>>> On 07.09.18 at 16:17, <andrew.coop...@citrix.com> wrote: > On 05/03/18 09:27, Jan Beulich wrote: >>>>> On 27.02.18 at 15:50, <andrew.coop...@citrix.com> wrote: >>> -compat_create_bounce_frame: >>> - ASSERT_INTERRUPTS_ENABLED >>> - mov %fs,%edi >>> - ASM_STAC >>> - testb $2,UREGS_cs+8(%rsp) >>> - jz 1f >>> - /* Push new frame at registered guest-OS stack base. */ >>> - movl VCPU_kernel_sp(%rbx),%esi >>> -.Lft1: mov VCPU_kernel_ss(%rbx),%fs >> Note how we did take into consideration the segment base here; >> pv_create_bounce_frame() doesn't. Hence while the patch here >> is >> Reviewed-by: Jan Beulich <jbeul...@suse.com> >> I'm afraid I have to withdraw the respective tag for the earlier one >> (despite realizing that there are other places where we [wrongly] >> assume stack segments to be flat). > > For the failsafe callback, %ss is set to be flat, and then a bounce > frame is created at the current kernel_sp. > > Despite the impression the API might give, a 32bit PV kernel cannot use > a non-flat stack segment. No PV guest (not even MiniOS) uses a non-flat > layout, so while this is a change of behaviour, its not going to break > anything.
Well, at the very least such a change in behavior needs to be called out very prominently in the description. Jan _______________________________________________ Xen-devel mailing list Xen-devel@lists.xenproject.org https://lists.xenproject.org/mailman/listinfo/xen-devel
