Re: [Xen-devel] [for-4.7] x86/emulate: synchronize LOCKed instruction emulation

["Jan Beulich" <jbeulich@xxxxxxxx>]

>>> Razvan Cojocaru <rcojocaru@xxxxxxxxxxxxxxx> 04/14/16 5:45 PM >>>
>On 04/14/2016 06:40 PM, Jan Beulich wrote:
>> To be honest, just having remembered that we do the write back for locked
>> instructions using CMPXCHG, I'd first of all like to see a proper description
>> of "the _whole_ issue".
>
>I believe at least part of the issue has to do with the comment on line
>1013 from xen/arch/x86/hvm/emulate.c:
>
 >994 static int hvmemul_cmpxchg(
 >995     enum x86_segment seg,
 >996     unsigned long offset,
 >997     void *p_old,
 >998     void *p_new,
 >999     unsigned int bytes,
>1000     struct x86_emulate_ctxt *ctxt)
>1001 {
>1002     struct hvm_emulate_ctxt *hvmemul_ctxt =
>1003         container_of(ctxt, struct hvm_emulate_ctxt, ctxt);
>1004
>1005     if ( unlikely(hvmemul_ctxt->set_context) )
>1006     {
>1007         int rc = set_context_data(p_new, bytes);
>1008
>1009         if ( rc != X86EMUL_OKAY )
>1010             return rc;
>1011     }
>1012
>1013     /* Fix this in case the guest is really relying on r-m-w atomicity. */
>1014     return hvmemul_write(seg, offset, p_new, bytes, ctxt);
>1015 }

Ah, so _that's_ where the problem wants to be fixed then (leaving - afaict -
PV emulation paths completely unaffected).

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel