Messages in this thread | | | Date | Wed, 22 Apr 2015 13:12:41 -0700 | From | Josh Triplett <> | Subject | Re: [PATCH 2/2] x86/asm/entry/32: Remove unnecessary optimization in stub32_clone |
| |
On Wed, Apr 22, 2015 at 11:22:02AM -0700, Linus Torvalds wrote: > On Wed, Apr 22, 2015 at 10:10 AM, Josh Triplett <josh@joshtriplett.org> wrote: > > > > I do think my two-patch HAVE_COPY_THREAD_TLS series should go in fixing > > this > > Ugh, I absolutely detesrt that patch. > > Don't make random crazy function signatures that depend on some config > option. That's just evil. The patch is a mess of #ifdef's and should > be shot in the head and staked with a silver stake to make sure it > never re-appears. > > Either: > > (a) make the change for every architecture > > (b) have side-by-side interfaces. With different names!
...that's exactly what I did. They're called copy_thread and copy_thread_tls; I very intentionally did not conditionally change the signature of copy_thread, for exactly that reason. Those functions are implemented in architecture-specific code, so the config option just specifies which of the two functions the architecture provides.
*sys_clone* has different function signatures based on config options, but I didn't touch that other than fixing the type of the tls argument. That's historical baggage that we can't throw away without breaking userspace.
- Josh Triplett
| |