linux/kernel
Daniel Borkmann a08dd0da53 bpf: fix regression on verifier pruning wrt map lookups
Commit 57a09bf0a4 ("bpf: Detect identical PTR_TO_MAP_VALUE_OR_NULL
registers") introduced a regression where existing programs stopped
loading due to reaching the verifier's maximum complexity limit,
whereas prior to this commit they were loading just fine; the affected
program has roughly 2k instructions.

What was found is that state pruning couldn't be performed effectively
anymore due to mismatches of the verifier's register state, in particular
in the id tracking. It doesn't mean that 57a09bf0a4 is incorrect per
se, but rather that verifier needs to perform a lot more work for the
same program with regards to involved map lookups.

Since commit 57a09bf0a4 is only about tracking registers with type
PTR_TO_MAP_VALUE_OR_NULL, the id is only needed to follow registers
until they are promoted through pattern matching with a NULL check to
either PTR_TO_MAP_VALUE or UNKNOWN_VALUE type. After that point, the
id becomes irrelevant for the transitioned types.

For UNKNOWN_VALUE, id is already reset to 0 via mark_reg_unknown_value(),
but not so for PTR_TO_MAP_VALUE where id is becoming stale. It's even
transferred further into other types that don't make use of it. Among
others, one example is where UNKNOWN_VALUE is set on function call
return with RET_INTEGER return type.

states_equal() will then fall through the memcmp() on register state;
note that the second memcmp() uses offsetofend(), so the id is part of
that since d2a4dd37f6 ("bpf: fix state equivalence"). But the bisect
pointed already to 57a09bf0a4, where we really reach beyond complexity
limit. What I found was that states_equal() often failed in this
case due to id mismatches in spilled regs with registers in type
PTR_TO_MAP_VALUE. Unlike non-spilled regs, spilled regs just perform
a memcmp() on their reg state and don't have any other optimizations
in place, therefore also id was relevant in this case for making a
pruning decision.

We can safely reset id to 0 as well when converting to PTR_TO_MAP_VALUE.
For the affected program, it resulted in a ~17 fold reduction of
complexity and let the program load fine again. Selftest suite also
runs fine. The only other place where env->id_gen is used currently is
through direct packet access, but for these cases id is long living, thus
a different scenario.

Also, the current logic in mark_map_regs() is not fully correct when
marking NULL branch with UNKNOWN_VALUE. We need to cache the destination
reg's id in any case. Otherwise, once we marked that reg as UNKNOWN_VALUE,
it's id is reset and any subsequent registers that hold the original id
and are of type PTR_TO_MAP_VALUE_OR_NULL won't be marked UNKNOWN_VALUE
anymore, since mark_map_reg() reuses the uncached regs[regno].id that
was just overridden. Note, we don't need to cache it outside of
mark_map_regs(), since it's called once on this_branch and the other
time on other_branch, which are both two independent verifier states.
A test case for this is added here, too.

Fixes: 57a09bf0a4 ("bpf: Detect identical PTR_TO_MAP_VALUE_OR_NULL registers")
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Thomas Graf <tgraf@suug.ch>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-12-17 10:51:31 -05:00
..
bpf bpf: fix regression on verifier pruning wrt map lookups 2016-12-17 10:51:31 -05:00
configs config: android: enable CONFIG_SECCOMP 2016-10-11 15:06:32 -07:00
debug
events Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net 2016-12-10 16:21:55 -05:00
gcov gcov: add support for gcc version >= 6 2016-07-15 14:54:27 +09:00
irq genirq: Use irq type from irqdata instead of irqdesc 2016-11-08 15:15:19 +01:00
livepatch livepatch/module: make TAINT_LIVEPATCH module-specific 2016-08-26 14:42:08 +02:00
locking lockdep: Fix report formatting 2016-12-06 10:40:08 +01:00
power PM / sleep: fix device reference leak in test_suspend 2016-11-02 05:10:04 +01:00
printk Revert "printk: make reading the kernel log flush pending lines" 2016-11-14 09:31:52 -08:00
rcu This adds a new gcc plugin named "latent_entropy". It is designed to 2016-10-15 10:03:15 -07:00
sched sched/autogroup: Fix 64-bit kernel nice level adjustment 2016-11-24 05:45:02 +01:00
time clocksource: export the clocks_calc_mult_shift to use by timestamp code 2016-12-07 11:13:48 -05:00
trace Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net 2016-11-22 13:27:16 -05:00
.gitignore
acct.c
async.c
audit_fsnotify.c
audit_tree.c audit: cleanup prune_tree_thread 2016-04-04 09:46:47 -04:00
audit_watch.c Merge branch 'stable-4.8' of git://git.infradead.org/users/pcmoore/audit 2016-09-01 15:55:56 -07:00
audit.c audit: remove useless synchronize_net() 2016-12-01 11:29:02 -05:00
audit.h Merge branch 'stable-4.8' of git://git.infradead.org/users/pcmoore/audit 2016-07-29 17:54:17 -07:00
auditfilter.c audit: add fields to exclude filter by reusing user filter 2016-06-27 11:01:00 -04:00
auditsc.c Merge branch 'stable-4.9' of git://git.infradead.org/users/pcmoore/audit 2016-10-04 14:21:41 -07:00
backtracetest.c
bounds.c
capability.c kernel: Add noaudit variant of ns_capable() 2016-06-06 20:16:18 +10:00
cgroup_freezer.c
cgroup_pids.c cgroup: Use lld instead of ld when printing pids controller events_limit 2016-06-21 15:03:36 -04:00
cgroup.c cgroup: add support for eBPF programs 2016-11-25 16:25:52 -05:00
compat.c
configs.c
context_tracking.c
cpu_pm.c
cpu.c cpu/hotplug: Use distinct name for cpu_hotplug.dep_map 2016-10-16 11:09:32 +02:00
cpuset.c Merge branch 'for-4.9' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup 2016-10-14 12:18:50 -07:00
crash_dump.c
cred.c cred: Reject inodes with invalid ids in set_create_file_as() 2016-06-30 18:05:09 -05:00
delayacct.c
dma.c
elfcore.c
exec_domain.c
exit.c sched/autogroup: Do not use autogroup->tg in zombie threads 2016-11-22 12:33:43 +01:00
extable.c
fork.c fork: Add task stack refcounting sanity check and prevent premature task stack freeing 2016-11-01 07:39:17 +01:00
freezer.c freezer, oom: check TIF_MEMDIE on the correct task 2016-07-28 16:07:41 -07:00
futex_compat.c
futex.c futex: Add some more function commentry 2016-09-05 17:20:18 +02:00
groups.c cred: simpler, 1D supplementary groups 2016-10-07 18:46:30 -07:00
hung_task.c hung_task: allow hung_task_panic when hung_task_warnings is 0 2016-10-11 15:06:33 -07:00
irq_work.c
jump_label.c powerpc updates for 4.8 #2 2016-08-05 09:00:54 -04:00
kallsyms.c kallsyms: add support for relative offsets in kallsyms address table 2016-03-15 16:55:16 -07:00
kcmp.c
Kconfig.freezer
Kconfig.hz
Kconfig.locks
Kconfig.preempt
kcov.c kcov: add missing #include <linux/sched.h> 2016-12-07 17:10:00 -08:00
kexec_core.c kexec: add restriction on kexec_load() segment sizes 2016-08-02 19:35:31 -04:00
kexec_file.c kexec: fix double-free when failing to relocate the purgatory 2016-09-01 17:52:01 -07:00
kexec_internal.h
kexec.c kexec: allow architectures to override boot mapping 2016-08-02 19:35:27 -04:00
kmod.c
kprobes.c kprobes: include <asm/sections.h> instead of <asm-generic/sections.h> 2016-10-11 15:06:31 -07:00
ksysfs.c kexec: add a kexec_crash_loaded() function 2016-08-02 19:35:30 -04:00
kthread.c kthread: better support freezable kthread workers 2016-10-11 15:06:33 -07:00
latencytop.c
Makefile userns: Add per user namespace sysctls. 2016-08-08 13:18:58 -05:00
membarrier.c
memremap.c mm: fix cache mode of dax pmd mappings 2016-09-09 17:34:46 -07:00
module_signing.c KEYS: Move the point of trust determination to __key_link() 2016-04-11 22:43:43 +01:00
module-internal.h
module.c Re-enable CONFIG_MODVERSIONS in a slightly weaker form 2016-11-29 16:01:30 -08:00
notifier.c
nsproxy.c
padata.c padata: Convert to hotplug state machine 2016-09-19 21:44:30 +02:00
panic.c x86/panic: replace smp_send_stop() with kdump friendly version in panic path 2016-10-11 15:06:32 -07:00
params.c
pid_namespace.c Merge branch 'nsfs-ioctls' into HEAD 2016-09-22 20:00:36 -05:00
pid.c remove lots of IS_ERR_VALUE abuses 2016-05-27 15:26:11 -07:00
profile.c profile: Convert to hotplug state machine 2016-07-15 10:41:42 +02:00
ptrace.c mm: replace access_process_vm() write parameter with gup_flags 2016-10-19 08:31:25 -07:00
range.c
reboot.c
relay.c relay: Use irq_work instead of plain timer for deferred wakeup 2016-10-11 15:06:32 -07:00
resource.c /proc/iomem: only expose physical resource addresses to privileged users 2016-04-14 12:56:09 -07:00
seccomp.c bpf: drop unnecessary context cast from BPF_PROG_RUN 2016-11-27 20:38:47 -05:00
signal.c x86/signal: Add SA_{X32,IA32}_ABI sa_flags 2016-09-14 21:28:11 +02:00
smp.c smp: Allocate smp_call_on_cpu() workqueue on stack too 2016-09-22 14:49:10 +02:00
smpboot.c kthread/smpboot: do not park in kthread_create_on_cpu() 2016-10-11 15:06:33 -07:00
smpboot.h
softirq.c softirq: Display IRQ_POLL for irq-poll statistics 2016-10-21 15:45:47 -06:00
stacktrace.c
stop_machine.c Merge branch 'sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip 2016-10-03 13:39:00 -07:00
sys_ni.c x86/pkeys: Fix pkeys build breakage for some non-x86 arches 2016-09-13 14:41:36 +02:00
sys.c prctl: make PR_SET_THP_DISABLE wait for mmap_sem killable 2016-05-23 17:04:14 -07:00
sysctl_binary.c kernel/sysctl_binary.c: use generic UUID library 2016-05-20 17:58:30 -07:00
sysctl.c Merge branch 'work.misc' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs 2016-10-10 13:04:49 -07:00
task_work.c task_work: use READ_ONCE/lockless_dereference, avoid pi_lock if !task_works 2016-08-02 19:35:02 -04:00
taskstats.c Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net 2016-11-15 10:54:36 -05:00
test_kprobes.c
torture.c torture: Convert torture_shutdown() to hrtimer 2016-08-22 10:01:49 -07:00
tracepoint.c kernel/...: convert pr_warning to pr_warn 2016-03-22 15:36:02 -07:00
tsacct.c
ucount.c mntns: Add a limit on the number of mount namespaces. 2016-08-31 07:28:35 -05:00
uid16.c cred: simpler, 1D supplementary groups 2016-10-07 18:46:30 -07:00
up.c smp: Add function to execute a function synchronously on a CPU 2016-09-05 13:52:39 +02:00
user_namespace.c Merge branch 'nsfs-ioctls' into HEAD 2016-09-22 20:00:36 -05:00
user-return-notifier.c
user.c
utsname_sysctl.c
utsname.c Merge branch 'nsfs-ioctls' into HEAD 2016-09-22 20:00:36 -05:00
watchdog.c Revert "perf/x86/intel, watchdog: Switch NMI watchdog to ref cycles on x86" 2016-07-10 20:58:36 +02:00
workqueue_internal.h
workqueue.c kthread: rename probe_kthread_data() to kthread_probe_data() 2016-10-11 15:06:33 -07:00