Re: [PATCH bpf-next 1/2] bpf: allow wide (u64) aligned stores for some fields of bpf_sock_addr

[Andrii Nakryiko <andrii.nakryiko@gmail.com>]

On Sat, Jun 29, 2019 at 10:53 PM Yonghong Song <yhs@fb.com> wrote:
>
>
>
> On 6/28/19 4:10 PM, Stanislav Fomichev wrote:
> > Since commit cd17d7770578 ("bpf/tools: sync bpf.h") clang decided
> > that it can do a single u64 store into user_ip6[2] instead of two
> > separate u32 ones:
> >
> >   #  17: (18) r2 = 0x100000000000000
> >   #  ; ctx->user_ip6[2] = bpf_htonl(DST_REWRITE_IP6_2);
> >   #  19: (7b) *(u64 *)(r1 +16) = r2
> >   #  invalid bpf_context access off=16 size=8
> >
> >  From the compiler point of view it does look like a correct thing
> > to do, so let's support it on the kernel side.
> >
> > Credit to Andrii Nakryiko for a proper implementation of
> > bpf_ctx_wide_store_ok.
> >
> > Cc: Andrii Nakryiko <andriin@fb.com>
> > Cc: Yonghong Song <yhs@fb.com>
> > Fixes: cd17d7770578 ("bpf/tools: sync bpf.h")
> > Reported-by: kernel test robot <rong.a.chen@intel.com>
> > Signed-off-by: Stanislav Fomichev <sdf@google.com>
>
> The change looks good to me with the following nits:
>    1. could you add a cover letter for the patch set?
>       typically if the number of patches is more than one,
>       it would be a good practice with a cover letter.
>       See bpf_devel_QA.rst .
>    2. with this change, the comments in uapi bpf.h
>       are not accurate any more.
>          __u32 user_ip6[4];      /* Allows 1,2,4-byte read an 4-byte write.
>                                   * Stored in network byte order.
>
>                                   */
>          __u32 msg_src_ip6[4];   /* Allows 1,2,4-byte read an 4-byte write.
>                                   * Stored in network byte order.
>                                   */
>       now for stores, aligned 8-byte write is permitted.
>       could you update this as well?
>
>  From the typical usage pattern, I did not see a need
> for 8-tye read of user_ip6 and msg_src_ip6 yet. So let
> us just deal with write for now.

But I guess it's still possible for clang to optimize two consecutive
4-byte reads into single 8-byte read in some circumstances? If that's
the case, maybe it's a good idea to have corresponding read checks as
well?

But overall this looks good to me:

Acked-by: Andrii Nakryiko <andriin@fb.com>

>
> With the above two nits,
> Acked-by: Yonghong Song <yhs@fb.com>
>
> > ---
> >   include/linux/filter.h |  6 ++++++
> >   net/core/filter.c      | 22 ++++++++++++++--------
> >   2 files changed, 20 insertions(+), 8 deletions(-)
> >
> > diff --git a/include/linux/filter.h b/include/linux/filter.h
> > index 340f7d648974..3901007e36f1 100644
> > --- a/include/linux/filter.h
> > +++ b/include/linux/filter.h
> > @@ -746,6 +746,12 @@ bpf_ctx_narrow_access_ok(u32 off, u32 size, u32 size_default)
> >       return size <= size_default && (size & (size - 1)) == 0;
> >   }
> >
> > +#define bpf_ctx_wide_store_ok(off, size, type, field)                        \
> > +     (size == sizeof(__u64) &&                                       \
> > +     off >= offsetof(type, field) &&                                 \
> > +     off + sizeof(__u64) <= offsetofend(type, field) &&              \
> > +     off % sizeof(__u64) == 0)
> > +
> >   #define bpf_classic_proglen(fprog) (fprog->len * sizeof(fprog->filter[0]))
> >
> >   static inline void bpf_prog_lock_ro(struct bpf_prog *fp)
> > diff --git a/net/core/filter.c b/net/core/filter.c
> > index dc8534be12fc..5d33f2146dab 100644
> > --- a/net/core/filter.c
> > +++ b/net/core/filter.c
> > @@ -6849,6 +6849,16 @@ static bool sock_addr_is_valid_access(int off, int size,
> >                       if (!bpf_ctx_narrow_access_ok(off, size, size_default))
> >                               return false;
> >               } else {
> > +                     if (bpf_ctx_wide_store_ok(off, size,
> > +                                               struct bpf_sock_addr,
> > +                                               user_ip6))
> > +                             return true;
> > +
> > +                     if (bpf_ctx_wide_store_ok(off, size,
> > +                                               struct bpf_sock_addr,
> > +                                               msg_src_ip6))
> > +                             return true;
> > +
> >                       if (size != size_default)
> >                               return false;
> >               }
> > @@ -7689,9 +7699,6 @@ static u32 xdp_convert_ctx_access(enum bpf_access_type type,
> >   /* SOCK_ADDR_STORE_NESTED_FIELD_OFF() has semantic similar to
> >    * SOCK_ADDR_LOAD_NESTED_FIELD_SIZE_OFF() but for store operation.
> >    *
> > - * It doesn't support SIZE argument though since narrow stores are not
> > - * supported for now.
> > - *
> >    * In addition it uses Temporary Field TF (member of struct S) as the 3rd
> >    * "register" since two registers available in convert_ctx_access are not
> >    * enough: we can't override neither SRC, since it contains value to store, nor
> > @@ -7699,7 +7706,7 @@ static u32 xdp_convert_ctx_access(enum bpf_access_type type,
> >    * instructions. But we need a temporary place to save pointer to nested
> >    * structure whose field we want to store to.
> >    */
> > -#define SOCK_ADDR_STORE_NESTED_FIELD_OFF(S, NS, F, NF, OFF, TF)                     \
> > +#define SOCK_ADDR_STORE_NESTED_FIELD_OFF(S, NS, F, NF, SIZE, OFF, TF)               \
> >       do {                                                                   \
> >               int tmp_reg = BPF_REG_9;                                       \
> >               if (si->src_reg == tmp_reg || si->dst_reg == tmp_reg)          \
> > @@ -7710,8 +7717,7 @@ static u32 xdp_convert_ctx_access(enum bpf_access_type type,
> >                                     offsetof(S, TF));                        \
> >               *insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF(S, F), tmp_reg,         \
> >                                     si->dst_reg, offsetof(S, F));            \
> > -             *insn++ = BPF_STX_MEM(                                         \
> > -                     BPF_FIELD_SIZEOF(NS, NF), tmp_reg, si->src_reg,        \
> > +             *insn++ = BPF_STX_MEM(SIZE, tmp_reg, si->src_reg,              \
> >                       bpf_target_off(NS, NF, FIELD_SIZEOF(NS, NF),           \
> >                                      target_size)                            \
> >                               + OFF);                                        \
> > @@ -7723,8 +7729,8 @@ static u32 xdp_convert_ctx_access(enum bpf_access_type type,
> >                                                     TF)                      \
> >       do {                                                                   \
> >               if (type == BPF_WRITE) {                                       \
> > -                     SOCK_ADDR_STORE_NESTED_FIELD_OFF(S, NS, F, NF, OFF,    \
> > -                                                      TF);                  \
> > +                     SOCK_ADDR_STORE_NESTED_FIELD_OFF(S, NS, F, NF, SIZE,   \
> > +                                                      OFF, TF);             \
> >               } else {                                                       \
> >                       SOCK_ADDR_LOAD_NESTED_FIELD_SIZE_OFF(                  \
> >                               S, NS, F, NF, SIZE, OFF);  \
> >