On Fri, 31 Mar 2017, Wei Liu wrote: > On Fri, Mar 31, 2017 at 02:53:55PM +0100, Punit Agrawal wrote: > [...] > > > > Correct! > > > > invalidate_icache() flushes the entire instruction cache which ends up > > being called each time flush_page_to_ram() is invoked from > > alloc_heap_pages(). The patch prevents repeated calls to > > invalidate_icache() and moves the responsibility to populate_physmap() > > which was the intention. > > > > It would help if you can explain why you think the changes are > > wrong. > > > > OK. So it turns out I misunderstood what you wanted to do here. > > You can do this in a less confusing way. You don't need to fiddle with > a->memflags. Just add '| MEMF_no_icache_flush' to alloc_domheap_page. > And then you always call invalidate_icache at the end. > > This would avoid people asking question whether the check should be > inverted...
Yes, good suggestion. _______________________________________________ Xen-devel mailing list Xen-devel@lists.xen.org https://lists.xen.org/xen-devel
