linux/kernel/bpf
Eduard Zingerman 71f656a501 bpf: Fix to preserve reg parent/live fields when copying range info
Register range information is copied in several places. The intent is
to transfer range/id information from one register/stack spill to
another. Currently this is done using direct register assignment, e.g.:

static void find_equal_scalars(..., struct bpf_reg_state *known_reg)
{
	...
	struct bpf_reg_state *reg;
	...
			*reg = *known_reg;
	...
}

However, such assignments also copy the following bpf_reg_state fields:

struct bpf_reg_state {
	...
	struct bpf_reg_state *parent;
	...
	enum bpf_reg_liveness live;
	...
};

Copying of these fields is accidental and incorrect, as could be
demonstrated by the following example:

     0: call ktime_get_ns()
     1: r6 = r0
     2: call ktime_get_ns()
     3: r7 = r0
     4: if r0 > r6 goto +1             ; r0 & r6 are unbound thus generated
                                       ; branch states are identical
     5: *(u64 *)(r10 - 8) = 0xdeadbeef ; 64-bit write to fp[-8]
    --- checkpoint ---
     6: r1 = 42                        ; r1 marked as written
     7: *(u8 *)(r10 - 8) = r1          ; 8-bit write, fp[-8] parent & live
                                       ; overwritten
     8: r2 = *(u64 *)(r10 - 8)
     9: r0 = 0
    10: exit

This example is unsafe because 64-bit write to fp[-8] at (5) is
conditional, thus not all bytes of fp[-8] are guaranteed to be set
when it is read at (8). However, currently the example passes
verification.

First, the execution path 1-10 is examined by verifier.
Suppose that a new checkpoint is created by is_state_visited() at (6).
After checkpoint creation:
- r1.parent points to checkpoint.r1,
- fp[-8].parent points to checkpoint.fp[-8].
At (6) the r1.live is set to REG_LIVE_WRITTEN.
At (7) the fp[-8].parent is set to r1.parent and fp[-8].live is set to
REG_LIVE_WRITTEN, because of the following code called in
check_stack_write_fixed_off():

static void save_register_state(struct bpf_func_state *state,
				int spi, struct bpf_reg_state *reg,
				int size)
{
	...
	state->stack[spi].spilled_ptr = *reg;  // <--- parent & live copied
	if (size == BPF_REG_SIZE)
		state->stack[spi].spilled_ptr.live |= REG_LIVE_WRITTEN;
	...
}

Note the intent to mark stack spill as written only if 8 bytes are
spilled to a slot, however this intent is spoiled by a 'live' field copy.
At (8) the checkpoint.fp[-8] should be marked as REG_LIVE_READ but
this does not happen:
- fp[-8] in a current state is already marked as REG_LIVE_WRITTEN;
- fp[-8].parent points to checkpoint.r1, parentage chain is used by
  mark_reg_read() to mark checkpoint states.
At (10) the verification is finished for path 1-10 and jump 4-6 is
examined. The checkpoint.fp[-8] never gets REG_LIVE_READ mark and this
spill is pruned from the cached states by clean_live_states(). Hence
verifier state obtained via path 1-4,6 is deemed identical to one
obtained via path 1-6 and program marked as safe.

Note: the example should be executed with BPF_F_TEST_STATE_FREQ flag
set to force creation of intermediate verifier states.

This commit revisits the locations where bpf_reg_state instances are
copied and replaces the direct copies with a call to a function
copy_register_state(dst, src) that preserves 'parent' and 'live'
fields of the 'dst'.

Fixes: 679c782de1 ("bpf/verifier: per-register parent pointers")
Signed-off-by: Eduard Zingerman <eddyz87@gmail.com>
Link: https://lore.kernel.org/r/20230106142214.1040390-2-eddyz87@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
2023-01-19 15:19:23 -08:00
..
preload bpf: iterators: Build and use lightweight bootstrap version of bpftool 2022-07-15 12:01:30 -07:00
arraymap.c bpf: Do btf_record_free outside map_free callback 2022-11-17 19:11:31 -08:00
bloom_filter.c treewide: use get_random_u32() when possible 2022-10-11 17:42:58 -06:00
bpf_cgrp_storage.c bpf: Fix a compilation failure with clang lto build 2022-11-30 17:13:25 -08:00
bpf_inode_storage.c bpf: Fix a compilation failure with clang lto build 2022-11-30 17:13:25 -08:00
bpf_iter.c bpf: Initialize the bpf_run_ctx in bpf_iter_run_prog() 2022-08-18 17:06:13 -07:00
bpf_local_storage.c Merge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net 2022-11-29 13:04:52 -08:00
bpf_lru_list.c
bpf_lru_list.h
bpf_lsm.c bpf: Define sock security related BTF IDs under CONFIG_SECURITY_NETWORK 2022-12-19 22:02:17 +01:00
bpf_struct_ops.c mm: Introduce set_memory_rox() 2022-12-15 10:37:26 -08:00
bpf_struct_ops_types.h
bpf_task_storage.c bpf: Fix a compilation failure with clang lto build 2022-11-30 17:13:25 -08:00
btf.c for-alexei-2022120701 2022-12-07 13:49:21 -08:00
cgroup.c Merge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net 2022-10-03 17:44:18 -07:00
cgroup_iter.c bpf: Pin the start cgroup in cgroup_iter_seq_init() 2022-11-21 17:40:42 +01:00
core.c Including fixes from bpf, netfilter and can. 2022-12-21 08:41:32 -08:00
cpumap.c bpf: Expand map key argument of bpf_redirect_map to u64 2022-11-15 09:00:27 -08:00
devmap.c bpf: Expand map key argument of bpf_redirect_map to u64 2022-11-15 09:00:27 -08:00
disasm.c
disasm.h
dispatcher.c bpf: Synchronize dispatcher update with bpf_dispatcher_xdp_func 2022-12-14 12:02:14 -08:00
hashtab.c bpf: hash map, avoid deadlock with suitable hash mask 2023-01-12 18:55:42 -08:00
helpers.c bpf: Use memmove for bpf_dynptr_{read,write} 2022-12-08 18:39:28 -08:00
inode.c
Kconfig
link_iter.c
local_storage.c bpf: Consolidate spin_lock, timer management into btf_record 2022-11-03 22:19:40 -07:00
lpm_trie.c bpf: Use bpf_map_area_alloc consistently on bpf map creation 2022-08-10 11:50:43 -07:00
Makefile bpf: Implement cgroup storage available to non-cgroup-attached bpf progs 2022-10-25 23:19:19 -07:00
map_in_map.c bpf: Add comments for map BTF matching requirement for bpf_list_head 2022-11-17 19:22:14 -08:00
map_in_map.h
map_iter.c
memalloc.c bpf: Fix off-by-one error in bpf_mem_cache_idx() 2023-01-18 18:36:26 -08:00
mmap_unlock_work.h
net_namespace.c
offload.c bpf: restore the ebpf program ID for BPF_AUDIT_UNLOAD and PERF_BPF_EVENT_PROG_UNLOAD 2023-01-09 19:47:58 -08:00
percpu_freelist.c bpf: Initialize same number of free nodes for each pcpu_freelist 2022-11-11 12:05:14 -08:00
percpu_freelist.h
prog_iter.c
queue_stack_maps.c bpf: Remove unneeded memset in queue_stack_map creation 2022-08-10 11:48:22 -07:00
reuseport_array.c net: Fix suspicious RCU usage in bpf_sk_reuseport_detach() 2022-08-17 16:42:59 -07:00
ringbuf.c bpf: Rename MEM_ALLOC to MEM_RINGBUF 2022-11-14 21:52:45 -08:00
stackmap.c perf/bpf: Always use perf callchains if exist 2022-09-13 15:03:22 +02:00
syscall.c bpf: remove the do_idr_lock parameter from bpf_prog_free_id() 2023-01-09 19:47:59 -08:00
sysfs_btf.c
task_iter.c bpf: keep a reference to the mm, in case the task is dead. 2022-12-28 14:11:48 -08:00
tnum.c
trampoline.c bpf: Fix panic due to wrong pageattr of im->image 2022-12-28 13:46:28 -08:00
verifier.c bpf: Fix to preserve reg parent/live fields when copying range info 2023-01-19 15:19:23 -08:00