In the refactoring of the coherency fabric assembly code, a function
called ll_get_cpuid() was created to factorize common logic between
functions adding CPU to the SMP coherency group, enabling and
disabling the coherency.
However, the name of the function is highly misleading: ll_get_cpuid()
makes one think tat it returns the ID of the CPU, i.e 0 for CPU0, 1
for CPU1, etc. In fact, this is not at all what this function returns:
it returns a CPU mask for the current CPU, usable for the coherency
fabric configuration and control registers.
Therefore this commit renames this function to
ll_get_coherency_cpumask(), and adds additional comments on top of the
function to explain in more details what it does, and also how the
endianess issue is handled.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Link: https://lkml.kernel.org/r/1400762882-10116-5-git-send-email-thomas.petazzoni@free-electrons.com
Acked-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
The Marvell Armada 375 and Armada 38x SOCs, which use the Cortex-A9
CPU core, the PL310 cache and the Marvell PCIe hardware block are
affected a L2/PCIe deadlock caused by a system erratum when hardware
I/O coherency is used.
This deadlock can be avoided by mapping the PCIe memory areas as
strongly-ordered (note: MT_UNCACHED is strongly-ordered), and by
removing the outer cache sync done in software. This is implemented in
this patch by:
* Registering a custom arch_ioremap_caller function that allows to
make sure PCI memory regions are mapped MT_UNCACHED.
* Adding at runtime the 'arm,io-coherent' property to the PL310 cache
controller. This cannot be done permanently in the DT, because the
hardware I/O coherency can only be enabled when CONFIG_SMP is
enabled, in the current kernel situation.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Link: https://lkml.kernel.org/r/1400165974-9059-4-git-send-email-thomas.petazzoni@free-electrons.com
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
Since the beginning of the introduction of hardware I/O coherency
support for Armada 370 and Armada XP, the special DMA operations
should have applied to all DMA capable devices. Unfortunately, while
the original code properly took into account platform devices, it
didn't take into account PCI devices, which can also be DMA masters.
This commit fixes that by registering a bus notifier on pci_bus_type,
to register our custom DMA operations, like is already done for
platform devices. While doing this, we also rename
mvebu_hwcc_platform_notifier() to mvebu_hwcc_notifier() and
mvebu_hwcc_platform_nb to mvebu_hwcc_nb because they are no longer
specific to platform devices.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Link: https://lkml.kernel.org/r/1399997070-11434-1-git-send-email-thomas.petazzoni@free-electrons.com
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
When compiling for multiplatform for both ARMv6 and ARMv7, the default compiler
flags are for ARMv6, and we will get:
/tmp/ccwDEzd0.s: Assembler messages:
/tmp/ccwDEzd0.s:639: Error: selected processor does not support ARM mode `isb '
/tmp/ccwDEzd0.s:645: Error: selected processor does not support ARM mode `isb '
/tmp/ccwDEzd0.s:646: Error: selected processor does not support ARM mode `dsb '
/tmp/ccwDEzd0.s:695: Error: selected processor does not support ARM mode `isb '
make[1]: *** [arch/arm/mach-mvebu/pmsu.o] Error 1
Fix this in a similar manner than done previously in commit
72533b77d3, by specifying ARMv7 flags for pmsu.o.
Signed-off-by: Vincent Stehlé <vincent.stehle@laposte.net>
Link: https://lkml.kernel.org/r/1399407782-29091-1-git-send-email-vincent.stehle@laposte.net
Cc: Jason Cooper <jason@lakedaemon.net>
Cc: Andrew Lunn <andrew@lunn.ch>
Cc: Gregory Clement <gregory.clement@free-electrons.com>
Cc: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
Cc: Russell King <linux@arm.linux.org.uk>
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
The Armada 375 SMP workaround only needs to be applied to the Z1
revision of the SoC. The A0 and later revisions have been fixed, and
no longer need this workaround.
Note that the initialization of the SMP workaround is delayed from
->smp_prepare_cpus() to ->smp_boot_secondary() because when
->smp_prepare_cpus() is called, the early initcalls have not be
called, so the mvebu-soc-id mechanism is not operational. Since the
workaround is anyway not needed before the secondary CPU is started,
we can delay its implementation until the ->smp_boot_secondary() call.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Link: https://lkml.kernel.org/r/1399302326-6917-5-git-send-email-thomas.petazzoni@free-electrons.com
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
Currently, the mvebu-soc-id logic is initialized through a
core_initcall(). However, we will soon need to know the SoC revision
before booting secondary CPUs, because a workaround affects Armada 375
Z1 steppings, but should not be applied on Armada 375 A0 steppings.
Unfortunately, core_initcall() are called way too late compared to the
SMP initialization. Therefore, the mvebu-soc-id initialization is move
to an early_initcall(), which is called before the SMP initialization.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Link: https://lkml.kernel.org/r/1399302326-6917-3-git-send-email-thomas.petazzoni@free-electrons.com
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
In commit 54fe26a900bc528f3df1e4235cb6b9ca5c6d4dc2 ('ARM: mvebu: Add
thermal quirk for the Armada 375 DB board'), a check on the Armada SoC
revision was added to decide whether a quirk for the thermal device
should be applied or not.
However, the quirk implementation has a bug: it assumes
mvebu_get_soc_id() returns true on success, but it returns
0. Therefore, the condition:
if (mvebu_get_soc_id(&dev, &rev) && rev > ARMADA_375_Z1_REV)
is always false (as long as mvebu-soc-id is properly initialized). As
a consequence, the quirk is always applied, even on A0 steppings, for
which the quirk should not be applied.
This was spotted by testing the thermal driver on Armada 375 A0, which
Ezequiel could not do since he does not have access to the A0 revision
of the SoC for the moment.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Link: https://lkml.kernel.org/r/1399302326-6917-2-git-send-email-thomas.petazzoni@free-electrons.com
Fixes: 54fe26a900bc528f3df1e4235cb6b9ca5c6d4dc2 ('ARM: mvebu: Add thermal quirk for the Armada 375 DB board')
Acked-by: Ezequiel Garcia <ezequiel.garcia@free-electrons.com>
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
The initial release of the Armada 375 DB board has an Armada 375
Z1 stepping silicon. This commit introduces a quirk that allows
to workaround a series of issues with the thermal sensor in this
stepping, but updating the devicetree:
* Updates the compatible string for the thermal, so the driver
can perform a specific initialization of the sensor.
* Moves the offset of the thermal control register. This quirk
allows to specifiy the correct (A0 stepping) offset in the
devicetree.
Signed-off-by: Ezequiel Garcia <ezequiel.garcia@free-electrons.com>
Link: https://lkml.kernel.org/r/1398371004-15807-9-git-send-email-ezequiel.garcia@free-electrons.com
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
Actually enabling coherency and adding a CPU on a SMP group are two
different operations which can be done separately. This patch splits
this in two functions.
Moreover as they use common pattern, this patch also creates local low
level functions (ll_get_coherency_base and ll_get_cpuid) to be used by
the exposed functions (ll_add_cpu_to_smp_group and
ll_enable_coherency)
Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Link: https://lkml.kernel.org/r/1397488214-20685-6-git-send-email-gregory.clement@free-electrons.com
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
In order to be able to deal with the MMU enabled and the MMU disabled
cases, the base address of the coherency registers was passed to the
function. The address by itself was not interesting as it can't change
for a given SoC, the only thing we need is to have a distinction
between the physical or the virtual address.
This patch add a check of the MMU bit to choose the accurate address,
then the calling function doesn't have to pass this information.
Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Link: https://lkml.kernel.org/r/1397488214-20685-3-git-send-email-gregory.clement@free-electrons.com
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
The pmsu.c driver contained an armada_xp_boot_cpu() function that sets
the boot address of a secondary CPUs and deasserts the reset. However,
the Armada 375 needs a slightly different logic, so it makes more
sense to move this code into the Armada XP specific platsmp.c.
In order to achieve this, the mvebu_pmsu_set_cpu_boot_addr() function
is exported. It will be needed for both the Armada XP and Armada 38x
SMP implementations.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Link: https://lkml.kernel.org/r/1397483648-26611-2-git-send-email-thomas.petazzoni@free-electrons.com
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
Non-DT irq handlers were working through irq causes from most-significant
to least-significant bit, while DT irqchip driver does it the other way
round. This revealed some more HW issues on Kirkwood peripheral IP, where
spurious sdio irqs can happen although irqs are masked.
Also, the generated binaries show that original non-DT order compared
to DT order save two instructions for each bit count check:
irqchip DT order with ffs():
60: e3a06001 mov r6, #1
64: e2643000 rsb r3, r4, #0
68: e0033004 and r3, r3, r4
6c: e16f3f13 clz r3, r3
70: e263301f rsb r3, r3, #31
74: e1c44316 bic r4, r4, r6, lsl r3
78: e5971004 ldr r1, [r7, #4]
Original non-DT order with fls():
60: e3a07001 mov r7, #1
64: e16f3f14 clz r3, r4
68: e263301f rsb r3, r3, #31
6c: e1c44317 bic r4, r4, r7, lsl r3
70: e5951004 ldr r1, [r5, #4]
Therefore, reverse irq bit handling back to original order by replacing
ffs() with fls().
Signed-off-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
Link: https://lkml.kernel.org/r/1398719528-23607-1-git-send-email-sebastian.hesselbarth@gmail.com
Acked-by: Jason Cooper <jason@lakedaemon.net>
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
Some irqchip initialization must be done on secondary CPUs. On mvebu
platforms, this is currently achieved by having the
arch/arm/mach-mvebu/platsmp.c code directly call into a function
exported by the irqchip driver, which isn't really nice.
This commit changes this by using the same solution as the one used in
the GIC driver: the irqchip driver registers a CPU notifier, which is
used to do the secondary CPU IRQ initialization. This way, the irqchip
driver is completely autonomous, and the function no longer needs to
be exposed from the irqchip driver to the SoC code.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Link: https://lkml.kernel.org/r/1397483648-26611-6-git-send-email-thomas.petazzoni@free-electrons.com
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
This commit changes the PMSU driver to no longer map itself the CPU
reset registers, and instead call into the CPU reset driver to
deassert the secondary CPUs for SMP booting.
In order to provide Device Tree backward compatibility, the CPU reset
driver is extended to not only support its official compatible string
"marvell,armada-370-cpu-reset", but to also look at the PMSU
compatible string "marvell,armada-370-xp-pmsu" to find the CPU reset
registers address. This allows old Device Tree to work correctly with
newer kernel versions. Therefore, the CPU reset driver implements the
following logic:
* If one of the normal compatible strings
"marvell,armada-370-cpu-reset" is found, then we map its first
memory resource as the CPU reset registers.
* Otherwise, if none of the normal compatible strings have been
found, we look for the "marvell,armada-370-xp-pmsu" compatible
string, and we map the second memory as the CPU reset registers.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Link: https://lkml.kernel.org/r/1397483433-25836-3-git-send-email-thomas.petazzoni@free-electrons.com
Acked-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
The Armada 370 and Armada XP have registers that allow to reset the
CPUs, which is particularly useful to take the secondary CPUs out of
reset in the context of the SMP support.
Unfortunately, an implementation mistake was originally made and the
support for these registers was integrated into the PMSU driver, which
is in fact completely unrelated. And it turns out that the Armada 375
has the same CPU reset registers, but does not have the PMSU
registers.
Therefore, this commit creates a small CPU reset driver. All it does
is provide a simple mvebu_cpu_reset_deassert() function that the SMP
support code can call to take secondary CPUs out of reset. As of this
commit, the driver isn't being used, it will be used through changes
in the following commits.
Note that we initially planned to use the 'reset controller'
framework, but it requires the addition of "resets" properties in the
Device Tree, which are causing too many problems if we want to keep
the Device Tree backward compatibility. Moreover, the 'reset
controller' framework is mainly useful when a device driver needs to
request a reset of its device from a separate reset controller. In our
case, the CPU reset handling and the SMP core code are both located in
arch/arm/mach-mvebu/ and are tightly linked together, so there's no
real benefit in going through a separate framework.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Link: https://lkml.kernel.org/r/1397483433-25836-2-git-send-email-thomas.petazzoni@free-electrons.com
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
The Armada 375, like the Armada 370 and Armada XP, has a coherency
unit. However, unlike the coherency unit of 370/XP which does both CPU
and I/O coherency, the one on Armada 735 only does I/O
coherency. Therefore, instead of having two sets of registers (the
first one being used mainly to register each CPU in the coherency
fabric, the second one being used for the I/O coherency barrier), it
has only one set of register (for the I/O coherency barrier).
This commit adds a new "marvell,armada-375-coherency-fabric"
compatible string for this variant of the coherency fabric. The custom
DMA operations, and the way of triggering an I/O barrier is the same
as Armada 370/XP, so the code changes are minimal. However, the
set_cpu_coherent() function is not needed on Armada 375 and will not
work.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Link: https://lkml.kernel.org/r/1397483228-25625-7-git-send-email-thomas.petazzoni@free-electrons.com
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
Contrary to the Armada 370 and XP that used the PJ4B Marvell cores,
the Armada 375 and Armada 38x use the ARM Cortex-A9. A consequence of
this is that the unit responsible for the coherency between CPUs is
now the ARM SCU, and not the Marvell coherency unit (which is still
present to do coherency with I/O devices).
Therefore this commit:
* Ensures that the selection of the Armada 375 or Armada 38x SoC
support enables the ARM SCU support in the kernel.
* Make sure to initialize the SCU at boot time.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Link: https://lkml.kernel.org/r/1397483228-25625-6-git-send-email-thomas.petazzoni@free-electrons.com
Signed-off-by: Jason Cooper <jason@lakedaemon.net>
