* [RFC bpf-next 1/3] bpf: add BPF_F_COMPARE flag and compare fields to map elem UAPI
2026-06-22 7:16 [RFC bpf-next 0/3] bpf: compare-and-delete (BPF_F_COMPARE) for hash maps Gyutae Bae
@ 2026-06-22 7:16 ` Gyutae Bae
2026-06-22 7:16 ` [RFC bpf-next 2/3] bpf: implement compare-and-delete (BPF_F_COMPARE) for BPF_MAP_TYPE_HASH Gyutae Bae
2026-06-22 7:16 ` [RFC bpf-next 3/3] selftests/bpf: test BPF_F_COMPARE compare-and-delete Gyutae Bae
2 siblings, 0 replies; 6+ messages in thread
From: Gyutae Bae @ 2026-06-22 7:16 UTC (permalink / raw)
To: Alexei Starovoitov, Daniel Borkmann, Andrii Nakryiko, bpf
Cc: John Fastabend, Eduard Zingerman, Kumar Kartikeya Dwivedi,
Martin KaFai Lau, Song Liu, Yonghong Song, Jiri Olsa,
Emil Tsalapatis, Shuah Khan, linux-kselftest, linux-kernel,
Minsu Jeon, Siwan Kim, Jonghyeon Kim, Gyutae Bae
From: Gyutae Bae <gyutae.bae@navercorp.com>
Introduce the UAPI surface for a compare-and-delete primitive: a new map
flag BPF_F_COMPARE plus compare/compare_offset/compare_size fields let
a delete request name an expected value region. Behaviour is wired up in
the following patch; this adds the flag and fields (kernel + tools/ copy)
only.
Co-developed-by: Minsu Jeon <minsu.jeon@navercorp.com>
Signed-off-by: Minsu Jeon <minsu.jeon@navercorp.com>
Co-developed-by: Siwan Kim <siwan.kim@navercorp.com>
Signed-off-by: Siwan Kim <siwan.kim@navercorp.com>
Co-developed-by: Jonghyeon Kim <jong-hyeon.kim@navercorp.com>
Signed-off-by: Jonghyeon Kim <jong-hyeon.kim@navercorp.com>
Signed-off-by: Gyutae Bae <gyutae.bae@navercorp.com>
---
include/uapi/linux/bpf.h | 6 +++++-
tools/include/uapi/linux/bpf.h | 6 +++++-
2 files changed, 10 insertions(+), 2 deletions(-)
diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
index 89b36de5fdbb..4705b02fb6a4 100644
--- a/include/uapi/linux/bpf.h
+++ b/include/uapi/linux/bpf.h
@@ -1397,7 +1397,7 @@ enum bpf_addr_space_cast {
BPF_ADDR_SPACE_CAST = 1,
};
-/* flags for BPF_MAP_UPDATE_ELEM command */
+/* flags for BPF_MAP_UPDATE_ELEM / BPF_MAP_DELETE_ELEM commands */
enum {
BPF_ANY = 0, /* create new element or update existing */
BPF_NOEXIST = 1, /* create new element if it didn't exist */
@@ -1405,6 +1405,7 @@ enum {
BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */
BPF_F_CPU = 8, /* cpu flag for percpu maps, upper 32-bit of flags is a cpu number */
BPF_F_ALL_CPUS = 16, /* update value across all CPUs for percpu maps */
+ BPF_F_COMPARE = 32, /* compare-and-delete: delete elem only if value region == compare */
};
/* flags for BPF_MAP_CREATE command */
@@ -1586,6 +1587,9 @@ union bpf_attr {
__aligned_u64 next_key;
};
__u64 flags;
+ __aligned_u64 compare; /* user ptr to expected bytes (BPF_F_COMPARE) */
+ __u32 compare_offset; /* start offset within value */
+ __u32 compare_size; /* bytes to compare; 0 == whole value */
};
struct { /* struct used by BPF_MAP_*_BATCH commands */
diff --git a/tools/include/uapi/linux/bpf.h b/tools/include/uapi/linux/bpf.h
index 89b36de5fdbb..4705b02fb6a4 100644
--- a/tools/include/uapi/linux/bpf.h
+++ b/tools/include/uapi/linux/bpf.h
@@ -1397,7 +1397,7 @@ enum bpf_addr_space_cast {
BPF_ADDR_SPACE_CAST = 1,
};
-/* flags for BPF_MAP_UPDATE_ELEM command */
+/* flags for BPF_MAP_UPDATE_ELEM / BPF_MAP_DELETE_ELEM commands */
enum {
BPF_ANY = 0, /* create new element or update existing */
BPF_NOEXIST = 1, /* create new element if it didn't exist */
@@ -1405,6 +1405,7 @@ enum {
BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */
BPF_F_CPU = 8, /* cpu flag for percpu maps, upper 32-bit of flags is a cpu number */
BPF_F_ALL_CPUS = 16, /* update value across all CPUs for percpu maps */
+ BPF_F_COMPARE = 32, /* compare-and-delete: delete elem only if value region == compare */
};
/* flags for BPF_MAP_CREATE command */
@@ -1586,6 +1587,9 @@ union bpf_attr {
__aligned_u64 next_key;
};
__u64 flags;
+ __aligned_u64 compare; /* user ptr to expected bytes (BPF_F_COMPARE) */
+ __u32 compare_offset; /* start offset within value */
+ __u32 compare_size; /* bytes to compare; 0 == whole value */
};
struct { /* struct used by BPF_MAP_*_BATCH commands */
--
2.39.5 (Apple Git-154)
^ permalink raw reply related [flat|nested] 6+ messages in thread* [RFC bpf-next 2/3] bpf: implement compare-and-delete (BPF_F_COMPARE) for BPF_MAP_TYPE_HASH
2026-06-22 7:16 [RFC bpf-next 0/3] bpf: compare-and-delete (BPF_F_COMPARE) for hash maps Gyutae Bae
2026-06-22 7:16 ` [RFC bpf-next 1/3] bpf: add BPF_F_COMPARE flag and compare fields to map elem UAPI Gyutae Bae
@ 2026-06-22 7:16 ` Gyutae Bae
2026-06-22 7:58 ` sashiko-bot
2026-06-22 7:16 ` [RFC bpf-next 3/3] selftests/bpf: test BPF_F_COMPARE compare-and-delete Gyutae Bae
2 siblings, 1 reply; 6+ messages in thread
From: Gyutae Bae @ 2026-06-22 7:16 UTC (permalink / raw)
To: Alexei Starovoitov, Daniel Borkmann, Andrii Nakryiko, bpf
Cc: John Fastabend, Eduard Zingerman, Kumar Kartikeya Dwivedi,
Martin KaFai Lau, Song Liu, Yonghong Song, Jiri Olsa,
Emil Tsalapatis, Shuah Khan, linux-kselftest, linux-kernel,
Minsu Jeon, Siwan Kim, Jonghyeon Kim, Gyutae Bae
From: Gyutae Bae <gyutae.bae@navercorp.com>
map_delete_elem() now honours BPF_F_COMPARE: it copies the expected
bytes, validates the compare window against value_size, and dispatches to a
new ->map_delete_elem_cmp(). The hash-map implementation, under the bucket
lock, memcmps the [off, off+size) window of the live value and unlinks only
on a match (mismatch -EBUSY, absent -ENOENT), making the compare and the
delete atomic w.r.t. every bucket-lock holder. Only BPF_MAP_TYPE_HASH
wires up the op; every other map type lacks it and returns -EOPNOTSUPP.
Atomicity boundary: the compare is read under the bucket lock, so it is
atomic against every other bucket-lock holder (concurrent deletes,
element-replacing updates). It is NOT mutual exclusion against a BPF
program that writes the value in place via the pointer from
bpf_map_lookup_elem(), which takes no bucket lock; against such a writer
the compare may observe a torn value within the locked section. For a
meaningful result the compared region should be a synchronization variable
-- written with a single aligned store such as a monotonic revision, or
replaced via an update -- not a multi-field blob mutated in place.
It compares the raw stored value, so maps whose value carries BTF-managed
fields (bpf_spin_lock, bpf_timer, kptr, ...) are also rejected with
-EOPNOTSUPP: those bytes are sanitised on lookup, so a raw compare would
never match the caller's snapshot and could expose kernel-internal bytes.
BPF_MAP_DELETE_ELEM_LAST_FIELD now reaches compare_size, so the compare*
fields pass CHECK_ATTR even without BPF_F_COMPARE. Reject that combination
with -EINVAL: a dropped flag must not silently turn a compare-and-delete
into an unconditional delete.
Co-developed-by: Minsu Jeon <minsu.jeon@navercorp.com>
Signed-off-by: Minsu Jeon <minsu.jeon@navercorp.com>
Co-developed-by: Siwan Kim <siwan.kim@navercorp.com>
Signed-off-by: Siwan Kim <siwan.kim@navercorp.com>
Co-developed-by: Jonghyeon Kim <jong-hyeon.kim@navercorp.com>
Signed-off-by: Jonghyeon Kim <jong-hyeon.kim@navercorp.com>
Signed-off-by: Gyutae Bae <gyutae.bae@navercorp.com>
---
include/linux/bpf.h | 2 ++
kernel/bpf/hashtab.c | 39 ++++++++++++++++++++++++++++++++
kernel/bpf/syscall.c | 54 +++++++++++++++++++++++++++++++++++++++++---
3 files changed, 92 insertions(+), 3 deletions(-)
diff --git a/include/linux/bpf.h b/include/linux/bpf.h
index 7719f6528445..2e858272f7a7 100644
--- a/include/linux/bpf.h
+++ b/include/linux/bpf.h
@@ -110,6 +110,8 @@ struct bpf_map_ops {
void *(*map_lookup_elem)(struct bpf_map *map, void *key);
long (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags);
long (*map_delete_elem)(struct bpf_map *map, void *key);
+ long (*map_delete_elem_cmp)(struct bpf_map *map, void *key,
+ const void *compare, u32 off, u32 size);
long (*map_push_elem)(struct bpf_map *map, void *value, u64 flags);
long (*map_pop_elem)(struct bpf_map *map, void *value);
long (*map_peek_elem)(struct bpf_map *map, void *value);
diff --git a/kernel/bpf/hashtab.c b/kernel/bpf/hashtab.c
index 9f394e1aa2e8..8fc3ab938b9e 100644
--- a/kernel/bpf/hashtab.c
+++ b/kernel/bpf/hashtab.c
@@ -1534,6 +1534,44 @@ static long htab_map_delete_elem(struct bpf_map *map, void *key)
return ret;
}
+static long htab_map_delete_elem_cmp(struct bpf_map *map, void *key,
+ const void *compare, u32 off, u32 size)
+{
+ struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
+ struct hlist_nulls_head *head;
+ struct bucket *b;
+ struct htab_elem *l;
+ unsigned long flags;
+ u32 hash, key_size;
+ int ret;
+
+ WARN_ON_ONCE(!bpf_rcu_lock_held());
+
+ key_size = map->key_size;
+
+ hash = htab_map_hash(key, key_size, htab->hashrnd);
+ b = __select_bucket(htab, hash);
+ head = &b->head;
+
+ ret = htab_lock_bucket(b, &flags);
+ if (ret)
+ return ret;
+
+ l = lookup_elem_raw(head, hash, key, key_size);
+ if (!l)
+ ret = -ENOENT;
+ else if (memcmp(htab_elem_value(l, key_size) + off, compare, size) != 0)
+ ret = -EBUSY;
+ else
+ hlist_nulls_del_rcu(&l->hash_node);
+
+ htab_unlock_bucket(b, flags);
+
+ if (!ret)
+ free_htab_elem(htab, l); /* ret==0 implies we unlinked l */
+ return ret;
+}
+
static long htab_lru_map_delete_elem(struct bpf_map *map, void *key)
{
struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
@@ -2366,6 +2404,7 @@ const struct bpf_map_ops htab_map_ops = {
.map_lookup_and_delete_elem = htab_map_lookup_and_delete_elem,
.map_update_elem = htab_map_update_elem,
.map_delete_elem = htab_map_delete_elem,
+ .map_delete_elem_cmp = htab_map_delete_elem_cmp,
.map_gen_lookup = htab_map_gen_lookup,
.map_seq_show_elem = htab_map_seq_show_elem,
.map_set_for_each_callback_args = map_set_for_each_callback_args,
diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
index b44106c8ea75..42fc432f82e1 100644
--- a/kernel/bpf/syscall.c
+++ b/kernel/bpf/syscall.c
@@ -1892,18 +1892,30 @@ static int map_update_elem(union bpf_attr *attr, bpfptr_t uattr)
return err;
}
-#define BPF_MAP_DELETE_ELEM_LAST_FIELD key
+#define BPF_MAP_DELETE_ELEM_LAST_FIELD compare_size
static int map_delete_elem(union bpf_attr *attr, bpfptr_t uattr)
{
bpfptr_t ukey = make_bpfptr(attr->key, uattr.is_kernel);
struct bpf_map *map;
- void *key;
+ void *key, *compare = NULL;
+ u32 off = 0, size = 0;
int err;
if (CHECK_ATTR(BPF_MAP_DELETE_ELEM))
return -EINVAL;
+ if (attr->flags & ~BPF_F_COMPARE)
+ return -EINVAL;
+
+ /* The compare* fields are meaningful only with BPF_F_COMPARE. Reject them
+ * when the flag is absent so a dropped BPF_F_COMPARE cannot silently turn a
+ * compare-and-delete into an unconditional delete.
+ */
+ if (!(attr->flags & BPF_F_COMPARE) &&
+ (attr->compare || attr->compare_offset || attr->compare_size))
+ return -EINVAL;
+
CLASS(fd, f)(attr->map_fd);
map = __bpf_map_get(f);
if (IS_ERR(map))
@@ -1920,6 +1932,38 @@ static int map_delete_elem(union bpf_attr *attr, bpfptr_t uattr)
goto err_put;
}
+ if (attr->flags & BPF_F_COMPARE) {
+ bpfptr_t ucmp = make_bpfptr(attr->compare, uattr.is_kernel);
+
+ off = attr->compare_offset;
+ size = attr->compare_size ?: map->value_size;
+ /* off + size must fit in value_size, overflow-safe */
+ if (size > map->value_size || off > map->value_size - size) {
+ err = -EINVAL;
+ goto out;
+ }
+ if (!map->ops->map_delete_elem_cmp) {
+ err = -EOPNOTSUPP;
+ goto out;
+ }
+ /* Compare-and-delete reads the raw stored value. Maps whose value carries
+ * BTF-managed fields (bpf_spin_lock, bpf_timer, kptr, ...) have
+ * those bytes sanitised on lookup, so a raw compare would never
+ * match the caller's snapshot and could expose kernel-internal
+ * bytes. Reject such maps for now.
+ */
+ if (!IS_ERR_OR_NULL(map->record)) {
+ err = -EOPNOTSUPP;
+ goto out;
+ }
+ compare = kvmemdup_bpfptr(ucmp, size);
+ if (IS_ERR(compare)) {
+ err = PTR_ERR(compare);
+ compare = NULL;
+ goto out;
+ }
+ }
+
if (bpf_map_is_offloaded(map)) {
err = bpf_map_offload_delete_elem(map, key);
goto out;
@@ -1932,12 +1976,16 @@ static int map_delete_elem(union bpf_attr *attr, bpfptr_t uattr)
bpf_disable_instrumentation();
rcu_read_lock();
- err = map->ops->map_delete_elem(map, key);
+ if (compare)
+ err = map->ops->map_delete_elem_cmp(map, key, compare, off, size);
+ else
+ err = map->ops->map_delete_elem(map, key);
rcu_read_unlock();
bpf_enable_instrumentation();
if (!err)
maybe_wait_bpf_programs(map);
out:
+ kvfree(compare);
kvfree(key);
err_put:
bpf_map_write_active_dec(map);
--
2.39.5 (Apple Git-154)
^ permalink raw reply related [flat|nested] 6+ messages in thread* Re: [RFC bpf-next 2/3] bpf: implement compare-and-delete (BPF_F_COMPARE) for BPF_MAP_TYPE_HASH
2026-06-22 7:16 ` [RFC bpf-next 2/3] bpf: implement compare-and-delete (BPF_F_COMPARE) for BPF_MAP_TYPE_HASH Gyutae Bae
@ 2026-06-22 7:58 ` sashiko-bot
0 siblings, 0 replies; 6+ messages in thread
From: sashiko-bot @ 2026-06-22 7:58 UTC (permalink / raw)
To: Gyutae Bae; +Cc: bpf
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] BPF_MAP_DELETE_ELEM no longer validates that the `value` field is zero, silently allowing uninitialized garbage and breaking BPF's strict UAPI zero-initialization guarantees.
--
commit 513ae67888e0e87d8de22021be59f6117444ee43
Author: Gyutae Bae <gyutae.bae@navercorp.com>
bpf: implement compare-and-delete (BPF_F_COMPARE) for BPF_MAP_TYPE_HASH
This commit introduces the BPF_F_COMPARE flag for map_delete_elem(), allowing
atomic compare-and-delete operations on hash maps. It adds a new map operation
and updates the syscall handler to process the new compare fields.
> diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
> --- a/kernel/bpf/syscall.c
> +++ b/kernel/bpf/syscall.c
[ ... ]
> @@ -1892,18 +1892,30 @@ static int map_update_elem(union bpf_attr *attr, bpfptr_t uattr)
> return err;
> }
>
> -#define BPF_MAP_DELETE_ELEM_LAST_FIELD key
> +#define BPF_MAP_DELETE_ELEM_LAST_FIELD compare_size
>
> static int map_delete_elem(union bpf_attr *attr, bpfptr_t uattr)
> {
> bpfptr_t ukey = make_bpfptr(attr->key, uattr.is_kernel);
> struct bpf_map *map;
> - void *key;
> + void *key, *compare = NULL;
> + u32 off = 0, size = 0;
> int err;
>
> if (CHECK_ATTR(BPF_MAP_DELETE_ELEM))
> return -EINVAL;
>
> + if (attr->flags & ~BPF_F_COMPARE)
> + return -EINVAL;
> +
> + /* The compare* fields are meaningful only with BPF_F_COMPARE. Reject them
> + * when the flag is absent so a dropped BPF_F_COMPARE cannot silently turn a
> + * compare-and-delete into an unconditional delete.
> + */
> + if (!(attr->flags & BPF_F_COMPARE) &&
> + (attr->compare || attr->compare_offset || attr->compare_size))
> + return -EINVAL;
[Severity: High]
Does changing BPF_MAP_DELETE_ELEM_LAST_FIELD to compare_size bypass
zero-validation for attr->value?
CHECK_ATTR now only validates that fields after compare_size are zero.
The patch manually adds validation for flags and the compare* fields to ensure
they are zero when not in use, but it does not appear to add a similar check
for attr->value.
Could this allow userspace to pass uninitialized stack garbage in the value
field without being rejected, permanently preventing the safe repurposing of
the value field for future BPF_MAP_DELETE_ELEM extensions?
--
Sashiko AI review · https://sashiko.dev/#/patchset/20260622071649.31541-1-gyutae.opensource@navercorp.com?part=2
^ permalink raw reply [flat|nested] 6+ messages in thread
* [RFC bpf-next 3/3] selftests/bpf: test BPF_F_COMPARE compare-and-delete
2026-06-22 7:16 [RFC bpf-next 0/3] bpf: compare-and-delete (BPF_F_COMPARE) for hash maps Gyutae Bae
2026-06-22 7:16 ` [RFC bpf-next 1/3] bpf: add BPF_F_COMPARE flag and compare fields to map elem UAPI Gyutae Bae
2026-06-22 7:16 ` [RFC bpf-next 2/3] bpf: implement compare-and-delete (BPF_F_COMPARE) for BPF_MAP_TYPE_HASH Gyutae Bae
@ 2026-06-22 7:16 ` Gyutae Bae
2026-06-22 7:53 ` sashiko-bot
2 siblings, 1 reply; 6+ messages in thread
From: Gyutae Bae @ 2026-06-22 7:16 UTC (permalink / raw)
To: Alexei Starovoitov, Daniel Borkmann, Andrii Nakryiko, bpf
Cc: John Fastabend, Eduard Zingerman, Kumar Kartikeya Dwivedi,
Martin KaFai Lau, Song Liu, Yonghong Song, Jiri Olsa,
Emil Tsalapatis, Shuah Khan, linux-kselftest, linux-kernel,
Minsu Jeon, Siwan Kim, Jonghyeon Kim, Gyutae Bae
From: Gyutae Bae <gyutae.bae@navercorp.com>
Add a test for compare-and-delete on a hash map. There is no libbpf
wrapper yet, so it issues the request through the raw bpf() syscall and
compares against a revision field stored in the value.
A matching expected value deletes the element, a stale one leaves it in
place and returns -EBUSY, and deleting an already-removed key returns
-ENOENT. A compare window that reaches past value_size is rejected with
-EINVAL, as are compare fields passed without BPF_F_COMPARE, so a dropped
flag cannot silently turn into an unconditional delete. Running the same
call on an array map, which has no compare-and-delete handler, returns
-EOPNOTSUPP.
Co-developed-by: Minsu Jeon <minsu.jeon@navercorp.com>
Signed-off-by: Minsu Jeon <minsu.jeon@navercorp.com>
Co-developed-by: Siwan Kim <siwan.kim@navercorp.com>
Signed-off-by: Siwan Kim <siwan.kim@navercorp.com>
Co-developed-by: Jonghyeon Kim <jong-hyeon.kim@navercorp.com>
Signed-off-by: Jonghyeon Kim <jong-hyeon.kim@navercorp.com>
Signed-off-by: Gyutae Bae <gyutae.bae@navercorp.com>
---
.../selftests/bpf/prog_tests/map_cmp_delete.c | 106 ++++++++++++++++++
1 file changed, 106 insertions(+)
create mode 100644 tools/testing/selftests/bpf/prog_tests/map_cmp_delete.c
diff --git a/tools/testing/selftests/bpf/prog_tests/map_cmp_delete.c b/tools/testing/selftests/bpf/prog_tests/map_cmp_delete.c
new file mode 100644
index 000000000000..6d6df13adcc4
--- /dev/null
+++ b/tools/testing/selftests/bpf/prog_tests/map_cmp_delete.c
@@ -0,0 +1,106 @@
+// SPDX-License-Identifier: GPL-2.0
+
+/*
+ * Copyright (c) 2026 NAVER Corp.
+ *
+ * Author: Gyutae Bae <gyutae.bae@navercorp.com>
+ */
+
+#include <test_progs.h>
+#include <sys/syscall.h>
+
+struct val {
+ __u64 revision; /* synchronization field we compare on */
+ __u64 payload;
+};
+
+/* libbpf has no wrapper that passes the compare-and-delete fields, so
+ * issue bpf() directly. 'flags' is a parameter so tests can also
+ * exercise the no-BPF_F_COMPARE case.
+ */
+static int cmp_delete_flags(int fd, const void *key, const void *compare,
+ __u32 off, __u32 size, __u64 flags)
+{
+ union bpf_attr attr;
+ int ret;
+
+ memset(&attr, 0, sizeof(attr));
+ attr.map_fd = fd;
+ attr.key = (__u64)(unsigned long)key;
+ attr.flags = flags;
+ attr.compare = (__u64)(unsigned long)compare;
+ attr.compare_offset = off;
+ attr.compare_size = size;
+ ret = syscall(__NR_bpf, BPF_MAP_DELETE_ELEM, &attr, sizeof(attr));
+ return ret < 0 ? -errno : ret;
+}
+
+static int cmp_delete(int fd, const void *key, const void *compare,
+ __u32 off, __u32 size)
+{
+ return cmp_delete_flags(fd, key, compare, off, size, BPF_F_COMPARE);
+}
+
+static int hash_map(void)
+{
+ return bpf_map_create(BPF_MAP_TYPE_HASH, "cmp_hash",
+ sizeof(__u32), sizeof(struct val), 64, NULL);
+}
+
+static void test_contract(void)
+{
+ const __u32 off = offsetof(struct val, revision);
+ const __u32 sz = sizeof(((struct val *)0)->revision);
+ struct val v = { .revision = 10, .payload = 0xabc };
+ struct val snap = v;
+ __u32 k = 1;
+ int fd = hash_map();
+
+ if (!ASSERT_GE(fd, 0, "map_create"))
+ return;
+ ASSERT_OK(bpf_map_update_elem(fd, &k, &v, BPF_ANY), "insert");
+
+ /* mismatch -> -EBUSY, entry preserved */
+ snap.revision = 9;
+ ASSERT_EQ(cmp_delete(fd, &k, &snap, off, sz), -EBUSY, "mismatch_ebusy");
+ ASSERT_OK(bpf_map_lookup_elem(fd, &k, &v), "still_present");
+
+ /* compare fields set but BPF_F_COMPARE omitted -> -EINVAL, entry preserved
+ * (a dropped flag must not silently become an unconditional delete)
+ */
+ ASSERT_EQ(cmp_delete_flags(fd, &k, &snap, off, sz, 0), -EINVAL,
+ "cmp_fields_no_flag_einval");
+ ASSERT_OK(bpf_map_lookup_elem(fd, &k, &v), "preserved_no_flag");
+
+ /* match -> 0, entry gone */
+ snap.revision = 10;
+ ASSERT_OK(cmp_delete(fd, &k, &snap, off, sz), "match_deletes");
+ ASSERT_EQ(bpf_map_lookup_elem(fd, &k, &v), -ENOENT, "gone");
+
+ /* absent -> -ENOENT */
+ ASSERT_EQ(cmp_delete(fd, &k, &snap, off, sz), -ENOENT, "absent_enoent");
+
+ /* bad window -> -EINVAL */
+ ASSERT_EQ(cmp_delete(fd, &k, &snap, 0, sizeof(struct val) + 8), -EINVAL,
+ "oversize_einval");
+
+ /* unsupported map type (no map_delete_elem_cmp op) -> -EOPNOTSUPP */
+ {
+ int afd = bpf_map_create(BPF_MAP_TYPE_ARRAY, "cmp_arr",
+ sizeof(__u32), sizeof(struct val), 4, NULL);
+ __u32 ak = 0;
+
+ if (ASSERT_GE(afd, 0, "array_create")) {
+ ASSERT_EQ(cmp_delete(afd, &ak, &snap, off, sz),
+ -EOPNOTSUPP, "array_eopnotsupp");
+ close(afd);
+ }
+ }
+ close(fd);
+}
+
+void test_map_cmp_delete(void)
+{
+ if (test__start_subtest("contract"))
+ test_contract();
+}
--
2.39.5 (Apple Git-154)
^ permalink raw reply related [flat|nested] 6+ messages in thread* Re: [RFC bpf-next 3/3] selftests/bpf: test BPF_F_COMPARE compare-and-delete
2026-06-22 7:16 ` [RFC bpf-next 3/3] selftests/bpf: test BPF_F_COMPARE compare-and-delete Gyutae Bae
@ 2026-06-22 7:53 ` sashiko-bot
0 siblings, 0 replies; 6+ messages in thread
From: sashiko-bot @ 2026-06-22 7:53 UTC (permalink / raw)
To: Gyutae Bae; +Cc: bpf
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Low] Multi-line comments do not follow the mandatory BPF subsystem style.
--
commit d9d64a94ef921683b52276aed902be3aa7f68eed
Author: Gyutae Bae <gyutae.bae@navercorp.com>
selftests/bpf: test BPF_F_COMPARE compare-and-delete
This patch introduces a selftest for the compare-and-delete operation on a
hash map. Since there is no libbpf wrapper, it uses the raw bpf syscall to
verify the deletion behavior based on a revision field.
> diff --git a/tools/testing/selftests/bpf/prog_tests/map_cmp_delete.c b/tools/testing/selftests/bpf/prog_tests/map_cmp_delete.c
> new file mode 100644
> index 0000000000000..6d6df13adcc42
> --- /dev/null
> +++ b/tools/testing/selftests/bpf/prog_tests/map_cmp_delete.c
[ ... ]
> +struct val {
> + __u64 revision; /* synchronization field we compare on */
> + __u64 payload;
> +};
> +
> +/* libbpf has no wrapper that passes the compare-and-delete fields, so
> + * issue bpf() directly. 'flags' is a parameter so tests can also
> + * exercise the no-BPF_F_COMPARE case.
> + */
[Severity: Low]
This isn't a bug, but does this multi-line comment follow the preferred BPF
subsystem style? The guidelines specify that the opening /* should be on its
own line, with the text starting on the following line.
> +static int cmp_delete_flags(int fd, const void *key, const void *compare,
> + __u32 off, __u32 size, __u64 flags)
[ ... ]
> +static void test_contract(void)
> +{
[ ... ]
> + /* mismatch -> -EBUSY, entry preserved */
> + snap.revision = 9;
> + ASSERT_EQ(cmp_delete(fd, &k, &snap, off, sz), -EBUSY, "mismatch_ebusy");
> + ASSERT_OK(bpf_map_lookup_elem(fd, &k, &v), "still_present");
> +
> + /* compare fields set but BPF_F_COMPARE omitted -> -EINVAL, entry preserved
> + * (a dropped flag must not silently become an unconditional delete)
> + */
[Severity: Low]
This isn't a bug, but is this multi-line comment using the preferred
formatting? Similar to the comment above, the opening /* should typically
be placed on its own empty line.
> + ASSERT_EQ(cmp_delete_flags(fd, &k, &snap, off, sz, 0), -EINVAL,
> + "cmp_fields_no_flag_einval");
> + ASSERT_OK(bpf_map_lookup_elem(fd, &k, &v), "preserved_no_flag");
--
Sashiko AI review · https://sashiko.dev/#/patchset/20260622071649.31541-1-gyutae.opensource@navercorp.com?part=3
^ permalink raw reply [flat|nested] 6+ messages in thread