bpf: fix mlock precharge on arraymaps
Subsystems (3)
Daniel Borkmann
maintainerPATCH 1/1 bpf: fix mlock precharge on arraymaps
syzkaller recently triggered OOM during percpu map allocation; while there is work in progress by Dennis Zhou to add __GFP_NORETRY semantics for percpu allocator under pressure, there seems also a missing bpf_map_precharge_memlock() check in array map allocation.
Given today the actual bpf_map_charge_memlock() happens after the find_and_alloc_map() in syscall path, the bpf_map_precharge_memlock() is there to bail out early before we go and do the map setup work when we find that we hit the limits anyway. Therefore add this for array map as well.
--- kernel/bpf/arraymap.c | 28 ++++++++++++++++------------ 1 file changed, 16 insertions(+), 12 deletions(-)
kernel/bpf/arraymap.c +16 -12
@@ -73,11 +73,11 @@ static int array_map_alloc_check(union bpf_attr *attr)static struct bpf_map *array_map_alloc (union bpf_attr *attr) {bool percpu = attr->map_type ==BPF_MAP_TYPE_PERCPU_ARRAY ; -int numa_node =bpf_map_attr_numa_node (attr); +int ret, numa_node =bpf_map_attr_numa_node (attr);u32 elem_size, index_mask, max_entries;bool unpriv = !capable (CAP_SYS_ADMIN ); +u64 cost, array_size, mask64;struct bpf_array *array; -u64 array_size, mask64; elem_size =round_up (attr->value_size,8 );
@@ -109,8 +109,19 @@ static struct bpf_map *array_map_alloc(union bpf_attr *attr) array_size += (u64 ) max_entries * elem_size;/* make sure there is no u32 overflow later in round_up() */ -if (array_size >=U32_MAX -PAGE_SIZE ) + cost = array_size; +if (cost >=U32_MAX -PAGE_SIZE )return ERR_PTR (-ENOMEM ); +if (percpu) { + cost += (u64 )attr->max_entries * elem_size *num_possible_cpus (); +if (cost >=U32_MAX -PAGE_SIZE ) +return ERR_PTR (-ENOMEM ); + } + cost =round_up (cost,PAGE_SIZE ) >>PAGE_SHIFT ; + + ret =bpf_map_precharge_memlock (cost); +if (ret <0 ) +return ERR_PTR (ret);/* allocate all map elements and zero-initialize them */ array =bpf_map_area_alloc (array_size, numa_node);
@@ -121,20 +132,13 @@ static struct bpf_map *array_map_alloc(union bpf_attr *attr)/* copy mandatory map attributes */ bpf_map_init_from_attr (&array->map, attr); + array->map.pages = cost; array->elem_size = elem_size; -if (!percpu) - goto out; - - array_size += (u64 ) attr->max_entries * elem_size *num_possible_cpus (); - -if (array_size >=U32_MAX -PAGE_SIZE || -bpf_array_alloc_percpu (array)) { +if (percpu &&bpf_array_alloc_percpu (array)) {bpf_map_area_free (array);return ERR_PTR (-ENOMEM ); } -out: - array->map.pages =round_up (array_size,PAGE_SIZE ) >>PAGE_SHIFT ;return &array->map; }
Signature · 1 line
2.9.5
Alexei Starovoitov
AppliedQuoted 11 lines · Daniel Borkmann
syzkaller recently triggered OOM during percpu map allocation; while there is work in progress by Dennis Zhou to add
__GFP_NORETRYsemantics for percpu allocator under pressure, there seems also a missingbpf_map_precharge_memlock()check in array map allocation.Given today the actual
bpf_map_charge_memlock()happens after thefind_and_alloc_map()in syscall path, thebpf_map_precharge_memlock()is there to bail out early before we go and do the map setup work when we find that we hit the limits anyway. Therefore add this for array map as well.
Quoted 5 lines
Fixes 6c9059817432 ("bpf: pre-allocate hash map elements")Fixes a10423b87a7e ("bpf: introduce BPF_MAP_TYPE_PERCPU_ARRAY map")Reported-by [email protected]Signed-off-by Daniel Borkmann <[email protected]>Cc Dennis Zhou <[email protected]>
Applied to bpf tree, Thanks Daniel.