linux.git/kernel/bpf, branch v5.13.1

bpf: Fix leakage under speculation on mispredicted branches

2021-06-14T21:06:10+00:00

The verifier only enumerates valid control-flow paths and skips paths that
are unreachable in the non-speculative domain. And so it can miss issues
under speculative execution on mispredicted branches.

For example, a type confusion has been demonstrated with the following
crafted program:

  // r0 = pointer to a map array entry
  // r6 = pointer to readable stack slot
  // r9 = scalar controlled by attacker
  1: r0 = *(u64 *)(r0) // cache miss
  2: if r0 != 0x0 goto line 4
  3: r6 = r9
  4: if r0 != 0x1 goto line 6
  5: r9 = *(u8 *)(r6)
  6: // leak r9

Since line 3 runs iff r0 == 0 and line 5 runs iff r0 == 1, the verifier
concludes that the pointer dereference on line 5 is safe. But: if the
attacker trains both the branches to fall-through, such that the following
is speculatively executed ...

  r6 = r9
  r9 = *(u8 *)(r6)
  // leak r9

... then the program will dereference an attacker-controlled value and could
leak its content under speculative execution via side-channel. This requires
to mistrain the branch predictor, which can be rather tricky, because the
branches are mutually exclusive. However such training can be done at
congruent addresses in user space using different branches that are not
mutually exclusive. That is, by training branches in user space ...

  A:  if r0 != 0x0 goto line C
  B:  ...
  C:  if r0 != 0x0 goto line D
  D:  ...

... such that addresses A and C collide to the same CPU branch prediction
entries in the PHT (pattern history table) as those of the BPF program's
lines 2 and 4, respectively. A non-privileged attacker could simply brute
force such collisions in the PHT until observing the attack succeeding.

Alternative methods to mistrain the branch predictor are also possible that
avoid brute forcing the collisions in the PHT. A reliable attack has been
demonstrated, for example, using the following crafted program:

  // r0 = pointer to a [control] map array entry
  // r7 = *(u64 *)(r0 + 0), training/attack phase
  // r8 = *(u64 *)(r0 + 8), oob address
  // [...]
  // r0 = pointer to a [data] map array entry
  1: if r7 == 0x3 goto line 3
  2: r8 = r0
  // crafted sequence of conditional jumps to separate the conditional
  // branch in line 193 from the current execution flow
  3: if r0 != 0x0 goto line 5
  4: if r0 == 0x0 goto exit
  5: if r0 != 0x0 goto line 7
  6: if r0 == 0x0 goto exit
  [...]
  187: if r0 != 0x0 goto line 189
  188: if r0 == 0x0 goto exit
  // load any slowly-loaded value (due to cache miss in phase 3) ...
  189: r3 = *(u64 *)(r0 + 0x1200)
  // ... and turn it into known zero for verifier, while preserving slowly-
  // loaded dependency when executing:
  190: r3 &= 1
  191: r3 &= 2
  // speculatively bypassed phase dependency
  192: r7 += r3
  193: if r7 == 0x3 goto exit
  194: r4 = *(u8 *)(r8 + 0)
  // leak r4

As can be seen, in training phase (phase != 0x3), the condition in line 1
turns into false and therefore r8 with the oob address is overridden with
the valid map value address, which in line 194 we can read out without
issues. However, in attack phase, line 2 is skipped, and due to the cache
miss in line 189 where the map value is (zeroed and later) added to the
phase register, the condition in line 193 takes the fall-through path due
to prior branch predictor training, where under speculation, it'll load the
byte at oob address r8 (unknown scalar type at that point) which could then
be leaked via side-channel.

One way to mitigate these is to 'branch off' an unreachable path, meaning,
the current verification path keeps following the is_branch_taken() path
and we push the other branch to the verification stack. Given this is
unreachable from the non-speculative domain, this branch's vstate is
explicitly marked as speculative. This is needed for two reasons: i) if
this path is solely seen from speculative execution, then we later on still
want the dead code elimination to kick in in order to sanitize these
instructions with jmp-1s, and ii) to ensure that paths walked in the
non-speculative domain are not pruned from earlier walks of paths walked in
the speculative domain. Additionally, for robustness, we mark the registers
which have been part of the conditional as unknown in the speculative path
given there should be no assumptions made on their content.

The fix in here mitigates type confusion attacks described earlier due to
i) all code paths in the BPF program being explored and ii) existing
verifier logic already ensuring that given memory access instruction
references one specific data structure.

An alternative to this fix that has also been looked at in this scope was to
mark aux->alu_state at the jump instruction with a BPF_JMP_TAKEN state as
well as direction encoding (always-goto, always-fallthrough, unknown), such
that mixing of different always-* directions themselves as well as mixing of
always-* with unknown directions would cause a program rejection by the
verifier, e.g. programs with constructs like 'if ([...]) { x = 0; } else
{ x = 1; }' with subsequent 'if (x == 1) { [...] }'. For unprivileged, this
would result in only single direction always-* taken paths, and unknown taken
paths being allowed, such that the former could be patched from a conditional
jump to an unconditional jump (ja). Compared to this approach here, it would
have two downsides: i) valid programs that otherwise are not performing any
pointer arithmetic, etc, would potentially be rejected/broken, and ii) we are
required to turn off path pruning for unprivileged, where both can be avoided
in this work through pushing the invalid branch to the verification stack.

The issue was originally discovered by Adam and Ofek, and later independently
discovered and reported as a result of Benedict and Piotr's research work.

Fixes: b2157399cc98 ("bpf: prevent out-of-bounds speculation")
Reported-by: Adam Morrison 
Reported-by: Ofek Kirzner 
Reported-by: Benedict Schlueter 
Reported-by: Piotr Krysiuk 
Signed-off-by: Daniel Borkmann 
Reviewed-by: John Fastabend 
Reviewed-by: Benedict Schlueter 
Reviewed-by: Piotr Krysiuk 
Acked-by: Alexei Starovoitov

bpf: Do not mark insn as seen under speculative path verification

2021-06-14T21:06:06+00:00

... in such circumstances, we do not want to mark the instruction as seen given
the goal is still to jmp-1 rewrite/sanitize dead code, if it is not reachable
from the non-speculative path verification. We do however want to verify it for
safety regardless.

With the patch as-is all the insns that have been marked as seen before the
patch will also be marked as seen after the patch (just with a potentially
different non-zero count). An upcoming patch will also verify paths that are
unreachable in the non-speculative domain, hence this extension is needed.

Signed-off-by: Daniel Borkmann 
Reviewed-by: John Fastabend 
Reviewed-by: Benedict Schlueter 
Reviewed-by: Piotr Krysiuk 
Acked-by: Alexei Starovoitov

bpf: Inherit expanded/patched seen count from old aux data

2021-06-14T21:06:00+00:00

Instead of relying on current env->pass_cnt, use the seen count from the
old aux data in adjust_insn_aux_data(), and expand it to the new range of
patched instructions. This change is valid given we always expand 1:n
with n>=1, so what applies to the old/original instruction needs to apply
for the replacement as well.

Not relying on env->pass_cnt is a prerequisite for a later change where we
want to avoid marking an instruction seen when verified under speculative
execution path.

Signed-off-by: Daniel Borkmann 
Reviewed-by: John Fastabend 
Reviewed-by: Benedict Schlueter 
Reviewed-by: Piotr Krysiuk 
Acked-by: Alexei Starovoitov

bpf, lockdown, audit: Fix buggy SELinux lockdown permission checks

2021-06-02T19:59:22+00:00

Commit 59438b46471a ("security,lockdown,selinux: implement SELinux lockdown")
added an implementation of the locked_down LSM hook to SELinux, with the aim
to restrict which domains are allowed to perform operations that would breach
lockdown. This is indirectly also getting audit subsystem involved to report
events. The latter is problematic, as reported by Ondrej and Serhei, since it
can bring down the whole system via audit:

  1) The audit events that are triggered due to calls to security_locked_down()
     can OOM kill a machine, see below details [0].

  2) It also seems to be causing a deadlock via avc_has_perm()/slow_avc_audit()
     when trying to wake up kauditd, for example, when using trace_sched_switch()
     tracepoint, see details in [1]. Triggering this was not via some hypothetical
     corner case, but with existing tools like runqlat & runqslower from bcc, for
     example, which make use of this tracepoint. Rough call sequence goes like:

     rq_lock(rq) -> -------------------------+
       trace_sched_switch() ->               |
         bpf_prog_xyz() ->                   +-> deadlock
           selinux_lockdown() ->             |
             audit_log_end() ->              |
               wake_up_interruptible() ->    |
                 try_to_wake_up() ->         |
                   rq_lock(rq) --------------+

What's worse is that the intention of 59438b46471a to further restrict lockdown
settings for specific applications in respect to the global lockdown policy is
completely broken for BPF. The SELinux policy rule for the current lockdown check
looks something like this:

  allow   : lockdown {  };

However, this doesn't match with the 'current' task where the security_locked_down()
is executed, example: httpd does a syscall. There is a tracing program attached
to the syscall which triggers a BPF program to run, which ends up doing a
bpf_probe_read_kernel{,_str}() helper call. The selinux_lockdown() hook does
the permission check against 'current', that is, httpd in this example. httpd
has literally zero relation to this tracing program, and it would be nonsensical
having to write an SELinux policy rule against httpd to let the tracing helper
pass. The policy in this case needs to be against the entity that is installing
the BPF program. For example, if bpftrace would generate a histogram of syscall
counts by user space application:

  bpftrace -e 'tracepoint:raw_syscalls:sys_enter { @[comm] = count(); }'

bpftrace would then go and generate a BPF program from this internally. One way
of doing it [for the sake of the example] could be to call bpf_get_current_task()
helper and then access current->comm via one of bpf_probe_read_kernel{,_str}()
helpers. So the program itself has nothing to do with httpd or any other random
app doing a syscall here. The BPF program _explicitly initiated_ the lockdown
check. The allow/deny policy belongs in the context of bpftrace: meaning, you
want to grant bpftrace access to use these helpers, but other tracers on the
system like my_random_tracer _not_.

Therefore fix all three issues at the same time by taking a completely different
approach for the security_locked_down() hook, that is, move the check into the
program verification phase where we actually retrieve the BPF func proto. This
also reliably gets the task (current) that is trying to install the BPF tracing
program, e.g. bpftrace/bcc/perf/systemtap/etc, and it also fixes the OOM since
we're moving this out of the BPF helper's fast-path which can be called several
millions of times per second.

The check is then also in line with other security_locked_down() hooks in the
system where the enforcement is performed at open/load time, for example,
open_kcore() for /proc/kcore access or module_sig_check() for module signatures
just to pick few random ones. What's out of scope in the fix as well as in
other security_locked_down() hook locations /outside/ of BPF subsystem is that
if the lockdown policy changes on the fly there is no retrospective action.
This requires a different discussion, potentially complex infrastructure, and
it's also not clear whether this can be solved generically. Either way, it is
out of scope for a suitable stable fix which this one is targeting. Note that
the breakage is specifically on 59438b46471a where it started to rely on 'current'
as UAPI behavior, and _not_ earlier infrastructure such as 9d1f8be5cf42 ("bpf:
Restrict bpf when kernel lockdown is in confidentiality mode").

[0] https://bugzilla.redhat.com/show_bug.cgi?id=1955585, Jakub Hrozek says:

  I starting seeing this with F-34. When I run a container that is traced with
  BPF to record the syscalls it is doing, auditd is flooded with messages like:

  type=AVC msg=audit(1619784520.593:282387): avc:  denied  { confidentiality }
    for pid=476 comm="auditd" lockdown_reason="use of bpf to read kernel RAM"
      scontext=system_u:system_r:auditd_t:s0 tcontext=system_u:system_r:auditd_t:s0
        tclass=lockdown permissive=0

  This seems to be leading to auditd running out of space in the backlog buffer
  and eventually OOMs the machine.

  [...]
  auditd running at 99% CPU presumably processing all the messages, eventually I get:
  Apr 30 12:20:42 fedora kernel: audit: backlog limit exceeded
  Apr 30 12:20:42 fedora kernel: audit: backlog limit exceeded
  Apr 30 12:20:42 fedora kernel: audit: audit_backlog=2152579 > audit_backlog_limit=64
  Apr 30 12:20:42 fedora kernel: audit: audit_backlog=2152626 > audit_backlog_limit=64
  Apr 30 12:20:42 fedora kernel: audit: audit_backlog=2152694 > audit_backlog_limit=64
  Apr 30 12:20:42 fedora kernel: audit: audit_lost=6878426 audit_rate_limit=0 audit_backlog_limit=64
  Apr 30 12:20:45 fedora kernel: oci-seccomp-bpf invoked oom-killer: gfp_mask=0x100cca(GFP_HIGHUSER_MOVABLE), order=0, oom_score_adj=-1000
  Apr 30 12:20:45 fedora kernel: CPU: 0 PID: 13284 Comm: oci-seccomp-bpf Not tainted 5.11.12-300.fc34.x86_64 #1
  Apr 30 12:20:45 fedora kernel: Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.13.0-2.fc32 04/01/2014
  [...]

[1] https://lore.kernel.org/linux-audit/CANYvDQN7H5tVp47fbYcRasv4XF07eUbsDwT_eDCHXJUj43J7jQ@mail.gmail.com/,
    Serhei Makarov says:

  Upstream kernel 5.11.0-rc7 and later was found to deadlock during a
  bpf_probe_read_compat() call within a sched_switch tracepoint. The problem
  is reproducible with the reg_alloc3 testcase from SystemTap's BPF backend
  testsuite on x86_64 as well as the runqlat, runqslower tools from bcc on
  ppc64le. Example stack trace:

  [...]
  [  730.868702] stack backtrace:
  [  730.869590] CPU: 1 PID: 701 Comm: in:imjournal Not tainted, 5.12.0-0.rc2.20210309git144c79ef3353.166.fc35.x86_64 #1
  [  730.871605] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.13.0-2.fc32 04/01/2014
  [  730.873278] Call Trace:
  [  730.873770]  dump_stack+0x7f/0xa1
  [  730.874433]  check_noncircular+0xdf/0x100
  [  730.875232]  __lock_acquire+0x1202/0x1e10
  [  730.876031]  ? __lock_acquire+0xfc0/0x1e10
  [  730.876844]  lock_acquire+0xc2/0x3a0
  [  730.877551]  ? __wake_up_common_lock+0x52/0x90
  [  730.878434]  ? lock_acquire+0xc2/0x3a0
  [  730.879186]  ? lock_is_held_type+0xa7/0x120
  [  730.880044]  ? skb_queue_tail+0x1b/0x50
  [  730.880800]  _raw_spin_lock_irqsave+0x4d/0x90
  [  730.881656]  ? __wake_up_common_lock+0x52/0x90
  [  730.882532]  __wake_up_common_lock+0x52/0x90
  [  730.883375]  audit_log_end+0x5b/0x100
  [  730.884104]  slow_avc_audit+0x69/0x90
  [  730.884836]  avc_has_perm+0x8b/0xb0
  [  730.885532]  selinux_lockdown+0xa5/0xd0
  [  730.886297]  security_locked_down+0x20/0x40
  [  730.887133]  bpf_probe_read_compat+0x66/0xd0
  [  730.887983]  bpf_prog_250599c5469ac7b5+0x10f/0x820
  [  730.888917]  trace_call_bpf+0xe9/0x240
  [  730.889672]  perf_trace_run_bpf_submit+0x4d/0xc0
  [  730.890579]  perf_trace_sched_switch+0x142/0x180
  [  730.891485]  ? __schedule+0x6d8/0xb20
  [  730.892209]  __schedule+0x6d8/0xb20
  [  730.892899]  schedule+0x5b/0xc0
  [  730.893522]  exit_to_user_mode_prepare+0x11d/0x240
  [  730.894457]  syscall_exit_to_user_mode+0x27/0x70
  [  730.895361]  entry_SYSCALL_64_after_hwframe+0x44/0xae
  [...]

Fixes: 59438b46471a ("security,lockdown,selinux: implement SELinux lockdown")
Reported-by: Ondrej Mosnacek 
Reported-by: Jakub Hrozek 
Reported-by: Serhei Makarov 
Reported-by: Jiri Olsa 
Signed-off-by: Daniel Borkmann 
Acked-by: Alexei Starovoitov 
Tested-by: Jiri Olsa 
Cc: Paul Moore 
Cc: James Morris 
Cc: Jerome Marchand 
Cc: Frank Eigler 
Cc: Linus Torvalds 
Link: https://lore.kernel.org/bpf/01135120-8bf7-df2e-cff0-1d73f1f841c3@iogearbox.net

bpf: No need to simulate speculative domain for immediates

2021-05-25T20:08:53+00:00

In 801c6058d14a ("bpf: Fix leakage of uninitialized bpf stack under
speculation") we replaced masking logic with direct loads of immediates
if the register is a known constant. Given in this case we do not apply
any masking, there is also no reason for the operation to be truncated
under the speculative domain.

Therefore, there is also zero reason for the verifier to branch-off and
simulate this case, it only needs to do it for unknown but bounded scalars.
As a side-effect, this also enables few test cases that were previously
rejected due to simulation under zero truncation.

Signed-off-by: Daniel Borkmann 
Reviewed-by: Piotr Krysiuk 
Acked-by: Alexei Starovoitov

bpf: Fix mask direction swap upon off reg sign change

2021-05-25T20:08:53+00:00

Masking direction as indicated via mask_to_left is considered to be
calculated once and then used to derive pointer limits. Thus, this
needs to be placed into bpf_sanitize_info instead so we can pass it
to sanitize_ptr_alu() call after the pointer move. Piotr noticed a
corner case where the off reg causes masking direction change which
then results in an incorrect final aux->alu_limit.

Fixes: 7fedb63a8307 ("bpf: Tighten speculative pointer arithmetic mask")
Reported-by: Piotr Krysiuk 
Signed-off-by: Daniel Borkmann 
Reviewed-by: Piotr Krysiuk 
Acked-by: Alexei Starovoitov

bpf: Wrap aux data inside bpf_sanitize_info container

2021-05-25T20:08:53+00:00

Add a container structure struct bpf_sanitize_info which holds
the current aux info, and update call-sites to sanitize_ptr_alu()
to pass it in. This is needed for passing in additional state
later on.

Signed-off-by: Daniel Borkmann 
Reviewed-by: Piotr Krysiuk 
Acked-by: Alexei Starovoitov

bpf: Fix BPF_LSM kconfig symbol dependency

2021-05-25T19:16:23+00:00

Similarly as 6bdacdb48e94 ("bpf: Fix BPF_JIT kconfig symbol dependency") we
need to detangle the hard BPF_LSM dependency on NET. This was previously
implicit by its dependency on BPF_JIT which itself was dependent on NET (but
without any actual/real hard dependency code-wise). Given the latter was
lifted, so should be the former as BPF_LSMs could well exist on net-less
systems. This therefore also fixes a randconfig build error recently reported
by Randy:

  ld: kernel/bpf/bpf_lsm.o: in function `bpf_lsm_func_proto':
  bpf_lsm.c:(.text+0x1a0): undefined reference to `bpf_sk_storage_get_proto'
  ld: bpf_lsm.c:(.text+0x1b8): undefined reference to `bpf_sk_storage_delete_proto'
  [...]

Fixes: b24abcff918a ("bpf, kconfig: Add consolidated menu entry for bpf with core options")
Reported-by: Randy Dunlap 
Signed-off-by: Daniel Borkmann 
Acked-by: Randy Dunlap 
Tested-by: Randy Dunlap

bpf, offload: Reorder offload callback 'prepare' in verifier

2021-05-20T21:51:52+00:00

Commit 4976b718c355 ("bpf: Introduce pseudo_btf_id") switched the
order of resolve_pseudo_ldimm(), in which some pseudo instructions
are rewritten. Thus those rewritten instructions cannot be passed
to driver via 'prepare' offload callback.

Reorder the 'prepare' offload callback to fix it.

Fixes: 4976b718c355 ("bpf: Introduce pseudo_btf_id")
Signed-off-by: Yinjun Zhang 
Signed-off-by: Simon Horman 
Signed-off-by: Daniel Borkmann 
Acked-by: Song Liu 
Link: https://lore.kernel.org/bpf/20210520085834.15023-1-simon.horman@netronome.com

bpf: Avoid using ARRAY_SIZE on an uninitialized pointer

2021-05-20T21:48:38+00:00

The cppcheck static code analysis reported the following error:

    if (WARN_ON_ONCE(nest_level > ARRAY_SIZE(bufs->tmp_bufs))) {
                                             ^
ARRAY_SIZE is a macro that expands to sizeofs, so bufs is not actually
dereferenced at runtime, and the code is actually safe. But to keep
things tidy, this patch removes the need for a call to ARRAY_SIZE by
extracting the size of the array into a macro. Cppcheck should no longer
be confused and the code ends up being a bit cleaner.

Fixes: e2d5b2bb769f ("bpf: Fix nested bpf_bprintf_prepare with more per-cpu buffers")
Reported-by: kernel test robot 
Signed-off-by: Florent Revest 
Signed-off-by: Daniel Borkmann 
Acked-by: Song Liu 
Link: https://lore.kernel.org/bpf/20210517092830.1026418-2-revest@chromium.org