[openib-general] Re: [PATCH 53 of 53] ipath - add memory barrier when waiting for writes

[ralphc at pathscale.com]

>     ralphc> We had a power failure here so I'm not able to reproduce
>     ralphc> the assembly code at the moment.  What I remember from
>     ralphc> looking at the code is that the code for
>     ralphc> ipath_read_kreg32() was present in i2c_wait_for_writes()
>     ralphc> when compiled -Os so it didn't look like a compiler bug.
>     ralphc> I probably could put the mb() at the end of i2c_gpio_set()
>     ralphc> if that makes you more comfortable.  The mb() is
>     ralphc> definitely needed though.
>
> Is it the mb()?  Or is just a barrier() enough?  In other words do you
> really need the mfence, or do you just need to stop the compiler from
> reordering things?
>
>  - R.

I didn't try calling barrier() so I don't know the answer.
When power is restored, I can try it.
My guess is that it's a timing issue and not a code reordering
issue.