Skip to contentSkip to conversation

bpf: Reject struct_ops registration that uses module ptr and the module btf_id is missing

PATCH Applied Martin KaFai Lau 3 participants 4 messages Dec 20, 2024 – Jan 7, 2025 BPF

lore
Share & tools
lore

Martin KaFai Lau

author reviewer
#0 lore raw

PATCH 1/1 bpf: Reject struct_ops registration that uses module ptr and the module btf_id is missing

From: Martin KaFai Lau <[email protected]>

There is a UAF report in the bpf_struct_ops when CONFIG_MODULES=n. In particular, the report is on tcp_congestion_ops that has a "struct module *owner" member.

For struct_ops that has a "struct module *owner" member, it can be extended either by the regular kernel module or by the bpf_struct_ops. bpf_try_module_get() will be used to do the refcounting and different refcount is done based on the owner pointer. When CONFIG_MODULES=n, the btf_id of the "struct module" is missing:

WARN: resolve_btfids: unresolved symbol module

Thus, the bpf_try_module_get() cannot do the correct refcounting.

Not all subsystem's struct_ops requires the "struct module *owner" member. e.g. the recent sched_ext_ops.

This patch is to disable bpf_struct_ops registration if the struct_ops has the "struct module *" member and the "struct module" btf_id is missing. The btf_type_is_fwd() helper is moved to the btf.h header file for this test.

This has happened since the beginning of bpf_struct_ops which has gone through many changes. The Fixes tag is set to a recent commit that this patch can apply cleanly. Considering CONFIG_MODULES=n is not common and the age of the issue, targeting for bpf-next also.

Fixes 1611603537a4 ("bpf: Create argument information for nullable arguments.")
---
 include/linux/btf.h         |  5 +++++
 kernel/bpf/bpf_struct_ops.c | 21 +++++++++++++++++++++
 kernel/bpf/btf.c            |  5 -----
 3 files changed, 26 insertions(+), 5 deletions(-)
include/linux/btf.h +5 -0
index 4214e76c9168..2a08a2b55592 100644
@@ -353,6 +353,11 @@ static inline bool btf_type_is_scalar(const struct btf_type *t)
 	return btf_type_is_int(t) || btf_type_is_enum(t);
 }
 
+static inline bool btf_type_is_fwd(const struct btf_type *t)
+{
+	return BTF_INFO_KIND(t->info) == BTF_KIND_FWD;
+}
+
 static inline bool btf_type_is_typedef(const struct btf_type *t)
 {
 	return BTF_INFO_KIND(t->info) == BTF_KIND_TYPEDEF;
kernel/bpf/bpf_struct_ops.c +21 -0
index 606efe32485a..040fb1cd840b 100644
@@ -310,6 +310,20 @@ void bpf_struct_ops_desc_release(struct bpf_struct_ops_desc *st_ops_desc)
 	kfree(arg_info);
 }
 
+static bool is_module_member(const struct btf *btf, u32 id)
+{
+	const struct btf_type *t;
+
+	t = btf_type_resolve_ptr(btf, id, NULL);
+	if (!t)
+		return false;
+
+	if (!__btf_type_is_struct(t) && !btf_type_is_fwd(t))
+		return false;
+
+	return !strcmp(btf_name_by_offset(btf, t->name_off), "module");
+}
+
 int bpf_struct_ops_desc_init(struct bpf_struct_ops_desc *st_ops_desc,
 			     struct btf *btf,
 			     struct bpf_verifier_log *log)
@@ -389,6 +403,13 @@ int bpf_struct_ops_desc_init(struct bpf_struct_ops_desc *st_ops_desc,
 			goto errout;
 		}
 
+		if (!st_ops_ids[IDX_MODULE_ID] && is_module_member(btf, member->type)) {
+			pr_warn("'struct module' btf id not found. Is CONFIG_MODULES enabled? bpf_struct_ops '%s' needs module support.\n",
+				st_ops->name);
+			err = -EOPNOTSUPP;
+			goto errout;
+		}
+
 		func_proto = btf_type_resolve_func_ptr(btf,
 						       member->type,
 						       NULL);
kernel/bpf/btf.c +0 -5
index 28246c59e12e..8396ce1d0fba 100644
@@ -498,11 +498,6 @@ bool btf_type_is_void(const struct btf_type *t)
 	return t == &btf_void;
 }
 
-static bool btf_type_is_fwd(const struct btf_type *t)
-{
-	return BTF_INFO_KIND(t->info) == BTF_KIND_FWD;
-}
-
 static bool btf_type_is_datasec(const struct btf_type *t)
 {
 	return BTF_INFO_KIND(t->info) == BTF_KIND_DATASEC;
Signature · 1 line
2.43.5
maintainer Tested-by Acked-by on v1 1/1
#1 lore raw
Quoted 30 lines · Martin KaFai Lau

From: Martin KaFai Lau <[email protected]>

There is a UAF report in the bpf_struct_ops when CONFIG_MODULES=n. In particular, the report is on tcp_congestion_ops that has a "struct module *owner" member.

For struct_ops that has a "struct module *owner" member, it can be extended either by the regular kernel module or by the bpf_struct_ops. bpf_try_module_get() will be used to do the refcounting and different refcount is done based on the owner pointer. When CONFIG_MODULES=n, the btf_id of the "struct module" is missing:

WARN: resolve_btfids: unresolved symbol module

Thus, the bpf_try_module_get() cannot do the correct refcounting.

Not all subsystem's struct_ops requires the "struct module *owner" member. e.g. the recent sched_ext_ops.

This patch is to disable bpf_struct_ops registration if the struct_ops has the "struct module *" member and the "struct module" btf_id is missing. The btf_type_is_fwd() helper is moved to the btf.h header file for this test.

This has happened since the beginning of bpf_struct_ops which has gone through many changes. The Fixes tag is set to a recent commit that this patch can apply cleanly. Considering CONFIG_MODULES=n is not common and the age of the issue, targeting for bpf-next also.

Quoted 5 lines
Fixes 1611603537a4 ("bpf: Create argument information for nullable arguments.")

---

Looks like this fix had not landed yet.
I tried it and id does fix the error reported in the "closes" link.

It was a bit hard for me to figure out what went wrong from the description, could you please double-check my understanding below?
- when struct_ops program is attached, bpf_struct_ops_map_update_elem() scans every member of specific struct_ops type (e.g. struct tcp_congestion_ops) looking for fields with type 'struct module *';
- to find these fields BTF id of 'struct module' is used, this id does not exist when CONFIG_MODULES=n, bpf_struct_ops_map_update_elem() does not check if 'struct module' BTF id is non-zero;
- bpf_struct_ops_map_update_elem() initializes 'struct module *' fields using a magic value BPF_MODULE_OWNER, this initialization would not happen if fields are not found;
- later bpf_try_module_get() is called by code specific to particular struct_ops, e.g. from tcp_cong.c:tcp_assign_congestion_control();
- the bpf_try_module_get() is implemented as follows:

    static inline bool bpf_try_module_get(const void *data, struct module *owner)
    {
    	if (owner == BPF_MODULE_OWNER)
    		return bpf_struct_ops_get(data);
    	else
    		return try_module_get(owner);
    }

if 'struct module *' fields are not correctly initialized as BPF_MODULE_OWNER the bpf_try_module_get() executes try_module_get() passing a bogus pointer to it.

Assuming the above is correct, the fix lgtm.

[...]

author reviewer
#2 lore raw
Quoted 4 lines · Eduard Zingerman

It was a bit hard for me to figure out what went wrong from the description, could you please double-check my understanding below?
- when struct_ops program is attached, bpf_struct_ops_map_update_elem() scans every member of specific

Quoted 8 lines

struct_ops type (e.g. struct tcp_congestion_ops) looking for fields with type 'struct module *';
- to find these fields BTF id of 'struct module' is used, this id does not exist when CONFIG_MODULES=n, bpf_struct_ops_map_update_elem() does not check if 'struct module' BTF id is non-zero;
- bpf_struct_ops_map_update_elem() initializes 'struct module *' fields using a magic value BPF_MODULE_OWNER, this initialization would not happen if fields are not found;

Thanks for reviewing and testing it!

Yes, the understanding is correct. These are pretty much the only two places where st_ops_ids[IDX_MODULE_ID] and BPF_MODULE_OWNER are used.

#3 lore raw
In reply to Martin KaFai Lau
This patch was applied to bpf/bpf-next.git (master) by Alexei Starovoitov <[email protected]>:

Hello:

This patch was applied to bpf/bpf-next.git (master) by Alexei Starovoitov <[email protected]>:

Quoted 6 lines · On Fri, 20 Dec 2024 12:18:18 -0800 you wrote:

From: Martin KaFai Lau <[email protected]>

There is a UAF report in the bpf_struct_ops when CONFIG_MODULES=n. In particular, the report is on tcp_congestion_ops that has a "struct module *owner" member.

Quoted 8 lines

For struct_ops that has a "struct module *owner" member, it can be extended either by the regular kernel module or by the bpf_struct_ops. bpf_try_module_get() will be used to do the refcounting and different refcount is done based on the owner pointer. When CONFIG_MODULES=n, the btf_id of the "struct module" is missing:

[...]

Here is the summary with links:
- [bpf-next] bpf: Reject struct_ops registration that uses module ptr and the module btf_id is missing

    https://git.kernel.org/bpf/bpf-next/c/96ea081ed52b

You are awesome, thank you!

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