All of lore.kernel.org
 help / color / mirror / Atom feed
From: "Toke Høiland-Jørgensen" <toke@redhat.com>
To: Jakub Kicinski <kuba@kernel.org>
Cc: John Fastabend <john.fastabend@gmail.com>,
	Alexei Starovoitov <alexei.starovoitov@gmail.com>,
	Lorenzo Bianconi <lorenzo@kernel.org>, bpf <bpf@vger.kernel.org>,
	Network Development <netdev@vger.kernel.org>,
	Lorenzo Bianconi <lorenzo.bianconi@redhat.com>,
	"David S. Miller" <davem@davemloft.net>,
	Alexei Starovoitov <ast@kernel.org>,
	Daniel Borkmann <daniel@iogearbox.net>,
	shayagr@amazon.com, David Ahern <dsahern@kernel.org>,
	Jesper Dangaard Brouer <brouer@redhat.com>,
	Eelco Chaudron <echaudro@redhat.com>,
	Jason Wang <jasowang@redhat.com>,
	Alexander Duyck <alexander.duyck@gmail.com>,
	Saeed Mahameed <saeed@kernel.org>,
	"Fijalkowski, Maciej" <maciej.fijalkowski@intel.com>,
	"Karlsson, Magnus" <magnus.karlsson@intel.com>,
	tirthendu.sarkar@intel.com
Subject: Re: [PATCH v14 bpf-next 00/18] mvneta: introduce XDP multi-buffer support
Date: Mon, 20 Sep 2021 23:01:48 +0200	[thread overview]
Message-ID: <87lf3r3qrn.fsf@toke.dk> (raw)
In-Reply-To: <20210920110216.4c54c9a3@kicinski-fedora-pc1c0hjn.dhcp.thefacebook.com>

Jakub Kicinski <kuba@kernel.org> writes:

> On Sat, 18 Sep 2021 13:53:35 +0200 Toke Høiland-Jørgensen wrote:
>> I'm OK with a bpf_header_pointer()-type helper - I quite like the
>> in-kernel version of this for SKBs, so replicating it as a BPF helper
>> would be great. But I'm a little worried about taking a performance hit.
>> 
>> I.e., if you do:
>> 
>> ptr = bpf_header_pointer(pkt, offset, len, stack_ptr)
>> *ptr = xxx;
>> 
>> then, if the helper ended up copying the data into the stack pointer,
>> you didn't actually change anything in the packet, so you need to do a
>> writeback.
>> 
>> Jakub suggested up-thread that this should be done with some kind of
>> flush() helper. But you don't know whether the header_pointer()-helper
>> copied the data, so you always need to call the flush() helper, which
>> will incur overhead. If the verifier can in-line the helpers that will
>> lower it, but will it be enough to make it negligible?
>
> Depends on the assumptions the program otherwise makes, right?
>
> For reading I'd expect a *layout-independent* TC program would 
> replace approximately:
>
> ptr = <some_ptr>;
> if (ptr + CONST >= md->ptr_end)
> 	if (bpf_pull_data(md, off + CONST))
> 		return DROP;
> 	ptr = <some_ptr>;
> 	if (ptr + CONST >= md->ptr_end)
> 		return DROP; /* da hell? */
> }
>
> With this (pre-inlining):
>
> ptr = bpf_header_pointer(md, offset, len, stack);
> if (!ptr)
> 	return DROP;
>
> Post-inlining (assuming static validation of args to prevent wraps):
>
> if (md->ptr + args->off + args->len < md->ptr_end)
> 	ptr = md->ptr + args->off;
> else
> 	ptr = __bpf_header_pointer(md, offset, len, stack);
> if (!ptr)
> 	return DROP;
>
> But that's based on guesswork so perhaps I'm off base.

Yeah, that's more or less what I was thinking...

> Regarding the flush() I was expecting that most progs will not modify
> the packet (or at least won't modify most headers they load) so no
> point paying the price of tracking changes auto-magically.

... but I guess my mental model assumed that packet writes would be just
as frequent as reads, in which case it would be a win if you could amend
your example like:

> In fact I don't think there is anything infra can do better for
> flushing than the prog itself:
>
> 	bool mod = false;
>
> 	ptr = bpf_header_pointer(...);
> 	...
> 	if (some_cond(...)) {
> 		change_packet(...);
> 		mod = true;
> 	}
> 	...
> 	if (mod)

to have an additional check like:

if (mod && ptr == stack)

(or something to that effect). No?

-Toke

> 		bpf_header_pointer_flush();
>
>
> is simple enough.. to me.


  reply	other threads:[~2021-09-20 21:03 UTC|newest]

Thread overview: 58+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-09-10 16:14 [PATCH v14 bpf-next 00/18] mvneta: introduce XDP multi-buffer support Lorenzo Bianconi
2021-09-10 16:14 ` [PATCH v14 bpf-next 01/18] net: skbuff: add size metadata to skb_shared_info for xdp Lorenzo Bianconi
2021-09-10 16:18   ` Jesper Dangaard Brouer
2021-09-10 16:14 ` [PATCH v14 bpf-next 02/18] xdp: introduce flags field in xdp_buff/xdp_frame Lorenzo Bianconi
2021-09-10 16:19   ` Jesper Dangaard Brouer
2021-09-10 16:14 ` [PATCH v14 bpf-next 03/18] net: mvneta: update mb bit before passing the xdp buffer to eBPF layer Lorenzo Bianconi
2021-09-20  8:25   ` Shay Agroskin
2021-09-20  8:37     ` Lorenzo Bianconi
2021-09-20  8:45       ` Shay Agroskin
2021-09-20  9:00         ` Lorenzo Bianconi
2021-09-10 16:14 ` [PATCH v14 bpf-next 04/18] net: mvneta: simplify mvneta_swbm_add_rx_fragment management Lorenzo Bianconi
2021-09-10 16:14 ` [PATCH v14 bpf-next 05/18] net: xdp: add xdp_update_skb_shared_info utility routine Lorenzo Bianconi
2021-09-10 16:14 ` [PATCH v14 bpf-next 06/18] net: marvell: rely on " Lorenzo Bianconi
2021-09-10 16:14 ` [PATCH v14 bpf-next 07/18] xdp: add multi-buff support to xdp_return_{buff/frame} Lorenzo Bianconi
2021-09-10 16:14 ` [PATCH v14 bpf-next 08/18] net: mvneta: add multi buffer support to XDP_TX Lorenzo Bianconi
2021-09-10 16:14 ` [PATCH v14 bpf-next 09/18] net: mvneta: enable jumbo frames for XDP Lorenzo Bianconi
2021-09-10 16:14 ` [PATCH v14 bpf-next 10/18] bpf: add multi-buff support to the bpf_xdp_adjust_tail() API Lorenzo Bianconi
2021-09-16 16:55   ` Jakub Kicinski
2021-09-17 10:02     ` Lorenzo Bianconi
2021-09-17 13:03       ` Jakub Kicinski
2021-09-10 16:14 ` [PATCH v14 bpf-next 11/18] bpf: introduce bpf_xdp_get_buff_len helper Lorenzo Bianconi
2021-09-10 16:14 ` [PATCH v14 bpf-next 12/18] bpf: add multi-buffer support to xdp copy helpers Lorenzo Bianconi
2021-09-10 16:14 ` [PATCH v14 bpf-next 13/18] bpf: move user_size out of bpf_test_init Lorenzo Bianconi
2021-09-10 16:14 ` [PATCH v14 bpf-next 14/18] bpf: introduce multibuff support to bpf_prog_test_run_xdp() Lorenzo Bianconi
2021-09-10 16:14 ` [PATCH v14 bpf-next 15/18] bpf: test_run: add xdp_shared_info pointer in bpf_test_finish signature Lorenzo Bianconi
2021-09-10 16:14 ` [PATCH v14 bpf-next 16/18] bpf: update xdp_adjust_tail selftest to include multi-buffer Lorenzo Bianconi
2021-09-10 16:14 ` [PATCH v14 bpf-next 17/18] net: xdp: introduce bpf_xdp_adjust_data helper Lorenzo Bianconi
2021-09-10 16:14 ` [PATCH v14 bpf-next 18/18] bpf: add bpf_xdp_adjust_data selftest Lorenzo Bianconi
2021-09-16 16:55 ` [PATCH v14 bpf-next 00/18] mvneta: introduce XDP multi-buffer support Jakub Kicinski
2021-09-17 14:51   ` Lorenzo Bianconi
2021-09-17 18:33     ` Jakub Kicinski
2021-09-17 18:43       ` Alexei Starovoitov
2021-09-17 19:00         ` Jakub Kicinski
2021-09-17 19:10           ` Alexei Starovoitov
2021-09-17 22:07             ` John Fastabend
2021-09-18 11:53               ` Toke Høiland-Jørgensen
2021-09-20 18:02                 ` Jakub Kicinski
2021-09-20 21:01                   ` Toke Høiland-Jørgensen [this message]
2021-09-20 21:25                     ` Jakub Kicinski
2021-09-20 22:44                       ` Toke Høiland-Jørgensen
2021-09-21 10:03                         ` Eelco Chaudron
2021-09-28 11:48                           ` Magnus Karlsson
2021-09-29 10:36                         ` Lorenz Bauer
2021-09-29 12:25                           ` Toke Høiland-Jørgensen
2021-09-29 12:32                             ` Lorenz Bauer
2021-09-29 17:48                             ` Jakub Kicinski
2021-09-29 17:46                           ` Jakub Kicinski
2021-09-29 10:41   ` Lorenz Bauer
2021-09-29 12:10     ` Toke Høiland-Jørgensen
2021-09-29 12:38       ` Lorenz Bauer
2021-09-29 18:54         ` Alexei Starovoitov
2021-09-29 19:22           ` Jakub Kicinski
2021-09-29 20:39             ` Toke Høiland-Jørgensen
2021-10-01  9:03               ` Lorenzo Bianconi
2021-10-01 18:35                 ` Jakub Kicinski
2021-10-06  9:32                   ` Lorenzo Bianconi
2021-10-06 10:08                     ` Eelco Chaudron
2021-10-06 12:15                       ` Lorenzo Bianconi

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=87lf3r3qrn.fsf@toke.dk \
    --to=toke@redhat.com \
    --cc=alexander.duyck@gmail.com \
    --cc=alexei.starovoitov@gmail.com \
    --cc=ast@kernel.org \
    --cc=bpf@vger.kernel.org \
    --cc=brouer@redhat.com \
    --cc=daniel@iogearbox.net \
    --cc=davem@davemloft.net \
    --cc=dsahern@kernel.org \
    --cc=echaudro@redhat.com \
    --cc=jasowang@redhat.com \
    --cc=john.fastabend@gmail.com \
    --cc=kuba@kernel.org \
    --cc=lorenzo.bianconi@redhat.com \
    --cc=lorenzo@kernel.org \
    --cc=maciej.fijalkowski@intel.com \
    --cc=magnus.karlsson@intel.com \
    --cc=netdev@vger.kernel.org \
    --cc=saeed@kernel.org \
    --cc=shayagr@amazon.com \
    --cc=tirthendu.sarkar@intel.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.