On Thu, Dec 12, 2019 at 2:46 AM Peter Zijlstra [off-list ref] wrote:

+#ifdef GCC_VERSION < 40800

Where does that 4.8 version check come from, and why?

Yeah, I know, but this really wants a comment. Sadly it looks like gcc
bugzilla is down, so

   https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58145

currently gives an "Internal Server Error" for me.

[ Delete the horrid code we have because of gcc bugs ]

+#else /* GCC_VERSION < 40800 */
+
+#define READ_ONCE_NOCHECK(x)                                           \
+({                                                                     \
+       typeof(x) __x = *(volatile typeof(x))&(x);                      \

I think we can/should just do this unconditionally if it helps th eissue.

Maybe add a warning about how gcc < 4.8 might mis-compile the kernel -
those versions are getting close to being unacceptable for kernel
builds anyway.

We could also look at being stricter for the normal READ/WRITE_ONCE(),
and require that they are

 (a) regular integer types

 (b) fit in an atomic word

We actually did (b) for a while, until we noticed that we do it on
loff_t's etc and relaxed the rules. But maybe we could have a
"non-atomic" version of READ/WRITE_ONCE() that is used for the
questionable cases?

              Linus