commit 0803278b0b upstream.
Syzkaller hit 'KASAN: use-after-free Write in sanitize_ptr_alu' bug.
Call trace:
dump_stack+0xbf/0x12e
print_address_description+0x6a/0x280
kasan_report+0x237/0x360
sanitize_ptr_alu+0x85a/0x8d0
adjust_ptr_min_max_vals+0x8f2/0x1ca0
adjust_reg_min_max_vals+0x8ed/0x22e0
do_check+0x1ca6/0x5d00
bpf_check+0x9ca/0x2570
bpf_prog_load+0xc91/0x1030
__se_sys_bpf+0x61e/0x1f00
do_syscall_64+0xc8/0x550
entry_SYSCALL_64_after_hwframe+0x49/0xbe
Fault injection trace:
kfree+0xea/0x290
free_func_state+0x4a/0x60
free_verifier_state+0x61/0xe0
push_stack+0x216/0x2f0 <- inject failslab
sanitize_ptr_alu+0x2b1/0x8d0
adjust_ptr_min_max_vals+0x8f2/0x1ca0
adjust_reg_min_max_vals+0x8ed/0x22e0
do_check+0x1ca6/0x5d00
bpf_check+0x9ca/0x2570
bpf_prog_load+0xc91/0x1030
__se_sys_bpf+0x61e/0x1f00
do_syscall_64+0xc8/0x550
entry_SYSCALL_64_after_hwframe+0x49/0xbe
When kzalloc() fails in push_stack(), free_verifier_state() will free
current verifier state. As push_stack() returns, dst_reg was restored
if ptr_is_dst_reg is false. However, as member of the cur_state,
dst_reg is also freed, and error occurs when dereferencing dst_reg.
Simply fix it by testing ret of push_stack() before restoring dst_reg.
Fixes: 979d63d50c ("bpf: prevent out of bounds speculation on pointer arithmetic")
Signed-off-by: Xu Yu <xuyu@linux.alibaba.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 7c0cdf0b39 ]
trie_delete_elem() was deleting an entry even though it was not matching
if the prefixlen was correct. This patch adds a check on matchlen.
Reproducer:
$ sudo bpftool map create /sys/fs/bpf/mylpm type lpm_trie key 8 value 1 entries 128 name mylpm flags 1
$ sudo bpftool map update pinned /sys/fs/bpf/mylpm key hex 10 00 00 00 aa bb cc dd value hex 01
$ sudo bpftool map dump pinned /sys/fs/bpf/mylpm
key: 10 00 00 00 aa bb cc dd value: 01
Found 1 element
$ sudo bpftool map delete pinned /sys/fs/bpf/mylpm key hex 10 00 00 00 ff ff ff ff
$ echo $?
0
$ sudo bpftool map dump pinned /sys/fs/bpf/mylpm
Found 0 elements
A similar reproducer is added in the selftests.
Without the patch:
$ sudo ./tools/testing/selftests/bpf/test_lpm_map
test_lpm_map: test_lpm_map.c:485: test_lpm_delete: Assertion `bpf_map_delete_elem(map_fd, key) == -1 && errno == ENOENT' failed.
Aborted
With the patch: test_lpm_map runs without errors.
Fixes: e454cf5958 ("bpf: Implement map_delete_elem for BPF_MAP_TYPE_LPM_TRIE")
Cc: Craig Gallek <kraig@google.com>
Signed-off-by: Alban Crequy <alban@kinvolk.io>
Acked-by: Craig Gallek <kraig@google.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 7c4cd051ad ]
The map_lookup_elem used to not acquiring spinlock
in order to optimize the reader.
It was true until commit 557c0c6e7d ("bpf: convert stackmap to pre-allocation")
The syscall's map_lookup_elem(stackmap) calls bpf_stackmap_copy().
bpf_stackmap_copy() may find the elem no longer needed after the copy is done.
If that is the case, pcpu_freelist_push() saves this elem for reuse later.
This push requires a spinlock.
If a tracing bpf_prog got run in the middle of the syscall's
map_lookup_elem(stackmap) and this tracing bpf_prog is calling
bpf_get_stackid(stackmap) which also requires the same pcpu_freelist's
spinlock, it may end up with a dead lock situation as reported by
Eric Dumazet in https://patchwork.ozlabs.org/patch/1030266/
The situation is the same as the syscall's map_update_elem() which
needs to acquire the pcpu_freelist's spinlock and could race
with tracing bpf_prog. Hence, this patch fixes it by protecting
bpf_stackmap_copy() with this_cpu_inc(bpf_prog_active)
to prevent tracing bpf_prog from running.
A later syscall's map_lookup_elem commit f1a2e44a3a ("bpf: add queue and stack maps")
also acquires a spinlock and races with tracing bpf_prog similarly.
Hence, this patch is forward looking and protects the majority
of the map lookups. bpf_map_offload_lookup_elem() is the exception
since it is for network bpf_prog only (i.e. never called by tracing
bpf_prog).
Fixes: 557c0c6e7d ("bpf: convert stackmap to pre-allocation")
Reported-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 3612af783c upstream.
Marek reported that he saw an issue with the below snippet in that
timing measurements where off when loaded as unpriv while results
were reasonable when loaded as privileged:
[...]
uint64_t a = bpf_ktime_get_ns();
uint64_t b = bpf_ktime_get_ns();
uint64_t delta = b - a;
if ((int64_t)delta > 0) {
[...]
Turns out there is a bug where a corner case is missing in the fix
d3bd7413e0 ("bpf: fix sanitation of alu op with pointer / scalar
type from different paths"), namely fixup_bpf_calls() only checks
whether aux has a non-zero alu_state, but it also needs to test for
the case of BPF_ALU_NON_POINTER since in both occasions we need to
skip the masking rewrite (as there is nothing to mask).
Fixes: d3bd7413e0 ("bpf: fix sanitation of alu op with pointer / scalar type from different paths")
Reported-by: Marek Majkowski <marek@cloudflare.com>
Reported-by: Arthur Fabre <afabre@cloudflare.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/netdev/CAJPywTJqP34cK20iLM5YmUMz9KXQOdu1-+BZrGMAGgLuBWz7fg@mail.gmail.com/T/
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 4af396ae48 ]
When returning BPF_STACK_BUILD_ID_IP from stack_map_get_build_id_offset,
make sure that build_id field is empty. Since we are using percpu
free list, there is a possibility that we might reuse some previous
bpf_stack_build_id with non-zero build_id.
Fixes: 615755a77b ("bpf: extend stackmap to save binary_build_id+offset instead of address")
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Stanislav Fomichev <sdf@google.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 0b698005a9 ]
Build-id length is not fixed to 20, it can be (`man ld` /--build-id):
* 128-bit (uuid)
* 160-bit (sha1)
* any length specified in ld --build-id=0xhexstring
To fix the issue of missing BPF_STACK_BUILD_ID_VALID for shorter build-ids,
assume that build-id is somewhere in the range of 1 .. 20.
Set the remaining bytes to zero.
v2:
* don't introduce new "len = min(BPF_BUILD_ID_SIZE, nhdr->n_descsz)",
we already know that nhdr->n_descsz <= BPF_BUILD_ID_SIZE if we enter
this 'if' condition
Fixes: 615755a77b ("bpf: extend stackmap to save binary_build_id+offset instead of address")
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Stanislav Fomichev <sdf@google.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit beaf3d1901 ]
As Naresh reported, test_stacktrace_build_id() causes panic on i386 and
arm32 systems. This is caused by page_address() returns NULL in certain
cases.
This patch fixes this error by using kmap_atomic/kunmap_atomic instead
of page_address.
Fixes: 615755a77b (" bpf: extend stackmap to save binary_build_id+offset instead of address")
Reported-by: Naresh Kamboju <naresh.kamboju@linaro.org>
Signed-off-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ commit d3bd7413e0 upstream ]
While 979d63d50c ("bpf: prevent out of bounds speculation on pointer
arithmetic") took care of rejecting alu op on pointer when e.g. pointer
came from two different map values with different map properties such as
value size, Jann reported that a case was not covered yet when a given
alu op is used in both "ptr_reg += reg" and "numeric_reg += reg" from
different branches where we would incorrectly try to sanitize based
on the pointer's limit. Catch this corner case and reject the program
instead.
Fixes: 979d63d50c ("bpf: prevent out of bounds speculation on pointer arithmetic")
Reported-by: Jann Horn <jannh@google.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ commit 979d63d50c upstream ]
Jann reported that the original commit back in b2157399cc
("bpf: prevent out-of-bounds speculation") was not sufficient
to stop CPU from speculating out of bounds memory access:
While b2157399cc only focussed on masking array map access
for unprivileged users for tail calls and data access such
that the user provided index gets sanitized from BPF program
and syscall side, there is still a more generic form affected
from BPF programs that applies to most maps that hold user
data in relation to dynamic map access when dealing with
unknown scalars or "slow" known scalars as access offset, for
example:
- Load a map value pointer into R6
- Load an index into R7
- Do a slow computation (e.g. with a memory dependency) that
loads a limit into R8 (e.g. load the limit from a map for
high latency, then mask it to make the verifier happy)
- Exit if R7 >= R8 (mispredicted branch)
- Load R0 = R6[R7]
- Load R0 = R6[R0]
For unknown scalars there are two options in the BPF verifier
where we could derive knowledge from in order to guarantee
safe access to the memory: i) While </>/<=/>= variants won't
allow to derive any lower or upper bounds from the unknown
scalar where it would be safe to add it to the map value
pointer, it is possible through ==/!= test however. ii) another
option is to transform the unknown scalar into a known scalar,
for example, through ALU ops combination such as R &= <imm>
followed by R |= <imm> or any similar combination where the
original information from the unknown scalar would be destroyed
entirely leaving R with a constant. The initial slow load still
precedes the latter ALU ops on that register, so the CPU
executes speculatively from that point. Once we have the known
scalar, any compare operation would work then. A third option
only involving registers with known scalars could be crafted
as described in [0] where a CPU port (e.g. Slow Int unit)
would be filled with many dependent computations such that
the subsequent condition depending on its outcome has to wait
for evaluation on its execution port and thereby executing
speculatively if the speculated code can be scheduled on a
different execution port, or any other form of mistraining
as described in [1], for example. Given this is not limited
to only unknown scalars, not only map but also stack access
is affected since both is accessible for unprivileged users
and could potentially be used for out of bounds access under
speculation.
In order to prevent any of these cases, the verifier is now
sanitizing pointer arithmetic on the offset such that any
out of bounds speculation would be masked in a way where the
pointer arithmetic result in the destination register will
stay unchanged, meaning offset masked into zero similar as
in array_index_nospec() case. With regards to implementation,
there are three options that were considered: i) new insn
for sanitation, ii) push/pop insn and sanitation as inlined
BPF, iii) reuse of ax register and sanitation as inlined BPF.
Option i) has the downside that we end up using from reserved
bits in the opcode space, but also that we would require
each JIT to emit masking as native arch opcodes meaning
mitigation would have slow adoption till everyone implements
it eventually which is counter-productive. Option ii) and iii)
have both in common that a temporary register is needed in
order to implement the sanitation as inlined BPF since we
are not allowed to modify the source register. While a push /
pop insn in ii) would be useful to have in any case, it
requires once again that every JIT needs to implement it
first. While possible, amount of changes needed would also
be unsuitable for a -stable patch. Therefore, the path which
has fewer changes, less BPF instructions for the mitigation
and does not require anything to be changed in the JITs is
option iii) which this work is pursuing. The ax register is
already mapped to a register in all JITs (modulo arm32 where
it's mapped to stack as various other BPF registers there)
and used in constant blinding for JITs-only so far. It can
be reused for verifier rewrites under certain constraints.
The interpreter's tmp "register" has therefore been remapped
into extending the register set with hidden ax register and
reusing that for a number of instructions that needed the
prior temporary variable internally (e.g. div, mod). This
allows for zero increase in stack space usage in the interpreter,
and enables (restricted) generic use in rewrites otherwise as
long as such a patchlet does not make use of these instructions.
The sanitation mask is dynamic and relative to the offset the
map value or stack pointer currently holds.
There are various cases that need to be taken under consideration
for the masking, e.g. such operation could look as follows:
ptr += val or val += ptr or ptr -= val. Thus, the value to be
sanitized could reside either in source or in destination
register, and the limit is different depending on whether
the ALU op is addition or subtraction and depending on the
current known and bounded offset. The limit is derived as
follows: limit := max_value_size - (smin_value + off). For
subtraction: limit := umax_value + off. This holds because
we do not allow any pointer arithmetic that would
temporarily go out of bounds or would have an unknown
value with mixed signed bounds where it is unclear at
verification time whether the actual runtime value would
be either negative or positive. For example, we have a
derived map pointer value with constant offset and bounded
one, so limit based on smin_value works because the verifier
requires that statically analyzed arithmetic on the pointer
must be in bounds, and thus it checks if resulting
smin_value + off and umax_value + off is still within map
value bounds at time of arithmetic in addition to time of
access. Similarly, for the case of stack access we derive
the limit as follows: MAX_BPF_STACK + off for subtraction
and -off for the case of addition where off := ptr_reg->off +
ptr_reg->var_off.value. Subtraction is a special case for
the masking which can be in form of ptr += -val, ptr -= -val,
or ptr -= val. In the first two cases where we know that
the value is negative, we need to temporarily negate the
value in order to do the sanitation on a positive value
where we later swap the ALU op, and restore original source
register if the value was in source.
The sanitation of pointer arithmetic alone is still not fully
sufficient as is, since a scenario like the following could
happen ...
PTR += 0x1000 (e.g. K-based imm)
PTR -= BIG_NUMBER_WITH_SLOW_COMPARISON
PTR += 0x1000
PTR -= BIG_NUMBER_WITH_SLOW_COMPARISON
[...]
... which under speculation could end up as ...
PTR += 0x1000
PTR -= 0 [ truncated by mitigation ]
PTR += 0x1000
PTR -= 0 [ truncated by mitigation ]
[...]
... and therefore still access out of bounds. To prevent such
case, the verifier is also analyzing safety for potential out
of bounds access under speculative execution. Meaning, it is
also simulating pointer access under truncation. We therefore
"branch off" and push the current verification state after the
ALU operation with known 0 to the verification stack for later
analysis. Given the current path analysis succeeded it is
likely that the one under speculation can be pruned. In any
case, it is also subject to existing complexity limits and
therefore anything beyond this point will be rejected. In
terms of pruning, it needs to be ensured that the verification
state from speculative execution simulation must never prune
a non-speculative execution path, therefore, we mark verifier
state accordingly at the time of push_stack(). If verifier
detects out of bounds access under speculative execution from
one of the possible paths that includes a truncation, it will
reject such program.
Given we mask every reg-based pointer arithmetic for
unprivileged programs, we've been looking into how it could
affect real-world programs in terms of size increase. As the
majority of programs are targeted for privileged-only use
case, we've unconditionally enabled masking (with its alu
restrictions on top of it) for privileged programs for the
sake of testing in order to check i) whether they get rejected
in its current form, and ii) by how much the number of
instructions and size will increase. We've tested this by
using Katran, Cilium and test_l4lb from the kernel selftests.
For Katran we've evaluated balancer_kern.o, Cilium bpf_lxc.o
and an older test object bpf_lxc_opt_-DUNKNOWN.o and l4lb
we've used test_l4lb.o as well as test_l4lb_noinline.o. We
found that none of the programs got rejected by the verifier
with this change, and that impact is rather minimal to none.
balancer_kern.o had 13,904 bytes (1,738 insns) xlated and
7,797 bytes JITed before and after the change. Most complex
program in bpf_lxc.o had 30,544 bytes (3,817 insns) xlated
and 18,538 bytes JITed before and after and none of the other
tail call programs in bpf_lxc.o had any changes either. For
the older bpf_lxc_opt_-DUNKNOWN.o object we found a small
increase from 20,616 bytes (2,576 insns) and 12,536 bytes JITed
before to 20,664 bytes (2,582 insns) and 12,558 bytes JITed
after the change. Other programs from that object file had
similar small increase. Both test_l4lb.o had no change and
remained at 6,544 bytes (817 insns) xlated and 3,401 bytes
JITed and for test_l4lb_noinline.o constant at 5,080 bytes
(634 insns) xlated and 3,313 bytes JITed. This can be explained
in that LLVM typically optimizes stack based pointer arithmetic
by using K-based operations and that use of dynamic map access
is not overly frequent. However, in future we may decide to
optimize the algorithm further under known guarantees from
branch and value speculation. Latter seems also unclear in
terms of prediction heuristics that today's CPUs apply as well
as whether there could be collisions in e.g. the predictor's
Value History/Pattern Table for triggering out of bounds access,
thus masking is performed unconditionally at this point but could
be subject to relaxation later on. We were generally also
brainstorming various other approaches for mitigation, but the
blocker was always lack of available registers at runtime and/or
overhead for runtime tracking of limits belonging to a specific
pointer. Thus, we found this to be minimally intrusive under
given constraints.
With that in place, a simple example with sanitized access on
unprivileged load at post-verification time looks as follows:
# bpftool prog dump xlated id 282
[...]
28: (79) r1 = *(u64 *)(r7 +0)
29: (79) r2 = *(u64 *)(r7 +8)
30: (57) r1 &= 15
31: (79) r3 = *(u64 *)(r0 +4608)
32: (57) r3 &= 1
33: (47) r3 |= 1
34: (2d) if r2 > r3 goto pc+19
35: (b4) (u32) r11 = (u32) 20479 |
36: (1f) r11 -= r2 | Dynamic sanitation for pointer
37: (4f) r11 |= r2 | arithmetic with registers
38: (87) r11 = -r11 | containing bounded or known
39: (c7) r11 s>>= 63 | scalars in order to prevent
40: (5f) r11 &= r2 | out of bounds speculation.
41: (0f) r4 += r11 |
42: (71) r4 = *(u8 *)(r4 +0)
43: (6f) r4 <<= r1
[...]
For the case where the scalar sits in the destination register
as opposed to the source register, the following code is emitted
for the above example:
[...]
16: (b4) (u32) r11 = (u32) 20479
17: (1f) r11 -= r2
18: (4f) r11 |= r2
19: (87) r11 = -r11
20: (c7) r11 s>>= 63
21: (5f) r2 &= r11
22: (0f) r2 += r0
23: (61) r0 = *(u32 *)(r2 +0)
[...]
JIT blinding example with non-conflicting use of r10:
[...]
d5: je 0x0000000000000106 _
d7: mov 0x0(%rax),%edi |
da: mov $0xf153246,%r10d | Index load from map value and
e0: xor $0xf153259,%r10 | (const blinded) mask with 0x1f.
e7: and %r10,%rdi |_
ea: mov $0x2f,%r10d |
f0: sub %rdi,%r10 | Sanitized addition. Both use r10
f3: or %rdi,%r10 | but do not interfere with each
f6: neg %r10 | other. (Neither do these instructions
f9: sar $0x3f,%r10 | interfere with the use of ax as temp
fd: and %r10,%rdi | in interpreter.)
100: add %rax,%rdi |_
103: mov 0x0(%rdi),%eax
[...]
Tested that it fixes Jann's reproducer, and also checked that test_verifier
and test_progs suite with interpreter, JIT and JIT with hardening enabled
on x86-64 and arm64 runs successfully.
[0] Speculose: Analyzing the Security Implications of Speculative
Execution in CPUs, Giorgi Maisuradze and Christian Rossow,
https://arxiv.org/pdf/1801.04084.pdf
[1] A Systematic Evaluation of Transient Execution Attacks and
Defenses, Claudio Canella, Jo Van Bulck, Michael Schwarz,
Moritz Lipp, Benjamin von Berg, Philipp Ortner, Frank Piessens,
Dmitry Evtyushkin, Daniel Gruss,
https://arxiv.org/pdf/1811.05441.pdf
Fixes: b2157399cc ("bpf: prevent out-of-bounds speculation")
Reported-by: Jann Horn <jannh@google.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ commit b7137c4eab upstream ]
In check_map_access() we probe actual bounds through __check_map_access()
with offset of reg->smin_value + off for lower bound and offset of
reg->umax_value + off for the upper bound. However, even though the
reg->smin_value could have a negative value, the final result of the
sum with off could be positive when pointer arithmetic with known and
unknown scalars is combined. In this case we reject the program with
an error such as "R<x> min value is negative, either use unsigned index
or do a if (index >=0) check." even though the access itself would be
fine. Therefore extend the check to probe whether the actual resulting
reg->smin_value + off is less than zero.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ commit 9d7eceede7 upstream ]
For unknown scalars of mixed signed bounds, meaning their smin_value is
negative and their smax_value is positive, we need to reject arithmetic
with pointer to map value. For unprivileged the goal is to mask every
map pointer arithmetic and this cannot reliably be done when it is
unknown at verification time whether the scalar value is negative or
positive. Given this is a corner case, the likelihood of breaking should
be very small.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ commit e4298d2583 upstream ]
Restrict stack pointer arithmetic for unprivileged users in that
arithmetic itself must not go out of bounds as opposed to the actual
access later on. Therefore after each adjust_ptr_min_max_vals() with
a stack pointer as a destination we simulate a check_stack_access()
of 1 byte on the destination and once that fails the program is
rejected for unprivileged program loads. This is analog to map
value pointer arithmetic and needed for masking later on.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ commit 0d6303db79 upstream ]
Restrict map value pointer arithmetic for unprivileged users in that
arithmetic itself must not go out of bounds as opposed to the actual
access later on. Therefore after each adjust_ptr_min_max_vals() with a
map value pointer as a destination it will simulate a check_map_access()
of 1 byte on the destination and once that fails the program is rejected
for unprivileged program loads. We use this later on for masking any
pointer arithmetic with the remainder of the map value space. The
likelihood of breaking any existing real-world unprivileged eBPF
program is very small for this corner case.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ commit 9b73bfdd08 upstream ]
Right now we are using BPF ax register in JIT for constant blinding as
well as in interpreter as temporary variable. Verifier will not be able
to use it simply because its use will get overridden from the former in
bpf_jit_blind_insn(). However, it can be made to work in that blinding
will be skipped if there is prior use in either source or destination
register on the instruction. Taking constraints of ax into account, the
verifier is then open to use it in rewrites under some constraints. Note,
ax register already has mappings in every eBPF JIT.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ commit 144cd91c4c upstream ]
This change moves the on-stack 64 bit tmp variable in ___bpf_prog_run()
into the hidden ax register. The latter is currently only used in JITs
for constant blinding as a temporary scratch register, meaning the BPF
interpreter will never see the use of ax. Therefore it is safe to use
it for the cases where tmp has been used earlier. This is needed to later
on allow restricted hidden use of ax in both interpreter and JITs.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ commit c08435ec7f upstream ]
Move prev_insn_idx and insn_idx from the do_check() function into
the verifier environment, so they can be read inside the various
helper functions for handling the instructions. It's easier to put
this into the environment rather than changing all call-sites only
to pass it along. insn_idx is useful in particular since this later
on allows to hold state in env->insn_aux_data[env->insn_idx].
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ commit ceefbc96fa upstream ]
malicious bpf program may try to force the verifier to remember
a lot of distinct verifier states.
Put a limit to number of per-insn 'struct bpf_verifier_state'.
Note that hitting the limit doesn't reject the program.
It potentially makes the verifier do more steps to analyze the program.
It means that malicious programs will hit BPF_COMPLEXITY_LIMIT_INSNS sooner
instead of spending cpu time walking long link list.
The limit of BPF_COMPLEXITY_LIMIT_STATES==64 affects cilium progs
with slight increase in number of "steps" it takes to successfully verify
the programs:
before after
bpf_lb-DLB_L3.o 1940 1940
bpf_lb-DLB_L4.o 3089 3089
bpf_lb-DUNKNOWN.o 1065 1065
bpf_lxc-DDROP_ALL.o 28052 | 28162
bpf_lxc-DUNKNOWN.o 35487 | 35541
bpf_netdev.o 10864 10864
bpf_overlay.o 6643 6643
bpf_lcx_jit.o 38437 38437
But it also makes malicious program to be rejected in 0.4 seconds vs 6.5
Hence apply this limit to unprivileged programs only.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ commit 4f7b3e8258 upstream ]
pathological bpf programs may try to force verifier to explode in
the number of branch states:
20: (d5) if r1 s<= 0x24000028 goto pc+0
21: (b5) if r0 <= 0xe1fa20 goto pc+2
22: (d5) if r1 s<= 0x7e goto pc+0
23: (b5) if r0 <= 0xe880e000 goto pc+0
24: (c5) if r0 s< 0x2100ecf4 goto pc+0
25: (d5) if r1 s<= 0xe880e000 goto pc+1
26: (c5) if r0 s< 0xf4041810 goto pc+0
27: (d5) if r1 s<= 0x1e007e goto pc+0
28: (b5) if r0 <= 0xe86be000 goto pc+0
29: (07) r0 += 16614
30: (c5) if r0 s< 0x6d0020da goto pc+0
31: (35) if r0 >= 0x2100ecf4 goto pc+0
Teach verifier to recognize always taken and always not taken branches.
This analysis is already done for == and != comparison.
Expand it to all other branches.
It also helps real bpf programs to be verified faster:
before after
bpf_lb-DLB_L3.o 2003 1940
bpf_lb-DLB_L4.o 3173 3089
bpf_lb-DUNKNOWN.o 1080 1065
bpf_lxc-DDROP_ALL.o 29584 28052
bpf_lxc-DUNKNOWN.o 36916 35487
bpf_netdev.o 11188 10864
bpf_overlay.o 6679 6643
bpf_lcx_jit.o 39555 38437
Reported-by: Anatoly Trosinenko <anatoly.trosinenko@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit e434b8cdf7 ]
Currently, the destination register is marked as unknown for 32-bit
sub-register move (BPF_MOV | BPF_ALU) whenever the source register type is
SCALAR_VALUE.
This is too conservative that some valid cases will be rejected.
Especially, this may turn a constant scalar value into unknown value that
could break some assumptions of verifier.
For example, test_l4lb_noinline.c has the following C code:
struct real_definition *dst
1: if (!get_packet_dst(&dst, &pckt, vip_info, is_ipv6))
2: return TC_ACT_SHOT;
3:
4: if (dst->flags & F_IPV6) {
get_packet_dst is responsible for initializing "dst" into valid pointer and
return true (1), otherwise return false (0). The compiled instruction
sequence using alu32 will be:
412: (54) (u32) r7 &= (u32) 1
413: (bc) (u32) r0 = (u32) r7
414: (95) exit
insn 413, a BPF_MOV | BPF_ALU, however will turn r0 into unknown value even
r7 contains SCALAR_VALUE 1.
This causes trouble when verifier is walking the code path that hasn't
initialized "dst" inside get_packet_dst, for which case 0 is returned and
we would then expect verifier concluding line 1 in the above C code pass
the "if" check, therefore would skip fall through path starting at line 4.
Now, because r0 returned from callee has became unknown value, so verifier
won't skip analyzing path starting at line 4 and "dst->flags" requires
dereferencing the pointer "dst" which actually hasn't be initialized for
this path.
This patch relaxed the code marking sub-register move destination. For a
SCALAR_VALUE, it is safe to just copy the value from source then truncate
it into 32-bit.
A unit test also included to demonstrate this issue. This test will fail
before this patch.
This relaxation could let verifier skipping more paths for conditional
comparison against immediate. It also let verifier recording a more
accurate/strict value for one register at one state, if this state end up
with going through exit without rejection and it is used for state
comparison later, then it is possible an inaccurate/permissive value is
better. So the real impact on verifier processed insn number is complex.
But in all, without this fix, valid program could be rejected.
>From real benchmarking on kernel selftests and Cilium bpf tests, there is
no impact on processed instruction number when tests ares compiled with
default compilation options. There is slightly improvements when they are
compiled with -mattr=+alu32 after this patch.
Also, test_xdp_noinline/-mattr=+alu32 now passed verification. It is
rejected before this fix.
Insn processed before/after this patch:
default -mattr=+alu32
Kernel selftest
===
test_xdp.o 371/371 369/369
test_l4lb.o 6345/6345 5623/5623
test_xdp_noinline.o 2971/2971 rejected/2727
test_tcp_estates.o 429/429 430/430
Cilium bpf
===
bpf_lb-DLB_L3.o: 2085/2085 1685/1687
bpf_lb-DLB_L4.o: 2287/2287 1986/1982
bpf_lb-DUNKNOWN.o: 690/690 622/622
bpf_lxc.o: 95033/95033 N/A
bpf_netdev.o: 7245/7245 N/A
bpf_overlay.o: 2898/2898 3085/2947
NOTE:
- bpf_lxc.o and bpf_netdev.o compiled by -mattr=+alu32 are rejected by
verifier due to another issue inside verifier on supporting alu32
binary.
- Each cilium bpf program could generate several processed insn number,
above number is sum of them.
v1->v2:
- Restrict the change on SCALAR_VALUE.
- Update benchmark numbers on Cilium bpf tests.
Signed-off-by: Jiong Wang <jiong.wang@netronome.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit 46f53a65d2 ]
Currently BPF verifier allows narrow loads for a context field only with
offset zero. E.g. if there is a __u32 field then only the following
loads are permitted:
* off=0, size=1 (narrow);
* off=0, size=2 (narrow);
* off=0, size=4 (full).
On the other hand LLVM can generate a load with offset different than
zero that make sense from program logic point of view, but verifier
doesn't accept it.
E.g. tools/testing/selftests/bpf/sendmsg4_prog.c has code:
#define DST_IP4 0xC0A801FEU /* 192.168.1.254 */
...
if ((ctx->user_ip4 >> 24) == (bpf_htonl(DST_IP4) >> 24) &&
where ctx is struct bpf_sock_addr.
Some versions of LLVM can produce the following byte code for it:
8: 71 12 07 00 00 00 00 00 r2 = *(u8 *)(r1 + 7)
9: 67 02 00 00 18 00 00 00 r2 <<= 24
10: 18 03 00 00 00 00 00 fe 00 00 00 00 00 00 00 00 r3 = 4261412864 ll
12: 5d 32 07 00 00 00 00 00 if r2 != r3 goto +7 <LBB0_6>
where `*(u8 *)(r1 + 7)` means narrow load for ctx->user_ip4 with size=1
and offset=3 (7 - sizeof(ctx->user_family) = 3). This load is currently
rejected by verifier.
Verifier code that rejects such loads is in bpf_ctx_narrow_access_ok()
what means any is_valid_access implementation, that uses the function,
works this way, e.g. bpf_skb_is_valid_access() for __sk_buff or
sock_addr_is_valid_access() for bpf_sock_addr.
The patch makes such loads supported. Offset can be in [0; size_default)
but has to be multiple of load size. E.g. for __u32 field the following
loads are supported now:
* off=0, size=1 (narrow);
* off=1, size=1 (narrow);
* off=2, size=1 (narrow);
* off=3, size=1 (narrow);
* off=0, size=2 (narrow);
* off=2, size=2 (narrow);
* off=0, size=4 (full).
Reported-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Andrey Ignatov <rdna@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
[ Upstream commit c3494801cd ]
Malicious user space may try to force the verifier to use as much cpu
time and memory as possible. Hence check for pending signals
while verifying the program.
Note that suspend of sys_bpf(PROG_LOAD) syscall will lead to EAGAIN,
since the kernel has to release the resources used for program verification.
Reported-by: Anatoly Trosinenko <anatoly.trosinenko@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit afd5942408 upstream.
When patching in a new sequence for the first insn of a subprog, the start
of that subprog does not change (it's the first insn of the sequence), so
adjust_subprog_starts should check start <= off (rather than < off).
Also added a test to test_verifier.c (it's essentially the syz reproducer).
Fixes: cc8b0b92a1 ("bpf: introduce function calls (function boundaries)")
Reported-by: syzbot+4fc427c7af994b0948be@syzkaller.appspotmail.com
Signed-off-by: Edward Cree <ecree@solarflare.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 28c2fae726 ]
While dbecd73884 ("bpf: get kernel symbol addresses via syscall")
zeroed info.nr_jited_ksyms in bpf_prog_get_info_by_fd() for queries
from unprivileged users, commit 815581c11c ("bpf: get JITed image
lengths of functions via syscall") forgot about doing so and therefore
returns the #elems of the user set up buffer which is incorrect. It
also needs to indicate a info.nr_jited_func_lens of zero.
Fixes: 815581c11c ("bpf: get JITed image lengths of functions via syscall")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Cc: Sandipan Das <sandipan@linux.vnet.ibm.com>
Cc: Song Liu <songliubraving@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
commit 1ae80cf319 upstream.
The map-in-map frequently serves as a mechanism for atomic
snapshotting of state that a BPF program might record. The current
implementation is dangerous to use in this way, however, since
userspace has no way of knowing when all programs that might have
retrieved the "old" value of the map may have completed.
This change ensures that map update operations on map-in-map map types
always wait for all references to the old map to drop before returning
to userspace.
Signed-off-by: Daniel Colascione <dancol@google.com>
Reviewed-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Chenbo Feng <fengc@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit a9c676bc8f ]
Edward Cree says:
In check_mem_access(), for the PTR_TO_CTX case, after check_ctx_access()
has supplied a reg_type, the other members of the register state are set
appropriately. Previously reg.range was set to 0, but as it is in a
union with reg.map_ptr, which is larger, upper bytes of the latter were
left in place. This then caused the memcmp() in regsafe() to fail,
preventing some branches from being pruned (and occasionally causing the
same program to take a varying number of processed insns on repeated
verifier runs).
Fix the instability by clearing bpf_reg_state in __mark_reg_[un]known()
Fixes: f1174f77b5 ("bpf/verifier: rework value tracking")
Debugged-by: Edward Cree <ecree@solarflare.com>
Acked-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 0962590e55 upstream.
ALU operations on pointers such as scalar_reg += map_value_ptr are
handled in adjust_ptr_min_max_vals(). Problem is however that map_ptr
and range in the register state share a union, so transferring state
through dst_reg->range = ptr_reg->range is just buggy as any new
map_ptr in the dst_reg is then truncated (or null) for subsequent
checks. Fix this by adding a raw member and use it for copying state
over to dst_reg.
Fixes: f1174f77b5 ("bpf/verifier: rework value tracking")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Cc: Edward Cree <ecree@solarflare.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
The XSKMAP update and delete functions called synchronize_net(), which
can sleep. It is not allowed to sleep during an RCU read section.
Instead we need to make sure that the sock sk_destruct (xsk_destruct)
function is asynchronously called after an RCU grace period. Setting
the SOCK_RCU_FREE flag for XDP sockets takes care of this.
Fixes: fbfc504a24 ("bpf: introduce new bpf AF_XDP map type BPF_MAP_TYPE_XSKMAP")
Reported-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Björn Töpel <bjorn.topel@intel.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
When I wrote commit 468f6eafa6 ("bpf: fix 32-bit ALU op verification"), I
assumed that, in order to emulate 64-bit arithmetic with 32-bit logic, it
is sufficient to just truncate the output to 32 bits; and so I just moved
the register size coercion that used to be at the start of the function to
the end of the function.
That assumption is true for almost every op, but not for 32-bit right
shifts, because those can propagate information towards the least
significant bit. Fix it by always truncating inputs for 32-bit ops to 32
bits.
Also get rid of the coerce_reg_to_size() after the ALU op, since that has
no effect.
Fixes: 468f6eafa6 ("bpf: fix 32-bit ALU op verification")
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Jann Horn <jannh@google.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Explicitly forbid creating cgroup local storage maps with zero value
size, as it makes no sense and might even cause a panic.
Reported-by: syzbot+18628320d3b14a5c459c@syzkaller.appspotmail.com
Signed-off-by: Roman Gushchin <guro@fb.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
cgroup_storage_update_elem() shouldn't accept any flags
argument values except BPF_ANY and BPF_EXIST to guarantee
the backward compatibility, had a new flag value been added.
Fixes: de9cbbaadb ("bpf: introduce cgroup storage maps")
Signed-off-by: Roman Gushchin <guro@fb.com>
Reported-by: Daniel Borkmann <daniel@iogearbox.net>
Cc: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
It is possible (via shutdown()) for TCP socks to go trough TCP_CLOSE
state via tcp_disconnect() without actually calling tcp_close which
would then call our bpf_tcp_close() callback. Because of this a user
could disconnect a socket then put it in a LISTEN state which would
break our assumptions about sockets always being ESTABLISHED state.
To resolve this rely on the unhash hook, which is called in the
disconnect case, to remove the sock from the sockmap.
Reported-by: Eric Dumazet <edumazet@google.com>
Fixes: 1aa12bdf1b ("bpf: sockmap, add sock close() hook to remove socks")
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
After this patch we only allow socks that are in ESTABLISHED state or
are being added via a sock_ops event that is transitioning into an
ESTABLISHED state. By allowing sock_ops events we allow users to
manage sockmaps directly from sock ops programs. The two supported
sock_ops ops are BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB and
BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB.
Similar to TLS ULP this ensures sk_user_data is correct.
Reported-by: Eric Dumazet <edumazet@google.com>
Fixes: 1aa12bdf1b ("bpf: sockmap, add sock close() hook to remove socks")
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Subtraction of pointers was accidentally allowed for unpriv programs
by commit 82abbf8d2f. Revert that part of commit.
Fixes: 82abbf8d2f ("bpf: do not allow root to mangle valid pointers")
Reported-by: Jann Horn <jannh@google.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
The end boundary math for type section is incorrect in
btf_check_all_metas(). It just happens that hdr->type_off
is always 0 for now because there are only two sections
(type and string) and string section must be at the end (ensured
in btf_parse_str_sec).
However, type_off may not be 0 if a new section would be added later.
This patch fixes it.
Fixes: f80442a4cd ("bpf: btf: Change how section is supported in btf_header")
Reported-by: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Acked-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Currently we check sk_user_data is non NULL to determine if the sk
exists in a map. However, this is not sufficient to ensure the psock
or the ULP ops are not in use by another user, such as kcm or TLS. To
avoid this when adding a sock to a map also verify it is of the
correct ULP type. Additionally, when releasing a psock verify that
it is the TCP_ULP_BPF type before releasing the ULP. The error case
where we abort an update due to ULP collision can cause this error
path.
For example,
__sock_map_ctx_update_elem()
[...]
err = tcp_set_ulp_id(sock, TCP_ULP_BPF) <- collides with TLS
if (err) <- so err out here
goto out_free
[...]
out_free:
smap_release_sock() <- calling tcp_cleanup_ulp releases the
TLS ULP incorrectly.
Fixes: 2f857d0460 ("bpf: sockmap, remove STRPARSER map_flags and add multi-map support")
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Currently, when a redirect occurs in sockmap and an error occurs in
the redirect call we unwind the scatterlist once in the error path
of bpf_tcp_sendmsg_do_redirect() and then again in sendmsg(). Then
in the error path of sendmsg we decrement the copied count by the
send size.
However, its possible we partially sent data before the error was
generated. This can happen if do_tcp_sendpages() partially sends the
scatterlist before encountering a memory pressure error. If this
happens we need to decrement the copied value (the value tracking
how many bytes were actually sent to TCP stack) by the number of
remaining bytes _not_ the entire send size. Otherwise we risk
confusing userspace.
Also we don't need two calls to free the scatterlist one is
good enough. So remove the one in bpf_tcp_sendmsg_do_redirect() and
then properly reduce copied by the number of remaining bytes which
may in fact be the entire send size if no bytes were sent.
To do this use bool to indicate if free_start_sg() should do mem
accounting or not.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
In bpf_tcp_recvmsg() we first took a reference on the psock, however
once we find that there are skbs in the normal socket's receive queue
we return with processing them through tcp_recvmsg(). Problem is that
we leak the taken reference on the psock in that path. Given we don't
really do anything with the psock at this point, move the skb_queue_empty()
test before we fetch the psock to fix this case.
Fixes: 8934ce2fd0 ("bpf: sockmap redirect ingress support")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
bpf_tcp_close() we pop the psock linkage to a map via psock_map_pop().
A parallel update on the sock hash map can happen between psock_map_pop()
and lookup_elem_raw() where we override the element under link->hash /
link->key. In bpf_tcp_close()'s lookup_elem_raw() we subsequently only
test whether an element is present, but we do not test whether the
element is infact the element we were looking for.
We lock the sock in bpf_tcp_close() during that time, so do we hold
the lock in sock_hash_update_elem(). However, the latter locks the
sock which is newly updated, not the one we're purging from the hash
table. This means that while one CPU is doing the lookup from bpf_tcp_close(),
another CPU is doing the map update in parallel, dropped our sock from
the hlist and released the psock.
Subsequently the first CPU will find the new sock and attempts to drop
and release the old sock yet another time. Fix is that we need to check
the elements for a match after lookup, similar as we do in the sock map.
Note that the hash tab elems are freed via RCU, so access to their
link->hash / link->key is fine since we're under RCU read side there.
Fixes: e9db4ef6bf ("bpf: sockhash fix omitted bucket lock in sock_close")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
All BPF hash and LRU maps currently have a known and global seed
we feed into jhash() which is 0. This is suboptimal, thus fix it
by generating a random seed upon hashtab setup time which we can
later on feed into jhash() on lookup, update and deletions.
Fixes: 0f8e4bd8a1 ("bpf: add hashtable type of eBPF maps")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Song Liu <songliubraving@fb.com>
Reviewed-by: Eduardo Valentin <eduval@amazon.com>
When sockmap code is using the stream parser it also handles the write
space events in order to handle the case where (a) verdict redirects
skb to another socket and (b) the sockmap then sends the skb but due
to memory constraints (or other EAGAIN errors) needs to do a retry.
But the initial code missed a third case where the
skb_send_sock_locked() triggers an sk_wait_event(). A typically case
would be when sndbuf size is exceeded. If this happens because we
do not pass the write_space event to the lower layers we never wake
up the event and it will wait for sndtimeo. Which as noted in ktls
fix may be rather large and look like a hang to the user.
To reproduce the best test is to reduce the sndbuf size and send
1B data chunks to stress the memory handling. To fix this pass the
event from the upper layer to the lower layer.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
When we try to allocate a new sock hash entry and the allocation
fails, then sock hash map fails to reduce the map element counter,
meaning we keep accounting this element although it was never used.
Fix it by dropping the element counter on error.
Fixes: 8111038444 ("bpf: sockmap, add hash map support")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Currently, it is possible to create a sock hash map with key size
of 0 and have the kernel return a fd back to user space. This is
invalid for hash maps (and kernel also hasn't been tested for zero
key size support in general at this point). Thus, reject such
configuration.
Fixes: 8111038444 ("bpf: sockmap, add hash map support")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Song Liu <songliubraving@fb.com>
Commits 109980b894 ("bpf: don't select potentially stale ri->map
from buggy xdp progs") and 7c30013133 ("bpf: fix ri->map_owner
pointer on bpf_prog_realloc") tried to mitigate that buggy programs
using bpf_redirect_map() helper call do not leave stale maps behind.
Idea was to add a map_owner cookie into the per CPU struct redirect_info
which was set to prog->aux by the prog making the helper call as a
proof that the map is not stale since the prog is implicitly holding
a reference to it. This owner cookie could later on get compared with
the program calling into BPF whether they match and therefore the
redirect could proceed with processing the map safely.
In (obvious) hindsight, this approach breaks down when tail calls are
involved since the original caller's prog->aux pointer does not have
to match the one from one of the progs out of the tail call chain,
and therefore the xdp buffer will be dropped instead of redirected.
A way around that would be to fix the issue differently (which also
allows to remove related work in fast path at the same time): once
the life-time of a redirect map has come to its end we use it's map
free callback where we need to wait on synchronize_rcu() for current
outstanding xdp buffers and remove such a map pointer from the
redirect info if found to be present. At that time no program is
using this map anymore so we simply invalidate the map pointers to
NULL iff they previously pointed to that instance while making sure
that the redirect path only reads out the map once.
Fixes: 97f91a7cf0 ("bpf: add bpf_redirect_map helper routine")
Fixes: 109980b894 ("bpf: don't select potentially stale ri->map from buggy xdp progs")
Reported-by: Sebastiano Miano <sebastiano.miano@polito.it>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The current code in sock_map_ctx_update_elem() allows for BPF_EXIST
and BPF_NOEXIST map update flags. While on array-like maps this approach
is rather uncommon, e.g. bpf_fd_array_map_update_elem() and others
enforce map update flags to be BPF_ANY such that xchg() can be used
directly, the current implementation in sock map does not guarantee
that such operation with BPF_EXIST / BPF_NOEXIST is atomic.
The initial test does a READ_ONCE(stab->sock_map[i]) to fetch the
socket from the slot which is then tested for NULL / non-NULL. However
later after __sock_map_ctx_update_elem(), the actual update is done
through osock = xchg(&stab->sock_map[i], sock). Problem is that in
the meantime a different CPU could have updated / deleted a socket
on that specific slot and thus flag contraints won't hold anymore.
I've been thinking whether best would be to just break UAPI and do
an enforcement of BPF_ANY to check if someone actually complains,
however trouble is that already in BPF kselftest we use BPF_NOEXIST
for the map update, and therefore it might have been copied into
applications already. The fix to keep the current behavior intact
would be to add a map lock similar to the sock hash bucket lock only
for covering the whole map.
Fixes: 174a79ff95 ("bpf: sockmap with sk redirect support")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The smap_start_sock() and smap_stop_sock() are each protected under
the sock->sk_callback_lock from their call-sites except in the case
of sock_map_delete_elem() where we drop the old socket from the map
slot. This is racy because the same sock could be part of multiple
sock maps, so we run smap_stop_sock() in parallel, and given at that
point psock->strp_enabled might be true on both CPUs, we might for
example wrongly restore the sk->sk_data_ready / sk->sk_write_space.
Therefore, hold the sock->sk_callback_lock as well on delete. Looks
like 2f857d0460 ("bpf: sockmap, remove STRPARSER map_flags and add
multi-map support") had this right, but later on e9db4ef6bf ("bpf:
sockhash fix omitted bucket lock in sock_close") removed it again
from delete leaving this smap_stop_sock() instance unprotected.
Fixes: e9db4ef6bf ("bpf: sockhash fix omitted bucket lock in sock_close")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
While working on sockmap I noticed that we do not always kfree the
struct smap_psock_map_entry list elements which track psocks attached
to maps. In the case of sock_hash_ctx_update_elem(), these map entries
are allocated outside of __sock_map_ctx_update_elem() with their
linkage to the socket hash table filled. In the case of sock array,
the map entries are allocated inside of __sock_map_ctx_update_elem()
and added with their linkage to the psock->maps. Both additions are
under psock->maps_lock each.
Now, we drop these elements from their psock->maps list in a few
occasions: i) in sock array via smap_list_map_remove() when an entry
is either deleted from the map from user space, or updated via
user space or BPF program where we drop the old socket at that map
slot, or the sock array is freed via sock_map_free() and drops all
its elements; ii) for sock hash via smap_list_hash_remove() in exactly
the same occasions as just described for sock array; iii) in the
bpf_tcp_close() where we remove the elements from the list via
psock_map_pop() and iterate over them dropping themselves from either
sock array or sock hash; and last but not least iv) once again in
smap_gc_work() which is a callback for deferring the work once the
psock refcount hit zero and thus the socket is being destroyed.
Problem is that the only case where we kfree() the list entry is
in case iv), which at that point should have an empty list in
normal cases. So in cases from i) to iii) we unlink the elements
without freeing where they go out of reach from us. Hence fix is
to properly kfree() them as well to stop the leakage. Given these
are all handled under psock->maps_lock there is no need for deferred
RCU freeing.
I later also ran with kmemleak detector and it confirmed the finding
as well where in the state before the fix the object goes unreferenced
while after the patch no kmemleak report related to BPF showed up.
[...]
unreferenced object 0xffff880378eadae0 (size 64):
comm "test_sockmap", pid 2225, jiffies 4294720701 (age 43.504s)
hex dump (first 32 bytes):
00 01 00 00 00 00 ad de 00 02 00 00 00 00 ad de ................
50 4d 75 5d 03 88 ff ff 00 00 00 00 00 00 00 00 PMu]............
backtrace:
[<000000005225ac3c>] sock_map_ctx_update_elem.isra.21+0xd8/0x210
[<0000000045dd6d3c>] bpf_sock_map_update+0x29/0x60
[<00000000877723aa>] ___bpf_prog_run+0x1e1f/0x4960
[<000000002ef89e83>] 0xffffffffffffffff
unreferenced object 0xffff880378ead240 (size 64):
comm "test_sockmap", pid 2225, jiffies 4294720701 (age 43.504s)
hex dump (first 32 bytes):
00 01 00 00 00 00 ad de 00 02 00 00 00 00 ad de ................
00 44 75 5d 03 88 ff ff 00 00 00 00 00 00 00 00 .Du]............
backtrace:
[<000000005225ac3c>] sock_map_ctx_update_elem.isra.21+0xd8/0x210
[<0000000030e37a3a>] sock_map_update_elem+0x125/0x240
[<000000002e5ce36e>] map_update_elem+0x4eb/0x7b0
[<00000000db453cc9>] __x64_sys_bpf+0x1f9/0x360
[<0000000000763660>] do_syscall_64+0x9a/0x300
[<00000000422a2bb2>] entry_SYSCALL_64_after_hwframe+0x44/0xa9
[<000000002ef89e83>] 0xffffffffffffffff
[...]
Fixes: e9db4ef6bf ("bpf: sockhash fix omitted bucket lock in sock_close")
Fixes: 54fedb42c6 ("bpf: sockmap, fix smap_list_map_remove when psock is in many maps")
Fixes: 2f857d0460 ("bpf: sockmap, remove STRPARSER map_flags and add multi-map support")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
