From: Luis Gerhorst <luis.gerhorst@fau.de>
To: Ryan Roberts <ryan.roberts@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>,
Will Deacon <will@kernel.org>,
Mark Rutland <mark.rutland@arm.com>,
Alexei Starovoitov <ast@kernel.org>,
Daniel Borkmann <daniel@iogearbox.net>,
John Fastabend <john.fastabend@gmail.com>,
Andrew Morton <akpm@linux-foundation.org>,
Alexander Viro <viro@zeniv.linux.org.uk>,
linux-arm-kernel@lists.infradead.org,
linux-kernel@vger.kernel.org, bpf@vger.kernel.org,
stable@vger.kernel.org
Subject: Re: [PATCH v1] arm64: Mitigate Spectre v1 in the BPF interpreter
Date: Wed, 18 Nov 2026 15:27:25 +0100 [thread overview]
Message-ID: <871p8ib4fm.fsf@fau.de> (raw)
In-Reply-To: <20261117172341.3152407-1-ryan.roberts@arm.com>
Ryan Roberts <ryan.roberts@arm.com> writes:
> The BPF verifier can accept programs with unsafe speculative paths by
> inserting BPF_NOSPEC instructions instead of rejecting them. It relies
> on those instructions to prevent execution of the unsafe paths.
>
> The arm64 JIT implements BPF_NOSPEC with a speculation barrier, but the
> interpreter calls barrier_nospec(), which falls back to a no-op on
> arm64. Programs executed by the interpreter therefore lack the
> Spectre-v1 barriers assumed by the verifier. Contrary to the code
> comment, kernel-wide SSBD does not provide this protection, since it
> mitigates Spectre v4 rather than Spectre v1.
>
> Rename spec_bar() to barrier_nospec() so that the interpreter and other
> callers receive SB if supported, or DSB NSH followed by ISB otherwise.
>
> Defining barrier_nospec() would also add barriers to the generic
> copy_from_user() and copy_from_user_iter() paths. These are unnecessary
> on arm64: its uaccess primitives mask user pointers so that they cannot
> address kernel memory even when an access_ok() check is mispredicted.
>
> Introduce user_access_may_speculate_kernel_addr(), defaulting to true,
> and have arm64 return false. Then update the uaccess sites to call
> barrier_nospec() conditionally on the macro.
>
> Cc: stable@vger.kernel.org
> Fixes: d6f1c85f2253 ("bpf: Fall back to nospec for Spectre v1")
> Signed-off-by: Ryan Roberts <ryan.roberts@arm.com>
Thanks! Tested that this prevents the attack (just recovery of OOB
value, did not aim for real secret) on Cavium ThunderX2 (no FEAT_SB) as
expected.
Tested-by: Luis Gerhorst <luis.gerhorst@fau.de>
> ---
> arch/arm64/include/asm/barrier.h | 6 +++---
> arch/arm64/include/asm/uaccess.h | 3 +++
> arch/arm64/kernel/proton-pack.c | 2 +-
> include/linux/uaccess.h | 7 ++++++-
> kernel/bpf/core.c | 9 +--------
> lib/iov_iter.c | 3 ++-
> 6 files changed, 16 insertions(+), 14 deletions(-)
>
> diff --git a/arch/arm64/include/asm/barrier.h b/arch/arm64/include/asm/barrier.h
> index 9495c4441a46..088ee29d36f4 100644
> --- a/arch/arm64/include/asm/barrier.h
> +++ b/arch/arm64/include/asm/barrier.h
> @@ -40,9 +40,9 @@
> */
> #define dgh() asm volatile("hint #6" : : : "memory")
>
> -#define spec_bar() asm volatile(ALTERNATIVE("dsb nsh\nisb\n", \
> - SB_BARRIER_INSN"nop\n", \
> - ARM64_HAS_SB))
> +#define barrier_nospec() asm volatile(ALTERNATIVE("dsb nsh\nisb\n", \
> + SB_BARRIER_INSN"nop\n", \
> + ARM64_HAS_SB) : : : "memory")
>
> #define gsb_ack() asm volatile(GSB_ACK_BARRIER_INSN : : : "memory")
> #define gsb_sys() asm volatile(GSB_SYS_BARRIER_INSN : : : "memory")
> diff --git a/arch/arm64/include/asm/uaccess.h b/arch/arm64/include/asm/uaccess.h
> index 9f5bd9c69c24..6fd03dc1ae76 100644
> --- a/arch/arm64/include/asm/uaccess.h
> +++ b/arch/arm64/include/asm/uaccess.h
> @@ -174,6 +174,9 @@ static inline void __user *__uaccess_mask_ptr(const void __user *ptr)
> return safe_ptr;
> }
>
> +/* __uaccess_mask_ptr() prevents speculative accesses to kernel addresses. */
> +#define user_access_may_speculate_kernel_addr() false
> +
> /*
> * The "__xxx" versions of the user access functions do not verify the address
> * space - it must have been done previously with a separate "access_ok()"
> diff --git a/arch/arm64/kernel/proton-pack.c b/arch/arm64/kernel/proton-pack.c
> index 3bcf86154d95..c96e93eca3b9 100644
> --- a/arch/arm64/kernel/proton-pack.c
> +++ b/arch/arm64/kernel/proton-pack.c
> @@ -559,7 +559,7 @@ static enum mitigation_state spectre_v4_enable_hw_mitigation(void)
> * could mis-speculate branches and bypass a conditional barrier.
> */
> if (IS_ENABLED(CONFIG_ARM64_ERRATUM_3194386))
> - spec_bar();
> + barrier_nospec();
>
> return SPECTRE_MITIGATED;
> }
> diff --git a/include/linux/uaccess.h b/include/linux/uaccess.h
> index 67cdb0a4f583..5a1fa4f48c77 100644
> --- a/include/linux/uaccess.h
> +++ b/include/linux/uaccess.h
> @@ -50,6 +50,10 @@
> #define mask_user_address(src) (src)
> #endif
>
> +#ifndef user_access_may_speculate_kernel_addr
> +#define user_access_may_speculate_kernel_addr() true
> +#endif
> +
> /*
> * Architectures should provide two primitives (raw_copy_{to,from}_user())
> * and get rid of their private instances of copy_{to,from}_user() and
> @@ -179,7 +183,8 @@ _inline_copy_from_user(void *to, const void __user *from, unsigned long n)
> * lead to nasty side effects *after* the copy is
> * finished:
> */
> - barrier_nospec();
> + if (user_access_may_speculate_kernel_addr())
> + barrier_nospec();
> }
> instrument_copy_from_user_before(to, from, n);
> res = raw_copy_from_user(to, from, n);
> diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c
> index 0b8456f2dce9..abd938d310bb 100644
> --- a/kernel/bpf/core.c
> +++ b/kernel/bpf/core.c
> @@ -2250,14 +2250,7 @@ static u64 ___bpf_prog_run(u64 *regs, const struct bpf_insn *insn)
> /* ST, STX and LDX*/
> ST_NOSPEC:
> /* Speculation barrier for mitigating Speculative Store Bypass,
> - * Bounds-Check Bypass and Type Confusion. In case of arm64, we
> - * rely on the firmware mitigation as controlled via the ssbd
> - * kernel parameter. Whenever the mitigation is enabled, it
> - * works for all of the kernel code with no need to provide any
> - * additional instructions here. In case of x86, we use 'lfence'
> - * insn for mitigation. We reuse preexisting logic from Spectre
> - * v1 mitigation that happens to produce the required code on
> - * x86 for v4 as well.
> + * Bounds-Check Bypass and Type Confusion.
> */
> barrier_nospec();
> CONT;
> diff --git a/lib/iov_iter.c b/lib/iov_iter.c
> index fa4fe2e92383..abd27189beff 100644
> --- a/lib/iov_iter.c
> +++ b/lib/iov_iter.c
> @@ -60,7 +60,8 @@ size_t copy_from_user_iter(void __user *iter_from, size_t progress,
> * lead to nasty side effects *after* the copy is
> * finished:
> */
> - barrier_nospec();
> + if (user_access_may_speculate_kernel_addr())
> + barrier_nospec();
> }
> to += progress;
> instrument_copy_from_user_before(to, iter_from, len);
prev parent reply other threads:[~2026-11-18 14:28 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-17 17:23 [PATCH v1] arm64: Mitigate Spectre v1 in the BPF interpreter Ryan Roberts
2026-11-18 14:27 ` Luis Gerhorst [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=871p8ib4fm.fsf@fau.de \
--to=luis.gerhorst@fau.de \
--cc=akpm@linux-foundation.org \
--cc=ast@kernel.org \
--cc=bpf@vger.kernel.org \
--cc=catalin.marinas@arm.com \
--cc=daniel@iogearbox.net \
--cc=john.fastabend@gmail.com \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mark.rutland@arm.com \
--cc=ryan.roberts@arm.com \
--cc=stable@vger.kernel.org \
--cc=viro@zeniv.linux.org.uk \
--cc=will@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox