[PATCH] sparseirq: Enable early irq_desc allocation.

[PATCH] sparseirq: Enable early irq_desc allocation.

[Paul Mundt]

Presently non-legacy IRQs have their irq_desc allocated with
kzalloc_node(). This assumes that all callers of irq_to_desc_cpu_alloc()
will be sufficiently late in the boot process that kmalloc is available.

While porting sparseirq support to sh this blew up immediately, as at the
time that we register the CPU's interrupt vector map only bootmem is
available.

This adds in a simple after_bootmem check to see where the allocation
needs to come from, which is likewise provided by all of the platforms
that support sparse irq today. :-)

Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>

---

 kernel/irq/handle.c |   17 ++++++++++++-----
 1 file changed, 12 insertions(+), 5 deletions(-)

diff --git a/kernel/irq/handle.c b/kernel/irq/handle.c
index d82142b..5fb3a5c 100644
--- a/kernel/irq/handle.c
+++ b/kernel/irq/handle.c
@@ -19,7 +19,7 @@
 #include <linux/hash.h>
 #include <trace/irq.h>
 #include <linux/bootmem.h>
-
+#include <linux/mm.h>
 #include "internals.h"
 
 /*
@@ -81,13 +81,17 @@ static struct irq_desc irq_desc_init = {
 	.lock       = __SPIN_LOCK_UNLOCKED(irq_desc_init.lock),
 };
 
-void init_kstat_irqs(struct irq_desc *desc, int cpu, int nr)
+void __ref init_kstat_irqs(struct irq_desc *desc, int cpu, int nr)
 {
 	int node;
 	void *ptr;
 
 	node = cpu_to_node(cpu);
-	ptr = kzalloc_node(nr * sizeof(*desc->kstat_irqs), GFP_ATOMIC, node);
+	if (after_bootmem)
+		ptr = kzalloc_node(nr * sizeof(*desc->kstat_irqs), GFP_ATOMIC, node);
+	else
+		ptr = alloc_bootmem_node(NODE_DATA(node),
+				nr * sizeof(*desc->kstat_irqs));
 
 	/*
 	 * don't overwite if can not get new one
@@ -187,7 +191,7 @@ struct irq_desc *irq_to_desc(unsigned int irq)
 	return NULL;
 }
 
-struct irq_desc *irq_to_desc_alloc_cpu(unsigned int irq, int cpu)
+struct irq_desc * __ref irq_to_desc_alloc_cpu(unsigned int irq, int cpu)
 {
 	struct irq_desc *desc;
 	unsigned long flags;
@@ -211,7 +215,10 @@ struct irq_desc *irq_to_desc_alloc_cpu(unsigned int irq, int cpu)
 		goto out_unlock;
 
 	node = cpu_to_node(cpu);
-	desc = kzalloc_node(sizeof(*desc), GFP_ATOMIC, node);
+	if (after_bootmem)
+		desc = kzalloc_node(sizeof(*desc), GFP_ATOMIC, node);
+	else
+		desc = alloc_bootmem_node(NODE_DATA(node), sizeof(*desc));
 	printk(KERN_DEBUG "  alloc irq_desc for %d on cpu %d node %d\n",
 		 irq, cpu, node);
 	if (!desc) {

Re: [PATCH] sparseirq: Enable early irq_desc allocation.

[Yinghai Lu]

Paul Mundt wrote:
> Presently non-legacy IRQs have their irq_desc allocated with
> kzalloc_node(). This assumes that all callers of irq_to_desc_cpu_alloc()
> will be sufficiently late in the boot process that kmalloc is available.
> 
> While porting sparseirq support to sh this blew up immediately, as at the
> time that we register the CPU's interrupt vector map only bootmem is
> available.
> 
> This adds in a simple after_bootmem check to see where the allocation
> needs to come from, which is likewise provided by all of the platforms
> that support sparse irq today. :-)
> 
> Cc: Yinghai Lu <yinghai@kernel.org>
> Cc: Ingo Molnar <mingo@elte.hu>
> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
> 
> ---
> 
>  kernel/irq/handle.c |   17 ++++++++++++-----
>  1 file changed, 12 insertions(+), 5 deletions(-)
> 
> diff --git a/kernel/irq/handle.c b/kernel/irq/handle.c
> index d82142b..5fb3a5c 100644
> --- a/kernel/irq/handle.c
> +++ b/kernel/irq/handle.c
> @@ -19,7 +19,7 @@
>  #include <linux/hash.h>
>  #include <trace/irq.h>
>  #include <linux/bootmem.h>
> -
> +#include <linux/mm.h>
>  #include "internals.h"
>  
>  /*
> @@ -81,13 +81,17 @@ static struct irq_desc irq_desc_init = {
>  	.lock       = __SPIN_LOCK_UNLOCKED(irq_desc_init.lock),
>  };
>  
> -void init_kstat_irqs(struct irq_desc *desc, int cpu, int nr)
> +void __ref init_kstat_irqs(struct irq_desc *desc, int cpu, int nr)
>  {
>  	int node;
>  	void *ptr;
>  
>  	node = cpu_to_node(cpu);
> -	ptr = kzalloc_node(nr * sizeof(*desc->kstat_irqs), GFP_ATOMIC, node);
> +	if (after_bootmem)
> +		ptr = kzalloc_node(nr * sizeof(*desc->kstat_irqs), GFP_ATOMIC, node);
> +	else
> +		ptr = alloc_bootmem_node(NODE_DATA(node),
> +				nr * sizeof(*desc->kstat_irqs));
>  
>  	/*
>  	 * don't overwite if can not get new one
> @@ -187,7 +191,7 @@ struct irq_desc *irq_to_desc(unsigned int irq)
>  	return NULL;
>  }
>  
> -struct irq_desc *irq_to_desc_alloc_cpu(unsigned int irq, int cpu)
> +struct irq_desc * __ref irq_to_desc_alloc_cpu(unsigned int irq, int cpu)
>  {
>  	struct irq_desc *desc;
>  	unsigned long flags;
> @@ -211,7 +215,10 @@ struct irq_desc *irq_to_desc_alloc_cpu(unsigned int irq, int cpu)
>  		goto out_unlock;
>  
>  	node = cpu_to_node(cpu);
> -	desc = kzalloc_node(sizeof(*desc), GFP_ATOMIC, node);
> +	if (after_bootmem)
> +		desc = kzalloc_node(sizeof(*desc), GFP_ATOMIC, node);
> +	else
> +		desc = alloc_bootmem_node(NODE_DATA(node), sizeof(*desc));
>  	printk(KERN_DEBUG "  alloc irq_desc for %d on cpu %d node %d\n",
>  		 irq, cpu, node);
>  	if (!desc) {

can you check tip?
we change _cpu to node already.

also only sh have after_bootmem now.
arch/sh/mm/init.c:int after_bootmem = 0;
arch/sh/mm/init.c:      after_bootmem = 1;
arch/sh/mm/ioremap_64.c:        extern int after_bootmem;
arch/sh/mm/ioremap_64.c:        if (after_bootmem) {
include/linux/mm.h:extern int after_bootmem;

for x86 we have bootmem_state ...
arch/x86/include/asm/page_types.h:enum bootmem_state {
arch/x86/include/asm/page_types.h:extern enum bootmem_state bootmem_state;
arch/x86/kernel/setup.c:        bootmem_state = DURING_BOOTMEM;
arch/x86/mm/init.c:enum bootmem_state bootmem_state = BEFORE_BOOTMEM;
arch/x86/mm/init.c:     if (bootmem_state == BEFORE_BOOTMEM)
arch/x86/mm/init.c:     if (bootmem_state == BEFORE_BOOTMEM)
arch/x86/mm/init.c:     if (bootmem_state == BEFORE_BOOTMEM && !start) {
arch/x86/mm/init.c:     if (bootmem_state == BEFORE_BOOTMEM &&
arch/x86/mm/init.c:     if (bootmem_state == BEFORE_BOOTMEM)

Andrew,
do we need to move bootmem_state back to linux/mm.h?

YH




YH

Re: [PATCH] sparseirq: Enable early irq_desc allocation.

[Paul Mundt]

On Thu, May 21, 2009 at 01:26:26PM -0700, Yinghai Lu wrote:
> can you check tip?
> we change _cpu to node already.
> 
> also only sh have after_bootmem now.
> arch/sh/mm/init.c:int after_bootmem = 0;
> arch/sh/mm/init.c:      after_bootmem = 1;
> arch/sh/mm/ioremap_64.c:        extern int after_bootmem;
> arch/sh/mm/ioremap_64.c:        if (after_bootmem) {
> include/linux/mm.h:extern int after_bootmem;
> 
> for x86 we have bootmem_state ...
> arch/x86/include/asm/page_types.h:enum bootmem_state {
> arch/x86/include/asm/page_types.h:extern enum bootmem_state bootmem_state;
> arch/x86/kernel/setup.c:        bootmem_state = DURING_BOOTMEM;
> arch/x86/mm/init.c:enum bootmem_state bootmem_state = BEFORE_BOOTMEM;
> arch/x86/mm/init.c:     if (bootmem_state == BEFORE_BOOTMEM)
> arch/x86/mm/init.c:     if (bootmem_state == BEFORE_BOOTMEM)
> arch/x86/mm/init.c:     if (bootmem_state == BEFORE_BOOTMEM && !start) {
> arch/x86/mm/init.c:     if (bootmem_state == BEFORE_BOOTMEM &&
> arch/x86/mm/init.c:     if (bootmem_state == BEFORE_BOOTMEM)
> 
> Andrew,
> do we need to move bootmem_state back to linux/mm.h?

No need, I just switched to slab_is_available() in the respin. Here is
v2, against -tip:

--

sparseirq: Enable early irq_desc allocation, v2.

Presently non-legacy IRQs have their irq_desc allocated with
kzalloc_node(). This assumes that all callers of irq_to_desc_node_alloc()
will be sufficiently late in the boot process that kmalloc is available.

While porting sparseirq support to sh this blew up immediately, as at the
time that we register the CPU's interrupt vector map only bootmem is
available. Check slab_is_available() to work out which path to use.

Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>

---

 V2 updated for -tip, and switched to slab_is_available().

 kernel/irq/handle.c |   19 ++++++++++++++-----
 1 file changed, 14 insertions(+), 5 deletions(-)

diff --git a/kernel/irq/handle.c b/kernel/irq/handle.c
index 33ff8f5..b5a47d6 100644
--- a/kernel/irq/handle.c
+++ b/kernel/irq/handle.c
@@ -18,8 +18,8 @@
 #include <linux/rculist.h>
 #include <linux/hash.h>
 #include <linux/bootmem.h>
+#include <linux/slab.h>
 #include <trace/events/irq.h>
-
 #include "internals.h"
 
 /*
@@ -81,11 +81,16 @@ static struct irq_desc irq_desc_init = {
 	.lock       = __SPIN_LOCK_UNLOCKED(irq_desc_init.lock),
 };
 
-void init_kstat_irqs(struct irq_desc *desc, int node, int nr)
+void __ref init_kstat_irqs(struct irq_desc *desc, int node, int nr)
 {
 	void *ptr;
 
-	ptr = kzalloc_node(nr * sizeof(*desc->kstat_irqs), GFP_ATOMIC, node);
+	if (slab_is_available())
+		ptr = kzalloc_node(nr * sizeof(*desc->kstat_irqs),
+				   GFP_ATOMIC, node);
+	else
+		ptr = alloc_bootmem_node(NODE_DATA(node),
+				nr * sizeof(*desc->kstat_irqs));
 
 	/*
 	 * don't overwite if can not get new one
@@ -186,7 +191,7 @@ struct irq_desc *irq_to_desc(unsigned int irq)
 	return NULL;
 }
 
-struct irq_desc *irq_to_desc_alloc_node(unsigned int irq, int node)
+struct irq_desc * __ref irq_to_desc_alloc_node(unsigned int irq, int node)
 {
 	struct irq_desc *desc;
 	unsigned long flags;
@@ -208,7 +213,11 @@ struct irq_desc *irq_to_desc_alloc_node(unsigned int irq, int node)
 	if (desc)
 		goto out_unlock;
 
-	desc = kzalloc_node(sizeof(*desc), GFP_ATOMIC, node);
+	if (slab_is_available())
+		desc = kzalloc_node(sizeof(*desc), GFP_ATOMIC, node);
+	else
+		desc = alloc_bootmem_node(NODE_DATA(node), sizeof(*desc));
+
 	printk(KERN_DEBUG "  alloc irq_desc for %d on node %d\n", irq, node);
 	if (!desc) {
 		printk(KERN_ERR "can not alloc irq_desc\n");

Re: [PATCH] sparseirq: Enable early irq_desc allocation.

[Ingo Molnar]

* Paul Mundt <lethal@linux-sh.org> wrote:

> On Thu, May 21, 2009 at 01:26:26PM -0700, Yinghai Lu wrote:
> > can you check tip?
> > we change _cpu to node already.
> > 
> > also only sh have after_bootmem now.
> > arch/sh/mm/init.c:int after_bootmem = 0;
> > arch/sh/mm/init.c:      after_bootmem = 1;
> > arch/sh/mm/ioremap_64.c:        extern int after_bootmem;
> > arch/sh/mm/ioremap_64.c:        if (after_bootmem) {
> > include/linux/mm.h:extern int after_bootmem;
> > 
> > for x86 we have bootmem_state ...
> > arch/x86/include/asm/page_types.h:enum bootmem_state {
> > arch/x86/include/asm/page_types.h:extern enum bootmem_state bootmem_state;
> > arch/x86/kernel/setup.c:        bootmem_state = DURING_BOOTMEM;
> > arch/x86/mm/init.c:enum bootmem_state bootmem_state = BEFORE_BOOTMEM;
> > arch/x86/mm/init.c:     if (bootmem_state == BEFORE_BOOTMEM)
> > arch/x86/mm/init.c:     if (bootmem_state == BEFORE_BOOTMEM)
> > arch/x86/mm/init.c:     if (bootmem_state == BEFORE_BOOTMEM && !start) {
> > arch/x86/mm/init.c:     if (bootmem_state == BEFORE_BOOTMEM &&
> > arch/x86/mm/init.c:     if (bootmem_state == BEFORE_BOOTMEM)
> > 
> > Andrew,
> > do we need to move bootmem_state back to linux/mm.h?
> 
> No need, I just switched to slab_is_available() in the respin. Here is
> v2, against -tip:

thanks.

I'm a bit unhappy about allocators here (and this is not your 
fault). There's not two but _three_ states for allocation: bootmem, 
pagealloc, kmalloc. There is a (short) stage during bootup where 
kmalloc is not enabled yet and bootmem is already deactivated. If we 
ever use slab_is_available() in such a state it will blow up.

Fixing all this allocator bootstrapping mess we have in Linux is a 
lot larger project though.

	Ingo

[tip:irq/numa] sparseirq: Allow early irq_desc allocation

[tip-bot for Paul Mundt]

Commit-ID:  948cd52906baf1f92aeea2f9b5c515db1b2e592a
Gitweb:     http://git.kernel.org/tip/948cd52906baf1f92aeea2f9b5c515db1b2e592a
Author:     Paul Mundt <lethal@linux-sh.org>
AuthorDate: Fri, 22 May 2009 10:40:09 +0900
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Sat, 23 May 2009 14:55:24 +0200

sparseirq: Allow early irq_desc allocation

Presently non-legacy IRQs have their irq_desc allocated with
kzalloc_node(). This assumes that all callers of irq_to_desc_node_alloc()
will be sufficiently late in the boot process that kmalloc is available.

While porting sparseirq support to sh this blew up immediately, as at the
time that we register the CPU's interrupt vector map only bootmem is
available. Check slab_is_available() to work out which path to use.

[ Impact: fix SH early boot crash with sparseirq enabled ]

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Acked-by: Yinghai Lu <yinghai@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Mel Gorman <mel@csn.ul.ie>
LKML-Reference: <20090522014008.GA2806@linux-sh.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>


---
 kernel/irq/handle.c |   18 ++++++++++++++----
 1 files changed, 14 insertions(+), 4 deletions(-)

diff --git a/kernel/irq/handle.c b/kernel/irq/handle.c
index a3c671e..18041a2 100644
--- a/kernel/irq/handle.c
+++ b/kernel/irq/handle.c
@@ -11,6 +11,7 @@
  */
 
 #include <linux/irq.h>
+#include <linux/slab.h>
 #include <linux/module.h>
 #include <linux/random.h>
 #include <linux/interrupt.h>
@@ -81,11 +82,16 @@ static struct irq_desc irq_desc_init = {
 	.lock       = __SPIN_LOCK_UNLOCKED(irq_desc_init.lock),
 };
 
-void init_kstat_irqs(struct irq_desc *desc, int node, int nr)
+void __ref init_kstat_irqs(struct irq_desc *desc, int node, int nr)
 {
 	void *ptr;
 
-	ptr = kzalloc_node(nr * sizeof(*desc->kstat_irqs), GFP_ATOMIC, node);
+	if (slab_is_available())
+		ptr = kzalloc_node(nr * sizeof(*desc->kstat_irqs),
+				   GFP_ATOMIC, node);
+	else
+		ptr = alloc_bootmem_node(NODE_DATA(node),
+				nr * sizeof(*desc->kstat_irqs));
 
 	/*
 	 * don't overwite if can not get new one
@@ -186,7 +192,7 @@ struct irq_desc *irq_to_desc(unsigned int irq)
 	return NULL;
 }
 
-struct irq_desc *irq_to_desc_alloc_node(unsigned int irq, int node)
+struct irq_desc * __ref irq_to_desc_alloc_node(unsigned int irq, int node)
 {
 	struct irq_desc *desc;
 	unsigned long flags;
@@ -208,7 +214,11 @@ struct irq_desc *irq_to_desc_alloc_node(unsigned int irq, int node)
 	if (desc)
 		goto out_unlock;
 
-	desc = kzalloc_node(sizeof(*desc), GFP_ATOMIC, node);
+	if (slab_is_available())
+		desc = kzalloc_node(sizeof(*desc), GFP_ATOMIC, node);
+	else
+		desc = alloc_bootmem_node(NODE_DATA(node), sizeof(*desc));
+
 	printk(KERN_DEBUG "  alloc irq_desc for %d on node %d\n", irq, node);
 	if (!desc) {
 		printk(KERN_ERR "can not alloc irq_desc\n");