New kernels

[Roland]

I think I found the source of the boot problem!

I was able to boot succesfully from an older disk having ms-dos partition table!

All my newer disks have gpt, and they do not boot.

Christian, what partition table did you use when you tested the new uInitrd?

[xeno74]

I think I found the source of the boot problem!

I was able to boot succesfully from an older disk having ms-dos partition table!

All my newer disks have gpt, and they do not boot.

Christian, what partition table did you use when you tested the new uInitrd?

MBR and RDB.

OK, we need EFI GUID Partition Table (GPT) support in our kernels.

I will add CONFIG_EFI_PARTITION=y to the next kernels. Thank you for the hint.

[Roland]

OK, we need EFI GUID Partition Table (GPT) support in our kernels.

I will add CONFIG_EFI_PARTITION=y to the next kernels. Thank you for the hint.

Thank you!-) But why did it work until 6.10 alpha3? Was the support just turned off accidentally in RC1?

AFAIK, there is no reason to use MBR anymore, unless you have some compatibility issues on dual-boot systems (windows etc.).

[xeno74]

I have released the stable long-term kernel 6.12.6 with EFI GUID Partition Table (GPT) support.

Download and further information: github.com

Please test it.

[Roland]

I have released the stable long-term kernel 6.12.6 with EFI GUID Partition Table (GPT) support.

Thank you! Tested it with uInitrd 6.12 - booting works now ok .

[xeno74]

Thank you! Tested it with uInitrd 6.12 - booting works now ok .

Great, that it works.

CONFIG_EFI_PARTITION is also activated in the X5000 kernel config of the stable long-term kernel 5.10.

[xeno74]

KVM: PPC: Book3E: KVM HV host module doesn't work anymore since the alpha3 of kernel 6.13. (alpha3: PowerPC updates 6.13-1 and PowerPC MM updates)

The kernel doesn't boot. There is no output via the serial connection and no output via the graphics card.

I reported this issue to the kernel developers today.

Link: KVM: PPC: Book3E: KVM HV host module doesn't work anymore

I bisected today because of this issue.
...

• git bisect bad
  Output:

  Code: Select all

  Bisecting: 51 revisions left to test after this (roughly 6 steps)
  [c9be85dabb376299504e0d391d15662c0edf8273] KVM: PPC: e500: Mark "struct page" dirty in kvmppc_e500_shadow_map()

• git bisect bad (It seems bad because the host system doesn't boot - Maybe another issue)
  Output:

  Code: Select all

  Bisecting: 25 revisions left to test after this (roughly 5 steps)
  [12fac89950996e556a99eb16f0359307ed4c2566] KVM: Use plain "struct page" pointer instead of single-entry array

• git bisect bad (It seems bad because the host system doesn't boot - Maybe another issue)
  Output:

  Code: Select all

  Bisecting: 12 revisions left to test after this (roughly 4 steps)
  [eec1e5db464eba928273b27246122d157fd0eff8] KVM: Annotate that all paths in hva_to_pfn() might sleep

  I have to stop bisecting because the X5000 doesn't boot with the last 2 test kernels and I don't really know if these kernels are bad or good.

For commits where you're unsure, ie. the host system doesn't boot, use git bisect skip. That will avoid the bisect going off and chasing a different bug. It may mean at the end the result is "the bug is somewhere in this range we skipped", but that's OK.

You'll have to restart the bisect, but you can just replay the existing known good/bad commits without testing them again, eg:

I will skip KVM: PPC: e500: Mark "struct page" dirty in kvmppc_e500_shadow_map().

• git bisect skip
  Output:

  Code: Select all

  Bisecting: 51 revisions left to test after this (roughly 6 steps)
  [8f57adac39167de7416c7eae91b4621454def3cd] RISC-V: KVM: Break down the __kvm_riscv_switch_to() into macros
  

• git bisect good
  Output:

  Code: Select all

  Bisecting: 48 revisions left to test after this (roughly 6 steps)
  [dc06193532af4ba88ed20daeef88f22b053ebb91] KVM: Move x86's API to release a faultin page to common KVM
  

• git bisect skip
  Output:

  Code: Select all

  Bisecting: 48 revisions left to test after this (roughly 6 steps)
  [334511d468e5b61e9b4b1432145e51dbce365888] KVM: RISC-V: Use kvm_faultin_pfn() when mapping pfns into the guest
  

• git bisect skip
  Output:

  Code: Select all

  Bisecting: 48 revisions left to test after this (roughly 6 steps)
  [eec1e5db464eba928273b27246122d157fd0eff8] KVM: Annotate that all paths in hva_to_pfn() might sleep
  

• git bisect skip
  Output:

  Code: Select all

  Bisecting: 48 revisions left to test after this (roughly 6 steps)
  [7103853952957cad58f957142c72bf7a4644b7a9] KVM: x86/mmu: Add helper to "finish" handling a guest page fault
  

• git bisect skip
  Output:

  Code: Select all

  Bisecting: 48 revisions left to test after this (roughly 6 steps)
  [b176f4b41775f8b2a7c642f87ccd5e3f405e5191] KVM: Drop extra GUP (via check_user_page_hwpoison()) to detect poisoned page
  

• git bisect skip
  Output:

  Code: Select all

  Bisecting: 48 revisions left to test after this (roughly 6 steps)
  [0cad68cab1135b60348a7b81814c8d2ff4b959ea] KVM: x86/mmu: Mark pages/folios dirty at the origin of make_spte()
  

• git bisect skip
  Output:

  Code: Select all

  Bisecting: 48 revisions left to test after this (roughly 6 steps)
  [31fccdd21263f12e8b701ad90a78e31e789cb780] KVM: Make kvm_follow_pfn.refcounted_page a required field
  

• git bisect skip
  Output:

  Code: Select all

  Bisecting: 48 revisions left to test after this (roughly 6 steps)
  [28991c91d577c39bbd9f1b5424554890c3aa351b] KVM: arm64: Mark "struct page" pfns accessed/dirty before dropping mmu_lock
  

• git bisect skip
  Output:

  Code: Select all

  Bisecting: 48 revisions left to test after this (roughly 6 steps)
  [cccefb0a0d3b4f7b41b1921538087dd7031876ac] KVM: Drop unused "hva" pointer from __gfn_to_pfn_memslot()
  

• git bisect skip
  Output:

  Code: Select all

  Bisecting: 48 revisions left to test after this (roughly 6 steps)
  [54ba8c98a2589f816c52545c9cd3db6665ee0c67] KVM: x86/mmu: Convert page fault paths to kvm_faultin_pfn()
  

• git bisect skip
  Output:

  Code: Select all

  Bisecting: 48 revisions left to test after this (roughly 6 steps)
  [0b139b877b1462ccdecb5146493c68a0b94ccdc8] KVM: Remove pointless sanity check on @map param to kvm_vcpu_(un)map()
  

• git bisect skip
  Output:

  Code: Select all

  Bisecting: 48 revisions left to test after this (roughly 6 steps)
  [84cf78dcd9d65c45ab73998d4ad50f433d53fb93] KVM: PPC: e500: Mark "struct page" pfn accessed before dropping mmu_lock
  

• git bisect skip
  Output:

  Code: Select all

  Bisecting: 48 revisions left to test after this (roughly 6 steps)
  [f42e289a2095f61755e6ca5fd1370d441bf589d5] KVM: Add support for read-only usage of gfn_to_page()
  

[xeno74]

KVM: PPC: Book3E: KVM HV host module doesn't work anymore since the alpha3 of kernel 6.13. (alpha3: PowerPC updates 6.13-1 and PowerPC MM updates)

The guest kernel doesn't boot. There is no output via the serial connection and no output via the graphics card.

We have found the first bad commit.

Links:

• KVM: PPC: Book3E: KVM HV host module doesn't work anymore #4 -- github.com
• KVM: PPC: Book3E: KVM HV host module doesn't work anymore -- Linuxppc-dev mailing list

[xeno74]

We have found the first bad commit.

Links:

• KVM: PPC: Book3E: KVM HV host module doesn't work anymore #4 -- github.com
• KVM: PPC: Book3E: KVM HV host module doesn't work anymore -- Linuxppc-dev mailing list

Here is the first bad commit: KVM: PPC: e500: Use __kvm_faultin_pfn() to handle page faults

[Roland]

@xeno74

There is another problem with the kernel 6.12... The driver for Soundblaster Live! (EMU10k) seems to be missing, as my card is not 'seen' at all. I did copy the kernel modules dir into lib/modules, but it did not have any effect.

With kernel 5.10 there is no problem with the named sound card. Also 6.9 works (even without the modules dir).