accessing stack of non-current task

accessing stack of non-current task

[Uma shankar]

Hi,

Is it possible for the kernel to access the user-stack data of a
task different from "current" ? ( This is needed for stack-dump as
well as backtrace. )

I thought the answer is "no". ( Kernel sees memory through the
page-table of "current" )

But I found few places in kernel where this is done. ( eg:
debug_rt_mutex_print_deadlock() in rtmutex-debug.c )

What is the explanation ?

                                        thanks
                                        shankar

--
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: accessing stack of non-current task

[Johannes Weiner]

On Tue, Apr 20, 2010 at 06:08:14PM +0530, Uma shankar wrote:
> Hi,
> 
> Is it possible for the kernel to access the user-stack data of a
> task different from "current" ? ( This is needed for stack-dump as
> well as backtrace. )

Yes, have a look at __get_user_pages() in mm/memory.c.

> I thought the answer is "no". ( Kernel sees memory through the
> page-table of "current" )

That is correct when using virtual addresses and letting the MMU
do the page table lookup in the currently active page tables.

But if you have the task_struct of another process, you can easily
get to its vmas and page tables (task->mm) and walk them in software.

--
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: accessing stack of non-current task

[Uma shankar]

On Tue, Apr 20, 2010 at 7:13 PM, Johannes Weiner <hannes@cmpxchg.org> wrote:
> On Tue, Apr 20, 2010 at 06:08:14PM +0530, Uma shankar wrote:
>> Hi,
>>
>> Is it possible for the kernel to access the user-stack data of a
>> task different from "current" ? ( This is needed for stack-dump as
>> well as backtrace. )
>
> Yes, have a look at __get_user_pages() in mm/memory.c.
>

Yes,  I understand this.

But  have a look  at  "void show_stack(struct task_struct *tsk,
unsigned long *sp)  "  in traps.c (  arch specific  ).

Is there a implicit assumption that  "tsk"  and "current"  are threads
sharing same  "mm_strct"  ?

--
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: accessing stack of non-current task

[Johannes Weiner]

On Tue, Apr 20, 2010 at 07:30:15PM +0530, Uma shankar wrote:
> On Tue, Apr 20, 2010 at 7:13 PM, Johannes Weiner <hannes@cmpxchg.org> wrote:
> > On Tue, Apr 20, 2010 at 06:08:14PM +0530, Uma shankar wrote:
> >> Hi,
> >>
> >> Is it possible for the kernel to access the user-stack data of a
> >> task different from "current" ? ( This is needed for stack-dump as
> >> well as backtrace. )
> >
> > Yes, have a look at __get_user_pages() in mm/memory.c.
> >
> 
> Yes,  I understand this.
> 
> But  have a look  at  "void show_stack(struct task_struct *tsk,
> unsigned long *sp)  "  in traps.c (  arch specific  ).
> 
> Is there a implicit assumption that  "tsk"  and "current"  are threads
> sharing same  "mm_strct"  ?

No, this is dumping the _kernel stack_ of a process, not the user stack.

The mm_struct does not matter.

--
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>