On 10.03.20 16:08, Alexander Duyck wrote: > On Tue, Mar 10, 2020 at 5:13 AM David Hildenbrand <david@xxxxxxxxxx> wrote: >> >> On 10.03.20 13:02, Anup Patel wrote: >>> On Tue, Mar 10, 2020 at 3:56 PM David Hildenbrand <david@xxxxxxxxxx> wrote: >>>> >>>> On 10.03.20 11:22, David Hildenbrand wrote: >>>>> On 10.03.20 03:19, Alexander Duyck wrote: >>>>>> On Sun, Mar 8, 2020 at 7:54 PM kbuild test robot <lkp@xxxxxxxxx> wrote: >>>>>>> >>>>>>> tree: https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git master >>>>>>> head: b86a6a241b7c60ca7a6ca4fb3c0d2aedbbf2c1b6 >>>>>>> commit: 120359931943d2b801ce51a1a045dcc0a5d1a55b [7003/7050] Merge branch 'akpm-current/current' >>>>>>> config: riscv-randconfig-a001-20200308 (attached as .config) >>>>>>> compiler: riscv64-linux-gcc (GCC) 7.5.0 >>>>>>> reproduce: >>>>>>> wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross >>>>>>> chmod +x ~/bin/make.cross >>>>>>> git checkout 120359931943d2b801ce51a1a045dcc0a5d1a55b >>>>>>> # save the attached .config to linux build tree >>>>>>> GCC_VERSION=7.5.0 make.cross ARCH=riscv >>>>>>> >>>>>>> If you fix the issue, kindly add following tag >>>>>>> Reported-by: kbuild test robot <lkp@xxxxxxxxx> >>>>>>> >>>>>>> All errors (new ones prefixed by >>): >>>>>>> >>>>>>> drivers/virtio/virtio_balloon.o: In function `leak_balloon': >>>>>>> drivers/virtio/virtio_balloon.c:281: undefined reference to `balloon_page_dequeue' >>>>>>> drivers/virtio/virtio_balloon.o: In function `__virtio_test_bit': >>>>>>>>> include/linux/virtio_config.h:113: undefined reference to `page_reporting_unregister' >>>>>>> drivers/virtio/virtio_balloon.o: In function `virtio_cread32': >>>>>>> include/linux/virtio_config.h:423: undefined reference to `balloon_page_alloc' >>>>>>> drivers/virtio/virtio_balloon.o: In function `fill_balloon': >>>>>>>>> drivers/virtio/virtio_balloon.c:233: undefined reference to `balloon_page_enqueue' >>>>>>> drivers/virtio/virtio_balloon.o: In function `__virtio_test_bit': >>>>>>>>> include/linux/virtio_config.h:113: undefined reference to `page_reporting_register' >>>>>>> drivers/gpu/drm/virtio/virtgpu_drv.o: In function `virtio_gpu_remove': >>>>>>> drivers/gpu/drm/virtio/virtgpu_drv.c:139: undefined reference to `drm_dev_unplug' >>>>>>> drivers/gpu/drm/virtio/virtgpu_drv.c:140: undefined reference to `drm_atomic_helper_shutdown' >>>>>>> drivers/gpu/drm/virtio/virtgpu_drv.c:141: undefined reference to `drm_dev_put' >>>>>>> drivers/gpu/drm/virtio/virtgpu_drv.o: In function `virtio_gpu_probe': >>>>>> >>>>>> It looks like somehow the config has CONFIG_VIRTIO_BALLOON enabled, >>>>>> but didn't select CONFIG_MEMORY_BALLOON nor CONFIG_PAGE_REPORTING That >>>>>> shouldn't be possible since in drivers/virtio/Kconfig we have the >>>>>> following: >>>>>> config VIRTIO_BALLOON >>>>>> tristate "Virtio balloon driver" >>>>>> depends on VIRTIO >>>>>> select MEMORY_BALLOON >>>>>> select PAGE_REPORTING >>>>> >>>>> IMHO that's perfectly valid. You might want to have ballooning without >>>>> page compaction (MEMORY_BALLOON). Same with PAGE_REPORTING. >>>>> >>>>> AFAIK, "select" will still allow you to disable these things - which >>>>> used to work fine with MEMORY_BALLOON (did that myself when debugging a >>>>> compaction issue). >>>>> >>>> >>>> Correction: I disabled BALLOON_COMPACTION back then ... so ignore my >>>> comment regarding that :) >>>> >>>> But we should be able to build VIRTIO_BALLOON without PAGE_REPORTING >>>> somehow IMHO. >>> >>> I have moved all VIRTIO driver selection to defconfigs for RISC-V which was >>> the case before. >>> >>> Here's the patch: >>> https://patchwork.kernel.org/patch/11429217/ >> >> Make perfect sense. Still I think we might want to have the option to >> build without page reporting in the future, similar to building without >> balloon compaction (which is another discussion). > > As far as page reporting it would be pretty simple to make it so that > we can build without having it enabled. It would just depend on how we > want to go about it. > > My past experience has been that you end up having to create a third > Kconfig option that depends on VIRTIO_BALLOON and selects the other > config item. Then it is just a matter of figuring out if we want to Yes, came to the same conclusion. > use the new define as a wrapper in the virtio balloon driver, or if we > want to create stubbed-out versions of page reporting and balloon > compaction. At least balloon compaction has stubs but still requires special handling in the balloon drivers (e.g., skip balloon mount etc.) Nothing urgent. -- Thanks, David / dhildenb
