Re: [PATCH bpf-next 1/2] libbpf: Support uniform BTF-defined key/value specification across all BPF maps
From: Andrii Nakryiko <hidden>
Date: 2021-09-09 04:26:20
On Sun, Sep 5, 2021 at 3:09 AM Hengqi Chen [off-list ref] wrote:
A bunch of BPF maps do not support specifying types for key and value.
s/types/BTF types/, it's a bit confusing otherwise
This is non-uniform and inconvenient[0]. Currently, libbpf uses a retry logic which removes BTF type IDs when BPF map creation failed. Instead of retrying, this commit recognizes those specialized map and removes
s/map/maps/
BTF type IDs when creating BPF map. [0] Closes: https://github.com/libbpf/libbpf/issues/355 Signed-off-by: Hengqi Chen <hengqi.chen@gmail.com> ---
For patch sets consisting of two or more patches, we ask for a cover letter, so for the next revision please provide a cover letter with an overall description of what the series is about.
quoted hunk ↗ jump to hunk
tools/lib/bpf/libbpf.c | 35 ++++++++++++++++++++--------------- 1 file changed, 20 insertions(+), 15 deletions(-)diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c index 88d8825fc6f6..7068c4d07337 100644 --- a/tools/lib/bpf/libbpf.c +++ b/tools/lib/bpf/libbpf.c@@ -4613,6 +4613,26 @@ static int bpf_object__create_map(struct bpf_object *obj, struct bpf_map *map, b create_attr.inner_map_fd = map->inner_map_fd; } + if (def->type == BPF_MAP_TYPE_PERF_EVENT_ARRAY || + def->type == BPF_MAP_TYPE_STACK_TRACE || + def->type == BPF_MAP_TYPE_CGROUP_ARRAY || + def->type == BPF_MAP_TYPE_ARRAY_OF_MAPS || + def->type == BPF_MAP_TYPE_HASH_OF_MAPS || + def->type == BPF_MAP_TYPE_DEVMAP || + def->type == BPF_MAP_TYPE_SOCKMAP || + def->type == BPF_MAP_TYPE_CPUMAP || + def->type == BPF_MAP_TYPE_XSKMAP || + def->type == BPF_MAP_TYPE_SOCKHASH || + def->type == BPF_MAP_TYPE_QUEUE || + def->type == BPF_MAP_TYPE_STACK || + def->type == BPF_MAP_TYPE_DEVMAP_HASH) { + create_attr.btf_fd = 0; + create_attr.btf_key_type_id = 0; + create_attr.btf_value_type_id = 0; + map->btf_key_type_id = 0; + map->btf_value_type_id = 0; + }
Let's do this as a more succinct switch statement. Consider also slightly rearranging entries to keep "related" map types together: - SOCKMAP + SOCKHASH - DEVMAP + DEVMAP_HASH + CPUMAP + XSKMAP Thanks!
quoted hunk ↗ jump to hunk
+ if (obj->gen_loader) { bpf_gen__map_create(obj->gen_loader, &create_attr, is_inner ? -1 : map - obj->maps); /* Pretend to have valid FD to pass various fd >= 0 checks.@@ -4622,21 +4642,6 @@ static int bpf_object__create_map(struct bpf_object *obj, struct bpf_map *map, b } else { map->fd = bpf_create_map_xattr(&create_attr); } - if (map->fd < 0 && (create_attr.btf_key_type_id || - create_attr.btf_value_type_id)) { - char *cp, errmsg[STRERR_BUFSIZE]; - - err = -errno; - cp = libbpf_strerror_r(err, errmsg, sizeof(errmsg)); - pr_warn("Error in bpf_create_map_xattr(%s):%s(%d). Retrying without BTF.\n", - map->name, cp, err); - create_attr.btf_fd = 0; - create_attr.btf_key_type_id = 0; - create_attr.btf_value_type_id = 0; - map->btf_key_type_id = 0; - map->btf_value_type_id = 0; - map->fd = bpf_create_map_xattr(&create_attr); - }
Please don't remove this fallback logic. There are multiple situations where libbpf might need to retry map creation without BTF.
err = map->fd < 0 ? -errno : 0; -- 2.25.1