On Tue, 5 Oct 2010, Paul Walmsley wrote: > __raw_writew(v, oh->_mpu_rt_va + reg_offs); > 28: 16ff3070 uxthne r3, r0 > 2c: 118130b2 strhne r3, [r1, r2] > > gcc for some reason decides to zero-extend it first, but that is not your > bug to fix. Unlikely anyone cares too much about this, but just to follow up, this appears to be caused by using -march=armv6 or -march=armv7-a on the gcc command line. Dropping the -march or using -march=armv5 avoids the uxt. That's with arm-linux-gcc (Sourcery G++ Lite 2008q3-72) 4.3.2. Minimal case follows. This appears to be gcc bug 40893: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40893 - Paul void foo(unsigned int v) { *(volatile unsigned short *)0xabcdefab = (v); } arm-linux-gcc -O2 -march=armv7-a -c test.c; arm-linux-objdump -DS test.o | less 00000000 <foo>: 0: e30e3fff movw r3, #61439 ; 0xefff 4: e34a3bcd movt r3, #43981 ; 0xabcd 8: e6ff0070 uxth r0, r0 c: e14305b4 strh r0, [r3, #-84] 10: e12fff1e bx lr -- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
