bpftool: Add struct_ops support
This patch adds struct_ops support to the bpftool.
To recap a bit on the recent bpf_struct_ops feature on the kernel side:
It currently supports "struct tcp_congestion_ops" to be implemented
in bpf. At a high level, bpf_struct_ops is struct_ops map populated
with a number of bpf progs. bpf_struct_ops currently supports the
"struct tcp_congestion_ops". However, the bpf_struct_ops design is
generic enough that other kernel struct ops can be supported in
the future.
Although struct_ops is map+progs at a high lever, there are differences
in details. For example,
1) After registering a struct_ops, the struct_ops is held by the kernel
subsystem (e.g. tcp-cc). Thus, there is no need to pin a
struct_ops map or its progs in order to keep them around.
2) To iterate all struct_ops in a system, it iterates all maps
in type BPF_MAP_TYPE_STRUCT_OPS. BPF_MAP_TYPE_STRUCT_OPS is
the current usual filter. In the future, it may need to
filter by other struct_ops specific properties. e.g. filter by
tcp_congestion_ops or other kernel subsystem ops in the future.
3) struct_ops requires the running kernel having BTF info. That allows
more flexibility in handling other kernel structs. e.g. it can
always dump the latest bpf_map_info.
4) Also, "struct_ops" command is not intended to repeat all features
already provided by "map" or "prog". For example, if there really
is a need to pin the struct_ops map, the user can use the "map" cmd
to do that.
While the first attempt was to reuse parts from map/prog.c, it ended up
not a lot to share. The only obvious item is the map_parse_fds() but
that still requires modifications to accommodate struct_ops map specific
filtering (for the immediate and the future needs). Together with the
earlier mentioned differences, it is better to part away from map/prog.c.
The initial set of subcmds are, register, unregister, show, and dump.
For register, it registers all struct_ops maps that can be found in an
obj file. Option can be added in the future to specify a particular
struct_ops map. Also, the common bpf_tcp_cc is stateless (e.g.
bpf_cubic.c and bpf_dctcp.c). The "reuse map" feature is not
implemented in this patch and it can be considered later also.
For other subcmds, please see the man doc for details.
A sample output of dump:
[root@arch-fb-vm1 bpf]# bpftool struct_ops dump name cubic
[{
"bpf_map_info": {
"type": 26,
"id": 64,
"key_size": 4,
"value_size": 256,
"max_entries": 1,
"map_flags": 0,
"name": "cubic",
"ifindex": 0,
"btf_vmlinux_value_type_id": 18452,
"netns_dev": 0,
"netns_ino": 0,
"btf_id": 52,
"btf_key_type_id": 0,
"btf_value_type_id": 0
}
},{
"bpf_struct_ops_tcp_congestion_ops": {
"refcnt": {
"refs": {
"counter": 1
}
},
"state": "BPF_STRUCT_OPS_STATE_INUSE",
"data": {
"list": {
"next": 0,
"prev": 0
},
"key": 0,
"flags": 0,
"init": "void (struct sock *) bictcp_init/prog_id:138",
"release": "void (struct sock *) 0",
"ssthresh": "u32 (struct sock *) bictcp_recalc_ssthresh/prog_id:141",
"cong_avoid": "void (struct sock *, u32, u32) bictcp_cong_avoid/prog_id:140",
"set_state": "void (struct sock *, u8) bictcp_state/prog_id:142",
"cwnd_event": "void (struct sock *, enum tcp_ca_event) bictcp_cwnd_event/prog_id:139",
"in_ack_event": "void (struct sock *, u32) 0",
"undo_cwnd": "u32 (struct sock *) tcp_reno_undo_cwnd/prog_id:144",
"pkts_acked": "void (struct sock *, const struct ack_sample *) bictcp_acked/prog_id:143",
"min_tso_segs": "u32 (struct sock *) 0",
"sndbuf_expand": "u32 (struct sock *) 0",
"cong_control": "void (struct sock *, const struct rate_sample *) 0",
"get_info": "size_t (struct sock *, u32, int *, union tcp_cc_info *) 0",
"name": "bpf_cubic",
"owner": 0
}
}
}
]
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20200318171656.129650-1-kafai@fb.com
2020-03-19 01:16:56 +08:00
|
|
|
// SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
|
|
|
|
/* Copyright (C) 2020 Facebook */
|
|
|
|
|
|
|
|
#include <errno.h>
|
|
|
|
#include <stdio.h>
|
|
|
|
#include <unistd.h>
|
|
|
|
|
|
|
|
#include <linux/err.h>
|
|
|
|
|
|
|
|
#include <bpf/bpf.h>
|
|
|
|
#include <bpf/btf.h>
|
|
|
|
#include <bpf/libbpf.h>
|
|
|
|
|
|
|
|
#include "json_writer.h"
|
|
|
|
#include "main.h"
|
|
|
|
|
|
|
|
#define STRUCT_OPS_VALUE_PREFIX "bpf_struct_ops_"
|
|
|
|
|
|
|
|
static const struct btf_type *map_info_type;
|
|
|
|
static __u32 map_info_alloc_len;
|
|
|
|
static struct btf *btf_vmlinux;
|
|
|
|
static __s32 map_info_type_id;
|
|
|
|
|
|
|
|
struct res {
|
|
|
|
unsigned int nr_maps;
|
|
|
|
unsigned int nr_errs;
|
|
|
|
};
|
|
|
|
|
|
|
|
static const struct btf *get_btf_vmlinux(void)
|
|
|
|
{
|
|
|
|
if (btf_vmlinux)
|
|
|
|
return btf_vmlinux;
|
|
|
|
|
|
|
|
btf_vmlinux = libbpf_find_kernel_btf();
|
2022-11-20 19:26:32 +08:00
|
|
|
if (!btf_vmlinux)
|
bpftool: Add struct_ops support
This patch adds struct_ops support to the bpftool.
To recap a bit on the recent bpf_struct_ops feature on the kernel side:
It currently supports "struct tcp_congestion_ops" to be implemented
in bpf. At a high level, bpf_struct_ops is struct_ops map populated
with a number of bpf progs. bpf_struct_ops currently supports the
"struct tcp_congestion_ops". However, the bpf_struct_ops design is
generic enough that other kernel struct ops can be supported in
the future.
Although struct_ops is map+progs at a high lever, there are differences
in details. For example,
1) After registering a struct_ops, the struct_ops is held by the kernel
subsystem (e.g. tcp-cc). Thus, there is no need to pin a
struct_ops map or its progs in order to keep them around.
2) To iterate all struct_ops in a system, it iterates all maps
in type BPF_MAP_TYPE_STRUCT_OPS. BPF_MAP_TYPE_STRUCT_OPS is
the current usual filter. In the future, it may need to
filter by other struct_ops specific properties. e.g. filter by
tcp_congestion_ops or other kernel subsystem ops in the future.
3) struct_ops requires the running kernel having BTF info. That allows
more flexibility in handling other kernel structs. e.g. it can
always dump the latest bpf_map_info.
4) Also, "struct_ops" command is not intended to repeat all features
already provided by "map" or "prog". For example, if there really
is a need to pin the struct_ops map, the user can use the "map" cmd
to do that.
While the first attempt was to reuse parts from map/prog.c, it ended up
not a lot to share. The only obvious item is the map_parse_fds() but
that still requires modifications to accommodate struct_ops map specific
filtering (for the immediate and the future needs). Together with the
earlier mentioned differences, it is better to part away from map/prog.c.
The initial set of subcmds are, register, unregister, show, and dump.
For register, it registers all struct_ops maps that can be found in an
obj file. Option can be added in the future to specify a particular
struct_ops map. Also, the common bpf_tcp_cc is stateless (e.g.
bpf_cubic.c and bpf_dctcp.c). The "reuse map" feature is not
implemented in this patch and it can be considered later also.
For other subcmds, please see the man doc for details.
A sample output of dump:
[root@arch-fb-vm1 bpf]# bpftool struct_ops dump name cubic
[{
"bpf_map_info": {
"type": 26,
"id": 64,
"key_size": 4,
"value_size": 256,
"max_entries": 1,
"map_flags": 0,
"name": "cubic",
"ifindex": 0,
"btf_vmlinux_value_type_id": 18452,
"netns_dev": 0,
"netns_ino": 0,
"btf_id": 52,
"btf_key_type_id": 0,
"btf_value_type_id": 0
}
},{
"bpf_struct_ops_tcp_congestion_ops": {
"refcnt": {
"refs": {
"counter": 1
}
},
"state": "BPF_STRUCT_OPS_STATE_INUSE",
"data": {
"list": {
"next": 0,
"prev": 0
},
"key": 0,
"flags": 0,
"init": "void (struct sock *) bictcp_init/prog_id:138",
"release": "void (struct sock *) 0",
"ssthresh": "u32 (struct sock *) bictcp_recalc_ssthresh/prog_id:141",
"cong_avoid": "void (struct sock *, u32, u32) bictcp_cong_avoid/prog_id:140",
"set_state": "void (struct sock *, u8) bictcp_state/prog_id:142",
"cwnd_event": "void (struct sock *, enum tcp_ca_event) bictcp_cwnd_event/prog_id:139",
"in_ack_event": "void (struct sock *, u32) 0",
"undo_cwnd": "u32 (struct sock *) tcp_reno_undo_cwnd/prog_id:144",
"pkts_acked": "void (struct sock *, const struct ack_sample *) bictcp_acked/prog_id:143",
"min_tso_segs": "u32 (struct sock *) 0",
"sndbuf_expand": "u32 (struct sock *) 0",
"cong_control": "void (struct sock *, const struct rate_sample *) 0",
"get_info": "size_t (struct sock *, u32, int *, union tcp_cc_info *) 0",
"name": "bpf_cubic",
"owner": 0
}
}
}
]
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20200318171656.129650-1-kafai@fb.com
2020-03-19 01:16:56 +08:00
|
|
|
p_err("struct_ops requires kernel CONFIG_DEBUG_INFO_BTF=y");
|
|
|
|
|
|
|
|
return btf_vmlinux;
|
|
|
|
}
|
|
|
|
|
|
|
|
static const char *get_kern_struct_ops_name(const struct bpf_map_info *info)
|
|
|
|
{
|
|
|
|
const struct btf *kern_btf;
|
|
|
|
const struct btf_type *t;
|
|
|
|
const char *st_ops_name;
|
|
|
|
|
|
|
|
kern_btf = get_btf_vmlinux();
|
2022-11-20 19:26:32 +08:00
|
|
|
if (!kern_btf)
|
bpftool: Add struct_ops support
This patch adds struct_ops support to the bpftool.
To recap a bit on the recent bpf_struct_ops feature on the kernel side:
It currently supports "struct tcp_congestion_ops" to be implemented
in bpf. At a high level, bpf_struct_ops is struct_ops map populated
with a number of bpf progs. bpf_struct_ops currently supports the
"struct tcp_congestion_ops". However, the bpf_struct_ops design is
generic enough that other kernel struct ops can be supported in
the future.
Although struct_ops is map+progs at a high lever, there are differences
in details. For example,
1) After registering a struct_ops, the struct_ops is held by the kernel
subsystem (e.g. tcp-cc). Thus, there is no need to pin a
struct_ops map or its progs in order to keep them around.
2) To iterate all struct_ops in a system, it iterates all maps
in type BPF_MAP_TYPE_STRUCT_OPS. BPF_MAP_TYPE_STRUCT_OPS is
the current usual filter. In the future, it may need to
filter by other struct_ops specific properties. e.g. filter by
tcp_congestion_ops or other kernel subsystem ops in the future.
3) struct_ops requires the running kernel having BTF info. That allows
more flexibility in handling other kernel structs. e.g. it can
always dump the latest bpf_map_info.
4) Also, "struct_ops" command is not intended to repeat all features
already provided by "map" or "prog". For example, if there really
is a need to pin the struct_ops map, the user can use the "map" cmd
to do that.
While the first attempt was to reuse parts from map/prog.c, it ended up
not a lot to share. The only obvious item is the map_parse_fds() but
that still requires modifications to accommodate struct_ops map specific
filtering (for the immediate and the future needs). Together with the
earlier mentioned differences, it is better to part away from map/prog.c.
The initial set of subcmds are, register, unregister, show, and dump.
For register, it registers all struct_ops maps that can be found in an
obj file. Option can be added in the future to specify a particular
struct_ops map. Also, the common bpf_tcp_cc is stateless (e.g.
bpf_cubic.c and bpf_dctcp.c). The "reuse map" feature is not
implemented in this patch and it can be considered later also.
For other subcmds, please see the man doc for details.
A sample output of dump:
[root@arch-fb-vm1 bpf]# bpftool struct_ops dump name cubic
[{
"bpf_map_info": {
"type": 26,
"id": 64,
"key_size": 4,
"value_size": 256,
"max_entries": 1,
"map_flags": 0,
"name": "cubic",
"ifindex": 0,
"btf_vmlinux_value_type_id": 18452,
"netns_dev": 0,
"netns_ino": 0,
"btf_id": 52,
"btf_key_type_id": 0,
"btf_value_type_id": 0
}
},{
"bpf_struct_ops_tcp_congestion_ops": {
"refcnt": {
"refs": {
"counter": 1
}
},
"state": "BPF_STRUCT_OPS_STATE_INUSE",
"data": {
"list": {
"next": 0,
"prev": 0
},
"key": 0,
"flags": 0,
"init": "void (struct sock *) bictcp_init/prog_id:138",
"release": "void (struct sock *) 0",
"ssthresh": "u32 (struct sock *) bictcp_recalc_ssthresh/prog_id:141",
"cong_avoid": "void (struct sock *, u32, u32) bictcp_cong_avoid/prog_id:140",
"set_state": "void (struct sock *, u8) bictcp_state/prog_id:142",
"cwnd_event": "void (struct sock *, enum tcp_ca_event) bictcp_cwnd_event/prog_id:139",
"in_ack_event": "void (struct sock *, u32) 0",
"undo_cwnd": "u32 (struct sock *) tcp_reno_undo_cwnd/prog_id:144",
"pkts_acked": "void (struct sock *, const struct ack_sample *) bictcp_acked/prog_id:143",
"min_tso_segs": "u32 (struct sock *) 0",
"sndbuf_expand": "u32 (struct sock *) 0",
"cong_control": "void (struct sock *, const struct rate_sample *) 0",
"get_info": "size_t (struct sock *, u32, int *, union tcp_cc_info *) 0",
"name": "bpf_cubic",
"owner": 0
}
}
}
]
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20200318171656.129650-1-kafai@fb.com
2020-03-19 01:16:56 +08:00
|
|
|
return "<btf_vmlinux_not_found>";
|
|
|
|
|
|
|
|
t = btf__type_by_id(kern_btf, info->btf_vmlinux_value_type_id);
|
|
|
|
st_ops_name = btf__name_by_offset(kern_btf, t->name_off);
|
|
|
|
st_ops_name += strlen(STRUCT_OPS_VALUE_PREFIX);
|
|
|
|
|
|
|
|
return st_ops_name;
|
|
|
|
}
|
|
|
|
|
|
|
|
static __s32 get_map_info_type_id(void)
|
|
|
|
{
|
|
|
|
const struct btf *kern_btf;
|
|
|
|
|
|
|
|
if (map_info_type_id)
|
|
|
|
return map_info_type_id;
|
|
|
|
|
|
|
|
kern_btf = get_btf_vmlinux();
|
2022-11-20 19:26:07 +08:00
|
|
|
if (!kern_btf)
|
|
|
|
return 0;
|
bpftool: Add struct_ops support
This patch adds struct_ops support to the bpftool.
To recap a bit on the recent bpf_struct_ops feature on the kernel side:
It currently supports "struct tcp_congestion_ops" to be implemented
in bpf. At a high level, bpf_struct_ops is struct_ops map populated
with a number of bpf progs. bpf_struct_ops currently supports the
"struct tcp_congestion_ops". However, the bpf_struct_ops design is
generic enough that other kernel struct ops can be supported in
the future.
Although struct_ops is map+progs at a high lever, there are differences
in details. For example,
1) After registering a struct_ops, the struct_ops is held by the kernel
subsystem (e.g. tcp-cc). Thus, there is no need to pin a
struct_ops map or its progs in order to keep them around.
2) To iterate all struct_ops in a system, it iterates all maps
in type BPF_MAP_TYPE_STRUCT_OPS. BPF_MAP_TYPE_STRUCT_OPS is
the current usual filter. In the future, it may need to
filter by other struct_ops specific properties. e.g. filter by
tcp_congestion_ops or other kernel subsystem ops in the future.
3) struct_ops requires the running kernel having BTF info. That allows
more flexibility in handling other kernel structs. e.g. it can
always dump the latest bpf_map_info.
4) Also, "struct_ops" command is not intended to repeat all features
already provided by "map" or "prog". For example, if there really
is a need to pin the struct_ops map, the user can use the "map" cmd
to do that.
While the first attempt was to reuse parts from map/prog.c, it ended up
not a lot to share. The only obvious item is the map_parse_fds() but
that still requires modifications to accommodate struct_ops map specific
filtering (for the immediate and the future needs). Together with the
earlier mentioned differences, it is better to part away from map/prog.c.
The initial set of subcmds are, register, unregister, show, and dump.
For register, it registers all struct_ops maps that can be found in an
obj file. Option can be added in the future to specify a particular
struct_ops map. Also, the common bpf_tcp_cc is stateless (e.g.
bpf_cubic.c and bpf_dctcp.c). The "reuse map" feature is not
implemented in this patch and it can be considered later also.
For other subcmds, please see the man doc for details.
A sample output of dump:
[root@arch-fb-vm1 bpf]# bpftool struct_ops dump name cubic
[{
"bpf_map_info": {
"type": 26,
"id": 64,
"key_size": 4,
"value_size": 256,
"max_entries": 1,
"map_flags": 0,
"name": "cubic",
"ifindex": 0,
"btf_vmlinux_value_type_id": 18452,
"netns_dev": 0,
"netns_ino": 0,
"btf_id": 52,
"btf_key_type_id": 0,
"btf_value_type_id": 0
}
},{
"bpf_struct_ops_tcp_congestion_ops": {
"refcnt": {
"refs": {
"counter": 1
}
},
"state": "BPF_STRUCT_OPS_STATE_INUSE",
"data": {
"list": {
"next": 0,
"prev": 0
},
"key": 0,
"flags": 0,
"init": "void (struct sock *) bictcp_init/prog_id:138",
"release": "void (struct sock *) 0",
"ssthresh": "u32 (struct sock *) bictcp_recalc_ssthresh/prog_id:141",
"cong_avoid": "void (struct sock *, u32, u32) bictcp_cong_avoid/prog_id:140",
"set_state": "void (struct sock *, u8) bictcp_state/prog_id:142",
"cwnd_event": "void (struct sock *, enum tcp_ca_event) bictcp_cwnd_event/prog_id:139",
"in_ack_event": "void (struct sock *, u32) 0",
"undo_cwnd": "u32 (struct sock *) tcp_reno_undo_cwnd/prog_id:144",
"pkts_acked": "void (struct sock *, const struct ack_sample *) bictcp_acked/prog_id:143",
"min_tso_segs": "u32 (struct sock *) 0",
"sndbuf_expand": "u32 (struct sock *) 0",
"cong_control": "void (struct sock *, const struct rate_sample *) 0",
"get_info": "size_t (struct sock *, u32, int *, union tcp_cc_info *) 0",
"name": "bpf_cubic",
"owner": 0
}
}
}
]
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20200318171656.129650-1-kafai@fb.com
2020-03-19 01:16:56 +08:00
|
|
|
|
|
|
|
map_info_type_id = btf__find_by_name_kind(kern_btf, "bpf_map_info",
|
|
|
|
BTF_KIND_STRUCT);
|
|
|
|
if (map_info_type_id < 0) {
|
|
|
|
p_err("can't find bpf_map_info from btf_vmlinux");
|
|
|
|
return map_info_type_id;
|
|
|
|
}
|
|
|
|
map_info_type = btf__type_by_id(kern_btf, map_info_type_id);
|
|
|
|
|
|
|
|
/* Ensure map_info_alloc() has at least what the bpftool needs */
|
|
|
|
map_info_alloc_len = map_info_type->size;
|
|
|
|
if (map_info_alloc_len < sizeof(struct bpf_map_info))
|
|
|
|
map_info_alloc_len = sizeof(struct bpf_map_info);
|
|
|
|
|
|
|
|
return map_info_type_id;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* If the subcmd needs to print out the bpf_map_info,
|
|
|
|
* it should always call map_info_alloc to allocate
|
|
|
|
* a bpf_map_info object instead of allocating it
|
|
|
|
* on the stack.
|
|
|
|
*
|
|
|
|
* map_info_alloc() will take the running kernel's btf
|
|
|
|
* into account. i.e. it will consider the
|
|
|
|
* sizeof(struct bpf_map_info) of the running kernel.
|
|
|
|
*
|
|
|
|
* It will enable the "struct_ops" cmd to print the latest
|
|
|
|
* "struct bpf_map_info".
|
|
|
|
*
|
|
|
|
* [ Recall that "struct_ops" requires the kernel's btf to
|
|
|
|
* be available ]
|
|
|
|
*/
|
|
|
|
static struct bpf_map_info *map_info_alloc(__u32 *alloc_len)
|
|
|
|
{
|
|
|
|
struct bpf_map_info *info;
|
|
|
|
|
|
|
|
if (get_map_info_type_id() < 0)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
info = calloc(1, map_info_alloc_len);
|
|
|
|
if (!info)
|
|
|
|
p_err("mem alloc failed");
|
|
|
|
else
|
|
|
|
*alloc_len = map_info_alloc_len;
|
|
|
|
|
|
|
|
return info;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* It iterates all struct_ops maps of the system.
|
|
|
|
* It returns the fd in "*res_fd" and map_info in "*info".
|
|
|
|
* In the very first iteration, info->id should be 0.
|
|
|
|
* An optional map "*name" filter can be specified.
|
|
|
|
* The filter can be made more flexible in the future.
|
|
|
|
* e.g. filter by kernel-struct-ops-name, regex-name, glob-name, ...etc.
|
|
|
|
*
|
|
|
|
* Return value:
|
|
|
|
* 1: A struct_ops map found. It is returned in "*res_fd" and "*info".
|
|
|
|
* The caller can continue to call get_next in the future.
|
|
|
|
* 0: No struct_ops map is returned.
|
|
|
|
* All struct_ops map has been found.
|
|
|
|
* -1: Error and the caller should abort the iteration.
|
|
|
|
*/
|
|
|
|
static int get_next_struct_ops_map(const char *name, int *res_fd,
|
|
|
|
struct bpf_map_info *info, __u32 info_len)
|
|
|
|
{
|
|
|
|
__u32 id = info->id;
|
|
|
|
int err, fd;
|
|
|
|
|
|
|
|
while (true) {
|
|
|
|
err = bpf_map_get_next_id(id, &id);
|
|
|
|
if (err) {
|
|
|
|
if (errno == ENOENT)
|
|
|
|
return 0;
|
|
|
|
p_err("can't get next map: %s", strerror(errno));
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
fd = bpf_map_get_fd_by_id(id);
|
|
|
|
if (fd < 0) {
|
|
|
|
if (errno == ENOENT)
|
|
|
|
continue;
|
|
|
|
p_err("can't get map by id (%u): %s",
|
|
|
|
id, strerror(errno));
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
2023-02-15 07:12:16 +08:00
|
|
|
err = bpf_map_get_info_by_fd(fd, info, &info_len);
|
bpftool: Add struct_ops support
This patch adds struct_ops support to the bpftool.
To recap a bit on the recent bpf_struct_ops feature on the kernel side:
It currently supports "struct tcp_congestion_ops" to be implemented
in bpf. At a high level, bpf_struct_ops is struct_ops map populated
with a number of bpf progs. bpf_struct_ops currently supports the
"struct tcp_congestion_ops". However, the bpf_struct_ops design is
generic enough that other kernel struct ops can be supported in
the future.
Although struct_ops is map+progs at a high lever, there are differences
in details. For example,
1) After registering a struct_ops, the struct_ops is held by the kernel
subsystem (e.g. tcp-cc). Thus, there is no need to pin a
struct_ops map or its progs in order to keep them around.
2) To iterate all struct_ops in a system, it iterates all maps
in type BPF_MAP_TYPE_STRUCT_OPS. BPF_MAP_TYPE_STRUCT_OPS is
the current usual filter. In the future, it may need to
filter by other struct_ops specific properties. e.g. filter by
tcp_congestion_ops or other kernel subsystem ops in the future.
3) struct_ops requires the running kernel having BTF info. That allows
more flexibility in handling other kernel structs. e.g. it can
always dump the latest bpf_map_info.
4) Also, "struct_ops" command is not intended to repeat all features
already provided by "map" or "prog". For example, if there really
is a need to pin the struct_ops map, the user can use the "map" cmd
to do that.
While the first attempt was to reuse parts from map/prog.c, it ended up
not a lot to share. The only obvious item is the map_parse_fds() but
that still requires modifications to accommodate struct_ops map specific
filtering (for the immediate and the future needs). Together with the
earlier mentioned differences, it is better to part away from map/prog.c.
The initial set of subcmds are, register, unregister, show, and dump.
For register, it registers all struct_ops maps that can be found in an
obj file. Option can be added in the future to specify a particular
struct_ops map. Also, the common bpf_tcp_cc is stateless (e.g.
bpf_cubic.c and bpf_dctcp.c). The "reuse map" feature is not
implemented in this patch and it can be considered later also.
For other subcmds, please see the man doc for details.
A sample output of dump:
[root@arch-fb-vm1 bpf]# bpftool struct_ops dump name cubic
[{
"bpf_map_info": {
"type": 26,
"id": 64,
"key_size": 4,
"value_size": 256,
"max_entries": 1,
"map_flags": 0,
"name": "cubic",
"ifindex": 0,
"btf_vmlinux_value_type_id": 18452,
"netns_dev": 0,
"netns_ino": 0,
"btf_id": 52,
"btf_key_type_id": 0,
"btf_value_type_id": 0
}
},{
"bpf_struct_ops_tcp_congestion_ops": {
"refcnt": {
"refs": {
"counter": 1
}
},
"state": "BPF_STRUCT_OPS_STATE_INUSE",
"data": {
"list": {
"next": 0,
"prev": 0
},
"key": 0,
"flags": 0,
"init": "void (struct sock *) bictcp_init/prog_id:138",
"release": "void (struct sock *) 0",
"ssthresh": "u32 (struct sock *) bictcp_recalc_ssthresh/prog_id:141",
"cong_avoid": "void (struct sock *, u32, u32) bictcp_cong_avoid/prog_id:140",
"set_state": "void (struct sock *, u8) bictcp_state/prog_id:142",
"cwnd_event": "void (struct sock *, enum tcp_ca_event) bictcp_cwnd_event/prog_id:139",
"in_ack_event": "void (struct sock *, u32) 0",
"undo_cwnd": "u32 (struct sock *) tcp_reno_undo_cwnd/prog_id:144",
"pkts_acked": "void (struct sock *, const struct ack_sample *) bictcp_acked/prog_id:143",
"min_tso_segs": "u32 (struct sock *) 0",
"sndbuf_expand": "u32 (struct sock *) 0",
"cong_control": "void (struct sock *, const struct rate_sample *) 0",
"get_info": "size_t (struct sock *, u32, int *, union tcp_cc_info *) 0",
"name": "bpf_cubic",
"owner": 0
}
}
}
]
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20200318171656.129650-1-kafai@fb.com
2020-03-19 01:16:56 +08:00
|
|
|
if (err) {
|
|
|
|
p_err("can't get map info: %s", strerror(errno));
|
|
|
|
close(fd);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (info->type == BPF_MAP_TYPE_STRUCT_OPS &&
|
|
|
|
(!name || !strcmp(name, info->name))) {
|
|
|
|
*res_fd = fd;
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
close(fd);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static int cmd_retval(const struct res *res, bool must_have_one_map)
|
|
|
|
{
|
|
|
|
if (res->nr_errs || (!res->nr_maps && must_have_one_map))
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* "data" is the work_func private storage */
|
|
|
|
typedef int (*work_func)(int fd, const struct bpf_map_info *info, void *data,
|
|
|
|
struct json_writer *wtr);
|
|
|
|
|
|
|
|
/* Find all struct_ops map in the system.
|
|
|
|
* Filter out by "name" (if specified).
|
|
|
|
* Then call "func(fd, info, data, wtr)" on each struct_ops map found.
|
|
|
|
*/
|
|
|
|
static struct res do_search(const char *name, work_func func, void *data,
|
|
|
|
struct json_writer *wtr)
|
|
|
|
{
|
|
|
|
struct bpf_map_info *info;
|
|
|
|
struct res res = {};
|
|
|
|
__u32 info_len;
|
|
|
|
int fd, err;
|
|
|
|
|
|
|
|
info = map_info_alloc(&info_len);
|
|
|
|
if (!info) {
|
|
|
|
res.nr_errs++;
|
|
|
|
return res;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (wtr)
|
|
|
|
jsonw_start_array(wtr);
|
|
|
|
while ((err = get_next_struct_ops_map(name, &fd, info, info_len)) == 1) {
|
|
|
|
res.nr_maps++;
|
|
|
|
err = func(fd, info, data, wtr);
|
|
|
|
if (err)
|
|
|
|
res.nr_errs++;
|
|
|
|
close(fd);
|
|
|
|
}
|
|
|
|
if (wtr)
|
|
|
|
jsonw_end_array(wtr);
|
|
|
|
|
|
|
|
if (err)
|
|
|
|
res.nr_errs++;
|
|
|
|
|
|
|
|
if (!wtr && name && !res.nr_errs && !res.nr_maps)
|
|
|
|
/* It is not printing empty [].
|
|
|
|
* Thus, needs to specifically say nothing found
|
|
|
|
* for "name" here.
|
|
|
|
*/
|
|
|
|
p_err("no struct_ops found for %s", name);
|
|
|
|
else if (!wtr && json_output && !res.nr_errs)
|
|
|
|
/* The "func()" above is not writing any json (i.e. !wtr
|
|
|
|
* test here).
|
|
|
|
*
|
|
|
|
* However, "-j" is enabled and there is no errs here,
|
|
|
|
* so call json_null() as the current convention of
|
|
|
|
* other cmds.
|
|
|
|
*/
|
|
|
|
jsonw_null(json_wtr);
|
|
|
|
|
|
|
|
free(info);
|
|
|
|
return res;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct res do_one_id(const char *id_str, work_func func, void *data,
|
|
|
|
struct json_writer *wtr)
|
|
|
|
{
|
|
|
|
struct bpf_map_info *info;
|
|
|
|
struct res res = {};
|
|
|
|
unsigned long id;
|
|
|
|
__u32 info_len;
|
|
|
|
char *endptr;
|
|
|
|
int fd;
|
|
|
|
|
|
|
|
id = strtoul(id_str, &endptr, 0);
|
|
|
|
if (*endptr || !id || id > UINT32_MAX) {
|
|
|
|
p_err("invalid id %s", id_str);
|
|
|
|
res.nr_errs++;
|
|
|
|
return res;
|
|
|
|
}
|
|
|
|
|
|
|
|
fd = bpf_map_get_fd_by_id(id);
|
2021-11-02 06:43:54 +08:00
|
|
|
if (fd < 0) {
|
bpftool: Add struct_ops support
This patch adds struct_ops support to the bpftool.
To recap a bit on the recent bpf_struct_ops feature on the kernel side:
It currently supports "struct tcp_congestion_ops" to be implemented
in bpf. At a high level, bpf_struct_ops is struct_ops map populated
with a number of bpf progs. bpf_struct_ops currently supports the
"struct tcp_congestion_ops". However, the bpf_struct_ops design is
generic enough that other kernel struct ops can be supported in
the future.
Although struct_ops is map+progs at a high lever, there are differences
in details. For example,
1) After registering a struct_ops, the struct_ops is held by the kernel
subsystem (e.g. tcp-cc). Thus, there is no need to pin a
struct_ops map or its progs in order to keep them around.
2) To iterate all struct_ops in a system, it iterates all maps
in type BPF_MAP_TYPE_STRUCT_OPS. BPF_MAP_TYPE_STRUCT_OPS is
the current usual filter. In the future, it may need to
filter by other struct_ops specific properties. e.g. filter by
tcp_congestion_ops or other kernel subsystem ops in the future.
3) struct_ops requires the running kernel having BTF info. That allows
more flexibility in handling other kernel structs. e.g. it can
always dump the latest bpf_map_info.
4) Also, "struct_ops" command is not intended to repeat all features
already provided by "map" or "prog". For example, if there really
is a need to pin the struct_ops map, the user can use the "map" cmd
to do that.
While the first attempt was to reuse parts from map/prog.c, it ended up
not a lot to share. The only obvious item is the map_parse_fds() but
that still requires modifications to accommodate struct_ops map specific
filtering (for the immediate and the future needs). Together with the
earlier mentioned differences, it is better to part away from map/prog.c.
The initial set of subcmds are, register, unregister, show, and dump.
For register, it registers all struct_ops maps that can be found in an
obj file. Option can be added in the future to specify a particular
struct_ops map. Also, the common bpf_tcp_cc is stateless (e.g.
bpf_cubic.c and bpf_dctcp.c). The "reuse map" feature is not
implemented in this patch and it can be considered later also.
For other subcmds, please see the man doc for details.
A sample output of dump:
[root@arch-fb-vm1 bpf]# bpftool struct_ops dump name cubic
[{
"bpf_map_info": {
"type": 26,
"id": 64,
"key_size": 4,
"value_size": 256,
"max_entries": 1,
"map_flags": 0,
"name": "cubic",
"ifindex": 0,
"btf_vmlinux_value_type_id": 18452,
"netns_dev": 0,
"netns_ino": 0,
"btf_id": 52,
"btf_key_type_id": 0,
"btf_value_type_id": 0
}
},{
"bpf_struct_ops_tcp_congestion_ops": {
"refcnt": {
"refs": {
"counter": 1
}
},
"state": "BPF_STRUCT_OPS_STATE_INUSE",
"data": {
"list": {
"next": 0,
"prev": 0
},
"key": 0,
"flags": 0,
"init": "void (struct sock *) bictcp_init/prog_id:138",
"release": "void (struct sock *) 0",
"ssthresh": "u32 (struct sock *) bictcp_recalc_ssthresh/prog_id:141",
"cong_avoid": "void (struct sock *, u32, u32) bictcp_cong_avoid/prog_id:140",
"set_state": "void (struct sock *, u8) bictcp_state/prog_id:142",
"cwnd_event": "void (struct sock *, enum tcp_ca_event) bictcp_cwnd_event/prog_id:139",
"in_ack_event": "void (struct sock *, u32) 0",
"undo_cwnd": "u32 (struct sock *) tcp_reno_undo_cwnd/prog_id:144",
"pkts_acked": "void (struct sock *, const struct ack_sample *) bictcp_acked/prog_id:143",
"min_tso_segs": "u32 (struct sock *) 0",
"sndbuf_expand": "u32 (struct sock *) 0",
"cong_control": "void (struct sock *, const struct rate_sample *) 0",
"get_info": "size_t (struct sock *, u32, int *, union tcp_cc_info *) 0",
"name": "bpf_cubic",
"owner": 0
}
}
}
]
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20200318171656.129650-1-kafai@fb.com
2020-03-19 01:16:56 +08:00
|
|
|
p_err("can't get map by id (%lu): %s", id, strerror(errno));
|
|
|
|
res.nr_errs++;
|
|
|
|
return res;
|
|
|
|
}
|
|
|
|
|
|
|
|
info = map_info_alloc(&info_len);
|
|
|
|
if (!info) {
|
|
|
|
res.nr_errs++;
|
|
|
|
goto done;
|
|
|
|
}
|
|
|
|
|
2023-02-15 07:12:16 +08:00
|
|
|
if (bpf_map_get_info_by_fd(fd, info, &info_len)) {
|
bpftool: Add struct_ops support
This patch adds struct_ops support to the bpftool.
To recap a bit on the recent bpf_struct_ops feature on the kernel side:
It currently supports "struct tcp_congestion_ops" to be implemented
in bpf. At a high level, bpf_struct_ops is struct_ops map populated
with a number of bpf progs. bpf_struct_ops currently supports the
"struct tcp_congestion_ops". However, the bpf_struct_ops design is
generic enough that other kernel struct ops can be supported in
the future.
Although struct_ops is map+progs at a high lever, there are differences
in details. For example,
1) After registering a struct_ops, the struct_ops is held by the kernel
subsystem (e.g. tcp-cc). Thus, there is no need to pin a
struct_ops map or its progs in order to keep them around.
2) To iterate all struct_ops in a system, it iterates all maps
in type BPF_MAP_TYPE_STRUCT_OPS. BPF_MAP_TYPE_STRUCT_OPS is
the current usual filter. In the future, it may need to
filter by other struct_ops specific properties. e.g. filter by
tcp_congestion_ops or other kernel subsystem ops in the future.
3) struct_ops requires the running kernel having BTF info. That allows
more flexibility in handling other kernel structs. e.g. it can
always dump the latest bpf_map_info.
4) Also, "struct_ops" command is not intended to repeat all features
already provided by "map" or "prog". For example, if there really
is a need to pin the struct_ops map, the user can use the "map" cmd
to do that.
While the first attempt was to reuse parts from map/prog.c, it ended up
not a lot to share. The only obvious item is the map_parse_fds() but
that still requires modifications to accommodate struct_ops map specific
filtering (for the immediate and the future needs). Together with the
earlier mentioned differences, it is better to part away from map/prog.c.
The initial set of subcmds are, register, unregister, show, and dump.
For register, it registers all struct_ops maps that can be found in an
obj file. Option can be added in the future to specify a particular
struct_ops map. Also, the common bpf_tcp_cc is stateless (e.g.
bpf_cubic.c and bpf_dctcp.c). The "reuse map" feature is not
implemented in this patch and it can be considered later also.
For other subcmds, please see the man doc for details.
A sample output of dump:
[root@arch-fb-vm1 bpf]# bpftool struct_ops dump name cubic
[{
"bpf_map_info": {
"type": 26,
"id": 64,
"key_size": 4,
"value_size": 256,
"max_entries": 1,
"map_flags": 0,
"name": "cubic",
"ifindex": 0,
"btf_vmlinux_value_type_id": 18452,
"netns_dev": 0,
"netns_ino": 0,
"btf_id": 52,
"btf_key_type_id": 0,
"btf_value_type_id": 0
}
},{
"bpf_struct_ops_tcp_congestion_ops": {
"refcnt": {
"refs": {
"counter": 1
}
},
"state": "BPF_STRUCT_OPS_STATE_INUSE",
"data": {
"list": {
"next": 0,
"prev": 0
},
"key": 0,
"flags": 0,
"init": "void (struct sock *) bictcp_init/prog_id:138",
"release": "void (struct sock *) 0",
"ssthresh": "u32 (struct sock *) bictcp_recalc_ssthresh/prog_id:141",
"cong_avoid": "void (struct sock *, u32, u32) bictcp_cong_avoid/prog_id:140",
"set_state": "void (struct sock *, u8) bictcp_state/prog_id:142",
"cwnd_event": "void (struct sock *, enum tcp_ca_event) bictcp_cwnd_event/prog_id:139",
"in_ack_event": "void (struct sock *, u32) 0",
"undo_cwnd": "u32 (struct sock *) tcp_reno_undo_cwnd/prog_id:144",
"pkts_acked": "void (struct sock *, const struct ack_sample *) bictcp_acked/prog_id:143",
"min_tso_segs": "u32 (struct sock *) 0",
"sndbuf_expand": "u32 (struct sock *) 0",
"cong_control": "void (struct sock *, const struct rate_sample *) 0",
"get_info": "size_t (struct sock *, u32, int *, union tcp_cc_info *) 0",
"name": "bpf_cubic",
"owner": 0
}
}
}
]
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20200318171656.129650-1-kafai@fb.com
2020-03-19 01:16:56 +08:00
|
|
|
p_err("can't get map info: %s", strerror(errno));
|
|
|
|
res.nr_errs++;
|
|
|
|
goto done;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (info->type != BPF_MAP_TYPE_STRUCT_OPS) {
|
|
|
|
p_err("%s id %u is not a struct_ops map", info->name, info->id);
|
|
|
|
res.nr_errs++;
|
|
|
|
goto done;
|
|
|
|
}
|
|
|
|
|
|
|
|
res.nr_maps++;
|
|
|
|
|
|
|
|
if (func(fd, info, data, wtr))
|
|
|
|
res.nr_errs++;
|
|
|
|
else if (!wtr && json_output)
|
|
|
|
/* The "func()" above is not writing any json (i.e. !wtr
|
|
|
|
* test here).
|
|
|
|
*
|
|
|
|
* However, "-j" is enabled and there is no errs here,
|
|
|
|
* so call json_null() as the current convention of
|
|
|
|
* other cmds.
|
|
|
|
*/
|
|
|
|
jsonw_null(json_wtr);
|
|
|
|
|
|
|
|
done:
|
|
|
|
free(info);
|
|
|
|
close(fd);
|
|
|
|
|
|
|
|
return res;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct res do_work_on_struct_ops(const char *search_type,
|
|
|
|
const char *search_term,
|
|
|
|
work_func func, void *data,
|
|
|
|
struct json_writer *wtr)
|
|
|
|
{
|
|
|
|
if (search_type) {
|
|
|
|
if (is_prefix(search_type, "id"))
|
|
|
|
return do_one_id(search_term, func, data, wtr);
|
|
|
|
else if (!is_prefix(search_type, "name"))
|
|
|
|
usage();
|
|
|
|
}
|
|
|
|
|
|
|
|
return do_search(search_term, func, data, wtr);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int __do_show(int fd, const struct bpf_map_info *info, void *data,
|
|
|
|
struct json_writer *wtr)
|
|
|
|
{
|
|
|
|
if (wtr) {
|
|
|
|
jsonw_start_object(wtr);
|
|
|
|
jsonw_uint_field(wtr, "id", info->id);
|
|
|
|
jsonw_string_field(wtr, "name", info->name);
|
|
|
|
jsonw_string_field(wtr, "kernel_struct_ops",
|
|
|
|
get_kern_struct_ops_name(info));
|
|
|
|
jsonw_end_object(wtr);
|
|
|
|
} else {
|
|
|
|
printf("%u: %-15s %-32s\n", info->id, info->name,
|
|
|
|
get_kern_struct_ops_name(info));
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int do_show(int argc, char **argv)
|
|
|
|
{
|
|
|
|
const char *search_type = NULL, *search_term = NULL;
|
|
|
|
struct res res;
|
|
|
|
|
|
|
|
if (argc && argc != 2)
|
|
|
|
usage();
|
|
|
|
|
|
|
|
if (argc == 2) {
|
|
|
|
search_type = GET_ARG();
|
|
|
|
search_term = GET_ARG();
|
|
|
|
}
|
|
|
|
|
|
|
|
res = do_work_on_struct_ops(search_type, search_term, __do_show,
|
|
|
|
NULL, json_wtr);
|
|
|
|
|
|
|
|
return cmd_retval(&res, !!search_term);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int __do_dump(int fd, const struct bpf_map_info *info, void *data,
|
|
|
|
struct json_writer *wtr)
|
|
|
|
{
|
|
|
|
struct btf_dumper *d = (struct btf_dumper *)data;
|
|
|
|
const struct btf_type *struct_ops_type;
|
|
|
|
const struct btf *kern_btf = d->btf;
|
|
|
|
const char *struct_ops_name;
|
|
|
|
int zero = 0;
|
|
|
|
void *value;
|
|
|
|
|
|
|
|
/* note: d->jw == wtr */
|
|
|
|
|
|
|
|
kern_btf = d->btf;
|
|
|
|
|
|
|
|
/* The kernel supporting BPF_MAP_TYPE_STRUCT_OPS must have
|
|
|
|
* btf_vmlinux_value_type_id.
|
|
|
|
*/
|
|
|
|
struct_ops_type = btf__type_by_id(kern_btf,
|
|
|
|
info->btf_vmlinux_value_type_id);
|
|
|
|
struct_ops_name = btf__name_by_offset(kern_btf,
|
|
|
|
struct_ops_type->name_off);
|
|
|
|
value = calloc(1, info->value_size);
|
|
|
|
if (!value) {
|
|
|
|
p_err("mem alloc failed");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (bpf_map_lookup_elem(fd, &zero, value)) {
|
|
|
|
p_err("can't lookup struct_ops map %s id %u",
|
|
|
|
info->name, info->id);
|
|
|
|
free(value);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
jsonw_start_object(wtr);
|
|
|
|
jsonw_name(wtr, "bpf_map_info");
|
|
|
|
btf_dumper_type(d, map_info_type_id, (void *)info);
|
|
|
|
jsonw_end_object(wtr);
|
|
|
|
|
|
|
|
jsonw_start_object(wtr);
|
|
|
|
jsonw_name(wtr, struct_ops_name);
|
|
|
|
btf_dumper_type(d, info->btf_vmlinux_value_type_id, value);
|
|
|
|
jsonw_end_object(wtr);
|
|
|
|
|
|
|
|
free(value);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int do_dump(int argc, char **argv)
|
|
|
|
{
|
|
|
|
const char *search_type = NULL, *search_term = NULL;
|
|
|
|
json_writer_t *wtr = json_wtr;
|
|
|
|
const struct btf *kern_btf;
|
|
|
|
struct btf_dumper d = {};
|
|
|
|
struct res res;
|
|
|
|
|
|
|
|
if (argc && argc != 2)
|
|
|
|
usage();
|
|
|
|
|
|
|
|
if (argc == 2) {
|
|
|
|
search_type = GET_ARG();
|
|
|
|
search_term = GET_ARG();
|
|
|
|
}
|
|
|
|
|
|
|
|
kern_btf = get_btf_vmlinux();
|
2022-11-20 19:26:32 +08:00
|
|
|
if (!kern_btf)
|
bpftool: Add struct_ops support
This patch adds struct_ops support to the bpftool.
To recap a bit on the recent bpf_struct_ops feature on the kernel side:
It currently supports "struct tcp_congestion_ops" to be implemented
in bpf. At a high level, bpf_struct_ops is struct_ops map populated
with a number of bpf progs. bpf_struct_ops currently supports the
"struct tcp_congestion_ops". However, the bpf_struct_ops design is
generic enough that other kernel struct ops can be supported in
the future.
Although struct_ops is map+progs at a high lever, there are differences
in details. For example,
1) After registering a struct_ops, the struct_ops is held by the kernel
subsystem (e.g. tcp-cc). Thus, there is no need to pin a
struct_ops map or its progs in order to keep them around.
2) To iterate all struct_ops in a system, it iterates all maps
in type BPF_MAP_TYPE_STRUCT_OPS. BPF_MAP_TYPE_STRUCT_OPS is
the current usual filter. In the future, it may need to
filter by other struct_ops specific properties. e.g. filter by
tcp_congestion_ops or other kernel subsystem ops in the future.
3) struct_ops requires the running kernel having BTF info. That allows
more flexibility in handling other kernel structs. e.g. it can
always dump the latest bpf_map_info.
4) Also, "struct_ops" command is not intended to repeat all features
already provided by "map" or "prog". For example, if there really
is a need to pin the struct_ops map, the user can use the "map" cmd
to do that.
While the first attempt was to reuse parts from map/prog.c, it ended up
not a lot to share. The only obvious item is the map_parse_fds() but
that still requires modifications to accommodate struct_ops map specific
filtering (for the immediate and the future needs). Together with the
earlier mentioned differences, it is better to part away from map/prog.c.
The initial set of subcmds are, register, unregister, show, and dump.
For register, it registers all struct_ops maps that can be found in an
obj file. Option can be added in the future to specify a particular
struct_ops map. Also, the common bpf_tcp_cc is stateless (e.g.
bpf_cubic.c and bpf_dctcp.c). The "reuse map" feature is not
implemented in this patch and it can be considered later also.
For other subcmds, please see the man doc for details.
A sample output of dump:
[root@arch-fb-vm1 bpf]# bpftool struct_ops dump name cubic
[{
"bpf_map_info": {
"type": 26,
"id": 64,
"key_size": 4,
"value_size": 256,
"max_entries": 1,
"map_flags": 0,
"name": "cubic",
"ifindex": 0,
"btf_vmlinux_value_type_id": 18452,
"netns_dev": 0,
"netns_ino": 0,
"btf_id": 52,
"btf_key_type_id": 0,
"btf_value_type_id": 0
}
},{
"bpf_struct_ops_tcp_congestion_ops": {
"refcnt": {
"refs": {
"counter": 1
}
},
"state": "BPF_STRUCT_OPS_STATE_INUSE",
"data": {
"list": {
"next": 0,
"prev": 0
},
"key": 0,
"flags": 0,
"init": "void (struct sock *) bictcp_init/prog_id:138",
"release": "void (struct sock *) 0",
"ssthresh": "u32 (struct sock *) bictcp_recalc_ssthresh/prog_id:141",
"cong_avoid": "void (struct sock *, u32, u32) bictcp_cong_avoid/prog_id:140",
"set_state": "void (struct sock *, u8) bictcp_state/prog_id:142",
"cwnd_event": "void (struct sock *, enum tcp_ca_event) bictcp_cwnd_event/prog_id:139",
"in_ack_event": "void (struct sock *, u32) 0",
"undo_cwnd": "u32 (struct sock *) tcp_reno_undo_cwnd/prog_id:144",
"pkts_acked": "void (struct sock *, const struct ack_sample *) bictcp_acked/prog_id:143",
"min_tso_segs": "u32 (struct sock *) 0",
"sndbuf_expand": "u32 (struct sock *) 0",
"cong_control": "void (struct sock *, const struct rate_sample *) 0",
"get_info": "size_t (struct sock *, u32, int *, union tcp_cc_info *) 0",
"name": "bpf_cubic",
"owner": 0
}
}
}
]
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20200318171656.129650-1-kafai@fb.com
2020-03-19 01:16:56 +08:00
|
|
|
return -1;
|
|
|
|
|
|
|
|
if (!json_output) {
|
|
|
|
wtr = jsonw_new(stdout);
|
|
|
|
if (!wtr) {
|
|
|
|
p_err("can't create json writer");
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
jsonw_pretty(wtr, true);
|
|
|
|
}
|
|
|
|
|
|
|
|
d.btf = kern_btf;
|
|
|
|
d.jw = wtr;
|
|
|
|
d.is_plain_text = !json_output;
|
|
|
|
d.prog_id_as_func_ptr = true;
|
|
|
|
|
|
|
|
res = do_work_on_struct_ops(search_type, search_term, __do_dump, &d,
|
|
|
|
wtr);
|
|
|
|
|
|
|
|
if (!json_output)
|
|
|
|
jsonw_destroy(&wtr);
|
|
|
|
|
|
|
|
return cmd_retval(&res, !!search_term);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int __do_unregister(int fd, const struct bpf_map_info *info, void *data,
|
|
|
|
struct json_writer *wtr)
|
|
|
|
{
|
|
|
|
int zero = 0;
|
|
|
|
|
|
|
|
if (bpf_map_delete_elem(fd, &zero)) {
|
|
|
|
p_err("can't unload %s %s id %u: %s",
|
|
|
|
get_kern_struct_ops_name(info), info->name,
|
|
|
|
info->id, strerror(errno));
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
p_info("Unregistered %s %s id %u",
|
|
|
|
get_kern_struct_ops_name(info), info->name,
|
|
|
|
info->id);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int do_unregister(int argc, char **argv)
|
|
|
|
{
|
|
|
|
const char *search_type, *search_term;
|
|
|
|
struct res res;
|
|
|
|
|
|
|
|
if (argc != 2)
|
|
|
|
usage();
|
|
|
|
|
|
|
|
search_type = GET_ARG();
|
|
|
|
search_term = GET_ARG();
|
|
|
|
|
|
|
|
res = do_work_on_struct_ops(search_type, search_term,
|
|
|
|
__do_unregister, NULL, NULL);
|
|
|
|
|
|
|
|
return cmd_retval(&res, true);
|
|
|
|
}
|
|
|
|
|
2023-04-20 08:28:21 +08:00
|
|
|
static int pin_link(struct bpf_link *link, const char *pindir,
|
|
|
|
const char *name)
|
|
|
|
{
|
|
|
|
char pinfile[PATH_MAX];
|
|
|
|
int err;
|
|
|
|
|
|
|
|
err = pathname_concat(pinfile, sizeof(pinfile), pindir, name);
|
|
|
|
if (err)
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
return bpf_link__pin(link, pinfile);
|
|
|
|
}
|
|
|
|
|
bpftool: Add struct_ops support
This patch adds struct_ops support to the bpftool.
To recap a bit on the recent bpf_struct_ops feature on the kernel side:
It currently supports "struct tcp_congestion_ops" to be implemented
in bpf. At a high level, bpf_struct_ops is struct_ops map populated
with a number of bpf progs. bpf_struct_ops currently supports the
"struct tcp_congestion_ops". However, the bpf_struct_ops design is
generic enough that other kernel struct ops can be supported in
the future.
Although struct_ops is map+progs at a high lever, there are differences
in details. For example,
1) After registering a struct_ops, the struct_ops is held by the kernel
subsystem (e.g. tcp-cc). Thus, there is no need to pin a
struct_ops map or its progs in order to keep them around.
2) To iterate all struct_ops in a system, it iterates all maps
in type BPF_MAP_TYPE_STRUCT_OPS. BPF_MAP_TYPE_STRUCT_OPS is
the current usual filter. In the future, it may need to
filter by other struct_ops specific properties. e.g. filter by
tcp_congestion_ops or other kernel subsystem ops in the future.
3) struct_ops requires the running kernel having BTF info. That allows
more flexibility in handling other kernel structs. e.g. it can
always dump the latest bpf_map_info.
4) Also, "struct_ops" command is not intended to repeat all features
already provided by "map" or "prog". For example, if there really
is a need to pin the struct_ops map, the user can use the "map" cmd
to do that.
While the first attempt was to reuse parts from map/prog.c, it ended up
not a lot to share. The only obvious item is the map_parse_fds() but
that still requires modifications to accommodate struct_ops map specific
filtering (for the immediate and the future needs). Together with the
earlier mentioned differences, it is better to part away from map/prog.c.
The initial set of subcmds are, register, unregister, show, and dump.
For register, it registers all struct_ops maps that can be found in an
obj file. Option can be added in the future to specify a particular
struct_ops map. Also, the common bpf_tcp_cc is stateless (e.g.
bpf_cubic.c and bpf_dctcp.c). The "reuse map" feature is not
implemented in this patch and it can be considered later also.
For other subcmds, please see the man doc for details.
A sample output of dump:
[root@arch-fb-vm1 bpf]# bpftool struct_ops dump name cubic
[{
"bpf_map_info": {
"type": 26,
"id": 64,
"key_size": 4,
"value_size": 256,
"max_entries": 1,
"map_flags": 0,
"name": "cubic",
"ifindex": 0,
"btf_vmlinux_value_type_id": 18452,
"netns_dev": 0,
"netns_ino": 0,
"btf_id": 52,
"btf_key_type_id": 0,
"btf_value_type_id": 0
}
},{
"bpf_struct_ops_tcp_congestion_ops": {
"refcnt": {
"refs": {
"counter": 1
}
},
"state": "BPF_STRUCT_OPS_STATE_INUSE",
"data": {
"list": {
"next": 0,
"prev": 0
},
"key": 0,
"flags": 0,
"init": "void (struct sock *) bictcp_init/prog_id:138",
"release": "void (struct sock *) 0",
"ssthresh": "u32 (struct sock *) bictcp_recalc_ssthresh/prog_id:141",
"cong_avoid": "void (struct sock *, u32, u32) bictcp_cong_avoid/prog_id:140",
"set_state": "void (struct sock *, u8) bictcp_state/prog_id:142",
"cwnd_event": "void (struct sock *, enum tcp_ca_event) bictcp_cwnd_event/prog_id:139",
"in_ack_event": "void (struct sock *, u32) 0",
"undo_cwnd": "u32 (struct sock *) tcp_reno_undo_cwnd/prog_id:144",
"pkts_acked": "void (struct sock *, const struct ack_sample *) bictcp_acked/prog_id:143",
"min_tso_segs": "u32 (struct sock *) 0",
"sndbuf_expand": "u32 (struct sock *) 0",
"cong_control": "void (struct sock *, const struct rate_sample *) 0",
"get_info": "size_t (struct sock *, u32, int *, union tcp_cc_info *) 0",
"name": "bpf_cubic",
"owner": 0
}
}
}
]
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20200318171656.129650-1-kafai@fb.com
2020-03-19 01:16:56 +08:00
|
|
|
static int do_register(int argc, char **argv)
|
|
|
|
{
|
2021-12-10 03:38:40 +08:00
|
|
|
LIBBPF_OPTS(bpf_object_open_opts, open_opts);
|
2023-04-20 08:28:21 +08:00
|
|
|
__u32 link_info_len = sizeof(struct bpf_link_info);
|
|
|
|
struct bpf_link_info link_info = {};
|
bpftool: Add struct_ops support
This patch adds struct_ops support to the bpftool.
To recap a bit on the recent bpf_struct_ops feature on the kernel side:
It currently supports "struct tcp_congestion_ops" to be implemented
in bpf. At a high level, bpf_struct_ops is struct_ops map populated
with a number of bpf progs. bpf_struct_ops currently supports the
"struct tcp_congestion_ops". However, the bpf_struct_ops design is
generic enough that other kernel struct ops can be supported in
the future.
Although struct_ops is map+progs at a high lever, there are differences
in details. For example,
1) After registering a struct_ops, the struct_ops is held by the kernel
subsystem (e.g. tcp-cc). Thus, there is no need to pin a
struct_ops map or its progs in order to keep them around.
2) To iterate all struct_ops in a system, it iterates all maps
in type BPF_MAP_TYPE_STRUCT_OPS. BPF_MAP_TYPE_STRUCT_OPS is
the current usual filter. In the future, it may need to
filter by other struct_ops specific properties. e.g. filter by
tcp_congestion_ops or other kernel subsystem ops in the future.
3) struct_ops requires the running kernel having BTF info. That allows
more flexibility in handling other kernel structs. e.g. it can
always dump the latest bpf_map_info.
4) Also, "struct_ops" command is not intended to repeat all features
already provided by "map" or "prog". For example, if there really
is a need to pin the struct_ops map, the user can use the "map" cmd
to do that.
While the first attempt was to reuse parts from map/prog.c, it ended up
not a lot to share. The only obvious item is the map_parse_fds() but
that still requires modifications to accommodate struct_ops map specific
filtering (for the immediate and the future needs). Together with the
earlier mentioned differences, it is better to part away from map/prog.c.
The initial set of subcmds are, register, unregister, show, and dump.
For register, it registers all struct_ops maps that can be found in an
obj file. Option can be added in the future to specify a particular
struct_ops map. Also, the common bpf_tcp_cc is stateless (e.g.
bpf_cubic.c and bpf_dctcp.c). The "reuse map" feature is not
implemented in this patch and it can be considered later also.
For other subcmds, please see the man doc for details.
A sample output of dump:
[root@arch-fb-vm1 bpf]# bpftool struct_ops dump name cubic
[{
"bpf_map_info": {
"type": 26,
"id": 64,
"key_size": 4,
"value_size": 256,
"max_entries": 1,
"map_flags": 0,
"name": "cubic",
"ifindex": 0,
"btf_vmlinux_value_type_id": 18452,
"netns_dev": 0,
"netns_ino": 0,
"btf_id": 52,
"btf_key_type_id": 0,
"btf_value_type_id": 0
}
},{
"bpf_struct_ops_tcp_congestion_ops": {
"refcnt": {
"refs": {
"counter": 1
}
},
"state": "BPF_STRUCT_OPS_STATE_INUSE",
"data": {
"list": {
"next": 0,
"prev": 0
},
"key": 0,
"flags": 0,
"init": "void (struct sock *) bictcp_init/prog_id:138",
"release": "void (struct sock *) 0",
"ssthresh": "u32 (struct sock *) bictcp_recalc_ssthresh/prog_id:141",
"cong_avoid": "void (struct sock *, u32, u32) bictcp_cong_avoid/prog_id:140",
"set_state": "void (struct sock *, u8) bictcp_state/prog_id:142",
"cwnd_event": "void (struct sock *, enum tcp_ca_event) bictcp_cwnd_event/prog_id:139",
"in_ack_event": "void (struct sock *, u32) 0",
"undo_cwnd": "u32 (struct sock *) tcp_reno_undo_cwnd/prog_id:144",
"pkts_acked": "void (struct sock *, const struct ack_sample *) bictcp_acked/prog_id:143",
"min_tso_segs": "u32 (struct sock *) 0",
"sndbuf_expand": "u32 (struct sock *) 0",
"cong_control": "void (struct sock *, const struct rate_sample *) 0",
"get_info": "size_t (struct sock *, u32, int *, union tcp_cc_info *) 0",
"name": "bpf_cubic",
"owner": 0
}
}
}
]
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20200318171656.129650-1-kafai@fb.com
2020-03-19 01:16:56 +08:00
|
|
|
struct bpf_map_info info = {};
|
|
|
|
__u32 info_len = sizeof(info);
|
|
|
|
int nr_errs = 0, nr_maps = 0;
|
2023-04-20 08:28:21 +08:00
|
|
|
const char *linkdir = NULL;
|
bpftool: Add struct_ops support
This patch adds struct_ops support to the bpftool.
To recap a bit on the recent bpf_struct_ops feature on the kernel side:
It currently supports "struct tcp_congestion_ops" to be implemented
in bpf. At a high level, bpf_struct_ops is struct_ops map populated
with a number of bpf progs. bpf_struct_ops currently supports the
"struct tcp_congestion_ops". However, the bpf_struct_ops design is
generic enough that other kernel struct ops can be supported in
the future.
Although struct_ops is map+progs at a high lever, there are differences
in details. For example,
1) After registering a struct_ops, the struct_ops is held by the kernel
subsystem (e.g. tcp-cc). Thus, there is no need to pin a
struct_ops map or its progs in order to keep them around.
2) To iterate all struct_ops in a system, it iterates all maps
in type BPF_MAP_TYPE_STRUCT_OPS. BPF_MAP_TYPE_STRUCT_OPS is
the current usual filter. In the future, it may need to
filter by other struct_ops specific properties. e.g. filter by
tcp_congestion_ops or other kernel subsystem ops in the future.
3) struct_ops requires the running kernel having BTF info. That allows
more flexibility in handling other kernel structs. e.g. it can
always dump the latest bpf_map_info.
4) Also, "struct_ops" command is not intended to repeat all features
already provided by "map" or "prog". For example, if there really
is a need to pin the struct_ops map, the user can use the "map" cmd
to do that.
While the first attempt was to reuse parts from map/prog.c, it ended up
not a lot to share. The only obvious item is the map_parse_fds() but
that still requires modifications to accommodate struct_ops map specific
filtering (for the immediate and the future needs). Together with the
earlier mentioned differences, it is better to part away from map/prog.c.
The initial set of subcmds are, register, unregister, show, and dump.
For register, it registers all struct_ops maps that can be found in an
obj file. Option can be added in the future to specify a particular
struct_ops map. Also, the common bpf_tcp_cc is stateless (e.g.
bpf_cubic.c and bpf_dctcp.c). The "reuse map" feature is not
implemented in this patch and it can be considered later also.
For other subcmds, please see the man doc for details.
A sample output of dump:
[root@arch-fb-vm1 bpf]# bpftool struct_ops dump name cubic
[{
"bpf_map_info": {
"type": 26,
"id": 64,
"key_size": 4,
"value_size": 256,
"max_entries": 1,
"map_flags": 0,
"name": "cubic",
"ifindex": 0,
"btf_vmlinux_value_type_id": 18452,
"netns_dev": 0,
"netns_ino": 0,
"btf_id": 52,
"btf_key_type_id": 0,
"btf_value_type_id": 0
}
},{
"bpf_struct_ops_tcp_congestion_ops": {
"refcnt": {
"refs": {
"counter": 1
}
},
"state": "BPF_STRUCT_OPS_STATE_INUSE",
"data": {
"list": {
"next": 0,
"prev": 0
},
"key": 0,
"flags": 0,
"init": "void (struct sock *) bictcp_init/prog_id:138",
"release": "void (struct sock *) 0",
"ssthresh": "u32 (struct sock *) bictcp_recalc_ssthresh/prog_id:141",
"cong_avoid": "void (struct sock *, u32, u32) bictcp_cong_avoid/prog_id:140",
"set_state": "void (struct sock *, u8) bictcp_state/prog_id:142",
"cwnd_event": "void (struct sock *, enum tcp_ca_event) bictcp_cwnd_event/prog_id:139",
"in_ack_event": "void (struct sock *, u32) 0",
"undo_cwnd": "u32 (struct sock *) tcp_reno_undo_cwnd/prog_id:144",
"pkts_acked": "void (struct sock *, const struct ack_sample *) bictcp_acked/prog_id:143",
"min_tso_segs": "u32 (struct sock *) 0",
"sndbuf_expand": "u32 (struct sock *) 0",
"cong_control": "void (struct sock *, const struct rate_sample *) 0",
"get_info": "size_t (struct sock *, u32, int *, union tcp_cc_info *) 0",
"name": "bpf_cubic",
"owner": 0
}
}
}
]
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20200318171656.129650-1-kafai@fb.com
2020-03-19 01:16:56 +08:00
|
|
|
struct bpf_object *obj;
|
|
|
|
struct bpf_link *link;
|
|
|
|
struct bpf_map *map;
|
|
|
|
const char *file;
|
|
|
|
|
2023-04-20 08:28:21 +08:00
|
|
|
if (argc != 1 && argc != 2)
|
bpftool: Add struct_ops support
This patch adds struct_ops support to the bpftool.
To recap a bit on the recent bpf_struct_ops feature on the kernel side:
It currently supports "struct tcp_congestion_ops" to be implemented
in bpf. At a high level, bpf_struct_ops is struct_ops map populated
with a number of bpf progs. bpf_struct_ops currently supports the
"struct tcp_congestion_ops". However, the bpf_struct_ops design is
generic enough that other kernel struct ops can be supported in
the future.
Although struct_ops is map+progs at a high lever, there are differences
in details. For example,
1) After registering a struct_ops, the struct_ops is held by the kernel
subsystem (e.g. tcp-cc). Thus, there is no need to pin a
struct_ops map or its progs in order to keep them around.
2) To iterate all struct_ops in a system, it iterates all maps
in type BPF_MAP_TYPE_STRUCT_OPS. BPF_MAP_TYPE_STRUCT_OPS is
the current usual filter. In the future, it may need to
filter by other struct_ops specific properties. e.g. filter by
tcp_congestion_ops or other kernel subsystem ops in the future.
3) struct_ops requires the running kernel having BTF info. That allows
more flexibility in handling other kernel structs. e.g. it can
always dump the latest bpf_map_info.
4) Also, "struct_ops" command is not intended to repeat all features
already provided by "map" or "prog". For example, if there really
is a need to pin the struct_ops map, the user can use the "map" cmd
to do that.
While the first attempt was to reuse parts from map/prog.c, it ended up
not a lot to share. The only obvious item is the map_parse_fds() but
that still requires modifications to accommodate struct_ops map specific
filtering (for the immediate and the future needs). Together with the
earlier mentioned differences, it is better to part away from map/prog.c.
The initial set of subcmds are, register, unregister, show, and dump.
For register, it registers all struct_ops maps that can be found in an
obj file. Option can be added in the future to specify a particular
struct_ops map. Also, the common bpf_tcp_cc is stateless (e.g.
bpf_cubic.c and bpf_dctcp.c). The "reuse map" feature is not
implemented in this patch and it can be considered later also.
For other subcmds, please see the man doc for details.
A sample output of dump:
[root@arch-fb-vm1 bpf]# bpftool struct_ops dump name cubic
[{
"bpf_map_info": {
"type": 26,
"id": 64,
"key_size": 4,
"value_size": 256,
"max_entries": 1,
"map_flags": 0,
"name": "cubic",
"ifindex": 0,
"btf_vmlinux_value_type_id": 18452,
"netns_dev": 0,
"netns_ino": 0,
"btf_id": 52,
"btf_key_type_id": 0,
"btf_value_type_id": 0
}
},{
"bpf_struct_ops_tcp_congestion_ops": {
"refcnt": {
"refs": {
"counter": 1
}
},
"state": "BPF_STRUCT_OPS_STATE_INUSE",
"data": {
"list": {
"next": 0,
"prev": 0
},
"key": 0,
"flags": 0,
"init": "void (struct sock *) bictcp_init/prog_id:138",
"release": "void (struct sock *) 0",
"ssthresh": "u32 (struct sock *) bictcp_recalc_ssthresh/prog_id:141",
"cong_avoid": "void (struct sock *, u32, u32) bictcp_cong_avoid/prog_id:140",
"set_state": "void (struct sock *, u8) bictcp_state/prog_id:142",
"cwnd_event": "void (struct sock *, enum tcp_ca_event) bictcp_cwnd_event/prog_id:139",
"in_ack_event": "void (struct sock *, u32) 0",
"undo_cwnd": "u32 (struct sock *) tcp_reno_undo_cwnd/prog_id:144",
"pkts_acked": "void (struct sock *, const struct ack_sample *) bictcp_acked/prog_id:143",
"min_tso_segs": "u32 (struct sock *) 0",
"sndbuf_expand": "u32 (struct sock *) 0",
"cong_control": "void (struct sock *, const struct rate_sample *) 0",
"get_info": "size_t (struct sock *, u32, int *, union tcp_cc_info *) 0",
"name": "bpf_cubic",
"owner": 0
}
}
}
]
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20200318171656.129650-1-kafai@fb.com
2020-03-19 01:16:56 +08:00
|
|
|
usage();
|
|
|
|
|
|
|
|
file = GET_ARG();
|
2023-04-20 08:28:21 +08:00
|
|
|
if (argc == 1)
|
|
|
|
linkdir = GET_ARG();
|
|
|
|
|
2023-05-06 11:07:19 +08:00
|
|
|
if (linkdir && mount_bpffs_for_pin(linkdir, true)) {
|
2023-04-20 08:28:21 +08:00
|
|
|
p_err("can't mount bpffs for pinning");
|
|
|
|
return -1;
|
|
|
|
}
|
bpftool: Add struct_ops support
This patch adds struct_ops support to the bpftool.
To recap a bit on the recent bpf_struct_ops feature on the kernel side:
It currently supports "struct tcp_congestion_ops" to be implemented
in bpf. At a high level, bpf_struct_ops is struct_ops map populated
with a number of bpf progs. bpf_struct_ops currently supports the
"struct tcp_congestion_ops". However, the bpf_struct_ops design is
generic enough that other kernel struct ops can be supported in
the future.
Although struct_ops is map+progs at a high lever, there are differences
in details. For example,
1) After registering a struct_ops, the struct_ops is held by the kernel
subsystem (e.g. tcp-cc). Thus, there is no need to pin a
struct_ops map or its progs in order to keep them around.
2) To iterate all struct_ops in a system, it iterates all maps
in type BPF_MAP_TYPE_STRUCT_OPS. BPF_MAP_TYPE_STRUCT_OPS is
the current usual filter. In the future, it may need to
filter by other struct_ops specific properties. e.g. filter by
tcp_congestion_ops or other kernel subsystem ops in the future.
3) struct_ops requires the running kernel having BTF info. That allows
more flexibility in handling other kernel structs. e.g. it can
always dump the latest bpf_map_info.
4) Also, "struct_ops" command is not intended to repeat all features
already provided by "map" or "prog". For example, if there really
is a need to pin the struct_ops map, the user can use the "map" cmd
to do that.
While the first attempt was to reuse parts from map/prog.c, it ended up
not a lot to share. The only obvious item is the map_parse_fds() but
that still requires modifications to accommodate struct_ops map specific
filtering (for the immediate and the future needs). Together with the
earlier mentioned differences, it is better to part away from map/prog.c.
The initial set of subcmds are, register, unregister, show, and dump.
For register, it registers all struct_ops maps that can be found in an
obj file. Option can be added in the future to specify a particular
struct_ops map. Also, the common bpf_tcp_cc is stateless (e.g.
bpf_cubic.c and bpf_dctcp.c). The "reuse map" feature is not
implemented in this patch and it can be considered later also.
For other subcmds, please see the man doc for details.
A sample output of dump:
[root@arch-fb-vm1 bpf]# bpftool struct_ops dump name cubic
[{
"bpf_map_info": {
"type": 26,
"id": 64,
"key_size": 4,
"value_size": 256,
"max_entries": 1,
"map_flags": 0,
"name": "cubic",
"ifindex": 0,
"btf_vmlinux_value_type_id": 18452,
"netns_dev": 0,
"netns_ino": 0,
"btf_id": 52,
"btf_key_type_id": 0,
"btf_value_type_id": 0
}
},{
"bpf_struct_ops_tcp_congestion_ops": {
"refcnt": {
"refs": {
"counter": 1
}
},
"state": "BPF_STRUCT_OPS_STATE_INUSE",
"data": {
"list": {
"next": 0,
"prev": 0
},
"key": 0,
"flags": 0,
"init": "void (struct sock *) bictcp_init/prog_id:138",
"release": "void (struct sock *) 0",
"ssthresh": "u32 (struct sock *) bictcp_recalc_ssthresh/prog_id:141",
"cong_avoid": "void (struct sock *, u32, u32) bictcp_cong_avoid/prog_id:140",
"set_state": "void (struct sock *, u8) bictcp_state/prog_id:142",
"cwnd_event": "void (struct sock *, enum tcp_ca_event) bictcp_cwnd_event/prog_id:139",
"in_ack_event": "void (struct sock *, u32) 0",
"undo_cwnd": "u32 (struct sock *) tcp_reno_undo_cwnd/prog_id:144",
"pkts_acked": "void (struct sock *, const struct ack_sample *) bictcp_acked/prog_id:143",
"min_tso_segs": "u32 (struct sock *) 0",
"sndbuf_expand": "u32 (struct sock *) 0",
"cong_control": "void (struct sock *, const struct rate_sample *) 0",
"get_info": "size_t (struct sock *, u32, int *, union tcp_cc_info *) 0",
"name": "bpf_cubic",
"owner": 0
}
}
}
]
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20200318171656.129650-1-kafai@fb.com
2020-03-19 01:16:56 +08:00
|
|
|
|
2021-12-10 03:38:40 +08:00
|
|
|
if (verifier_logs)
|
|
|
|
/* log_level1 + log_level2 + stats, but not stable UAPI */
|
|
|
|
open_opts.kernel_log_level = 1 + 2 + 4;
|
|
|
|
|
|
|
|
obj = bpf_object__open_file(file, &open_opts);
|
2022-11-20 19:26:32 +08:00
|
|
|
if (!obj)
|
bpftool: Add struct_ops support
This patch adds struct_ops support to the bpftool.
To recap a bit on the recent bpf_struct_ops feature on the kernel side:
It currently supports "struct tcp_congestion_ops" to be implemented
in bpf. At a high level, bpf_struct_ops is struct_ops map populated
with a number of bpf progs. bpf_struct_ops currently supports the
"struct tcp_congestion_ops". However, the bpf_struct_ops design is
generic enough that other kernel struct ops can be supported in
the future.
Although struct_ops is map+progs at a high lever, there are differences
in details. For example,
1) After registering a struct_ops, the struct_ops is held by the kernel
subsystem (e.g. tcp-cc). Thus, there is no need to pin a
struct_ops map or its progs in order to keep them around.
2) To iterate all struct_ops in a system, it iterates all maps
in type BPF_MAP_TYPE_STRUCT_OPS. BPF_MAP_TYPE_STRUCT_OPS is
the current usual filter. In the future, it may need to
filter by other struct_ops specific properties. e.g. filter by
tcp_congestion_ops or other kernel subsystem ops in the future.
3) struct_ops requires the running kernel having BTF info. That allows
more flexibility in handling other kernel structs. e.g. it can
always dump the latest bpf_map_info.
4) Also, "struct_ops" command is not intended to repeat all features
already provided by "map" or "prog". For example, if there really
is a need to pin the struct_ops map, the user can use the "map" cmd
to do that.
While the first attempt was to reuse parts from map/prog.c, it ended up
not a lot to share. The only obvious item is the map_parse_fds() but
that still requires modifications to accommodate struct_ops map specific
filtering (for the immediate and the future needs). Together with the
earlier mentioned differences, it is better to part away from map/prog.c.
The initial set of subcmds are, register, unregister, show, and dump.
For register, it registers all struct_ops maps that can be found in an
obj file. Option can be added in the future to specify a particular
struct_ops map. Also, the common bpf_tcp_cc is stateless (e.g.
bpf_cubic.c and bpf_dctcp.c). The "reuse map" feature is not
implemented in this patch and it can be considered later also.
For other subcmds, please see the man doc for details.
A sample output of dump:
[root@arch-fb-vm1 bpf]# bpftool struct_ops dump name cubic
[{
"bpf_map_info": {
"type": 26,
"id": 64,
"key_size": 4,
"value_size": 256,
"max_entries": 1,
"map_flags": 0,
"name": "cubic",
"ifindex": 0,
"btf_vmlinux_value_type_id": 18452,
"netns_dev": 0,
"netns_ino": 0,
"btf_id": 52,
"btf_key_type_id": 0,
"btf_value_type_id": 0
}
},{
"bpf_struct_ops_tcp_congestion_ops": {
"refcnt": {
"refs": {
"counter": 1
}
},
"state": "BPF_STRUCT_OPS_STATE_INUSE",
"data": {
"list": {
"next": 0,
"prev": 0
},
"key": 0,
"flags": 0,
"init": "void (struct sock *) bictcp_init/prog_id:138",
"release": "void (struct sock *) 0",
"ssthresh": "u32 (struct sock *) bictcp_recalc_ssthresh/prog_id:141",
"cong_avoid": "void (struct sock *, u32, u32) bictcp_cong_avoid/prog_id:140",
"set_state": "void (struct sock *, u8) bictcp_state/prog_id:142",
"cwnd_event": "void (struct sock *, enum tcp_ca_event) bictcp_cwnd_event/prog_id:139",
"in_ack_event": "void (struct sock *, u32) 0",
"undo_cwnd": "u32 (struct sock *) tcp_reno_undo_cwnd/prog_id:144",
"pkts_acked": "void (struct sock *, const struct ack_sample *) bictcp_acked/prog_id:143",
"min_tso_segs": "u32 (struct sock *) 0",
"sndbuf_expand": "u32 (struct sock *) 0",
"cong_control": "void (struct sock *, const struct rate_sample *) 0",
"get_info": "size_t (struct sock *, u32, int *, union tcp_cc_info *) 0",
"name": "bpf_cubic",
"owner": 0
}
}
}
]
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20200318171656.129650-1-kafai@fb.com
2020-03-19 01:16:56 +08:00
|
|
|
return -1;
|
|
|
|
|
Revert "bpftool: Use libbpf 1.0 API mode instead of RLIMIT_MEMLOCK"
This reverts commit a777e18f1bcd32528ff5dfd10a6629b655b05eb8.
In commit a777e18f1bcd ("bpftool: Use libbpf 1.0 API mode instead of
RLIMIT_MEMLOCK"), we removed the rlimit bump in bpftool, because the
kernel has switched to memcg-based memory accounting. Thanks to the
LIBBPF_STRICT_AUTO_RLIMIT_MEMLOCK, we attempted to keep compatibility
with other systems and ask libbpf to raise the limit for us if
necessary.
How do we know if memcg-based accounting is supported? There is a probe
in libbpf to check this. But this probe currently relies on the
availability of a given BPF helper, bpf_ktime_get_coarse_ns(), which
landed in the same kernel version as the memory accounting change. This
works in the generic case, but it may fail, for example, if the helper
function has been backported to an older kernel. This has been observed
for Google Cloud's Container-Optimized OS (COS), where the helper is
available but rlimit is still in use. The probe succeeds, the rlimit is
not raised, and probing features with bpftool, for example, fails.
A patch was submitted [0] to update this probe in libbpf, based on what
the cilium/ebpf Go library does [1]. It would lower the soft rlimit to
0, attempt to load a BPF object, and reset the rlimit. But it may induce
some hard-to-debug flakiness if another process starts, or the current
application is killed, while the rlimit is reduced, and the approach was
discarded.
As a workaround to ensure that the rlimit bump does not depend on the
availability of a given helper, we restore the unconditional rlimit bump
in bpftool for now.
[0] https://lore.kernel.org/bpf/20220609143614.97837-1-quentin@isovalent.com/
[1] https://github.com/cilium/ebpf/blob/v0.9.0/rlimit/rlimit.go#L39
Signed-off-by: Quentin Monnet <quentin@isovalent.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Cc: Yafang Shao <laoar.shao@gmail.com>
Cc: Stanislav Fomichev <sdf@google.com>
Link: https://lore.kernel.org/bpf/20220610112648.29695-2-quentin@isovalent.com
2022-06-10 19:26:47 +08:00
|
|
|
set_max_rlimit();
|
|
|
|
|
2021-12-10 03:38:40 +08:00
|
|
|
if (bpf_object__load(obj)) {
|
bpftool: Add struct_ops support
This patch adds struct_ops support to the bpftool.
To recap a bit on the recent bpf_struct_ops feature on the kernel side:
It currently supports "struct tcp_congestion_ops" to be implemented
in bpf. At a high level, bpf_struct_ops is struct_ops map populated
with a number of bpf progs. bpf_struct_ops currently supports the
"struct tcp_congestion_ops". However, the bpf_struct_ops design is
generic enough that other kernel struct ops can be supported in
the future.
Although struct_ops is map+progs at a high lever, there are differences
in details. For example,
1) After registering a struct_ops, the struct_ops is held by the kernel
subsystem (e.g. tcp-cc). Thus, there is no need to pin a
struct_ops map or its progs in order to keep them around.
2) To iterate all struct_ops in a system, it iterates all maps
in type BPF_MAP_TYPE_STRUCT_OPS. BPF_MAP_TYPE_STRUCT_OPS is
the current usual filter. In the future, it may need to
filter by other struct_ops specific properties. e.g. filter by
tcp_congestion_ops or other kernel subsystem ops in the future.
3) struct_ops requires the running kernel having BTF info. That allows
more flexibility in handling other kernel structs. e.g. it can
always dump the latest bpf_map_info.
4) Also, "struct_ops" command is not intended to repeat all features
already provided by "map" or "prog". For example, if there really
is a need to pin the struct_ops map, the user can use the "map" cmd
to do that.
While the first attempt was to reuse parts from map/prog.c, it ended up
not a lot to share. The only obvious item is the map_parse_fds() but
that still requires modifications to accommodate struct_ops map specific
filtering (for the immediate and the future needs). Together with the
earlier mentioned differences, it is better to part away from map/prog.c.
The initial set of subcmds are, register, unregister, show, and dump.
For register, it registers all struct_ops maps that can be found in an
obj file. Option can be added in the future to specify a particular
struct_ops map. Also, the common bpf_tcp_cc is stateless (e.g.
bpf_cubic.c and bpf_dctcp.c). The "reuse map" feature is not
implemented in this patch and it can be considered later also.
For other subcmds, please see the man doc for details.
A sample output of dump:
[root@arch-fb-vm1 bpf]# bpftool struct_ops dump name cubic
[{
"bpf_map_info": {
"type": 26,
"id": 64,
"key_size": 4,
"value_size": 256,
"max_entries": 1,
"map_flags": 0,
"name": "cubic",
"ifindex": 0,
"btf_vmlinux_value_type_id": 18452,
"netns_dev": 0,
"netns_ino": 0,
"btf_id": 52,
"btf_key_type_id": 0,
"btf_value_type_id": 0
}
},{
"bpf_struct_ops_tcp_congestion_ops": {
"refcnt": {
"refs": {
"counter": 1
}
},
"state": "BPF_STRUCT_OPS_STATE_INUSE",
"data": {
"list": {
"next": 0,
"prev": 0
},
"key": 0,
"flags": 0,
"init": "void (struct sock *) bictcp_init/prog_id:138",
"release": "void (struct sock *) 0",
"ssthresh": "u32 (struct sock *) bictcp_recalc_ssthresh/prog_id:141",
"cong_avoid": "void (struct sock *, u32, u32) bictcp_cong_avoid/prog_id:140",
"set_state": "void (struct sock *, u8) bictcp_state/prog_id:142",
"cwnd_event": "void (struct sock *, enum tcp_ca_event) bictcp_cwnd_event/prog_id:139",
"in_ack_event": "void (struct sock *, u32) 0",
"undo_cwnd": "u32 (struct sock *) tcp_reno_undo_cwnd/prog_id:144",
"pkts_acked": "void (struct sock *, const struct ack_sample *) bictcp_acked/prog_id:143",
"min_tso_segs": "u32 (struct sock *) 0",
"sndbuf_expand": "u32 (struct sock *) 0",
"cong_control": "void (struct sock *, const struct rate_sample *) 0",
"get_info": "size_t (struct sock *, u32, int *, union tcp_cc_info *) 0",
"name": "bpf_cubic",
"owner": 0
}
}
}
]
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20200318171656.129650-1-kafai@fb.com
2020-03-19 01:16:56 +08:00
|
|
|
bpf_object__close(obj);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
bpf_object__for_each_map(map, obj) {
|
2022-01-08 08:42:15 +08:00
|
|
|
if (bpf_map__type(map) != BPF_MAP_TYPE_STRUCT_OPS)
|
bpftool: Add struct_ops support
This patch adds struct_ops support to the bpftool.
To recap a bit on the recent bpf_struct_ops feature on the kernel side:
It currently supports "struct tcp_congestion_ops" to be implemented
in bpf. At a high level, bpf_struct_ops is struct_ops map populated
with a number of bpf progs. bpf_struct_ops currently supports the
"struct tcp_congestion_ops". However, the bpf_struct_ops design is
generic enough that other kernel struct ops can be supported in
the future.
Although struct_ops is map+progs at a high lever, there are differences
in details. For example,
1) After registering a struct_ops, the struct_ops is held by the kernel
subsystem (e.g. tcp-cc). Thus, there is no need to pin a
struct_ops map or its progs in order to keep them around.
2) To iterate all struct_ops in a system, it iterates all maps
in type BPF_MAP_TYPE_STRUCT_OPS. BPF_MAP_TYPE_STRUCT_OPS is
the current usual filter. In the future, it may need to
filter by other struct_ops specific properties. e.g. filter by
tcp_congestion_ops or other kernel subsystem ops in the future.
3) struct_ops requires the running kernel having BTF info. That allows
more flexibility in handling other kernel structs. e.g. it can
always dump the latest bpf_map_info.
4) Also, "struct_ops" command is not intended to repeat all features
already provided by "map" or "prog". For example, if there really
is a need to pin the struct_ops map, the user can use the "map" cmd
to do that.
While the first attempt was to reuse parts from map/prog.c, it ended up
not a lot to share. The only obvious item is the map_parse_fds() but
that still requires modifications to accommodate struct_ops map specific
filtering (for the immediate and the future needs). Together with the
earlier mentioned differences, it is better to part away from map/prog.c.
The initial set of subcmds are, register, unregister, show, and dump.
For register, it registers all struct_ops maps that can be found in an
obj file. Option can be added in the future to specify a particular
struct_ops map. Also, the common bpf_tcp_cc is stateless (e.g.
bpf_cubic.c and bpf_dctcp.c). The "reuse map" feature is not
implemented in this patch and it can be considered later also.
For other subcmds, please see the man doc for details.
A sample output of dump:
[root@arch-fb-vm1 bpf]# bpftool struct_ops dump name cubic
[{
"bpf_map_info": {
"type": 26,
"id": 64,
"key_size": 4,
"value_size": 256,
"max_entries": 1,
"map_flags": 0,
"name": "cubic",
"ifindex": 0,
"btf_vmlinux_value_type_id": 18452,
"netns_dev": 0,
"netns_ino": 0,
"btf_id": 52,
"btf_key_type_id": 0,
"btf_value_type_id": 0
}
},{
"bpf_struct_ops_tcp_congestion_ops": {
"refcnt": {
"refs": {
"counter": 1
}
},
"state": "BPF_STRUCT_OPS_STATE_INUSE",
"data": {
"list": {
"next": 0,
"prev": 0
},
"key": 0,
"flags": 0,
"init": "void (struct sock *) bictcp_init/prog_id:138",
"release": "void (struct sock *) 0",
"ssthresh": "u32 (struct sock *) bictcp_recalc_ssthresh/prog_id:141",
"cong_avoid": "void (struct sock *, u32, u32) bictcp_cong_avoid/prog_id:140",
"set_state": "void (struct sock *, u8) bictcp_state/prog_id:142",
"cwnd_event": "void (struct sock *, enum tcp_ca_event) bictcp_cwnd_event/prog_id:139",
"in_ack_event": "void (struct sock *, u32) 0",
"undo_cwnd": "u32 (struct sock *) tcp_reno_undo_cwnd/prog_id:144",
"pkts_acked": "void (struct sock *, const struct ack_sample *) bictcp_acked/prog_id:143",
"min_tso_segs": "u32 (struct sock *) 0",
"sndbuf_expand": "u32 (struct sock *) 0",
"cong_control": "void (struct sock *, const struct rate_sample *) 0",
"get_info": "size_t (struct sock *, u32, int *, union tcp_cc_info *) 0",
"name": "bpf_cubic",
"owner": 0
}
}
}
]
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20200318171656.129650-1-kafai@fb.com
2020-03-19 01:16:56 +08:00
|
|
|
continue;
|
|
|
|
|
|
|
|
link = bpf_map__attach_struct_ops(map);
|
2022-11-20 19:26:18 +08:00
|
|
|
if (!link) {
|
bpftool: Add struct_ops support
This patch adds struct_ops support to the bpftool.
To recap a bit on the recent bpf_struct_ops feature on the kernel side:
It currently supports "struct tcp_congestion_ops" to be implemented
in bpf. At a high level, bpf_struct_ops is struct_ops map populated
with a number of bpf progs. bpf_struct_ops currently supports the
"struct tcp_congestion_ops". However, the bpf_struct_ops design is
generic enough that other kernel struct ops can be supported in
the future.
Although struct_ops is map+progs at a high lever, there are differences
in details. For example,
1) After registering a struct_ops, the struct_ops is held by the kernel
subsystem (e.g. tcp-cc). Thus, there is no need to pin a
struct_ops map or its progs in order to keep them around.
2) To iterate all struct_ops in a system, it iterates all maps
in type BPF_MAP_TYPE_STRUCT_OPS. BPF_MAP_TYPE_STRUCT_OPS is
the current usual filter. In the future, it may need to
filter by other struct_ops specific properties. e.g. filter by
tcp_congestion_ops or other kernel subsystem ops in the future.
3) struct_ops requires the running kernel having BTF info. That allows
more flexibility in handling other kernel structs. e.g. it can
always dump the latest bpf_map_info.
4) Also, "struct_ops" command is not intended to repeat all features
already provided by "map" or "prog". For example, if there really
is a need to pin the struct_ops map, the user can use the "map" cmd
to do that.
While the first attempt was to reuse parts from map/prog.c, it ended up
not a lot to share. The only obvious item is the map_parse_fds() but
that still requires modifications to accommodate struct_ops map specific
filtering (for the immediate and the future needs). Together with the
earlier mentioned differences, it is better to part away from map/prog.c.
The initial set of subcmds are, register, unregister, show, and dump.
For register, it registers all struct_ops maps that can be found in an
obj file. Option can be added in the future to specify a particular
struct_ops map. Also, the common bpf_tcp_cc is stateless (e.g.
bpf_cubic.c and bpf_dctcp.c). The "reuse map" feature is not
implemented in this patch and it can be considered later also.
For other subcmds, please see the man doc for details.
A sample output of dump:
[root@arch-fb-vm1 bpf]# bpftool struct_ops dump name cubic
[{
"bpf_map_info": {
"type": 26,
"id": 64,
"key_size": 4,
"value_size": 256,
"max_entries": 1,
"map_flags": 0,
"name": "cubic",
"ifindex": 0,
"btf_vmlinux_value_type_id": 18452,
"netns_dev": 0,
"netns_ino": 0,
"btf_id": 52,
"btf_key_type_id": 0,
"btf_value_type_id": 0
}
},{
"bpf_struct_ops_tcp_congestion_ops": {
"refcnt": {
"refs": {
"counter": 1
}
},
"state": "BPF_STRUCT_OPS_STATE_INUSE",
"data": {
"list": {
"next": 0,
"prev": 0
},
"key": 0,
"flags": 0,
"init": "void (struct sock *) bictcp_init/prog_id:138",
"release": "void (struct sock *) 0",
"ssthresh": "u32 (struct sock *) bictcp_recalc_ssthresh/prog_id:141",
"cong_avoid": "void (struct sock *, u32, u32) bictcp_cong_avoid/prog_id:140",
"set_state": "void (struct sock *, u8) bictcp_state/prog_id:142",
"cwnd_event": "void (struct sock *, enum tcp_ca_event) bictcp_cwnd_event/prog_id:139",
"in_ack_event": "void (struct sock *, u32) 0",
"undo_cwnd": "u32 (struct sock *) tcp_reno_undo_cwnd/prog_id:144",
"pkts_acked": "void (struct sock *, const struct ack_sample *) bictcp_acked/prog_id:143",
"min_tso_segs": "u32 (struct sock *) 0",
"sndbuf_expand": "u32 (struct sock *) 0",
"cong_control": "void (struct sock *, const struct rate_sample *) 0",
"get_info": "size_t (struct sock *, u32, int *, union tcp_cc_info *) 0",
"name": "bpf_cubic",
"owner": 0
}
}
}
]
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20200318171656.129650-1-kafai@fb.com
2020-03-19 01:16:56 +08:00
|
|
|
p_err("can't register struct_ops %s: %s",
|
2022-11-20 19:26:18 +08:00
|
|
|
bpf_map__name(map), strerror(errno));
|
bpftool: Add struct_ops support
This patch adds struct_ops support to the bpftool.
To recap a bit on the recent bpf_struct_ops feature on the kernel side:
It currently supports "struct tcp_congestion_ops" to be implemented
in bpf. At a high level, bpf_struct_ops is struct_ops map populated
with a number of bpf progs. bpf_struct_ops currently supports the
"struct tcp_congestion_ops". However, the bpf_struct_ops design is
generic enough that other kernel struct ops can be supported in
the future.
Although struct_ops is map+progs at a high lever, there are differences
in details. For example,
1) After registering a struct_ops, the struct_ops is held by the kernel
subsystem (e.g. tcp-cc). Thus, there is no need to pin a
struct_ops map or its progs in order to keep them around.
2) To iterate all struct_ops in a system, it iterates all maps
in type BPF_MAP_TYPE_STRUCT_OPS. BPF_MAP_TYPE_STRUCT_OPS is
the current usual filter. In the future, it may need to
filter by other struct_ops specific properties. e.g. filter by
tcp_congestion_ops or other kernel subsystem ops in the future.
3) struct_ops requires the running kernel having BTF info. That allows
more flexibility in handling other kernel structs. e.g. it can
always dump the latest bpf_map_info.
4) Also, "struct_ops" command is not intended to repeat all features
already provided by "map" or "prog". For example, if there really
is a need to pin the struct_ops map, the user can use the "map" cmd
to do that.
While the first attempt was to reuse parts from map/prog.c, it ended up
not a lot to share. The only obvious item is the map_parse_fds() but
that still requires modifications to accommodate struct_ops map specific
filtering (for the immediate and the future needs). Together with the
earlier mentioned differences, it is better to part away from map/prog.c.
The initial set of subcmds are, register, unregister, show, and dump.
For register, it registers all struct_ops maps that can be found in an
obj file. Option can be added in the future to specify a particular
struct_ops map. Also, the common bpf_tcp_cc is stateless (e.g.
bpf_cubic.c and bpf_dctcp.c). The "reuse map" feature is not
implemented in this patch and it can be considered later also.
For other subcmds, please see the man doc for details.
A sample output of dump:
[root@arch-fb-vm1 bpf]# bpftool struct_ops dump name cubic
[{
"bpf_map_info": {
"type": 26,
"id": 64,
"key_size": 4,
"value_size": 256,
"max_entries": 1,
"map_flags": 0,
"name": "cubic",
"ifindex": 0,
"btf_vmlinux_value_type_id": 18452,
"netns_dev": 0,
"netns_ino": 0,
"btf_id": 52,
"btf_key_type_id": 0,
"btf_value_type_id": 0
}
},{
"bpf_struct_ops_tcp_congestion_ops": {
"refcnt": {
"refs": {
"counter": 1
}
},
"state": "BPF_STRUCT_OPS_STATE_INUSE",
"data": {
"list": {
"next": 0,
"prev": 0
},
"key": 0,
"flags": 0,
"init": "void (struct sock *) bictcp_init/prog_id:138",
"release": "void (struct sock *) 0",
"ssthresh": "u32 (struct sock *) bictcp_recalc_ssthresh/prog_id:141",
"cong_avoid": "void (struct sock *, u32, u32) bictcp_cong_avoid/prog_id:140",
"set_state": "void (struct sock *, u8) bictcp_state/prog_id:142",
"cwnd_event": "void (struct sock *, enum tcp_ca_event) bictcp_cwnd_event/prog_id:139",
"in_ack_event": "void (struct sock *, u32) 0",
"undo_cwnd": "u32 (struct sock *) tcp_reno_undo_cwnd/prog_id:144",
"pkts_acked": "void (struct sock *, const struct ack_sample *) bictcp_acked/prog_id:143",
"min_tso_segs": "u32 (struct sock *) 0",
"sndbuf_expand": "u32 (struct sock *) 0",
"cong_control": "void (struct sock *, const struct rate_sample *) 0",
"get_info": "size_t (struct sock *, u32, int *, union tcp_cc_info *) 0",
"name": "bpf_cubic",
"owner": 0
}
}
}
]
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20200318171656.129650-1-kafai@fb.com
2020-03-19 01:16:56 +08:00
|
|
|
nr_errs++;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
nr_maps++;
|
|
|
|
|
2023-04-20 08:28:21 +08:00
|
|
|
if (bpf_map_get_info_by_fd(bpf_map__fd(map), &info,
|
|
|
|
&info_len)) {
|
bpftool: Add struct_ops support
This patch adds struct_ops support to the bpftool.
To recap a bit on the recent bpf_struct_ops feature on the kernel side:
It currently supports "struct tcp_congestion_ops" to be implemented
in bpf. At a high level, bpf_struct_ops is struct_ops map populated
with a number of bpf progs. bpf_struct_ops currently supports the
"struct tcp_congestion_ops". However, the bpf_struct_ops design is
generic enough that other kernel struct ops can be supported in
the future.
Although struct_ops is map+progs at a high lever, there are differences
in details. For example,
1) After registering a struct_ops, the struct_ops is held by the kernel
subsystem (e.g. tcp-cc). Thus, there is no need to pin a
struct_ops map or its progs in order to keep them around.
2) To iterate all struct_ops in a system, it iterates all maps
in type BPF_MAP_TYPE_STRUCT_OPS. BPF_MAP_TYPE_STRUCT_OPS is
the current usual filter. In the future, it may need to
filter by other struct_ops specific properties. e.g. filter by
tcp_congestion_ops or other kernel subsystem ops in the future.
3) struct_ops requires the running kernel having BTF info. That allows
more flexibility in handling other kernel structs. e.g. it can
always dump the latest bpf_map_info.
4) Also, "struct_ops" command is not intended to repeat all features
already provided by "map" or "prog". For example, if there really
is a need to pin the struct_ops map, the user can use the "map" cmd
to do that.
While the first attempt was to reuse parts from map/prog.c, it ended up
not a lot to share. The only obvious item is the map_parse_fds() but
that still requires modifications to accommodate struct_ops map specific
filtering (for the immediate and the future needs). Together with the
earlier mentioned differences, it is better to part away from map/prog.c.
The initial set of subcmds are, register, unregister, show, and dump.
For register, it registers all struct_ops maps that can be found in an
obj file. Option can be added in the future to specify a particular
struct_ops map. Also, the common bpf_tcp_cc is stateless (e.g.
bpf_cubic.c and bpf_dctcp.c). The "reuse map" feature is not
implemented in this patch and it can be considered later also.
For other subcmds, please see the man doc for details.
A sample output of dump:
[root@arch-fb-vm1 bpf]# bpftool struct_ops dump name cubic
[{
"bpf_map_info": {
"type": 26,
"id": 64,
"key_size": 4,
"value_size": 256,
"max_entries": 1,
"map_flags": 0,
"name": "cubic",
"ifindex": 0,
"btf_vmlinux_value_type_id": 18452,
"netns_dev": 0,
"netns_ino": 0,
"btf_id": 52,
"btf_key_type_id": 0,
"btf_value_type_id": 0
}
},{
"bpf_struct_ops_tcp_congestion_ops": {
"refcnt": {
"refs": {
"counter": 1
}
},
"state": "BPF_STRUCT_OPS_STATE_INUSE",
"data": {
"list": {
"next": 0,
"prev": 0
},
"key": 0,
"flags": 0,
"init": "void (struct sock *) bictcp_init/prog_id:138",
"release": "void (struct sock *) 0",
"ssthresh": "u32 (struct sock *) bictcp_recalc_ssthresh/prog_id:141",
"cong_avoid": "void (struct sock *, u32, u32) bictcp_cong_avoid/prog_id:140",
"set_state": "void (struct sock *, u8) bictcp_state/prog_id:142",
"cwnd_event": "void (struct sock *, enum tcp_ca_event) bictcp_cwnd_event/prog_id:139",
"in_ack_event": "void (struct sock *, u32) 0",
"undo_cwnd": "u32 (struct sock *) tcp_reno_undo_cwnd/prog_id:144",
"pkts_acked": "void (struct sock *, const struct ack_sample *) bictcp_acked/prog_id:143",
"min_tso_segs": "u32 (struct sock *) 0",
"sndbuf_expand": "u32 (struct sock *) 0",
"cong_control": "void (struct sock *, const struct rate_sample *) 0",
"get_info": "size_t (struct sock *, u32, int *, union tcp_cc_info *) 0",
"name": "bpf_cubic",
"owner": 0
}
}
}
]
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20200318171656.129650-1-kafai@fb.com
2020-03-19 01:16:56 +08:00
|
|
|
/* Not p_err. The struct_ops was attached
|
|
|
|
* successfully.
|
|
|
|
*/
|
|
|
|
p_info("Registered %s but can't find id: %s",
|
|
|
|
bpf_map__name(map), strerror(errno));
|
2023-04-20 08:28:21 +08:00
|
|
|
goto clean_link;
|
|
|
|
}
|
|
|
|
if (!(bpf_map__map_flags(map) & BPF_F_LINK)) {
|
|
|
|
p_info("Registered %s %s id %u",
|
|
|
|
get_kern_struct_ops_name(&info),
|
|
|
|
info.name,
|
|
|
|
info.id);
|
|
|
|
goto clean_link;
|
|
|
|
}
|
|
|
|
if (bpf_link_get_info_by_fd(bpf_link__fd(link),
|
|
|
|
&link_info,
|
|
|
|
&link_info_len)) {
|
|
|
|
p_err("Registered %s but can't find link id: %s",
|
|
|
|
bpf_map__name(map), strerror(errno));
|
|
|
|
nr_errs++;
|
|
|
|
goto clean_link;
|
|
|
|
}
|
|
|
|
if (linkdir && pin_link(link, linkdir, info.name)) {
|
|
|
|
p_err("can't pin link %u for %s: %s",
|
|
|
|
link_info.id, info.name,
|
|
|
|
strerror(errno));
|
|
|
|
nr_errs++;
|
|
|
|
goto clean_link;
|
|
|
|
}
|
|
|
|
p_info("Registered %s %s map id %u link id %u",
|
|
|
|
get_kern_struct_ops_name(&info),
|
|
|
|
info.name, info.id, link_info.id);
|
|
|
|
|
|
|
|
clean_link:
|
|
|
|
bpf_link__disconnect(link);
|
|
|
|
bpf_link__destroy(link);
|
bpftool: Add struct_ops support
This patch adds struct_ops support to the bpftool.
To recap a bit on the recent bpf_struct_ops feature on the kernel side:
It currently supports "struct tcp_congestion_ops" to be implemented
in bpf. At a high level, bpf_struct_ops is struct_ops map populated
with a number of bpf progs. bpf_struct_ops currently supports the
"struct tcp_congestion_ops". However, the bpf_struct_ops design is
generic enough that other kernel struct ops can be supported in
the future.
Although struct_ops is map+progs at a high lever, there are differences
in details. For example,
1) After registering a struct_ops, the struct_ops is held by the kernel
subsystem (e.g. tcp-cc). Thus, there is no need to pin a
struct_ops map or its progs in order to keep them around.
2) To iterate all struct_ops in a system, it iterates all maps
in type BPF_MAP_TYPE_STRUCT_OPS. BPF_MAP_TYPE_STRUCT_OPS is
the current usual filter. In the future, it may need to
filter by other struct_ops specific properties. e.g. filter by
tcp_congestion_ops or other kernel subsystem ops in the future.
3) struct_ops requires the running kernel having BTF info. That allows
more flexibility in handling other kernel structs. e.g. it can
always dump the latest bpf_map_info.
4) Also, "struct_ops" command is not intended to repeat all features
already provided by "map" or "prog". For example, if there really
is a need to pin the struct_ops map, the user can use the "map" cmd
to do that.
While the first attempt was to reuse parts from map/prog.c, it ended up
not a lot to share. The only obvious item is the map_parse_fds() but
that still requires modifications to accommodate struct_ops map specific
filtering (for the immediate and the future needs). Together with the
earlier mentioned differences, it is better to part away from map/prog.c.
The initial set of subcmds are, register, unregister, show, and dump.
For register, it registers all struct_ops maps that can be found in an
obj file. Option can be added in the future to specify a particular
struct_ops map. Also, the common bpf_tcp_cc is stateless (e.g.
bpf_cubic.c and bpf_dctcp.c). The "reuse map" feature is not
implemented in this patch and it can be considered later also.
For other subcmds, please see the man doc for details.
A sample output of dump:
[root@arch-fb-vm1 bpf]# bpftool struct_ops dump name cubic
[{
"bpf_map_info": {
"type": 26,
"id": 64,
"key_size": 4,
"value_size": 256,
"max_entries": 1,
"map_flags": 0,
"name": "cubic",
"ifindex": 0,
"btf_vmlinux_value_type_id": 18452,
"netns_dev": 0,
"netns_ino": 0,
"btf_id": 52,
"btf_key_type_id": 0,
"btf_value_type_id": 0
}
},{
"bpf_struct_ops_tcp_congestion_ops": {
"refcnt": {
"refs": {
"counter": 1
}
},
"state": "BPF_STRUCT_OPS_STATE_INUSE",
"data": {
"list": {
"next": 0,
"prev": 0
},
"key": 0,
"flags": 0,
"init": "void (struct sock *) bictcp_init/prog_id:138",
"release": "void (struct sock *) 0",
"ssthresh": "u32 (struct sock *) bictcp_recalc_ssthresh/prog_id:141",
"cong_avoid": "void (struct sock *, u32, u32) bictcp_cong_avoid/prog_id:140",
"set_state": "void (struct sock *, u8) bictcp_state/prog_id:142",
"cwnd_event": "void (struct sock *, enum tcp_ca_event) bictcp_cwnd_event/prog_id:139",
"in_ack_event": "void (struct sock *, u32) 0",
"undo_cwnd": "u32 (struct sock *) tcp_reno_undo_cwnd/prog_id:144",
"pkts_acked": "void (struct sock *, const struct ack_sample *) bictcp_acked/prog_id:143",
"min_tso_segs": "u32 (struct sock *) 0",
"sndbuf_expand": "u32 (struct sock *) 0",
"cong_control": "void (struct sock *, const struct rate_sample *) 0",
"get_info": "size_t (struct sock *, u32, int *, union tcp_cc_info *) 0",
"name": "bpf_cubic",
"owner": 0
}
}
}
]
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20200318171656.129650-1-kafai@fb.com
2020-03-19 01:16:56 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
bpf_object__close(obj);
|
|
|
|
|
|
|
|
if (nr_errs)
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
if (!nr_maps) {
|
|
|
|
p_err("no struct_ops found in %s", file);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (json_output)
|
|
|
|
jsonw_null(json_wtr);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int do_help(int argc, char **argv)
|
|
|
|
{
|
|
|
|
if (json_output) {
|
|
|
|
jsonw_null(json_wtr);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
fprintf(stderr,
|
2020-05-23 09:07:51 +08:00
|
|
|
"Usage: %1$s %2$s { show | list } [STRUCT_OPS_MAP]\n"
|
|
|
|
" %1$s %2$s dump [STRUCT_OPS_MAP]\n"
|
2023-04-20 08:28:21 +08:00
|
|
|
" %1$s %2$s register OBJ [LINK_DIR]\n"
|
2020-05-23 09:07:51 +08:00
|
|
|
" %1$s %2$s unregister STRUCT_OPS_MAP\n"
|
|
|
|
" %1$s %2$s help\n"
|
bpftool: Add struct_ops support
This patch adds struct_ops support to the bpftool.
To recap a bit on the recent bpf_struct_ops feature on the kernel side:
It currently supports "struct tcp_congestion_ops" to be implemented
in bpf. At a high level, bpf_struct_ops is struct_ops map populated
with a number of bpf progs. bpf_struct_ops currently supports the
"struct tcp_congestion_ops". However, the bpf_struct_ops design is
generic enough that other kernel struct ops can be supported in
the future.
Although struct_ops is map+progs at a high lever, there are differences
in details. For example,
1) After registering a struct_ops, the struct_ops is held by the kernel
subsystem (e.g. tcp-cc). Thus, there is no need to pin a
struct_ops map or its progs in order to keep them around.
2) To iterate all struct_ops in a system, it iterates all maps
in type BPF_MAP_TYPE_STRUCT_OPS. BPF_MAP_TYPE_STRUCT_OPS is
the current usual filter. In the future, it may need to
filter by other struct_ops specific properties. e.g. filter by
tcp_congestion_ops or other kernel subsystem ops in the future.
3) struct_ops requires the running kernel having BTF info. That allows
more flexibility in handling other kernel structs. e.g. it can
always dump the latest bpf_map_info.
4) Also, "struct_ops" command is not intended to repeat all features
already provided by "map" or "prog". For example, if there really
is a need to pin the struct_ops map, the user can use the "map" cmd
to do that.
While the first attempt was to reuse parts from map/prog.c, it ended up
not a lot to share. The only obvious item is the map_parse_fds() but
that still requires modifications to accommodate struct_ops map specific
filtering (for the immediate and the future needs). Together with the
earlier mentioned differences, it is better to part away from map/prog.c.
The initial set of subcmds are, register, unregister, show, and dump.
For register, it registers all struct_ops maps that can be found in an
obj file. Option can be added in the future to specify a particular
struct_ops map. Also, the common bpf_tcp_cc is stateless (e.g.
bpf_cubic.c and bpf_dctcp.c). The "reuse map" feature is not
implemented in this patch and it can be considered later also.
For other subcmds, please see the man doc for details.
A sample output of dump:
[root@arch-fb-vm1 bpf]# bpftool struct_ops dump name cubic
[{
"bpf_map_info": {
"type": 26,
"id": 64,
"key_size": 4,
"value_size": 256,
"max_entries": 1,
"map_flags": 0,
"name": "cubic",
"ifindex": 0,
"btf_vmlinux_value_type_id": 18452,
"netns_dev": 0,
"netns_ino": 0,
"btf_id": 52,
"btf_key_type_id": 0,
"btf_value_type_id": 0
}
},{
"bpf_struct_ops_tcp_congestion_ops": {
"refcnt": {
"refs": {
"counter": 1
}
},
"state": "BPF_STRUCT_OPS_STATE_INUSE",
"data": {
"list": {
"next": 0,
"prev": 0
},
"key": 0,
"flags": 0,
"init": "void (struct sock *) bictcp_init/prog_id:138",
"release": "void (struct sock *) 0",
"ssthresh": "u32 (struct sock *) bictcp_recalc_ssthresh/prog_id:141",
"cong_avoid": "void (struct sock *, u32, u32) bictcp_cong_avoid/prog_id:140",
"set_state": "void (struct sock *, u8) bictcp_state/prog_id:142",
"cwnd_event": "void (struct sock *, enum tcp_ca_event) bictcp_cwnd_event/prog_id:139",
"in_ack_event": "void (struct sock *, u32) 0",
"undo_cwnd": "u32 (struct sock *) tcp_reno_undo_cwnd/prog_id:144",
"pkts_acked": "void (struct sock *, const struct ack_sample *) bictcp_acked/prog_id:143",
"min_tso_segs": "u32 (struct sock *) 0",
"sndbuf_expand": "u32 (struct sock *) 0",
"cong_control": "void (struct sock *, const struct rate_sample *) 0",
"get_info": "size_t (struct sock *, u32, int *, union tcp_cc_info *) 0",
"name": "bpf_cubic",
"owner": 0
}
}
}
]
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20200318171656.129650-1-kafai@fb.com
2020-03-19 01:16:56 +08:00
|
|
|
"\n"
|
2020-05-23 09:07:51 +08:00
|
|
|
" STRUCT_OPS_MAP := [ id STRUCT_OPS_MAP_ID | name STRUCT_OPS_MAP_NAME ]\n"
|
2021-07-31 05:54:32 +08:00
|
|
|
" " HELP_SPEC_OPTIONS " }\n"
|
2020-05-23 09:07:51 +08:00
|
|
|
"",
|
bpftool: Add struct_ops support
This patch adds struct_ops support to the bpftool.
To recap a bit on the recent bpf_struct_ops feature on the kernel side:
It currently supports "struct tcp_congestion_ops" to be implemented
in bpf. At a high level, bpf_struct_ops is struct_ops map populated
with a number of bpf progs. bpf_struct_ops currently supports the
"struct tcp_congestion_ops". However, the bpf_struct_ops design is
generic enough that other kernel struct ops can be supported in
the future.
Although struct_ops is map+progs at a high lever, there are differences
in details. For example,
1) After registering a struct_ops, the struct_ops is held by the kernel
subsystem (e.g. tcp-cc). Thus, there is no need to pin a
struct_ops map or its progs in order to keep them around.
2) To iterate all struct_ops in a system, it iterates all maps
in type BPF_MAP_TYPE_STRUCT_OPS. BPF_MAP_TYPE_STRUCT_OPS is
the current usual filter. In the future, it may need to
filter by other struct_ops specific properties. e.g. filter by
tcp_congestion_ops or other kernel subsystem ops in the future.
3) struct_ops requires the running kernel having BTF info. That allows
more flexibility in handling other kernel structs. e.g. it can
always dump the latest bpf_map_info.
4) Also, "struct_ops" command is not intended to repeat all features
already provided by "map" or "prog". For example, if there really
is a need to pin the struct_ops map, the user can use the "map" cmd
to do that.
While the first attempt was to reuse parts from map/prog.c, it ended up
not a lot to share. The only obvious item is the map_parse_fds() but
that still requires modifications to accommodate struct_ops map specific
filtering (for the immediate and the future needs). Together with the
earlier mentioned differences, it is better to part away from map/prog.c.
The initial set of subcmds are, register, unregister, show, and dump.
For register, it registers all struct_ops maps that can be found in an
obj file. Option can be added in the future to specify a particular
struct_ops map. Also, the common bpf_tcp_cc is stateless (e.g.
bpf_cubic.c and bpf_dctcp.c). The "reuse map" feature is not
implemented in this patch and it can be considered later also.
For other subcmds, please see the man doc for details.
A sample output of dump:
[root@arch-fb-vm1 bpf]# bpftool struct_ops dump name cubic
[{
"bpf_map_info": {
"type": 26,
"id": 64,
"key_size": 4,
"value_size": 256,
"max_entries": 1,
"map_flags": 0,
"name": "cubic",
"ifindex": 0,
"btf_vmlinux_value_type_id": 18452,
"netns_dev": 0,
"netns_ino": 0,
"btf_id": 52,
"btf_key_type_id": 0,
"btf_value_type_id": 0
}
},{
"bpf_struct_ops_tcp_congestion_ops": {
"refcnt": {
"refs": {
"counter": 1
}
},
"state": "BPF_STRUCT_OPS_STATE_INUSE",
"data": {
"list": {
"next": 0,
"prev": 0
},
"key": 0,
"flags": 0,
"init": "void (struct sock *) bictcp_init/prog_id:138",
"release": "void (struct sock *) 0",
"ssthresh": "u32 (struct sock *) bictcp_recalc_ssthresh/prog_id:141",
"cong_avoid": "void (struct sock *, u32, u32) bictcp_cong_avoid/prog_id:140",
"set_state": "void (struct sock *, u8) bictcp_state/prog_id:142",
"cwnd_event": "void (struct sock *, enum tcp_ca_event) bictcp_cwnd_event/prog_id:139",
"in_ack_event": "void (struct sock *, u32) 0",
"undo_cwnd": "u32 (struct sock *) tcp_reno_undo_cwnd/prog_id:144",
"pkts_acked": "void (struct sock *, const struct ack_sample *) bictcp_acked/prog_id:143",
"min_tso_segs": "u32 (struct sock *) 0",
"sndbuf_expand": "u32 (struct sock *) 0",
"cong_control": "void (struct sock *, const struct rate_sample *) 0",
"get_info": "size_t (struct sock *, u32, int *, union tcp_cc_info *) 0",
"name": "bpf_cubic",
"owner": 0
}
}
}
]
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20200318171656.129650-1-kafai@fb.com
2020-03-19 01:16:56 +08:00
|
|
|
bin_name, argv[-2]);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static const struct cmd cmds[] = {
|
|
|
|
{ "show", do_show },
|
|
|
|
{ "list", do_show },
|
|
|
|
{ "register", do_register },
|
|
|
|
{ "unregister", do_unregister },
|
|
|
|
{ "dump", do_dump },
|
|
|
|
{ "help", do_help },
|
|
|
|
{ 0 }
|
|
|
|
};
|
|
|
|
|
|
|
|
int do_struct_ops(int argc, char **argv)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
|
|
|
|
err = cmd_select(cmds, argc, argv, do_help);
|
|
|
|
|
2022-11-20 19:26:32 +08:00
|
|
|
btf__free(btf_vmlinux);
|
2020-04-10 10:06:12 +08:00
|
|
|
|
bpftool: Add struct_ops support
This patch adds struct_ops support to the bpftool.
To recap a bit on the recent bpf_struct_ops feature on the kernel side:
It currently supports "struct tcp_congestion_ops" to be implemented
in bpf. At a high level, bpf_struct_ops is struct_ops map populated
with a number of bpf progs. bpf_struct_ops currently supports the
"struct tcp_congestion_ops". However, the bpf_struct_ops design is
generic enough that other kernel struct ops can be supported in
the future.
Although struct_ops is map+progs at a high lever, there are differences
in details. For example,
1) After registering a struct_ops, the struct_ops is held by the kernel
subsystem (e.g. tcp-cc). Thus, there is no need to pin a
struct_ops map or its progs in order to keep them around.
2) To iterate all struct_ops in a system, it iterates all maps
in type BPF_MAP_TYPE_STRUCT_OPS. BPF_MAP_TYPE_STRUCT_OPS is
the current usual filter. In the future, it may need to
filter by other struct_ops specific properties. e.g. filter by
tcp_congestion_ops or other kernel subsystem ops in the future.
3) struct_ops requires the running kernel having BTF info. That allows
more flexibility in handling other kernel structs. e.g. it can
always dump the latest bpf_map_info.
4) Also, "struct_ops" command is not intended to repeat all features
already provided by "map" or "prog". For example, if there really
is a need to pin the struct_ops map, the user can use the "map" cmd
to do that.
While the first attempt was to reuse parts from map/prog.c, it ended up
not a lot to share. The only obvious item is the map_parse_fds() but
that still requires modifications to accommodate struct_ops map specific
filtering (for the immediate and the future needs). Together with the
earlier mentioned differences, it is better to part away from map/prog.c.
The initial set of subcmds are, register, unregister, show, and dump.
For register, it registers all struct_ops maps that can be found in an
obj file. Option can be added in the future to specify a particular
struct_ops map. Also, the common bpf_tcp_cc is stateless (e.g.
bpf_cubic.c and bpf_dctcp.c). The "reuse map" feature is not
implemented in this patch and it can be considered later also.
For other subcmds, please see the man doc for details.
A sample output of dump:
[root@arch-fb-vm1 bpf]# bpftool struct_ops dump name cubic
[{
"bpf_map_info": {
"type": 26,
"id": 64,
"key_size": 4,
"value_size": 256,
"max_entries": 1,
"map_flags": 0,
"name": "cubic",
"ifindex": 0,
"btf_vmlinux_value_type_id": 18452,
"netns_dev": 0,
"netns_ino": 0,
"btf_id": 52,
"btf_key_type_id": 0,
"btf_value_type_id": 0
}
},{
"bpf_struct_ops_tcp_congestion_ops": {
"refcnt": {
"refs": {
"counter": 1
}
},
"state": "BPF_STRUCT_OPS_STATE_INUSE",
"data": {
"list": {
"next": 0,
"prev": 0
},
"key": 0,
"flags": 0,
"init": "void (struct sock *) bictcp_init/prog_id:138",
"release": "void (struct sock *) 0",
"ssthresh": "u32 (struct sock *) bictcp_recalc_ssthresh/prog_id:141",
"cong_avoid": "void (struct sock *, u32, u32) bictcp_cong_avoid/prog_id:140",
"set_state": "void (struct sock *, u8) bictcp_state/prog_id:142",
"cwnd_event": "void (struct sock *, enum tcp_ca_event) bictcp_cwnd_event/prog_id:139",
"in_ack_event": "void (struct sock *, u32) 0",
"undo_cwnd": "u32 (struct sock *) tcp_reno_undo_cwnd/prog_id:144",
"pkts_acked": "void (struct sock *, const struct ack_sample *) bictcp_acked/prog_id:143",
"min_tso_segs": "u32 (struct sock *) 0",
"sndbuf_expand": "u32 (struct sock *) 0",
"cong_control": "void (struct sock *, const struct rate_sample *) 0",
"get_info": "size_t (struct sock *, u32, int *, union tcp_cc_info *) 0",
"name": "bpf_cubic",
"owner": 0
}
}
}
]
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Quentin Monnet <quentin@isovalent.com>
Link: https://lore.kernel.org/bpf/20200318171656.129650-1-kafai@fb.com
2020-03-19 01:16:56 +08:00
|
|
|
return err;
|
|
|
|
}
|