Thank you for a clear explanation. I agree that the device is old and if the current driver needs to be rewritten, it is better to focus on more important tasks.
Best wishes for the New Year! Andrzej On Wed, Dec 30, 2015 at 05:11:37PM -0700, Theo de Raadt wrote: > The driver must be rewritten to use bus_dma(9). > > It will not be re-enabled as-is, since many architectures will never > have vtophys. In particular, amd64. > > "testing" won't help. Nor is there any other information needed. The > problem is obvious. To a kernel developer, the commit message speaks > far enough to the issue. > > If you don't have the skill to do it, well that's too bad... there > also appears to be a lack of people who want to rewrite the driver. > It is a very old and rare chip. There are far more important issues > than supporting such ancient devices. > > So you are better off getting some other device. > > >I got access to a D-Link DFE-580TX PCI network device (Sundance Technologies > >ST201 10/100 Ethernet device) recently and decided to do some preliminary > >experiments on amd64 involving the corresponding driver on OpenBSD. I will > >be using this device for about a week for research purposes. > > > >The driver seems to have been temporarily disabled in the default kernel > >configuration (GENERIC, rev 1.70) on amd64, while it remains enabled on > >i386. The respective commit description was: "remove support for sf and ste. > > vtophys is NOT a working solution. Do not re-enable these drivers until > >they are bus-dma'ified.". > > > >Currently, the following warning/error appears on amd64 after activating the > >driver: > > > >cc1: warnings being treated as errors > >../../../../dev/pci/if_ste.c: In function 'ste_newbuf': > >../../../../dev/pci/if_ste.c:963: warning: implicit declaration of function > >'vtophys' > >*** Error 1 in /usr/src/sys/arch/amd64/compile/GENERIC.MP (Makefile:938 > >'if_ste.o') > > > >I have verified (and thus can confirm) that the ste* driver in FreeBSD 10.2 > >RELEASE works on amd64. Apparently it does not use vtophys anymore though. > > > >I wonder if I could help with testing, or deliver specific information that > >might be necessary to improve the corresponding driver in OpenBSD (amd64)? > > > >Best wishes, > >Andrzej > > > >
