Re: [PATCH v2 bpf] bpf: Fix a data-race around bpf_jit_limit.

[Daniel Borkmann <daniel@iogearbox.net>]

On 8/23/22 8:12 PM, Kuniyuki Iwashima wrote:
> While reading bpf_jit_limit, it can be changed concurrently.
> Thus, we need to add READ_ONCE() to its reader.

For sake of a better/clearer commit message, please also provide data about the
WRITE_ONCE() pairing that this READ_ONCE() targets. This seems to be the case in
__do_proc_doulongvec_minmax() as far as I can see. For your 2nd sentence above
please also include load-tearing as main motivation for your fix.

> Fixes: ede95a63b5e8 ("bpf: add bpf_jit_limit knob to restrict unpriv allocations")
> Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
> ---
> v2:
>    * Drop other 3 patches (No change for this patch)
> 
> v1: https://lore.kernel.org/bpf/20220818042339.82992-1-kuniyu@amazon.com/
> ---
>   kernel/bpf/core.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c
> index c1e10d088dbb..3d9eb3ae334c 100644
> --- a/kernel/bpf/core.c
> +++ b/kernel/bpf/core.c
> @@ -971,7 +971,7 @@ pure_initcall(bpf_jit_charge_init);
>   
>   int bpf_jit_charge_modmem(u32 size)
>   {
> -	if (atomic_long_add_return(size, &bpf_jit_current) > bpf_jit_limit) {
> +	if (atomic_long_add_return(size, &bpf_jit_current) > READ_ONCE(bpf_jit_limit)) {
>   		if (!bpf_capable()) {
>   			atomic_long_sub(size, &bpf_jit_current);
>   			return -EPERM;
>