Re: [PATCH v2] elf: Make string tunables startup-only
Yury Khrustalev <[email protected]> Thu, 30 Jul 2026 13:58:37 +0100
| Newsgroups | gmane.comp.lib.glibc.alpha |
|---|---|
| Message-ID | <[email protected]> |
On Thu, Jul 30, 2026 at 08:47:12AM -0300, Adhemerval Zanella Netto wrote:
>
> On 30/07/26 03:41, Andreas Schwab wrote:
> > On Jul 29 2026, Adhemerval Zanella wrote:
> >
> >> +/* Set by __tunable_seal_strings once the values of the string tunables are
> >> + no longer valid. */
> >> +static bool tunables_strings_sealed attribute_relro;
> >> +
> >> +static void __attribute__ ((noreturn))
> >> +tunable_sealed_error (const tunable_t *cur, const char *action)
> >> +{
> >> + _dl_fatal_printf ("Fatal glibc error: %s: string tunable %s after "
> >> + "process initialization\n", cur->name, action);
> >
> > I think this should be presented as an internal error, ie. it is not due
> > to some kind of resource constraint, but a glibc programming error.
> >
>
> Maybe something like:
>
> "internal error: %s: string tunable %s after process initialization\n"
>
> ?
Both "internal error" and "fatal glibc error" are used in several
places, I'm not sure which one applies here best. I think it's not an
internal error per se though it probably depends on whether it is a bug
in Glibc itself or a program trying to use private Glibc interface.
I think "fatal glibc error" is the best in this context.