XDP hardware hints discussion mail archive
 help / color / mirror / Atom feed
From: Andrii Nakryiko <andrii.nakryiko@gmail.com>
To: Jesper Dangaard Brouer <jbrouer@redhat.com>
Cc: "Michal Swiatkowski" <michal.swiatkowski@linux.intel.com>,
	"Jesper Dangaard Brouer" <brouer@redhat.com>,
	"Toke Høiland-Jørgensen" <toke@redhat.com>,
	"Jakub Kicinski" <kuba@kernel.org>,
	"John Fastabend" <john.fastabend@gmail.com>,
	BPF-dev-list <bpf@vger.kernel.org>,
	"Magnus Karlsson" <magnus.karlsson@gmail.com>,
	"William Tu" <u9012063@gmail.com>,
	xdp-hints@xdp-project.net,
	"Zaremba Larysa" <larysa.zaremba@intel.com>,
	"Jiri Olsa" <jolsa@redhat.com>
Subject: Re: XDP-hints: Howto support multiple BTF types per packet basis?
Date: Thu, 9 Sep 2021 11:19:23 -0700	[thread overview]
Message-ID: <CAEf4BzabVVPgRB9V=DAFjzYSx-q59bmBsQQAupKYWy5eUxqVkw@mail.gmail.com> (raw)
In-Reply-To: <936bfbdf-e194-b676-d28a-acf526120155@redhat.com>

On Wed, Sep 8, 2021 at 6:28 AM Jesper Dangaard Brouer
<jbrouer@redhat.com> wrote:
>
>
> On 07/09/2021 08.27, Michal Swiatkowski wrote:
> > On Thu, Sep 02, 2021 at 11:17:43AM +0200, Jesper Dangaard Brouer wrote:
> >>
> >>
> >> On 02/09/2021 04.49, Michal Swiatkowski wrote:
> >>> On Fri, Jul 09, 2021 at 12:57:08PM +0200, Toke Høiland-Jørgensen wrote:
> >>>> Michal Swiatkowski <michal.swiatkowski@linux.intel.com> writes:
> >>>>
> >>>>>> I would expect that the program would decide ahead-of-time which BTF IDs
> >>>>>> it supports, by something like including the relevant structs from
> >>>>>> vmlinux.h. And then we need the BTF ID encoded into the packet metadata
> >>>>>> as well, so that it is possible to check at run-time which driver the
> >>>>>> packet came from (since a packet can be redirected, so you may end up
> >>>>>> having to deal with multiple formats in the same XDP program).
> >>>>>>
> >>>>>> Which would allow you to write code like:
> >>>>>>
> >>>>>> if (ctx->has_driver_meta) {
> >>>>>>     /* this should be at a well-known position, like first (or last) in meta area */
> >>>>>>     __u32 *meta_btf_id = ctx->data_meta;
> >>>>>>     if (*meta_btf_id == BTF_ID_MLX5) {
> >>>>>>       struct meta_mlx5 *meta = ctx->data_meta;
> >>>>>>       /* do something with meta */
> >>>>>>     } else if (meta_btf_id == BTF_ID_I40E) {
> >>>>>>       struct meta_i40e *meta = ctx->data_meta;
> >>>>>>       /* do something with meta */
> >>>>>>     } /* etc */
> >>>>>> }
> >>>>>>
> >>>>>> and libbpf could do relocations based on the different meta structs,
> >>>>>> even removing the code for the ones that don't exist on the running
> >>>>>> kernel.
> >>>>>
> >>>>> This looks nice. In this case we need defintions of struct meta_mlx5 and
> >>>>> struct meta_i40e at build time. How are we going to deliver this to bpf
> >>>>> core app? This will be available in /sys/kernel/btf/mlx5 and
> >>>>> /sys/kernel/btf/i40e (if drivers are loaded). Should we dump this to
> >>>>> vmlinux.h? Or a developer of the xdp program should add this definition
> >>>>> to his code?
> >>>>
> >>>> Well, if the driver just defines the struct, the BTF for it will be
> >>>> automatically part of the driver module BTF. BPF program developers
> >>>> would need to include this in their programs somehow (similar to how
> >>>> you'll need to get the type definitions from vmlinux.h today to use
> >>>> CO-RE); how they do this is up to them. Since this is a compile-time
> >>>> thing it will probably depend on the project (for instance, BCC includes
> >>>> a copy of vmlinux.h in their source tree, but you can also just pick out
> >>>> the structs you need).
> >>>>
> >>>>> Maybe create another /sys/kernel/btf/hints with vmlinux and hints from
> >>>>> all drivers which support hints?
> >>>>
> >>>> It may be useful to have a way for the kernel to export all the hints
> >>>> currently loaded, so libbpf can just use that when relocating. The
> >>>> problem of course being that this will only include drivers that are
> >>>> actually loaded, so users need to make sure to load all their network
> >>>> drivers before loading any XDP programs. I think it would be better if
> >>>> the loader could discover all modules *available* on the system, but I'm
> >>>> not sure if there's a good way to do that.
> >>>>
> >>>>> Previously in this thread someone mentioned this ___ use case in libbpf
> >>>>> and proposed creating something like mega xdp hints structure with all
> >>>>> available fields across all drivers. As I understand this could solve
> >>>>> the problem about defining correct structure at build time. But how will
> >>>>> it work when there will be more than one structures with the same name
> >>>>> before ___? I mean:
> >>>>> struct xdp_hints___mega defined only in core app
> >>>>> struct xdp_hints___mlx5 available when mlx5 driver is loaded
> >>>>> struct xdp_hints___i40e available when i40e driver is loaded
> >>>>>
> >>>>> When there will be only one driver loaded should libbpf do correct
> >>>>> reallocation of fields? What will happen when both of the drivers are
> >>>>> loaded?
> >>>>
> >>>> I think we definitely need to make this easy for developers so they
> >>>> don't have to go and manually track down the driver structs and write
> >>>> the disambiguation code etc. I.e., the example code I included above
> >>>> that checks the frame BTF ID and does the loading based on it should be
> >>>> auto-generated. We already have some precedence for auto-generated code
> >>>> in vmlinux.h and the bpftool skeletons. So maybe we could have a command
> >>>> like 'bpftool gen_xdp_meta <fields>' which would go and lookup all the
> >>>> available driver structs and generate a code helper function that will
> >>>> extract the driver structs and generate the loader code? So that if,
> >>>> say, you're interested in rxhash and tstamp you could do:
> >>>>
> >>>> bpftool gen_xdp_meta rxhash tstamp > my_meta.h
> >>>>
> >>>> which would then produce my_meta.h with content like:
> >>>>
> >>>> struct my_meta { /* contains fields specified on the command line */
> >>>>     u32 rxhash;
> >>>>     u32 tstamp;
> >>>> }
> >>>>
> >>>> struct meta_mlx5 {/*generated from kernel BTF */};
> >>>> struct meta_i40e {/*generated from kernel BTF */};
> >>>>
> >>>> static inline int get_xdp_meta(struct xdp_md *ctx, struct my_meta *meta)
> >>>> {
> >>>>    if (ctx->has_driver_meta) {
> >>>>      /* this should be at a well-known position, like first (or last) in meta area */
> >>>>      __u32 *meta_btf_id = ctx->data_meta;
> >>>>      if (*meta_btf_id == BTF_ID_MLX5) {
> >>>>        struct meta_mlx5 *meta = ctx->data_meta;
> >>>>        my_meta->rxhash = meta->rxhash;
> >>>>        my_meta->tstamp = meta->tstamp;
> >>>>        return 0;
> >>>>      } else if (meta_btf_id == BTF_ID_I40E) {
> >>>>        struct meta_i40e *meta = ctx->data_meta;
> >>>>        my_meta->rxhash = meta->rxhash;
> >>>>        my_meta->tstamp = meta->tstamp;
> >>>>        return 0;
> >>>>      } /* etc */
> >>>>    }
> >>>>    return -ENOENT;
> >>>> }
> >>>
> >>> According to meta_btf_id.
> >>
> >> In BPF-prog (that gets loaded by libbpf), the BTF_ID_MLX5 and BTF_ID_I40E
> >> should be replaced by bpf_core_type_id_kernel() calls.
> >>
> >> I have a code example here[1], where I use the triple-underscore to lookup
> >> btf_id = bpf_core_type_id_kernel(struct sk_buff___local).
> >>
> >> AFAIK (Andrii correctly me if I'm wrong) It is libbpf that does the bpf_id
> >> lookup before loading the BPF-prog into the kernel.
> >>
> >> For AF_XDP we need to code our own similar lookup of the btf_id. (In that
> >> process I imagine that userspace tools could/would read the BTF member
> >> offsets and check it against what they expected).
> >>
> >>
> >>   [1] https://github.com/xdp-project/bpf-examples/blob/master/ktrace-CO-RE/ktrace01_kern.c#L34-L57
> >>
> > Thanks a lot. I tested Your CO-RE example. For defines that are located
> > in vmlinux everything works fine (like for sk_buff). When I tried to get
> > btf id of structures defined in module (loaded module, structure can be
> > find in /sys/kerne/btf/module_name) I always get 0 (not found). Do You
> > know if bpf_core_type_id_kernel() should also support modules btf?
>
> This might be caused by git-submodule libbpf being too old in
> xdp-project/bpf-examples repo.  I will investigate closer.
>
>
> I did notice my bpftool (on my devel box) were tool old, as bpftool
> could not dump info in /sys/kerne/btf/module_name
>
>    # bpftool btf dump file /sys/kernel/btf/igc format raw
>    Error: failed to load BTF from /sys/kernel/btf/igc: Unknown error -22
>
> After updating bpftool it does work.
>
> > Based on:
> > [1] https://lore.kernel.org/bpf/20201110011932.3201430-5-andrii@kernel.org/
> > I assume that modules btfs also are marked as in-kernel, but I can't
> > make it works with bpf_core_type_id_kernel(). My clang version is
> > 12.0.1, so changes needed by modules btf should be there
> > [2] https://reviews.llvm.org/D91489
> >
> >>> Do You have an idea how driver should fill this field?
> >>
> >> (Andrii please correctly me as this is likely wrong:)
> >> I imagine that driver will have a pointer to a 'struct btf' object and the
> >> btf_id can be read via btf_obj_id() (that just return btf->id).
> >> As this also allows driver to take refcnt on the btf-object.
> >> Much like Ederson did in [2].
> >>
> >> Maybe I misunderstood the use of the 'struct btf' object ?
> >>
> >> Maybe it is the wrong approach? As the patchset[2] exports btf_obj_id() and
> >> introduced helper functions that can register/unregister btf objects[3],
> >> which I sense might not be needed today, as modules can get their own BTF
> >> info automatically today.
> >> Maybe this (btf->id) is not the ID we are looking for?
> >>
> >> [2] https://lore.kernel.org/all/20210803010331.39453-11-ederson.desouza@intel.com/
> >> [3]
> >> https://lore.kernel.org/all/20210803010331.39453-2-ederson.desouza@intel.com/
> >>
> >
> > As 'struct btf' object do You mean one module btf with all definitions
> > or specific structure btf object?
> >
> > In case of Your example [1]. If in driver side there will be call to get
> > btf id of sk_buff:
> > id = btf_find_by_name_kind(vmlinux_btf, "sk_buff", BTF_KIND_STRUCT);
> > this id will be the same as id from Your ktrace01 program. I think this
> > is id that we are looking for.
>
> Yes, I think this is the ID we should use.
>
> The 'struct btf' object ID is something else I suspect? Or it is also a
> valid BTF ID?
>
> I wanted to ask Andrii if the IDs are unique across vmlinux and modules?

Depending on what IDs we are talking about (sorry, I don't follow this
thread very closely, so if you are curious about some aspects of BTF
or libbpf APIs, it would be good to have a specific questions with
some context). BTF as kernel object has it's own ID allocated through
idr, so yes, they are unique. so vmlinux BTF object will have it's own
ID, while each module's BTF will have it's own.

But if we are talking about BTF type IDs, that's entirely different
thing. BTF type IDs start from 1 (0 is reserved for special 'VOID'
type) all the way to number of types in vmlinux BTF. Then each module
extends vmlinx BTF starting at N + 1 and going to N + M, where N is
number of BTF types in vmlinux BTF and M is number of added types in
module BTF.

So in that regard each module has BTF type IDs that are overlapping
with other modules, which is why for unique fetching of BTF types from
modules you also need BTF object FD or ID of a module BTF, and then
BTF type ID within that module. But as I said, I didn't follow along
closely, so not sure if I'm answering the right question, sorry.

>
> I suspect as it looks like kern uses the IDR infra[0]:
>   [0] https://www.kernel.org/doc/html/latest/core-api/idr.html
>
>
> >>> hints->btf_id = btf_id_by_name("struct meta_i40e"); /* fill btf id at
> >>> config time */
> >>
> >> Yes, at config time the btf_id can change (and maybe we want to cache the
> >> btf_obj_id() lookup to avoid a function call).
> >>
> >>> btf_id_by_name will get module btf (or vmlinux btf) and search for
> >>> correct name for each ids. Does this look correct?
> >>>
> >>> Is there any way in kernel to get btf id based on name or we have to
> >>> write functions for this? I haven't seen code for this case, but maybe I
> >>> missed it.
> >>
> >> There is a function named: btf_find_by_name_kind()
> >>
> > Thanks, this is what I needed.
>

  reply	other threads:[~2021-09-09 18:19 UTC|newest]

Thread overview: 31+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <20210526125848.1c7adbb0@carbon>
     [not found] ` <CAEf4BzYXUDyQaBjZmb_Q5-z3jw1-Uvdgxm+cfcQjSwb9oRoXnQ@mail.gmail.com>
     [not found]   ` <60aeb01ebcd10_fe49208b8@john-XPS-13-9370.notmuch>
     [not found]     ` <CAEf4Bza3m5dwZ_d0=zAWR+18f5RUjzv9=1NbhTKAO1uzWg_fzQ@mail.gmail.com>
     [not found]       ` <60aeeb5252147_19a622085a@john-XPS-13-9370.notmuch>
     [not found]         ` <CAEf4Bzb1OZHpHYagbVs7s9tMSk4wrbxzGeBCCBHQ-qCOgdu6EQ@mail.gmail.com>
     [not found]           ` <60b08442b18d5_1cf8208a0@john-XPS-13-9370.notmuch>
2021-05-28  9:16             ` Toke Høiland-Jørgensen
2021-05-28 10:38               ` Alexander Lobakin
2021-05-28 14:35               ` John Fastabend
2021-05-28 15:33                 ` Toke Høiland-Jørgensen
2021-05-28 16:02                 ` Jesper Dangaard Brouer
2021-05-28 17:29                   ` John Fastabend
2021-05-30  3:27                     ` Andrii Nakryiko
2021-05-31 11:03                     ` Toke Høiland-Jørgensen
2021-05-31 13:17                       ` Jesper Dangaard Brouer
2021-06-02  0:22                       ` John Fastabend
2021-06-02 16:18                         ` Jakub Kicinski
2021-06-22  7:44                           ` Michal Swiatkowski
2021-06-22 11:53                             ` Toke Høiland-Jørgensen
2021-06-23  8:32                               ` Michal Swiatkowski
2021-06-24 12:23                                 ` Toke Høiland-Jørgensen
2021-06-24 13:07                                   ` Magnus Karlsson
2021-06-24 14:58                                     ` Alexei Starovoitov
2021-06-24 15:11                                   ` Zvi Effron
2021-06-24 16:04                                     ` Toke Høiland-Jørgensen
2021-06-24 16:32                                       ` Zvi Effron
2021-06-24 16:45                                       ` Jesper Dangaard Brouer
2021-07-08  8:32                                   ` Michal Swiatkowski
2021-07-09 10:57                                     ` Toke Høiland-Jørgensen
2021-09-02  2:49                                       ` Michal Swiatkowski
2021-09-02  9:17                                         ` Jesper Dangaard Brouer
2021-09-07  6:27                                           ` Michal Swiatkowski
2021-09-08 13:28                                             ` Jesper Dangaard Brouer
2021-09-09 18:19                                               ` Andrii Nakryiko [this message]
2021-09-10 11:16                                                 ` Jesper Dangaard Brouer
     [not found]   ` <20210526222023.44f9b3c6@carbon>
     [not found]     ` <CAEf4BzZ+VSemxx7WFanw7DfLGN7w42G6ZC4dvOSB1zAsUgRQaw@mail.gmail.com>
2021-05-28 11:16       ` Jesper Dangaard Brouer
2021-05-30  3:24         ` Andrii Nakryiko

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

  List information: https://lists.xdp-project.net/postorius/lists/xdp-hints.xdp-project.net/

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to='CAEf4BzabVVPgRB9V=DAFjzYSx-q59bmBsQQAupKYWy5eUxqVkw@mail.gmail.com' \
    --to=andrii.nakryiko@gmail.com \
    --cc=bpf@vger.kernel.org \
    --cc=brouer@redhat.com \
    --cc=jbrouer@redhat.com \
    --cc=john.fastabend@gmail.com \
    --cc=jolsa@redhat.com \
    --cc=kuba@kernel.org \
    --cc=larysa.zaremba@intel.com \
    --cc=magnus.karlsson@gmail.com \
    --cc=michal.swiatkowski@linux.intel.com \
    --cc=toke@redhat.com \
    --cc=u9012063@gmail.com \
    --cc=xdp-hints@xdp-project.net \
    /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