Jiri Olsa reported a bug ([1]) in kernel where cgroup local
storage pointer may be NULL in bpf_get_local_storage() helper.
There are two issues uncovered by this bug:
(1). kprobe or tracepoint prog incorrectly sets cgroup local storage
before prog run,
(2). due to change from preempt_disable to migrate_disable,
preemption is possible and percpu storage might be overwritten
by other tasks.
This issue (1) is fixed in [2]. This patch tried to address issue (2).
The following shows how things can go wrong:
task 1: bpf_cgroup_storage_set() for percpu local storage
preemption happens
task 2: bpf_cgroup_storage_set() for percpu local storage
preemption happens
task 1: run bpf program
task 1 will effectively use the percpu local storage setting by task 2
which will be either NULL or incorrect ones.
Instead of just one common local storage per cpu, this patch fixed
the issue by permitting 8 local storages per cpu and each local
storage is identified by a task_struct pointer. This way, we
allow at most 8 nested preemption between bpf_cgroup_storage_set()
and bpf_cgroup_storage_unset(). The percpu local storage slot
is released (calling bpf_cgroup_storage_unset()) by the same task
after bpf program finished running.
bpf_test_run() is also fixed to use the new bpf_cgroup_storage_set()
interface.
The patch is tested on top of [2] with reproducer in [1].
Without this patch, kernel will emit error in 2-3 minutes.
With this patch, after one hour, still no error.
[1] https://lore.kernel.org/bpf/CAKH8qBuXCfUz=w8L+Fj74OaUpbosO29niYwTki7e3Ag044_aww@mail.gmail.com/T
[2] https://lore.kernel.org/bpf/
20210309185028.
3763817-1-yhs@fb.com
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Roman Gushchin <guro@fb.com>
Link: https://lore.kernel.org/bpf/20210323055146.3334476-1-yhs@fb.com
struct bpf_cgroup_storage;
struct ctl_table;
struct ctl_table_header;
+struct task_struct;
#ifdef CONFIG_CGROUP_BPF
extern struct static_key_false cgroup_bpf_enabled_key[MAX_BPF_ATTACH_TYPE];
#define cgroup_bpf_enabled(type) static_branch_unlikely(&cgroup_bpf_enabled_key[type])
-DECLARE_PER_CPU(struct bpf_cgroup_storage*,
- bpf_cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]);
+#define BPF_CGROUP_STORAGE_NEST_MAX 8
+
+struct bpf_cgroup_storage_info {
+ struct task_struct *task;
+ struct bpf_cgroup_storage *storage[MAX_BPF_CGROUP_STORAGE_TYPE];
+};
+
+/* For each cpu, permit maximum BPF_CGROUP_STORAGE_NEST_MAX number of tasks
+ * to use bpf cgroup storage simultaneously.
+ */
+DECLARE_PER_CPU(struct bpf_cgroup_storage_info,
+ bpf_cgroup_storage_info[BPF_CGROUP_STORAGE_NEST_MAX]);
#define for_each_cgroup_storage_type(stype) \
for (stype = 0; stype < MAX_BPF_CGROUP_STORAGE_TYPE; stype++)
return BPF_CGROUP_STORAGE_SHARED;
}
-static inline void bpf_cgroup_storage_set(struct bpf_cgroup_storage
- *storage[MAX_BPF_CGROUP_STORAGE_TYPE])
+static inline int bpf_cgroup_storage_set(struct bpf_cgroup_storage
+ *storage[MAX_BPF_CGROUP_STORAGE_TYPE])
{
enum bpf_cgroup_storage_type stype;
+ int i, err = 0;
+
+ preempt_disable();
+ for (i = 0; i < BPF_CGROUP_STORAGE_NEST_MAX; i++) {
+ if (unlikely(this_cpu_read(bpf_cgroup_storage_info[i].task) != NULL))
+ continue;
+
+ this_cpu_write(bpf_cgroup_storage_info[i].task, current);
+ for_each_cgroup_storage_type(stype)
+ this_cpu_write(bpf_cgroup_storage_info[i].storage[stype],
+ storage[stype]);
+ goto out;
+ }
+ err = -EBUSY;
+ WARN_ON_ONCE(1);
+
+out:
+ preempt_enable();
+ return err;
+}
+
+static inline void bpf_cgroup_storage_unset(void)
+{
+ int i;
+
+ for (i = 0; i < BPF_CGROUP_STORAGE_NEST_MAX; i++) {
+ if (unlikely(this_cpu_read(bpf_cgroup_storage_info[i].task) != current))
+ continue;
- for_each_cgroup_storage_type(stype)
- this_cpu_write(bpf_cgroup_storage[stype], storage[stype]);
+ this_cpu_write(bpf_cgroup_storage_info[i].task, NULL);
+ return;
+ }
}
struct bpf_cgroup_storage *
return -EINVAL;
}
-static inline void bpf_cgroup_storage_set(
- struct bpf_cgroup_storage *storage[MAX_BPF_CGROUP_STORAGE_TYPE]) {}
+static inline int bpf_cgroup_storage_set(
+ struct bpf_cgroup_storage *storage[MAX_BPF_CGROUP_STORAGE_TYPE]) { return 0; }
+static inline void bpf_cgroup_storage_unset(void) {}
static inline int bpf_cgroup_storage_assign(struct bpf_prog_aux *aux,
struct bpf_map *map) { return 0; }
static inline struct bpf_cgroup_storage *bpf_cgroup_storage_alloc(
/* BPF program asks to set CN on the packet. */
#define BPF_RET_SET_CN (1 << 0)
+/* For BPF_PROG_RUN_ARRAY_FLAGS and __BPF_PROG_RUN_ARRAY,
+ * if bpf_cgroup_storage_set() failed, the rest of programs
+ * will not execute. This should be a really rare scenario
+ * as it requires BPF_CGROUP_STORAGE_NEST_MAX number of
+ * preemptions all between bpf_cgroup_storage_set() and
+ * bpf_cgroup_storage_unset() on the same cpu.
+ */
#define BPF_PROG_RUN_ARRAY_FLAGS(array, ctx, func, ret_flags) \
({ \
struct bpf_prog_array_item *_item; \
_array = rcu_dereference(array); \
_item = &_array->items[0]; \
while ((_prog = READ_ONCE(_item->prog))) { \
- bpf_cgroup_storage_set(_item->cgroup_storage); \
+ if (unlikely(bpf_cgroup_storage_set(_item->cgroup_storage))) \
+ break; \
func_ret = func(_prog, ctx); \
_ret &= (func_ret & 1); \
*(ret_flags) |= (func_ret >> 1); \
+ bpf_cgroup_storage_unset(); \
_item++; \
} \
rcu_read_unlock(); \
goto _out; \
_item = &_array->items[0]; \
while ((_prog = READ_ONCE(_item->prog))) { \
- if (set_cg_storage) \
- bpf_cgroup_storage_set(_item->cgroup_storage); \
- _ret &= func(_prog, ctx); \
+ if (!set_cg_storage) { \
+ _ret &= func(_prog, ctx); \
+ } else { \
+ if (unlikely(bpf_cgroup_storage_set(_item->cgroup_storage))) \
+ break; \
+ _ret &= func(_prog, ctx); \
+ bpf_cgroup_storage_unset(); \
+ } \
_item++; \
} \
_out: \
};
#ifdef CONFIG_CGROUP_BPF
-DECLARE_PER_CPU(struct bpf_cgroup_storage*,
- bpf_cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]);
+DECLARE_PER_CPU(struct bpf_cgroup_storage_info,
+ bpf_cgroup_storage_info[BPF_CGROUP_STORAGE_NEST_MAX]);
BPF_CALL_2(bpf_get_local_storage, struct bpf_map *, map, u64, flags)
{
* verifier checks that its value is correct.
*/
enum bpf_cgroup_storage_type stype = cgroup_storage_type(map);
- struct bpf_cgroup_storage *storage;
+ struct bpf_cgroup_storage *storage = NULL;
void *ptr;
+ int i;
- storage = this_cpu_read(bpf_cgroup_storage[stype]);
+ for (i = 0; i < BPF_CGROUP_STORAGE_NEST_MAX; i++) {
+ if (unlikely(this_cpu_read(bpf_cgroup_storage_info[i].task) != current))
+ continue;
+
+ storage = this_cpu_read(bpf_cgroup_storage_info[i].storage[stype]);
+ break;
+ }
if (stype == BPF_CGROUP_STORAGE_SHARED)
ptr = &READ_ONCE(storage->buf)->data[0];
#include <linux/slab.h>
#include <uapi/linux/btf.h>
-DEFINE_PER_CPU(struct bpf_cgroup_storage*, bpf_cgroup_storage[MAX_BPF_CGROUP_STORAGE_TYPE]);
-
#ifdef CONFIG_CGROUP_BPF
+DEFINE_PER_CPU(struct bpf_cgroup_storage_info,
+ bpf_cgroup_storage_info[BPF_CGROUP_STORAGE_NEST_MAX]);
+
#include "../cgroup/cgroup-internal.h"
#define LOCAL_STORAGE_CREATE_FLAG_MASK \
bpf_test_timer_enter(&t);
do {
- bpf_cgroup_storage_set(storage);
+ ret = bpf_cgroup_storage_set(storage);
+ if (ret)
+ break;
if (xdp)
*retval = bpf_prog_run_xdp(prog, ctx);
else
*retval = BPF_PROG_RUN(prog, ctx);
+
+ bpf_cgroup_storage_unset();
} while (bpf_test_timer_continue(&t, repeat, &ret, time));
bpf_test_timer_leave(&t);