From: Y Song <ys114321@gmail.com>
To: Alban Crequy <alban.crequy@gmail.com>
Cc: "John Fastabend" <john.fastabend@gmail.com>,
"Alexei Starovoitov" <ast@kernel.org>,
"Daniel Borkmann" <daniel@iogearbox.net>,
bpf <bpf@vger.kernel.org>, netdev <netdev@vger.kernel.org>,
LKML <linux-kernel@vger.kernel.org>,
"Alban Crequy" <alban@kinvolk.io>,
"Iago López Galeiras" <iago@kinvolk.io>
Subject: Re: [PATCH bpf-next v3 1/4] bpf: sock ops: add netns ino and dev in bpf context
Date: Sat, 27 Apr 2019 09:34:54 -0700 [thread overview]
Message-ID: <CAH3MdRViqmPWm9UaOEO3ZKa5AodL7AGZ4Bb0FzyDcoqRoDmqNw@mail.gmail.com> (raw)
In-Reply-To: <20190426154848.23490-1-alban@kinvolk.io>
On Fri, Apr 26, 2019 at 8:50 AM Alban Crequy <alban.crequy@gmail.com> wrote:
>
> From: Alban Crequy <alban@kinvolk.io>
>
> sockops programs can now access the network namespace inode and device
> via (struct bpf_sock_ops)->netns_ino and ->netns_dev. This can be useful
> to apply different policies on different network namespaces.
>
> In the unlikely case where network namespaces are not compiled in
> (CONFIG_NET_NS=n), the verifier will not allow access to ->netns_*.
>
> The generated BPF bytecode for netns_ino is loading the correct inode
> number at the time of execution.
>
> However, the generated BPF bytecode for netns_dev is loading an
> immediate value determined at BPF-load-time by looking at the initial
> network namespace. In practice, this works because all netns currently
> use the same virtual device. If this was to change, this code would need
> to be updated too.
>
> Signed-off-by: Alban Crequy <alban@kinvolk.io>
>
> ---
>
> Changes since v1:
> - add netns_dev (review from Alexei)
>
> Changes since v2:
> - replace __u64 by u64 in kernel code (review from Y Song)
> - remove unneeded #else branch: program would be rejected in
> is_valid_access (review from Y Song)
> - allow partial reads (<u64) (review from Y Song)
>
> Note: I have not been able to fully test partial reads on netns_dev.
> The following patches check partial reads in the verifier but it does
> not actually execute the program to check if partial reads generate the
> correct value. I tried to write a BPF program in C and declare the
> struct bpf_sock_ops as a volatile variable and I could get llvm to
> generate the BPF instructions to do partial loads. But then, I get the
> verifier error "dereference of modified ctx ptr R2 off=184 disallowed",
> explained in https://www.spinics.net/lists/netdev/msg531582.html
> What do you think should be done here?
You added partial read tests in test_verifier with raw asm codes.
It should be good enough.
For the compiler generated code causing verifier error, will take
a detailed look later.
Also I did not see a cover letter. For a series with 4 patches, it would be
the best if you can provide a separate cover letter.
> ---
> include/uapi/linux/bpf.h | 2 +
> net/core/filter.c | 94 ++++++++++++++++++++++++++++++++++++++++
> 2 files changed, 96 insertions(+)
>
> diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> index eaf2d3284248..f4f841dde42c 100644
> --- a/include/uapi/linux/bpf.h
> +++ b/include/uapi/linux/bpf.h
> @@ -3213,6 +3213,8 @@ struct bpf_sock_ops {
> __u32 sk_txhash;
> __u64 bytes_received;
> __u64 bytes_acked;
> + __u64 netns_dev;
> + __u64 netns_ino;
> };
>
> /* Definitions for bpf_sock_ops_cb_flags */
> diff --git a/net/core/filter.c b/net/core/filter.c
> index 2f88baf39cc2..9c77464b1501 100644
> --- a/net/core/filter.c
> +++ b/net/core/filter.c
> @@ -75,6 +75,8 @@
> #include <net/seg6_local.h>
> #include <net/lwtunnel.h>
> #include <net/ipv6_stubs.h>
> +#include <linux/kdev_t.h>
> +#include <linux/proc_ns.h>
>
> /**
> * sk_filter_trim_cap - run a packet through a socket filter
> @@ -6810,6 +6812,24 @@ static bool sock_ops_is_valid_access(int off, int size,
> }
> } else {
> switch (off) {
> + case offsetof(struct bpf_sock_ops, netns_dev) ...
> + offsetof(struct bpf_sock_ops, netns_dev) + sizeof(u64) - 1:
> +#ifdef CONFIG_NET_NS
> + if (off - offsetof(struct bpf_sock_ops, netns_dev)
> + + size > sizeof(u64))
This will allow something off = 1, size = 4. This is not what we want as
the access is not properly aligned.
You can look at function bpf_skb_is_valid_access(), esp. the two lines below:
bpf_ctx_record_field_size(info, size_default);
if (!bpf_ctx_narrow_access_ok(off, size, size_default))
return false;
> + return false;
> +#else
> + return false;
> +#endif
> + break;
> + case offsetof(struct bpf_sock_ops, netns_ino):
> +#ifdef CONFIG_NET_NS
> + if (size != sizeof(u64))
> + return false;
> +#else
> + return false;
> +#endif
> + break;
> case bpf_ctx_range_till(struct bpf_sock_ops, bytes_received,
> bytes_acked):
> if (size != sizeof(__u64))
> @@ -7727,6 +7747,11 @@ static u32 sock_addr_convert_ctx_access(enum bpf_access_type type,
> return insn - insn_buf;
> }
>
> +static struct ns_common *sockops_netns_cb(void *private_data)
> +{
> + return &init_net.ns;
> +}
> +
> static u32 sock_ops_convert_ctx_access(enum bpf_access_type type,
> const struct bpf_insn *si,
> struct bpf_insn *insn_buf,
> @@ -7735,6 +7760,10 @@ static u32 sock_ops_convert_ctx_access(enum bpf_access_type type,
> {
> struct bpf_insn *insn = insn_buf;
> int off;
> + struct inode *ns_inode;
> + struct path ns_path;
> + u64 netns_dev;
> + void *res;
>
> /* Helper macro for adding read access to tcp_sock or sock fields. */
> #define SOCK_OPS_GET_FIELD(BPF_FIELD, OBJ_FIELD, OBJ) \
> @@ -7981,6 +8010,71 @@ static u32 sock_ops_convert_ctx_access(enum bpf_access_type type,
> SOCK_OPS_GET_OR_SET_FIELD(sk_txhash, sk_txhash,
> struct sock, type);
> break;
> +
> + case offsetof(struct bpf_sock_ops, netns_dev) ...
> + offsetof(struct bpf_sock_ops, netns_dev) + sizeof(u64) - 1:
> +#ifdef CONFIG_NET_NS
> + /* We get the netns_dev at BPF-load-time and not at
> + * BPF-exec-time. We assume that netns_dev is a constant.
> + */
> + res = ns_get_path_cb(&ns_path, sockops_netns_cb, NULL);
> + if (IS_ERR(res)) {
> + netns_dev = 0;
> + } else {
> + ns_inode = ns_path.dentry->d_inode;
> + netns_dev = new_encode_dev(ns_inode->i_sb->s_dev);
> + }
> + off = si->off;
> + off -= offsetof(struct bpf_sock_ops, netns_dev);
> + switch (BPF_LDST_BYTES(si)) {
> + case sizeof(u64):
> + *insn++ = BPF_MOV64_IMM(si->dst_reg, netns_dev);
> + break;
> + case sizeof(u32):
> + netns_dev = *(u32 *)(((char *)&netns_dev) + off);
> + *insn++ = BPF_MOV32_IMM(si->dst_reg, netns_dev);
> + break;
> + case sizeof(u16):
> + netns_dev = *(u16 *)(((char *)&netns_dev) + off);
> + *insn++ = BPF_MOV32_IMM(si->dst_reg, netns_dev);
> + break;
> + case sizeof(u8):
> + netns_dev = *(u8 *)(((char *)&netns_dev) + off);
> + *insn++ = BPF_MOV32_IMM(si->dst_reg, netns_dev);
> + break;
> + }
> +#endif
> + break;
> +
> + case offsetof(struct bpf_sock_ops, netns_ino):
> +#ifdef CONFIG_NET_NS
> + /* Loading: sk_ops->sk->__sk_common.skc_net.net->ns.inum
> + * Type: (struct bpf_sock_ops_kern *)
> + * ->(struct sock *)
> + * ->(struct sock_common)
> + * .possible_net_t
> + * .(struct net *)
> + * ->(struct ns_common)
> + * .(unsigned int)
> + */
> + BUILD_BUG_ON(offsetof(struct sock, __sk_common) != 0);
> + BUILD_BUG_ON(offsetof(possible_net_t, net) != 0);
> + *insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF(
> + struct bpf_sock_ops_kern, sk),
> + si->dst_reg, si->src_reg,
> + offsetof(struct bpf_sock_ops_kern, sk));
> + *insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF(
> + possible_net_t, net),
> + si->dst_reg, si->dst_reg,
> + offsetof(struct sock_common, skc_net));
> + *insn++ = BPF_LDX_MEM(BPF_FIELD_SIZEOF(
> + struct ns_common, inum),
> + si->dst_reg, si->dst_reg,
> + offsetof(struct net, ns) +
> + offsetof(struct ns_common, inum));
> +#endif
> + break;
> +
> }
> return insn - insn_buf;
> }
> --
> 2.20.1
>
next prev parent reply other threads:[~2019-04-27 16:35 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-04-26 15:48 [PATCH bpf-next v3 1/4] bpf: sock ops: add netns ino and dev in bpf context Alban Crequy
2019-04-26 15:48 ` [PATCH bpf-next v3 2/4] bpf: sync bpf.h to tools/ for bpf_sock_ops->netns* Alban Crequy
2019-04-26 15:48 ` [PATCH bpf-next v3 3/4] selftests: bpf: read netns_ino from struct bpf_sock_ops Alban Crequy
2019-04-26 15:48 ` [PATCH bpf-next v3 4/4] selftests: bpf: verifier: read netns_dev and " Alban Crequy
2019-04-26 21:03 ` [PATCH bpf-next v3 1/4] bpf: sock ops: add netns ino and dev in bpf context Jakub Kicinski
2019-04-27 10:48 ` Alban Crequy
2019-04-27 18:39 ` Jakub Kicinski
2019-04-29 15:42 ` Alban Crequy
2019-04-27 16:34 ` Y Song [this message]
2019-04-29 15:34 ` Alban Crequy
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=CAH3MdRViqmPWm9UaOEO3ZKa5AodL7AGZ4Bb0FzyDcoqRoDmqNw@mail.gmail.com \
--to=ys114321@gmail.com \
--cc=alban.crequy@gmail.com \
--cc=alban@kinvolk.io \
--cc=ast@kernel.org \
--cc=bpf@vger.kernel.org \
--cc=daniel@iogearbox.net \
--cc=iago@kinvolk.io \
--cc=john.fastabend@gmail.com \
--cc=linux-kernel@vger.kernel.org \
--cc=netdev@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).