Messages in this thread | | | From | Andy Lutomirski <> | Date | Thu, 13 Mar 2014 12:35:30 -0700 | Subject | Re: [PATCH 3/3] x86, vdso32: handle 32 bit vDSO larger one page |
| |
On Thu, Mar 13, 2014 at 12:21 PM, Linus Torvalds <torvalds@linux-foundation.org> wrote: > > On Mar 13, 2014 12:13 PM, "Andy Lutomirski" <luto@amacapital.net> wrote: >> >> How? We either set null in AT_SYSINFO_EHDR, we set something simple >> that SuSE is okay with (and compile separate images?) or we set >> something fancy and not relocated, and SuSE crashes. > > Yeah, I was thinking that "something simple is the fixmap space that suse > would be happy with" > > Keyword being "simple", ie so simple that it's not a maintenance headache > even if it's different from the "real" vdso. > > But I suspect we don't really need it.
I agree. Also, the syscall instruction selection is currently rather complicated, and I don't think it can ever be *that* simple :(
> >> On a related note: is it okay to keep the option called >> CONFIG_COMPAT_VDSO but change the default to n? > > Yes. If you start from scratch with a config, you have to get the options > right. It's just the "I upgraded the kernel, did an oldconfig, and it > doesn't work" case that we should avoid. > > The alternative would be to make a new config name, but then it has to > default to the compatible state (iow no vdso at all by default), because > that's the rule: make oldconfig should work for people. > > Basically, kernel upgrades should be nobrainers and "just work".
OK. I'll send out new patches.
--Andy
| |