Re: [Qemu-devel] [PATCH 29/43] linux-user: Fix broken m68k signal handling on 64 bit hosts

[Peter Maydell]

On 25 February 2015 at 20:58, Laurent Vivier <address@hidden> wrote:
>
>> Le 25 février 2015 à 12:28, Peter Maydell <address@hidden> a
>> écrit :
>>
>>
>> On 25 February 2015 at 17:39, Laurent Vivier <address@hidden> wrote:
>> > Hi,
>> >
>> > I think you should use abi_long instead of uint32_t.
>> >
>> > abi_long has an "aligned" attribute, and on m68k long are aligned on a
>> > short
>> > boundary.
>> >
>> >
>> > #ifdef TARGET_M68K
>> > #define ABI_INT_ALIGNMENT 2
>> > #define ABI_LONG_ALIGNMENT 2
>> > #define ABI_LLONG_ALIGNMENT 2
>> > #endif
>> >
>> > typedef uint32_t abi_ulong __attribute__((aligned(ABI_LONG_ALIGNMENT)));
>>
>> In this instance it doesn't matter because get_sigframe() aligns
>> the pointer we're going to write the sigframe to, and the elements
>> in the struct before retcode[] won't make it worse than 4-aligned,
>> I think?
>
> Yes, I agree. But the aim of the abi_* types is to define the target ABI.
> Thus, for consistency it should better to use the abi_long (or abi_ulong)
> instead of uint32_t.

Well, strictly speaking this is writing instructions into a
byte array rather than ABI longs, so it ought to do it
byte-at-a-time to respect the definition of the struct.
Anyway, if you want to submit a patch I don't object.

-- PMM