On Thursday 03 November 2016 03:18 PM, Michael Ellerman wrote: > Ravi Bangoria <ravi.bango...@linux.vnet.ibm.com> writes: > >> On Thursday 03 November 2016 02:34 AM, Anton Blanchard wrote: >>> Hi Ravi, >>> >>>> emulate_step() uses a number of underlying kernel functions that were >>>> initially not enabled for LE. This has been rectified since. So, fix >>>> emulate_step() for LE for the corresponding instructions. >>> Thanks. Should this be queued up for stable? >> Thanks Anton. Yes, this should go in stable. > It's fairly big for stable. Does it fix an actual bug? If so what, and > how bad is it, what's the user impact.
Hi Michael, Yes, kernel-space hw-breakpoint feature is broken on LE without this. Actually, there is no specific commit that introduced this. Back in 2010, Paul Mackerras has added emulation support for load/store instructions for BE. hw-breakpoint was also develpoed by K.Prasad in the same timeframe. Kernel-space hw-breakpoint emulates causative instruction before notifying to user. As emulate_step is never enabled for LE, kernel- space hw-breakpoint is always broken on LE. -Ravi > Can you also pinpoint which commit it "fixes"? > > cheers >
