On Wed, Feb 26, 2025 at 03:31:03PM +0800, Huacai Chen wrote:
On Wed, Feb 26, 2025 at 4:41 AM Kees Cook kees@kernel.org wrote:
On Tue, Feb 25, 2025 at 07:35:13PM +0800, Huacai Chen wrote:
I have investigated deeper, and then found it is an arch-specific problem (at least for LoongArch), and the correct solution is here: https://lore.kernel.org/loongarch/20250225111812.3065545-1-chenhuacai@loongs...
Ah-ha, so it seems like some system start was being incorrectly shared between restoration image and hibernated image? Yeah, that's important to fix.
But I don't know how to fix arm64.
Is arm64 broken in this same way?
ARM64 is broken but I don't know whether it is in the same way, I just know this patch can solve ARM64's problem: https://lore.kernel.org/linux-mm/CAAhV-H7i=WJmdFCCtY5DgE2eN657ddJwJwHGK1jgLK...
Let's Cc ARM folks :)
A bit of context: LoongArch and ARM64 crash when resuming from hibernation if CONFIG_RANDOM_KMALLOC_CACHES=y. Initially it was argued that kmalloc randomization should be disabled during early boot. Kees and Rafael thinks [1] randomization features should not be the root cause of the bug. Later, Huacai fixed [2] the issue on LoongArch side by addressing LoongArch-specific problem.
A similar crash related to the kmalloc randomization feature was reported [3] on a ARM64-based laptop. ARM64 might be broken in a similar way, but we don't know for sure yet.
[1] https://lore.kernel.org/linux-mm/CAJZ5v0hZZdRPwp=OgPw4w8r9X=VbL6Hn6R4ZX6ZujN... [2] https://lore.kernel.org/loongarch/20250225111812.3065545-1-chenhuacai@loongs... [3] https://lore.kernel.org/linux-mm/CAAhV-H7i=WJmdFCCtY5DgE2eN657ddJwJwHGK1jgLK...
Huacai
-- Kees Cook