This patch adds watchdog support by installing shmobile_boot_vector_gen2
to ICRAM1 when enough memory is available, in which case we also keep a
copy of MPIDR to complete the reset vector logic.
Signed-off-by: Fabrizio Castro <fabrizio.castro@bp.renesas.com>
Signed-off-by: Ramesh Shanmugasundaram <ramesh.shanmugasundaram@bp.renesas.com>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
On Cortex-A7, the arch timer CNTVOFF register is uninitialized.
Ideally it should be initialized by the boot loader, but it isn't.
For the boot CPU, CNTVOFF is initialized by Linux since commit
9ce3fa6816 ("ARM: shmobile: rcar-gen2: Add CA7 arch_timer
initialization for r8a7794").
For secondary CPU cores, no such initialization is done.
Hence when enabling SMP on r8a7794, the kernel log is spammed with:
WARNING: Underflow in clocksource 'arch_sys_counter' observed, time update ignored.
Please report this, consider using a different clocksource, if possible.
Your kernel is probably still fine.
As Marc Zyngier pointed out that Cortex-A15 and Cortex-A7 are similar with
respect to CNTVOFF, we have been very lucky this just worked on R-Car
Gen2 SoCs with Cortex-A15 cores.
To fix this:
- Move the existing inline asm code to initialize CNTVOFF to an
assembler source file (adding comments and replacing hardcoded
constants by definitions in the process), so it can be reused,
- Perform the initialization of CNTVOFF on the boot CPU (Cortex-A15 or
Cortex-A7) on all R-Car Gen2 and RZ/G1 parts,
- Wrap the standard secondary_startup() routine inside a routine which
initializes CNTVOFF.
Based on patches by Hisashi Nakamura in the BSP.
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Tested-by: Fabrizio Castro <fabrizio.castro@bp.renesas.com>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Pull ARM updates from Russell King:
- nommu updates from Afzal Mohammed cleaning up the vectors support
- allow DMA memory "mapping" for nommu Benjamin Gaignard
- fixing a correctness issue with R_ARM_PREL31 relocations in the
module linker
- add strlen() prototype for the decompressor
- support for DEBUG_VIRTUAL from Florian Fainelli
- adjusting memory bounds after memory reservations have been
registered
- unipher cache handling updates from Masahiro Yamada
- initrd and Thumb Kconfig cleanups
* 'for-linus' of git://git.armlinux.org.uk/~rmk/linux-arm: (23 commits)
ARM: mm: round the initrd reservation to page boundaries
ARM: mm: clean up initrd initialisation
ARM: mm: move initrd init code out of arm_memblock_init()
ARM: 8655/1: improve NOMMU definition of pgprot_*()
ARM: 8654/1: decompressor: add strlen prototype
ARM: 8652/1: cache-uniphier: clean up active way setup code
ARM: 8651/1: cache-uniphier: include <linux/errno.h> instead of <linux/types.h>
ARM: 8650/1: module: handle negative R_ARM_PREL31 addends correctly
ARM: 8649/2: nommu: remove Hivecs configuration is asm
ARM: 8648/2: nommu: display vectors base
ARM: 8647/2: nommu: dynamic exception base address setting
ARM: 8646/1: mmu: decouple VECTORS_BASE from Kconfig
ARM: 8644/1: Reduce "CPU: shutdown" message to debug level
ARM: 8641/1: treewide: Replace uses of virt_to_phys with __pa_symbol
ARM: 8640/1: Add support for CONFIG_DEBUG_VIRTUAL
ARM: 8639/1: Define KERNEL_START and KERNEL_END
ARM: 8638/1: mtd: lart: Rename partition defines to be prefixed with PART_
ARM: 8637/1: Adjust memory boundaries after reservations
ARM: 8636/1: Cleanup sanity_check_meminfo
ARM: add CPU_THUMB_CAPABLE to indicate possible Thumb support
...
All low-level PM/SMP code using virt_to_phys() should actually use
__pa_symbol() against kernel symbols. Update code where relevant to move
away from virt_to_phys().
Acked-by: Russell King <rmk+kernel@armlinux.org.uk>
Reviewed-by: Laura Abbott <labbott@redhat.com>
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Now debug resource reset is handled properly, allow booting secondary
CPU cores when hardware debug mode is enabled (MD21=1) on SoCs using the
"renesas,apmu" enable method.
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Tested-by: Hiep Cao Minh <cm-hiep@jinso.co.jp>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
In debug mode (MD21=1), reset requests derived from power-shutoff to the
AP-system CPU cores must be enabled before the AP-system CPU cores
resume from power-shutoff for the first time. Else resume may fail,
causing the system to hang during boot.
As setting these bits is a no-op in normal mode, there's no need to
check the actual state of MD21 first.
Inspired by CPU-specific patches in the BSP by Hisashi Nakamura
<hisashi.nakamura.ak@renesas.com>.
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Tested-by: Hiep Cao Minh <cm-hiep@jinso.co.jp>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Allow DT configuration of the APMU hardware in the case when the APMU is
pointed out in the DTB via the enable-method. The ability to configure
the APMU via C code is still kept intact to prevent DTB breakage for older
SoCs that do not rely on the enable-method for SMP support.
Signed-off-by: Magnus Damm <damm+renesas@opensource.se>
[geert: Fix CONFIG_SMP=n build]
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
shmobile_smp_apmu_prepare_cpus() is used only if CONFIG_SMP=y.
Hence move the #ifdef to cover shmobile_smp_apmu_prepare_cpus() and all
functions only called by it (apmu_init_cpu() and apmu_parse_cfg()).
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Ulrich Hecht <ulrich.hecht+renesas@gmail.com>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
CPU boot configuration writes to shmobile_boot_arg, which is located in
the .text section, and thus should not be written to.
As of commit 1d33a354bb ("ARM: shmobile: Per-CPU SMP boot / sleep
code for SCU SoCs"), and ignoring accidental remainings,
shmobile_boot_arg is always set to MPIDR_HWID_BITMASK by C code.
Hence we can just hardcode this in the assembler code, and remove the
variable, and thus also remove the need to write to this variable.
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Acked-by: Nicolas Pitre <nico@linaro.org>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
CPU id can be negative, so it cannot be assigned to unsigned variable.
The problem has been detected using proposed semantic patch
scripts/coccinelle/tests/unsigned_lesser_than_zero.cocci [1].
[1]: http://permalink.gmane.org/gmane.linux.kernel/2038576
Signed-off-by: Andrzej Hajda <a.hajda@samsung.com>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
shmobile_smp_apmu_cpu_shutdown() is used inside platsmp-apmu.c only.
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
With shmobile_defconfig but SMP=n && SUSPEND=n, I get:
arch/arm/mach-shmobile/platsmp-apmu.c:49:12: warning: 'apmu_power_off' defined but not used [-Wunused-function]
arch/arm/mach-shmobile/platsmp-apmu.c:70:12: warning: 'apmu_wrap' defined but not used [-Wunused-function]
Annotate those functions like the functions around it.
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
All ARMv5 and older CPUs invalidate their caches in the early assembly
setup function, prior to enabling the MMU. This is because the L1
cache should not contain any data relevant to the execution of the
kernel at this point; all data should have been flushed out to memory.
This requirement should also be true for ARMv6 and ARMv7 CPUs - indeed,
these typically do not search their caches when caching is disabled (as
it needs to be when the MMU is disabled) so this change should be safe.
ARMv7 allows there to be CPUs which search their caches while caching is
disabled, and it's permitted that the cache is uninitialised at boot;
for these, the architecture reference manual requires that an
implementation specific code sequence is used immediately after reset
to ensure that the cache is placed into a sane state. Such
functionality is definitely outside the remit of the Linux kernel, and
must be done by the SoC's firmware before _any_ CPU gets to the Linux
kernel.
Changing the data cache clean+invalidate to a mere invalidate allows us
to get rid of a lot of platform specific hacks around this issue for
their secondary CPU bringup paths - some of which were buggy.
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Tested-by: Florian Fainelli <f.fainelli@gmail.com>
Tested-by: Heiko Stuebner <heiko@sntech.de>
Tested-by: Dinh Nguyen <dinguyen@opensource.altera.com>
Acked-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
Tested-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
Acked-by: Shawn Guo <shawn.guo@linaro.org>
Tested-by: Thierry Reding <treding@nvidia.com>
Acked-by: Thierry Reding <treding@nvidia.com>
Tested-by: Geert Uytterhoeven <geert+renesas@glider.be>
Tested-by: Michal Simek <michal.simek@xilinx.com>
Tested-by: Wei Xu <xuwei5@hisilicon.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
APMU resources are not common to all R-Car SoCs so don't share this data.
A subsequent patch will correct the CPU cores for the r8a7791.
Signed-off-by: Hisashi Nakamura <hisashi.nakamura.ak@renesas.com>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Patch d6d757c9a4 ("ARM: shmobile: APMU: Add Core-Standby-state for
Suspend to RAM") added both an inline wrapper for shmobile_smp_apmu_suspend_init
and an empty function in arch/arm/mach-shmobile/platsmp-apmu.c. We get a
build failure when both are present, so this patch removes the one in the .c
file and keeps the inline version.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
The function shmobile_smp_apmu_suspend_init() should
be put into the init section to not trigger section
mismatch warnings.
Signed-off-by: Magnus Damm <damm+renesas@opensource.se>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Adjust the APMU code to allow build when CONFIG_SMP=n.
Signed-off-by: Magnus Damm <damm+renesas@opensource.se>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
This patch add Core-Standby-state for Suspend to RAM.
Signed-off-by: Keita Kobayashi <keita.kobayashi.ym@renesas.com>
Acked-by: Magnus Damm <damm+renesas@opensource.se>
[horms+renesas@verge.net.au: rebase]
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Change location for common.h so it can be used as #include "common.h"
instead of the old style #include <mach/common.h>.
Signed-off-by: Magnus Damm <damm+renesas@opensource.se>
Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Use the %pr printk specifier to print resource variables. This fixes
warnings on platforms where resource_size_t has a different size than
int.
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Add information to the shared APMU code regarding
the APMU instance used to control the CA7 cores.
This can be used on r8a7790 and r8a73a4, but should
most likely be converted to DT in the future.
Signed-off-by: Magnus Damm <damm@opensource.se>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Extend the APMU code with a check to only allow boot
of CPU cores that sit in the same cluster as CPU0.
This makes it possible for people to use the r8a790
CA7 boot mode with CA7-cores only. The default CA15
boot mode will enable CA15 cores only. This is an
intentional software limitation to cope with lacking
scheduler support.
By removing this patch it is possible to run all 8 cores
in parallel, but this is not recommended without out of tree
scheduler modfications or custom user space code to control
the CPU affinitiy.
Signed-off-by: Magnus Damm <damm@opensource.se>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Introduce shared APMU SMP code for mach-shmobile. Both SMP boot up
and CPU Hotplug is supported. This version does not use DT but
if needed this will be added as an incremental feature patch.
The code is designed around CONFIG_NR_CPUS and should in theory support
any number of APMUs, however due to the current DT-less static design
only a single APMU is supported.
Signed-off-by: Magnus Damm <damm@opensource.se>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Fabrizio Castro