On 05/03/2025 9:23 am, Jan Beulich wrote: > On 04.03.2025 00:29, Andrew Cooper wrote: >> struct domain is always a page aligned allocation. Update it's type to >> reflect this, so we can safely reuse the lower bits in the pointer for >> auxiliary information. >> >> No functional change. >> >> Signed-off-by: Andrew Cooper <andrew.coop...@citrix.com> >> --- >> CC: Anthony PERARD <anthony.per...@vates.tech> >> CC: Michal Orzel <michal.or...@amd.com> >> CC: Jan Beulich <jbeul...@suse.com> >> CC: Julien Grall <jul...@xen.org> >> CC: Roger Pau Monné <roger....@citrix.com> >> CC: Stefano Stabellini <sstabell...@kernel.org> >> CC: Dario Faggioli <dfaggi...@suse.com> >> CC: Juergen Gross <jgr...@suse.com> >> CC: George Dunlap <g...@xenproject.org> >> >> v2: >> * New >> >> Interestingly this does cause two changes in the resulting binary, both >> caused >> by GCC electing to use a more complicated addressing mode to save one ADD >> instruction. > That's on x86, I suppose?
Yes. > >> --- a/xen/include/xen/sched.h >> +++ b/xen/include/xen/sched.h >> @@ -645,7 +645,7 @@ struct domain >> unsigned int num_llc_colors; >> const unsigned int *llc_colors; >> #endif >> -}; >> +} __aligned(PAGE_SIZE); >> >> static inline struct page_list_head *page_to_list( >> struct domain *d, const struct page_info *pg) > I understand struct domain is where you need the annotation right away, but is > there a reason not to do the same for struct vcpu right away? struct vcpu is more complicated. It's multi-page on ARM, and I have a strong suspicion that alignment is going to have to change away from PAGE_SIZE for architectures wanting to use larger page sizes. e.g. with 64k pagetables, I expect these will still want 4k alignment, and therefore they'll all want changing. i.e. I'm probably creating less work for someone in the future by not annotating struct vcpu. ~Andrew
