Re: [Qemu-ppc] [PATCH 5/7] target-ppc: gdbstub: fix altivec registers for little-endian guests

[David Gibson]

On Tue, Jan 19, 2016 at 10:59:20AM +0100, Greg Kurz wrote:
> On Mon, 18 Jan 2016 13:25:19 +1100
> David Gibson <address@hidden> wrote:
> 
> > On Fri, Jan 15, 2016 at 04:00:38PM +0100, Greg Kurz wrote:
> > > Altivec registers are 128-bit wide. They are stored in memory as two
> > > 64-bit values that must be byteswapped when the guest is little-endian.
> > > Let's reuse the ppc_maybe_bswap_register() helper for this.
> > > 
> > > We also need to fix the ordering of the 64-bit elements according to
> > > the target endianness, for both system and user mode.
> > > 
> > > Signed-off-by: Greg Kurz <address@hidden>  
> > 
> > What bothers me about this is that avr_need_swap() now depends on both
> > host and guest endianness.  However the VSCR and VRSAVE swap - like
> > the swaps for GPRs and FPRs - uses ppc_maybe_bswap_register() which
> > depends only on guest endianness.
> > 
> > Why does altivec depend on the host endianness?
> > 
> 
> This has always been the case:
> 
> commit b4f8d821e5211bbb51a278ba0fc4a4db2d581221
> Author: aurel32 <address@hidden>
> Date:   Sat Jan 24 15:08:09 2009 +0000
> 
>     target-ppc: Add Altivec register read/write using XML
> 
> [...]
> 
> +static int gdb_get_avr_reg(CPUState *env, uint8_t *mem_buf, int n)
> +{
> +    if (n < 32) {
> +#ifdef WORDS_BIGENDIAN
> +        stq_p(mem_buf, env->avr[n].u64[0]);
> +        stq_p(mem_buf+8, env->avr[n].u64[1]);
> +#else
> +        stq_p(mem_buf, env->avr[n].u64[1]);
> +        stq_p(mem_buf+8, env->avr[n].u64[0]);
> +#endif
> +        return 16;
> +    }
> 
> My understanding is that gdb expects registers to be presented with
> the target endianness but QEMU have them in host endianness.
> 
> The ppc_maybe_bswap_register() helper is needed to fix 64-bit values
> according to the target effective endianness because stq_p() always
> consider both ppc64 and ppc64le to be big endian.
> 
> Here, we have a 128-bit register that we break into two 64-bit values
> in memory. Each quad word has to be fixed by ppc_maybe_bswap_register().
> But we also have to reorder these quad words if the host endianness
> differs from the target's one. This is the purpose of avr_need_swap().

Ok, understood.  I've merged the series to ppc-for-2.6

-- 
David Gibson                    | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au  | minimalist, thank you.  NOT _the_ _other_
                                | _way_ _around_!
http://www.ozlabs.org/~dgibson

signature.asc
Description: PGP signature