/r/20190829161635.25067-3-jbrunet@baylibre.com
We don't need dev_err() messages when platform_get_irq() fails now that
platform_get_irq() prints an error message itself when something goes
wrong. Let's remove these prints with a simple semantic patch.
// <smpl>
@@
expression ret;
struct platform_device *E;
@@
ret =
(
platform_get_irq(E, ...)
|
platform_get_irq_byname(E, ...)
);
if ( \( ret < 0 \| ret <= 0 \) )
{
(
-if (ret != -EPROBE_DEFER)
-{ ...
-dev_err(...);
-... }
|
...
-dev_err(...);
)
...
}
// </smpl>
While we're here, remove braces on if statements that only have one
statement (manually).
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Jason Cooper <jason@lakedaemon.net>
Cc: Marc Zyngier <marc.zyngier@arm.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Using a default domain on DT platforms is unnecessary, as the firmware
tables describe the full topology, and nothing is implicit.
Signed-off-by: Lubomir Rintel <lkundrak@v3.sk>
[maz: wrote an actual changelog]
Signed-off-by: Marc Zyngier <maz@kernel.org>
