On Thu, Sep 4, 2025 at 9:56 AM Hengqi Chen <hengqi.chen@xxxxxxxxx> wrote: > > On Wed, Sep 3, 2025 at 10:56 PM Huacai Chen <chenhuacai@xxxxxxxxxx> wrote: > > > > Hi, Hengqi, > > > > On Wed, Sep 3, 2025 at 8:06 PM Hengqi Chen <hengqi.chen@xxxxxxxxx> wrote: > > > > > > Currently, arch_alloc_bpf_trampoline() use bpf_prog_pack_alloc() > > > which will pack multiple trampolines into a huge page. So no need > > > to assume the trampoline size is page aligned. > > We do the alignment because larch_insn_text_copy() changes page attrs. > > If there is other data and BPF trampoline is in the same page, > > changing page attrs may cause errors. > > > > Doesn't stop_machine() prevent this side effect? I don't think so. Consider such a case: BPF trampoline and another vmalloc area share the same page, larch_insn_text_copy() makes the whole page be "rox", even if stop_machine() completes. But "rox" is only needed by BPF trampoline, and the other vmalloc area isn't. Huacai > > > Huacai > > > > > > > > Signed-off-by: Hengqi Chen <hengqi.chen@xxxxxxxxx> > > > --- > > > arch/loongarch/net/bpf_jit.c | 3 +-- > > > 1 file changed, 1 insertion(+), 2 deletions(-) > > > > > > diff --git a/arch/loongarch/net/bpf_jit.c b/arch/loongarch/net/bpf_jit.c > > > index 35b13d91a979..43628b5e1553 100644 > > > --- a/arch/loongarch/net/bpf_jit.c > > > +++ b/arch/loongarch/net/bpf_jit.c > > > @@ -1747,8 +1747,7 @@ int arch_bpf_trampoline_size(const struct btf_func_model *m, u32 flags, > > > > > > ret = __arch_prepare_bpf_trampoline(&ctx, &im, m, tlinks, func_addr, flags); > > > > > > - /* Page align */ > > > - return ret < 0 ? ret : round_up(ret * LOONGARCH_INSN_SIZE, PAGE_SIZE); > > > + return ret < 0 ? ret : ret * LOONGARCH_INSN_SIZE; > > > } > > > > > > struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *prog) > > > -- > > > 2.43.5 > > > >
