Skip to contentSkip to conversation

Address more libbpf deprecations

PATCH Applied David Ahern 4 participants 9 messages Apr 23, 2022 – May 2, 2022

Subsystems (3)
lore
Share & tools
lore
#1 lore raw

PATCH 1/3 libbpf: Use bpf_object__load instead of bpf_object__load_xattr

In reply to David Ahern

bpf_object__load_xattr is deprecated as of v0.8+; remove it in favor of bpf_object__load.

---
 lib/bpf_libbpf.c | 7 +------
 1 file changed, 1 insertion(+), 6 deletions(-)
lib/bpf_libbpf.c +1 -6
index f4f98caa1e58..f723f6310c28 100644
@@ -248,7 +248,6 @@ static int handle_legacy_maps(struct bpf_object *obj)
 
 static int load_bpf_object(struct bpf_cfg_in *cfg)
 {
-	struct bpf_object_load_attr attr = {};
 	struct bpf_program *p, *prog = NULL;
 	struct bpf_object *obj;
 	char root_path[PATH_MAX];
@@ -305,11 +304,7 @@ static int load_bpf_object(struct bpf_cfg_in *cfg)
 	if (ret)
 		goto unload_obj;
 
-	attr.obj = obj;
-	if (cfg->verbose)
-		attr.log_level = 2;
-
-	ret = bpf_object__load_xattr(&attr);
+	ret = bpf_object__load(obj);
 	if (ret)
 		goto unload_obj;
 
Signature · 1 line
2.24.3 (Apple Git-128)
#0 lore raw

cover Address more libbpf deprecations

Another round of changes to handle libbpf deprecations. Compiles are clean as of libbpf commit 533c7666eb72 ("Fix downloads formats").

David Ahern (3):
libbpf: Use bpf_object__load instead of bpf_object__load_xattr
libbpf: Remove use of bpf_program__set_priv and bpf_program__priv
libbpf: Remove use of bpf_map_is_offload_neutral

 lib/bpf_libbpf.c | 30 +++++++++++++++++-------------
 1 file changed, 17 insertions(+), 13 deletions(-)
Signature · 1 line
2.24.3 (Apple Git-128)
#6 lore raw

PATCH 3/3 libbpf: Remove use of bpf_map_is_offload_neutral

bpf_map_is_offload_neutral is deprecated as of v0.8+; import definition to maintain backwards compatibility.

---
 lib/bpf_libbpf.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
lib/bpf_libbpf.c +6 -1
index 7dd1faf536f4..7b16ee71589d 100644
@@ -249,6 +249,11 @@ static int handle_legacy_maps(struct bpf_object *obj)
 	return ret;
 }
 
+static bool bpf_map_is_offload_neutral(const struct bpf_map *map)
+{
+	return bpf_map__type(map) == BPF_MAP_TYPE_PERF_EVENT_ARRAY;
+}
+
 static int load_bpf_object(struct bpf_cfg_in *cfg)
 {
 	struct bpf_program *p, *prog = NULL;
@@ -294,7 +299,7 @@ static int load_bpf_object(struct bpf_cfg_in *cfg)
 	}
 
 	bpf_object__for_each_map(map, obj) {
-		if (!bpf_map__is_offload_neutral(map))
+		if (!bpf_map_is_offload_neutral(map))
 			bpf_map__set_ifindex(map, cfg->ifindex);
 	}
 
Signature · 1 line
2.24.3 (Apple Git-128)
#5 lore raw

PATCH 2/3 libbpf: Remove use of bpf_program__set_priv and bpf_program__priv

In reply to David Ahern

bpf_program__set_priv and bpf_program__priv are deprecated as of libbpf v0.7+. Rather than store the map as priv on the program, change find_legacy_tail_calls to take an argument to return a reference to the map.

find_legacy_tail_calls is invoked twice from load_bpf_object - the first time to check for programs that should be loaded. In this case a reference to the map is not needed, but it does validate the map exists. The second is invoked from update_legacy_tail_call_maps where the map pointer is needed.

---
 lib/bpf_libbpf.c | 16 ++++++++++------
 1 file changed, 10 insertions(+), 6 deletions(-)
lib/bpf_libbpf.c +10 -6
index f723f6310c28..7dd1faf536f4 100644
@@ -151,7 +151,8 @@ handle_legacy_map_in_map(struct bpf_object *obj, struct bpf_map *inner_map,
 	return ret;
 }
 
-static int find_legacy_tail_calls(struct bpf_program *prog, struct bpf_object *obj)
+static int find_legacy_tail_calls(struct bpf_program *prog, struct bpf_object *obj,
+				  struct bpf_map **pmap)
 {
 	unsigned int map_id, key_id;
 	const char *sec_name;
@@ -175,8 +176,8 @@ static int find_legacy_tail_calls(struct bpf_program *prog, struct bpf_object *o
 	if (!map)
 		return -1;
 
-	/* Save the map here for later updating */
-	bpf_program__set_priv(prog, map, NULL);
+	if (pmap)
+		*pmap = map;
 
 	return 0;
 }
@@ -190,8 +191,10 @@ static int update_legacy_tail_call_maps(struct bpf_object *obj)
 	struct bpf_map *map;
 
 	bpf_object__for_each_program(prog, obj) {
-		map = bpf_program__priv(prog);
-		if (!map)
+		/* load_bpf_object has already verified find_legacy_tail_calls
+		 * succeeds when it should
+		 */
+		if (find_legacy_tail_calls(prog, obj, &map) < 0)
 			continue;
 
 		prog_fd = bpf_program__fd(prog);
@@ -275,7 +278,8 @@ static int load_bpf_object(struct bpf_cfg_in *cfg)
 
 		/* Only load the programs that will either be subsequently
 		 * attached or inserted into a tail call map */
-		if (find_legacy_tail_calls(p, obj) < 0 && !prog_to_attach) {
+		if (find_legacy_tail_calls(p, obj, NULL) < 0 &&
+		    !prog_to_attach) {
 			ret = bpf_program__set_autoload(p, false);
 			if (ret)
 				return -EINVAL;
Signature · 1 line
2.24.3 (Apple Git-128)
Acked-by on the v1 series
#7 lore raw
In reply to David Ahern
Quoted 10 lines · David Ahern

Another round of changes to handle libbpf deprecations. Compiles are clean as of libbpf commit 533c7666eb72 ("Fix downloads formats").

David Ahern (3):
libbpf: Use bpf_object__load instead of bpf_object__load_xattr
libbpf: Remove use of bpf_program__set_priv and bpf_program__priv
libbpf: Remove use of bpf_map_is_offload_neutral

 lib/bpf_libbpf.c | 30 +++++++++++++++++-------------
 1 file changed, 17 insertions(+), 13 deletions(-)

For the series:

#2 lore raw

Re: [PATCH iproute2-next 1/3] libbpf: Use bpf_object__load instead of bpf_object__load_xattr

In reply to David Ahern

Hi David,

This patch revert c04e45d0 lib/bpf: fix verbose flag when using libbpf, Should we set prog->log_level directly before it loaded, like bpf_program__set_log_level() does?

Thanks
Hangbin

Quoted 36 lines · David Ahern

bpf_object__load_xattr is deprecated as of v0.8+; remove it in favor of bpf_object__load.

---
 lib/bpf_libbpf.c | 7 +------
 1 file changed, 1 insertion(+), 6 deletions(-)
lib/bpf_libbpf.c +1 -6
index f4f98caa1e58..f723f6310c28 100644
@@ -248,7 +248,6 @@ static int handle_legacy_maps(struct bpf_object *obj)
 
 static int load_bpf_object(struct bpf_cfg_in *cfg)
 {
-	struct bpf_object_load_attr attr = {};
 	struct bpf_program *p, *prog = NULL;
 	struct bpf_object *obj;
 	char root_path[PATH_MAX];
@@ -305,11 +304,7 @@ static int load_bpf_object(struct bpf_cfg_in *cfg)
 	if (ret)
 		goto unload_obj;
 
-	attr.obj = obj;
-	if (cfg->verbose)
-		attr.log_level = 2;
-
-	ret = bpf_object__load_xattr(&attr);
+	ret = bpf_object__load(obj);
 	if (ret)
 		goto unload_obj;
 
Signature · 1 line
2.24.3 (Apple Git-128)
#3 lore raw

Re: [PATCH iproute2-next 1/3] libbpf: Use bpf_object__load instead of bpf_object__load_xattr

Quoted 6 lines · Hangbin Liu

Hi David,

This patch revert c04e45d0 lib/bpf: fix verbose flag when using libbpf, Should we set prog->log_level directly before it loaded, like bpf_program__set_log_level() does?

that API is new - Dec 2021 so it will not be present across relevant libbpf versions. Detecting what exists in a libbpf version and adding compat wrappers needs to be added. That's an undertaking I do not have time for at the moment. If you or someone else does it would be appreciated.

#4 lore raw

Re: [PATCH iproute2-next 1/3] libbpf: Use bpf_object__load instead of bpf_object__load_xattr

Quoted 8 lines · David Ahern
Quoted 6 lines · Hangbin Liu

Hi David,

This patch revert c04e45d0 lib/bpf: fix verbose flag when using libbpf, Should we set prog->log_level directly before it loaded, like bpf_program__set_log_level() does?

Quoted 5 lines

that API is new - Dec 2021 so it will not be present across relevant libbpf versions. Detecting what exists in a libbpf version and adding compat wrappers needs to be added. That's an undertaking I do not have time for at the moment. If you or someone else does it would be appreciated.

Ah, yes, I forgot we can't set prog log_level directly. Looks we need to add a new flag similar with HAVE_LIBBPF_SECTION_NAME... It's really a pain to add more flags for libbfp...

Hi Paul, will you do that?

Thanks
Hangbin

#8 lore raw
In reply to David Ahern
This series was applied to iproute2/iproute2-next.git (main) by David Ahern <[email protected]>:

Hello:

This series was applied to iproute2/iproute2-next.git (main) by David Ahern <[email protected]>:

Quoted 9 lines · On Sat, 23 Apr 2022 09:22:57 -0600 you wrote:

Another round of changes to handle libbpf deprecations. Compiles are clean as of libbpf commit 533c7666eb72 ("Fix downloads formats").

David Ahern (3):
libbpf: Use bpf_object__load instead of bpf_object__load_xattr
libbpf: Remove use of bpf_program__set_priv and bpf_program__priv
libbpf: Remove use of bpf_map_is_offload_neutral

[...]

Here is the summary with links:
- [iproute2-next,1/3] libbpf: Use bpf_object__load instead of bpf_object__load_xattr

    https://git.kernel.org/pub/scm/network/iproute2/iproute2-next.git/commit/?id=ba6519cbcb28
  - [iproute2-next,2/3] libbpf: Remove use of bpf_program__set_priv and bpf_program__priv
    https://git.kernel.org/pub/scm/network/iproute2/iproute2-next.git/commit/?id=64e5ed779f5d
  - [iproute2-next,3/3] libbpf: Remove use of bpf_map_is_offload_neutral
    https://git.kernel.org/pub/scm/network/iproute2/iproute2-next.git/commit/?id=837294e45252

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html