powerpc/32s: Fix boot failure with KASAN + SMP + JUMP_LABEL_FEATURE_CHECK_DEBUG
authorChristophe Leroy <christophe.leroy@csgroup.eu>
Fri, 1 Jul 2022 06:06:15 +0000 (08:06 +0200)
committerMichael Ellerman <mpe@ellerman.id.au>
Wed, 27 Jul 2022 11:36:05 +0000 (21:36 +1000)
commit6042a1652d643d1d34fa89bb314cb102960c0800
tree8c839cea25acf577091bc5c8adca53255c428ba8
parent2a0fb3c155c97c75176e557d61f8e66c1bd9b735
powerpc/32s: Fix boot failure with KASAN + SMP + JUMP_LABEL_FEATURE_CHECK_DEBUG

Since commit 4291d085b0b0 ("powerpc/32s: Make pte_update() non
atomic on 603 core"), pte_update() has been using
mmu_has_feature(MMU_FTR_HPTE_TABLE) to avoid a useless atomic
operation on 603 cores.

When kasan_early_init() sets up the early zero shadow, it uses
__set_pte_at(). On book3s/32, __set_pte_at() calls pte_update()
when CONFIG_SMP is selected in order to ensure the preservation of
_PAGE_HASHPTE in case of concurrent update of the PTE. But that's
too early for mmu_has_feature(), so when
CONFIG_JUMP_LABEL_FEATURE_CHECK_DEBUG is selected, mmu_has_feature()
calls printk(). That's too early to call printk() because KASAN
early zero shadow page is not set up yet. It leads to a deadlock.

However, when kasan_early_init() is called, there is only one CPU
running and no risk of concurrent PTE update. So __set_pte_at() can
be called with the 'percpu' flag. With that flag set, the PTE is
written directly instead of being written via pte_update().

Fixes: 4291d085b0b0 ("powerpc/32s: Make pte_update() non atomic on 603 core")
Reported-by: Erhard Furtner <erhard_f@mailbox.org>
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/2ee707512b8b212b079b877f4ceb525a1606a3fb.1656655567.git.christophe.leroy@csgroup.eu
arch/powerpc/mm/kasan/init_32.c