This is the mail archive of the
binutils@sourceware.org
mailing list for the binutils project.
Re: [PATCH] gold: pad code segments out to page size for isolate_execinstr targets
- From: Roland McGrath <mcgrathr at google dot com>
- To: Cary Coutant <ccoutant at google dot com>
- Cc: "binutils at sourceware dot org" <binutils at sourceware dot org>
- Date: Tue, 27 Aug 2013 14:55:26 -0700
- Subject: Re: [PATCH] gold: pad code segments out to page size for isolate_execinstr targets
- Authentication-results: sourceware.org; auth=none
- References: <CAB=4xhrmv4R_ZMM-8PMXQUJwz1fho=wUF6qPzUcu6xDbBmL0Dw at mail dot gmail dot com> <CAHACq4qR34DkzdW=M9+Rrm-e6vU89m6Xona6ZHmLrsU8CCT8Ww at mail dot gmail dot com>
Committed.
I like the simplicity of one-byte nops and don't really know what the
benefit of longer ones is.
But moreover, I intend to keep gold's behavior in partiy with BFD
ld's, and in BFDland it is more hassle to change the signature of the
hook so I didn't consider bothering with it.
I considered doing code fill with HLT, but hesitated for two reasons.
First, while I cannot really imagine a sensical situation in which
code fill instructions would be executed, I'm not completely confident
it never could validly be executed.
Second, I'd like to keep x86 and ARM at parity in this sort of regard,
and in BFDland ARM does default zero-fill (which is not quite a nop in
ARM, but a harmless enough instruction) so it would be a bunch more
work to change code fill for arm-nacl in BFD (which I'd do before
making gold match).
Thanks,
Roland