[lkml]   [2006]   [Sep]   [8]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: Opinion on ordering of writel vs. stores to RAM

On Sat, 9 Sep 2006, Paul Mackerras wrote:
> Do you have an opinion about whether the MMIO write in writel() should
> be ordered with respect to preceding writes to normal memory?

It shouldn't. It's too expensive. The fact that PC's have nice memory
consistency models means that most of the testing is going to be with the
PC memory ordering, but the same way we have "smp_wmb()" (which is also a
no-op on x86) we should probably have a "mmiowb()" there.

Gaah. Right now, mmiowb() is actually broken on x86 (it's an empty define,
so it may cause compiler warnings about statements with no effects or

I don't think anyting but a few SCSI drivers that are used on some ia64
boxes use mmiowb(). And it's currently a no-op not only on x86 but also on
powerpc. Probably because it's defined to be a barrier between _two_ MMIO
operations, while we should probably have things like

.. regular mem store ..
.. IOMEM store ..

.. IOMEM store ..
.. regular mem store ..

although it's quite possible that (a) never makes any sense at all.

That said, it's also entirely possible that what you _should_ do is to
just make sure that the "sync" is always _before_ the IO op. But that
would require that you have one before an IO load too. Do you? I'm too
lazy to check.

(Keeping it inside a spinlock, I don't know. Spinlocks aren't _supposed_
to order IO, so I don't _think_ that's necessarily an argument for doing
so. So your rationale seems strange. Even on x86, a spinlock release by
_no_ means would mean that an IO write would be "done").

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
More majordomo info at
Please read the FAQ at

 \ /
  Last update: 2009-11-18 23:46    [W:0.085 / U:13.436 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site