Arch Linux User Repository

Pull requests are welcome on github.

Please report bugs on upstream GitLab.

Hi @chrhasse,

Thanks for research! I had added patched you mentioned, it works for me now.

I was finally able to find the issue about the black screen/kernel hang with GCC 10. It's right here and the fix can be found here. I also found the actual issue for the compilation failure I was experiencing here and the fix for it here.

Adding both the commits as patches to my PKGBUILD results in a working kernel compiled with gcc 10.1.0. I'm not sure if you're interested in including them in this repo @yurikoles, or if we should let Alex Deucher know these commits fix our builds and hope he's able to merge them fairly quickly.

@chrhasse,

I am experiencing the exact same issue. Build fails with the same error and re-compiling with your patch results in a blackscreen.

I'm actually getting a completely different error:

/usr/bin/ld: scripts/dtc/dtc-parser.tab.o:(.bss+0x20): multiple definition of `yylloc'; scripts/dtc/dtc-lexer.lex.o:(.bss+0x0): first defined here
collect2: error: ld returned 1 exit status
make[1]:  [scripts/Makefile.host:116: scripts/dtc/dtc] Error 1
  CC       /tmp/makepkg/linux-amd-git/src/linux-amd-git/tools/objtool/sigchain.o
make:  [Makefile:1260: scripts_dtc] Error 2

@chrhasse I've heard that there was an issue with black screens on boot upstream but that it was patched, I guess the issue hasn't been fixed in the agd5f repo. I don't know what patch fixed it, you'll have to wait until they merge the patch I guess. Or contact Alex at AMD and ask him about it. I think it's AlexDeucher@amd.com. He's the main AMD Linux dev, and agd5f is his handle on gitlab and other places. So he would be the person to talk to. But yeah I'm pretty sure this is a bug, not anything wrong with your setup.

Is anyone else experiencing build failures with GCC 10.1? I'm getting

arch/x86/kvm/../../../virt/kvm/kvm_main.c: In function ‘__kvm_gfn_to_hva_cache_init’:
arch/x86/kvm/../../../virt/kvm/kvm_main.c:2236:42: error: ‘nr_pages_avail’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
 2236 |  for ( ; start_gfn <= end_gfn; start_gfn += nr_pages_avail) {
      |                                ~~~~~~~~~~^~~~~~~~~~~~~~~~~

and I can solve that error and get it to compile with the patch

diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index 70f03ce0e5c1..dc8a67ad082d 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -2219,7 +2219,7 @@ static int __kvm_gfn_to_hva_cache_init(struct kvm_memslots *slots,
    gfn_t start_gfn = gpa >> PAGE_SHIFT;
    gfn_t end_gfn = (gpa + len - 1) >> PAGE_SHIFT;
    gfn_t nr_pages_needed = end_gfn - start_gfn + 1;
-   gfn_t nr_pages_avail;
+   gfn_t nr_pages_avail = 0;

    /* Update ghc->generation before performing any error checks. */
    ghc->generation = slots->generation;

but even though it builds with that patch, I get a black screen on boot.

The build is fixed

This has already been patched in the regular kernel: https://git.archlinux.org/svntogit/packages.git/tree/trunk/sphinx-workaround.patch?h=packages/linux&id=08a7fa88eaa91ef8b0e1f71b496d4097c3702e8d @yurikoles Could you pick that up?

What's going on with the sphinx nonsense? Do we just have to wait for whether they decide to fix it or not before this will work? Because he closed the issue, and it sounded like they're going to allow alternative implementations, but not fix what's broken.

The build is failing for me at the moment with: Extension error: Could not import extension cdomain (exception: cannot import name 'c_funcptr_sig_re' from 'sphinx.domains.c' (/usr/lib/python3.8/site-packages/sphinx/domains/c.py))

This points to this bug: https://github.com/sphinx-doc/sphinx/issues/7421

Not sure if the mentioned workaround would make sense at the moment or not: https://github.com/sphinx-doc/sphinx/issues/7421#issuecomment-609809210