From: Alexei Starovoitov <alexei.starovoitov@gmail.com>
To: Dave Marchevsky <davemarchevsky@fb.com>
Cc: bpf@vger.kernel.org, Alexei Starovoitov <ast@kernel.org>,
Daniel Borkmann <daniel@iogearbox.net>,
Andrii Nakryiko <andrii@kernel.org>,
Martin KaFai Lau <martin.lau@kernel.org>,
Kernel Team <kernel-team@fb.com>
Subject: Re: [PATCH v2 bpf-next 5/9] bpf: Migrate bpf_rbtree_add and bpf_list_push_{front,back} to possibly fail
Date: Sat, 15 Apr 2023 18:11:06 -0700 [thread overview]
Message-ID: <20230416011106.hloajgkq5c7tctju@macbook-pro-6.dhcp.thefacebook.com> (raw)
In-Reply-To: <20230415201811.343116-6-davemarchevsky@fb.com>
On Sat, Apr 15, 2023 at 01:18:07PM -0700, Dave Marchevsky wrote:
> -extern void bpf_rbtree_add(struct bpf_rb_root *root, struct bpf_rb_node *node,
> - bool (less)(struct bpf_rb_node *a, const struct bpf_rb_node *b)) __ksym;
> +extern int bpf_rbtree_add_impl(struct bpf_rb_root *root, struct bpf_rb_node *node,
> + bool (less)(struct bpf_rb_node *a, const struct bpf_rb_node *b),
> + void *meta, __u64 off) __ksym;
> +
> +/* Convenience macro to wrap over bpf_rbtree_add_impl */
> +#define bpf_rbtree_add(head, node, less) bpf_rbtree_add_impl(head, node, less, NULL, 0)
Applied, but can we do better here?
It's not a new issue. We have the same inefficiency in bpf_obj_drop.
BPF program populates 1 or 2 extra registers, but the verifier patches the call insn
with necessary values for R4 and R5 for bpf_rbtree_add_impl or R2 for bpf_obj_drop_impl.
So one/two register assignments by bpf prog is a dead code.
Can we come up with a way to avoid this unnecessary register assignment in bpf prog?
Can we keep
extern void bpf_rbtree_add(root, node, less) __ksym; ?
Both in the kernel and in bpf_experimental.h so that libbpf's
bpf_object__resolve_ksym_func_btf_id() -> bpf_core_types_are_compat() check will succeed,
but the kernel bpf_rbtree_add will actually have 5 arguments?
Maybe always_inline or __attribute__((alias(..))) trick we can use?
Or define both and patch bpf code to use _impl later ?
@@ -2053,6 +2053,12 @@ __bpf_kfunc int bpf_rbtree_add_impl(struct bpf_rb_root *root, struct bpf_rb_node
return __bpf_rbtree_add(root, node, (void *)less, meta ? meta->record : NULL, off);
}
+__bpf_kfunc notrace int bpf_rbtree_add(struct bpf_rb_root *root, struct bpf_rb_node *node,
+ bool (less)(struct bpf_rb_node *a, const struct bpf_rb_node *b))
+{
+ return 0;
+}
Only wastes 3 bytes of .text on x86 and extra BTF_KIND_FUNC in vmlinux BTF,
but will save two registers assignment at run-time ?
next prev parent reply other threads:[~2023-04-16 1:11 UTC|newest]
Thread overview: 22+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-04-15 20:18 [PATCH v2 bpf-next 0/9] Shared ownership for local kptrs Dave Marchevsky
2023-04-15 20:18 ` [PATCH v2 bpf-next 1/9] bpf: Remove btf_field_offs, use btf_record's fields instead Dave Marchevsky
2023-04-15 20:18 ` [PATCH v2 bpf-next 2/9] bpf: Introduce opaque bpf_refcount struct and add btf_record plumbing Dave Marchevsky
2023-04-15 20:18 ` [PATCH v2 bpf-next 3/9] bpf: Support refcounted local kptrs in existing semantics Dave Marchevsky
2023-04-15 20:18 ` [PATCH v2 bpf-next 4/9] bpf: Add bpf_refcount_acquire kfunc Dave Marchevsky
2023-04-15 20:18 ` [PATCH v2 bpf-next 5/9] bpf: Migrate bpf_rbtree_add and bpf_list_push_{front,back} to possibly fail Dave Marchevsky
2023-04-16 1:11 ` Alexei Starovoitov [this message]
2023-04-17 18:08 ` Dave Marchevsky
2023-04-15 20:18 ` [PATCH v2 bpf-next 6/9] selftests/bpf: Modify linked_list tests to work with macro-ified inserts Dave Marchevsky
2023-04-15 20:18 ` [PATCH v2 bpf-next 7/9] bpf: Migrate bpf_rbtree_remove to possibly fail Dave Marchevsky
2023-04-15 20:18 ` [PATCH v2 bpf-next 8/9] bpf: Centralize btf_field-specific initialization logic Dave Marchevsky
2023-04-15 20:18 ` [PATCH v2 bpf-next 9/9] selftests/bpf: Add refcounted_kptr tests Dave Marchevsky
2023-04-21 22:17 ` Kumar Kartikeya Dwivedi
2023-04-21 23:49 ` Alexei Starovoitov
2023-04-22 2:06 ` Kumar Kartikeya Dwivedi
2023-04-22 2:18 ` Alexei Starovoitov
2023-04-25 6:53 ` Dave Marchevsky
2023-04-16 0:50 ` [PATCH v2 bpf-next 0/9] Shared ownership for local kptrs patchwork-bot+netdevbpf
2023-04-22 2:03 ` Kumar Kartikeya Dwivedi
2023-04-22 3:21 ` Alexei Starovoitov
2023-04-22 18:42 ` Kumar Kartikeya Dwivedi
2023-04-22 21:25 ` Alexei Starovoitov
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=20230416011106.hloajgkq5c7tctju@macbook-pro-6.dhcp.thefacebook.com \
--to=alexei.starovoitov@gmail.com \
--cc=andrii@kernel.org \
--cc=ast@kernel.org \
--cc=bpf@vger.kernel.org \
--cc=daniel@iogearbox.net \
--cc=davemarchevsky@fb.com \
--cc=kernel-team@fb.com \
--cc=martin.lau@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).