From: Arnaldo Carvalho de Melo <acme@kernel.org>
To: Emil Tsalapatis <emil@etsalapatis.com>
Cc: Vineet Gupta <vineet.gupta@linux.dev>,
dwarves@vger.kernel.org, bpf@vger.kernel.org,
Andrii Nakryiko <andrii@kernel.org>,
Alan Maguire <alan.maguire@oracle.com>,
jose.marchesi@oracle.com, David Faust <david.faust@oracle.com>,
Yonghong Song <yonghong.song@linux.dev>
Subject: Re: [PAHOLE v5 5/5] tests: Add btf_type_tag ordering test
Date: Thu, 18 Jun 2026 10:25:49 -0300 [thread overview]
Message-ID: <ajPx3Y-LVhwN0jM4@x1> (raw)
In-Reply-To: <DJBWIKFFVWDX.124XRXJK8VA48@etsalapatis.com>
On Thu, Jun 18, 2026 at 12:47:30AM -0400, Emil Tsalapatis wrote:
> On Wed Jun 17, 2026 at 8:57 PM EDT, Vineet Gupta wrote:
> > From: Alan Maguire <alan.maguire@oracle.com>
> >
> > Add a test that compiles a pointer carrying two btf_type_tags
> > (__tag(outer) __tag(inner)) and checks the BTF that pahole emits to
> > ensure the tag chain is PTR -> TYPE_TAG inner -> TYPE_TAG outer ->
> > struct sample, i.e. the innermost tag is closest to the pointer.
> >
> > This exercises the type_tag ordering for both encodings: LLVM emits the
> > annotations as child DIEs in source order (reversed when building the
> > tag list), while GCC chains them via DW_AT_GNU_annotation already in BTF
> > order. The test runs with gcc and clang when they support btf_type_tag,
> > and is skipped when bpftool is unavailable.
> >
> > Signed-off-by: Alan Maguire <alan.maguire@oracle.com>
> > Signed-off-by: Vineet Gupta <vineet.gupta@linux.dev>
> > ---
>
> Reviewed-by: Emil Tsalapatis <emil@etsalapatis.com>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
- Arnaldo
> > tests/btf_type_tag_order.sh | 179 ++++++++++++++++++++++++++++++++++++
> > 1 file changed, 179 insertions(+)
> > create mode 100755 tests/btf_type_tag_order.sh
> >
> > diff --git a/tests/btf_type_tag_order.sh b/tests/btf_type_tag_order.sh
> > new file mode 100755
> > index 000000000000..09d1ac346000
> > --- /dev/null
> > +++ b/tests/btf_type_tag_order.sh
> > @@ -0,0 +1,179 @@
> > +#!/bin/bash
> > +# SPDX-License-Identifier: GPL-2.0-only
> > +
> > +# Check that pahole preserves btf_type_tag order when emitting BTF from DWARF.
> > +
> > +source test_lib.sh
> > +
> > +outdir=$(make_tmpdir)
> > +
> > +# Comment this out to save test data.
> > +trap cleanup EXIT
> > +
> > +title_log "Check BTF type tag order."
> > +
> > +GCC=${GCC:-gcc}
> > +CLANG=${CLANG:-clang}
> > +PAHOLE=${PAHOLE:-pahole}
> > +BPFTOOL=${BPFTOOL:-bpftool}
> > +
> > +if ! command -v "$BPFTOOL" > /dev/null; then
> > + info_log "skip: bpftool not available"
> > + test_skip
> > +fi
> > +
> > +compiler_has_btf_type_tag()
> > +{
> > + local compiler=$1
> > +
> > + if ! command -v "$compiler" > /dev/null; then
> > + return 1
> > + fi
> > +
> > + "$compiler" -x c -E -P - <<'EOF' 2>/dev/null | grep -qx 1
> > +#ifndef __has_attribute
> > +#define __has_attribute(x) 0
> > +#endif
> > +#if __has_attribute(btf_type_tag)
> > +1
> > +#else
> > +0
> > +#endif
> > +EOF
> > +}
> > +
> > +use_gcc=0
> > +if compiler_has_btf_type_tag "$GCC"; then
> > + use_gcc=1
> > +fi
> > +
> > +use_clang=0
> > +if compiler_has_btf_type_tag "$CLANG"; then
> > + use_clang=1
> > +fi
> > +
> > +if [ "$use_gcc" -eq 0 ] && [ "$use_clang" -eq 0 ]; then
> > + error_log "Need gcc or clang with btf_type_tag support for test $0"
> > + test_fail
> > +fi
> > +
> > +src=$(cat <<EOF
> > +#define __tag(x) __attribute__((btf_type_tag(#x)))
> > +
> > +struct sample {
> > + int value;
> > +};
> > +
> > +struct sample __tag(outer) __tag(inner) *global_ptr;
> > +
> > +EOF
> > +)
> > +
> > +check_type_tag_order()
> > +{
> > + local btf=$1
> > + local dump
> > +
> > + if ! dump=$("$BPFTOOL" btf dump file "$btf"); then
> > + return 1
> > + fi
> > +
> > + printf '%s\n' "$dump" | awk '
> > + function parse_id(line, m) {
> > + if (match(line, /^\[([0-9]+)\]/, m))
> > + return m[1]
> > + return 0
> > + }
> > + function parse_name(line, m) {
> > + if (match(line, /\047([^\047]*)\047/, m))
> > + return m[1]
> > + return ""
> > + }
> > + function parse_type(line, m) {
> > + if (match(line, /type_id=([0-9]+)/, m))
> > + return m[1]
> > + return 0
> > + }
> > + function check_ptr(ptr, id, tags, seen) {
> > + id = type[ptr]
> > + while (id != 0 && !seen[id]) {
> > + seen[id] = 1
> > + if (kind[id] == "TYPE_TAG") {
> > + tags = tags (tags == "" ? "" : " -> ") name[id]
> > + id = type[id]
> > + continue
> > + }
> > + if (kind[id] == "STRUCT" && name[id] == "sample") {
> > + if (tags == "inner -> outer")
> > + exit 0
> > + candidates = candidates (candidates == "" ? "" : ", ") tags
> > + }
> > + return
> > + }
> > + }
> > + /^\[[0-9]+\]/ {
> > + id = parse_id($0)
> > + kind[id] = $2
> > + name[id] = parse_name($0)
> > + type[id] = parse_type($0)
> > + if (kind[id] == "PTR")
> > + ptrs[++nr_ptrs] = id
> > + }
> > + END {
> > + for (i = 1; i <= nr_ptrs; i++)
> > + check_ptr(ptrs[i])
> > + if (candidates != "") {
> > + print "type tag order mismatch; expected inner -> outer, found " candidates > "/dev/stderr"
> > + exit 1
> > + }
> > + print "could not find tagged pointer to struct sample" > "/dev/stderr"
> > + exit 1
> > + }'
> > +}
> > +
> > +run_test()
> > +{
> > + local compiler=$1
> > + local tmpobj=$2
> > + local btf=$3
> > +
> > + info_log "Testing with $compiler"
> > +
> > + if ! echo "$src" | "$compiler" -g -c -x c -o "$tmpobj" - 2>/dev/null; then
> > + error_log "Could not compile type tag order test with $compiler"
> > + return 1
> > + fi
> > +
> > + if ! "$PAHOLE" --btf_features=+type_tag --btf_encode_detached="$btf" "$tmpobj" 2>/dev/null; then
> > + error_log "Could not encode BTF for $tmpobj"
> > + return 1
> > + fi
> > +
> > + if check_type_tag_order "$btf"; then
> > + info_log " passed"
> > + return 0
> > + fi
> > +
> > + error_log "BTF type tag order does not match expected order ($compiler)"
> > + return 1
> > +}
> > +
> > +failed=0
> > +
> > +if [ "$use_gcc" -eq 1 ]; then
> > + tmpobj=$(make_tmpobj)
> > + btf=${tmpobj%.o}.btf
> > + run_test "$GCC" "$tmpobj" "$btf" || failed=1
> > +fi
> > +
> > +if [ "$use_clang" -eq 1 ]; then
> > + tmpobj=$(make_tmpobj)
> > + btf=${tmpobj%.o}.btf
> > + run_test "$CLANG" "$tmpobj" "$btf" || failed=1
> > +fi
> > +
> > +if [ "$failed" -eq 0 ]; then
> > + test_pass
> > +else
> > + test_fail
> > +fi
prev parent reply other threads:[~2026-06-18 13:25 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-06-18 0:57 [PAHOLE v5 0/5] support for DW_TAG_GNU_annotation and fixes Vineet Gupta
2026-06-18 0:57 ` [PAHOLE v5 1/5] btf_loader: Handle decl tag component_idx for parameters Vineet Gupta
2026-06-18 4:46 ` Emil Tsalapatis
2026-06-18 13:27 ` Arnaldo Carvalho de Melo
2026-06-18 14:38 ` Alan Maguire
2026-06-18 17:18 ` Vineet Gupta
2026-06-18 0:57 ` [PAHOLE v5 2/5] dwarf_loader: Extract die__add_btf_type_tag() helper [NFC] Vineet Gupta
2026-06-18 0:57 ` [PAHOLE v5 3/5] dwarf_loader: Add support for DW_TAG_GNU_annotation Vineet Gupta
2026-06-18 0:57 ` [PAHOLE v5 4/5] tests: Support GCC in pfunct-btf-decl-tags test Vineet Gupta
2026-06-18 0:57 ` [PAHOLE v5 5/5] tests: Add btf_type_tag ordering test Vineet Gupta
2026-06-18 4:47 ` Emil Tsalapatis
2026-06-18 13:25 ` Arnaldo Carvalho de Melo [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=ajPx3Y-LVhwN0jM4@x1 \
--to=acme@kernel.org \
--cc=alan.maguire@oracle.com \
--cc=andrii@kernel.org \
--cc=bpf@vger.kernel.org \
--cc=david.faust@oracle.com \
--cc=dwarves@vger.kernel.org \
--cc=emil@etsalapatis.com \
--cc=jose.marchesi@oracle.com \
--cc=vineet.gupta@linux.dev \
--cc=yonghong.song@linux.dev \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox