27 Mar
2014
27 Mar
'14
10:52 p.m.
On Wed, 2014-03-26 at 10:01 +0100, Wolfgang Denk wrote:
I'm not an expert for ARM, but this indeed looks suspiscious - thanks for reporting this.
FYI I made the change which prompted this and the resulting code was the same see https://groups.google.com/forum/#!topic/linux-sunxi/REZ18q0wcDY The barriers were only ever compile barrier() type, not cpu barriers.
So the open coded thing was effectively: v = read(); barrier(); v = fiddlebits(b) barrier(); write(v);
Whereas the code using clrsetbits is basically: write(fiddlebits(read()))
Which I think is OK in this case at least. Not sure about the general case.
Ian.