Re: [PATCH 6.1.y 3/6] bpf: Restrict JIT predictor flush to cBPF

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] Spectre v2 Mitigation Bypass for Unprivileged Native eBPF
--

--- Patch [3]: [PATCH 6.1.y 3/6] bpf: Restrict JIT predictor flush to cBPF ---
commit 77faa1df216dd9c68cbac9a596c9d2f4b8b6ffef
Author: Pawan Gupta <pawan.kumar.gupta@linux.intel.com>

bpf: Restrict JIT predictor flush to cBPF

This patch restricts the predictor flush on memory reuse to cBPF programs
only, aiming to reduce performance overhead on busy systems. It skips the
flush for native eBPF allocations under the assumption that eBPF is
privileged.

> diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c
> index 4643554c6e3a9..f78ee18bab3b5 100644
> --- a/kernel/bpf/core.c
> +++ b/kernel/bpf/core.c
[ ... ]
> @@ -947,7 +947,9 @@ void *bpf_prog_pack_alloc(u32 size, bpf_jit_fill_hole_t bpf_fill_ill_insns)
>  	pos = 0;
>  
>  found_free_area:
> -	static_call_cond(bpf_arch_pred_flush)();
> +	/* Flush only for cBPF as it may contain a crafted gadget */
> +	if (static_branch_unlikely(&bpf_pred_flush_enabled) && was_classic)
> +		static_call_cond(bpf_arch_pred_flush)();

[Severity: High]
Does skipping the predictor flush here for native eBPF bypass the Spectre v2
JIT mitigation for unprivileged users?

The commit message notes that eBPF is privileged by default, but unprivileged
local users can still load native eBPF programs (such as
BPF_PROG_TYPE_SOCKET_FILTER) on systems where kernel.unprivileged_bpf_disabled
is set to 0.

Since native eBPF sets was_classic to false, wouldn't bpf_prog_pack_alloc()
now skip the flush and leave unprivileged native eBPF programs unprotected
against branch target poisoning?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260727-cbpf-jit-spray-hardening-6-1-y-v1-0-eb80dcf1eb6e@linux.intel.com?part=3