Re: [PATCH 6.12 0000/1102] 6.12.111-rc1 review

[Sasha Levin <sashal@kernel.org>]

On Thu, Sep 17, 2026 at 10:19:33PM +0200, Peter Schneider wrote:
> When I revert 22ae428b0045981bd19f8c75f700a67198062f32 from rc1, the
> boot crash goes away and my machine boots successfully.

Thanks for confirming, and for the crash photo from the Ivy Bridge box in your
earlier mail - that is the same bug. Dropped from the 6.12 queue:

  b1b7c045e808 ("mm/hugetlb: initialize gigantic bootmem hugepage
                 struct pages earlier")

The backport relocates gather_bootmem_prealloc() into page_alloc_init_late(),
which on 6.12 runs before anything has initialized the per-node
huge_boot_pages[] list heads - upstream that is done unconditionally in
hugetlb_bootmem_alloc() since v6.15, which 6.12 does not have. Any
CONFIG_HUGETLB_PAGE=y kernel booted without hugepagesz=/hugepages= walks an
all-zero list head and faults.

> However, then I have another weirdo in the dmesg output:
> [   65.573307] WARNING: CPU: 14 PID: 0 at kernel/sched/fair.c:5303
> place_entity+0x13b/0x140

That one is a separate 6.12 problem and your follow-up bisect nailed it - I've
replied there with what was dropped for it.

-- 
Thanks,
Sasha