On 13.03.2025 15:11, Andrew Cooper wrote: > On 13/03/2025 1:05 pm, Jan Beulich wrote: >> On 12.03.2025 18:45, Andrew Cooper wrote: >>> Now that there's a common stub implementation TLB clocks, there's no need >>> for >>> architectures to provide their own. >>> >>> Repeatedly zeroing page->tlbflush_timestamp is no use, so provide an even >>> more >>> empty common stub for page_set_tlbflush_timestamp(). >> At which point the field itself could in principle go away. There are three >> printk()s (accompanying BUG()s) which use it; surely we can find a way to >> abstract that out. This may then still be enough of a reason to introduce >> HAS_TLB_CLOCK. > > I wanted to remove the field, but it wasn't trivial, and I've probably > spent more time than I can afford on this.
I can understand this. It'll remain to be seen how useful HAS_TLB_CLOCK is with patch 4 corrected. And of course it's ... > I'm tempted to leave a TODO in tlb-clock.h to make it clear that there's > more that ought to be done. ... kind of okay to leave parts to be done later, as long as it's at least halfway clear what it is that wants doing. Jan
