Am 07.08.2017 um 09:47 schrieb Phil Elwell:
> Hi Stefan,
>
> On 06/08/2017 23:32, Stefan Wahren wrote:
>> In order to easily provide SMP for BCM2837 on 32-bit and 64-bit
>> the SMP startup code was placed in irq-bcm2836. That's not the
>> right approach. So move this code where it belongs.
>>
>> Signed-off-by: Stefan Wahren <stefan.wahren@xxxxxxxx>
>> Fixes: 41f4988cc287 ("irqchip/bcm2836: Add SMP support for the 2836")
>> ---
>> arch/arm/mach-bcm/Makefile | 3 ++
>> arch/arm/mach-bcm/platsmp.c | 35 ++++++++++++++++
>> drivers/irqchip/irq-bcm2836.c | 79 +------------------------------------
>> include/linux/irqchip/irq-bcm2836.h | 70 ++++++++++++++++++++++++++++++++
>> 4 files changed, 110 insertions(+), 77 deletions(-)
>> create mode 100644 include/linux/irqchip/irq-bcm2836.h
>>
>> diff --git a/arch/arm/mach-bcm/Makefile b/arch/arm/mach-bcm/Makefile
>> index 980f585..637cd9e 100644
>> --- a/arch/arm/mach-bcm/Makefile
>> +++ b/arch/arm/mach-bcm/Makefile
>> @@ -43,6 +43,9 @@ endif
>>
>> # BCM2835
>> obj-$(CONFIG_ARCH_BCM2835) += board_bcm2835.o
>> +ifeq ($(CONFIG_ARCH_BCM2835),y)
>> +obj-$(CONFIG_SMP) += platsmp.o
>> +endif
>>
>> # BCM5301X
>> obj-$(CONFIG_ARCH_BCM_5301X) += bcm_5301x.o
>> diff --git a/arch/arm/mach-bcm/platsmp.c b/arch/arm/mach-bcm/platsmp.c
>> index 9e3f275..3fa1a99 100644
>> --- a/arch/arm/mach-bcm/platsmp.c
>> +++ b/arch/arm/mach-bcm/platsmp.c
>> @@ -17,6 +17,7 @@
>> #include <linux/errno.h>
>> #include <linux/init.h>
>> #include <linux/io.h>
>> +#include <linux/irqchip/irq-bcm2836.h>
>> #include <linux/jiffies.h>
>> #include <linux/of.h>
>> #include <linux/of_address.h>
>> @@ -287,6 +288,35 @@ static int nsp_boot_secondary(unsigned int cpu, struct task_struct *idle)
>> return ret;
>> }
>>
>> +static int bcm2836_boot_secondary(unsigned int cpu, struct task_struct *idle)
>> +{
>> + void __iomem *intc_base;
>> + struct device_node *dn;
>> + char *name;
>> +
>> + name = "brcm,bcm2836-l1-intc";
>> + dn = of_find_compatible_node(NULL, NULL, name);
>> + if (!dn) {
>> + pr_err("unable to find intc node\n");
>> + return -ENODEV;
>> + }
>> +
>> + intc_base = of_iomap(dn, 0);
>> + of_node_put(dn);
>> +
>> + if (!intc_base) {
>> + pr_err("unable to remap intc base register\n");
>> + return -ENOMEM;
>> + }
>> +
>> + writel(virt_to_phys(secondary_startup),
>> + intc_base + LOCAL_MAILBOX3_SET0 + 16 * cpu);
> I was going to upstream the patch to add dsb+sev today, but now the goalposts are
> (literally) moving. Should I produce two patchsets - one for before your patch is
> appied, and one for after - or will you add the required lines here?:
>
> + dsb(sy); /* Ensure write has completed before waking the other CPUs */
> + sev();
> +
Since Marc make it clear that he won't accept any patches before this
cleanup, please base your patch on top of this repo:
https://github.com/anholt/linux/commits/bcm2835-soc-smp-cleanup
Btw Eric reported problems [1] about this change, is this fixed now?
[1] -
https://github.com/raspberrypi/linux/issues/1989#issuecomment-318180597
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
