On Mon, Aug 29, 2022 at 9:24 PM Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> wrote: > > On Mon, 29 Aug 2022 16:57:31 +0200 Alexander Potapenko <glider@xxxxxxxxxx> wrote: > > > On Sat, Aug 27, 2022 at 6:17 AM Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> wrote: > > > > > > On Fri, 26 Aug 2022 17:07:27 +0200 Alexander Potapenko <glider@xxxxxxxxxx> wrote: > > > > > > > Use hooks from instrumented.h to notify bug detection tools about > > > > usercopy events in variations of get_user() and put_user(). > > > > > > And this one blows up x86_64 allmodconfig builds. > > > > How do I reproduce this? > > I tried running `make mrproper; make allmodconfig; make -j64` (or > > allyesconfig, allnoconfig) on both KMSAN tree > > (https://github.com/google/kmsan/commit/ac3859c02d7f40f59992737d63afcacda0a972ec, > > which is Linux v6.0-rc2 plus the 44 KMSAN patches) and > > linux-mm/mm-stable @ec6624452e36158d0813758d837f7a2263a4109d with > > KMSAN patches applied on top of it. > > All builds were successful. > > > > I then tried to cherry-pick just the first 4 commits to mm-stable and > > see if allmodconfig works - it resulted in numerous "implicit > > declaration of function ‘instrument_get_user’" errors (quite silly of > > me), but nothing looking like the errors you posted. > > I'll try to build-test every patch in the series after fixing the > > missing declarations, but so far I don't see other problems. > > > > Could you share the mmotm commit id which resulted in the failures? > > I just pushed out a tree which exhibits this with gcc-12.1.1 and with > gcc-11.1.0. Tag is mm-everything-2022-08-29-19-17. > > The problem is introduced by d0d9a44d2210 ("kmsan: add KMSAN runtime core") > > make mrproper > make allmodconfig > make init/do_mounts.o > > In file included from ./include/linux/kernel.h:22, > from ./arch/x86/include/asm/percpu.h:27, > from ./arch/x86/include/asm/nospec-branch.h:14, > from ./arch/x86/include/asm/paravirt_types.h:40, > from ./arch/x86/include/asm/ptrace.h:97, > from ./arch/x86/include/asm/math_emu.h:5, > from ./arch/x86/include/asm/processor.h:13, > from ./arch/x86/include/asm/timex.h:5, > from ./include/linux/timex.h:67, > from ./include/linux/time32.h:13, > from ./include/linux/time.h:60, > from ./include/linux/stat.h:19, > from ./include/linux/module.h:13, > from init/do_mounts.c:2: > ./include/linux/page-flags.h: In function ‘page_fixed_fake_head’: > ./include/linux/page-flags.h:226:36: error: invalid use of undefined type ‘const struct page’ > 226 | test_bit(PG_head, &page->flags)) { > | ^~ > ./include/linux/bitops.h:50:44: note: in definition of macro ‘bitop’ > 50 | __builtin_constant_p((uintptr_t)(addr) != (uintptr_t)NULL) && \ > | ^~~~ > ./include/linux/page-flags.h:226:13: note: in expansion of macro ‘test_bit’ > 226 | test_bit(PG_head, &page->flags)) { > | ^~~~~~~~ > ... Gotcha, this is a circular dependency: mm_types.h -> sched.h -> kmsan.h -> gfp.h -> mmzone.h -> page-flags.h -> mm_types.h, where the inclusion of sched.h into mm_types.h was only introduced in "mm: multi-gen LRU: support page table walks" - that's why the problem was missing in other trees. In fact sched.h only needs the definitions of `struct kmsan_context_state` and `struct kmsan_ctx` from kmsan.h, so I am splitting them off into kmsan_types.h to break this circle. Doing so also helped catch a couple of missing/incorrect inclusions of KMSAN headers in subsystems. I'll fix those and do more testing. -- Alexander Potapenko Software Engineer Google Germany GmbH Erika-Mann-Straße, 33 80636 München Geschäftsführer: Paul Manicle, Liana Sebastian Registergericht und -nummer: Hamburg, HRB 86891 Sitz der Gesellschaft: Hamburg
