[PATCH] hugetlb: add locking for overcommit sysctl

[PATCH] hugetlb: add locking for overcommit sysctl

[Nishanth Aravamudan]

When I replaced hugetlb_dynamic_pool with nr_overcommit_hugepages I used
proc_doulongvec_minmax() directly. However, hugetlb.c's locking rules
require that all counter modifications occur under the hugetlb_lock. Add
a callback into the hugetlb code similar to the one for nr_hugepages.
Grab the lock around the manipulation of nr_overcommit_hugepages in
proc_doulongvec_minmax().

Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>

---
I noticed that the nr_hugepages sysctl uses a helper variable,
max_huge_pages to do the same thing. Is that because of locking
requirements (undocmented) for proc_doulongvec_minmax() or because of
the more complicated manipulation of the pool state for that sysctl? If
the former, I will need to modify this patch.

diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
index 30d606a..7ca198b 100644
--- a/include/linux/hugetlb.h
+++ b/include/linux/hugetlb.h
@@ -17,6 +17,7 @@ static inline int is_vm_hugetlb_page(struct vm_area_struct *vma)
 }
 
 int hugetlb_sysctl_handler(struct ctl_table *, int, struct file *, void __user *, size_t *, loff_t *);
+int hugetlb_overcommit_handler(struct ctl_table *, int, struct file *, void __user *, size_t *, loff_t *);
 int hugetlb_treat_movable_handler(struct ctl_table *, int, struct file *, void __user *, size_t *, loff_t *);
 int copy_hugetlb_page_range(struct mm_struct *, struct mm_struct *, struct vm_area_struct *);
 int follow_hugetlb_page(struct mm_struct *, struct vm_area_struct *, struct page **, struct vm_area_struct **, unsigned long *, int *, int, int);
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index 5e2ad5b..92fef8d 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -973,7 +973,7 @@ static struct ctl_table vm_table[] = {
 		.data		= &nr_overcommit_huge_pages,
 		.maxlen		= sizeof(nr_overcommit_huge_pages),
 		.mode		= 0644,
-		.proc_handler	= &proc_doulongvec_minmax,
+		.proc_handler	= &hugetlb_overcommit_handler,
 	},
 #endif
 	{
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index 1a56420..d9a3803 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -605,6 +605,16 @@ int hugetlb_treat_movable_handler(struct ctl_table *table, int write,
 	return 0;
 }
 
+int hugetlb_overcommit_handler(struct ctl_table *table, int write,
+			struct file *file, void __user *buffer,
+			size_t *length, loff_t *ppos)
+{
+	spin_lock(&hugetlb_lock);
+	proc_doulongvec_minmax(table, write, file, buffer, length, ppos);
+	spin_unlock(&hugetlb_lock);
+	return 0;
+}
+
 #endif /* CONFIG_SYSCTL */
 
 int hugetlb_report_meminfo(char *buf)

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH] hugetlb: add locking for overcommit sysctl

[Adam Litke]

On Wed, 2008-02-06 at 15:02 -0800, Nishanth Aravamudan wrote:
> When I replaced hugetlb_dynamic_pool with nr_overcommit_hugepages I used
> proc_doulongvec_minmax() directly. However, hugetlb.c's locking rules
> require that all counter modifications occur under the hugetlb_lock. Add
> a callback into the hugetlb code similar to the one for nr_hugepages.
> Grab the lock around the manipulation of nr_overcommit_hugepages in
> proc_doulongvec_minmax().
> 
> Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Acked-by: Adam Litke <agl@us.ibm.com>

> ---
> I noticed that the nr_hugepages sysctl uses a helper variable,
> max_huge_pages to do the same thing. Is that because of locking
> requirements (undocmented) for proc_doulongvec_minmax() or because of
> the more complicated manipulation of the pool state for that sysctl? If
> the former, I will need to modify this patch.
> 
> diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
> index 30d606a..7ca198b 100644
> --- a/include/linux/hugetlb.h
> +++ b/include/linux/hugetlb.h
> @@ -17,6 +17,7 @@ static inline int is_vm_hugetlb_page(struct vm_area_struct *vma)
>  }
> 
>  int hugetlb_sysctl_handler(struct ctl_table *, int, struct file *, void __user *, size_t *, loff_t *);
> +int hugetlb_overcommit_handler(struct ctl_table *, int, struct file *, void __user *, size_t *, loff_t *);
>  int hugetlb_treat_movable_handler(struct ctl_table *, int, struct file *, void __user *, size_t *, loff_t *);
>  int copy_hugetlb_page_range(struct mm_struct *, struct mm_struct *, struct vm_area_struct *);
>  int follow_hugetlb_page(struct mm_struct *, struct vm_area_struct *, struct page **, struct vm_area_struct **, unsigned long *, int *, int, int);
> diff --git a/kernel/sysctl.c b/kernel/sysctl.c
> index 5e2ad5b..92fef8d 100644
> --- a/kernel/sysctl.c
> +++ b/kernel/sysctl.c
> @@ -973,7 +973,7 @@ static struct ctl_table vm_table[] = {
>  		.data		= &nr_overcommit_huge_pages,
>  		.maxlen		= sizeof(nr_overcommit_huge_pages),
>  		.mode		= 0644,
> -		.proc_handler	= &proc_doulongvec_minmax,
> +		.proc_handler	= &hugetlb_overcommit_handler,
>  	},
>  #endif
>  	{
> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> index 1a56420..d9a3803 100644
> --- a/mm/hugetlb.c
> +++ b/mm/hugetlb.c
> @@ -605,6 +605,16 @@ int hugetlb_treat_movable_handler(struct ctl_table *table, int write,
>  	return 0;
>  }
> 
> +int hugetlb_overcommit_handler(struct ctl_table *table, int write,
> +			struct file *file, void __user *buffer,
> +			size_t *length, loff_t *ppos)
> +{
> +	spin_lock(&hugetlb_lock);
> +	proc_doulongvec_minmax(table, write, file, buffer, length, ppos);
> +	spin_unlock(&hugetlb_lock);
> +	return 0;
> +}
> +
>  #endif /* CONFIG_SYSCTL */
> 
>  int hugetlb_report_meminfo(char *buf)
> 
-- 
Adam Litke - (agl at us.ibm.com)
IBM Linux Technology Center

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH] hugetlb: add locking for overcommit sysctl

[Nishanth Aravamudan]

On 07.02.2008 [09:19:44 -0600], Adam Litke wrote:
> 
> On Wed, 2008-02-06 at 15:02 -0800, Nishanth Aravamudan wrote:
> > When I replaced hugetlb_dynamic_pool with nr_overcommit_hugepages I used
> > proc_doulongvec_minmax() directly. However, hugetlb.c's locking rules
> > require that all counter modifications occur under the hugetlb_lock. Add
> > a callback into the hugetlb code similar to the one for nr_hugepages.
> > Grab the lock around the manipulation of nr_overcommit_hugepages in
> > proc_doulongvec_minmax().
> > 
> > Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
> Acked-by: Adam Litke <agl@us.ibm.com>

Thanks, Adam.

Andrew, this is a bugfix for 2.6.25 and presumably should be backported
to -stable once upstream (I'll keep my eye out for the commit message if
it goes there).

> > ---
> > I noticed that the nr_hugepages sysctl uses a helper variable,
> > max_huge_pages to do the same thing. Is that because of locking
> > requirements (undocmented) for proc_doulongvec_minmax() or because of
> > the more complicated manipulation of the pool state for that sysctl? If
> > the former, I will need to modify this patch.
> > 
> > diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
> > index 30d606a..7ca198b 100644
> > --- a/include/linux/hugetlb.h
> > +++ b/include/linux/hugetlb.h
> > @@ -17,6 +17,7 @@ static inline int is_vm_hugetlb_page(struct vm_area_struct *vma)
> >  }
> > 
> >  int hugetlb_sysctl_handler(struct ctl_table *, int, struct file *, void __user *, size_t *, loff_t *);
> > +int hugetlb_overcommit_handler(struct ctl_table *, int, struct file *, void __user *, size_t *, loff_t *);
> >  int hugetlb_treat_movable_handler(struct ctl_table *, int, struct file *, void __user *, size_t *, loff_t *);
> >  int copy_hugetlb_page_range(struct mm_struct *, struct mm_struct *, struct vm_area_struct *);
> >  int follow_hugetlb_page(struct mm_struct *, struct vm_area_struct *, struct page **, struct vm_area_struct **, unsigned long *, int *, int, int);
> > diff --git a/kernel/sysctl.c b/kernel/sysctl.c
> > index 5e2ad5b..92fef8d 100644
> > --- a/kernel/sysctl.c
> > +++ b/kernel/sysctl.c
> > @@ -973,7 +973,7 @@ static struct ctl_table vm_table[] = {
> >  		.data		= &nr_overcommit_huge_pages,
> >  		.maxlen		= sizeof(nr_overcommit_huge_pages),
> >  		.mode		= 0644,
> > -		.proc_handler	= &proc_doulongvec_minmax,
> > +		.proc_handler	= &hugetlb_overcommit_handler,
> >  	},
> >  #endif
> >  	{
> > diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> > index 1a56420..d9a3803 100644
> > --- a/mm/hugetlb.c
> > +++ b/mm/hugetlb.c
> > @@ -605,6 +605,16 @@ int hugetlb_treat_movable_handler(struct ctl_table *table, int write,
> >  	return 0;
> >  }
> > 
> > +int hugetlb_overcommit_handler(struct ctl_table *table, int write,
> > +			struct file *file, void __user *buffer,
> > +			size_t *length, loff_t *ppos)
> > +{
> > +	spin_lock(&hugetlb_lock);
> > +	proc_doulongvec_minmax(table, write, file, buffer, length, ppos);
> > +	spin_unlock(&hugetlb_lock);
> > +	return 0;
> > +}
> > +
> >  #endif /* CONFIG_SYSCTL */
> > 
> >  int hugetlb_report_meminfo(char *buf)
> > 
> -- 
> Adam Litke - (agl at us.ibm.com)
> IBM Linux Technology Center
> 

-- 
Nishanth Aravamudan <nacc@us.ibm.com>
IBM Linux Technology Center

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>