This is the mail archive of the
binutils@sources.redhat.com
mailing list for the binutils project.
Re: Bug on syscall error test for 64 bits apps
- From: Segher Boessenkool <segher at kernel dot crashing dot org>
- To: Alan Modra <amodra at bigpond dot net dot au>
- Cc: binutils at sources dot redhat dot com, linuxppc64-dev <linuxppc64-dev at lists dot linuxppc dot org>, Benjamin Herrenschmidt <benh at kernel dot crashing dot org>
- Date: Tue, 30 Mar 2004 08:54:45 +0200
- Subject: Re: Bug on syscall error test for 64 bits apps
- References: <1080284301.1389.11.camel@gaston> <780A91EC-8166-11D8-BC8C-000A95A4DC02@kernel.crashing.org> <20040330005037.GD23980@bubble.modra.org>
On 30-mrt-04, at 2:50, Alan Modra wrote:
On Mon, Mar 29, 2004 at 11:50:10AM +0200, Segher Boessenkool wrote:
Also, cmpl with three arguments is not a legal PowerPC instruction,
only a POWER instruction. I can't find any GAS flag to stop it
assembling POWER-only instructions/mnemonics, though.
binutils bug.
feature :)
Bug, actually.
It refuses all other POWER insns (I used -mppc64, but anything else
that refuses POWER insns will do).
Looking deeper into it, the cause is that the L field is defined
as optional. But it is not. (Ref.: Book I and/or the PEM v2).
The L field (in binutils) is only used for one other insn: tlbie.
But tlbie always has two operands (says Book III).
But tlbie always has one operand (says the PEM).
Furthermore, tlbiel is defined as having only one operand.
But tlbiel always has two operands (says Book III).
But tlbiel doesn't exist (says the PEM).
I suggest for now, we fix the cmp[l] thing, and make both
tlbie and tlbiel take either one or two operands.
If that's okay with you, I'll send some patches shortly.
Segher