Merge branch 'bpf-rewrite-value-tracking-in-verifier'

Edward Cree says:

====================
bpf: rewrite value tracking in verifier

This series simplifies alignment tracking, generalises bounds tracking
and fixes some bounds-tracking bugs in the BPF verifier.  Pointer
arithmetic on packet pointers, stack pointers, map value pointers and
context pointers has been unified, and bounds on these pointers are
only checked when the pointer is dereferenced.

Operations on pointers which destroy all relation to the original
pointer (such as multiplies and shifts) are disallowed if
!env->allow_ptr_leaks, otherwise they convert the pointer to an
unknown scalar and feed it to the normal scalar arithmetic handling.

Pointer types have been unified with the corresponding
adjusted-pointer types where those existed
(e.g. PTR_TO_MAP_VALUE[_ADJ] or FRAME_PTR vs PTR_TO_STACK); similarly,
CONST_IMM and UNKNOWN_VALUE have been unified into SCALAR_VALUE.

Pointer types (except CONST_PTR_TO_MAP, PTR_TO_MAP_VALUE_OR_NULL and
PTR_TO_PACKET_END, which do not allow arithmetic) have a 'fixed
offset' and a 'variable offset'; the former is used when e.g. adding
an immediate or a known-constant register, as long as it does not
overflow.  Otherwise the latter is used, and any operation creating a
new variable offset creates a new 'id' (and, for PTR_TO_PACKET, clears
the 'range').  SCALAR_VALUEs use the 'variable offset' fields to track
the range of possible values; the 'fixed offset' should never be set
on a scalar.
====================

Acked-by: Daniel Borkmann <[email protected]>
Signed-off-by: David S. Miller <[email protected]>

Author: davem330

Documentation/networking/filter.txt

@@ -793,7 +793,7 @@ Some core changes of the new internal format:
     bpf_exit
 
   After the call the registers R1-R5 contain junk values and cannot be read.
-  In the future an eBPF verifier can be used to validate internal BPF programs.
+  An in-kernel eBPF verifier is used to validate internal BPF programs.
 
 Also in the new design, eBPF is limited to 4096 insns, which means that any
 program will terminate quickly and will only call a fixed number of kernel
@@ -1017,7 +1017,7 @@ At the start of the program the register R1 contains a pointer to context
 and has type PTR_TO_CTX.
 If verifier sees an insn that does R2=R1, then R2 has now type
 PTR_TO_CTX as well and can be used on the right hand side of expression.
-If R1=PTR_TO_CTX and insn is R2=R1+R1, then R2=UNKNOWN_VALUE,
+If R1=PTR_TO_CTX and insn is R2=R1+R1, then R2=SCALAR_VALUE,
 since addition of two valid pointers makes invalid pointer.
 (In 'secure' mode verifier will reject any type of pointer arithmetic to make
 sure that kernel addresses don't leak to unprivileged users)
@@ -1039,7 +1039,7 @@ is a correct program. If there was R1 instead of R6, it would have
 been rejected.
 
 load/store instructions are allowed only with registers of valid types, which
-are PTR_TO_CTX, PTR_TO_MAP, FRAME_PTR. They are bounds and alignment checked.
+are PTR_TO_CTX, PTR_TO_MAP, PTR_TO_STACK. They are bounds and alignment checked.
 For example:
  bpf_mov R1 = 1
  bpf_mov R2 = 2
@@ -1058,7 +1058,7 @@ intends to load a word from address R6 + 8 and store it into R0
 If R6=PTR_TO_CTX, via is_valid_access() callback the verifier will know
 that offset 8 of size 4 bytes can be accessed for reading, otherwise
 the verifier will reject the program.
-If R6=FRAME_PTR, then access should be aligned and be within
+If R6=PTR_TO_STACK, then access should be aligned and be within
 stack bounds, which are [-MAX_BPF_STACK, 0). In this example offset is 8,
 so it will fail verification, since it's out of bounds.
 
@@ -1069,7 +1069,7 @@ For example:
   bpf_ld R0 = *(u32 *)(R10 - 4)
   bpf_exit
 is invalid program.
-Though R10 is correct read-only register and has type FRAME_PTR
+Though R10 is correct read-only register and has type PTR_TO_STACK
 and R10 - 4 is within stack bounds, there were no stores into that location.
 
 Pointer register spill/fill is tracked as well, since four (R6-R9)
@@ -1094,6 +1094,71 @@ all use cases.
 
 See details of eBPF verifier in kernel/bpf/verifier.c
 
+Register value tracking
+-----------------------
+In order to determine the safety of an eBPF program, the verifier must track
+the range of possible values in each register and also in each stack slot.
+This is done with 'struct bpf_reg_state', defined in include/linux/
+bpf_verifier.h, which unifies tracking of scalar and pointer values.  Each
+register state has a type, which is either NOT_INIT (the register has not been
+written to), SCALAR_VALUE (some value which is not usable as a pointer), or a
+pointer type.  The types of pointers describe their base, as follows:
+    PTR_TO_CTX          Pointer to bpf_context.
+    CONST_PTR_TO_MAP    Pointer to struct bpf_map.  "Const" because arithmetic
+                        on these pointers is forbidden.
+    PTR_TO_MAP_VALUE    Pointer to the value stored in a map element.
+    PTR_TO_MAP_VALUE_OR_NULL
+                        Either a pointer to a map value, or NULL; map accesses
+                        (see section 'eBPF maps', below) return this type,
+                        which becomes a PTR_TO_MAP_VALUE when checked != NULL.
+                        Arithmetic on these pointers is forbidden.
+    PTR_TO_STACK        Frame pointer.
+    PTR_TO_PACKET       skb->data.
+    PTR_TO_PACKET_END   skb->data + headlen; arithmetic forbidden.
+However, a pointer may be offset from this base (as a result of pointer
+arithmetic), and this is tracked in two parts: the 'fixed offset' and 'variable
+offset'.  The former is used when an exactly-known value (e.g. an immediate
+operand) is added to a pointer, while the latter is used for values which are
+not exactly known.  The variable offset is also used in SCALAR_VALUEs, to track
+the range of possible values in the register.
+The verifier's knowledge about the variable offset consists of:
+* minimum and maximum values as unsigned
+* minimum and maximum values as signed
+* knowledge of the values of individual bits, in the form of a 'tnum': a u64
+'mask' and a u64 'value'.  1s in the mask represent bits whose value is unknown;
+1s in the value represent bits known to be 1.  Bits known to be 0 have 0 in both
+mask and value; no bit should ever be 1 in both.  For example, if a byte is read
+into a register from memory, the register's top 56 bits are known zero, while
+the low 8 are unknown - which is represented as the tnum (0x0; 0xff).  If we
+then OR this with 0x40, we get (0x40; 0xcf), then if we add 1 we get (0x0;
+0x1ff), because of potential carries.
+Besides arithmetic, the register state can also be updated by conditional
+branches.  For instance, if a SCALAR_VALUE is compared > 8, in the 'true' branch
+it will have a umin_value (unsigned minimum value) of 9, whereas in the 'false'
+branch it will have a umax_value of 8.  A signed compare (with BPF_JSGT or
+BPF_JSGE) would instead update the signed minimum/maximum values.  Information
+from the signed and unsigned bounds can be combined; for instance if a value is
+first tested < 8 and then tested s> 4, the verifier will conclude that the value
+is also > 4 and s< 8, since the bounds prevent crossing the sign boundary.
+PTR_TO_PACKETs with a variable offset part have an 'id', which is common to all
+pointers sharing that same variable offset.  This is important for packet range
+checks: after adding some variable to a packet pointer, if you then copy it to
+another register and (say) add a constant 4, both registers will share the same
+'id' but one will have a fixed offset of +4.  Then if it is bounds-checked and
+found to be less than a PTR_TO_PACKET_END, the other register is now known to
+have a safe range of at least 4 bytes.  See 'Direct packet access', below, for
+more on PTR_TO_PACKET ranges.
+The 'id' field is also used on PTR_TO_MAP_VALUE_OR_NULL, common to all copies of
+the pointer returned from a map lookup.  This means that when one copy is
+checked and found to be non-NULL, all copies can become PTR_TO_MAP_VALUEs.
+As well as range-checking, the tracked information is also used for enforcing
+alignment of pointer accesses.  For instance, on most systems the packet pointer
+is 2 bytes after a 4-byte alignment.  If a program adds 14 bytes to that to jump
+over the Ethernet header, then reads IHL and addes (IHL * 4), the resulting
+pointer will have a variable offset known to be 4n+2 for some n, so adding the 2
+bytes (NET_IP_ALIGN) gives a 4-byte alignment and so word-sized accesses through
+that pointer are safe.
+
 Direct packet access
 --------------------
 In cls_bpf and act_bpf programs the verifier allows direct access to the packet
@@ -1121,7 +1186,7 @@ it now points to 'skb->data + 14' and accessible range is [R5, R5 + 14 - 14)
 which is zero bytes.
 
 More complex packet access may look like:
- R0=imm1 R1=ctx R3=pkt(id=0,off=0,r=14) R4=pkt_end R5=pkt(id=0,off=14,r=14) R10=fp
+ R0=inv1 R1=ctx R3=pkt(id=0,off=0,r=14) R4=pkt_end R5=pkt(id=0,off=14,r=14) R10=fp
  6:  r0 = *(u8 *)(r3 +7) /* load 7th byte from the packet */
  7:  r4 = *(u8 *)(r3 +12)
  8:  r4 *= 14
@@ -1135,26 +1200,31 @@ More complex packet access may look like:
 16:  r2 += 8
 17:  r1 = *(u32 *)(r1 +80) /* load skb->data_end */
 18:  if r2 > r1 goto pc+2
- R0=inv56 R1=pkt_end R2=pkt(id=2,off=8,r=8) R3=pkt(id=2,off=0,r=8) R4=inv52 R5=pkt(id=0,off=14,r=14) R10=fp
+ R0=inv(id=0,umax_value=255,var_off=(0x0; 0xff)) R1=pkt_end R2=pkt(id=2,off=8,r=8) R3=pkt(id=2,off=0,r=8) R4=inv(id=0,umax_value=3570,var_off=(0x0; 0xfffe)) R5=pkt(id=0,off=14,r=14) R10=fp
 19:  r1 = *(u8 *)(r3 +4)
 The state of the register R3 is R3=pkt(id=2,off=0,r=8)
 id=2 means that two 'r3 += rX' instructions were seen, so r3 points to some
 offset within a packet and since the program author did
 'if (r3 + 8 > r1) goto err' at insn #18, the safe range is [R3, R3 + 8).
-The verifier only allows 'add' operation on packet registers. Any other
-operation will set the register state to 'unknown_value' and it won't be
+The verifier only allows 'add'/'sub' operations on packet registers. Any other
+operation will set the register state to 'SCALAR_VALUE' and it won't be
 available for direct packet access.
 Operation 'r3 += rX' may overflow and become less than original skb->data,
-therefore the verifier has to prevent that. So it tracks the number of
-upper zero bits in all 'uknown_value' registers, so when it sees
-'r3 += rX' instruction and rX is more than 16-bit value, it will error as:
-"cannot add integer value with N upper zero bits to ptr_to_packet"
+therefore the verifier has to prevent that.  So when it sees 'r3 += rX'
+instruction and rX is more than 16-bit value, any subsequent bounds-check of r3
+against skb->data_end will not give us 'range' information, so attempts to read
+through the pointer will give "invalid access to packet" error.
 Ex. after insn 'r4 = *(u8 *)(r3 +12)' (insn #7 above) the state of r4 is
-R4=inv56 which means that upper 56 bits on the register are guaranteed
-to be zero. After insn 'r4 *= 14' the state becomes R4=inv52, since
-multiplying 8-bit value by constant 14 will keep upper 52 bits as zero.
-Similarly 'r2 >>= 48' will make R2=inv48, since the shift is not sign
-extending. This logic is implemented in evaluate_reg_alu() function.
+R4=inv(id=0,umax_value=255,var_off=(0x0; 0xff)) which means that upper 56 bits
+of the register are guaranteed to be zero, and nothing is known about the lower
+8 bits. After insn 'r4 *= 14' the state becomes
+R4=inv(id=0,umax_value=3570,var_off=(0x0; 0xfffe)), since multiplying an 8-bit
+value by constant 14 will keep upper 52 bits as zero, also the least significant
+bit will be zero as 14 is even.  Similarly 'r2 >>= 48' will make
+R2=inv(id=0,umax_value=65535,var_off=(0x0; 0xffff)), since the shift is not sign
+extending.  This logic is implemented in adjust_reg_min_max_vals() function,
+which calls adjust_ptr_min_max_vals() for adding pointer to scalar (or vice
+versa) and adjust_scalar_min_max_vals() for operations on two scalars.
 
 The end result is that bpf program author can access packet directly
 using normal C code as:
@@ -1214,6 +1284,22 @@ The map is defined by:
   . key size in bytes
   . value size in bytes
 
+Pruning
+-------
+The verifier does not actually walk all possible paths through the program.  For
+each new branch to analyse, the verifier looks at all the states it's previously
+been in when at this instruction.  If any of them contain the current state as a
+subset, the branch is 'pruned' - that is, the fact that the previous state was
+accepted implies the current state would be as well.  For instance, if in the
+previous state, r1 held a packet-pointer, and in the current state, r1 holds a
+packet-pointer with a range as long or longer and at least as strict an
+alignment, then r1 is safe.  Similarly, if r2 was NOT_INIT before then it can't
+have been used by any path from that point, so any value in r2 (including
+another NOT_INIT) is safe.  The implementation is in the function regsafe().
+Pruning considers not only the registers but also the stack (and any spilled
+registers it may hold).  They must all be safe for the branch to be pruned.
+This is implemented in states_equal().
+
 Understanding eBPF verifier messages
 ------------------------------------
 

drivers/net/ethernet/netronome/nfp/bpf/verifier.c

@@ -79,28 +79,32 @@ nfp_bpf_check_exit(struct nfp_prog *nfp_prog,
 		   const struct bpf_verifier_env *env)
 {
 	const struct bpf_reg_state *reg0 = &env->cur_state.regs[0];
+	u64 imm;
 
 	if (nfp_prog->act == NN_ACT_XDP)
 		return 0;
 
-	if (reg0->type != CONST_IMM) {
-		pr_info("unsupported exit state: %d, imm: %llx\n",
-			reg0->type, reg0->imm);
+	if (!(reg0->type == SCALAR_VALUE && tnum_is_const(reg0->var_off))) {
+		char tn_buf[48];
+
+		tnum_strn(tn_buf, sizeof(tn_buf), reg0->var_off);
+		pr_info("unsupported exit state: %d, var_off: %s\n",
+			reg0->type, tn_buf);
 		return -EINVAL;
 	}
 
-	if (nfp_prog->act != NN_ACT_DIRECT &&
-	    reg0->imm != 0 && (reg0->imm & ~0U) != ~0U) {
+	imm = reg0->var_off.value;
+	if (nfp_prog->act != NN_ACT_DIRECT && imm != 0 && (imm & ~0U) != ~0U) {
 		pr_info("unsupported exit state: %d, imm: %llx\n",
-			reg0->type, reg0->imm);
+			reg0->type, imm);
 		return -EINVAL;
 	}
 
-	if (nfp_prog->act == NN_ACT_DIRECT && reg0->imm <= TC_ACT_REDIRECT &&
-	    reg0->imm != TC_ACT_SHOT && reg0->imm != TC_ACT_STOLEN &&
-	    reg0->imm != TC_ACT_QUEUED) {
+	if (nfp_prog->act == NN_ACT_DIRECT && imm <= TC_ACT_REDIRECT &&
+	    imm != TC_ACT_SHOT && imm != TC_ACT_STOLEN &&
+	    imm != TC_ACT_QUEUED) {
 		pr_info("unsupported exit state: %d, imm: %llx\n",
-			reg0->type, reg0->imm);
+			reg0->type, imm);
 		return -EINVAL;
 	}
 

include/linux/bpf.h

@@ -117,35 +117,25 @@ enum bpf_access_type {
 };
 
 /* types of values stored in eBPF registers */
+/* Pointer types represent:
+ * pointer
+ * pointer + imm
+ * pointer + (u16) var
+ * pointer + (u16) var + imm
+ * if (range > 0) then [ptr, ptr + range - off) is safe to access
+ * if (id > 0) means that some 'var' was added
+ * if (off > 0) means that 'imm' was added
+ */
 enum bpf_reg_type {
 	NOT_INIT = 0,		 /* nothing was written into register */
-	UNKNOWN_VALUE,		 /* reg doesn't contain a valid pointer */
+	SCALAR_VALUE,		 /* reg doesn't contain a valid pointer */
 	PTR_TO_CTX,		 /* reg points to bpf_context */
 	CONST_PTR_TO_MAP,	 /* reg points to struct bpf_map */
 	PTR_TO_MAP_VALUE,	 /* reg points to map element value */
 	PTR_TO_MAP_VALUE_OR_NULL,/* points to map elem value or NULL */
-	FRAME_PTR,		 /* reg == frame_pointer */
-	PTR_TO_STACK,		 /* reg == frame_pointer + imm */
-	CONST_IMM,		 /* constant integer value */
-
-	/* PTR_TO_PACKET represents:
-	 * skb->data
-	 * skb->data + imm
-	 * skb->data + (u16) var
-	 * skb->data + (u16) var + imm
-	 * if (range > 0) then [ptr, ptr + range - off) is safe to access
-	 * if (id > 0) means that some 'var' was added
-	 * if (off > 0) menas that 'imm' was added
-	 */
-	PTR_TO_PACKET,
+	PTR_TO_STACK,		 /* reg == frame_pointer + offset */
+	PTR_TO_PACKET,		 /* reg points to skb->data */
 	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;

include/linux/bpf_verifier.h

@@ -9,41 +9,54 @@
 
 #include <linux/bpf.h> /* for enum bpf_reg_type */
 #include <linux/filter.h> /* for MAX_BPF_STACK */
+#include <linux/tnum.h>
 
- /* 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 -1
+/* Maximum variable offset umax_value permitted when resolving memory accesses.
+ * In practice this is far bigger than any realistic pointer offset; this limit
+ * ensures that umax_value + (int)off + (int)size cannot overflow a u64.
+ */
+#define BPF_MAX_VAR_OFF	(1ULL << 31)
+/* Maximum variable size permitted for ARG_CONST_SIZE[_OR_ZERO].  This ensures
+ * that converting umax_value to int cannot overflow.
+ */
+#define BPF_MAX_VAR_SIZ	INT_MAX
 
 struct bpf_reg_state {
 	enum bpf_reg_type type;
 	union {
-		/* valid when type == CONST_IMM | PTR_TO_STACK | UNKNOWN_VALUE */
-		s64 imm;
-
-		/* valid when type == PTR_TO_PACKET* */
-		struct {
-			u16 off;
-			u16 range;
-		};
+		/* valid when type == PTR_TO_PACKET */
+		u16 range;
 
 		/* valid when type == CONST_PTR_TO_MAP | PTR_TO_MAP_VALUE |
 		 *   PTR_TO_MAP_VALUE_OR_NULL
 		 */
 		struct bpf_map *map_ptr;
 	};
+	/* Fixed part of pointer offset, pointer types only */
+	s32 off;
+	/* For PTR_TO_PACKET, used to find other pointers with the same variable
+	 * offset, so they can share range knowledge.
+	 * For PTR_TO_MAP_VALUE_OR_NULL this is used to share which map value we
+	 * came from, when one is tested for != NULL.
+	 */
 	u32 id;
+	/* These five fields must be last.  See states_equal() */
+	/* For scalar types (SCALAR_VALUE), this represents our knowledge of
+	 * the actual value.
+	 * For pointer types, this represents the variable part of the offset
+	 * from the pointed-to object, and is shared with all bpf_reg_states
+	 * with the same id as us.
+	 */
+	struct tnum var_off;
 	/* Used to determine if any memory access using this register will
-	 * result in a bad access. These two fields must be last.
-	 * See states_equal()
+	 * result in a bad access.
+	 * These refer to the same value as var_off, not necessarily the actual
+	 * contents of the register.
 	 */
-	s64 min_value;
-	u64 max_value;
-	u32 min_align;
-	u32 aux_off;
-	u32 aux_off_align;
-	bool value_from_signed;
+	s64 smin_value; /* minimum possible (s64)value */
+	s64 smax_value; /* maximum possible (s64)value */
+	u64 umin_value; /* minimum possible (u64)value */
+	u64 umax_value; /* maximum possible (u64)value */
 };
 
 enum bpf_stack_slot_type {