[Bug 3533] [IcedTea8] HotSpot generates code with unaligned stack, crashes on SSE operations

[bugzilla-daemon at icedtea.classpath.org]

https://icedtea.classpath.org/bugzilla/show_bug.cgi?id=3533

--- Comment #22 from Andrew John Hughes <gnu.andrew at redhat.com> ---
(In reply to Maciej S. Szmigiero from comment #21)
> I've tried "-mstackrealign" but it didn't add the required
> stack pointer realignment prologue to every entry function.
> 

I've been able to do a full bootstrap with -mstackrealign on a setup that would
reliably crash without that option. The fact that Oracle added it for the same
issue on Mac OS X suggests that it's the right workaround and that's also the
opinion of other engineers I've spoken to.

What issues do you still see with this enabled?

> From the Gentoo bug:
> > That's why an other solution is necessary in form of
> > "-mincoming-stack-boundary=2" {C,CXX}FLAG for icedtea (not
> > "-mstackrealign" since it does not work in our case as this flag seems
> > to be for leaf - and not intermediate - functions).
> > It is a big hammer but it allows building of a working icedtea (with
> > glibc compiled by gcc-7.3.0) without any patching of the HotSpot code.
> 
> Gentoo now adds "-mincoming-stack-boundary=2" when building IcedTea for
> 32-bit x86 target.

I'm dubious about this, especially as 2 means 4-byte alignment, which is a
reduction from the default 16-byte alignment and could cause issues with some
values.

"
       -mincoming-stack-boundary=num
           Assume the incoming stack is aligned to a 2 raised to num byte
boundary.  If -mincoming-stack-boundary is not specified, the one specified by
-mpreferred-stack-boundary is used. On Pentium and Pentium Pro, "double" and
"long double" values should be aligned to an 8-byte boundary (see
-malign-double) or suffer significant run time performance penalties.  On
Pentium III, the Streaming SIMD Extension (SSE) data type "__m128" may not work
properly if it is not 16-byte aligned."

whereas -mstackrealign sounds like exactly the workaround we want (adjustment
as appropriate):

"
       -mstackrealign
           Realign the stack at entry.  On the x86, the -mstackrealign option
generates an alternate prologue and epilogue that realigns the run-time stack
if necessary.  This supports mixing legacy codes that keep 4-byte stack
alignment with modern codes that keep 16-byte stack alignment for SSE
compatibility.  See
also the attribute "force_align_arg_pointer", applicable to individual
functions."

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/distro-pkg-dev/attachments/20180517/8a2fe47e/attachment.html>