Re: [GIT PULL] locking changes for v4.4

[Linus Torvalds]

On Tue, Nov 3, 2015 at 1:16 AM, Ingo Molnar <mingo@xxxxxxxxxx> wrote:
>
>   - More gradual enhancements to atomic ops: new atomic*_read_ctrl() ops,
>     synchronize atomic_{read,set}() ordering requirements between architectures,
>     add atomic_long_t bitops. (Peter Zijlstra)

>From another thread: those new "atomic*_read_ctrl()" operations are
complete voodoo programming, and should never ever be used.

Those helpers seem to be based entirely on a mis-reading of alpha
memory ordering that is actually not possible in a universe where
causality exists.

It's not a new disease - we already have READ_ONCE_CTRL(), but it is
currently only actually used in one single place (but mentioned many
times in documentation that is looking less and less like reality).

But this pull request seems to make that thing be institutional, and
spreads this mis-understanding out, and tries to "document" it as some
kind of actual truth.

                  Linus
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/