cvs commit: src/sys/platform/pc32/i386 busdma_machdep.c
dillon at apollo.backplane.com
Tue Jun 5 11:01:44 PDT 2007
:>> This fixes NATA's UDMA device support. The UDMA dma controller is
:>> to a transfer size of 65536 bytes and address incrementation might not
:>> carry over to bit 16.
:> Good catch! Well hidden though.
:Indeed, I was staring myself blind on interrupts...
: Thomas E. Spanjaard
: tgen at netphreax.net
I'll just say this is why sometimes being able to reproduce a problem
locally, without having to debug it over email, can make a huge
difference. I've actually built DMA engines out of digital logic
(like 15 years ago at this point!), so the moment I saw the corruption
was on a page boundary and in page-sized chunks instead of in
I/O-sized chunks I knew what the problem had to be.
More information about the Commits