[OpenWrt-Devel] [PATCH] mac80211: switch to upstream owl-loader driver

[Russell Senior]

This commit broke wifi on the Buffalo WZR600DHP.  See:
https://bugs.openwrt.org/index.php?do=details&task_id=2668

On Fri, Nov 22, 2019 at 12:00 PM Christian Lamparter <chunkeey at gmail.com>
wrote:

> On Monday, 18 November 2019 00:34:01 CET Hauke Mehrtens wrote:
> > > +--- a/drivers/net/wireless/ath/ath9k/ath9k_pci_owl_loader.c
> > > ++++ b/drivers/net/wireless/ath/ath9k/ath9k_pci_owl_loader.c
> > > +@@ -84,6 +84,10 @@
> > > +                   val = swahb32(val);
> > > +           }
> > > +
> > > ++#ifdef CONFIG_LANTIQ
> > > ++          val = swab32(val);
> > > ++#endif
> >
> > Lantiq is big endian, are there other big endian system which do not
> > need this byte swap?
>
> From what I vaguely remember (I know that Mathias explained it to me
> once.),
> that special hack was necessary due to Lantiq's pci(e?)-host silicon doing
> byteswaps just for 32-bit writes. The only other system that uses the
> owl-loader
> is ath79/ar71xx. This is a big-endian MIPS as well that didn't need the
> swap.
>
> (That said, I don't remember what was the reason for going with
> __raw_writel
> rather than "iowrite32" though. At least ath9k is using it for the pci
> access
> just fine everywhere.)
>
> Anyone fancy checking out lantiq and ath79 devices with a AR92XX without
> the
> swap above and the __raw_writel replaced by iowrite32?
>
> > > ++
> > > +           __raw_writel(val, mem + reg);
> > > +           usleep_range(100, 120);
> > > +   }
>
> Regards,
> Christian
>
>
>
> _______________________________________________
> openwrt-devel mailing list
> openwrt-devel at lists.openwrt.org
> https://lists.openwrt.org/mailman/listinfo/openwrt-devel
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.infradead.org/pipermail/openwrt-devel/attachments/20191211/02068383/attachment.htm>