> rearnsha@buzzard.freeserve.co.uk (Richard Earnshaw) writes:
> > This 
> > isn't a major problem if momentary stack unalignemnt is OK, but issues 
> > such as asynchronous events have to be considered (for example, if a 
> > signal were received when the stack was not 8-byte aligned, would the 
> > signal handler function correctly -- in particular, would it run on its 
> > own stack, or on the stack of the current process?).
> 
> That particular one is easy: if 8-byte alignment on function entry is
> mandated, the kernel could be sure to align the SP to 8 bytes before
> invoking the user signal handler...
> 
> (It seems that if ATPCS requires 8 byte alignment on entry & exit, if
> we want to support that -- i dunno anything about it -- then we've
> gotta make sure the stack is properly aligned on signal handler entry
> as well...)

Sure, so the issue is that the kernel signal despatching code has to be 
sure to re-align the stack before posting the signal (and to restore the 
correct value if the handler returns).

R.