A small bogon sneaked into the ppc64 lockdep support. A test is branching slightly off causing a clobbered register value to overwrite the irq state under some circumstances.
Signed-off-by: Benjamin Herrenschmidt <[EMAIL PROTECTED]> --- This explains some of the reported weird lockdep errors, maybe even all of them, we'll see. We should probably backport that to 2.6.26-stable too, any candidate ? arch/powerpc/include/asm/irqflags.h | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) --- linux-work.orig/arch/powerpc/include/asm/irqflags.h 2008-08-15 16:48:43.000000000 +1000 +++ linux-work/arch/powerpc/include/asm/irqflags.h 2008-08-15 16:51:02.000000000 +1000 @@ -20,7 +20,7 @@ #define TRACE_ENABLE_INTS bl .trace_hardirqs_on #define TRACE_DISABLE_INTS bl .trace_hardirqs_off #define TRACE_AND_RESTORE_IRQ_PARTIAL(en,skip) \ - cmpdi en, 0; \ + cmpdi en,0; \ bne 95f; \ stb en,PACASOFTIRQEN(r13); \ bl .trace_hardirqs_off; \ @@ -29,7 +29,8 @@ li en,1; #define TRACE_AND_RESTORE_IRQ(en) \ TRACE_AND_RESTORE_IRQ_PARTIAL(en,96f); \ -96: stb en,PACASOFTIRQEN(r13) + stb en,PACASOFTIRQEN(r13); \ +96: #else #define TRACE_ENABLE_INTS #define TRACE_DISABLE_INTS _______________________________________________ Linuxppc-dev mailing list Linuxppc-dev@ozlabs.org https://ozlabs.org/mailman/listinfo/linuxppc-dev