bpf: allow access into map value arrays
authorJosef Bacik <jbacik@fb.com>
Wed, 28 Sep 2016 14:54:32 +0000 (10:54 -0400)
committerDavid S. Miller <davem@davemloft.net>
Thu, 29 Sep 2016 05:35:35 +0000 (01:35 -0400)
Suppose you have a map array value that is something like this

struct foo {
unsigned iter;
int array[SOME_CONSTANT];
};

You can easily insert this into an array, but you cannot modify the contents of
foo->array[] after the fact.  This is because we have no way to verify we won't
go off the end of the array at verification time.  This patch provides a start
for this work.  We accomplish this by keeping track of a minimum and maximum
value a register could be while we're checking the code.  Then at the time we
try to do an access into a MAP_VALUE we verify that the maximum offset into that
region is a valid access into that memory region.  So in practice, code such as
this

unsigned index = 0;

if (foo->iter >= SOME_CONSTANT)
foo->iter = index;
else
index = foo->iter++;
foo->array[index] = bar;

would be allowed, as we can verify that index will always be between 0 and
SOME_CONSTANT-1.  If you wish to use signed values you'll have to have an extra
check to make sure the index isn't less than 0, or do something like index %=
SOME_CONSTANT.

Signed-off-by: Josef Bacik <jbacik@fb.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
include/linux/bpf.h
include/linux/bpf_verifier.h
kernel/bpf/verifier.c
samples/bpf/libbpf.h
samples/bpf/test_verifier.c

index 5691fdc..c201017 100644 (file)
@@ -139,6 +139,13 @@ enum bpf_reg_type {
         */
        PTR_TO_PACKET,
        PTR_TO_PACKET_END,       /* skb->data + headlen */
+
+       /* PTR_TO_MAP_VALUE_ADJ is used for doing pointer math inside of a map
+        * elem value.  We only allow this if we can statically verify that
+        * access from this register are going to fall within the size of the
+        * map element.
+        */
+       PTR_TO_MAP_VALUE_ADJ,
 };
 
 struct bpf_prog;
index c5cb661..7035b99 100644 (file)
 #include <linux/bpf.h> /* for enum bpf_reg_type */
 #include <linux/filter.h> /* for MAX_BPF_STACK */
 
+ /* Just some arbitrary values so we can safely do math without overflowing and
+  * are obviously wrong for any sort of memory access.
+  */
+#define BPF_REGISTER_MAX_RANGE (1024 * 1024 * 1024)
+#define BPF_REGISTER_MIN_RANGE -(1024 * 1024 * 1024)
+
 struct bpf_reg_state {
        enum bpf_reg_type type;
+       /*
+        * Used to determine if any memory access using this register will
+        * result in a bad access.
+        */
+       u64 min_value, max_value;
        union {
                /* valid when type == CONST_IMM | PTR_TO_STACK | UNKNOWN_VALUE */
                s64 imm;
@@ -81,6 +92,7 @@ struct bpf_verifier_env {
        u32 id_gen;                     /* used to generate unique reg IDs */
        bool allow_ptr_leaks;
        bool seen_direct_write;
+       bool varlen_map_value_access;
        struct bpf_insn_aux_data *insn_aux_data; /* array of per-insn state */
 };
 
index 7ada315..99a7e5b 100644 (file)
@@ -182,6 +182,7 @@ static const char * const reg_type_str[] = {
        [CONST_PTR_TO_MAP]      = "map_ptr",
        [PTR_TO_MAP_VALUE]      = "map_value",
        [PTR_TO_MAP_VALUE_OR_NULL] = "map_value_or_null",
+       [PTR_TO_MAP_VALUE_ADJ]  = "map_value_adj",
        [FRAME_PTR]             = "fp",
        [PTR_TO_STACK]          = "fp",
        [CONST_IMM]             = "imm",
@@ -209,10 +210,17 @@ static void print_verifier_state(struct bpf_verifier_state *state)
                else if (t == UNKNOWN_VALUE && reg->imm)
                        verbose("%lld", reg->imm);
                else if (t == CONST_PTR_TO_MAP || t == PTR_TO_MAP_VALUE ||
-                        t == PTR_TO_MAP_VALUE_OR_NULL)
+                        t == PTR_TO_MAP_VALUE_OR_NULL ||
+                        t == PTR_TO_MAP_VALUE_ADJ)
                        verbose("(ks=%d,vs=%d)",
                                reg->map_ptr->key_size,
                                reg->map_ptr->value_size);
+               if (reg->min_value != BPF_REGISTER_MIN_RANGE)
+                       verbose(",min_value=%llu",
+                               (unsigned long long)reg->min_value);
+               if (reg->max_value != BPF_REGISTER_MAX_RANGE)
+                       verbose(",max_value=%llu",
+                               (unsigned long long)reg->max_value);
        }
        for (i = 0; i < MAX_BPF_STACK; i += BPF_REG_SIZE) {
                if (state->stack_slot_type[i] == STACK_SPILL)
@@ -424,6 +432,8 @@ static void init_reg_state(struct bpf_reg_state *regs)
        for (i = 0; i < MAX_BPF_REG; i++) {
                regs[i].type = NOT_INIT;
                regs[i].imm = 0;
+               regs[i].min_value = BPF_REGISTER_MIN_RANGE;
+               regs[i].max_value = BPF_REGISTER_MAX_RANGE;
        }
 
        /* frame pointer */
@@ -440,6 +450,12 @@ static void mark_reg_unknown_value(struct bpf_reg_state *regs, u32 regno)
        regs[regno].imm = 0;
 }
 
+static void reset_reg_range_values(struct bpf_reg_state *regs, u32 regno)
+{
+       regs[regno].min_value = BPF_REGISTER_MIN_RANGE;
+       regs[regno].max_value = BPF_REGISTER_MAX_RANGE;
+}
+
 enum reg_arg_type {
        SRC_OP,         /* register is used as source operand */
        DST_OP,         /* register is used as destination operand */
@@ -665,7 +681,7 @@ static bool is_pointer_value(struct bpf_verifier_env *env, int regno)
 static int check_ptr_alignment(struct bpf_verifier_env *env,
                               struct bpf_reg_state *reg, int off, int size)
 {
-       if (reg->type != PTR_TO_PACKET) {
+       if (reg->type != PTR_TO_PACKET && reg->type != PTR_TO_MAP_VALUE_ADJ) {
                if (off % size != 0) {
                        verbose("misaligned access off %d size %d\n",
                                off, size);
@@ -675,16 +691,6 @@ static int check_ptr_alignment(struct bpf_verifier_env *env,
                }
        }
 
-       switch (env->prog->type) {
-       case BPF_PROG_TYPE_SCHED_CLS:
-       case BPF_PROG_TYPE_SCHED_ACT:
-       case BPF_PROG_TYPE_XDP:
-               break;
-       default:
-               verbose("verifier is misconfigured\n");
-               return -EACCES;
-       }
-
        if (IS_ENABLED(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS))
                /* misaligned access to packet is ok on x86,arm,arm64 */
                return 0;
@@ -695,7 +701,8 @@ static int check_ptr_alignment(struct bpf_verifier_env *env,
        }
 
        /* skb->data is NET_IP_ALIGN-ed */
-       if ((NET_IP_ALIGN + reg->off + off) % size != 0) {
+       if (reg->type == PTR_TO_PACKET &&
+           (NET_IP_ALIGN + reg->off + off) % size != 0) {
                verbose("misaligned packet access off %d+%d+%d size %d\n",
                        NET_IP_ALIGN, reg->off, off, size);
                return -EACCES;
@@ -728,12 +735,52 @@ static int check_mem_access(struct bpf_verifier_env *env, u32 regno, int off,
        if (err)
                return err;
 
-       if (reg->type == PTR_TO_MAP_VALUE) {
+       if (reg->type == PTR_TO_MAP_VALUE ||
+           reg->type == PTR_TO_MAP_VALUE_ADJ) {
                if (t == BPF_WRITE && value_regno >= 0 &&
                    is_pointer_value(env, value_regno)) {
                        verbose("R%d leaks addr into map\n", value_regno);
                        return -EACCES;
                }
+
+               /* If we adjusted the register to this map value at all then we
+                * need to change off and size to min_value and max_value
+                * respectively to make sure our theoretical access will be
+                * safe.
+                */
+               if (reg->type == PTR_TO_MAP_VALUE_ADJ) {
+                       if (log_level)
+                               print_verifier_state(state);
+                       env->varlen_map_value_access = true;
+                       /* The minimum value is only important with signed
+                        * comparisons where we can't assume the floor of a
+                        * value is 0.  If we are using signed variables for our
+                        * index'es we need to make sure that whatever we use
+                        * will have a set floor within our range.
+                        */
+                       if ((s64)reg->min_value < 0) {
+                               verbose("R%d min value is negative, either use unsigned index or do a if (index >=0) check.\n",
+                                       regno);
+                               return -EACCES;
+                       }
+                       err = check_map_access(env, regno, reg->min_value + off,
+                                              size);
+                       if (err) {
+                               verbose("R%d min value is outside of the array range\n",
+                                       regno);
+                               return err;
+                       }
+
+                       /* If we haven't set a max value then we need to bail
+                        * since we can't be sure we won't do bad things.
+                        */
+                       if (reg->max_value == BPF_REGISTER_MAX_RANGE) {
+                               verbose("R%d unbounded memory access, make sure to bounds check any array access into a map\n",
+                                       regno);
+                               return -EACCES;
+                       }
+                       off += reg->max_value;
+               }
                err = check_map_access(env, regno, off, size);
                if (!err && t == BPF_READ && value_regno >= 0)
                        mark_reg_unknown_value(state->regs, value_regno);
@@ -1195,6 +1242,7 @@ static int check_call(struct bpf_verifier_env *env, int func_id)
                regs[BPF_REG_0].type = NOT_INIT;
        } else if (fn->ret_type == RET_PTR_TO_MAP_VALUE_OR_NULL) {
                regs[BPF_REG_0].type = PTR_TO_MAP_VALUE_OR_NULL;
+               regs[BPF_REG_0].max_value = regs[BPF_REG_0].min_value = 0;
                /* remember map_ptr, so that check_map_access()
                 * can check 'value_size' boundary of memory access
                 * to map element returned from bpf_map_lookup_elem()
@@ -1416,6 +1464,106 @@ static int evaluate_reg_imm_alu(struct bpf_verifier_env *env,
        return 0;
 }
 
+static void check_reg_overflow(struct bpf_reg_state *reg)
+{
+       if (reg->max_value > BPF_REGISTER_MAX_RANGE)
+               reg->max_value = BPF_REGISTER_MAX_RANGE;
+       if ((s64)reg->min_value < BPF_REGISTER_MIN_RANGE)
+               reg->min_value = BPF_REGISTER_MIN_RANGE;
+}
+
+static void adjust_reg_min_max_vals(struct bpf_verifier_env *env,
+                                   struct bpf_insn *insn)
+{
+       struct bpf_reg_state *regs = env->cur_state.regs, *dst_reg;
+       u64 min_val = BPF_REGISTER_MIN_RANGE, max_val = BPF_REGISTER_MAX_RANGE;
+       bool min_set = false, max_set = false;
+       u8 opcode = BPF_OP(insn->code);
+
+       dst_reg = &regs[insn->dst_reg];
+       if (BPF_SRC(insn->code) == BPF_X) {
+               check_reg_overflow(&regs[insn->src_reg]);
+               min_val = regs[insn->src_reg].min_value;
+               max_val = regs[insn->src_reg].max_value;
+
+               /* If the source register is a random pointer then the
+                * min_value/max_value values represent the range of the known
+                * accesses into that value, not the actual min/max value of the
+                * register itself.  In this case we have to reset the reg range
+                * values so we know it is not safe to look at.
+                */
+               if (regs[insn->src_reg].type != CONST_IMM &&
+                   regs[insn->src_reg].type != UNKNOWN_VALUE) {
+                       min_val = BPF_REGISTER_MIN_RANGE;
+                       max_val = BPF_REGISTER_MAX_RANGE;
+               }
+       } else if (insn->imm < BPF_REGISTER_MAX_RANGE &&
+                  (s64)insn->imm > BPF_REGISTER_MIN_RANGE) {
+               min_val = max_val = insn->imm;
+               min_set = max_set = true;
+       }
+
+       /* We don't know anything about what was done to this register, mark it
+        * as unknown.
+        */
+       if (min_val == BPF_REGISTER_MIN_RANGE &&
+           max_val == BPF_REGISTER_MAX_RANGE) {
+               reset_reg_range_values(regs, insn->dst_reg);
+               return;
+       }
+
+       switch (opcode) {
+       case BPF_ADD:
+               dst_reg->min_value += min_val;
+               dst_reg->max_value += max_val;
+               break;
+       case BPF_SUB:
+               dst_reg->min_value -= min_val;
+               dst_reg->max_value -= max_val;
+               break;
+       case BPF_MUL:
+               dst_reg->min_value *= min_val;
+               dst_reg->max_value *= max_val;
+               break;
+       case BPF_AND:
+               /* & is special since it could end up with 0 bits set. */
+               dst_reg->min_value &= min_val;
+               dst_reg->max_value = max_val;
+               break;
+       case BPF_LSH:
+               /* Gotta have special overflow logic here, if we're shifting
+                * more than MAX_RANGE then just assume we have an invalid
+                * range.
+                */
+               if (min_val > ilog2(BPF_REGISTER_MAX_RANGE))
+                       dst_reg->min_value = BPF_REGISTER_MIN_RANGE;
+               else
+                       dst_reg->min_value <<= min_val;
+
+               if (max_val > ilog2(BPF_REGISTER_MAX_RANGE))
+                       dst_reg->max_value = BPF_REGISTER_MAX_RANGE;
+               else
+                       dst_reg->max_value <<= max_val;
+               break;
+       case BPF_RSH:
+               dst_reg->min_value >>= min_val;
+               dst_reg->max_value >>= max_val;
+               break;
+       case BPF_MOD:
+               /* % is special since it is an unsigned modulus, so the floor
+                * will always be 0.
+                */
+               dst_reg->min_value = 0;
+               dst_reg->max_value = max_val - 1;
+               break;
+       default:
+               reset_reg_range_values(regs, insn->dst_reg);
+               break;
+       }
+
+       check_reg_overflow(dst_reg);
+}
+
 /* check validity of 32-bit and 64-bit arithmetic operations */
 static int check_alu_op(struct bpf_verifier_env *env, struct bpf_insn *insn)
 {
@@ -1479,6 +1627,11 @@ static int check_alu_op(struct bpf_verifier_env *env, struct bpf_insn *insn)
                if (err)
                        return err;
 
+               /* we are setting our register to something new, we need to
+                * reset its range values.
+                */
+               reset_reg_range_values(regs, insn->dst_reg);
+
                if (BPF_SRC(insn->code) == BPF_X) {
                        if (BPF_CLASS(insn->code) == BPF_ALU64) {
                                /* case: R1 = R2
@@ -1500,6 +1653,8 @@ static int check_alu_op(struct bpf_verifier_env *env, struct bpf_insn *insn)
                         */
                        regs[insn->dst_reg].type = CONST_IMM;
                        regs[insn->dst_reg].imm = insn->imm;
+                       regs[insn->dst_reg].max_value = insn->imm;
+                       regs[insn->dst_reg].min_value = insn->imm;
                }
 
        } else if (opcode > BPF_END) {
@@ -1552,6 +1707,9 @@ static int check_alu_op(struct bpf_verifier_env *env, struct bpf_insn *insn)
 
                dst_reg = &regs[insn->dst_reg];
 
+               /* first we want to adjust our ranges. */
+               adjust_reg_min_max_vals(env, insn);
+
                /* pattern match 'bpf_add Rx, imm' instruction */
                if (opcode == BPF_ADD && BPF_CLASS(insn->code) == BPF_ALU64 &&
                    dst_reg->type == FRAME_PTR && BPF_SRC(insn->code) == BPF_K) {
@@ -1586,8 +1744,17 @@ static int check_alu_op(struct bpf_verifier_env *env, struct bpf_insn *insn)
                        return -EACCES;
                }
 
-               /* mark dest operand */
-               mark_reg_unknown_value(regs, insn->dst_reg);
+               /* If we did pointer math on a map value then just set it to our
+                * PTR_TO_MAP_VALUE_ADJ type so we can deal with any stores or
+                * loads to this register appropriately, otherwise just mark the
+                * register as unknown.
+                */
+               if (env->allow_ptr_leaks &&
+                   (dst_reg->type == PTR_TO_MAP_VALUE ||
+                    dst_reg->type == PTR_TO_MAP_VALUE_ADJ))
+                       dst_reg->type = PTR_TO_MAP_VALUE_ADJ;
+               else
+                       mark_reg_unknown_value(regs, insn->dst_reg);
        }
 
        return 0;
@@ -1642,6 +1809,104 @@ static void find_good_pkt_pointers(struct bpf_verifier_state *state,
        }
 }
 
+/* Adjusts the register min/max values in the case that the dst_reg is the
+ * variable register that we are working on, and src_reg is a constant or we're
+ * simply doing a BPF_K check.
+ */
+static void reg_set_min_max(struct bpf_reg_state *true_reg,
+                           struct bpf_reg_state *false_reg, u64 val,
+                           u8 opcode)
+{
+       switch (opcode) {
+       case BPF_JEQ:
+               /* If this is false then we know nothing Jon Snow, but if it is
+                * true then we know for sure.
+                */
+               true_reg->max_value = true_reg->min_value = val;
+               break;
+       case BPF_JNE:
+               /* If this is true we know nothing Jon Snow, but if it is false
+                * we know the value for sure;
+                */
+               false_reg->max_value = false_reg->min_value = val;
+               break;
+       case BPF_JGT:
+               /* Unsigned comparison, the minimum value is 0. */
+               false_reg->min_value = 0;
+       case BPF_JSGT:
+               /* If this is false then we know the maximum val is val,
+                * otherwise we know the min val is val+1.
+                */
+               false_reg->max_value = val;
+               true_reg->min_value = val + 1;
+               break;
+       case BPF_JGE:
+               /* Unsigned comparison, the minimum value is 0. */
+               false_reg->min_value = 0;
+       case BPF_JSGE:
+               /* If this is false then we know the maximum value is val - 1,
+                * otherwise we know the mimimum value is val.
+                */
+               false_reg->max_value = val - 1;
+               true_reg->min_value = val;
+               break;
+       default:
+               break;
+       }
+
+       check_reg_overflow(false_reg);
+       check_reg_overflow(true_reg);
+}
+
+/* Same as above, but for the case that dst_reg is a CONST_IMM reg and src_reg
+ * is the variable reg.
+ */
+static void reg_set_min_max_inv(struct bpf_reg_state *true_reg,
+                               struct bpf_reg_state *false_reg, u64 val,
+                               u8 opcode)
+{
+       switch (opcode) {
+       case BPF_JEQ:
+               /* If this is false then we know nothing Jon Snow, but if it is
+                * true then we know for sure.
+                */
+               true_reg->max_value = true_reg->min_value = val;
+               break;
+       case BPF_JNE:
+               /* If this is true we know nothing Jon Snow, but if it is false
+                * we know the value for sure;
+                */
+               false_reg->max_value = false_reg->min_value = val;
+               break;
+       case BPF_JGT:
+               /* Unsigned comparison, the minimum value is 0. */
+               true_reg->min_value = 0;
+       case BPF_JSGT:
+               /*
+                * If this is false, then the val is <= the register, if it is
+                * true the register <= to the val.
+                */
+               false_reg->min_value = val;
+               true_reg->max_value = val - 1;
+               break;
+       case BPF_JGE:
+               /* Unsigned comparison, the minimum value is 0. */
+               true_reg->min_value = 0;
+       case BPF_JSGE:
+               /* If this is false then constant < register, if it is true then
+                * the register < constant.
+                */
+               false_reg->min_value = val + 1;
+               true_reg->max_value = val;
+               break;
+       default:
+               break;
+       }
+
+       check_reg_overflow(false_reg);
+       check_reg_overflow(true_reg);
+}
+
 static int check_cond_jmp_op(struct bpf_verifier_env *env,
                             struct bpf_insn *insn, int *insn_idx)
 {
@@ -1708,6 +1973,23 @@ static int check_cond_jmp_op(struct bpf_verifier_env *env,
        if (!other_branch)
                return -EFAULT;
 
+       /* detect if we are comparing against a constant value so we can adjust
+        * our min/max values for our dst register.
+        */
+       if (BPF_SRC(insn->code) == BPF_X) {
+               if (regs[insn->src_reg].type == CONST_IMM)
+                       reg_set_min_max(&other_branch->regs[insn->dst_reg],
+                                       dst_reg, regs[insn->src_reg].imm,
+                                       opcode);
+               else if (dst_reg->type == CONST_IMM)
+                       reg_set_min_max_inv(&other_branch->regs[insn->src_reg],
+                                           &regs[insn->src_reg], dst_reg->imm,
+                                           opcode);
+       } else {
+               reg_set_min_max(&other_branch->regs[insn->dst_reg],
+                                       dst_reg, insn->imm, opcode);
+       }
+
        /* detect if R == 0 where R is returned from bpf_map_lookup_elem() */
        if (BPF_SRC(insn->code) == BPF_K &&
            insn->imm == 0 && (opcode == BPF_JEQ || opcode == BPF_JNE) &&
@@ -2144,7 +2426,8 @@ static bool compare_ptrs_to_packet(struct bpf_reg_state *old,
  * whereas register type in current state is meaningful, it means that
  * the current state will reach 'bpf_exit' instruction safely
  */
-static bool states_equal(struct bpf_verifier_state *old,
+static bool states_equal(struct bpf_verifier_env *env,
+                        struct bpf_verifier_state *old,
                         struct bpf_verifier_state *cur)
 {
        struct bpf_reg_state *rold, *rcur;
@@ -2157,6 +2440,13 @@ static bool states_equal(struct bpf_verifier_state *old,
                if (memcmp(rold, rcur, sizeof(*rold)) == 0)
                        continue;
 
+               /* If the ranges were not the same, but everything else was and
+                * we didn't do a variable access into a map then we are a-ok.
+                */
+               if (!env->varlen_map_value_access &&
+                   rold->type == rcur->type && rold->imm == rcur->imm)
+                       continue;
+
                if (rold->type == NOT_INIT ||
                    (rold->type == UNKNOWN_VALUE && rcur->type != NOT_INIT))
                        continue;
@@ -2213,7 +2503,7 @@ static int is_state_visited(struct bpf_verifier_env *env, int insn_idx)
                return 0;
 
        while (sl != STATE_LIST_MARK) {
-               if (states_equal(&sl->state, &env->cur_state))
+               if (states_equal(env, &sl->state, &env->cur_state))
                        /* reached equivalent register/stack state,
                         * prune the search
                         */
@@ -2259,6 +2549,7 @@ static int do_check(struct bpf_verifier_env *env)
 
        init_reg_state(regs);
        insn_idx = 0;
+       env->varlen_map_value_access = false;
        for (;;) {
                struct bpf_insn *insn;
                u8 class;
@@ -2339,6 +2630,7 @@ static int do_check(struct bpf_verifier_env *env)
                        if (err)
                                return err;
 
+                       reset_reg_range_values(regs, insn->dst_reg);
                        if (BPF_SIZE(insn->code) != BPF_W &&
                            BPF_SIZE(insn->code) != BPF_DW) {
                                insn_idx++;
@@ -2509,6 +2801,7 @@ process_bpf_exit:
                                verbose("invalid BPF_LD mode\n");
                                return -EINVAL;
                        }
+                       reset_reg_range_values(regs, insn->dst_reg);
                } else {
                        verbose("unknown insn class %d\n", class);
                        return -EINVAL;
index 364582b..ac6edb6 100644 (file)
@@ -85,6 +85,14 @@ extern char bpf_log_buf[LOG_BUF_SIZE];
                .off   = 0,                                     \
                .imm   = IMM })
 
+#define BPF_MOV32_IMM(DST, IMM)                                        \
+       ((struct bpf_insn) {                                    \
+               .code  = BPF_ALU | BPF_MOV | BPF_K,             \
+               .dst_reg = DST,                                 \
+               .src_reg = 0,                                   \
+               .off   = 0,                                     \
+               .imm   = IMM })
+
 /* BPF_LD_IMM64 macro encodes single 'load 64-bit immediate' insn */
 #define BPF_LD_IMM64(DST, IMM)                                 \
        BPF_LD_IMM64_RAW(DST, 0, IMM)
index ac590d4..369ffaa 100644 (file)
@@ -29,6 +29,7 @@ struct bpf_test {
        struct bpf_insn insns[MAX_INSNS];
        int fixup[MAX_FIXUPS];
        int prog_array_fixup[MAX_FIXUPS];
+       int test_val_map_fixup[MAX_FIXUPS];
        const char *errstr;
        const char *errstr_unpriv;
        enum {
@@ -39,6 +40,19 @@ struct bpf_test {
        enum bpf_prog_type prog_type;
 };
 
+/* Note we want this to be 64 bit aligned so that the end of our array is
+ * actually the end of the structure.
+ */
+#define MAX_ENTRIES 11
+struct test_val {
+       unsigned index;
+       int foo[MAX_ENTRIES];
+};
+
+struct other_val {
+       unsigned int action[32];
+};
+
 static struct bpf_test tests[] = {
        {
                "add+sub+mul",
@@ -2163,6 +2177,212 @@ static struct bpf_test tests[] = {
                .errstr = "invalid access to packet",
                .prog_type = BPF_PROG_TYPE_SCHED_CLS,
        },
+       {
+               "valid map access into an array with a constant",
+               .insns = {
+                       BPF_ST_MEM(BPF_DW, BPF_REG_10, -8, 0),
+                       BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
+                       BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -8),
+                       BPF_LD_MAP_FD(BPF_REG_1, 0),
+                       BPF_RAW_INSN(BPF_JMP | BPF_CALL, 0, 0, 0, BPF_FUNC_map_lookup_elem),
+                       BPF_JMP_IMM(BPF_JEQ, BPF_REG_0, 0, 1),
+                       BPF_ST_MEM(BPF_DW, BPF_REG_0, 0, offsetof(struct test_val, foo)),
+                       BPF_EXIT_INSN(),
+               },
+               .test_val_map_fixup = {3},
+               .errstr_unpriv = "R0 leaks addr",
+               .result_unpriv = REJECT,
+               .result = ACCEPT,
+       },
+       {
+               "valid map access into an array with a register",
+               .insns = {
+                       BPF_ST_MEM(BPF_DW, BPF_REG_10, -8, 0),
+                       BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
+                       BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -8),
+                       BPF_LD_MAP_FD(BPF_REG_1, 0),
+                       BPF_RAW_INSN(BPF_JMP | BPF_CALL, 0, 0, 0, BPF_FUNC_map_lookup_elem),
+                       BPF_JMP_IMM(BPF_JEQ, BPF_REG_0, 0, 4),
+                       BPF_MOV64_IMM(BPF_REG_1, 4),
+                       BPF_ALU64_IMM(BPF_LSH, BPF_REG_1, 2),
+                       BPF_ALU64_REG(BPF_ADD, BPF_REG_0, BPF_REG_1),
+                       BPF_ST_MEM(BPF_DW, BPF_REG_0, 0, offsetof(struct test_val, foo)),
+                       BPF_EXIT_INSN(),
+               },
+               .test_val_map_fixup = {3},
+               .errstr_unpriv = "R0 leaks addr",
+               .result_unpriv = REJECT,
+               .result = ACCEPT,
+       },
+       {
+               "valid map access into an array with a variable",
+               .insns = {
+                       BPF_ST_MEM(BPF_DW, BPF_REG_10, -8, 0),
+                       BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
+                       BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -8),
+                       BPF_LD_MAP_FD(BPF_REG_1, 0),
+                       BPF_RAW_INSN(BPF_JMP | BPF_CALL, 0, 0, 0, BPF_FUNC_map_lookup_elem),
+                       BPF_JMP_IMM(BPF_JEQ, BPF_REG_0, 0, 5),
+                       BPF_LDX_MEM(BPF_W, BPF_REG_1, BPF_REG_0, 0),
+                       BPF_JMP_IMM(BPF_JGE, BPF_REG_1, MAX_ENTRIES, 3),
+                       BPF_ALU64_IMM(BPF_LSH, BPF_REG_1, 2),
+                       BPF_ALU64_REG(BPF_ADD, BPF_REG_0, BPF_REG_1),
+                       BPF_ST_MEM(BPF_DW, BPF_REG_0, 0, offsetof(struct test_val, foo)),
+                       BPF_EXIT_INSN(),
+               },
+               .test_val_map_fixup = {3},
+               .errstr_unpriv = "R0 leaks addr",
+               .result_unpriv = REJECT,
+               .result = ACCEPT,
+       },
+       {
+               "valid map access into an array with a signed variable",
+               .insns = {
+                       BPF_ST_MEM(BPF_DW, BPF_REG_10, -8, 0),
+                       BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
+                       BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -8),
+                       BPF_LD_MAP_FD(BPF_REG_1, 0),
+                       BPF_RAW_INSN(BPF_JMP | BPF_CALL, 0, 0, 0, BPF_FUNC_map_lookup_elem),
+                       BPF_JMP_IMM(BPF_JEQ, BPF_REG_0, 0, 9),
+                       BPF_LDX_MEM(BPF_W, BPF_REG_1, BPF_REG_0, 0),
+                       BPF_JMP_IMM(BPF_JSGT, BPF_REG_1, 0xffffffff, 1),
+                       BPF_MOV32_IMM(BPF_REG_1, 0),
+                       BPF_MOV32_IMM(BPF_REG_2, MAX_ENTRIES),
+                       BPF_JMP_REG(BPF_JSGT, BPF_REG_2, BPF_REG_1, 1),
+                       BPF_MOV32_IMM(BPF_REG_1, 0),
+                       BPF_ALU32_IMM(BPF_LSH, BPF_REG_1, 2),
+                       BPF_ALU64_REG(BPF_ADD, BPF_REG_0, BPF_REG_1),
+                       BPF_ST_MEM(BPF_DW, BPF_REG_0, 0, offsetof(struct test_val, foo)),
+                       BPF_EXIT_INSN(),
+               },
+               .test_val_map_fixup = {3},
+               .errstr_unpriv = "R0 leaks addr",
+               .result_unpriv = REJECT,
+               .result = ACCEPT,
+       },
+       {
+               "invalid map access into an array with a constant",
+               .insns = {
+                       BPF_ST_MEM(BPF_DW, BPF_REG_10, -8, 0),
+                       BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
+                       BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -8),
+                       BPF_LD_MAP_FD(BPF_REG_1, 0),
+                       BPF_RAW_INSN(BPF_JMP | BPF_CALL, 0, 0, 0, BPF_FUNC_map_lookup_elem),
+                       BPF_JMP_IMM(BPF_JEQ, BPF_REG_0, 0, 1),
+                       BPF_ST_MEM(BPF_DW, BPF_REG_0, (MAX_ENTRIES + 1) << 2,
+                                  offsetof(struct test_val, foo)),
+                       BPF_EXIT_INSN(),
+               },
+               .test_val_map_fixup = {3},
+               .errstr = "invalid access to map value, value_size=48 off=48 size=8",
+               .result = REJECT,
+       },
+       {
+               "invalid map access into an array with a register",
+               .insns = {
+                       BPF_ST_MEM(BPF_DW, BPF_REG_10, -8, 0),
+                       BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
+                       BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -8),
+                       BPF_LD_MAP_FD(BPF_REG_1, 0),
+                       BPF_RAW_INSN(BPF_JMP | BPF_CALL, 0, 0, 0, BPF_FUNC_map_lookup_elem),
+                       BPF_JMP_IMM(BPF_JEQ, BPF_REG_0, 0, 4),
+                       BPF_MOV64_IMM(BPF_REG_1, MAX_ENTRIES + 1),
+                       BPF_ALU64_IMM(BPF_LSH, BPF_REG_1, 2),
+                       BPF_ALU64_REG(BPF_ADD, BPF_REG_0, BPF_REG_1),
+                       BPF_ST_MEM(BPF_DW, BPF_REG_0, 0, offsetof(struct test_val, foo)),
+                       BPF_EXIT_INSN(),
+               },
+               .test_val_map_fixup = {3},
+               .errstr = "R0 min value is outside of the array range",
+               .result = REJECT,
+       },
+       {
+               "invalid map access into an array with a variable",
+               .insns = {
+                       BPF_ST_MEM(BPF_DW, BPF_REG_10, -8, 0),
+                       BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
+                       BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -8),
+                       BPF_LD_MAP_FD(BPF_REG_1, 0),
+                       BPF_RAW_INSN(BPF_JMP | BPF_CALL, 0, 0, 0, BPF_FUNC_map_lookup_elem),
+                       BPF_JMP_IMM(BPF_JEQ, BPF_REG_0, 0, 4),
+                       BPF_LDX_MEM(BPF_W, BPF_REG_1, BPF_REG_0, 0),
+                       BPF_ALU64_IMM(BPF_LSH, BPF_REG_1, 2),
+                       BPF_ALU64_REG(BPF_ADD, BPF_REG_0, BPF_REG_1),
+                       BPF_ST_MEM(BPF_DW, BPF_REG_0, 0, offsetof(struct test_val, foo)),
+                       BPF_EXIT_INSN(),
+               },
+               .test_val_map_fixup = {3},
+               .errstr = "R0 min value is negative, either use unsigned index or do a if (index >=0) check.",
+               .result = REJECT,
+       },
+       {
+               "invalid map access into an array with no floor check",
+               .insns = {
+                       BPF_ST_MEM(BPF_DW, BPF_REG_10, -8, 0),
+                       BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
+                       BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -8),
+                       BPF_LD_MAP_FD(BPF_REG_1, 0),
+                       BPF_RAW_INSN(BPF_JMP | BPF_CALL, 0, 0, 0, BPF_FUNC_map_lookup_elem),
+                       BPF_JMP_IMM(BPF_JEQ, BPF_REG_0, 0, 7),
+                       BPF_LDX_MEM(BPF_W, BPF_REG_1, BPF_REG_0, 0),
+                       BPF_MOV32_IMM(BPF_REG_2, MAX_ENTRIES),
+                       BPF_JMP_REG(BPF_JSGT, BPF_REG_2, BPF_REG_1, 1),
+                       BPF_MOV32_IMM(BPF_REG_1, 0),
+                       BPF_ALU32_IMM(BPF_LSH, BPF_REG_1, 2),
+                       BPF_ALU64_REG(BPF_ADD, BPF_REG_0, BPF_REG_1),
+                       BPF_ST_MEM(BPF_DW, BPF_REG_0, 0, offsetof(struct test_val, foo)),
+                       BPF_EXIT_INSN(),
+               },
+               .test_val_map_fixup = {3},
+               .errstr = "R0 min value is negative, either use unsigned index or do a if (index >=0) check.",
+               .result = REJECT,
+       },
+       {
+               "invalid map access into an array with a invalid max check",
+               .insns = {
+                       BPF_ST_MEM(BPF_DW, BPF_REG_10, -8, 0),
+                       BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
+                       BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -8),
+                       BPF_LD_MAP_FD(BPF_REG_1, 0),
+                       BPF_RAW_INSN(BPF_JMP | BPF_CALL, 0, 0, 0, BPF_FUNC_map_lookup_elem),
+                       BPF_JMP_IMM(BPF_JEQ, BPF_REG_0, 0, 7),
+                       BPF_LDX_MEM(BPF_W, BPF_REG_1, BPF_REG_0, 0),
+                       BPF_MOV32_IMM(BPF_REG_2, MAX_ENTRIES + 1),
+                       BPF_JMP_REG(BPF_JGT, BPF_REG_2, BPF_REG_1, 1),
+                       BPF_MOV32_IMM(BPF_REG_1, 0),
+                       BPF_ALU32_IMM(BPF_LSH, BPF_REG_1, 2),
+                       BPF_ALU64_REG(BPF_ADD, BPF_REG_0, BPF_REG_1),
+                       BPF_ST_MEM(BPF_DW, BPF_REG_0, 0, offsetof(struct test_val, foo)),
+                       BPF_EXIT_INSN(),
+               },
+               .test_val_map_fixup = {3},
+               .errstr = "invalid access to map value, value_size=48 off=44 size=8",
+               .result = REJECT,
+       },
+       {
+               "invalid map access into an array with a invalid max check",
+               .insns = {
+                       BPF_ST_MEM(BPF_DW, BPF_REG_10, -8, 0),
+                       BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
+                       BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -8),
+                       BPF_LD_MAP_FD(BPF_REG_1, 0),
+                       BPF_RAW_INSN(BPF_JMP | BPF_CALL, 0, 0, 0, BPF_FUNC_map_lookup_elem),
+                       BPF_JMP_IMM(BPF_JEQ, BPF_REG_0, 0, 10),
+                       BPF_MOV64_REG(BPF_REG_8, BPF_REG_0),
+                       BPF_ST_MEM(BPF_DW, BPF_REG_10, -8, 0),
+                       BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
+                       BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -8),
+                       BPF_LD_MAP_FD(BPF_REG_1, 0),
+                       BPF_RAW_INSN(BPF_JMP | BPF_CALL, 0, 0, 0, BPF_FUNC_map_lookup_elem),
+                       BPF_JMP_IMM(BPF_JEQ, BPF_REG_0, 0, 2),
+                       BPF_ALU64_REG(BPF_ADD, BPF_REG_0, BPF_REG_8),
+                       BPF_LDX_MEM(BPF_W, BPF_REG_0, BPF_REG_0, offsetof(struct test_val, foo)),
+                       BPF_EXIT_INSN(),
+               },
+               .test_val_map_fixup = {3, 11},
+               .errstr = "R0 min value is negative, either use unsigned index or do a if (index >=0) check.",
+               .result = REJECT,
+       },
 };
 
 static int probe_filter_length(struct bpf_insn *fp)
@@ -2176,12 +2396,12 @@ static int probe_filter_length(struct bpf_insn *fp)
        return len + 1;
 }
 
-static int create_map(void)
+static int create_map(size_t val_size, int num)
 {
        int map_fd;
 
        map_fd = bpf_create_map(BPF_MAP_TYPE_HASH,
-                               sizeof(long long), sizeof(long long), 1024, 0);
+                               sizeof(long long), val_size, num, 0);
        if (map_fd < 0)
                printf("failed to create map '%s'\n", strerror(errno));
 
@@ -2211,12 +2431,13 @@ static int test(void)
                int prog_len = probe_filter_length(prog);
                int *fixup = tests[i].fixup;
                int *prog_array_fixup = tests[i].prog_array_fixup;
+               int *test_val_map_fixup = tests[i].test_val_map_fixup;
                int expected_result;
                const char *expected_errstr;
-               int map_fd = -1, prog_array_fd = -1;
+               int map_fd = -1, prog_array_fd = -1, test_val_map_fd = -1;
 
                if (*fixup) {
-                       map_fd = create_map();
+                       map_fd = create_map(sizeof(long long), 1024);
 
                        do {
                                prog[*fixup].imm = map_fd;
@@ -2231,6 +2452,18 @@ static int test(void)
                                prog_array_fixup++;
                        } while (*prog_array_fixup);
                }
+               if (*test_val_map_fixup) {
+                       /* Unprivileged can't create a hash map.*/
+                       if (unpriv)
+                               continue;
+                       test_val_map_fd = create_map(sizeof(struct test_val),
+                                                    256);
+                       do {
+                               prog[*test_val_map_fixup].imm = test_val_map_fd;
+                               test_val_map_fixup++;
+                       } while (*test_val_map_fixup);
+               }
+
                printf("#%d %s ", i, tests[i].descr);
 
                prog_fd = bpf_prog_load(prog_type ?: BPF_PROG_TYPE_SOCKET_FILTER,
@@ -2277,6 +2510,8 @@ fail:
                        close(map_fd);
                if (prog_array_fd >= 0)
                        close(prog_array_fd);
+               if (test_val_map_fd >= 0)
+                       close(test_val_map_fd);
                close(prog_fd);
 
        }