Skip to contentSkip to conversation

bpf: Relax tracing prog recursive attach rules

RFC PATCH Superseded Dmitrii Dolgov 2 participants 4 messages Nov 14, 2023 – Nov 21, 2023

Subsystems (3)

v1 of 12 Superseded Latest is v12 →

lore
Share & tools
lore
#0 lore raw

RFC PATCH 1/1 bpf: Relax tracing prog recursive attach rules

Currently, it's not allowed to attach an fentry/fexit prog to another one of the same type. At the same time it's not uncommon to see a tracing program with lots of logic in use, and the attachment limitation prevents usage of fentry/fexit for performance analysis (e.g. with "bpftool prog profile" command) in this case. An example could be falcosecurity libs project that uses tp_btf tracing programs.

Following the corresponding discussion [1], the reason for that is to avoid tracing progs call cycles without introducing more complex solutions. Relax "no same type" requirement to "no progs that are already an attach target themselves" for the tracing type. In this way only a standalone tracing program (without any other progs attached to it) could be attached to another one, and no cycle could be formed. To implement, add a new field into bpf_prog_aux to track the fact of attachment in the target prog.

As a side effect of this change alone, one could create an unbounded chain of tracing progs attached to each other. Similar issues between fentry/fexit and extend progs are addressed via forbidding certain combinations that could lead to similar chains. Introduce an attach_depth field to limit the attachment chain, and display it in bpftool.

[1]: https://lore.kernel.org/bpf/[email protected]/

---
 include/linux/bpf.h            |  2 ++
 include/uapi/linux/bpf.h       |  1 +
 kernel/bpf/syscall.c           |  8 +++++++-
 kernel/bpf/verifier.c          | 19 ++++++++++++++++---
 tools/bpf/bpftool/prog.c       |  2 ++
 tools/include/uapi/linux/bpf.h |  1 +
 6 files changed, 29 insertions(+), 4 deletions(-)
include/linux/bpf.h +2 -0
index 4001d11be151..9fddb420cfe2 100644
@@ -1394,6 +1394,7 @@ struct bpf_prog_aux {
 	u32 real_func_cnt; /* includes hidden progs, only used for JIT and freeing progs */
 	u32 func_idx; /* 0 for non-func prog, the index in func array for func prog */
 	u32 attach_btf_id; /* in-kernel BTF type id to attach to */
+	u32 attach_depth; /* position of the prog in the attachment chain */
 	u32 ctx_arg_info_size;
 	u32 max_rdonly_access;
 	u32 max_rdwr_access;
@@ -1414,6 +1415,7 @@ struct bpf_prog_aux {
 	bool xdp_has_frags;
 	bool exception_cb;
 	bool exception_boundary;
+	bool followed; /* Another program is attached to it. */
 	/* BTF_KIND_FUNC_PROTO for valid attach_btf_id */
 	const struct btf_type *attach_func_proto;
 	/* function name for valid attach_btf_id */
include/uapi/linux/bpf.h +1 -0
index 7cf8bcf9f6a2..aa6614547ef6 100644
@@ -6465,6 +6465,7 @@ struct bpf_prog_info {
 	__u32 verified_insns;
 	__u32 attach_btf_obj_id;
 	__u32 attach_btf_id;
+	__u32 attach_depth;
 } __attribute__((aligned(8)));
 
 struct bpf_map_info {
kernel/bpf/syscall.c +7 -1
index 0ed286b8a0f0..fe2b4ef82623 100644
@@ -3038,9 +3038,12 @@ static void bpf_tracing_link_release(struct bpf_link *link)
 
 	bpf_trampoline_put(tr_link->trampoline);
 
+	link->prog->aux->attach_depth--;
 	/* tgt_prog is NULL if target is a kernel function */
-	if (tr_link->tgt_prog)
+	if (tr_link->tgt_prog) {
+		tr_link->tgt_prog->aux->followed = false;
 		bpf_prog_put(tr_link->tgt_prog);
+	}
 }
 
 static void bpf_tracing_link_dealloc(struct bpf_link *link)
@@ -3235,6 +3238,8 @@ static int bpf_tracing_prog_attach(struct bpf_prog *prog,
 	if (err)
 		goto out_unlock;
 
+	tgt_prog->aux->followed = true;
+	prog->aux->attach_depth = tgt_prog->aux->attach_depth + 1;
 	err = bpf_trampoline_link_prog(&link->link, tr);
 	if (err) {
 		bpf_link_cleanup(&link_primer);
@@ -4509,6 +4514,7 @@ static int bpf_prog_get_info_by_fd(struct file *file,
 	if (prog->aux->btf)
 		info.btf_id = btf_obj_id(prog->aux->btf);
 	info.attach_btf_id = prog->aux->attach_btf_id;
+	info.attach_depth = prog->aux->attach_depth;
 	if (attach_btf)
 		info.attach_btf_obj_id = btf_obj_id(attach_btf);
 
kernel/bpf/verifier.c +16 -3
index 9ae6eae13471..2c569083874f 100644
@@ -20329,6 +20329,12 @@ int bpf_check_attach_target(struct bpf_verifier_log *log,
 	if (tgt_prog) {
 		struct bpf_prog_aux *aux = tgt_prog->aux;
 
+		if (aux->attach_depth >= 32) {
+			bpf_log(log, "Target program attach depth is %d. Too large\n",
+					aux->attach_depth);
+			return -EINVAL;
+		}
+
 		if (bpf_prog_is_dev_bound(prog->aux) &&
 		    !bpf_prog_dev_bound_match(prog, tgt_prog)) {
 			bpf_log(log, "Target program bound device mismatch");
@@ -20367,9 +20373,16 @@ int bpf_check_attach_target(struct bpf_verifier_log *log,
 			bpf_log(log, "Can attach to only JITed progs\n");
 			return -EINVAL;
 		}
-		if (tgt_prog->type == prog->type) {
-			/* Cannot fentry/fexit another fentry/fexit program.
-			 * Cannot attach program extension to another extension.
+		if (tgt_prog->type == prog->type &&
+			(prog_extension || prog->aux->followed)) {
+			/*
+			 * To avoid potential call chain cycles, prevent attaching programs
+			 * of the same type. The only exception is standalone fentry/fexit
+			 * programs that themselves are not attachment targets.
+			 * That means:
+			 *  - Cannot attach followed fentry/fexit to another
+			 *    fentry/fexit program.
+			 *  - Cannot attach program extension to another extension.
 			 * It's ok to attach fentry/fexit to extension program.
 			 */
 			bpf_log(log, "Cannot recursively attach\n");
tools/bpf/bpftool/prog.c +2 -0
index 7ec4f5671e7a..24565e8bb825 100644
@@ -554,6 +554,8 @@ static void print_prog_plain(struct bpf_prog_info *info, int fd)
 		printf("  memlock %sB", memlock);
 	free(memlock);
 
+	printf("  attach depth %d", info->attach_depth);
+
 	if (info->nr_map_ids)
 		show_prog_maps(fd, info->nr_map_ids);
 
tools/include/uapi/linux/bpf.h +1 -0
index 7cf8bcf9f6a2..aa6614547ef6 100644
@@ -6465,6 +6465,7 @@ struct bpf_prog_info {
 	__u32 verified_insns;
 	__u32 attach_btf_obj_id;
 	__u32 attach_btf_id;
+	__u32 attach_depth;
 } __attribute__((aligned(8)));
 
 struct bpf_map_info {

base-commit: 100888fb6d8a185866b1520031ee7e3182b173de

Signature · 1 line
2.41.0
#1 lore raw
Quoted 17 lines · Dmitrii Dolgov

Currently, it's not allowed to attach an fentry/fexit prog to another one of the same type. At the same time it's not uncommon to see a tracing program with lots of logic in use, and the attachment limitation prevents usage of fentry/fexit for performance analysis (e.g. with "bpftool prog profile" command) in this case. An example could be falcosecurity libs project that uses tp_btf tracing programs.

Following the corresponding discussion [1], the reason for that is to avoid tracing progs call cycles without introducing more complex solutions. Relax "no same type" requirement to "no progs that are already an attach target themselves" for the tracing type. In this way only a standalone tracing program (without any other progs attached to it) could be attached to another one, and no cycle could be formed. To implement, add a new field into bpf_prog_aux to track the fact of attachment in the target prog.

Quoted 6 lines

As a side effect of this change alone, one could create an unbounded chain of tracing progs attached to each other. Similar issues between fentry/fexit and extend progs are addressed via forbidding certain combinations that could lead to similar chains. Introduce an attach_depth field to limit the attachment chain, and display it in bpftool.

From what I see currently it's not possible to achieve such a call cycle with tracing programs (even without the verifier check addressed in this patch), so I had to test this change on a modified kernel with a couple of other checks disabled as well. But otherwise I would appreciate feedback about whether it's a reasonable thing to do.

#2 lore raw
In reply to Dmitrii Dolgov

Hi Dmitrii,

kernel test robot noticed the following build warnings:

url:    https://github.com/intel-lab-lkp/linux/commits/Dmitrii-Dolgov/bpf-Relax-tracing-prog-recursive-attach-rules/20231114-164631
base:   100888fb6d8a185866b1520031ee7e3182b173de
patch link:    https://lore.kernel.org/r/20231114084118.11095-1-9erthalion6%40gmail.com
patch subject: [RFC PATCH bpf-next] bpf: Relax tracing prog recursive attach rules
config: i386-randconfig-141-20231119 (https://download.01.org/0day-ci/archive/20231119/[email protected]/config)
compiler: gcc-12 (Debian 12.2.0-14) 12.2.0
reproduce: (https://download.01.org/0day-ci/archive/20231119/[email protected]/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags

| Reported-by: kernel test robot <[email protected]>
| Reported-by: Dan Carpenter <[email protected]>
| Closes: https://lore.kernel.org/r/[email protected]/

smatch warnings:
kernel/bpf/syscall.c:3241 bpf_tracing_prog_attach() error: we previously assumed 'tgt_prog' could be null (see line 3186)

vim +/tgt_prog +3241 kernel/bpf/syscall.c

4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3096  static int bpf_tracing_prog_attach(struct bpf_prog *prog,
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3097  				   int tgt_prog_fd,
2fcc82411e74e5 Kui-Feng Lee           2022-05-10  3098  				   u32 btf_id,
2fcc82411e74e5 Kui-Feng Lee           2022-05-10  3099  				   u64 bpf_cookie)
fec56f5890d93f Alexei Starovoitov     2019-11-14  3100  {
a3b80e1078943d Andrii Nakryiko        2020-04-28  3101  	struct bpf_link_primer link_primer;
3aac1ead5eb6b7 Toke Høiland-Jørgensen 2020-09-29  3102  	struct bpf_prog *tgt_prog = NULL;
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3103  	struct bpf_trampoline *tr = NULL;
70ed506c3bbcfa Andrii Nakryiko        2020-03-02  3104  	struct bpf_tracing_link *link;
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3105  	u64 key = 0;
a3b80e1078943d Andrii Nakryiko        2020-04-28  3106  	int err;
fec56f5890d93f Alexei Starovoitov     2019-11-14  3107  
9e4e01dfd3254c KP Singh               2020-03-29  3108  	switch (prog->type) {
9e4e01dfd3254c KP Singh               2020-03-29  3109  	case BPF_PROG_TYPE_TRACING:
fec56f5890d93f Alexei Starovoitov     2019-11-14  3110  		if (prog->expected_attach_type != BPF_TRACE_FENTRY &&
be8704ff07d237 Alexei Starovoitov     2020-01-20  3111  		    prog->expected_attach_type != BPF_TRACE_FEXIT &&
9e4e01dfd3254c KP Singh               2020-03-29  3112  		    prog->expected_attach_type != BPF_MODIFY_RETURN) {
9e4e01dfd3254c KP Singh               2020-03-29  3113  			err = -EINVAL;
9e4e01dfd3254c KP Singh               2020-03-29  3114  			goto out_put_prog;
9e4e01dfd3254c KP Singh               2020-03-29  3115  		}
9e4e01dfd3254c KP Singh               2020-03-29  3116  		break;
9e4e01dfd3254c KP Singh               2020-03-29  3117  	case BPF_PROG_TYPE_EXT:
9e4e01dfd3254c KP Singh               2020-03-29  3118  		if (prog->expected_attach_type != 0) {
9e4e01dfd3254c KP Singh               2020-03-29  3119  			err = -EINVAL;
9e4e01dfd3254c KP Singh               2020-03-29  3120  			goto out_put_prog;
9e4e01dfd3254c KP Singh               2020-03-29  3121  		}
9e4e01dfd3254c KP Singh               2020-03-29  3122  		break;
9e4e01dfd3254c KP Singh               2020-03-29  3123  	case BPF_PROG_TYPE_LSM:
9e4e01dfd3254c KP Singh               2020-03-29  3124  		if (prog->expected_attach_type != BPF_LSM_MAC) {
9e4e01dfd3254c KP Singh               2020-03-29  3125  			err = -EINVAL;
9e4e01dfd3254c KP Singh               2020-03-29  3126  			goto out_put_prog;
9e4e01dfd3254c KP Singh               2020-03-29  3127  		}
9e4e01dfd3254c KP Singh               2020-03-29  3128  		break;
9e4e01dfd3254c KP Singh               2020-03-29  3129  	default:
fec56f5890d93f Alexei Starovoitov     2019-11-14  3130  		err = -EINVAL;
fec56f5890d93f Alexei Starovoitov     2019-11-14  3131  		goto out_put_prog;
fec56f5890d93f Alexei Starovoitov     2019-11-14  3132  	}
fec56f5890d93f Alexei Starovoitov     2019-11-14  3133  
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3134  	if (!!tgt_prog_fd != !!btf_id) {
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3135  		err = -EINVAL;
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3136  		goto out_put_prog;
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3137  	}
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3138  
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3139  	if (tgt_prog_fd) {
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3140  		/* For now we only allow new targets for BPF_PROG_TYPE_EXT */
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3141  		if (prog->type != BPF_PROG_TYPE_EXT) {
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3142  			err = -EINVAL;
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3143  			goto out_put_prog;
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3144  		}
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3145  
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3146  		tgt_prog = bpf_prog_get(tgt_prog_fd);
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3147  		if (IS_ERR(tgt_prog)) {
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3148  			err = PTR_ERR(tgt_prog);
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3149  			tgt_prog = NULL;
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3150  			goto out_put_prog;
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3151  		}
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3152  
22dc4a0f5ed11b Andrii Nakryiko        2020-12-03  3153  		key = bpf_trampoline_compute_key(tgt_prog, NULL, btf_id);
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3154  	}
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3155  
70ed506c3bbcfa Andrii Nakryiko        2020-03-02  3156  	link = kzalloc(sizeof(*link), GFP_USER);
70ed506c3bbcfa Andrii Nakryiko        2020-03-02  3157  	if (!link) {
70ed506c3bbcfa Andrii Nakryiko        2020-03-02  3158  		err = -ENOMEM;
70ed506c3bbcfa Andrii Nakryiko        2020-03-02  3159  		goto out_put_prog;
70ed506c3bbcfa Andrii Nakryiko        2020-03-02  3160  	}
f7e0beaf39d386 Kui-Feng Lee           2022-05-10  3161  	bpf_link_init(&link->link.link, BPF_LINK_TYPE_TRACING,
f2e10bff16a0fd Andrii Nakryiko        2020-04-28  3162  		      &bpf_tracing_link_lops, prog);
f2e10bff16a0fd Andrii Nakryiko        2020-04-28  3163  	link->attach_type = prog->expected_attach_type;
2fcc82411e74e5 Kui-Feng Lee           2022-05-10  3164  	link->link.cookie = bpf_cookie;
70ed506c3bbcfa Andrii Nakryiko        2020-03-02  3165  
3aac1ead5eb6b7 Toke Høiland-Jørgensen 2020-09-29  3166  	mutex_lock(&prog->aux->dst_mutex);
3aac1ead5eb6b7 Toke Høiland-Jørgensen 2020-09-29  3167  
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3168  	/* There are a few possible cases here:
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3169  	 *
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3170  	 * - if prog->aux->dst_trampoline is set, the program was just loaded
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3171  	 *   and not yet attached to anything, so we can use the values stored
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3172  	 *   in prog->aux
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3173  	 *
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3174  	 * - if prog->aux->dst_trampoline is NULL, the program has already been
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3175           *   attached to a target and its initial target was cleared (below)
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3176  	 *
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3177  	 * - if tgt_prog != NULL, the caller specified tgt_prog_fd +
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3178  	 *   target_btf_id using the link_create API.
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3179  	 *
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3180  	 * - if tgt_prog == NULL when this function was called using the old
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3181  	 *   raw_tracepoint_open API, and we need a target from prog->aux
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3182  	 *
f3a95075549e0e Jiri Olsa              2021-04-14  3183  	 * - if prog->aux->dst_trampoline and tgt_prog is NULL, the program
f3a95075549e0e Jiri Olsa              2021-04-14  3184  	 *   was detached and is going for re-attachment.

There is a giant comment here explaining how tgt_prog can be NULL. I read it but I lack some context.

4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29 3185 */

4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29 @3186  	if (!prog->aux->dst_trampoline && !tgt_prog) {
                                                                                                  ^^^^^^^^^
f3a95075549e0e Jiri Olsa              2021-04-14  3187  		/*
f3a95075549e0e Jiri Olsa              2021-04-14  3188  		 * Allow re-attach for TRACING and LSM programs. If it's
f3a95075549e0e Jiri Olsa              2021-04-14  3189  		 * currently linked, bpf_trampoline_link_prog will fail.
f3a95075549e0e Jiri Olsa              2021-04-14  3190  		 * EXT programs need to specify tgt_prog_fd, so they
f3a95075549e0e Jiri Olsa              2021-04-14  3191  		 * re-attach in separate code path.
f3a95075549e0e Jiri Olsa              2021-04-14  3192  		 */
f3a95075549e0e Jiri Olsa              2021-04-14  3193  		if (prog->type != BPF_PROG_TYPE_TRACING &&
f3a95075549e0e Jiri Olsa              2021-04-14  3194  		    prog->type != BPF_PROG_TYPE_LSM) {
f3a95075549e0e Jiri Olsa              2021-04-14  3195  			err = -EINVAL;
3aac1ead5eb6b7 Toke Høiland-Jørgensen 2020-09-29  3196  			goto out_unlock;
babf3164095b06 Andrii Nakryiko        2020-03-09  3197  		}
f3a95075549e0e Jiri Olsa              2021-04-14  3198  		btf_id = prog->aux->attach_btf_id;
f3a95075549e0e Jiri Olsa              2021-04-14  3199  		key = bpf_trampoline_compute_key(NULL, prog->aux->attach_btf, btf_id);
f3a95075549e0e Jiri Olsa              2021-04-14  3200  	}
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3201  
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3202  	if (!prog->aux->dst_trampoline ||
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3203  	    (key && key != prog->aux->dst_trampoline->key)) {
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3204  		/* If there is no saved target, or the specified target is
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3205  		 * different from the destination specified at load time, we
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3206  		 * need a new trampoline and a check for compatibility
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3207  		 */
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3208  		struct bpf_attach_target_info tgt_info = {};
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3209  
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3210  		err = bpf_check_attach_target(NULL, prog, tgt_prog, btf_id,
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3211  					      &tgt_info);
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3212  		if (err)
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3213  			goto out_unlock;
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3214  
31bf1dbccfb0a9 Viktor Malik           2023-03-10  3215  		if (tgt_info.tgt_mod) {
31bf1dbccfb0a9 Viktor Malik           2023-03-10  3216  			module_put(prog->aux->mod);
31bf1dbccfb0a9 Viktor Malik           2023-03-10  3217  			prog->aux->mod = tgt_info.tgt_mod;
31bf1dbccfb0a9 Viktor Malik           2023-03-10  3218  		}
31bf1dbccfb0a9 Viktor Malik           2023-03-10  3219  
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3220  		tr = bpf_trampoline_get(key, &tgt_info);
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3221  		if (!tr) {
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3222  			err = -ENOMEM;
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3223  			goto out_unlock;
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3224  		}
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3225  	} else {
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3226  		/* The caller didn't specify a target, or the target was the
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3227  		 * same as the destination supplied during program load. This
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3228  		 * means we can reuse the trampoline and reference from program
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3229  		 * load time, and there is no need to allocate a new one. This
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3230  		 * can only happen once for any program, as the saved values in
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3231  		 * prog->aux are cleared below.
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3232  		 */
3aac1ead5eb6b7 Toke Høiland-Jørgensen 2020-09-29  3233  		tr = prog->aux->dst_trampoline;
3aac1ead5eb6b7 Toke Høiland-Jørgensen 2020-09-29  3234  		tgt_prog = prog->aux->dst_prog;

tgt_prog is re-assigned here. Maybe the this covers all the NULL cases.

4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29 3235 }
3aac1ead5eb6b7 Toke Høiland-Jørgensen 2020-09-29 3236
f7e0beaf39d386 Kui-Feng Lee 2022-05-10 3237 err = bpf_link_prime(&link->link.link, &link_primer);

3aac1ead5eb6b7 Toke Høiland-Jørgensen 2020-09-29  3238  	if (err)
3aac1ead5eb6b7 Toke Høiland-Jørgensen 2020-09-29  3239  		goto out_unlock;
fec56f5890d93f Alexei Starovoitov     2019-11-14  3240  
26a0833438cdde Dmitrii Dolgov         2023-11-14 @3241  	tgt_prog->aux->followed = true;
                                                                ^^^^^^^^^^^^^^^^^^^^^^^

Unchecked dereference

26a0833438cdde Dmitrii Dolgov 2023-11-14 3242 prog->aux->attach_depth = tgt_prog->aux->attach_depth + 1;

Second unchecked dereference here

f7e0beaf39d386 Kui-Feng Lee           2022-05-10  3243  	err = bpf_trampoline_link_prog(&link->link, tr);
babf3164095b06 Andrii Nakryiko        2020-03-09  3244  	if (err) {
a3b80e1078943d Andrii Nakryiko        2020-04-28  3245  		bpf_link_cleanup(&link_primer);
3aac1ead5eb6b7 Toke Høiland-Jørgensen 2020-09-29  3246  		link = NULL;
3aac1ead5eb6b7 Toke Høiland-Jørgensen 2020-09-29  3247  		goto out_unlock;
fec56f5890d93f Alexei Starovoitov     2019-11-14  3248  	}
babf3164095b06 Andrii Nakryiko        2020-03-09  3249  
3aac1ead5eb6b7 Toke Høiland-Jørgensen 2020-09-29  3250  	link->tgt_prog = tgt_prog;
3aac1ead5eb6b7 Toke Høiland-Jørgensen 2020-09-29  3251  	link->trampoline = tr;
3aac1ead5eb6b7 Toke Høiland-Jørgensen 2020-09-29  3252  
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3253  	/* Always clear the trampoline and target prog from prog->aux to make
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3254  	 * sure the original attach destination is not kept alive after a
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3255  	 * program is (re-)attached to another target.
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3256  	 */
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3257  	if (prog->aux->dst_prog &&
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3258  	    (tgt_prog_fd || tr != prog->aux->dst_trampoline))
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3259  		/* got extra prog ref from syscall, or attaching to different prog */
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3260  		bpf_prog_put(prog->aux->dst_prog);
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3261  	if (prog->aux->dst_trampoline && tr != prog->aux->dst_trampoline)
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3262  		/* we allocated a new trampoline, so free the old one */
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3263  		bpf_trampoline_put(prog->aux->dst_trampoline);
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3264  
3aac1ead5eb6b7 Toke Høiland-Jørgensen 2020-09-29  3265  	prog->aux->dst_prog = NULL;
3aac1ead5eb6b7 Toke Høiland-Jørgensen 2020-09-29  3266  	prog->aux->dst_trampoline = NULL;
3aac1ead5eb6b7 Toke Høiland-Jørgensen 2020-09-29  3267  	mutex_unlock(&prog->aux->dst_mutex);
3aac1ead5eb6b7 Toke Høiland-Jørgensen 2020-09-29  3268  
a3b80e1078943d Andrii Nakryiko        2020-04-28  3269  	return bpf_link_settle(&link_primer);
3aac1ead5eb6b7 Toke Høiland-Jørgensen 2020-09-29  3270  out_unlock:
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3271  	if (tr && tr != prog->aux->dst_trampoline)
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3272  		bpf_trampoline_put(tr);
3aac1ead5eb6b7 Toke Høiland-Jørgensen 2020-09-29  3273  	mutex_unlock(&prog->aux->dst_mutex);
3aac1ead5eb6b7 Toke Høiland-Jørgensen 2020-09-29  3274  	kfree(link);
fec56f5890d93f Alexei Starovoitov     2019-11-14  3275  out_put_prog:
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3276  	if (tgt_prog_fd && tgt_prog)
4a1e7c0c63e02d Toke Høiland-Jørgensen 2020-09-29  3277  		bpf_prog_put(tgt_prog);
fec56f5890d93f Alexei Starovoitov     2019-11-14  3278  	return err;
fec56f5890d93f Alexei Starovoitov     2019-11-14  3279  }
fec56f5890d93f Alexei Starovoitov     2019-11-14  3280  
Signature · 2 lines
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
#3 lore raw
Quoted 15 lines · Dan Carpenter

Hi Dmitrii,

kernel test robot noticed the following build warnings:

url:    https://github.com/intel-lab-lkp/linux/commits/Dmitrii-Dolgov/bpf-Relax-tracing-prog-recursive-attach-rules/20231114-164631
base:   100888fb6d8a185866b1520031ee7e3182b173de
patch link:    https://lore.kernel.org/r/20231114084118.11095-1-9erthalion6%40gmail.com
patch subject: [RFC PATCH bpf-next] bpf: Relax tracing prog recursive attach rules
config: i386-randconfig-141-20231119 (https://download.01.org/0day-ci/archive/20231119/[email protected]/config)
compiler: gcc-12 (Debian 12.2.0-14) 12.2.0
reproduce: (https://download.01.org/0day-ci/archive/20231119/[email protected]/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags

Quoted 8 lines
| Reported-by: kernel test robot <[email protected]>
| Reported-by: Dan Carpenter <[email protected]>
| Closes: https://lore.kernel.org/r/[email protected]/

smatch warnings:
kernel/bpf/syscall.c:3241 bpf_tracing_prog_attach() error: we previously assumed 'tgt_prog' could be null (see line 3186)

vim +/tgt_prog +3241 kernel/bpf/syscall.c

Right, my mistake, thanks for letting me know. Looks like the condition branch where tgt_prog is assigned covers only reattachment, and it will blow when using old raw_tracepoint API. I'm going to send a new version of the patch soon.