linux-kernel - Re: [PATCH v2] riscv: VMAP_STACK overflow detection thread-safe

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20221125213503.GB2037421@debug.ba.rivosinc.com>
Date:   Fri, 25 Nov 2022 13:35:03 -0800
From:   Deepak Gupta <debug@...osinc.com>
To:     Jisheng Zhang <jszhang@...nel.org>
Cc:     palmer@...belt.com, linux-kernel@...r.kernel.org,
        linux-riscv@...ts.infradead.org, Guo Ren <guoren@...nel.org>
Subject: Re: [PATCH v2] riscv: VMAP_STACK overflow detection thread-safe

On Thu, Nov 24, 2022 at 11:26:42PM +0800, Jisheng Zhang wrote:
>On Thu, Nov 24, 2022 at 01:48:45AM -0800, Deepak Gupta wrote:
>> commit 31da94c25aea ("riscv: add VMAP_STACK overflow detection") added
>> support for CONFIG_VMAP_STACK. If overflow is detected, CPU switches to
>> `shadow_stack` temporarily before switching finally to per-cpu
>> `overflow_stack`.
>>
>> If two CPUs/harts are racing and end up in over flowing kernel stack, one
>> or both will end up corrupting each other state because `shadow_stack` is
>> not per-cpu. This patch optimizes per-cpu overflow stack switch by
>> directly picking per-cpu `overflow_stack` and gets rid of `shadow_stack`.
>>
>> Following are the changes in this patch
>>
>>  - Defines an asm macro to obtain per-cpu symbols in destination
>>    register.
>>  - In entry.S, when overflow is detected, per-cpu overflow stack is
>>    located using per-cpu asm macro. Computing per-cpu symbol requires
>>    a temporary register. x31 is saved away into CSR_SCRATCH
>
>This only works if CSR_SCRATCH doesn't contain any valid reg saving,
>but.. see below.
>
>>    (CSR_SCRATCH is anyways zero since we're in kernel).
>>
>
>To be honest, before [1] I have similar idea to keep the percpu usage,
>however, the solution doesn't work. The key here is that there's
>another VMAP_STACK bug in current riscv implementation: it only checks
>vmap stack overflow when comming from kernelspace, but vmap should
>check when comming from both kernelspace and userspace. So we can't

Why do we need to check if space is available or not when coming from user space.
Kernel stack is fresh and just starting out it's life.

>assume CSR_SCRATCH is always zero and free to use. The only available
>solution is my fix[1] which only makes use of tp. But since[1] modifies
>lots of code, it's not idea to merge it as a fix, so [2] is suggested
>and sent out.
>
>PS: I planed to send a fix for the missing FROM_USERSPACE after the
>race fix is merged.
>
>
>[1]https://lore.kernel.org/linux-riscv/20220925175356.681-1-jszhang@kernel.org/T/#t
>[2]https://lore.kernel.org/linux-riscv/Y347B0x4VUNOd6V7@xhacker/T/#t
>