Messages in this thread | | | Date | Mon, 30 Sep 2013 10:25:10 -0700 | Subject | Re: [PATCH, v2] anon_vmas: Convert the rwsem to an rwlock_t | From | Linus Torvalds <> |
| |
On Mon, Sep 30, 2013 at 10:03 AM, Andrew Morton <akpm@linux-foundation.org> wrote: > > Let's convince ourselves that we won't be undoing things which will > return to bite us?
Umm. We call them regressions, and we fix them.
As already mentioned, the original switch to a mutex didn't even have any good explanation for it, much less actual data.
When we make mistakes, we'd better realize it and fix them, rather than assume that they were fixes just because they were made.
The performance numbers are pretty damn compelling for this having been a major mistake.
(Of course, the current performance numbers also contain the conversion to a rwlock_t rather than back to a spinlock, which may actually help wrt the original situation too)
Linus
| |