The required protection is that the caller cannot be migrated to a
different CPU as these functions end up in places which take either a hash
bucket lock or might trigger a kprobe inside the memory allocator. Both
scenarios can lead to deadlocks. The deadlock prevention is per CPU by
incrementing a per CPU variable which temporarily blocks the invocation of
BPF programs from perf and kprobes.
Replace the preempt_disable/enable() pairs with migrate_disable/enable()
pairs to prepare BPF to work on PREEMPT_RT enabled kernels. On a non-RT
kernel this maps to preempt_disable/enable(), i.e. no functional change.
Signed-off-by: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
---
kernel/bpf/syscall.c | 16 ++++++++--------
1 file changed, 8 insertions(+), 8 deletions(-)
--- a/kernel/bpf/syscall.c
+++ b/kernel/bpf/syscall.c
@@ -174,7 +174,7 @@ static int bpf_map_update_value(struct b
/* must increment bpf_prog_active to avoid kprobe+bpf triggering from
* inside bpf map update or delete otherwise deadlocks are possible
*/
- preempt_disable();
+ migrate_disable();
__this_cpu_inc(bpf_prog_active);
if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH ||
map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH) {
@@ -207,7 +207,7 @@ static int bpf_map_update_value(struct b
rcu_read_unlock();
}
__this_cpu_dec(bpf_prog_active);
- preempt_enable();
+ migrate_enable();
maybe_wait_bpf_programs(map);
return err;
@@ -222,7 +222,7 @@ static int bpf_map_copy_value(struct bpf
if (bpf_map_is_dev_bound(map))
return bpf_map_offload_lookup_elem(map, key, value);
- preempt_disable();
+ migrate_disable();
this_cpu_inc(bpf_prog_active);
if (map->map_type == BPF_MAP_TYPE_PERCPU_HASH ||
map->map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH) {
@@ -269,7 +269,7 @@ static int bpf_map_copy_value(struct bpf
}
this_cpu_dec(bpf_prog_active);
- preempt_enable();
+ migrate_enable();
maybe_wait_bpf_programs(map);
return err;
@@ -1136,13 +1136,13 @@ static int map_delete_elem(union bpf_att
goto out;
}
- preempt_disable();
+ migrate_disable();
__this_cpu_inc(bpf_prog_active);
rcu_read_lock();
err = map->ops->map_delete_elem(map, key);
rcu_read_unlock();
__this_cpu_dec(bpf_prog_active);
- preempt_enable();
+ migrate_enable();
maybe_wait_bpf_programs(map);
out:
kfree(key);
@@ -1254,13 +1254,13 @@ int generic_map_delete_batch(struct bpf_
break;
}
- preempt_disable();
+ migrate_disable();
__this_cpu_inc(bpf_prog_active);
rcu_read_lock();
err = map->ops->map_delete_elem(map, key);
rcu_read_unlock();
__this_cpu_dec(bpf_prog_active);
- preempt_enable();
+ migrate_enable();
maybe_wait_bpf_programs(map);
if (err)
break;
