Re: [Qemu-devel] [PATCH 1/3] memory_ldst: Add atomic ops for PTE updates

[Benjamin Herrenschmidt]

On Thu, 2018-12-13 at 21:01 -0600, Richard Henderson wrote:
> On 12/13/18 5:58 PM, Benjamin Herrenschmidt wrote:
> > +#ifdef CONFIG_ATOMIC64
> > +/* This is meant to be used for atomic PTE updates under MT-TCG */
> > +uint32_t glue(address_space_cmpxchgq_notdirty, SUFFIX)(ARG1_DECL,
> > +    hwaddr addr, uint64_t old, uint64_t new, MemTxAttrs attrs, MemTxResult 
> > *result)
> > +{
> > +    uint8_t *ptr;
> > +    MemoryRegion *mr;
> > +    hwaddr l = 8;
> > +    hwaddr addr1;
> > +    MemTxResult r;
> > +    uint8_t dirty_log_mask;
> > +
> > +    /* Must test result */
> > +    assert(result);
> > +
> > +    RCU_READ_LOCK();
> > +    mr = TRANSLATE(addr, &addr1, &l, true, attrs);
> > +    if (l < 8 || !memory_access_is_direct(mr, true)) {
> > +        r = MEMTX_ERROR;
> > +    } else {
> > +        uint32_t orig = old;
> > +
> > +        ptr = qemu_map_ram_ptr(mr->ram_block, addr1);
> > +        old = atomic_cmpxchg(ptr, orig, new);
> > +
> 
> I think you need atomic_cmpxchg__nocheck here.
> 
> Failure would be with a 32-bit host that supports ATOMIC64.
> E.g. i686.

I'm confused by this and the comments around the definition of
ATOMIC_REG_SIZE :)

So would we have CONFIG_ATOMIC64 in that case and if yes why if all the
atomic_* end up barfing ?

Or rather, why set CONFIG_ATOMIC64 if we ought not to use 64-bit
atomics ?

Also we should probably define ATOMIC_REG_SIZE to 8 for ppc64...

Cheers
Ben.
> 
> r~