[PATCH 0/4] selftests/bpf: A few edge case fixes

[PATCH 0/4] selftests/bpf: A few edge case fixes

[Ricardo B. Marlière]

Hi,

Please consider merging the following fixes, found by means of targeting
BPF selftests through openQA [1].

[1]: e.g. https://openqa.opensuse.org/tests/6087703

1/4 ("libbpf: Search /lib64 and /lib in resolve_full_path()")
- Hit by testing an older SLES root filesystem.

2/4 ("selftests/bpf: Fix lsm_bdev dev_t encoding mismatch")
- https://openqa.opensuse.org/tests/6087703/logfile?filename=test_progs.tap.txt#line-2706

3/4 ("selftests/bpf: Install resolve_btfids.test.o.BTF where the test expects it")
- https://openqa.opensuse.org/tests/6087703/logfile?filename=test_progs.tap.txt#line-3332

4/4 ("selftests/bpf: Install bpftool where test_progs expects to find it")
- https://openqa.opensuse.org/tests/6087703/logfile?filename=test_progs.tap.txt#line-806
- https://openqa.opensuse.org/tests/6087703#step/test_bpftool_build_sh/1
- https://openqa.opensuse.org/tests/6087703#step/test_doc_build_sh/1

The build process is done at runtime with BPF_STRICT_BUILD=0:
https://openqa.opensuse.org/tests/6087703/logfile?filename=serial_terminal.txt#line-1937

Assisted-by: Claude Sonnet 5
Signed-off-by: Ricardo B. Marlière <rbm@suse.com>
---
Ricardo B. Marlière (4):
      selftests/bpf: Install bpftool where test_progs expects to find it
      selftests/bpf: Install resolve_btfids.test.o.BTF where the test expects it
      selftests/bpf: Fix lsm_bdev dev_t encoding mismatch
      libbpf: Search /lib64 and /lib in resolve_full_path()

 tools/lib/bpf/libbpf.c                            | 3 ++-
 tools/testing/selftests/bpf/Makefile              | 5 ++++-
 tools/testing/selftests/bpf/prog_tests/lsm_bdev.c | 3 ++-
 3 files changed, 8 insertions(+), 3 deletions(-)
---
base-commit: a635d6748234582ea287c5ffeae28b9b23f91c7e
change-id: 20260710-selftests-bpf_fixes-06e9a2eb3c67

Best regards,
--  
Ricardo B. Marlière <rbm@suse.com>

[PATCH 1/4] selftests/bpf: Install bpftool where test_progs expects to find it

[Ricardo B. Marlière]

bpftool_helpers.c looks for bpftool at "tools/sbin/bpftool" or
"../tools/sbin/bpftool" relative to the running test_progs binary, matching
the build directory layout where bpftool sits at
$(OUTPUT)/tools/sbin/bpftool.

INSTALL_RULE installs bpftool through the generic TEST_GEN_PROGS_EXTENDED
mechanism, which flattens it straight into $(INSTALL_PATH) and loses the
tools/sbin/ prefix. After "make install", every flavor of test_progs,
including the default one, fails to locate bpftool, and bpftool_maps_access
and bpftool_metadata error out with "Failed to detect bpftool path", even
though bpftool itself built and installed fine.

Explicitly install bpftool under tools/sbin/ in the installed tree
so its relative location matches what detect_bpftool_path() already
expects, restoring the pre-install layout instead of introducing a
new one.

Fixes: f21fae577446 ("selftests/bpf: Add a few helpers for bpftool testing")
Signed-off-by: Ricardo B. Marlière <rbm@suse.com>
---
 tools/testing/selftests/bpf/Makefile | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/tools/testing/selftests/bpf/Makefile b/tools/testing/selftests/bpf/Makefile
index b642ee489ea6..47911c9b4977 100644
--- a/tools/testing/selftests/bpf/Makefile
+++ b/tools/testing/selftests/bpf/Makefile
@@ -1055,6 +1055,8 @@ endif
 DEFAULT_INSTALL_RULE := $(INSTALL_RULE)
 override define INSTALL_RULE
 	$(DEFAULT_INSTALL_RULE)
+	@mkdir -p $(INSTALL_PATH)/tools/sbin
+	@rsync -a $(if $(PERMISSIVE),--ignore-missing-args) $(TRUNNER_BPFTOOL) $(INSTALL_PATH)/tools/sbin/
 	@for DIR in $(TEST_INST_SUBDIRS); do				  \
 		mkdir -p $(INSTALL_PATH)/$$DIR;				  \
 		rsync -a $(if $(PERMISSIVE),--ignore-missing-args)	  \

-- 
2.54.0

Re: [PATCH 1/4] selftests/bpf: Install bpftool where test_progs expects to find it

[Alexis Lothoré]

Hi,

On Sat Jul 11, 2026 at 12:09 AM CEST, Ricardo B. Marlière wrote:
> bpftool_helpers.c looks for bpftool at "tools/sbin/bpftool" or
> "../tools/sbin/bpftool" relative to the running test_progs binary, matching
> the build directory layout where bpftool sits at
> $(OUTPUT)/tools/sbin/bpftool.
>
> INSTALL_RULE installs bpftool through the generic TEST_GEN_PROGS_EXTENDED
> mechanism, which flattens it straight into $(INSTALL_PATH) and loses the
> tools/sbin/ prefix. After "make install", every flavor of test_progs,
> including the default one, fails to locate bpftool, and bpftool_maps_access
> and bpftool_metadata error out with "Failed to detect bpftool path", even
> though bpftool itself built and installed fine.
>
> Explicitly install bpftool under tools/sbin/ in the installed tree
> so its relative location matches what detect_bpftool_path() already
> expects, restoring the pre-install layout instead of introducing a
> new one.
>
> Fixes: f21fae577446 ("selftests/bpf: Add a few helpers for bpftool testing")
> Signed-off-by: Ricardo B. Marlière <rbm@suse.com>

I did not face the issue because I am running the selftests in-tree, and
the CI seems to do the same, but for users using selftests after running
`make install`, I guess the change makes sense.

That's not the first fix aiming to improve bpftool
detection/installation (eg the BPFTOOL env variable, see [1]), so it
makes me wonder if we shouldn't just make bpftool part of
TRUNNER_EXTRA_FILES for all flavours of test_progs, similarly to other
binaries needed during tests (eg: urandom_read), instead of just adding
some bpftool-specific details in this generic install rule, and just
make the test runners assume that the needed binary is in the same
directory.

Alexis

[1] https://lore.kernel.org/all/20260223191118.655185-2-ihor.solodrai@linux.dev/

> ---
>  tools/testing/selftests/bpf/Makefile | 2 ++
>  1 file changed, 2 insertions(+)
>
> diff --git a/tools/testing/selftests/bpf/Makefile b/tools/testing/selftests/bpf/Makefile
> index b642ee489ea6..47911c9b4977 100644
> --- a/tools/testing/selftests/bpf/Makefile
> +++ b/tools/testing/selftests/bpf/Makefile
> @@ -1055,6 +1055,8 @@ endif
>  DEFAULT_INSTALL_RULE := $(INSTALL_RULE)
>  override define INSTALL_RULE
>  	$(DEFAULT_INSTALL_RULE)
> +	@mkdir -p $(INSTALL_PATH)/tools/sbin
> +	@rsync -a $(if $(PERMISSIVE),--ignore-missing-args) $(TRUNNER_BPFTOOL) $(INSTALL_PATH)/tools/sbin/
>  	@for DIR in $(TEST_INST_SUBDIRS); do				  \
>  		mkdir -p $(INSTALL_PATH)/$$DIR;				  \
>  		rsync -a $(if $(PERMISSIVE),--ignore-missing-args)	  \




-- 
Alexis Lothoré, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com

[PATCH 2/4] selftests/bpf: Install resolve_btfids.test.o.BTF where the test expects it

[Ricardo B. Marlière]

prog_tests/resolve_btfids.c looks for "resolve_btfids.test.o.BTF" relative
to the running test_progs binary. resolve_btfids always writes
<obj.path>.BTF next to the object it processes, so building
resolve_btfids.test.o also produces resolve_btfids.test.o.BTF in the build
output directory as a side effect.

INSTALL_RULE never carries it over: it is not part of TEST_FILES,
TEST_GEN_FILES or TRUNNER_EXTRA_FILES, and it is not copied by the *.bpf.o
rsync used for TEST_INST_SUBDIRS. After "make install", every flavor of
test_progs, including the default one, fails with "Failed to load BTF from
resolve_btfids.test.o.BTF", even though the file built fine and BTF ID
resolution itself works correctly.

Install it next to the rest of the per-flavor build output, using an
explicit rsync in INSTALL_RULE like the one already used for bpftool.

Fixes: 522397d05e7d ("resolve_btfids: Change in-place update with raw binary output")
Signed-off-by: Ricardo B. Marlière <rbm@suse.com>
---
 tools/testing/selftests/bpf/Makefile | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/tools/testing/selftests/bpf/Makefile b/tools/testing/selftests/bpf/Makefile
index 47911c9b4977..7260d672acf7 100644
--- a/tools/testing/selftests/bpf/Makefile
+++ b/tools/testing/selftests/bpf/Makefile
@@ -1057,10 +1057,11 @@ override define INSTALL_RULE
 	$(DEFAULT_INSTALL_RULE)
 	@mkdir -p $(INSTALL_PATH)/tools/sbin
 	@rsync -a $(if $(PERMISSIVE),--ignore-missing-args) $(TRUNNER_BPFTOOL) $(INSTALL_PATH)/tools/sbin/
+	@rsync -a $(if $(PERMISSIVE),--ignore-missing-args) $(OUTPUT)/resolve_btfids.test.o.BTF $(INSTALL_PATH)/
 	@for DIR in $(TEST_INST_SUBDIRS); do				  \
 		mkdir -p $(INSTALL_PATH)/$$DIR;				  \
 		rsync -a $(if $(PERMISSIVE),--ignore-missing-args)	  \
-			$(OUTPUT)/$$DIR/*.bpf.o				  \
+			$(OUTPUT)/$$DIR/*.bpf.o $(OUTPUT)/$$DIR/resolve_btfids.test.o.BTF \
 			$(INSTALL_PATH)/$$DIR;				  \
 	done
 endef

-- 
2.54.0

[PATCH 3/4] selftests/bpf: Fix lsm_bdev dev_t encoding mismatch

[Ricardo B. Marlière]

progs/lsm_bdev.c keys its verity_devices hashmap with the raw kernel dev_t
read straight off bdev->bd_dev, i.e. MKDEV(major, minor) = (major << 20) |
minor. prog_tests/lsm_bdev.c instead builds its lookup key with dev_key =
(__u32)st.st_rdev from stat(2), but the stat(2) syscall fills st_rdev via
the kernel's new_encode_dev(), a different bit layout: (minor & 0xff) |
(major << 8) | ((minor & ~0xff) << 12).

For any device with a non-trivial major these two values differ, so the
lookup can never find what the BPF program stored, and test_lsm_bdev()
always fails with:

  test_lsm_bdev:FAIL:map lookup unexpected error: -2 (errno 2)

Reconstruct the raw kernel dev_t from the decoded major/minor instead of
casting st_rdev directly, restoring the layout the BPF program actually
reads.

Fixes: 96f4c251a087 ("selftests/bpf: add block device management selftests")
Signed-off-by: Ricardo B. Marlière <rbm@suse.com>
---
 tools/testing/selftests/bpf/prog_tests/lsm_bdev.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/tools/testing/selftests/bpf/prog_tests/lsm_bdev.c b/tools/testing/selftests/bpf/prog_tests/lsm_bdev.c
index a970798e1173..28bc4b117f41 100644
--- a/tools/testing/selftests/bpf/prog_tests/lsm_bdev.c
+++ b/tools/testing/selftests/bpf/prog_tests/lsm_bdev.c
@@ -17,6 +17,7 @@
 #include <stdlib.h>
 #include <string.h>
 #include <sys/stat.h>
+#include <sys/sysmacros.h>
 #include <sys/types.h>
 #include <unistd.h>
 #include "lsm_bdev.skel.h"
@@ -172,7 +173,7 @@ void test_lsm_bdev(void)
 	if (!ASSERT_OK(stat(DM_DEV_PATH, &st), "stat dm dev"))
 		goto remove_dm;
 
-	dev_key = (__u32)st.st_rdev;
+	dev_key = (major(st.st_rdev) << 20) | minor(st.st_rdev);
 
 	/* Look up the device in the BPF map and verify. */
 	err = bpf_map__lookup_elem(skel->maps.verity_devices,

-- 
2.54.0

[PATCH 4/4] libbpf: Search /lib64 and /lib in resolve_full_path()

[Ricardo B. Marlière]

attach_probe/uprobe-lib and uprobe_autoattach selftests fail with "failed
to resolve full path for libc.so.6" on older non-usrmerged distros, where
libc.so.6 lives under a top-level /lib64 or /lib rather than /usr/lib64 or
/usr/lib. Add /lib64:/lib to the search paths, alongside the existing
/usr/lib64:/usr/lib and Debian multiarch entries.

Signed-off-by: Ricardo B. Marlière <rbm@suse.com>
---
 tools/lib/bpf/libbpf.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
index 1368752aa13c..1ab939dfb7f0 100644
--- a/tools/lib/bpf/libbpf.c
+++ b/tools/lib/bpf/libbpf.c
@@ -12973,13 +12973,14 @@ static const char *arch_specific_lib_paths(void)
 /* Get full path to program/shared library. */
 static int resolve_full_path(const char *file, char *result, size_t result_sz)
 {
-	const char *search_paths[3] = {};
+	const char *search_paths[4] = {};
 	int i, perm;
 
 	if (str_has_sfx(file, ".so") || strstr(file, ".so.")) {
 		search_paths[0] = getenv("LD_LIBRARY_PATH");
 		search_paths[1] = "/usr/lib64:/usr/lib";
 		search_paths[2] = arch_specific_lib_paths();
+		search_paths[3] = "/lib64:/lib";
 		perm = R_OK;
 	} else {
 		search_paths[0] = getenv("PATH");

-- 
2.54.0

Re: [PATCH 4/4] libbpf: Search /lib64 and /lib in resolve_full_path()

[bot+bpf-ci]

[-- Attachment #1: Type: text/plain, Size: 1266 bytes --]

>     libbpf: Search /lib64 and /lib in resolve_full_path()
>
>     attach_probe/uprobe-lib and uprobe_autoattach selftests fail with "failed
>     to resolve full path for libc.so.6" on older non-usrmerged distros, where
>     libc.so.6 lives under a top-level /lib64 or /lib rather than /usr/lib64 or
>     /usr/lib. Add /lib64:/lib to the search paths, alongside the existing
>     /usr/lib64:/usr/lib and Debian multiarch entries.
>
>     Signed-off-by: Ricardo B. Marlière <rbm@suse.com>

This looks like a bug fix for the shared-library search path.  Should
this include a Fixes: tag?

resolve_full_path() was introduced with a search-path list of only
LD_LIBRARY_PATH and "/usr/lib64:/usr/lib", which omitted the top-level
/lib64 and /lib directories from the start.  The later addition of the
Debian multiarch /lib/<triple> paths did not add /lib64:/lib either, so
the omission has existed since the function was created:

  Fixes: 1ce3a60e3c28 ("libbpf: auto-resolve programs/libraries when necessary for uprobes")


---
AI reviewed your patch. Please fix the bug or email reply why it's not a bug.
See: https://github.com/kernel-patches/vmtest/blob/master/ci/claude/README.md

CI run summary: https://github.com/kernel-patches/bpf/actions/runs/29127668708