y...@mwd.biglobe.ne.jp (YAMAMOTO Takashi) wrote: > > Not sure what do you mean by leak. The first page is unmapped and uarea > > layout in x86 has not changed (yet). Probably I am missing something? > > the page in question used to be hanged on the kernel aobj. > now you made it wired and there's no way to reach it once unmapped.
Not sure what "hanged on the kernel aobj" means. :) But I see the problem. By wired you mean mapping is unmanaged now, therefore redzone page unmap loses the track of physical page, and thus it gets leaked. I wonder what would be a good way to fix this. Perhaps allocate VA and manually enter physical pages (leaving the gap for redzone)? Would get a little bit more complex (as for diagnostic case). On the other hand, I think it would be good to setup redzone at MI level, instead of strictly keeping it MD. Also, is there a reason to have x86 uarea layout different from other ports? Thanks a lot for reviewing this! -- Mindaugas
