This patch adds support for VAR and DATASEC in btf_dedup(). VAR/DATASEC
are never deduplicated, but they need to be processed anyway as types
they refer to might need to be remapped due to deduplication and
compaction.
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Yonghong Song <yhs@fb.com>
Cc: Alexei Starovoitov <ast@fb.com>
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
1. There is one commit ("libbpf: fix btf_dedup equivalence check
handling of different kinds"), which was originally committed into bpf
tree and manually synced into libbpf (it was a critical bug fix). Now
that bpf and bpf-next trees were merged, this causes duplicate commits in
libbpf. To address this, check is added to see if libbpf repo already has
a commit with the same subject and allow to skip it.
2. Just checking by subject is not good enough, as we currently have few
different commits with the same subject "bpf: Sync bpf.h to tools/".
To address this, script now will ask user whether to skip or cherry-pick
commit. This allows for human to double-check the commit and make
informed devision.
3. There is a commit ('bfb35c27c65f bpf: fix whitespace for ENCAP_L2
defines in bpf.h'), which modifies both include/uapi/linux/bpf.h and
tools/include/uapi/linux/bpf.h in the same commit. This break the
sync script, as it doesn't cherry-pick commits that hasn't modified
libbpf files. So at the time that commit bfb35c27c65f is applied, linux
repo is missing some preceding changes to include/uapi/linux/bpf.h. They
would normally be ignored later during repo filtering, but to do
cherry-pick they are necessary. It's hard to fix it automatically, so
script will wait for user to resolve the issue and continue.
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
btf_dedup_is_equiv() used to compare btf_type->info fields, before doing
kind-specific equivalence check. This comparsion implicitly verified
that candidate and canonical types are of the same kind. With enum fwd
resolution logic this check couldn't be done generically anymore, as for
enums info contains vlen, which differs between enum fwd and
fully-defined enum, so this check was subsumed by kind-specific
equivalence checks.
This change caused btf_dedup_is_equiv() to let through VOID vs other
types check to reach switch, which was never meant to be handing VOID
kind, as VOID kind is always pre-resolved to itself and is only
equivalent to itself, which is checked early in btf_dedup_is_equiv().
This change adds back BTF kind equality check in place of more generic
btf_type->info check, still defering further kind-specific checks to
a per-kind switch.
Fixes: 9768095ba97c ("btf: resolve enum fwds in btf_dedup")
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Merges are common. `git cherry-pick` can't cherry-pick merge commit and
requires specifying which of the parents should be cherry-picked, so
it's a pain. But merges are also commonly empty, so it's safe to skip them.
This change goes through all merges touching libbpf files, checks that
they don't have any conflict resolution diffs. If there are any, abort,
if not, proceed as usual, ignoring merges.
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
GCC and clang support enum forward declarations as an extension. Such
forward-declared enums will be represented as normal BTF_KIND_ENUM types with
vlen=0. This patch adds ability to resolve such enums to their corresponding
fully defined enums. This helps to avoid duplicated BTF type graphs which only
differ by some types referencing forward-declared enum vs full enum.
One such example in kernel is enum irqchip_irq_state, defined in
include/linux/interrupt.h and forward-declared in include/linux/irq.h. This
causes entire struct task_struct and all referenced types to be duplicated in
btf_dedup output. This patch eliminates such duplication cases.
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
This patch splits and cleans up error handling logic for loading BTF data.
Previously, if BTF data was parsed successfully, but failed to load into
kernel, we'd report nonsensical error code, instead of error returned from
btf__load(). Now btf__new() and btf__load() are handled separately with proper
cleanup and warning reporting.
Fixes: d29d87f7e612 ("btf: separate btf creation and loading")
Reported-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
With xsk.{o,h} changes we started including Linux UAPI's if_xdp.h
header. We need to sync it along the other UAPI headers.
Also updated README to reflect this.
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
This patch makes sure we build AF_XDP-related code as part of libbpf. This
also required copying few uapi/linux headers and adding few used definitions
in include headers.
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
When checking available canonical candidates for struct/union algorithm
utilizes btf_dedup_is_equiv to determine if candidate is suitable. This
check is not enough when candidate is corresponding FWD for that
struct/union, because according to equivalence logic they are
equivalent. When it so happens that FWD and STRUCT/UNION end in hashing
to the same bucket, it's possible to create remapping loop from FWD to
STRUCT and STRUCT to same FWD, which will cause btf_dedup() to loop
forever.
This patch fixes the issue by additionally checking that type and
canonical candidate are strictly equal (utilizing btf_equal_struct).
Fixes: d5caef5b5655 ("btf: add BTF types deduplication algorithm")
Reported-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Acked-by: Song Liu <songliubraving@fb.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Default size of dedup table (16k) is good enough for most binaries, even
typical vmlinux images. But there are cases of binaries with huge amount
of BTF types (e.g., allyesconfig variants of kernel), which benefit from
having bigger dedup table size to lower amount of unnecessary hash
collisions. Tools like pahole, thus, can tune this parameter to reach
optimal performance.
This change also serves double purpose of allowing tests to force hash
collisions to test some corner cases, used in follow up patch.
Signed-off-by: Andrii Nakryiko <andriin@fb.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Andrii Nakryiko