Re: [PATCH v3 09/51] x86/dumpstack: fix x86_32 kernel_stack_pointer() previous stack access

[Andy Lutomirski]

On Mon, Aug 15, 2016 at 10:22 AM, Josh Poimboeuf <jpoimboe@xxxxxxxxxx> wrote:
> On Mon, Aug 15, 2016 at 10:05:58AM -0500, Josh Poimboeuf wrote:
>> On Sun, Aug 14, 2016 at 12:26:29AM -0700, Andy Lutomirski wrote:
>> > On Fri, Aug 12, 2016 at 7:28 AM, Josh Poimboeuf <jpoimboe@xxxxxxxxxx> wrote:
>> > > On x86_32, when an interrupt happens from kernel space, SS and SP aren't
>> > > pushed and the existing stack is used.  So pt_regs is effectively two
>> > > words shorter, and the previous stack pointer is normally the memory
>> > > after the shortened pt_regs, aka '&regs->sp'.
>> > >
>> > > But in the rare case where the interrupt hits right after the stack
>> > > pointer has been changed to point to an empty stack, like for example
>> > > when call_on_stack() is used, the address immediately after the
>> > > shortened pt_regs is no longer on the stack.  In that case, instead of
>> > > '&regs->sp', the previous stack pointer should be retrieved from the
>> > > beginning of the current stack page.
>> > >
>> > > kernel_stack_pointer() wants to do that, but it forgets to dereference
>> > > the pointer.  So instead of returning a pointer to the previous stack,
>> > > it returns a pointer to the beginning of the current stack.
>> > >
>> > > Fixes: 0788aa6a23cb ("x86: Prepare removal of previous_esp from i386 thread_info structure")
>> > > Signed-off-by: Josh Poimboeuf <jpoimboe@xxxxxxxxxx>
>> >
>> > This seems like a valid fix, but I'm not sure I agree with the intent
>> > of the code.  &regs->sp really is the previous stack pointer in the
>> > sense that the stack pointer was &regs->sp when the entry happened.
>> > From an unwinder's perspective, how is:
>> >
>> > movl [whatever], $esp
>> > <-- interrupt
>> >
>> > any different from:
>> >
>> > movl [whatever], $esp
>> > pushl [something]
>> > <-- interrupt
>>
>> In the first case, the stack is empty, so reading the value pointed to
>> by %esp would result in accessing outside the bounds of the stack.
>
> ...but maybe your point is that following the previous stack pointer is
> outside the scope of kernel_stack_pointer() and should instead be done
> by its caller.  Especially considering the fact that the x86_64 version
> of this function doesn't have this "feature".  In which case I think I
> would agree.

Yes, especially since your code seems to know how to find the previous
stack already.

>
> However I think fixing that is outside the scope of this
> already-way-too-big patch set.

Agreed.

>
> --
> Josh



-- 
Andy Lutomirski
AMA Capital Management, LLC