Re: [rfc] x86, ptrace: memory accounting for branch tracing

[Andrew Morton <akpm@linux-foundation.org>]

On Tue, 16 Dec 2008 18:30:59 +0100 Ingo Molnar <mingo@elte.hu> wrote:

> 
> * Markus Metzger <markus.t.metzger@intel.com> wrote:
> 
> > Account memory allocated for the BTS buffer to the traced task's 
> > total_vm and locked_vm.
> 
> Andrew, is this the right (and preferred) way to attach BTS buffer 
> allocation overhead to the RLIMIT_MEMLOCK bucket:

Close.  I suspect we could refactor mlock.c to avoid all the code
duplication we have there.

There's (almost) nothing BTS-specific in this code, and it would be
better if it lived in mm/mlock.c.  Hopefully in a
usable-by-other-parts-of-mlock.c fashion.

> > +static int ptrace_bts_allocate_buffer(struct task_struct *child, size_t size)
> > +{
> > +	unsigned long rlim, vm, pgsz;
> > +	int error = -ENOMEM;
> > +
> > +	pgsz = PAGE_ALIGN(size) >> PAGE_SHIFT;
> > +
> > +	down_write(&child->mm->mmap_sem);
> > +
> > +	rlim = child->signal->rlim[RLIMIT_AS].rlim_cur >> PAGE_SHIFT;
> > +	vm   = child->mm->total_vm + pgsz;
> > +	if (rlim < vm)

This is off-by-one, I think.  Should be

	if (vm > rmlim)

> > +		goto out;
> > +
> > +	rlim = child->signal->rlim[RLIMIT_MEMLOCK].rlim_cur >> PAGE_SHIFT;
> > +	vm   = child->mm->locked_vm + pgsz;
> > +	if (rlim < vm)

ditto

> > +		goto out;
> > +
> > +	child->bts_buffer = kzalloc(size, GFP_KERNEL);
> > +	if (!child->bts_buffer)
> > +		goto out;
> > +
> > +	child->bts_size = size;
> > +
> > +	child->mm->total_vm  += pgsz;
> > +	child->mm->locked_vm += pgsz;
> > +
> > +	error = 0;
> > + out:
> > +	up_write(&child->mm->mmap_sem);
> > +	return error;
> 
> ?
> 
> 	Ingo