[PATCH v4 4/6] infiniband: cxgb4: Eliminate duplicate barriers on weakly-ordered archs

[jgg@ziepe.ca (Jason Gunthorpe)]

On Thu, Mar 22, 2018 at 10:28:13AM -0400, Sinan Kaya wrote:
>    On 03/22/2018 08:48 AM, [1]okaya at codeaurora.org wrote:
> 
>      Jason,
>      Can you remove the writeq change if it is too late for me to fix?
>      This is an infrastructural issue on xtensa arch.
>      Probably, it won't get fixed today.

I was able to drop the patch, please resend.

>      AFAIS, even writeq won't compile on this arch. I started questioning
>      this build test.

>    I found out that the solution is this:
> #include <linux/io-64-nonatomic-hi-lo.h>

Yuk, what an ugly API..

>    [2]https://patchwork.ozlabs.org/patch/511801/
>    I did a compile test with this change on xtensa and it passed. I'll
>    repost with the added diff.
>    +++ b/drivers/infiniband/hw/cxgb4/iw_cxgb4.h
>    @@ -46,7 +46,7 @@
>     #include <linux/timer.h>
>     #include <linux/io.h>
>     #include <linux/workqueue.h>
>    -
>    +#include <linux/io-64-nonatomic-hi-lo.h>

I think this is the wrong one. I would expect all PCI-E devices should
use lo-hi, eg writes are done in address increasing order on the bus.

This is what the PCI-E spec would require if a write were to be
fragmented so I would expect devices to handle it properly.

Steve? Any idea of specific things for this HW?

Jason