On Fri, 16 Aug 2019, Andy Lutomirski wrote: > On 8/15/19 9:38 AM, Dmitry Safonov wrote: > > As it has been discussed on timens RFC, adding a new conditional branch > > `if (inside_time_ns)` on VDSO for all processes is undesirable. > > It will add a penalty for everybody as branch predictor may mispredict > > the jump. Also there are instruction cache lines wasted on cmp/jmp. > > > > Those effects of introducing time namespace are very much unwanted > > having in mind how much work have been spent on micro-optimisation > > vdso code. > > > > The propose is to allocate a second vdso code with dynamically > > patched out (disabled by static_branch) timens code on boot time. > > > > Allocate another vdso and copy original code. > > > I'm unconvinced that any of this magic is wise. I think you should make a > special timens vvar page that causes the normal fastpath to fail (using a > special vclock mode, a special seq value, or a special "last" value) and then > make the failure path detect that timens is in use and use the timens path. My initial suggestion still stands. Do that at compile time. It really does not matter whether we have another 2 or 3 variants of vdso binaries. Use it and be done with it. No special magic, just straight forward decisions to use a timens capable VDSO or not. Thanks, tglx
