Re: + procfs-provide-stack-information-for-threads-v08.patch added to -mm tree

Re: + procfs-provide-stack-information-for-threads-v08.patch added to -mm tree

[Alexey Dobriyan]

On Mon, Jun 15, 2009 at 03:02:05PM -0700, akpm@linux-foundation.org wrote:
>      procfs-provide-stack-information-for-threads-v08.patch
> --- a/fs/proc/array.c~procfs-provide-stack-information-for-threads-v08

> +++ a/fs/proc/array.c
> @@ -321,6 +321,54 @@ static inline void task_context_switch_c
>  			p->nivcsw);
>  }
>  
> +static inline unsigned long get_stack_usage_in_bytes(struct vm_area_struct *vma,
> +					struct task_struct *p)
> +{
> +	unsigned long	i;
> +	struct page	*page;
> +	unsigned long	stkpage;
> +
> +	stkpage = KSTK_ESP(p) & PAGE_MASK;
> +
> +#ifdef CONFIG_STACK_GROWSUP
> +	for (i = vma->vm_end; i-PAGE_SIZE > stkpage; i -= PAGE_SIZE) {
> +
> +		page = follow_page(vma, i-PAGE_SIZE, 0);

How can this work?

If stack page got swapped out, you'll get smaller than actual result.

> +
> +		if (!IS_ERR(page) && page)
> +			break;
> +	}
> +	return i - (p->stack_start & PAGE_MASK);
> +#else
> +	for (i = vma->vm_start; i+PAGE_SIZE <= stkpage; i += PAGE_SIZE) {
> +
> +		page = follow_page(vma, i, 0);
> +
> +		if (!IS_ERR(page) && page)
> +			break;
> +	}
> +	return (p->stack_start & PAGE_MASK) - i + PAGE_SIZE;
> +#endif
> +}
> +
> +static inline void task_show_stack_usage(struct seq_file *m,
> +						struct task_struct *task)
> +{
> +	struct vm_area_struct	*vma;
> +	struct mm_struct	*mm = get_task_mm(task);
> +
> +	if (mm) {
> +		down_read(&mm->mmap_sem);
> +		vma = find_vma(mm, task->stack_start);
> +		if (vma)
> +			seq_printf(m, "Stack usage:\t%lu kB\n",
> +				get_stack_usage_in_bytes(vma, task) >> 10);
> +
> +		up_read(&mm->mmap_sem);
> +		mmput(mm);
> +	}
> +}

Re: + procfs-provide-stack-information-for-threads-v08.patch added to -mm tree

[Stefani Seibold]

Am Dienstag, den 16.06.2009, 02:33 +0400 schrieb Alexey Dobriyan:
> On Mon, Jun 15, 2009 at 03:02:05PM -0700, akpm@linux-foundation.org wrote:
> >      procfs-provide-stack-information-for-threads-v08.patch
> > --- a/fs/proc/array.c~procfs-provide-stack-information-for-threads-v08
> 
> > +++ a/fs/proc/array.c
> > @@ -321,6 +321,54 @@ static inline void task_context_switch_c
> >  			p->nivcsw);
> >  }
> >  
> > +static inline unsigned long get_stack_usage_in_bytes(struct vm_area_struct *vma,
> > +					struct task_struct *p)
> > +{
> > +	unsigned long	i;
> > +	struct page	*page;
> > +	unsigned long	stkpage;
> > +
> > +	stkpage = KSTK_ESP(p) & PAGE_MASK;
> > +
> > +#ifdef CONFIG_STACK_GROWSUP
> > +	for (i = vma->vm_end; i-PAGE_SIZE > stkpage; i -= PAGE_SIZE) {
> > +
> > +		page = follow_page(vma, i-PAGE_SIZE, 0);
> 
> How can this work?
> 
> If stack page got swapped out, you'll get smaller than actual result.
> 

If you tell me how to do it in the right way, i can fix it!

Re: + procfs-provide-stack-information-for-threads-v08.patch added to -mm tree

[Eric W. Biederman]

Stefani Seibold <stefani@seibold.net> writes:

> Am Dienstag, den 16.06.2009, 02:33 +0400 schrieb Alexey Dobriyan:
>> On Mon, Jun 15, 2009 at 03:02:05PM -0700, akpm@linux-foundation.org wrote:
>> >      procfs-provide-stack-information-for-threads-v08.patch
>> > --- a/fs/proc/array.c~procfs-provide-stack-information-for-threads-v08
>> 
>> > +++ a/fs/proc/array.c
>> > @@ -321,6 +321,54 @@ static inline void task_context_switch_c
>> >  			p->nivcsw);
>> >  }
>> >  
>> > +static inline unsigned long get_stack_usage_in_bytes(struct vm_area_struct *vma,
>> > +					struct task_struct *p)
>> > +{
>> > +	unsigned long	i;
>> > +	struct page	*page;
>> > +	unsigned long	stkpage;
>> > +
>> > +	stkpage = KSTK_ESP(p) & PAGE_MASK;
>> > +
>> > +#ifdef CONFIG_STACK_GROWSUP
>> > +	for (i = vma->vm_end; i-PAGE_SIZE > stkpage; i -= PAGE_SIZE) {
>> > +
>> > +		page = follow_page(vma, i-PAGE_SIZE, 0);
>> 
>> How can this work?
>> 
>> If stack page got swapped out, you'll get smaller than actual result.
>> 
>
> If you tell me how to do it in the right way, i can fix it!

You are attempting to answer two questions here.
1) Where do the thread stacks reside.
2) What is the maximum stack space that has been used.

Just listing the thread stacks seems like a small O(1) change.

Computing how much stack space has been used looks trickier.
Perhaps you could map them with MAP_GROWSDOWN?  Of course that
has the problem that you don't stop growing your stack until
it bumps into something else.  Not ideal for a thread stack.

Eric

Re: + procfs-provide-stack-information-for-threads-v08.patch added to -mm tree

[Alexey Dobriyan]

On Tue, Jun 16, 2009 at 12:39:28PM -0700, Eric W. Biederman wrote:
> Stefani Seibold <stefani@seibold.net> writes:
> 
> > Am Dienstag, den 16.06.2009, 02:33 +0400 schrieb Alexey Dobriyan:
> >> On Mon, Jun 15, 2009 at 03:02:05PM -0700, akpm@linux-foundation.org wrote:
> >> >      procfs-provide-stack-information-for-threads-v08.patch
> >> > --- a/fs/proc/array.c~procfs-provide-stack-information-for-threads-v08
> >> 
> >> > +++ a/fs/proc/array.c
> >> > @@ -321,6 +321,54 @@ static inline void task_context_switch_c
> >> >  			p->nivcsw);
> >> >  }
> >> >  
> >> > +static inline unsigned long get_stack_usage_in_bytes(struct vm_area_struct *vma,
> >> > +					struct task_struct *p)
> >> > +{
> >> > +	unsigned long	i;
> >> > +	struct page	*page;
> >> > +	unsigned long	stkpage;
> >> > +
> >> > +	stkpage = KSTK_ESP(p) & PAGE_MASK;
> >> > +
> >> > +#ifdef CONFIG_STACK_GROWSUP
> >> > +	for (i = vma->vm_end; i-PAGE_SIZE > stkpage; i -= PAGE_SIZE) {
> >> > +
> >> > +		page = follow_page(vma, i-PAGE_SIZE, 0);
> >> 
> >> How can this work?
> >> 
> >> If stack page got swapped out, you'll get smaller than actual result.
> >> 
> >
> > If you tell me how to do it in the right way, i can fix it!
> 
> You are attempting to answer two questions here.
> 1) Where do the thread stacks reside.

Kernel doesn't know. Application stack is somewhere below/above
some SP register.

> 2) What is the maximum stack space that has been used.
> 
> Just listing the thread stacks seems like a small O(1) change.
> 
> Computing how much stack space has been used looks trickier.
> Perhaps you could map them with MAP_GROWSDOWN?  Of course that
> has the problem that you don't stop growing your stack until
> it bumps into something else.  Not ideal for a thread stack.