On Tue, Jul 25, 2017 at 3:52 PM, Ingo Molnar <mingo@xxxxxxxxxx> wrote: > > * Ingo Molnar <mingo@xxxxxxxxxx> wrote: > >> >> * Dmitry Vyukov <dvyukov@xxxxxxxxxx> wrote: >> >> > On Sat, Jul 8, 2017 at 7:56 PM, Mike Galbraith <efault@xxxxxx> wrote: >> > > Greetings, >> > > >> > > The commit below results in early boot silent death with virgin tip >> > > v4.12-1134-g813b2061457c and the attached config in my garden variety >> > > i7-4790 box. A kvm box (w. full tune-for-max-buildtime config) boots >> > > just fine, it's only toxic on bare metal. >> > >> > Hi Mike, >> > >> > Yes, there is a known issue with this commit. I think it was reverted >> > now. See this thread for details: >> > https://groups.google.com/forum/#!topic/kasan-dev/ZJl66N7smmk >> >> It was not reverted - it was moved into WIP.locking/atomics, but still present in >> tip:master. >> >> Is there a fix for WIP.locking/atomics (if yes then please resend it) - if not >> I'll drop the branch from tip:master as well. > > Since there haven't been any updates let's re-try this again, ok? > > I had these commits in WIP.locking/atomics: > > 4b47cc154eed: locking/atomic/x86, asm-generic: Add comments for atomic instrumentation > 35787d9d7ca4: locking/atomics, asm-generic: Add KASAN instrumentation to atomic operations > 68c1ed1fdb0a: kasan: Allow kasan_check_read/write() to accept pointers to volatiles > f1c3049f6729: locking/atomic/x86: Switch atomic.h to use atomic-instrumented.h > d079eebb3958: locking/atomic: Add asm-generic/atomic-instrumented.h > 007d185b4462: locking/atomic/x86: Use 's64 *' for 'old' argument of atomic64_try_cmpxchg() > ba1c9f83f633: locking/atomic/x86: Un-macro-ify atomic ops implementation > > Thomas objected to the ugly instrumentation approach - so I'll keep only these two > patches: > > 007d185b4462: locking/atomic/x86: Use 's64 *' for 'old' argument of atomic64_try_cmpxchg() > ba1c9f83f633: locking/atomic/x86: Un-macro-ify atomic ops implementation > > because they appear to be uncontroversial. > > I've pushed these two out into tip:locking/core, future (re-)submissions should be > based on that tree. Hi Ingo, It's still in my sight, but was preempted by other tasks. Also I am somewhat scared by the amount of changes for the final plan. Thanks for taking these 2. That's a good start. Can you please also merge: 68c1ed1fdb0a: kasan: Allow kasan_check_read/write() to accept pointers to volatiles It's small and harmless and will be needed anyway. For the rest of the changes: will it be ok to just resolve the cmpxchg macro mess (that caused the crash) and then rebase the current changes on top of that? Or we need to do the full overhaul across all arches? If we go with full overhaul, would somebody with more kernel expertise be willing to do the overhaul? -- To unsubscribe from this list: send the line "unsubscribe linux-tip-commits" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
|