Re: [PATCH bpf-next v2 08/10] tools/libbpf: Add support for BPF_PROG_TYPE_LSM

[KP Singh <kpsingh@chromium.org>]

On 16-Jan 13:49, KP Singh wrote:
> Thanks for the review Andrii!
> 
> I will incorporate the fixes in the next revision.
> 
> On 15-Jan 13:19, Andrii Nakryiko wrote:
> > On Wed, Jan 15, 2020 at 9:13 AM KP Singh <kpsingh@chromium.org> wrote:
> > >
> > > From: KP Singh <kpsingh@google.com>
> > >
> > > * Add functionality in libbpf to attach eBPF program to LSM hooks
> > > * Lookup the index of the LSM hook in security_hook_heads and pass it in
> > >   attr->lsm_hook_index
> > >
> > > Signed-off-by: KP Singh <kpsingh@google.com>
> > > ---
> > >  tools/lib/bpf/bpf.c      |   6 +-
> > >  tools/lib/bpf/bpf.h      |   1 +
> > >  tools/lib/bpf/libbpf.c   | 143 ++++++++++++++++++++++++++++++++++-----
> > >  tools/lib/bpf/libbpf.h   |   4 ++
> > >  tools/lib/bpf/libbpf.map |   3 +
> > >  5 files changed, 138 insertions(+), 19 deletions(-)
> > >
> > > diff --git a/tools/lib/bpf/bpf.c b/tools/lib/bpf/bpf.c
> > > index 500afe478e94..b138d98ff862 100644
> > > --- a/tools/lib/bpf/bpf.c
> > > +++ b/tools/lib/bpf/bpf.c
> > > @@ -235,7 +235,10 @@ int bpf_load_program_xattr(const struct bpf_load_program_attr *load_attr,
> > >         memset(&attr, 0, sizeof(attr));
> > >         attr.prog_type = load_attr->prog_type;
> > >         attr.expected_attach_type = load_attr->expected_attach_type;
> > > -       if (attr.prog_type == BPF_PROG_TYPE_STRUCT_OPS) {
> > > +
> > > +       if (attr.prog_type == BPF_PROG_TYPE_LSM) {
> > > +               attr.lsm_hook_index = load_attr->lsm_hook_index;
> > > +       } else if (attr.prog_type == BPF_PROG_TYPE_STRUCT_OPS) {
> > >                 attr.attach_btf_id = load_attr->attach_btf_id;
> > >         } else if (attr.prog_type == BPF_PROG_TYPE_TRACING) {
> > >                 attr.attach_btf_id = load_attr->attach_btf_id;
> > > @@ -244,6 +247,7 @@ int bpf_load_program_xattr(const struct bpf_load_program_attr *load_attr,
> > >                 attr.prog_ifindex = load_attr->prog_ifindex;
> > >                 attr.kern_version = load_attr->kern_version;
> > >         }
> > > +
> > >         attr.insn_cnt = (__u32)load_attr->insns_cnt;
> > >         attr.insns = ptr_to_u64(load_attr->insns);
> > >         attr.license = ptr_to_u64(load_attr->license);
> > > diff --git a/tools/lib/bpf/bpf.h b/tools/lib/bpf/bpf.h
> > > index 56341d117e5b..54458a102939 100644
> > > --- a/tools/lib/bpf/bpf.h
> > > +++ b/tools/lib/bpf/bpf.h
> > > @@ -86,6 +86,7 @@ struct bpf_load_program_attr {
> > >                 __u32 prog_ifindex;
> > >                 __u32 attach_btf_id;
> > >         };
> > > +       __u32 lsm_hook_index;
> > 
> > 
> > this is changing memory layout of struct bpf_load_program_attr, which
> > is part of public API, so breaking backward compatibility. But I think
> > you intended to put it inside union along the attach_btf_id?
> > 
> 
> Correct, I moved it to the union.
> 
> > also, we use idx for index pretty consistently (apart from ifindex),
> > so maybe lsm_hook_idx?
> 
> Renamed all usages of index -> idx.
> 
> > 
> > >         __u32 prog_btf_fd;
> > >         __u32 func_info_rec_size;
> > >         const void *func_info;
> > > diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
> > > index 0c229f00a67e..60737559a9a6 100644
> > > --- a/tools/lib/bpf/libbpf.c
> > > +++ b/tools/lib/bpf/libbpf.c
> > > @@ -229,6 +229,7 @@ struct bpf_program {
> > >         enum bpf_attach_type expected_attach_type;
> > >         __u32 attach_btf_id;
> > >         __u32 attach_prog_fd;
> > > +       __u32 lsm_hook_index
> > >         void *func_info;
> > >         __u32 func_info_rec_size;
> > >         __u32 func_info_cnt;
> > > @@ -4886,7 +4887,10 @@ load_program(struct bpf_program *prog, struct bpf_insn *insns, int insns_cnt,
> > >         load_attr.insns = insns;
> > >         load_attr.insns_cnt = insns_cnt;
> > >         load_attr.license = license;
> > > -       if (prog->type == BPF_PROG_TYPE_STRUCT_OPS) {
> > > +
> > > +       if (prog->type == BPF_PROG_TYPE_LSM) {
> > > +               load_attr.lsm_hook_index = prog->lsm_hook_index;
> > > +       } else if (prog->type == BPF_PROG_TYPE_STRUCT_OPS) {
> > >                 load_attr.attach_btf_id = prog->attach_btf_id;
> > >         } else if (prog->type == BPF_PROG_TYPE_TRACING) {
> > >                 load_attr.attach_prog_fd = prog->attach_prog_fd;
> > > @@ -4895,6 +4899,7 @@ load_program(struct bpf_program *prog, struct bpf_insn *insns, int insns_cnt,
> > >                 load_attr.kern_version = kern_version;
> > >                 load_attr.prog_ifindex = prog->prog_ifindex;
> > >         }
> > > +
> > >         /* if .BTF.ext was loaded, kernel supports associated BTF for prog */
> > >         if (prog->obj->btf_ext)
> > >                 btf_fd = bpf_object__btf_fd(prog->obj);
> > > @@ -4967,9 +4972,11 @@ static int libbpf_find_attach_btf_id(const char *name,
> > >                                      enum bpf_attach_type attach_type,
> > >                                      __u32 attach_prog_fd);
> > >
> > > +static __s32 btf__find_lsm_hook_index(const char *name);
> > > +
> > >  int bpf_program__load(struct bpf_program *prog, char *license, __u32 kern_ver)
> > >  {
> > > -       int err = 0, fd, i, btf_id;
> > > +       int err = 0, fd, i, btf_id, index;
> > >
> > >         if (prog->type == BPF_PROG_TYPE_TRACING) {
> > >                 btf_id = libbpf_find_attach_btf_id(prog->section_name,
> > > @@ -4980,6 +4987,13 @@ int bpf_program__load(struct bpf_program *prog, char *license, __u32 kern_ver)
> > >                 prog->attach_btf_id = btf_id;
> > >         }
> > >
> > > +       if (prog->type == BPF_PROG_TYPE_LSM) {
> > > +               index = btf__find_lsm_hook_index(prog->section_name);
> > > +               if (index < 0)
> > > +                       return index;
> > > +               prog->lsm_hook_index = index;
> > > +       }
> > > +
> > >         if (prog->instances.nr < 0 || !prog->instances.fds) {
> > >                 if (prog->preprocessor) {
> > >                         pr_warn("Internal error: can't load program '%s'\n",
> > > @@ -6207,6 +6221,7 @@ bool bpf_program__is_##NAME(const struct bpf_program *prog)       \
> > >  }                                                              \
> > >
> > >  BPF_PROG_TYPE_FNS(socket_filter, BPF_PROG_TYPE_SOCKET_FILTER);
> > > +BPF_PROG_TYPE_FNS(lsm, BPF_PROG_TYPE_LSM);
> > >  BPF_PROG_TYPE_FNS(kprobe, BPF_PROG_TYPE_KPROBE);
> > >  BPF_PROG_TYPE_FNS(sched_cls, BPF_PROG_TYPE_SCHED_CLS);
> > >  BPF_PROG_TYPE_FNS(sched_act, BPF_PROG_TYPE_SCHED_ACT);
> > > @@ -6272,6 +6287,8 @@ static struct bpf_link *attach_raw_tp(const struct bpf_sec_def *sec,
> > >                                       struct bpf_program *prog);
> > >  static struct bpf_link *attach_trace(const struct bpf_sec_def *sec,
> > >                                      struct bpf_program *prog);
> > > +static struct bpf_link *attach_lsm(const struct bpf_sec_def *sec,
> > > +                                  struct bpf_program *prog);
> > >
> > >  struct bpf_sec_def {
> > >         const char *sec;
> > > @@ -6315,12 +6332,17 @@ static const struct bpf_sec_def section_defs[] = {
> > >                 .expected_attach_type = BPF_TRACE_FEXIT,
> > >                 .is_attach_btf = true,
> > >                 .attach_fn = attach_trace),
> > > +       SEC_DEF("lsm/", LSM,
> > > +               .expected_attach_type = BPF_LSM_MAC,
> > > +               .attach_fn = attach_lsm),
> > >         BPF_PROG_SEC("xdp",                     BPF_PROG_TYPE_XDP),
> > >         BPF_PROG_SEC("perf_event",              BPF_PROG_TYPE_PERF_EVENT),
> > >         BPF_PROG_SEC("lwt_in",                  BPF_PROG_TYPE_LWT_IN),
> > >         BPF_PROG_SEC("lwt_out",                 BPF_PROG_TYPE_LWT_OUT),
> > >         BPF_PROG_SEC("lwt_xmit",                BPF_PROG_TYPE_LWT_XMIT),
> > >         BPF_PROG_SEC("lwt_seg6local",           BPF_PROG_TYPE_LWT_SEG6LOCAL),
> > > +       BPF_PROG_BTF("lsm/",                    BPF_PROG_TYPE_LSM,
> > > +                                               BPF_LSM_MAC),
> > 
> > This is just a duplicate of SEC_DEF above, remove?
> 
> Oops. Removed.
> 
> > 
> > >         BPF_APROG_SEC("cgroup_skb/ingress",     BPF_PROG_TYPE_CGROUP_SKB,
> > >                                                 BPF_CGROUP_INET_INGRESS),
> > >         BPF_APROG_SEC("cgroup_skb/egress",      BPF_PROG_TYPE_CGROUP_SKB,
> > > @@ -6576,32 +6598,80 @@ static int bpf_object__collect_struct_ops_map_reloc(struct bpf_object *obj,
> > >         return -EINVAL;
> > >  }
> > >
> > > -#define BTF_PREFIX "btf_trace_"
> > > +#define BTF_TRACE_PREFIX "btf_trace_"
> > > +
> > > +static inline int btf__find_by_prefix_kind(struct btf *btf, const char *name,
> > > +                                          const char *prefix, __u32 kind)
> > 
> > this is internal helper, not really BTF API, let's call it
> > find_btf_by_prefix_kind? Also const char *prefix more logically should
> > go before name argument?
> 
> Done.
> 
> > 
> > > +{
> > > +       char btf_type_name[128];
> > > +
> > > +       snprintf(btf_type_name, sizeof(btf_type_name), "%s%s", prefix, name);
> > 
> > check overflow?
> 
> Done.
> 
> > 
> > > +       return btf__find_by_name_kind(btf, btf_type_name, kind);
> > > +}
> > > +
> > > +static __s32 btf__find_lsm_hook_index(const char *name)
> > 
> > this name is violating libbpf naming guidelines. Just
> > `find_lsm_hook_idx` for now?
> 
> Cool. I think I finally hang of the naming convention now :). I mixed
> up btf.c and libbpf.c here.
> 
> > 
> > > +{
> > > +       struct btf *btf = bpf_find_kernel_btf();
> > 
> > ok, it's probably time to do this right. Let's ensure we load kernel
> > BTF just once, keep it inside bpf_object while we need it and then
> > release it after successful load. We are at the point where all the
> > new types of program is loading/releasing kernel BTF for every section
> > and it starts to feel very wasteful.
> 
> Sure, will give it a shot in v3.

Since this will be useful for other programs as well, I will send it
as a separate patch sooner than the v3 of the LSM patch-set.

- KP

> 
> > 
> > > +       const struct bpf_sec_def *sec_def;
> > > +       const struct btf_type *hl_type;
> > > +       struct btf_member *m;
> > > +       __u16 vlen;
> > > +       __s32 hl_id;
> > > +       int j;
> > 
> > j without having i used anywhere?...
> 
> Fixed.
> 
> > 
> > > +
> > > +       sec_def = find_sec_def(name);
> > > +       if (!sec_def)
> > > +               return -ESRCH;
> > > +
> > > +       name += sec_def->len;
> > > +
> > > +       hl_id = btf__find_by_name_kind(btf, "security_hook_heads",
> > > +                                      BTF_KIND_STRUCT);
> > > +       if (hl_id < 0) {
> > > +               pr_debug("security_hook_heads cannot be found in BTF\n");
> > 
> > "in vmlinux BTF" ?
> > 
> > and it should be pr_warn(), we don't really expect this, right?
> 
> Fixed both.
> 
> > 
> > and it should be (hl_id <= 0) with current btf__find_by_name_kind(),
> > and then return hl_id ? : -ESRCH, which further proves we need to
> > change btf__find_by_name_kind() as I suggested below.
> > 
> > > +               return hl_id;
> > > +       }
> > > +
> > > +       hl_type = btf__type_by_id(btf, hl_id);
> > > +       if (!hl_type) {
> > > +               pr_warn("Can't find type for security_hook_heads: %u\n", hl_id);
> > > +               return -EINVAL;
> > 
> > -ESRCH?
> 
> Done.
> 
> > 
> > > +       }
> > > +
> > > +       m = btf_members(hl_type);
> > > +       vlen = btf_vlen(hl_type);
> > > +
> > > +       for (j = 0; j < vlen; j++) {
> > 
> > can add succinct `, m++` here instead
> 
> Done.
> 
> > 
> > > +               if (!strcmp(btf__name_by_offset(btf, m->name_off), name))
> > > +                       return j + 1;
> > 
> > I looked briefly through kernel-side patch introducing lsm_hook_index,
> > but it didn't seem to explain why this index needs to be (unnaturally)
> > 1-based. So asking here first as I'm looking through libbpf changes?
> 
> The lsm_hook_idx is one-based as it makes it easy to validate the
> input. If we make it zero-based it's hard to check if the user
> intended to attach to the LSM hook at index 0 or did not set it.
> 
> We are then up to the verifier to reject the loaded program which
> may or may not match the signature of the hook at lsm_hook_idx = 0.
> 
> I will clarify this in the commit log as well.
> 
> > 
> > > +               m++;
> > > +       }
> > > +
> > > +       pr_warn("Cannot find offset for %s in security_hook_heads\n", name);
> > 
> > it's not offset, rather member index?
> 
> Correct, fixed.
> 
> > 
> > > +       return -ENOENT;
> > 
> > not entirely clear about distinction between ENOENT and ESRCH? So far
> > we typically used ESRCH, does ENOENT have more specific semantics?
> 
> Updated it to ESRCH, to be consistent with libbpf's convention.
> 
> > 
> > > +}
> > > +
> > >  int libbpf_find_vmlinux_btf_id(const char *name,
> > >                                enum bpf_attach_type attach_type)
> > >  {
> > >         struct btf *btf = bpf_find_kernel_btf();
> > > -       char raw_tp_btf[128] = BTF_PREFIX;
> > > -       char *dst = raw_tp_btf + sizeof(BTF_PREFIX) - 1;
> > > -       const char *btf_name;
> > >         int err = -EINVAL;
> > > -       __u32 kind;
> > >
> > >         if (IS_ERR(btf)) {
> > >                 pr_warn("vmlinux BTF is not found\n");
> > >                 return -EINVAL;
> > >         }
> > >
> > > -       if (attach_type == BPF_TRACE_RAW_TP) {
> > > -               /* prepend "btf_trace_" prefix per kernel convention */
> > > -               strncat(dst, name, sizeof(raw_tp_btf) - sizeof(BTF_PREFIX));
> > > -               btf_name = raw_tp_btf;
> > > -               kind = BTF_KIND_TYPEDEF;
> > > -       } else {
> > > -               btf_name = name;
> > > -               kind = BTF_KIND_FUNC;
> > > -       }
> > > -       err = btf__find_by_name_kind(btf, btf_name, kind);
> > > +       if (attach_type == BPF_TRACE_RAW_TP)
> > > +               err = btf__find_by_prefix_kind(btf, name, BTF_TRACE_PREFIX,
> > > +                                              BTF_KIND_TYPEDEF);
> > > +       else
> > > +               err = btf__find_by_name_kind(btf, name, BTF_KIND_FUNC);
> > > +
> > > +       /* err = 0 means void / UNKNOWN which is treated as an error */
> > > +       if (err == 0)
> > > +               err = -EINVAL;
> > 
> > I think it's actually less error-prone to make btf__find_by_name_kind
> > and btf__find_by_prefix_kind to return -ESRCH when type is not found,
> > instead of a valid type_id 0. I just checked, and struct_ops code
> > already is mishandling it, only checking for <0. Could you make this
> > change and just do a natural <0 check everywhere?
> > 
> > 
> > > +
> > >         btf__free(btf);
> > >         return err;
> > >  }
> > > @@ -6630,7 +6700,7 @@ static int libbpf_find_prog_btf_id(const char *name, __u32 attach_prog_fd)
> > >         }
> > >         err = btf__find_by_name_kind(btf, name, BTF_KIND_FUNC);
> > >         btf__free(btf);
> > > -       if (err <= 0) {
> > > +       if (err < 0) {
> > >                 pr_warn("%s is not found in prog's BTF\n", name);
> > >                 goto out;
> > >         }
> > > @@ -7395,6 +7465,43 @@ static struct bpf_link *attach_trace(const struct bpf_sec_def *sec,
> > >         return bpf_program__attach_trace(prog);
> > >  }
> > >
> > > +struct bpf_link *bpf_program__attach_lsm(struct bpf_program *prog)
> > > +{
> > > +       char errmsg[STRERR_BUFSIZE];
> > > +       struct bpf_link_fd *link;
> > > +       int prog_fd, pfd;
> > > +
> > > +       prog_fd = bpf_program__fd(prog);
> > > +       if (prog_fd < 0) {
> > > +               pr_warn("program '%s': can't attach before loaded\n",
> > > +                       bpf_program__title(prog, false));
> > > +               return ERR_PTR(-EINVAL);
> > > +       }
> > > +
> > > +       link = calloc(1, sizeof(*link));
> > > +       if (!link)
> > > +               return ERR_PTR(-ENOMEM);
> > > +       link->link.detach = &bpf_link__detach_fd;
> > > +
> > > +       pfd = bpf_prog_attach(prog_fd, 0, BPF_LSM_MAC,
> > > +                             BPF_F_ALLOW_OVERRIDE);
> > 
> > do we want to always specify ALLOW_OVERRIDE? Or should it be an option?
> 
> I think this is a relic from the duplicate attachment code. With the
> anonymous-fd + link which can be destroyed. The way the OVERRIDE
> should work is:
> 
> -  Destroy the link (detach)
> -  And attach.
> 
> We don't use the BPF_F_ALLOW_OVERRIDE in the attach logic in the LSM.
> So I will get rid of that.
> 
> > 
> > > +       if (pfd < 0) {
> > > +               pfd = -errno;
> > > +               pr_warn("program '%s': failed to attach: %s\n",
> > > +                       bpf_program__title(prog, false),
> > > +                       libbpf_strerror_r(pfd, errmsg, sizeof(errmsg)));
> > > +               return ERR_PTR(pfd);
> > 
> > leaking link here
> 
> Fixed.
> 
> > 
> > > +       }
> > > +       link->fd = pfd;
> > > +       return (struct bpf_link *)link;
> > > +}
> > > +
> > > +static struct bpf_link *attach_lsm(const struct bpf_sec_def *sec,
> > > +                                  struct bpf_program *prog)
> > > +{
> > > +       return bpf_program__attach_lsm(prog);
> > > +}
> > > +
> > >  struct bpf_link *bpf_program__attach(struct bpf_program *prog)
> > >  {
> > >         const struct bpf_sec_def *sec_def;
> > > diff --git a/tools/lib/bpf/libbpf.h b/tools/lib/bpf/libbpf.h
> > > index 01639f9a1062..a97e709a29e6 100644
> > > --- a/tools/lib/bpf/libbpf.h
> > > +++ b/tools/lib/bpf/libbpf.h
> > > @@ -241,6 +241,8 @@ LIBBPF_API struct bpf_link *
> > >  bpf_program__attach_trace(struct bpf_program *prog);
> > >  struct bpf_map;
> > >  LIBBPF_API struct bpf_link *bpf_map__attach_struct_ops(struct bpf_map *map);
> > > +LIBBPF_API struct bpf_link *
> > > +bpf_program__attach_lsm(struct bpf_program *prog);
> > 
> > nit: put it after attach_trace, so that program attaches and map
> > attaches are grouped together, not intermixed
> 
> Done.
> 
> > 
> > >  struct bpf_insn;
> > >
> > >  /*
> > > @@ -318,6 +320,7 @@ LIBBPF_API int bpf_program__set_xdp(struct bpf_program *prog);
> > >  LIBBPF_API int bpf_program__set_perf_event(struct bpf_program *prog);
> > >  LIBBPF_API int bpf_program__set_tracing(struct bpf_program *prog);
> > >  LIBBPF_API int bpf_program__set_struct_ops(struct bpf_program *prog);
> > > +LIBBPF_API int bpf_program__set_lsm(struct bpf_program *prog);
> > >
> > >  LIBBPF_API enum bpf_prog_type bpf_program__get_type(struct bpf_program *prog);
> > >  LIBBPF_API void bpf_program__set_type(struct bpf_program *prog,
> > > @@ -339,6 +342,7 @@ LIBBPF_API bool bpf_program__is_xdp(const struct bpf_program *prog);
> > >  LIBBPF_API bool bpf_program__is_perf_event(const struct bpf_program *prog);
> > >  LIBBPF_API bool bpf_program__is_tracing(const struct bpf_program *prog);
> > >  LIBBPF_API bool bpf_program__is_struct_ops(const struct bpf_program *prog);
> > > +LIBBPF_API bool bpf_program__is_lsm(const struct bpf_program *prog);
> > >
> > >  /*
> > >   * No need for __attribute__((packed)), all members of 'bpf_map_def'
> > > diff --git a/tools/lib/bpf/libbpf.map b/tools/lib/bpf/libbpf.map
> > > index a19f04e6e3d9..3da0452ce679 100644
> > > --- a/tools/lib/bpf/libbpf.map
> > > +++ b/tools/lib/bpf/libbpf.map
> > > @@ -227,4 +227,7 @@ LIBBPF_0.0.7 {
> > >                 bpf_program__is_struct_ops;
> > >                 bpf_program__set_struct_ops;
> > >                 btf__align_of;
> > > +               bpf_program__is_lsm;
> > > +               bpf_program__set_lsm;
> > > +               bpf_program__attach_lsm;
> > 
> > preserve alphabetical order, please
> 
> Sure.
> 
> > 
> > >  } LIBBPF_0.0.6;
> > 
> > > --
> > > 2.20.1
> > >