From: Jiri Olsa <jolsa@redhat.com>
To: Andrii Nakryiko <andrii.nakryiko@gmail.com>
Cc: Alexei Starovoitov <ast@kernel.org>,
Daniel Borkmann <daniel@iogearbox.net>,
Andrii Nakryiko <andrii@kernel.org>,
Masami Hiramatsu <mhiramat@kernel.org>,
Networking <netdev@vger.kernel.org>, bpf <bpf@vger.kernel.org>,
lkml <linux-kernel@vger.kernel.org>,
Martin KaFai Lau <kafai@fb.com>, Song Liu <songliubraving@fb.com>,
Yonghong Song <yhs@fb.com>,
John Fastabend <john.fastabend@gmail.com>,
KP Singh <kpsingh@chromium.org>,
Steven Rostedt <rostedt@goodmis.org>,
Jiri Olsa <olsajiri@gmail.com>
Subject: Re: [PATCH 3/8] bpf: Add bpf_cookie support to fprobe
Date: Tue, 8 Feb 2022 10:07:39 +0100 [thread overview]
Message-ID: <YgIy2yCzbNmKPoxv@krava> (raw)
In-Reply-To: <CAEf4BzbPeQbURZOD93TgPudOk3JD4odsZ9uwriNkrphes9V4dg@mail.gmail.com>
On Mon, Feb 07, 2022 at 10:59:21AM -0800, Andrii Nakryiko wrote:
> On Wed, Feb 2, 2022 at 5:54 AM Jiri Olsa <jolsa@redhat.com> wrote:
> >
> > Adding support to call bpf_get_attach_cookie helper from
> > kprobe program attached by fprobe link.
> >
> > The bpf_cookie is provided by array of u64 values, where
> > each value is paired with provided function address with
> > the same array index.
> >
> > Suggested-by: Andrii Nakryiko <andrii@kernel.org>
> > Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> > ---
> > include/linux/bpf.h | 2 +
> > include/uapi/linux/bpf.h | 1 +
> > kernel/bpf/syscall.c | 83 +++++++++++++++++++++++++++++++++-
> > kernel/trace/bpf_trace.c | 16 ++++++-
> > tools/include/uapi/linux/bpf.h | 1 +
> > 5 files changed, 100 insertions(+), 3 deletions(-)
> >
> > diff --git a/include/linux/bpf.h b/include/linux/bpf.h
> > index 6eb0b180d33b..7b65f05c0487 100644
> > --- a/include/linux/bpf.h
> > +++ b/include/linux/bpf.h
> > @@ -1301,6 +1301,8 @@ static inline void bpf_reset_run_ctx(struct bpf_run_ctx *old_ctx)
> > #endif
> > }
> >
> > +u64 bpf_fprobe_cookie(struct bpf_run_ctx *ctx, u64 ip);
> > +
> > /* BPF program asks to bypass CAP_NET_BIND_SERVICE in bind. */
> > #define BPF_RET_BIND_NO_CAP_NET_BIND_SERVICE (1 << 0)
> > /* BPF program asks to set CN on the packet. */
> > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> > index c0912f0a3dfe..0dc6aa4f9683 100644
> > --- a/include/uapi/linux/bpf.h
> > +++ b/include/uapi/linux/bpf.h
> > @@ -1484,6 +1484,7 @@ union bpf_attr {
> > __aligned_u64 addrs;
> > __u32 cnt;
> > __u32 flags;
> > + __aligned_u64 bpf_cookies;
>
> maybe put it right after addrs, they are closely related and cnt
> describes all of syms/addrs/cookies.
ok
>
> > } fprobe;
> > };
> > } link_create;
> > diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
> > index 0cfbb112c8e1..6c5e74bc43b6 100644
> > --- a/kernel/bpf/syscall.c
> > +++ b/kernel/bpf/syscall.c
> > @@ -33,6 +33,8 @@
> > #include <linux/rcupdate_trace.h>
> > #include <linux/memcontrol.h>
> > #include <linux/fprobe.h>
> > +#include <linux/bsearch.h>
> > +#include <linux/sort.h>
> >
> > #define IS_FD_ARRAY(map) ((map)->map_type == BPF_MAP_TYPE_PERF_EVENT_ARRAY || \
> > (map)->map_type == BPF_MAP_TYPE_CGROUP_ARRAY || \
> > @@ -3025,10 +3027,18 @@ static int bpf_perf_link_attach(const union bpf_attr *attr, struct bpf_prog *pro
> >
> > #ifdef CONFIG_FPROBE
> >
> > +struct bpf_fprobe_cookie {
> > + unsigned long addr;
> > + u64 bpf_cookie;
> > +};
> > +
> > struct bpf_fprobe_link {
> > struct bpf_link link;
> > struct fprobe fp;
> > unsigned long *addrs;
> > + struct bpf_run_ctx run_ctx;
> > + struct bpf_fprobe_cookie *bpf_cookies;
>
> you already have all the addrs above, why keeping a second copy of
> each addrs in bpf_fprobe_cookie. Let's have two arrays: addrs
> (unsigned long) and cookies (u64) and make sure that they are sorted
> together. Then lookup addrs, calculate index, use that index to fetch
> cookie.
>
> Seems like sort_r() provides exactly the interface you'd need to do
> this very easily. Having addrs separate from cookies also a bit
> advantageous in terms of TLB misses (if you need any more persuasion
> ;)
no persuation needed, I actually tried that but it turned out sort_r
is not ready yet ;-)
because you can't pass priv pointer to the swap callback, so we can't
swap the other array.. I did a change to allow that, but it's not trivial
and will need some bigger testing/review because the original sort
calls sort_r, and of course there are many 'sort' users ;-)
>
> > + u32 cnt;
> > };
> >
> > static void bpf_fprobe_link_release(struct bpf_link *link)
> > @@ -3045,6 +3055,7 @@ static void bpf_fprobe_link_dealloc(struct bpf_link *link)
> >
> > fprobe_link = container_of(link, struct bpf_fprobe_link, link);
> > kfree(fprobe_link->addrs);
> > + kfree(fprobe_link->bpf_cookies);
> > kfree(fprobe_link);
> > }
> >
> > @@ -3053,9 +3064,37 @@ static const struct bpf_link_ops bpf_fprobe_link_lops = {
> > .dealloc = bpf_fprobe_link_dealloc,
> > };
> >
> > +static int bpf_fprobe_cookie_cmp(const void *_a, const void *_b)
> > +{
> > + const struct bpf_fprobe_cookie *a = _a;
> > + const struct bpf_fprobe_cookie *b = _b;
> > +
> > + if (a->addr == b->addr)
> > + return 0;
> > + return a->addr < b->addr ? -1 : 1;
> > +}
> > +
> > +u64 bpf_fprobe_cookie(struct bpf_run_ctx *ctx, u64 ip)
> > +{
> > + struct bpf_fprobe_link *fprobe_link;
> > + struct bpf_fprobe_cookie *val, key = {
> > + .addr = (unsigned long) ip,
> > + };
> > +
> > + if (!ctx)
> > + return 0;
>
> is it allowed to have ctx == NULL?
nope, I was also thinking this is more 'WARN_ON[_ONCE]' check
>
> > + fprobe_link = container_of(ctx, struct bpf_fprobe_link, run_ctx);
> > + if (!fprobe_link->bpf_cookies)
> > + return 0;
> > + val = bsearch(&key, fprobe_link->bpf_cookies, fprobe_link->cnt,
> > + sizeof(key), bpf_fprobe_cookie_cmp);
> > + return val ? val->bpf_cookie : 0;
> > +}
> > +
> > static int fprobe_link_prog_run(struct bpf_fprobe_link *fprobe_link,
> > struct pt_regs *regs)
> > {
> > + struct bpf_run_ctx *old_run_ctx;
> > int err;
> >
> > if (unlikely(__this_cpu_inc_return(bpf_prog_active) != 1)) {
> > @@ -3063,12 +3102,16 @@ static int fprobe_link_prog_run(struct bpf_fprobe_link *fprobe_link,
> > goto out;
> > }
> >
> > + old_run_ctx = bpf_set_run_ctx(&fprobe_link->run_ctx);
> > +
> > rcu_read_lock();
> > migrate_disable();
> > err = bpf_prog_run(fprobe_link->link.prog, regs);
> > migrate_enable();
> > rcu_read_unlock();
> >
> > + bpf_reset_run_ctx(old_run_ctx);
> > +
> > out:
> > __this_cpu_dec(bpf_prog_active);
> > return err;
> > @@ -3161,10 +3204,12 @@ static int fprobe_resolve_syms(const void *usyms, u32 cnt,
> >
> > static int bpf_fprobe_link_attach(const union bpf_attr *attr, struct bpf_prog *prog)
> > {
> > + struct bpf_fprobe_cookie *bpf_cookies = NULL;
> > struct bpf_fprobe_link *link = NULL;
> > struct bpf_link_primer link_primer;
> > + void __user *ubpf_cookies;
> > + u32 flags, cnt, i, size;
> > unsigned long *addrs;
> > - u32 flags, cnt, size;
> > void __user *uaddrs;
> > void __user *usyms;
> > int err;
> > @@ -3205,6 +3250,37 @@ static int bpf_fprobe_link_attach(const union bpf_attr *attr, struct bpf_prog *p
> > goto error;
> > }
> >
> > + ubpf_cookies = u64_to_user_ptr(attr->link_create.fprobe.bpf_cookies);
>
> nit: let's call all this "cookies", this bpf_ prefix feels a bit
> redundant (I know about perf_event.bpf_cookie, but still).
ok
>
> > + if (ubpf_cookies) {
> > + u64 *tmp;
> > +
> > + err = -ENOMEM;
> > + tmp = kzalloc(size, GFP_KERNEL);
>
> kvmalloc?
ok
thanks,
jirka
next prev parent reply other threads:[~2022-02-08 9:07 UTC|newest]
Thread overview: 50+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-02-02 13:53 [PATCH 0/8] bpf: Add fprobe link Jiri Olsa
2022-02-02 13:53 ` [PATCH 1/8] bpf: Add support to attach kprobe program with fprobe Jiri Olsa
2022-02-07 18:59 ` Andrii Nakryiko
2022-02-08 8:56 ` Jiri Olsa
2022-02-02 13:53 ` [PATCH 2/8] bpf: Add bpf_get_func_ip kprobe helper for fprobe link Jiri Olsa
2022-02-07 18:59 ` Andrii Nakryiko
2022-02-07 21:01 ` Alexei Starovoitov
2022-02-09 15:01 ` Jiri Olsa
2022-02-09 16:05 ` Andrii Nakryiko
2022-02-09 19:14 ` Jiri Olsa
2022-02-02 13:53 ` [PATCH 3/8] bpf: Add bpf_cookie support to fprobe Jiri Olsa
2022-02-07 18:59 ` Andrii Nakryiko
2022-02-08 9:07 ` Jiri Olsa [this message]
2022-02-08 23:35 ` Andrii Nakryiko
2022-02-08 23:46 ` Jiri Olsa
2022-02-08 23:53 ` Andrii Nakryiko
2022-02-02 13:53 ` [PATCH 4/8] libbpf: Add libbpf__kallsyms_parse function Jiri Olsa
2022-02-07 18:59 ` Andrii Nakryiko
2022-02-08 9:08 ` Jiri Olsa
2022-02-02 13:53 ` [PATCH 5/8] libbpf: Add bpf_link_create support for multi kprobes Jiri Olsa
2022-02-02 13:53 ` [PATCH 6/8] libbpf: Add bpf_program__attach_kprobe_opts " Jiri Olsa
2022-02-07 18:59 ` Andrii Nakryiko
2022-02-08 9:12 ` Jiri Olsa
2022-02-02 13:53 ` [PATCH 7/8] selftest/bpf: Add fprobe attach test Jiri Olsa
2022-02-02 13:53 ` [PATCH 8/8] selftest/bpf: Add fprobe test for bpf_cookie values Jiri Olsa
2022-02-07 18:59 ` Andrii Nakryiko
2022-02-08 9:15 ` Jiri Olsa
2022-02-08 23:24 ` Andrii Nakryiko
2022-02-02 17:09 ` [PATCH 0/8] bpf: Add fprobe link Alexei Starovoitov
2022-02-02 17:24 ` Jiri Olsa
2022-02-02 17:30 ` Alexei Starovoitov
2022-02-03 15:06 ` Jiri Olsa
2022-02-04 0:46 ` Masami Hiramatsu
2022-02-04 1:34 ` Alexei Starovoitov
2022-02-04 2:07 ` Masami Hiramatsu
2022-02-04 2:12 ` Alexei Starovoitov
2022-02-04 2:19 ` Steven Rostedt
2022-02-04 2:42 ` Alexei Starovoitov
2022-02-04 3:17 ` Masami Hiramatsu
2022-02-04 3:59 ` Masami Hiramatsu
2022-02-15 13:21 ` Jiri Olsa
2022-02-16 18:27 ` Andrii Nakryiko
2022-02-17 14:03 ` Masami Hiramatsu
2022-02-17 22:01 ` Andrii Nakryiko
2022-02-18 4:07 ` Masami Hiramatsu
2022-02-18 19:46 ` Andrii Nakryiko
2022-02-19 2:10 ` Alexei Starovoitov
2022-02-21 7:18 ` Masami Hiramatsu
2022-02-22 12:42 ` Jiri Olsa
2022-02-04 3:14 ` Masami Hiramatsu
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=YgIy2yCzbNmKPoxv@krava \
--to=jolsa@redhat.com \
--cc=andrii.nakryiko@gmail.com \
--cc=andrii@kernel.org \
--cc=ast@kernel.org \
--cc=bpf@vger.kernel.org \
--cc=daniel@iogearbox.net \
--cc=john.fastabend@gmail.com \
--cc=kafai@fb.com \
--cc=kpsingh@chromium.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mhiramat@kernel.org \
--cc=netdev@vger.kernel.org \
--cc=olsajiri@gmail.com \
--cc=rostedt@goodmis.org \
--cc=songliubraving@fb.com \
--cc=yhs@fb.com \
/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).