ARM SMP/GIC/LDS patches for Xilinx Zynq remoteproc

ARM SMP/GIC/LDS patches for Xilinx Zynq remoteproc

[Michal Simek]

Hi,

I am sending bunch of patches which I have made for a Xilinx Zynq remoteproc AMP solution +
some patches around.

Xilinx Zynq is dual core ARM Cortex A9. Remoteproc in new interface for controlling other cpus
is the system. I have created solution where on one ARM runs Linux and the second ARM
runs RTOS or any other standalone firmware. It is AMP between two ARM cores.

Communication is done via shared memory with kicking via IPI interrupts. 
Based on that I had to do some changes in the kernel which I am sending. 

Thanks for your review,
Michal

ARM SMP/GIC/LDS patches for Xilinx Zynq remoteproc

[Michal Simek]

Hi,

I am sending bunch of patches which I have made for a Xilinx Zynq remoteproc AMP solution +
some patches around.

Xilinx Zynq is dual core ARM Cortex A9. Remoteproc in new interface for controlling other cpus
is the system. I have created solution where on one ARM runs Linux and the second ARM
runs RTOS or any other standalone firmware. It is AMP between two ARM cores.

Communication is done via shared memory with kicking via IPI interrupts. 
Based on that I had to do some changes in the kernel which I am sending. 

Thanks for your review,
Michal

[RFC PATCH 1/8] ARM: gic: Support forcing cpumask for possible cpus in the system

[Michal Simek]

gic_set_affinity doesn't support force option. Force option
should work with all possible cpus in the system not only with
online cpus.

Signed-off-by: Michal Simek <monstr@monstr.eu>
---
 arch/arm/common/gic.c |    7 ++++++-
 1 files changed, 6 insertions(+), 1 deletions(-)

diff --git a/arch/arm/common/gic.c b/arch/arm/common/gic.c
index aa52699..4003cc2 100644
--- a/arch/arm/common/gic.c
+++ b/arch/arm/common/gic.c
@@ -235,9 +235,14 @@ static int gic_set_affinity(struct irq_data *d, const struct cpumask *mask_val,
 {
 	void __iomem *reg = gic_dist_base(d) + GIC_DIST_TARGET + (gic_irq(d) & ~3);
 	unsigned int shift = (gic_irq(d) % 4) * 8;
-	unsigned int cpu = cpumask_any_and(mask_val, cpu_online_mask);
+	unsigned int cpu;
 	u32 val, mask, bit;
 
+	if (force)
+		cpu = cpumask_any_and(mask_val, cpu_possible_mask);
+	else
+		cpu = cpumask_any_and(mask_val, cpu_online_mask);
+
 	if (cpu >= 8 || cpu >= nr_cpu_ids)
 		return -EINVAL;
 
-- 
1.7.0.4

[RFC PATCH 1/8] ARM: gic: Support forcing cpumask for possible cpus in the system

[Michal Simek]

gic_set_affinity doesn't support force option. Force option
should work with all possible cpus in the system not only with
online cpus.

Signed-off-by: Michal Simek <monstr@monstr.eu>
---
 arch/arm/common/gic.c |    7 ++++++-
 1 files changed, 6 insertions(+), 1 deletions(-)

diff --git a/arch/arm/common/gic.c b/arch/arm/common/gic.c
index aa52699..4003cc2 100644
--- a/arch/arm/common/gic.c
+++ b/arch/arm/common/gic.c
@@ -235,9 +235,14 @@ static int gic_set_affinity(struct irq_data *d, const struct cpumask *mask_val,
 {
 	void __iomem *reg = gic_dist_base(d) + GIC_DIST_TARGET + (gic_irq(d) & ~3);
 	unsigned int shift = (gic_irq(d) % 4) * 8;
-	unsigned int cpu = cpumask_any_and(mask_val, cpu_online_mask);
+	unsigned int cpu;
 	u32 val, mask, bit;
 
+	if (force)
+		cpu = cpumask_any_and(mask_val, cpu_possible_mask);
+	else
+		cpu = cpumask_any_and(mask_val, cpu_online_mask);
+
 	if (cpu >= 8 || cpu >= nr_cpu_ids)
 		return -EINVAL;
 
-- 
1.7.0.4

Re: [RFC PATCH 1/8] ARM: gic: Support forcing cpumask for possible cpus in the system

[Russell King - ARM Linux]

On Mon, Jun 18, 2012 at 01:30:04PM +0200, Michal Simek wrote:
> gic_set_affinity doesn't support force option. Force option
> should work with all possible cpus in the system not only with
> online cpus.

Neither does x86.  Why do you want this?  Offline CPUs are unable to
receive the interrupt in any case (and must not receive interrupts in
the bringup path until they are online.)

[RFC PATCH 1/8] ARM: gic: Support forcing cpumask for possible cpus in the system

[Russell King - ARM Linux]

On Mon, Jun 18, 2012 at 01:30:04PM +0200, Michal Simek wrote:
> gic_set_affinity doesn't support force option. Force option
> should work with all possible cpus in the system not only with
> online cpus.

Neither does x86.  Why do you want this?  Offline CPUs are unable to
receive the interrupt in any case (and must not receive interrupts in
the bringup path until they are online.)

Re: [RFC PATCH 1/8] ARM: gic: Support forcing cpumask for possible cpus in the system

[Michal Simek]

On 06/18/2012 01:33 PM, Russell King - ARM Linux wrote:
> On Mon, Jun 18, 2012 at 01:30:04PM +0200, Michal Simek wrote:
>> gic_set_affinity doesn't support force option. Force option
>> should work with all possible cpus in the system not only with
>> online cpus.
>
> Neither does x86.  Why do you want this?  Offline CPUs are unable to
> receive the interrupt in any case (and must not receive interrupts in
> the bringup path until they are online.)

The whole reason for all these changes in AMP on dual core Cortex A9 where
on one cpu runs Linux and on the second RTOS/standlone.

It means that the second cpu is from Linux point of view offline all the time.
But GIC has to be (and can be) setup to forward IRQ directly to the second cpu.
GIC should be controlled by Linux.

Thanks,
Michal




-- 
Michal Simek, Ing. (M.Eng)
w: www.monstr.eu p: +42-0-721842854
Maintainer of Linux kernel 2.6 Microblaze Linux - http://www.monstr.eu/fdt/
Microblaze U-BOOT custodian

[RFC PATCH 1/8] ARM: gic: Support forcing cpumask for possible cpus in the system

[Michal Simek]

On 06/18/2012 01:33 PM, Russell King - ARM Linux wrote:
> On Mon, Jun 18, 2012 at 01:30:04PM +0200, Michal Simek wrote:
>> gic_set_affinity doesn't support force option. Force option
>> should work with all possible cpus in the system not only with
>> online cpus.
>
> Neither does x86.  Why do you want this?  Offline CPUs are unable to
> receive the interrupt in any case (and must not receive interrupts in
> the bringup path until they are online.)

The whole reason for all these changes in AMP on dual core Cortex A9 where
on one cpu runs Linux and on the second RTOS/standlone.

It means that the second cpu is from Linux point of view offline all the time.
But GIC has to be (and can be) setup to forward IRQ directly to the second cpu.
GIC should be controlled by Linux.

Thanks,
Michal




-- 
Michal Simek, Ing. (M.Eng)
w: www.monstr.eu p: +42-0-721842854
Maintainer of Linux kernel 2.6 Microblaze Linux - http://www.monstr.eu/fdt/
Microblaze U-BOOT custodian

[RFC PATCH 1/8] ARM: gic: Support forcing cpumask for possible cpus in the system

[Michal Simek]

2012/6/18 Michal Simek <monstr@monstr.eu>

> On 06/18/2012 01:33 PM, Russell King - ARM Linux wrote:
>
>> On Mon, Jun 18, 2012 at 01:30:04PM +0200, Michal Simek wrote:
>>
>>> gic_set_affinity doesn't support force option. Force option
>>> should work with all possible cpus in the system not only with
>>> online cpus.
>>>
>>
>> Neither does x86.  Why do you want this?  Offline CPUs are unable to
>> receive the interrupt in any case (and must not receive interrupts in
>> the bringup path until they are online.)
>>
>
> The whole reason for all these changes in AMP on dual core Cortex A9 where
> on one cpu runs Linux and on the second RTOS/standlone.
>
> It means that the second cpu is from Linux point of view offline all the
> time.
> But GIC has to be (and can be) setup to forward IRQ directly to the second
> cpu.
> GIC should be controlled by Linux.
>
>
Russel: Any other reaction?
Can you see a way how to get this feature to the mainline Linux kernel?
Some patches needs to be refactored a little bit but the point is if you
are willing to
support this type of configuration.

Not sure if Ohad has any information that someone else would like to use it
remoteproc(AMP) among ARM cores.

Thanks,
Michal
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20120620/5d1c190a/attachment-0001.html>

[RFC PATCH 2/8] ARM: gic: Export gic_raise_softirq function for kernel modules

[Michal Simek]

This function can be used by device driver for raising
software irqs. For example AMP(remoteproc) on Zynq.

Signed-off-by: Michal Simek <monstr@monstr.eu>
---
 arch/arm/common/gic.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

diff --git a/arch/arm/common/gic.c b/arch/arm/common/gic.c
index 4003cc2..b8e7202 100644
--- a/arch/arm/common/gic.c
+++ b/arch/arm/common/gic.c
@@ -22,6 +22,7 @@
  * As such, the enable set/clear, pending set/clear and active bit
  * registers are banked per-cpu for these sources.
  */
+#include <linux/module.h>
 #include <linux/init.h>
 #include <linux/kernel.h>
 #include <linux/err.h>
@@ -753,6 +754,7 @@ void gic_raise_softirq(const struct cpumask *mask, unsigned int irq)
 	/* this always happens on GIC0 */
 	writel_relaxed(map << 16 | irq, gic_data_dist_base(&gic_data[0]) + GIC_DIST_SOFTINT);
 }
+EXPORT_SYMBOL(gic_raise_softirq);
 #endif
 
 #ifdef CONFIG_OF
-- 
1.7.0.4

[RFC PATCH 2/8] ARM: gic: Export gic_raise_softirq function for kernel modules

[Michal Simek]

This function can be used by device driver for raising
software irqs. For example AMP(remoteproc) on Zynq.

Signed-off-by: Michal Simek <monstr@monstr.eu>
---
 arch/arm/common/gic.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

diff --git a/arch/arm/common/gic.c b/arch/arm/common/gic.c
index 4003cc2..b8e7202 100644
--- a/arch/arm/common/gic.c
+++ b/arch/arm/common/gic.c
@@ -22,6 +22,7 @@
  * As such, the enable set/clear, pending set/clear and active bit
  * registers are banked per-cpu for these sources.
  */
+#include <linux/module.h>
 #include <linux/init.h>
 #include <linux/kernel.h>
 #include <linux/err.h>
@@ -753,6 +754,7 @@ void gic_raise_softirq(const struct cpumask *mask, unsigned int irq)
 	/* this always happens on GIC0 */
 	writel_relaxed(map << 16 | irq, gic_data_dist_base(&gic_data[0]) + GIC_DIST_SOFTINT);
 }
+EXPORT_SYMBOL(gic_raise_softirq);
 #endif
 
 #ifdef CONFIG_OF
-- 
1.7.0.4

[RFC PATCH 3/8] ARM: gic: Introduce new gic_set_cpu

[Michal Simek]

gic_set_cpu enable option to forward specific IRQ to choosen
ARM core.

Signed-off-by: Michal Simek <monstr@monstr.eu>
---
 arch/arm/common/gic.c               |   11 +++++++++++
 arch/arm/include/asm/hardware/gic.h |    2 ++
 2 files changed, 13 insertions(+), 0 deletions(-)

diff --git a/arch/arm/common/gic.c b/arch/arm/common/gic.c
index b8e7202..2998d9e 100644
--- a/arch/arm/common/gic.c
+++ b/arch/arm/common/gic.c
@@ -257,6 +257,17 @@ static int gic_set_affinity(struct irq_data *d, const struct cpumask *mask_val,
 
 	return IRQ_SET_MASK_OK;
 }
+
+void gic_set_cpu(unsigned int cpu, unsigned int irq)
+{
+	struct irq_data *d = irq_get_irq_data(irq);
+	struct cpumask mask;
+
+	cpumask_clear(&mask);
+	cpumask_set_cpu(cpu, &mask);
+	gic_set_affinity(d, &mask, true);
+}
+EXPORT_SYMBOL(gic_set_cpu);
 #endif
 
 #ifdef CONFIG_PM
diff --git a/arch/arm/include/asm/hardware/gic.h b/arch/arm/include/asm/hardware/gic.h
index 4b1ce6c..8aa9631 100644
--- a/arch/arm/include/asm/hardware/gic.h
+++ b/arch/arm/include/asm/hardware/gic.h
@@ -46,6 +46,8 @@ void gic_handle_irq(struct pt_regs *regs);
 void gic_cascade_irq(unsigned int gic_nr, unsigned int irq);
 void gic_raise_softirq(const struct cpumask *mask, unsigned int irq);
 
+void gic_set_cpu(unsigned int cpu, unsigned int irq);
+
 static inline void gic_init(unsigned int nr, int start,
 			    void __iomem *dist , void __iomem *cpu)
 {
-- 
1.7.0.4

[RFC PATCH 3/8] ARM: gic: Introduce new gic_set_cpu

[Michal Simek]

gic_set_cpu enable option to forward specific IRQ to choosen
ARM core.

Signed-off-by: Michal Simek <monstr@monstr.eu>
---
 arch/arm/common/gic.c               |   11 +++++++++++
 arch/arm/include/asm/hardware/gic.h |    2 ++
 2 files changed, 13 insertions(+), 0 deletions(-)

diff --git a/arch/arm/common/gic.c b/arch/arm/common/gic.c
index b8e7202..2998d9e 100644
--- a/arch/arm/common/gic.c
+++ b/arch/arm/common/gic.c
@@ -257,6 +257,17 @@ static int gic_set_affinity(struct irq_data *d, const struct cpumask *mask_val,
 
 	return IRQ_SET_MASK_OK;
 }
+
+void gic_set_cpu(unsigned int cpu, unsigned int irq)
+{
+	struct irq_data *d = irq_get_irq_data(irq);
+	struct cpumask mask;
+
+	cpumask_clear(&mask);
+	cpumask_set_cpu(cpu, &mask);
+	gic_set_affinity(d, &mask, true);
+}
+EXPORT_SYMBOL(gic_set_cpu);
 #endif
 
 #ifdef CONFIG_PM
diff --git a/arch/arm/include/asm/hardware/gic.h b/arch/arm/include/asm/hardware/gic.h
index 4b1ce6c..8aa9631 100644
--- a/arch/arm/include/asm/hardware/gic.h
+++ b/arch/arm/include/asm/hardware/gic.h
@@ -46,6 +46,8 @@ void gic_handle_irq(struct pt_regs *regs);
 void gic_cascade_irq(unsigned int gic_nr, unsigned int irq);
 void gic_raise_softirq(const struct cpumask *mask, unsigned int irq);
 
+void gic_set_cpu(unsigned int cpu, unsigned int irq);
+
 static inline void gic_init(unsigned int nr, int start,
 			    void __iomem *dist , void __iomem *cpu)
 {
-- 
1.7.0.4

[RFC PATCH 4/8] ARM: smp: Move cpu initialization directly to ipi_cpu_stop

[Michal Simek]

This is just preparation patch for dynamic IPI allocation.

Signed-off-by: Michal Simek <monstr@monstr.eu>
---
 arch/arm/kernel/smp.c |    6 ++++--
 1 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/arch/arm/kernel/smp.c b/arch/arm/kernel/smp.c
index 2c7217d..47406ee 100644
--- a/arch/arm/kernel/smp.c
+++ b/arch/arm/kernel/smp.c
@@ -468,8 +468,10 @@ static DEFINE_RAW_SPINLOCK(stop_lock);
 /*
  * ipi_cpu_stop - handle IPI from smp_send_stop()
  */
-static void ipi_cpu_stop(unsigned int cpu)
+static void ipi_cpu_stop(void)
 {
+	unsigned int cpu = smp_processor_id();
+
 	if (system_state == SYSTEM_BOOTING ||
 	    system_state == SYSTEM_RUNNING) {
 		raw_spin_lock(&stop_lock);
@@ -528,7 +530,7 @@ void handle_IPI(int ipinr, struct pt_regs *regs)
 
 	case IPI_CPU_STOP:
 		irq_enter();
-		ipi_cpu_stop(cpu);
+		ipi_cpu_stop();
 		irq_exit();
 		break;
 
-- 
1.7.0.4

[RFC PATCH 4/8] ARM: smp: Move cpu initialization directly to ipi_cpu_stop

[Michal Simek]

This is just preparation patch for dynamic IPI allocation.

Signed-off-by: Michal Simek <monstr@monstr.eu>
---
 arch/arm/kernel/smp.c |    6 ++++--
 1 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/arch/arm/kernel/smp.c b/arch/arm/kernel/smp.c
index 2c7217d..47406ee 100644
--- a/arch/arm/kernel/smp.c
+++ b/arch/arm/kernel/smp.c
@@ -468,8 +468,10 @@ static DEFINE_RAW_SPINLOCK(stop_lock);
 /*
  * ipi_cpu_stop - handle IPI from smp_send_stop()
  */
-static void ipi_cpu_stop(unsigned int cpu)
+static void ipi_cpu_stop(void)
 {
+	unsigned int cpu = smp_processor_id();
+
 	if (system_state == SYSTEM_BOOTING ||
 	    system_state == SYSTEM_RUNNING) {
 		raw_spin_lock(&stop_lock);
@@ -528,7 +530,7 @@ void handle_IPI(int ipinr, struct pt_regs *regs)
 
 	case IPI_CPU_STOP:
 		irq_enter();
-		ipi_cpu_stop(cpu);
+		ipi_cpu_stop();
 		irq_exit();
 		break;
 
-- 
1.7.0.4

Re: [RFC PATCH 4/8] ARM: smp: Move cpu initialization directly to ipi_cpu_stop

[Russell King - ARM Linux]

On Mon, Jun 18, 2012 at 01:30:07PM +0200, Michal Simek wrote:
> This is just preparation patch for dynamic IPI allocation.

Why?

[RFC PATCH 4/8] ARM: smp: Move cpu initialization directly to ipi_cpu_stop

[Russell King - ARM Linux]

On Mon, Jun 18, 2012 at 01:30:07PM +0200, Michal Simek wrote:
> This is just preparation patch for dynamic IPI allocation.

Why?

Re: [RFC PATCH 4/8] ARM: smp: Move cpu initialization directly to ipi_cpu_stop

[Michal Simek]

On 06/18/2012 01:34 PM, Russell King - ARM Linux wrote:
> On Mon, Jun 18, 2012 at 01:30:07PM +0200, Michal Simek wrote:
>> This is just preparation patch for dynamic IPI allocation.
>
> Why?

Let's start with this one first.

The reason for dynamic IPI allocation is that IPI interrupts can be used (and I am using them)
for sending IRQ to the second cpu and the second cpu doesn't run Linux.
But still I need functionality for sending interrupts to the second core for Linux. A also need
to have any hook to be able to receive irq from the second core.
Based on that I have looked at this code where only static allocation is used and it is not possible
to use it in this model because no driver can allocate IPI handler.
Based on that I have changed it to model where dynamic IPIs can be allocated dynamically.
As you see from the code. Origin static IPI allocation is still there.

Maybe what it is not visible is that NR_IPI value. As you commented in 5/8 NR_IPI stores number of
IPI used by Linux and I am using is as maximum IPIs in the system.
I am completely fine with not changing NR_IPI value and introduce new one like MAX_NR_IPI value which
will be for Xilinx Zynq GIC setup to 16.

Hope I have clarify it a little bit.

Thanks,
Michal

-- 
Michal Simek, Ing. (M.Eng)
w: www.monstr.eu p: +42-0-721842854
Maintainer of Linux kernel 2.6 Microblaze Linux - http://www.monstr.eu/fdt/
Microblaze U-BOOT custodian

[RFC PATCH 4/8] ARM: smp: Move cpu initialization directly to ipi_cpu_stop

[Michal Simek]

On 06/18/2012 01:34 PM, Russell King - ARM Linux wrote:
> On Mon, Jun 18, 2012 at 01:30:07PM +0200, Michal Simek wrote:
>> This is just preparation patch for dynamic IPI allocation.
>
> Why?

Let's start with this one first.

The reason for dynamic IPI allocation is that IPI interrupts can be used (and I am using them)
for sending IRQ to the second cpu and the second cpu doesn't run Linux.
But still I need functionality for sending interrupts to the second core for Linux. A also need
to have any hook to be able to receive irq from the second core.
Based on that I have looked at this code where only static allocation is used and it is not possible
to use it in this model because no driver can allocate IPI handler.
Based on that I have changed it to model where dynamic IPIs can be allocated dynamically.
As you see from the code. Origin static IPI allocation is still there.

Maybe what it is not visible is that NR_IPI value. As you commented in 5/8 NR_IPI stores number of
IPI used by Linux and I am using is as maximum IPIs in the system.
I am completely fine with not changing NR_IPI value and introduce new one like MAX_NR_IPI value which
will be for Xilinx Zynq GIC setup to 16.

Hope I have clarify it a little bit.

Thanks,
Michal

-- 
Michal Simek, Ing. (M.Eng)
w: www.monstr.eu p: +42-0-721842854
Maintainer of Linux kernel 2.6 Microblaze Linux - http://www.monstr.eu/fdt/
Microblaze U-BOOT custodian

[RFC PATCH 5/8] AMP: smp: Extend number of IPIs

[Michal Simek]

GIC supports up to 16 IPIs.

Signed-off-by: Michal Simek <monstr@monstr.eu>

---
Please correct me if this code is also designed for different
interrupt controller which doesn't support 16 IPIs.
---
 arch/arm/include/asm/hardirq.h |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/arch/arm/include/asm/hardirq.h b/arch/arm/include/asm/hardirq.h
index 436e60b..3e9ded5 100644
--- a/arch/arm/include/asm/hardirq.h
+++ b/arch/arm/include/asm/hardirq.h
@@ -5,7 +5,7 @@
 #include <linux/threads.h>
 #include <asm/irq.h>
 
-#define NR_IPI	5
+#define NR_IPI	16
 
 typedef struct {
 	unsigned int __softirq_pending;
-- 
1.7.0.4

[RFC PATCH 5/8] AMP: smp: Extend number of IPIs

[Michal Simek]

GIC supports up to 16 IPIs.

Signed-off-by: Michal Simek <monstr@monstr.eu>

---
Please correct me if this code is also designed for different
interrupt controller which doesn't support 16 IPIs.
---
 arch/arm/include/asm/hardirq.h |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/arch/arm/include/asm/hardirq.h b/arch/arm/include/asm/hardirq.h
index 436e60b..3e9ded5 100644
--- a/arch/arm/include/asm/hardirq.h
+++ b/arch/arm/include/asm/hardirq.h
@@ -5,7 +5,7 @@
 #include <linux/threads.h>
 #include <asm/irq.h>
 
-#define NR_IPI	5
+#define NR_IPI	16
 
 typedef struct {
 	unsigned int __softirq_pending;
-- 
1.7.0.4

Re: [RFC PATCH 5/8] AMP: smp: Extend number of IPIs

[Russell King - ARM Linux]

On Mon, Jun 18, 2012 at 01:30:08PM +0200, Michal Simek wrote:
> GIC supports up to 16 IPIs.
> 
> Signed-off-by: Michal Simek <monstr@monstr.eu>
> 
> ---
> Please correct me if this code is also designed for different
> interrupt controller which doesn't support 16 IPIs.

This code is not specific to the GIC.  We use NR_IPI to indicate the
number of IPIs that the kernel actually uses, not the number which the
hardware provides.

[RFC PATCH 5/8] AMP: smp: Extend number of IPIs

[Russell King - ARM Linux]

On Mon, Jun 18, 2012 at 01:30:08PM +0200, Michal Simek wrote:
> GIC supports up to 16 IPIs.
> 
> Signed-off-by: Michal Simek <monstr@monstr.eu>
> 
> ---
> Please correct me if this code is also designed for different
> interrupt controller which doesn't support 16 IPIs.

This code is not specific to the GIC.  We use NR_IPI to indicate the
number of IPIs that the kernel actually uses, not the number which the
hardware provides.

Re: [RFC PATCH 5/8] AMP: smp: Extend number of IPIs

[Michal Simek]

On 06/18/2012 01:35 PM, Russell King - ARM Linux wrote:
> On Mon, Jun 18, 2012 at 01:30:08PM +0200, Michal Simek wrote:
>> GIC supports up to 16 IPIs.
>>
>> Signed-off-by: Michal Simek<monstr@monstr.eu>
>>
>> ---
>> Please correct me if this code is also designed for different
>> interrupt controller which doesn't support 16 IPIs.
>
> This code is not specific to the GIC.  We use NR_IPI to indicate the
> number of IPIs that the kernel actually uses, not the number which the
> hardware provides.

ok.

Thanks,
Michal


-- 
Michal Simek, Ing. (M.Eng)
w: www.monstr.eu p: +42-0-721842854
Maintainer of Linux kernel 2.6 Microblaze Linux - http://www.monstr.eu/fdt/
Microblaze U-BOOT custodian

[RFC PATCH 5/8] AMP: smp: Extend number of IPIs

[Michal Simek]

On 06/18/2012 01:35 PM, Russell King - ARM Linux wrote:
> On Mon, Jun 18, 2012 at 01:30:08PM +0200, Michal Simek wrote:
>> GIC supports up to 16 IPIs.
>>
>> Signed-off-by: Michal Simek<monstr@monstr.eu>
>>
>> ---
>> Please correct me if this code is also designed for different
>> interrupt controller which doesn't support 16 IPIs.
>
> This code is not specific to the GIC.  We use NR_IPI to indicate the
> number of IPIs that the kernel actually uses, not the number which the
> hardware provides.

ok.

Thanks,
Michal


-- 
Michal Simek, Ing. (M.Eng)
w: www.monstr.eu p: +42-0-721842854
Maintainer of Linux kernel 2.6 Microblaze Linux - http://www.monstr.eu/fdt/
Microblaze U-BOOT custodian

[RFC PATCH 6/8] ARM: smp: Use generic API for ipi

[Michal Simek]

Support dynamic IPI allocation. It supports cases
where user wants to use IPI for communication among cpus
and setup own handler for it.

New functions set_ipi_handler/clear_ipi_handler are used
for dynamic allocation.

Signed-off-by: Michal Simek <monstr@monstr.eu>
---
 arch/arm/include/asm/smp.h |    3 +
 arch/arm/kernel/smp.c      |  104 +++++++++++++++++++++++++------------------
 2 files changed, 63 insertions(+), 44 deletions(-)

diff --git a/arch/arm/include/asm/smp.h b/arch/arm/include/asm/smp.h
index ae29293..308cca9 100644
--- a/arch/arm/include/asm/smp.h
+++ b/arch/arm/include/asm/smp.h
@@ -93,4 +93,7 @@ extern void platform_cpu_enable(unsigned int cpu);
 extern void arch_send_call_function_single_ipi(int cpu);
 extern void arch_send_call_function_ipi_mask(const struct cpumask *mask);
 
+extern int set_ipi_handler(int ipinr, void *handler, char *desc);
+extern void clear_ipi_handler(int ipinr);
+
 #endif /* ifndef __ASM_ARM_SMP_H */
diff --git a/arch/arm/kernel/smp.c b/arch/arm/kernel/smp.c
index 47406ee..4af1679 100644
--- a/arch/arm/kernel/smp.c
+++ b/arch/arm/kernel/smp.c
@@ -51,7 +51,7 @@
 struct secondary_data secondary_data;
 
 enum ipi_msg_type {
-	IPI_TIMER = 2,
+	IPI_TIMER,
 	IPI_RESCHEDULE,
 	IPI_CALL_FUNC,
 	IPI_CALL_FUNC_SINGLE,
@@ -346,13 +346,23 @@ void arch_send_call_function_single_ipi(int cpu)
 	smp_cross_call(cpumask_of(cpu), IPI_CALL_FUNC_SINGLE);
 }
 
-static const char *ipi_types[NR_IPI] = {
-#define S(x,s)	[x - IPI_TIMER] = s
-	S(IPI_TIMER, "Timer broadcast interrupts"),
-	S(IPI_RESCHEDULE, "Rescheduling interrupts"),
-	S(IPI_CALL_FUNC, "Function call interrupts"),
-	S(IPI_CALL_FUNC_SINGLE, "Single function call interrupts"),
-	S(IPI_CPU_STOP, "CPU stop interrupts"),
+struct ipi {
+	const char *desc;
+	void (*handler)(void);
+};
+
+static void ipi_timer(void);
+static void ipi_cpu_stop(void);
+
+static struct ipi ipi_types[NR_IPI] = {
+#define S(x, s, f)	[x].desc = s, [x].handler = f
+	S(IPI_TIMER, "Timer broadcast interrupts", ipi_timer),
+	S(IPI_RESCHEDULE, "Rescheduling interrupts", scheduler_ipi),
+	S(IPI_CALL_FUNC, "Function call interrupts",
+					generic_smp_call_function_interrupt),
+	S(IPI_CALL_FUNC_SINGLE, "Single function call interrupts",
+				generic_smp_call_function_single_interrupt),
+	S(IPI_CPU_STOP, "CPU stop interrupts", ipi_cpu_stop),
 };
 
 void show_ipi_list(struct seq_file *p, int prec)
@@ -360,13 +370,15 @@ void show_ipi_list(struct seq_file *p, int prec)
 	unsigned int cpu, i;
 
 	for (i = 0; i < NR_IPI; i++) {
-		seq_printf(p, "%*s%u: ", prec - 1, "IPI", i);
+		if (ipi_types[i].handler) {
+			seq_printf(p, "%*s%u: ", prec - 1, "IPI", i);
 
-		for_each_present_cpu(cpu)
-			seq_printf(p, "%10u ",
-				   __get_irq_stat(cpu, ipi_irqs[i]));
+			for_each_present_cpu(cpu)
+				seq_printf(p, "%10u ",
+					__get_irq_stat(cpu, ipi_irqs[i]));
 
-		seq_printf(p, " %s\n", ipi_types[i]);
+			seq_printf(p, " %s\n", ipi_types[i].desc);
+		}
 	}
 }
 
@@ -502,45 +514,49 @@ void handle_IPI(int ipinr, struct pt_regs *regs)
 	unsigned int cpu = smp_processor_id();
 	struct pt_regs *old_regs = set_irq_regs(regs);
 
-	if (ipinr >= IPI_TIMER && ipinr < IPI_TIMER + NR_IPI)
-		__inc_irq_stat(cpu, ipi_irqs[ipinr - IPI_TIMER]);
-
-	switch (ipinr) {
-	case IPI_TIMER:
+	if (ipi_types[ipinr].handler) {
+		__inc_irq_stat(cpu, ipi_irqs[ipinr]);
 		irq_enter();
-		ipi_timer();
+		(*ipi_types[ipinr].handler)();
 		irq_exit();
-		break;
+	} else
+		pr_crit("CPU%u: Unknown IPI message 0x%x\n",
+		       cpu, ipinr);
 
-	case IPI_RESCHEDULE:
-		scheduler_ipi();
-		break;
+	set_irq_regs(old_regs);
+}
 
-	case IPI_CALL_FUNC:
-		irq_enter();
-		generic_smp_call_function_interrupt();
-		irq_exit();
-		break;
+/*
+ * set_ipi_handler:
+ * Interface provided for a kernel module to specify an IPI handler function.
+ */
+int set_ipi_handler(int ipinr, void *handler, char *desc)
+{
+	unsigned int cpu = smp_processor_id();
 
-	case IPI_CALL_FUNC_SINGLE:
-		irq_enter();
-		generic_smp_call_function_single_interrupt();
-		irq_exit();
-		break;
+	if (ipi_types[ipinr].handler) {
+		pr_crit("CPU%u: IPI handler 0x%x already registered to %pf\n",
+					cpu, ipinr, ipi_types[ipinr].handler);
+		return -1;
+	}
 
-	case IPI_CPU_STOP:
-		irq_enter();
-		ipi_cpu_stop();
-		irq_exit();
-		break;
+	ipi_types[ipinr].handler = handler;
+	ipi_types[ipinr].desc = desc;
 
-	default:
-		printk(KERN_CRIT "CPU%u: Unknown IPI message 0x%x\n",
-		       cpu, ipinr);
-		break;
-	}
-	set_irq_regs(old_regs);
+	return 0;
+}
+EXPORT_SYMBOL(set_ipi_handler);
+
+/*
+ * clear_ipi_handler:
+ * Interface provided for a kernel module to clear an IPI handler function.
+ */
+void clear_ipi_handler(int ipinr)
+{
+	ipi_types[ipinr].handler = NULL;
+	ipi_types[ipinr].desc = NULL;
 }
+EXPORT_SYMBOL(clear_ipi_handler);
 
 void smp_send_reschedule(int cpu)
 {
-- 
1.7.0.4

[RFC PATCH 6/8] ARM: smp: Use generic API for ipi

[Michal Simek]

Support dynamic IPI allocation. It supports cases
where user wants to use IPI for communication among cpus
and setup own handler for it.

New functions set_ipi_handler/clear_ipi_handler are used
for dynamic allocation.

Signed-off-by: Michal Simek <monstr@monstr.eu>
---
 arch/arm/include/asm/smp.h |    3 +
 arch/arm/kernel/smp.c      |  104 +++++++++++++++++++++++++------------------
 2 files changed, 63 insertions(+), 44 deletions(-)

diff --git a/arch/arm/include/asm/smp.h b/arch/arm/include/asm/smp.h
index ae29293..308cca9 100644
--- a/arch/arm/include/asm/smp.h
+++ b/arch/arm/include/asm/smp.h
@@ -93,4 +93,7 @@ extern void platform_cpu_enable(unsigned int cpu);
 extern void arch_send_call_function_single_ipi(int cpu);
 extern void arch_send_call_function_ipi_mask(const struct cpumask *mask);
 
+extern int set_ipi_handler(int ipinr, void *handler, char *desc);
+extern void clear_ipi_handler(int ipinr);
+
 #endif /* ifndef __ASM_ARM_SMP_H */
diff --git a/arch/arm/kernel/smp.c b/arch/arm/kernel/smp.c
index 47406ee..4af1679 100644
--- a/arch/arm/kernel/smp.c
+++ b/arch/arm/kernel/smp.c
@@ -51,7 +51,7 @@
 struct secondary_data secondary_data;
 
 enum ipi_msg_type {
-	IPI_TIMER = 2,
+	IPI_TIMER,
 	IPI_RESCHEDULE,
 	IPI_CALL_FUNC,
 	IPI_CALL_FUNC_SINGLE,
@@ -346,13 +346,23 @@ void arch_send_call_function_single_ipi(int cpu)
 	smp_cross_call(cpumask_of(cpu), IPI_CALL_FUNC_SINGLE);
 }
 
-static const char *ipi_types[NR_IPI] = {
-#define S(x,s)	[x - IPI_TIMER] = s
-	S(IPI_TIMER, "Timer broadcast interrupts"),
-	S(IPI_RESCHEDULE, "Rescheduling interrupts"),
-	S(IPI_CALL_FUNC, "Function call interrupts"),
-	S(IPI_CALL_FUNC_SINGLE, "Single function call interrupts"),
-	S(IPI_CPU_STOP, "CPU stop interrupts"),
+struct ipi {
+	const char *desc;
+	void (*handler)(void);
+};
+
+static void ipi_timer(void);
+static void ipi_cpu_stop(void);
+
+static struct ipi ipi_types[NR_IPI] = {
+#define S(x, s, f)	[x].desc = s, [x].handler = f
+	S(IPI_TIMER, "Timer broadcast interrupts", ipi_timer),
+	S(IPI_RESCHEDULE, "Rescheduling interrupts", scheduler_ipi),
+	S(IPI_CALL_FUNC, "Function call interrupts",
+					generic_smp_call_function_interrupt),
+	S(IPI_CALL_FUNC_SINGLE, "Single function call interrupts",
+				generic_smp_call_function_single_interrupt),
+	S(IPI_CPU_STOP, "CPU stop interrupts", ipi_cpu_stop),
 };
 
 void show_ipi_list(struct seq_file *p, int prec)
@@ -360,13 +370,15 @@ void show_ipi_list(struct seq_file *p, int prec)
 	unsigned int cpu, i;
 
 	for (i = 0; i < NR_IPI; i++) {
-		seq_printf(p, "%*s%u: ", prec - 1, "IPI", i);
+		if (ipi_types[i].handler) {
+			seq_printf(p, "%*s%u: ", prec - 1, "IPI", i);
 
-		for_each_present_cpu(cpu)
-			seq_printf(p, "%10u ",
-				   __get_irq_stat(cpu, ipi_irqs[i]));
+			for_each_present_cpu(cpu)
+				seq_printf(p, "%10u ",
+					__get_irq_stat(cpu, ipi_irqs[i]));
 
-		seq_printf(p, " %s\n", ipi_types[i]);
+			seq_printf(p, " %s\n", ipi_types[i].desc);
+		}
 	}
 }
 
@@ -502,45 +514,49 @@ void handle_IPI(int ipinr, struct pt_regs *regs)
 	unsigned int cpu = smp_processor_id();
 	struct pt_regs *old_regs = set_irq_regs(regs);
 
-	if (ipinr >= IPI_TIMER && ipinr < IPI_TIMER + NR_IPI)
-		__inc_irq_stat(cpu, ipi_irqs[ipinr - IPI_TIMER]);
-
-	switch (ipinr) {
-	case IPI_TIMER:
+	if (ipi_types[ipinr].handler) {
+		__inc_irq_stat(cpu, ipi_irqs[ipinr]);
 		irq_enter();
-		ipi_timer();
+		(*ipi_types[ipinr].handler)();
 		irq_exit();
-		break;
+	} else
+		pr_crit("CPU%u: Unknown IPI message 0x%x\n",
+		       cpu, ipinr);
 
-	case IPI_RESCHEDULE:
-		scheduler_ipi();
-		break;
+	set_irq_regs(old_regs);
+}
 
-	case IPI_CALL_FUNC:
-		irq_enter();
-		generic_smp_call_function_interrupt();
-		irq_exit();
-		break;
+/*
+ * set_ipi_handler:
+ * Interface provided for a kernel module to specify an IPI handler function.
+ */
+int set_ipi_handler(int ipinr, void *handler, char *desc)
+{
+	unsigned int cpu = smp_processor_id();
 
-	case IPI_CALL_FUNC_SINGLE:
-		irq_enter();
-		generic_smp_call_function_single_interrupt();
-		irq_exit();
-		break;
+	if (ipi_types[ipinr].handler) {
+		pr_crit("CPU%u: IPI handler 0x%x already registered to %pf\n",
+					cpu, ipinr, ipi_types[ipinr].handler);
+		return -1;
+	}
 
-	case IPI_CPU_STOP:
-		irq_enter();
-		ipi_cpu_stop();
-		irq_exit();
-		break;
+	ipi_types[ipinr].handler = handler;
+	ipi_types[ipinr].desc = desc;
 
-	default:
-		printk(KERN_CRIT "CPU%u: Unknown IPI message 0x%x\n",
-		       cpu, ipinr);
-		break;
-	}
-	set_irq_regs(old_regs);
+	return 0;
+}
+EXPORT_SYMBOL(set_ipi_handler);
+
+/*
+ * clear_ipi_handler:
+ * Interface provided for a kernel module to clear an IPI handler function.
+ */
+void clear_ipi_handler(int ipinr)
+{
+	ipi_types[ipinr].handler = NULL;
+	ipi_types[ipinr].desc = NULL;
 }
+EXPORT_SYMBOL(clear_ipi_handler);
 
 void smp_send_reschedule(int cpu)
 {
-- 
1.7.0.4

Re: [RFC PATCH 6/8] ARM: smp: Use generic API for ipi

[Russell King - ARM Linux]

On Mon, Jun 18, 2012 at 01:30:09PM +0200, Michal Simek wrote:
> Support dynamic IPI allocation. It supports cases
> where user wants to use IPI for communication among cpus
> and setup own handler for it.
> 
> New functions set_ipi_handler/clear_ipi_handler are used
> for dynamic allocation.

For what purpose?

[RFC PATCH 6/8] ARM: smp: Use generic API for ipi

[Russell King - ARM Linux]

On Mon, Jun 18, 2012 at 01:30:09PM +0200, Michal Simek wrote:
> Support dynamic IPI allocation. It supports cases
> where user wants to use IPI for communication among cpus
> and setup own handler for it.
> 
> New functions set_ipi_handler/clear_ipi_handler are used
> for dynamic allocation.

For what purpose?

Re: [RFC PATCH 6/8] ARM: smp: Use generic API for ipi

[Michal Simek]

On 06/18/2012 01:36 PM, Russell King - ARM Linux wrote:
> On Mon, Jun 18, 2012 at 01:30:09PM +0200, Michal Simek wrote:
>> Support dynamic IPI allocation. It supports cases
>> where user wants to use IPI for communication among cpus
>> and setup own handler for it.
>>
>> New functions set_ipi_handler/clear_ipi_handler are used
>> for dynamic allocation.
>
> For what purpose?

Please look at my reaction in 4/8. It is the part of dynamic IPI allocation.

Thanks,
Michal


-- 
Michal Simek, Ing. (M.Eng)
w: www.monstr.eu p: +42-0-721842854
Maintainer of Linux kernel 2.6 Microblaze Linux - http://www.monstr.eu/fdt/
Microblaze U-BOOT custodian

[RFC PATCH 6/8] ARM: smp: Use generic API for ipi

[Michal Simek]

On 06/18/2012 01:36 PM, Russell King - ARM Linux wrote:
> On Mon, Jun 18, 2012 at 01:30:09PM +0200, Michal Simek wrote:
>> Support dynamic IPI allocation. It supports cases
>> where user wants to use IPI for communication among cpus
>> and setup own handler for it.
>>
>> New functions set_ipi_handler/clear_ipi_handler are used
>> for dynamic allocation.
>
> For what purpose?

Please look at my reaction in 4/8. It is the part of dynamic IPI allocation.

Thanks,
Michal


-- 
Michal Simek, Ing. (M.Eng)
w: www.monstr.eu p: +42-0-721842854
Maintainer of Linux kernel 2.6 Microblaze Linux - http://www.monstr.eu/fdt/
Microblaze U-BOOT custodian

[RFC PATCH 7/8] ARM: vmlinux.lds: Setup physical load address not virtual

[Michal Simek]

Setup correct virtual and physical address in ELF LOAD section.

Signed-off-by: Michal Simek <monstr@monstr.eu>
---
 arch/arm/include/asm/page.h   |    2 ++
 arch/arm/kernel/vmlinux.lds.S |   28 ++++++++++++++--------------
 2 files changed, 16 insertions(+), 14 deletions(-)

diff --git a/arch/arm/include/asm/page.h b/arch/arm/include/asm/page.h
index ecf9019..4070209 100644
--- a/arch/arm/include/asm/page.h
+++ b/arch/arm/include/asm/page.h
@@ -15,6 +15,8 @@
 #define PAGE_SIZE		(_AC(1,UL) << PAGE_SHIFT)
 #define PAGE_MASK		(~(PAGE_SIZE-1))
 
+#define LOAD_OFFSET	(CONFIG_PAGE_OFFSET - CONFIG_PHYS_OFFSET)
+
 #ifndef __ASSEMBLY__
 
 #ifndef CONFIG_MMU
diff --git a/arch/arm/kernel/vmlinux.lds.S b/arch/arm/kernel/vmlinux.lds.S
index c887be0..b996bc0 100644
--- a/arch/arm/kernel/vmlinux.lds.S
+++ b/arch/arm/kernel/vmlinux.lds.S
@@ -3,11 +3,11 @@
  * Written by Martin Mares <mj@atrey.karlin.mff.cuni.cz>
  */
 
+#include <asm/page.h>
 #include <asm-generic/vmlinux.lds.h>
 #include <asm/cache.h>
 #include <asm/thread_info.h>
 #include <asm/memory.h>
-#include <asm/page.h>
 	
 #define PROC_INFO							\
 	. = ALIGN(4);							\
@@ -86,11 +86,11 @@ SECTIONS
 #else
 	. = PAGE_OFFSET + TEXT_OFFSET;
 #endif
-	.head.text : {
+	.head.text : AT(ADDR(.head.text) - LOAD_OFFSET) {
 		_text = .;
 		HEAD_TEXT
 	}
-	.text : {			/* Real text segment		*/
+	.text : AT(ADDR(.text) - LOAD_OFFSET){	/* Real text segment	*/
 		_stext = .;		/* Text and read-only data	*/
 			__exception_text_start = .;
 			*(.exception.text)
@@ -132,12 +132,12 @@ SECTIONS
 	 * Stack unwinding tables
 	 */
 	. = ALIGN(8);
-	.ARM.unwind_idx : {
+	.ARM.unwind_idx : AT(ADDR(.ARM.unwind_idx) - LOAD_OFFSET) {
 		__start_unwind_idx = .;
 		*(.ARM.exidx*)
 		__stop_unwind_idx = .;
 	}
-	.ARM.unwind_tab : {
+	.ARM.unwind_tab : AT(ADDR(.ARM.unwind_tab) - LOAD_OFFSET) {
 		__start_unwind_tab = .;
 		*(.ARM.extab*)
 		__stop_unwind_tab = .;
@@ -152,35 +152,35 @@ SECTIONS
 #endif
 
 	INIT_TEXT_SECTION(8)
-	.exit.text : {
+	.exit.text : AT(ADDR(.exit.text) - LOAD_OFFSET) {
 		ARM_EXIT_KEEP(EXIT_TEXT)
 	}
-	.init.proc.info : {
+	.init.proc.info : AT(ADDR(.init.proc.info) - LOAD_OFFSET) {
 		ARM_CPU_DISCARD(PROC_INFO)
 	}
-	.init.arch.info : {
+	.init.arch.info : AT(ADDR(.init.arch.info) - LOAD_OFFSET) {
 		__arch_info_begin = .;
 		*(.arch.info.init)
 		__arch_info_end = .;
 	}
-	.init.tagtable : {
+	.init.tagtable : AT(ADDR(.init.tagtable) - LOAD_OFFSET) {
 		__tagtable_begin = .;
 		*(.taglist.init)
 		__tagtable_end = .;
 	}
 #ifdef CONFIG_SMP_ON_UP
-	.init.smpalt : {
+	.init.smpalt : AT(ADDR(.init.smpalt) - LOAD_OFFSET) {
 		__smpalt_begin = .;
 		*(.alt.smp.init)
 		__smpalt_end = .;
 	}
 #endif
-	.init.pv_table : {
+	.init.pv_table : AT(ADDR(.init.pv_table) - LOAD_OFFSET) {
 		__pv_table_begin = .;
 		*(.pv_table)
 		__pv_table_end = .;
 	}
-	.init.data : {
+	.init.data : AT(ADDR(.init.data) - LOAD_OFFSET) {
 #ifndef CONFIG_XIP_KERNEL
 		INIT_DATA
 #endif
@@ -191,7 +191,7 @@ SECTIONS
 		INIT_RAM_FS
 	}
 #ifndef CONFIG_XIP_KERNEL
-	.exit.data : {
+	.exit.data : AT(ADDR(.exit.data) - LOAD_OFFSET) {
 		ARM_EXIT_KEEP(EXIT_DATA)
 	}
 #endif
@@ -207,7 +207,7 @@ SECTIONS
 	__data_loc = .;
 #endif
 
-	.data : AT(__data_loc) {
+	.data : AT(ADDR(.data) - LOAD_OFFSET) {
 		_data = .;		/* address in memory */
 		_sdata = .;
 
-- 
1.7.0.4

[RFC PATCH 7/8] ARM: vmlinux.lds: Setup physical load address not virtual

[Michal Simek]

Setup correct virtual and physical address in ELF LOAD section.

Signed-off-by: Michal Simek <monstr@monstr.eu>
---
 arch/arm/include/asm/page.h   |    2 ++
 arch/arm/kernel/vmlinux.lds.S |   28 ++++++++++++++--------------
 2 files changed, 16 insertions(+), 14 deletions(-)

diff --git a/arch/arm/include/asm/page.h b/arch/arm/include/asm/page.h
index ecf9019..4070209 100644
--- a/arch/arm/include/asm/page.h
+++ b/arch/arm/include/asm/page.h
@@ -15,6 +15,8 @@
 #define PAGE_SIZE		(_AC(1,UL) << PAGE_SHIFT)
 #define PAGE_MASK		(~(PAGE_SIZE-1))
 
+#define LOAD_OFFSET	(CONFIG_PAGE_OFFSET - CONFIG_PHYS_OFFSET)
+
 #ifndef __ASSEMBLY__
 
 #ifndef CONFIG_MMU
diff --git a/arch/arm/kernel/vmlinux.lds.S b/arch/arm/kernel/vmlinux.lds.S
index c887be0..b996bc0 100644
--- a/arch/arm/kernel/vmlinux.lds.S
+++ b/arch/arm/kernel/vmlinux.lds.S
@@ -3,11 +3,11 @@
  * Written by Martin Mares <mj@atrey.karlin.mff.cuni.cz>
  */
 
+#include <asm/page.h>
 #include <asm-generic/vmlinux.lds.h>
 #include <asm/cache.h>
 #include <asm/thread_info.h>
 #include <asm/memory.h>
-#include <asm/page.h>
 	
 #define PROC_INFO							\
 	. = ALIGN(4);							\
@@ -86,11 +86,11 @@ SECTIONS
 #else
 	. = PAGE_OFFSET + TEXT_OFFSET;
 #endif
-	.head.text : {
+	.head.text : AT(ADDR(.head.text) - LOAD_OFFSET) {
 		_text = .;
 		HEAD_TEXT
 	}
-	.text : {			/* Real text segment		*/
+	.text : AT(ADDR(.text) - LOAD_OFFSET){	/* Real text segment	*/
 		_stext = .;		/* Text and read-only data	*/
 			__exception_text_start = .;
 			*(.exception.text)
@@ -132,12 +132,12 @@ SECTIONS
 	 * Stack unwinding tables
 	 */
 	. = ALIGN(8);
-	.ARM.unwind_idx : {
+	.ARM.unwind_idx : AT(ADDR(.ARM.unwind_idx) - LOAD_OFFSET) {
 		__start_unwind_idx = .;
 		*(.ARM.exidx*)
 		__stop_unwind_idx = .;
 	}
-	.ARM.unwind_tab : {
+	.ARM.unwind_tab : AT(ADDR(.ARM.unwind_tab) - LOAD_OFFSET) {
 		__start_unwind_tab = .;
 		*(.ARM.extab*)
 		__stop_unwind_tab = .;
@@ -152,35 +152,35 @@ SECTIONS
 #endif
 
 	INIT_TEXT_SECTION(8)
-	.exit.text : {
+	.exit.text : AT(ADDR(.exit.text) - LOAD_OFFSET) {
 		ARM_EXIT_KEEP(EXIT_TEXT)
 	}
-	.init.proc.info : {
+	.init.proc.info : AT(ADDR(.init.proc.info) - LOAD_OFFSET) {
 		ARM_CPU_DISCARD(PROC_INFO)
 	}
-	.init.arch.info : {
+	.init.arch.info : AT(ADDR(.init.arch.info) - LOAD_OFFSET) {
 		__arch_info_begin = .;
 		*(.arch.info.init)
 		__arch_info_end = .;
 	}
-	.init.tagtable : {
+	.init.tagtable : AT(ADDR(.init.tagtable) - LOAD_OFFSET) {
 		__tagtable_begin = .;
 		*(.taglist.init)
 		__tagtable_end = .;
 	}
 #ifdef CONFIG_SMP_ON_UP
-	.init.smpalt : {
+	.init.smpalt : AT(ADDR(.init.smpalt) - LOAD_OFFSET) {
 		__smpalt_begin = .;
 		*(.alt.smp.init)
 		__smpalt_end = .;
 	}
 #endif
-	.init.pv_table : {
+	.init.pv_table : AT(ADDR(.init.pv_table) - LOAD_OFFSET) {
 		__pv_table_begin = .;
 		*(.pv_table)
 		__pv_table_end = .;
 	}
-	.init.data : {
+	.init.data : AT(ADDR(.init.data) - LOAD_OFFSET) {
 #ifndef CONFIG_XIP_KERNEL
 		INIT_DATA
 #endif
@@ -191,7 +191,7 @@ SECTIONS
 		INIT_RAM_FS
 	}
 #ifndef CONFIG_XIP_KERNEL
-	.exit.data : {
+	.exit.data : AT(ADDR(.exit.data) - LOAD_OFFSET) {
 		ARM_EXIT_KEEP(EXIT_DATA)
 	}
 #endif
@@ -207,7 +207,7 @@ SECTIONS
 	__data_loc = .;
 #endif
 
-	.data : AT(__data_loc) {
+	.data : AT(ADDR(.data) - LOAD_OFFSET) {
 		_data = .;		/* address in memory */
 		_sdata = .;
 
-- 
1.7.0.4

Re: [RFC PATCH 7/8] ARM: vmlinux.lds: Setup physical load address not virtual

[Nicolas Pitre]

On Mon, 18 Jun 2012, Michal Simek wrote:

> Setup correct virtual and physical address in ELF LOAD section.

We are moving to a single kernel binary for multiple targets, including 
targets with different physical load addresses.  The kernel code figures 
out at run time what the actual physical address is when 
CONFIG_ARM_PATCH_PHYS_VIRT is set which is the default these days.  In 
other words, we don't know the physical offset at build time in that 
case and CONFIG_PHYS_OFFSET is simply not defined.

Why do you need this change?  Your patch comments are lacking 
justification for them.


> Signed-off-by: Michal Simek <monstr@monstr.eu>
> ---
>  arch/arm/include/asm/page.h   |    2 ++
>  arch/arm/kernel/vmlinux.lds.S |   28 ++++++++++++++--------------
>  2 files changed, 16 insertions(+), 14 deletions(-)
> 
> diff --git a/arch/arm/include/asm/page.h b/arch/arm/include/asm/page.h
> index ecf9019..4070209 100644
> --- a/arch/arm/include/asm/page.h
> +++ b/arch/arm/include/asm/page.h
> @@ -15,6 +15,8 @@
>  #define PAGE_SIZE		(_AC(1,UL) << PAGE_SHIFT)
>  #define PAGE_MASK		(~(PAGE_SIZE-1))
>  
> +#define LOAD_OFFSET	(CONFIG_PAGE_OFFSET - CONFIG_PHYS_OFFSET)
> +
>  #ifndef __ASSEMBLY__
>  
>  #ifndef CONFIG_MMU
> diff --git a/arch/arm/kernel/vmlinux.lds.S b/arch/arm/kernel/vmlinux.lds.S
> index c887be0..b996bc0 100644
> --- a/arch/arm/kernel/vmlinux.lds.S
> +++ b/arch/arm/kernel/vmlinux.lds.S
> @@ -3,11 +3,11 @@
>   * Written by Martin Mares <mj@atrey.karlin.mff.cuni.cz>
>   */
>  
> +#include <asm/page.h>
>  #include <asm-generic/vmlinux.lds.h>
>  #include <asm/cache.h>
>  #include <asm/thread_info.h>
>  #include <asm/memory.h>
> -#include <asm/page.h>
>  	
>  #define PROC_INFO							\
>  	. = ALIGN(4);							\
> @@ -86,11 +86,11 @@ SECTIONS
>  #else
>  	. = PAGE_OFFSET + TEXT_OFFSET;
>  #endif
> -	.head.text : {
> +	.head.text : AT(ADDR(.head.text) - LOAD_OFFSET) {
>  		_text = .;
>  		HEAD_TEXT
>  	}
> -	.text : {			/* Real text segment		*/
> +	.text : AT(ADDR(.text) - LOAD_OFFSET){	/* Real text segment	*/
>  		_stext = .;		/* Text and read-only data	*/
>  			__exception_text_start = .;
>  			*(.exception.text)
> @@ -132,12 +132,12 @@ SECTIONS
>  	 * Stack unwinding tables
>  	 */
>  	. = ALIGN(8);
> -	.ARM.unwind_idx : {
> +	.ARM.unwind_idx : AT(ADDR(.ARM.unwind_idx) - LOAD_OFFSET) {
>  		__start_unwind_idx = .;
>  		*(.ARM.exidx*)
>  		__stop_unwind_idx = .;
>  	}
> -	.ARM.unwind_tab : {
> +	.ARM.unwind_tab : AT(ADDR(.ARM.unwind_tab) - LOAD_OFFSET) {
>  		__start_unwind_tab = .;
>  		*(.ARM.extab*)
>  		__stop_unwind_tab = .;
> @@ -152,35 +152,35 @@ SECTIONS
>  #endif
>  
>  	INIT_TEXT_SECTION(8)
> -	.exit.text : {
> +	.exit.text : AT(ADDR(.exit.text) - LOAD_OFFSET) {
>  		ARM_EXIT_KEEP(EXIT_TEXT)
>  	}
> -	.init.proc.info : {
> +	.init.proc.info : AT(ADDR(.init.proc.info) - LOAD_OFFSET) {
>  		ARM_CPU_DISCARD(PROC_INFO)
>  	}
> -	.init.arch.info : {
> +	.init.arch.info : AT(ADDR(.init.arch.info) - LOAD_OFFSET) {
>  		__arch_info_begin = .;
>  		*(.arch.info.init)
>  		__arch_info_end = .;
>  	}
> -	.init.tagtable : {
> +	.init.tagtable : AT(ADDR(.init.tagtable) - LOAD_OFFSET) {
>  		__tagtable_begin = .;
>  		*(.taglist.init)
>  		__tagtable_end = .;
>  	}
>  #ifdef CONFIG_SMP_ON_UP
> -	.init.smpalt : {
> +	.init.smpalt : AT(ADDR(.init.smpalt) - LOAD_OFFSET) {
>  		__smpalt_begin = .;
>  		*(.alt.smp.init)
>  		__smpalt_end = .;
>  	}
>  #endif
> -	.init.pv_table : {
> +	.init.pv_table : AT(ADDR(.init.pv_table) - LOAD_OFFSET) {
>  		__pv_table_begin = .;
>  		*(.pv_table)
>  		__pv_table_end = .;
>  	}
> -	.init.data : {
> +	.init.data : AT(ADDR(.init.data) - LOAD_OFFSET) {
>  #ifndef CONFIG_XIP_KERNEL
>  		INIT_DATA
>  #endif
> @@ -191,7 +191,7 @@ SECTIONS
>  		INIT_RAM_FS
>  	}
>  #ifndef CONFIG_XIP_KERNEL
> -	.exit.data : {
> +	.exit.data : AT(ADDR(.exit.data) - LOAD_OFFSET) {
>  		ARM_EXIT_KEEP(EXIT_DATA)
>  	}
>  #endif
> @@ -207,7 +207,7 @@ SECTIONS
>  	__data_loc = .;
>  #endif
>  
> -	.data : AT(__data_loc) {
> +	.data : AT(ADDR(.data) - LOAD_OFFSET) {
>  		_data = .;		/* address in memory */
>  		_sdata = .;
>  
> -- 
> 1.7.0.4
> 

[RFC PATCH 7/8] ARM: vmlinux.lds: Setup physical load address not virtual

[Nicolas Pitre]

On Mon, 18 Jun 2012, Michal Simek wrote:

> Setup correct virtual and physical address in ELF LOAD section.

We are moving to a single kernel binary for multiple targets, including 
targets with different physical load addresses.  The kernel code figures 
out at run time what the actual physical address is when 
CONFIG_ARM_PATCH_PHYS_VIRT is set which is the default these days.  In 
other words, we don't know the physical offset at build time in that 
case and CONFIG_PHYS_OFFSET is simply not defined.

Why do you need this change?  Your patch comments are lacking 
justification for them.


> Signed-off-by: Michal Simek <monstr@monstr.eu>
> ---
>  arch/arm/include/asm/page.h   |    2 ++
>  arch/arm/kernel/vmlinux.lds.S |   28 ++++++++++++++--------------
>  2 files changed, 16 insertions(+), 14 deletions(-)
> 
> diff --git a/arch/arm/include/asm/page.h b/arch/arm/include/asm/page.h
> index ecf9019..4070209 100644
> --- a/arch/arm/include/asm/page.h
> +++ b/arch/arm/include/asm/page.h
> @@ -15,6 +15,8 @@
>  #define PAGE_SIZE		(_AC(1,UL) << PAGE_SHIFT)
>  #define PAGE_MASK		(~(PAGE_SIZE-1))
>  
> +#define LOAD_OFFSET	(CONFIG_PAGE_OFFSET - CONFIG_PHYS_OFFSET)
> +
>  #ifndef __ASSEMBLY__
>  
>  #ifndef CONFIG_MMU
> diff --git a/arch/arm/kernel/vmlinux.lds.S b/arch/arm/kernel/vmlinux.lds.S
> index c887be0..b996bc0 100644
> --- a/arch/arm/kernel/vmlinux.lds.S
> +++ b/arch/arm/kernel/vmlinux.lds.S
> @@ -3,11 +3,11 @@
>   * Written by Martin Mares <mj@atrey.karlin.mff.cuni.cz>
>   */
>  
> +#include <asm/page.h>
>  #include <asm-generic/vmlinux.lds.h>
>  #include <asm/cache.h>
>  #include <asm/thread_info.h>
>  #include <asm/memory.h>
> -#include <asm/page.h>
>  	
>  #define PROC_INFO							\
>  	. = ALIGN(4);							\
> @@ -86,11 +86,11 @@ SECTIONS
>  #else
>  	. = PAGE_OFFSET + TEXT_OFFSET;
>  #endif
> -	.head.text : {
> +	.head.text : AT(ADDR(.head.text) - LOAD_OFFSET) {
>  		_text = .;
>  		HEAD_TEXT
>  	}
> -	.text : {			/* Real text segment		*/
> +	.text : AT(ADDR(.text) - LOAD_OFFSET){	/* Real text segment	*/
>  		_stext = .;		/* Text and read-only data	*/
>  			__exception_text_start = .;
>  			*(.exception.text)
> @@ -132,12 +132,12 @@ SECTIONS
>  	 * Stack unwinding tables
>  	 */
>  	. = ALIGN(8);
> -	.ARM.unwind_idx : {
> +	.ARM.unwind_idx : AT(ADDR(.ARM.unwind_idx) - LOAD_OFFSET) {
>  		__start_unwind_idx = .;
>  		*(.ARM.exidx*)
>  		__stop_unwind_idx = .;
>  	}
> -	.ARM.unwind_tab : {
> +	.ARM.unwind_tab : AT(ADDR(.ARM.unwind_tab) - LOAD_OFFSET) {
>  		__start_unwind_tab = .;
>  		*(.ARM.extab*)
>  		__stop_unwind_tab = .;
> @@ -152,35 +152,35 @@ SECTIONS
>  #endif
>  
>  	INIT_TEXT_SECTION(8)
> -	.exit.text : {
> +	.exit.text : AT(ADDR(.exit.text) - LOAD_OFFSET) {
>  		ARM_EXIT_KEEP(EXIT_TEXT)
>  	}
> -	.init.proc.info : {
> +	.init.proc.info : AT(ADDR(.init.proc.info) - LOAD_OFFSET) {
>  		ARM_CPU_DISCARD(PROC_INFO)
>  	}
> -	.init.arch.info : {
> +	.init.arch.info : AT(ADDR(.init.arch.info) - LOAD_OFFSET) {
>  		__arch_info_begin = .;
>  		*(.arch.info.init)
>  		__arch_info_end = .;
>  	}
> -	.init.tagtable : {
> +	.init.tagtable : AT(ADDR(.init.tagtable) - LOAD_OFFSET) {
>  		__tagtable_begin = .;
>  		*(.taglist.init)
>  		__tagtable_end = .;
>  	}
>  #ifdef CONFIG_SMP_ON_UP
> -	.init.smpalt : {
> +	.init.smpalt : AT(ADDR(.init.smpalt) - LOAD_OFFSET) {
>  		__smpalt_begin = .;
>  		*(.alt.smp.init)
>  		__smpalt_end = .;
>  	}
>  #endif
> -	.init.pv_table : {
> +	.init.pv_table : AT(ADDR(.init.pv_table) - LOAD_OFFSET) {
>  		__pv_table_begin = .;
>  		*(.pv_table)
>  		__pv_table_end = .;
>  	}
> -	.init.data : {
> +	.init.data : AT(ADDR(.init.data) - LOAD_OFFSET) {
>  #ifndef CONFIG_XIP_KERNEL
>  		INIT_DATA
>  #endif
> @@ -191,7 +191,7 @@ SECTIONS
>  		INIT_RAM_FS
>  	}
>  #ifndef CONFIG_XIP_KERNEL
> -	.exit.data : {
> +	.exit.data : AT(ADDR(.exit.data) - LOAD_OFFSET) {
>  		ARM_EXIT_KEEP(EXIT_DATA)
>  	}
>  #endif
> @@ -207,7 +207,7 @@ SECTIONS
>  	__data_loc = .;
>  #endif
>  
> -	.data : AT(__data_loc) {
> +	.data : AT(ADDR(.data) - LOAD_OFFSET) {
>  		_data = .;		/* address in memory */
>  		_sdata = .;
>  
> -- 
> 1.7.0.4
> 

[RFC PATCH 7/8] ARM: vmlinux.lds: Setup physical load address not virtual

[Michal Simek]

2012/6/18 Nicolas Pitre <nicolas.pitre@linaro.org>

> On Mon, 18 Jun 2012, Michal Simek wrote:
>
> > Setup correct virtual and physical address in ELF LOAD section.
>
> We are moving to a single kernel binary for multiple targets, including
> targets with different physical load addresses.  The kernel code figures
> out at run time what the actual physical address is when
> CONFIG_ARM_PATCH_PHYS_VIRT is set which is the default these days.  In
> other words, we don't know the physical offset at build time in that
> case and CONFIG_PHYS_OFFSET is simply not defined.
>

ok. good to know and nice features. In that case I expect that you are
using any binary format
and you copy it to memory and run it. Code find out where it runs and based
on that setup things.
(BTW: This is very interesting for me to implement it on Microblaze. Can
you point me to that kernel code part? or any doc?)

But can you do it with elf? From my experience elf loaders are reading
physical address from program headers
and load all load sections. At the end they read entry point address and
setup PC on it and run it
(readelf -l -h). If these values are setup incorrectly to virtual addresses
elf loader just load them to virtual memory address
or just failed.


Why do you need this change?  Your patch comments are lacking
> justification for them.
>
> The main reason is to elf loaders.
Please correct me if I am wrong IRC ARM Qemu uses zImage but it is no
problem to use elf file. For this case
Qemu elf loader also uses physical addresses from ELF and also entry point.

There is one more thing in connection with this patch which is that exactly
the same description is done
in asm-generic/vmlinux.lds.h macros which arm lds.S includes. For example
INIT_TEXT_SECTION.

If you look at vmlinux before and after my changes you will see some
differences which shouldn't affect your binary format
at all. Look below to readelf outputs.

I haven't played with CONFIG_ARM_PATCH_PHYS_VIRT option but what is elf
entry point for this case?

Thanks,
Michal



Before:

ELF Header:
  Magic:   7f 45 4c 46 01 01 01 00 00 00 00 00 00 00 00 00
  Class:                             ELF32
  Data:                              2's complement, little endian
  Version:                           1 (current)
  OS/ABI:                            UNIX - System V
  ABI Version:                       0
  Type:                              EXEC (Executable file)
  Machine:                           ARM
  Version:                           0x1
  Entry point address:               0x80008000
  Start of program headers:          52 (bytes into file)
  Start of section headers:          84873104 (bytes into file)
  Flags:                             0x5000002, has entry point, Version5
EABI
  Size of this header:               52 (bytes)
  Size of program headers:           32 (bytes)
  Number of program headers:         3
  Size of section headers:           40 (bytes)
  Number of section headers:         37
  Section header string table index: 34

Program Headers:
  Type           Offset   VirtAddr   PhysAddr   FileSiz MemSiz  Flg Align
  LOAD           0x008000 0x80008000 0x80008000 0x254a204 0x255e814 RWE
0x8000
  NOTE           0x25521e0 0x825521e0 0x825521e0 0x00024 0x00024 R E 0x4
  GNU_STACK      0x000000 0x00000000 0x00000000 0x00000 0x00000 RWE 0x4


After:

ELF Header:
  Magic:   7f 45 4c 46 01 01 01 00 00 00 00 00 00 00 00 00
  Class:                             ELF32
  Data:                              2's complement, little endian
  Version:                           1 (current)
  OS/ABI:                            UNIX - System V
  ABI Version:                       0
  Type:                              EXEC (Executable file)
  Machine:                           ARM
  Version:                           0x1
  Entry point address:               0x10008000
  Start of program headers:          52 (bytes into file)
  Start of section headers:          84873088 (bytes into file)
  Flags:                             0x5000002, has entry point, Version5
EABI
  Size of this header:               52 (bytes)
  Size of program headers:           32 (bytes)
  Number of program headers:         3
  Size of section headers:           40 (bytes)
  Number of section headers:         37
  Section header string table index: 34

Program Headers:
  Type           Offset   VirtAddr   PhysAddr   FileSiz MemSiz  Flg Align
  LOAD           0x008000 0x80008000 0x10008000 0x254a204 0x255e814 RWE
0x8000
  NOTE           0x25521e0 0x825521e0 0x125521e0 0x00024 0x00024 R E 0x4
  GNU_STACK      0x000000 0x00000000 0x00000000 0x00000 0x00000 RWE 0x4
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20120618/f66f580d/attachment-0001.html>

Re: [RFC PATCH 7/8] ARM: vmlinux.lds: Setup physical load address not virtual

[Nicolas Pitre]

On Mon, 18 Jun 2012, Michal Simek wrote:

> 2012/6/18 Nicolas Pitre <nicolas.pitre@linaro.org>
> 
> > On Mon, 18 Jun 2012, Michal Simek wrote:
> >
> > > Setup correct virtual and physical address in ELF LOAD section.
> >
> > We are moving to a single kernel binary for multiple targets, including
> > targets with different physical load addresses.  The kernel code figures
> > out at run time what the actual physical address is when
> > CONFIG_ARM_PATCH_PHYS_VIRT is set which is the default these days.  In
> > other words, we don't know the physical offset at build time in that
> > case and CONFIG_PHYS_OFFSET is simply not defined.
> >
> 
> ok. good to know and nice features. In that case I expect that you are 
> using any binary format and you copy it to memory and run it. Code 
> find out where it runs and based on that setup things.

Exact.

> (BTW: This is very interesting for me to implement it on Microblaze. Can
> you point me to that kernel code part? or any doc?)

You could have a look at https://wiki.ubuntu.com/Specs/ARMSingleKernel, 
especially the section "Optimized virt_to_phys() with a runtime 
determined PHYS_OFFSET".

Then the following commits should be of interest:

72a20e22f4 "ARM: P2V: eliminate head.S use of PHYS_OFFSET for !XIP_KERNEL"

dc21af99fa "ARM: P2V: introduce phys_to_virt/virt_to_phys runtime patching"

> But can you do it with elf?

No.

> Please correct me if I am wrong IRC ARM Qemu uses zImage but it is no
> problem to use elf file. For this case
> Qemu elf loader also uses physical addresses from ELF and also entry point.

If you want to use the ELF image, then you need to hardcode the physical 
memory location in its header, and that's something we want to get away 
from because it prevents a single kernel binary image from being used on 
different targets with different memory layouts.  Same issue with 
u-Boot's legacy uImage format.

> I haven't played with CONFIG_ARM_PATCH_PHYS_VIRT option but what is elf
> entry point for this case?

For the kernel image itself it is the kernel virtual address.

For the relocatable zImage decompressor wrapper it is 0.


Nicolas

[RFC PATCH 7/8] ARM: vmlinux.lds: Setup physical load address not virtual

[Nicolas Pitre]

On Mon, 18 Jun 2012, Michal Simek wrote:

> 2012/6/18 Nicolas Pitre <nicolas.pitre@linaro.org>
> 
> > On Mon, 18 Jun 2012, Michal Simek wrote:
> >
> > > Setup correct virtual and physical address in ELF LOAD section.
> >
> > We are moving to a single kernel binary for multiple targets, including
> > targets with different physical load addresses.  The kernel code figures
> > out at run time what the actual physical address is when
> > CONFIG_ARM_PATCH_PHYS_VIRT is set which is the default these days.  In
> > other words, we don't know the physical offset at build time in that
> > case and CONFIG_PHYS_OFFSET is simply not defined.
> >
> 
> ok. good to know and nice features. In that case I expect that you are 
> using any binary format and you copy it to memory and run it. Code 
> find out where it runs and based on that setup things.

Exact.

> (BTW: This is very interesting for me to implement it on Microblaze. Can
> you point me to that kernel code part? or any doc?)

You could have a look at https://wiki.ubuntu.com/Specs/ARMSingleKernel, 
especially the section "Optimized virt_to_phys() with a runtime 
determined PHYS_OFFSET".

Then the following commits should be of interest:

72a20e22f4 "ARM: P2V: eliminate head.S use of PHYS_OFFSET for !XIP_KERNEL"

dc21af99fa "ARM: P2V: introduce phys_to_virt/virt_to_phys runtime patching"

> But can you do it with elf?

No.

> Please correct me if I am wrong IRC ARM Qemu uses zImage but it is no
> problem to use elf file. For this case
> Qemu elf loader also uses physical addresses from ELF and also entry point.

If you want to use the ELF image, then you need to hardcode the physical 
memory location in its header, and that's something we want to get away 
from because it prevents a single kernel binary image from being used on 
different targets with different memory layouts.  Same issue with 
u-Boot's legacy uImage format.

> I haven't played with CONFIG_ARM_PATCH_PHYS_VIRT option but what is elf
> entry point for this case?

For the kernel image itself it is the kernel virtual address.

For the relocatable zImage decompressor wrapper it is 0.


Nicolas

Re: [RFC PATCH 7/8] ARM: vmlinux.lds: Setup physical load address not virtual

[Michal Simek]

On 06/18/2012 09:02 PM, Nicolas Pitre wrote:
> On Mon, 18 Jun 2012, Michal Simek wrote:
>
>> 2012/6/18 Nicolas Pitre<nicolas.pitre@linaro.org>
>>
>>> On Mon, 18 Jun 2012, Michal Simek wrote:
>>>
>>>> Setup correct virtual and physical address in ELF LOAD section.
>>>
>>> We are moving to a single kernel binary for multiple targets, including
>>> targets with different physical load addresses.  The kernel code figures
>>> out at run time what the actual physical address is when
>>> CONFIG_ARM_PATCH_PHYS_VIRT is set which is the default these days.  In
>>> other words, we don't know the physical offset at build time in that
>>> case and CONFIG_PHYS_OFFSET is simply not defined.
>>>
>>
>> ok. good to know and nice features. In that case I expect that you are
>> using any binary format and you copy it to memory and run it. Code
>> find out where it runs and based on that setup things.
>
> Exact.

ok. One more question about this feature. What is the alignment for it?
I mean I have done the similar feature for Xilinx ppc440 and alignment depends on
MMU TLB size.

>
>> (BTW: This is very interesting for me to implement it on Microblaze. Can
>> you point me to that kernel code part? or any doc?)
>
> You could have a look at https://wiki.ubuntu.com/Specs/ARMSingleKernel,
> especially the section "Optimized virt_to_phys() with a runtime
> determined PHYS_OFFSET".
>
> Then the following commits should be of interest:
>
> 72a20e22f4 "ARM: P2V: eliminate head.S use of PHYS_OFFSET for !XIP_KERNEL"
>
> dc21af99fa "ARM: P2V: introduce phys_to_virt/virt_to_phys runtime patching"

ok. Thanks will look at it.

>
>> But can you do it with elf?
>
> No.

ok

>
>> Please correct me if I am wrong IRC ARM Qemu uses zImage but it is no
>> problem to use elf file. For this case
>> Qemu elf loader also uses physical addresses from ELF and also entry point.
>
> If you want to use the ELF image, then you need to hardcode the physical
> memory location in its header, and that's something we want to get away
> from because it prevents a single kernel binary image from being used on
> different targets with different memory layouts.  Same issue with
> u-Boot's legacy uImage format.

And are you able to get these information away from ELF? Or you just need to set it up
to any default value. I would expect to 0x0.


>
>> I haven't played with CONFIG_ARM_PATCH_PHYS_VIRT option but what is elf
>> entry point for this case?
>
> For the kernel image itself it is the kernel virtual address.

Is it used somewhere? Or is it just setup to virtual address and no code use it?
I mean that kernel virtual entry point.

I expect that if CONFIG_ARM_PATCH_PHYS_VIRT=y then CONFIG_PHYS_OFFSET is not setup
Then LOAD_OFFSET should be equal to CONFIG_PAGE_OFFSET.

>
> For the relocatable zImage decompressor wrapper it is 0.

ok.

Thanks,
Michal


-- 
Michal Simek, Ing. (M.Eng)
w: www.monstr.eu p: +42-0-721842854
Maintainer of Linux kernel 2.6 Microblaze Linux - http://www.monstr.eu/fdt/
Microblaze U-BOOT custodian

[RFC PATCH 7/8] ARM: vmlinux.lds: Setup physical load address not virtual

[Michal Simek]

On 06/18/2012 09:02 PM, Nicolas Pitre wrote:
> On Mon, 18 Jun 2012, Michal Simek wrote:
>
>> 2012/6/18 Nicolas Pitre<nicolas.pitre@linaro.org>
>>
>>> On Mon, 18 Jun 2012, Michal Simek wrote:
>>>
>>>> Setup correct virtual and physical address in ELF LOAD section.
>>>
>>> We are moving to a single kernel binary for multiple targets, including
>>> targets with different physical load addresses.  The kernel code figures
>>> out at run time what the actual physical address is when
>>> CONFIG_ARM_PATCH_PHYS_VIRT is set which is the default these days.  In
>>> other words, we don't know the physical offset at build time in that
>>> case and CONFIG_PHYS_OFFSET is simply not defined.
>>>
>>
>> ok. good to know and nice features. In that case I expect that you are
>> using any binary format and you copy it to memory and run it. Code
>> find out where it runs and based on that setup things.
>
> Exact.

ok. One more question about this feature. What is the alignment for it?
I mean I have done the similar feature for Xilinx ppc440 and alignment depends on
MMU TLB size.

>
>> (BTW: This is very interesting for me to implement it on Microblaze. Can
>> you point me to that kernel code part? or any doc?)
>
> You could have a look at https://wiki.ubuntu.com/Specs/ARMSingleKernel,
> especially the section "Optimized virt_to_phys() with a runtime
> determined PHYS_OFFSET".
>
> Then the following commits should be of interest:
>
> 72a20e22f4 "ARM: P2V: eliminate head.S use of PHYS_OFFSET for !XIP_KERNEL"
>
> dc21af99fa "ARM: P2V: introduce phys_to_virt/virt_to_phys runtime patching"

ok. Thanks will look at it.

>
>> But can you do it with elf?
>
> No.

ok

>
>> Please correct me if I am wrong IRC ARM Qemu uses zImage but it is no
>> problem to use elf file. For this case
>> Qemu elf loader also uses physical addresses from ELF and also entry point.
>
> If you want to use the ELF image, then you need to hardcode the physical
> memory location in its header, and that's something we want to get away
> from because it prevents a single kernel binary image from being used on
> different targets with different memory layouts.  Same issue with
> u-Boot's legacy uImage format.

And are you able to get these information away from ELF? Or you just need to set it up
to any default value. I would expect to 0x0.


>
>> I haven't played with CONFIG_ARM_PATCH_PHYS_VIRT option but what is elf
>> entry point for this case?
>
> For the kernel image itself it is the kernel virtual address.

Is it used somewhere? Or is it just setup to virtual address and no code use it?
I mean that kernel virtual entry point.

I expect that if CONFIG_ARM_PATCH_PHYS_VIRT=y then CONFIG_PHYS_OFFSET is not setup
Then LOAD_OFFSET should be equal to CONFIG_PAGE_OFFSET.

>
> For the relocatable zImage decompressor wrapper it is 0.

ok.

Thanks,
Michal


-- 
Michal Simek, Ing. (M.Eng)
w: www.monstr.eu p: +42-0-721842854
Maintainer of Linux kernel 2.6 Microblaze Linux - http://www.monstr.eu/fdt/
Microblaze U-BOOT custodian

Re: [RFC PATCH 7/8] ARM: vmlinux.lds: Setup physical load address not virtual

[Nicolas Pitre]

On Mon, 18 Jun 2012, Michal Simek wrote:

> On 06/18/2012 09:02 PM, Nicolas Pitre wrote:
> > On Mon, 18 Jun 2012, Michal Simek wrote:
> > 
> > > 2012/6/18 Nicolas Pitre<nicolas.pitre@linaro.org>
> > > 
> > > > On Mon, 18 Jun 2012, Michal Simek wrote:
> > > > 
> > > > > Setup correct virtual and physical address in ELF LOAD section.
> > > > 
> > > > We are moving to a single kernel binary for multiple targets, including
> > > > targets with different physical load addresses.  The kernel code figures
> > > > out at run time what the actual physical address is when
> > > > CONFIG_ARM_PATCH_PHYS_VIRT is set which is the default these days.  In
> > > > other words, we don't know the physical offset at build time in that
> > > > case and CONFIG_PHYS_OFFSET is simply not defined.
> > > > 
> > > 
> > > ok. good to know and nice features. In that case I expect that you are
> > > using any binary format and you copy it to memory and run it. Code
> > > find out where it runs and based on that setup things.
> > 
> > Exact.
> 
> ok. One more question about this feature. What is the alignment for 
> it? I mean I have done the similar feature for Xilinx ppc440 and 
> alignment depends on MMU TLB size.

Currently the alignment is 16MB as this is the granularity that a single 
add or sub instruction can cover with an immediate argument.  This is 
highly specific to the ARM instruction set.  At some point we needed 2MB 
alignment which required two adds or subs, but that requirement has now 
been removed.

> > If you want to use the ELF image, then you need to hardcode the physical
> > memory location in its header, and that's something we want to get away
> > from because it prevents a single kernel binary image from being used on
> > different targets with different memory layouts.  Same issue with
> > u-Boot's legacy uImage format.
> 
> And are you able to get these information away from ELF? Or you just 
> need to set it up to any default value. I would expect to 0x0.

We just don't use the ELF image format with bootloaders.

> > > I haven't played with CONFIG_ARM_PATCH_PHYS_VIRT option but what is elf
> > > entry point for this case?
> > 
> > For the kernel image itself it is the kernel virtual address.
> 
> Is it used somewhere? Or is it just setup to virtual address and no 
> code use it? I mean that kernel virtual entry point.

Nothing uses it.  The official ARM kernel image format is a binary image 
and its entry point is the very beginning of it.

> I expect that if CONFIG_ARM_PATCH_PHYS_VIRT=y then CONFIG_PHYS_OFFSET 
> is not setup Then LOAD_OFFSET should be equal to CONFIG_PAGE_OFFSET.

No. CONFIG_PAGE_OFFSET is the virtual address that the kernel uses for 
the beginning of its memory view.  That is not where the kernel is 
loaded.


Nicolas

[RFC PATCH 7/8] ARM: vmlinux.lds: Setup physical load address not virtual

[Nicolas Pitre]

On Mon, 18 Jun 2012, Michal Simek wrote:

> On 06/18/2012 09:02 PM, Nicolas Pitre wrote:
> > On Mon, 18 Jun 2012, Michal Simek wrote:
> > 
> > > 2012/6/18 Nicolas Pitre<nicolas.pitre@linaro.org>
> > > 
> > > > On Mon, 18 Jun 2012, Michal Simek wrote:
> > > > 
> > > > > Setup correct virtual and physical address in ELF LOAD section.
> > > > 
> > > > We are moving to a single kernel binary for multiple targets, including
> > > > targets with different physical load addresses.  The kernel code figures
> > > > out at run time what the actual physical address is when
> > > > CONFIG_ARM_PATCH_PHYS_VIRT is set which is the default these days.  In
> > > > other words, we don't know the physical offset at build time in that
> > > > case and CONFIG_PHYS_OFFSET is simply not defined.
> > > > 
> > > 
> > > ok. good to know and nice features. In that case I expect that you are
> > > using any binary format and you copy it to memory and run it. Code
> > > find out where it runs and based on that setup things.
> > 
> > Exact.
> 
> ok. One more question about this feature. What is the alignment for 
> it? I mean I have done the similar feature for Xilinx ppc440 and 
> alignment depends on MMU TLB size.

Currently the alignment is 16MB as this is the granularity that a single 
add or sub instruction can cover with an immediate argument.  This is 
highly specific to the ARM instruction set.  At some point we needed 2MB 
alignment which required two adds or subs, but that requirement has now 
been removed.

> > If you want to use the ELF image, then you need to hardcode the physical
> > memory location in its header, and that's something we want to get away
> > from because it prevents a single kernel binary image from being used on
> > different targets with different memory layouts.  Same issue with
> > u-Boot's legacy uImage format.
> 
> And are you able to get these information away from ELF? Or you just 
> need to set it up to any default value. I would expect to 0x0.

We just don't use the ELF image format with bootloaders.

> > > I haven't played with CONFIG_ARM_PATCH_PHYS_VIRT option but what is elf
> > > entry point for this case?
> > 
> > For the kernel image itself it is the kernel virtual address.
> 
> Is it used somewhere? Or is it just setup to virtual address and no 
> code use it? I mean that kernel virtual entry point.

Nothing uses it.  The official ARM kernel image format is a binary image 
and its entry point is the very beginning of it.

> I expect that if CONFIG_ARM_PATCH_PHYS_VIRT=y then CONFIG_PHYS_OFFSET 
> is not setup Then LOAD_OFFSET should be equal to CONFIG_PAGE_OFFSET.

No. CONFIG_PAGE_OFFSET is the virtual address that the kernel uses for 
the beginning of its memory view.  That is not where the kernel is 
loaded.


Nicolas

[RFC PATCH 7/8] ARM: vmlinux.lds: Setup physical load address not virtual

[Michal Simek]

2012/6/18 Nicolas Pitre <nicolas.pitre@linaro.org>

> On Mon, 18 Jun 2012, Michal Simek wrote:
>
> > On 06/18/2012 09:02 PM, Nicolas Pitre wrote:
> > > On Mon, 18 Jun 2012, Michal Simek wrote:
> > >
> > > > 2012/6/18 Nicolas Pitre<nicolas.pitre@linaro.org>
> > > >
> > > > > On Mon, 18 Jun 2012, Michal Simek wrote:
> > > > >
> > > > > > Setup correct virtual and physical address in ELF LOAD section.
> > > > >
> > > > > We are moving to a single kernel binary for multiple targets,
> including
> > > > > targets with different physical load addresses.  The kernel code
> figures
> > > > > out at run time what the actual physical address is when
> > > > > CONFIG_ARM_PATCH_PHYS_VIRT is set which is the default these days.
>  In
> > > > > other words, we don't know the physical offset at build time in
> that
> > > > > case and CONFIG_PHYS_OFFSET is simply not defined.
> > > > >
> > > >
> > > > ok. good to know and nice features. In that case I expect that you
> are
> > > > using any binary format and you copy it to memory and run it. Code
> > > > find out where it runs and based on that setup things.
> > >
> > > Exact.
> >
> > ok. One more question about this feature. What is the alignment for
> > it? I mean I have done the similar feature for Xilinx ppc440 and
> > alignment depends on MMU TLB size.
>
> Currently the alignment is 16MB as this is the granularity that a single
> add or sub instruction can cover with an immediate argument.  This is
> highly specific to the ARM instruction set.  At some point we needed 2MB
> alignment which required two adds or subs, but that requirement has now
> been removed.
>

ok.


>
> > > If you want to use the ELF image, then you need to hardcode the
> physical
> > > memory location in its header, and that's something we want to get away
> > > from because it prevents a single kernel binary image from being used
> on
> > > different targets with different memory layouts.  Same issue with
> > > u-Boot's legacy uImage format.
> >
> > And are you able to get these information away from ELF? Or you just
> > need to set it up to any default value. I would expect to 0x0.
>
> We just don't use the ELF image format with bootloaders.
>


ok


>
> > > > I haven't played with CONFIG_ARM_PATCH_PHYS_VIRT option but what is
> elf
> > > > entry point for this case?
> > >
> > > For the kernel image itself it is the kernel virtual address.
> >
> > Is it used somewhere? Or is it just setup to virtual address and no
> > code use it? I mean that kernel virtual entry point.
>
> Nothing uses it.  The official ARM kernel image format is a binary image
> and its entry point is the very beginning of it.
>

ok.


>
> > I expect that if CONFIG_ARM_PATCH_PHYS_VIRT=y then CONFIG_PHYS_OFFSET
> > is not setup Then LOAD_OFFSET should be equal to CONFIG_PAGE_OFFSET.
>
> No. CONFIG_PAGE_OFFSET is the virtual address that the kernel uses for
> the beginning of its memory view.  That is not where the kernel is
> loaded.
>
>
ok. Correct. For ARM_PATCH_PHYS_VIRT default value should be used.

Thanks,
Michal
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20120619/d02b9795/attachment-0001.html>

[RFC PATCH 8/8] ARM: vmlinux.lds: Setup correct entry point to physical address

[Michal Simek]

linker script setups entry point to virtual address.

Signed-off-by: Michal Simek <monstr@monstr.eu>

---
Not sure what is the reason to setup entry point to virtual address.
---
 arch/arm/kernel/vmlinux.lds.S |    4 +++-
 1 files changed, 3 insertions(+), 1 deletions(-)

diff --git a/arch/arm/kernel/vmlinux.lds.S b/arch/arm/kernel/vmlinux.lds.S
index b996bc0..de43b01 100644
--- a/arch/arm/kernel/vmlinux.lds.S
+++ b/arch/arm/kernel/vmlinux.lds.S
@@ -39,7 +39,7 @@
 #endif
 
 OUTPUT_ARCH(arm)
-ENTRY(stext)
+ENTRY(arm_start)
 
 #ifndef __ARMEB__
 jiffies = jiffies_64;
@@ -83,8 +83,10 @@ SECTIONS
 
 #ifdef CONFIG_XIP_KERNEL
 	. = XIP_VIRT_ADDR(CONFIG_XIP_PHYS_ADDR);
+	arm_start = . ;
 #else
 	. = PAGE_OFFSET + TEXT_OFFSET;
+	arm_start = CONFIG_PHYS_OFFSET + TEXT_OFFSET;
 #endif
 	.head.text : AT(ADDR(.head.text) - LOAD_OFFSET) {
 		_text = .;
-- 
1.7.0.4

[RFC PATCH 8/8] ARM: vmlinux.lds: Setup correct entry point to physical address

[Michal Simek]

linker script setups entry point to virtual address.

Signed-off-by: Michal Simek <monstr@monstr.eu>

---
Not sure what is the reason to setup entry point to virtual address.
---
 arch/arm/kernel/vmlinux.lds.S |    4 +++-
 1 files changed, 3 insertions(+), 1 deletions(-)

diff --git a/arch/arm/kernel/vmlinux.lds.S b/arch/arm/kernel/vmlinux.lds.S
index b996bc0..de43b01 100644
--- a/arch/arm/kernel/vmlinux.lds.S
+++ b/arch/arm/kernel/vmlinux.lds.S
@@ -39,7 +39,7 @@
 #endif
 
 OUTPUT_ARCH(arm)
-ENTRY(stext)
+ENTRY(arm_start)
 
 #ifndef __ARMEB__
 jiffies = jiffies_64;
@@ -83,8 +83,10 @@ SECTIONS
 
 #ifdef CONFIG_XIP_KERNEL
 	. = XIP_VIRT_ADDR(CONFIG_XIP_PHYS_ADDR);
+	arm_start = . ;
 #else
 	. = PAGE_OFFSET + TEXT_OFFSET;
+	arm_start = CONFIG_PHYS_OFFSET + TEXT_OFFSET;
 #endif
 	.head.text : AT(ADDR(.head.text) - LOAD_OFFSET) {
 		_text = .;
-- 
1.7.0.4

Re: [RFC PATCH 8/8] ARM: vmlinux.lds: Setup correct entry point to physical address

[Russell King - ARM Linux]

On Mon, Jun 18, 2012 at 01:30:11PM +0200, Michal Simek wrote:
> linker script setups entry point to virtual address.
> 
> Signed-off-by: Michal Simek <monstr@monstr.eu>
> 
> ---
> Not sure what is the reason to setup entry point to virtual address.
> ---
>  arch/arm/kernel/vmlinux.lds.S |    4 +++-
>  1 files changed, 3 insertions(+), 1 deletions(-)
> 
> diff --git a/arch/arm/kernel/vmlinux.lds.S b/arch/arm/kernel/vmlinux.lds.S
> index b996bc0..de43b01 100644
> --- a/arch/arm/kernel/vmlinux.lds.S
> +++ b/arch/arm/kernel/vmlinux.lds.S
> @@ -39,7 +39,7 @@
>  #endif
>  
>  OUTPUT_ARCH(arm)
> -ENTRY(stext)
> +ENTRY(arm_start)

stext etc better stay as the virtual address through this change.  The
kernel makes use of all stext/etext etc symbols as virtual addresses.

[RFC PATCH 8/8] ARM: vmlinux.lds: Setup correct entry point to physical address

[Russell King - ARM Linux]

On Mon, Jun 18, 2012 at 01:30:11PM +0200, Michal Simek wrote:
> linker script setups entry point to virtual address.
> 
> Signed-off-by: Michal Simek <monstr@monstr.eu>
> 
> ---
> Not sure what is the reason to setup entry point to virtual address.
> ---
>  arch/arm/kernel/vmlinux.lds.S |    4 +++-
>  1 files changed, 3 insertions(+), 1 deletions(-)
> 
> diff --git a/arch/arm/kernel/vmlinux.lds.S b/arch/arm/kernel/vmlinux.lds.S
> index b996bc0..de43b01 100644
> --- a/arch/arm/kernel/vmlinux.lds.S
> +++ b/arch/arm/kernel/vmlinux.lds.S
> @@ -39,7 +39,7 @@
>  #endif
>  
>  OUTPUT_ARCH(arm)
> -ENTRY(stext)
> +ENTRY(arm_start)

stext etc better stay as the virtual address through this change.  The
kernel makes use of all stext/etext etc symbols as virtual addresses.

Re: [RFC PATCH 8/8] ARM: vmlinux.lds: Setup correct entry point to physical address

[Michal Simek]

On 06/18/2012 01:37 PM, Russell King - ARM Linux wrote:
> On Mon, Jun 18, 2012 at 01:30:11PM +0200, Michal Simek wrote:
>> linker script setups entry point to virtual address.
>>
>> Signed-off-by: Michal Simek<monstr@monstr.eu>
>>
>> ---
>> Not sure what is the reason to setup entry point to virtual address.
>> ---
>>   arch/arm/kernel/vmlinux.lds.S |    4 +++-
>>   1 files changed, 3 insertions(+), 1 deletions(-)
>>
>> diff --git a/arch/arm/kernel/vmlinux.lds.S b/arch/arm/kernel/vmlinux.lds.S
>> index b996bc0..de43b01 100644
>> --- a/arch/arm/kernel/vmlinux.lds.S
>> +++ b/arch/arm/kernel/vmlinux.lds.S
>> @@ -39,7 +39,7 @@
>>   #endif
>>
>>   OUTPUT_ARCH(arm)
>> -ENTRY(stext)
>> +ENTRY(arm_start)
>
> stext etc better stay as the virtual address through this change.  The
> kernel makes use of all stext/etext etc symbols as virtual addresses.

I am not changing stext to physical address. Just elf entry point.
What's it reason to have entry point in virtual address space?

Thanks,
Michal


-- 
Michal Simek, Ing. (M.Eng)
w: www.monstr.eu p: +42-0-721842854
Maintainer of Linux kernel 2.6 Microblaze Linux - http://www.monstr.eu/fdt/
Microblaze U-BOOT custodian

[RFC PATCH 8/8] ARM: vmlinux.lds: Setup correct entry point to physical address

[Michal Simek]

On 06/18/2012 01:37 PM, Russell King - ARM Linux wrote:
> On Mon, Jun 18, 2012 at 01:30:11PM +0200, Michal Simek wrote:
>> linker script setups entry point to virtual address.
>>
>> Signed-off-by: Michal Simek<monstr@monstr.eu>
>>
>> ---
>> Not sure what is the reason to setup entry point to virtual address.
>> ---
>>   arch/arm/kernel/vmlinux.lds.S |    4 +++-
>>   1 files changed, 3 insertions(+), 1 deletions(-)
>>
>> diff --git a/arch/arm/kernel/vmlinux.lds.S b/arch/arm/kernel/vmlinux.lds.S
>> index b996bc0..de43b01 100644
>> --- a/arch/arm/kernel/vmlinux.lds.S
>> +++ b/arch/arm/kernel/vmlinux.lds.S
>> @@ -39,7 +39,7 @@
>>   #endif
>>
>>   OUTPUT_ARCH(arm)
>> -ENTRY(stext)
>> +ENTRY(arm_start)
>
> stext etc better stay as the virtual address through this change.  The
> kernel makes use of all stext/etext etc symbols as virtual addresses.

I am not changing stext to physical address. Just elf entry point.
What's it reason to have entry point in virtual address space?

Thanks,
Michal


-- 
Michal Simek, Ing. (M.Eng)
w: www.monstr.eu p: +42-0-721842854
Maintainer of Linux kernel 2.6 Microblaze Linux - http://www.monstr.eu/fdt/
Microblaze U-BOOT custodian