[PATCH] x86: Add thresholds for "rep movsb/stosb" to tunables

Carlos O'Donell carlos@redhat.com
Mon Jun 1 20:56:58 GMT 2020


On Mon, Jun 1, 2020 at 4:48 PM Florian Weimer <fw@deneb.enyo.de> wrote:
>
> * H. J. Lu via Libc-alpha:
>
> > On Mon, Jun 1, 2020 at 1:15 PM H.J. Lu <hjl.tools@gmail.com> wrote:
> >>
> >> On Mon, Jun 1, 2020 at 12:38 PM Carlos O'Donell <carlos@redhat.com> wrote:
> >> >
> >> > On Mon, Jun 1, 2020 at 3:33 PM H.J. Lu <hjl.tools@gmail.com> wrote:
> >> > > Did you mean adding  --list-tunables to ld.so?  libc.so.6 doesn't take
> >> > > any arguments.
> >> >
> >> > Yes, I mean adding argument processing to libc.so.6, and handling
> >> > --list-tunables.
> >> >
> >> > We have enough infrastructure in place that wiring that up shouldn't be too bad?
> >> >
> >> > Then, even in trimmed down containers, you can just run
> >> > /lib64/libc.so.6 --list-tunables and get back the list of tunables and
> >> > their min, max, and security values.
> >>
> >> Adding an argument to libc.so.6 is difficult since argument passing is
> >> processor specific.  Adding --list-tunables to ld.so? is more doable.
> >
> > But tunables are in libc.so.
>
> If this is really a problem, we can load libc.so and call a
> GLIBC_PRIVATE function to print the information.

Agreed.

Please keep in mind the original problem we are trying to solve.

We want a tunable for a parameter that is difficult to explain to the user.

To make it easier for our users to use the tunable we are going to
provide them a way to look at the tunable settings in detail.

Yes, it requires a target system, but we can't avoid that in some cases.

Cheers,
Carlos.



More information about the Libc-alpha mailing list