Re: [RESEND][PATCH] Mark thread stack correctly in proc/<pid>/maps

On Wed, Feb 8, 2012 at 11:27 PM, KOSAKI Motohiro
[off-list ref] wrote:

Now, we are using some bit saving hack. example,

1) use ifdef

#ifndef CONFIG_TRANSPARENT_HUGEPAGE
#define VM_MAPPED_COPY 0x01000000      /* T if mapped copy of data (nommu
mmap) */
#else
#define VM_HUGEPAGE    0x01000000      /* MADV_HUGEPAGE marked this vma */
#endif

2) use bit combination

#define VM_STACK_INCOMPLETE_SETUP      (VM_RAND_READ | VM_SEQ_READ)


Maybe you can take a similar way. And of course, you can ban some useless
flag
bits.

I found the thread in which Linus rejected the idea of expanding vm_flags:

https://lkml.org/lkml/2011/11/10/522

and based on that, I don't think I can justify the need for a new flag
for this patch since it is purely for display purposes and has nothing
to do with the actual treatment of the vma. So I figured out another
way to identify a thread stack without changing the way the vma
properties (I should have done this in the first place I think) which
is by checking if the vma contains the stack pointer of the task.

With this change:

/proc/pid/task/tid/maps: will only mark the stack that the task uses

/proc/pid/maps: will mark all stacks. This path will be slower since
it will iterate through the entire thread group for each vma.

I'll test this and attach a new version of the patch.

Regards,
Siddhesh

-- 
Siddhesh Poyarekar
http://siddhesh.in
--
To unsubscribe from this list: send the line "unsubscribe linux-man" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html