Re: [PATCH] PowerPC: bp-permanent.exp, kill-after-signal fix

[will schmidt <will_schmidt@vnet.ibm.com>]

On Mon, 2022-05-02 at 15:10 +0000, Ulrich Weigand wrote:
> will schmidt <will_schmidt@vnet.ibm.com> wrote:
> 
> > So..  Is this directly related to the amount of kernel code
> > handling
> > the signals?  i.e. Would this need to be updated if another
> instruction
> > is added to the kernel code?
> 
> To clarify, this not *kernel* code, it's user space code, but code
> that is placed by the kernel into user space (in the vdso).
> 
> Many platforms today use the vdso in the signal *return* path, but
> Power seems special in that it is using a vdso instruction also to
> *enter* a signal handler in the first place.  (So instead of the
> kernel directly dispatching to the installed signal handler, it
> first dispatches to vdso code which in turn calls into the signal
> handler.)
> 
> Reading the kernel logs, this seems to have been a performance
> optimization to prevent the call/return stack from getting out
> of balance.
> 
> That vdso code path when entering a signal handler is just a
> single "brctl"; I don't see any particular reason why this would
> ever need to become any longer.

Ok, thanks for clarifying. 
I look forward to seeing the updated
comments in the next version of the patch by Carl.  :-)

Thanks
-Will


> 
> Bye,
> Ulrich
>