On Thu, May 26, 2022 at 05:45:25PM +0800, Xu Kuohai wrote: > On 5/25/2022 9:45 PM, Mark Rutland wrote: > > On Wed, May 18, 2022 at 09:16:35AM -0400, Xu Kuohai wrote: > >> BPF_TRAM_F_XXX flags are not used by user code and are almost constant > >> at compile time, so run time validation is a bit overkill. Remove > >> is_valid_bpf_tramp_flags() and add some usage comments. > >> > >> Signed-off-by: Xu Kuohai <xukuohai@xxxxxxxxxx> > >> Acked-by: Song Liu <songliubraving@xxxxxx> > > > > Am I right in thinking this is independent of the arm64-specific bits, and > > could be taken on its own now? > > > > Currenly is_valid_bpf_tramp_flags() is defined in x86 and called before > bpf trampoline is constructed. The check logic is irrelevant to the > architecture code. So we also need to call this function on arm64. But > as Alexei pointed out, the check is not requried, so it's better to > remove it before adding bpf trampoline to arm64. Cool. So this patch could be merged now, even if the rest of the series needs more work? Thanks, Mark. > >> --- > >> arch/x86/net/bpf_jit_comp.c | 20 -------------------- > >> kernel/bpf/bpf_struct_ops.c | 3 +++ > >> kernel/bpf/trampoline.c | 3 +++ > >> 3 files changed, 6 insertions(+), 20 deletions(-) > >> > >> diff --git a/arch/x86/net/bpf_jit_comp.c b/arch/x86/net/bpf_jit_comp.c > >> index a2b6d197c226..7698ef3b4821 100644 > >> --- a/arch/x86/net/bpf_jit_comp.c > >> +++ b/arch/x86/net/bpf_jit_comp.c > >> @@ -1922,23 +1922,6 @@ static int invoke_bpf_mod_ret(const struct btf_func_model *m, u8 **pprog, > >> return 0; > >> } > >> > >> -static bool is_valid_bpf_tramp_flags(unsigned int flags) > >> -{ > >> - if ((flags & BPF_TRAMP_F_RESTORE_REGS) && > >> - (flags & BPF_TRAMP_F_SKIP_FRAME)) > >> - return false; > >> - > >> - /* > >> - * BPF_TRAMP_F_RET_FENTRY_RET is only used by bpf_struct_ops, > >> - * and it must be used alone. > >> - */ > >> - if ((flags & BPF_TRAMP_F_RET_FENTRY_RET) && > >> - (flags & ~BPF_TRAMP_F_RET_FENTRY_RET)) > >> - return false; > >> - > >> - return true; > >> -} > >> - > >> /* Example: > >> * __be16 eth_type_trans(struct sk_buff *skb, struct net_device *dev); > >> * its 'struct btf_func_model' will be nr_args=2 > >> @@ -2017,9 +2000,6 @@ int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *image, void *i > >> if (nr_args > 6) > >> return -ENOTSUPP; > >> > >> - if (!is_valid_bpf_tramp_flags(flags)) > >> - return -EINVAL; > >> - > >> /* Generated trampoline stack layout: > >> * > >> * RBP + 8 [ return address ] > >> diff --git a/kernel/bpf/bpf_struct_ops.c b/kernel/bpf/bpf_struct_ops.c > >> index d9a3c9207240..0572cc5aeb28 100644 > >> --- a/kernel/bpf/bpf_struct_ops.c > >> +++ b/kernel/bpf/bpf_struct_ops.c > >> @@ -341,6 +341,9 @@ int bpf_struct_ops_prepare_trampoline(struct bpf_tramp_links *tlinks, > >> > >> tlinks[BPF_TRAMP_FENTRY].links[0] = link; > >> tlinks[BPF_TRAMP_FENTRY].nr_links = 1; > >> + /* BPF_TRAMP_F_RET_FENTRY_RET is only used by bpf_struct_ops, > >> + * and it must be used alone. > >> + */ > >> flags = model->ret_size > 0 ? BPF_TRAMP_F_RET_FENTRY_RET : 0; > >> return arch_prepare_bpf_trampoline(NULL, image, image_end, > >> model, flags, tlinks, NULL); > >> diff --git a/kernel/bpf/trampoline.c b/kernel/bpf/trampoline.c > >> index 93c7675f0c9e..bd3f2e673874 100644 > >> --- a/kernel/bpf/trampoline.c > >> +++ b/kernel/bpf/trampoline.c > >> @@ -358,6 +358,9 @@ static int bpf_trampoline_update(struct bpf_trampoline *tr) > >> > >> if (tlinks[BPF_TRAMP_FEXIT].nr_links || > >> tlinks[BPF_TRAMP_MODIFY_RETURN].nr_links) > >> + /* NOTE: BPF_TRAMP_F_RESTORE_REGS and BPF_TRAMP_F_SKIP_FRAME > >> + * should not be set together. > >> + */ > >> flags = BPF_TRAMP_F_CALL_ORIG | BPF_TRAMP_F_SKIP_FRAME; > >> > >> if (ip_arg) > >> -- > >> 2.30.2 > >> > > . >