bpf: use 32bit safe version of u64_stats
Subsystems (4)
v2 cover bpf: use 32bit safe version of u64_stats
From: Eric Dumazet <[email protected]>
Two first patches fix bugs added in 5.1 and 5.5
Third patch replaces the u64 fields in struct bpf_prog_stats with u64_stats_t ones to avoid possible sampling errors, in case of load/store stearing.
Eric Dumazet (3):
bpf: avoid races in __bpf_prog_run() for 32bit arches
bpf: fixes possible race in update_prog_stats() for 32bit arches
bpf: use u64_stats_t in struct bpf_prog_stats
include/linux/filter.h | 15 ++++++++------- kernel/bpf/syscall.c | 18 ++++++++++++------ kernel/bpf/trampoline.c | 12 +++++++----- 3 files changed, 27 insertions(+), 18 deletions(-)
Signature · 1 line
2.33.0.1079.g6e70778dc9-goog
PATCH v2 1/3 bpf: avoid races in __bpf_prog_run() for 32bit arches
From: Eric Dumazet <[email protected]>
__bpf_prog_run() can run from non IRQ contexts, meaning it could be re entered if interrupted.
This calls for the irq safe variant of u64_stats_update_{begin|end}, or risk a deadlock.
This patch is a nop on 64bit arches, fortunately.
syzbot report:
WARNING: inconsistent lock state
5.12.0-rc3-syzkaller #0 Not tainted
--------------------------------
inconsistent {IN-SOFTIRQ-W} -> {SOFTIRQ-ON-W} usage.
udevd/4013 [HC0[0]:SC0[0]:HE1:SE1] takes:ff7c9dec (&(&pstats->syncp)->seq){+.?.}-{0:0}, at: sk_filter include/linux/filter.h:867 [inline] ff7c9dec (&(&pstats->syncp)->seq){+.?.}-{0:0}, at: do_one_broadcast net/netlink/af_netlink.c:1468 [inline] ff7c9dec (&(&pstats->syncp)->seq){+.?.}-{0:0}, at: netlink_broadcast_filtered+0x27c/0x4fc net/netlink/af_netlink.c:1520 {IN-SOFTIRQ-W} state was registered at:
lock_acquire.part.0+0xf0/0x41c kernel/locking/lockdep.c:5510 lock_acquire+0x6c/0x74 kernel/locking/lockdep.c:5483 do_write_seqcount_begin_nested include/linux/seqlock.h:520 [inline] do_write_seqcount_begin include/linux/seqlock.h:545 [inline] u64_stats_update_begin include/linux/u64_stats_sync.h:129 [inline] bpf_prog_run_pin_on_cpu include/linux/filter.h:624 [inline] bpf_prog_run_clear_cb+0x1bc/0x270 include/linux/filter.h:755 run_filter+0xa0/0x17c net/packet/af_packet.c:2031 packet_rcv+0xc0/0x3e0 net/packet/af_packet.c:2104 dev_queue_xmit_nit+0x2bc/0x39c net/core/dev.c:2387 xmit_one net/core/dev.c:3588 [inline] dev_hard_start_xmit+0x94/0x518 net/core/dev.c:3609 sch_direct_xmit+0x11c/0x1f0 net/sched/sch_generic.c:313 qdisc_restart net/sched/sch_generic.c:376 [inline] __qdisc_run+0x194/0x7f8 net/sched/sch_generic.c:384 qdisc_run include/net/pkt_sched.h:136 [inline] qdisc_run include/net/pkt_sched.h:128 [inline] __dev_xmit_skb net/core/dev.c:3795 [inline] __dev_queue_xmit+0x65c/0xf84 net/core/dev.c:4150 dev_queue_xmit+0x14/0x18 net/core/dev.c:4215 neigh_resolve_output net/core/neighbour.c:1491 [inline] neigh_resolve_output+0x170/0x228 net/core/neighbour.c:1471 neigh_output include/net/neighbour.h:510 [inline] ip6_finish_output2+0x2e4/0x9fc net/ipv6/ip6_output.c:117 __ip6_finish_output net/ipv6/ip6_output.c:182 [inline] __ip6_finish_output+0x164/0x3f8 net/ipv6/ip6_output.c:161 ip6_finish_output+0x2c/0xb0 net/ipv6/ip6_output.c:192 NF_HOOK_COND include/linux/netfilter.h:290 [inline] ip6_output+0x74/0x294 net/ipv6/ip6_output.c:215 dst_output include/net/dst.h:448 [inline] NF_HOOK include/linux/netfilter.h:301 [inline] NF_HOOK include/linux/netfilter.h:295 [inline] mld_sendpack+0x2a8/0x7e4 net/ipv6/mcast.c:1679 mld_send_cr net/ipv6/mcast.c:1975 [inline] mld_ifc_timer_expire+0x1e8/0x494 net/ipv6/mcast.c:2474 call_timer_fn+0xd0/0x570 kernel/time/timer.c:1431 expire_timers kernel/time/timer.c:1476 [inline] __run_timers kernel/time/timer.c:1745 [inline] run_timer_softirq+0x2e4/0x384 kernel/time/timer.c:1758 __do_softirq+0x204/0x7ac kernel/softirq.c:345 do_softirq_own_stack include/asm-generic/softirq_stack.h:10 [inline] invoke_softirq kernel/softirq.c:228 [inline] __irq_exit_rcu+0x1d8/0x200 kernel/softirq.c:422 irq_exit+0x10/0x3c kernel/softirq.c:446 __handle_domain_irq+0xb4/0x120 kernel/irq/irqdesc.c:692 handle_domain_irq include/linux/irqdesc.h:176 [inline] gic_handle_irq+0x84/0xac drivers/irqchip/irq-gic.c:370 __irq_svc+0x5c/0x94 arch/arm/kernel/entry-armv.S:205 debug_smp_processor_id+0x0/0x24 lib/smp_processor_id.c:53 rcu_read_lock_held_common kernel/rcu/update.c:108 [inline] rcu_read_lock_sched_held+0x24/0x7c kernel/rcu/update.c:123 trace_lock_acquire+0x24c/0x278 include/trace/events/lock.h:13 lock_acquire+0x3c/0x74 kernel/locking/lockdep.c:5481 rcu_lock_acquire include/linux/rcupdate.h:267 [inline] rcu_read_lock include/linux/rcupdate.h:656 [inline] avc_has_perm_noaudit+0x6c/0x260 security/selinux/avc.c:1150 selinux_inode_permission+0x140/0x220 security/selinux/hooks.c:3141 security_inode_permission+0x44/0x60 security/security.c:1268 inode_permission.part.0+0x5c/0x13c fs/namei.c:521 inode_permission fs/namei.c:494 [inline] may_lookup fs/namei.c:1652 [inline] link_path_walk.part.0+0xd4/0x38c fs/namei.c:2208 link_path_walk fs/namei.c:2189 [inline] path_lookupat+0x3c/0x1b8 fs/namei.c:2419 filename_lookup+0xa8/0x1a4 fs/namei.c:2453 user_path_at_empty+0x74/0x90 fs/namei.c:2733 do_readlinkat+0x5c/0x12c fs/stat.c:417 __do_sys_readlink fs/stat.c:450 [inline] sys_readlink+0x24/0x28 fs/stat.c:447 ret_fast_syscall+0x0/0x2c arch/arm/mm/proc-v7.S:64 0x7eaa4974 irq event stamp: 298277 hardirqs last enabled at (298277): [<802000d0>] no_work_pending+0x4/0x34 hardirqs last disabled at (298276): [<8020c9b8>] do_work_pending+0x9c/0x648 arch/arm/kernel/signal.c:676 softirqs last enabled at (298216): [<8020167c>] __do_softirq+0x584/0x7ac kernel/softirq.c:372 softirqs last disabled at (298201): [<8024dff4>] do_softirq_own_stack include/asm-generic/softirq_stack.h:10 [inline] softirqs last disabled at (298201): [<8024dff4>] invoke_softirq kernel/softirq.c:228 [inline] softirqs last disabled at (298201): [<8024dff4>] __irq_exit_rcu+0x1d8/0x200 kernel/softirq.c:422
other info that might help us debug this:
Possible unsafe locking scenario:
CPU0
----
lock(&(&pstats->syncp)->seq);
<Interrupt>
lock(&(&pstats->syncp)->seq);*** DEADLOCK ***
1 lock held by udevd/4013:
#0: 82b09c5c (rcu_read_lock){....}-{1:2}, at: sk_filter_trim_cap+0x54/0x434 net/core/filter.c:139
stack backtrace: CPU: 1 PID: 4013 Comm: udevd Not tainted 5.12.0-rc3-syzkaller #0 Hardware name: ARM-Versatile Express Backtrace: [<81802550>] (dump_backtrace) from [<818027c4>] (show_stack+0x18/0x1c arch/arm/kernel/traps.c:252) r7:00000080 r6:600d0093 r5:00000000 r4:82b58344
[<818027ac>] (show_stack) from [<81809e98>] (__dump_stack lib/dump_stack.c:79 [inline])
[<818027ac>] (show_stack) from [<81809e98>] (dump_stack+0xb8/0xe8 lib/dump_stack.c:120)
[<81809de0>] (dump_stack) from [<81804a00>] (print_usage_bug.part.0+0x228/0x230 kernel/locking/lockdep.c:3806)
r7:86bcb768 r6:81a0326c r5:830f96a8 r4:86bcb0c0
[<818047d8>] (print_usage_bug.part.0) from [<802bb1b8>] (print_usage_bug kernel/locking/lockdep.c:3776 [inline]) [<818047d8>] (print_usage_bug.part.0) from [<802bb1b8>] (valid_state kernel/locking/lockdep.c:3818 [inline]) [<818047d8>] (print_usage_bug.part.0) from [<802bb1b8>] (mark_lock_irq kernel/locking/lockdep.c:4021 [inline]) [<818047d8>] (print_usage_bug.part.0) from [<802bb1b8>] (mark_lock.part.0+0xc34/0x136c kernel/locking/lockdep.c:4478)
r10:83278fe8 r9:82c6d748 r8:00000000 r7:82c6d2d4 r6:00000004 r5:86bcb768
r4:00000006
[<802ba584>] (mark_lock.part.0) from [<802bc644>] (mark_lock kernel/locking/lockdep.c:4442 [inline]) [<802ba584>] (mark_lock.part.0) from [<802bc644>] (mark_usage kernel/locking/lockdep.c:4391 [inline]) [<802ba584>] (mark_lock.part.0) from [<802bc644>] (__lock_acquire+0x9bc/0x3318 kernel/locking/lockdep.c:4854)
r10:86bcb768 r9:86bcb0c0 r8:00000001 r7:00040000 r6:0000075a r5:830f96a8
r4:00000000
[<802bbc88>] (__lock_acquire) from [<802bfb90>] (lock_acquire.part.0+0xf0/0x41c kernel/locking/lockdep.c:5510)
r10:00000000 r9:600d0013 r8:00000000 r7:00000000 r6:828a2680 r5:828a2680
r4:861e5bc8
[<802bfaa0>] (lock_acquire.part.0) from [<802bff28>] (lock_acquire+0x6c/0x74 kernel/locking/lockdep.c:5483)
r10:8146137c r9:00000000 r8:00000001 r7:00000000 r6:00000000 r5:00000000
r4:ff7c9dec
[<802bfebc>] (lock_acquire) from [<81381eb4>] (do_write_seqcount_begin_nested include/linux/seqlock.h:520 [inline]) [<802bfebc>] (lock_acquire) from [<81381eb4>] (do_write_seqcount_begin include/linux/seqlock.h:545 [inline]) [<802bfebc>] (lock_acquire) from [<81381eb4>] (u64_stats_update_begin include/linux/u64_stats_sync.h:129 [inline]) [<802bfebc>] (lock_acquire) from [<81381eb4>] (__bpf_prog_run_save_cb include/linux/filter.h:727 [inline]) [<802bfebc>] (lock_acquire) from [<81381eb4>] (bpf_prog_run_save_cb include/linux/filter.h:741 [inline]) [<802bfebc>] (lock_acquire) from [<81381eb4>] (sk_filter_trim_cap+0x26c/0x434 net/core/filter.c:149)
r10:a4095dd0 r9:ff7c9dd0 r8:e44be000 r7:8146137c r6:00000001 r5:8611ba80
r4:00000000
[<81381c48>] (sk_filter_trim_cap) from [<8146137c>] (sk_filter include/linux/filter.h:867 [inline]) [<81381c48>] (sk_filter_trim_cap) from [<8146137c>] (do_one_broadcast net/netlink/af_netlink.c:1468 [inline]) [<81381c48>] (sk_filter_trim_cap) from [<8146137c>] (netlink_broadcast_filtered+0x27c/0x4fc net/netlink/af_netlink.c:1520)
r10:00000001 r9:833d6b1c r8:00000000 r7:8572f864 r6:8611ba80 r5:8698d800
r4:8572f800
[<81461100>] (netlink_broadcast_filtered) from [<81463e60>] (netlink_broadcast net/netlink/af_netlink.c:1544 [inline]) [<81461100>] (netlink_broadcast_filtered) from [<81463e60>] (netlink_sendmsg+0x3d0/0x478 net/netlink/af_netlink.c:1925)
r10:00000000 r9:00000002 r8:8698d800 r7:000000b7 r6:8611b900 r5:861e5f50
r4:86aa3000
[<81463a90>] (netlink_sendmsg) from [<81321f54>] (sock_sendmsg_nosec net/socket.c:654 [inline]) [<81463a90>] (netlink_sendmsg) from [<81321f54>] (sock_sendmsg+0x3c/0x4c net/socket.c:674)
r10:00000000 r9:861e5dd4 r8:00000000 r7:86570000 r6:00000000 r5:86570000
r4:861e5f50
[<81321f18>] (sock_sendmsg) from [<813234d0>] (____sys_sendmsg+0x230/0x29c net/socket.c:2350)
r5:00000040 r4:861e5f50
[<813232a0>] (____sys_sendmsg) from [<8132549c>] (___sys_sendmsg+0xac/0xe4 net/socket.c:2404)
r10:00000128 r9:861e4000 r8:00000000 r7:00000000 r6:86570000 r5:861e5f50
r4:00000000
[<813253f0>] (___sys_sendmsg) from [<81325684>] (__sys_sendmsg net/socket.c:2433 [inline]) [<813253f0>] (___sys_sendmsg) from [<81325684>] (__do_sys_sendmsg net/socket.c:2442 [inline]) [<813253f0>] (___sys_sendmsg) from [<81325684>] (sys_sendmsg+0x58/0xa0 net/socket.c:2440)
r8:80200224 r7:00000128 r6:00000000 r5:7eaa541c r4:86570000
[<8132562c>] (sys_sendmsg) from [<80200060>] (ret_fast_syscall+0x0/0x2c arch/arm/mm/proc-v7.S:64)
Exception stack(0x861e5fa8 to 0x861e5ff0)
5fa0: 00000000 00000000 0000000c 7eaa541c 00000000 00000000
5fc0: 00000000 00000000 76fbf840 00000128 00000000 0000008f 7eaa541c 000563f8 5fe0: 00056110 7eaa53e0 00036cec 76c9bf44
r6:76fbf840 r5:00000000 r4:00000000
--- include/linux/filter.h | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)
include/linux/filter.h +3 -2
@@ -612,13 +612,14 @@ static __always_inline u32 __bpf_prog_run(const struct bpf_prog *prog,if (static_branch_unlikely (&bpf_stats_enabled_key)) {struct bpf_prog_stats *stats;u64 start =sched_clock (); +unsigned long flags; ret =dfunc (ctx, prog->insnsi, prog->bpf_func); stats =this_cpu_ptr (prog->stats); -u64_stats_update_begin (&stats->syncp); + flags =u64_stats_update_begin_irqsave (&stats->syncp); stats->cnt++; stats->nsecs +=sched_clock () - start; -u64_stats_update_end (&stats->syncp); +u64_stats_update_end_irqrestore (&stats->syncp, flags); }else { ret =dfunc (ctx, prog->insnsi, prog->bpf_func); }
Signature · 1 line
2.33.0.1079.g6e70778dc9-goog
PATCH v2 2/3 bpf: fixes possible race in update_prog_stats() for 32bit arches
In reply to Eric DumazetFrom: Eric Dumazet <[email protected]>
It seems update_prog_stats() suffers from same issue fixed in the prior patch:
As it can run while interrupts are enabled, it could be re-entered and the u64_stats syncp could be mangled.
--- kernel/bpf/trampoline.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)
kernel/bpf/trampoline.c +4 -2
@@ -586,11 +586,13 @@ static void notrace update_prog_stats(struct bpf_prog *prog,* Hence check that 'start' is valid. */ start > NO_START_TIME) { +unsigned long flags; + stats =this_cpu_ptr (prog->stats); -u64_stats_update_begin (&stats->syncp); + flags =u64_stats_update_begin_irqsave (&stats->syncp); stats->cnt++; stats->nsecs +=sched_clock () - start; -u64_stats_update_end (&stats->syncp); +u64_stats_update_end_irqrestore (&stats->syncp, flags); } }
Signature · 1 line
2.33.0.1079.g6e70778dc9-goog
PATCH v2 3/3 bpf: use u64_stats_t in struct bpf_prog_stats
In reply to Eric DumazetFrom: Eric Dumazet <[email protected]>
Commit 316580b69d0a ("u64_stats: provide u64_stats_t type") fixed possible load/store tearing on 64bit arches.
For instance the following C code
stats->nsecs += sched_clock() - start;
Could be rightfully implemented like this by a compiler, confusing concurrent readers a lot:
stats->nsecs += sched_clock();
// arbitrary delaystats->nsecs -= start;
--- include/linux/filter.h | 10 +++++----- kernel/bpf/syscall.c | 18 ++++++++++++------ kernel/bpf/trampoline.c | 6 +++--- 3 files changed, 20 insertions(+), 14 deletions(-)
include/linux/filter.h +5 -5
@@ -553,9 +553,9 @@ struct bpf_binary_header { };struct bpf_prog_stats { -u64 cnt; -u64 nsecs; -u64 misses; +u64_stats_t cnt; +u64_stats_t nsecs; +u64_stats_t misses;struct u64_stats_sync syncp; }__aligned(2 * sizeof(u64)) ;
@@ -617,8 +617,8 @@ static __always_inline u32 __bpf_prog_run(const struct bpf_prog *prog, ret =dfunc (ctx, prog->insnsi, prog->bpf_func); stats =this_cpu_ptr (prog->stats); flags =u64_stats_update_begin_irqsave (&stats->syncp); - stats->cnt++; - stats->nsecs +=sched_clock () - start; +u64_stats_inc (&stats->cnt); +u64_stats_add (&stats->nsecs,sched_clock () - start);u64_stats_update_end_irqrestore (&stats->syncp, flags); }else { ret =dfunc (ctx, prog->insnsi, prog->bpf_func);
kernel/bpf/syscall.c +12 -6
@@ -1804,8 +1804,14 @@ static int bpf_prog_release(struct inode *inode, struct file *filp)return 0 ; } +struct bpf_prog_kstats { +u64 nsecs; +u64 cnt; +u64 misses; +}; +static void bpf_prog_get_stats (const struct bpf_prog *prog, -struct bpf_prog_stats *stats) +struct bpf_prog_kstats *stats) {u64 nsecs =0 , cnt =0 , misses =0 ;int cpu;
@@ -1818,9 +1824,9 @@ static void bpf_prog_get_stats(const struct bpf_prog *prog, st =per_cpu_ptr (prog->stats, cpu);do { start =u64_stats_fetch_begin_irq (&st->syncp); - tnsecs = st->nsecs; - tcnt = st->cnt; - tmisses = st->misses; + tnsecs =u64_stats_read (&st->nsecs); + tcnt =u64_stats_read (&st->cnt); + tmisses =u64_stats_read (&st->misses); }while (u64_stats_fetch_retry_irq (&st->syncp, start)); nsecs += tnsecs; cnt += tcnt;
@@ -1836,7 +1842,7 @@ static void bpf_prog_show_fdinfo(struct seq_file *m, struct file *filp) {const struct bpf_prog *prog = filp->private_data;char prog_tag[sizeof (prog->tag) *2 +1 ] = { }; -struct bpf_prog_stats stats; +struct bpf_prog_kstats stats;bpf_prog_get_stats (prog, &stats);bin2hex (prog_tag, prog->tag,sizeof (prog->tag));
@@ -3575,7 +3581,7 @@ static int bpf_prog_get_info_by_fd(struct file *file,struct bpf_prog_info __user *uinfo =u64_to_user_ptr (attr->info.info);struct bpf_prog_info info;u32 info_len = attr->info.info_len; -struct bpf_prog_stats stats; +struct bpf_prog_kstats stats;char __user *uinsns;u32 ulen;int err;
kernel/bpf/trampoline.c +3 -3
@@ -545,7 +545,7 @@ static void notrace inc_misses_counter(struct bpf_prog *prog) stats =this_cpu_ptr (prog->stats);u64_stats_update_begin (&stats->syncp); - stats->misses++; +u64_stats_inc (&stats->misses);u64_stats_update_end (&stats->syncp); }
@@ -590,8 +590,8 @@ static void notrace update_prog_stats(struct bpf_prog *prog, stats =this_cpu_ptr (prog->stats); flags =u64_stats_update_begin_irqsave (&stats->syncp); - stats->cnt++; - stats->nsecs +=sched_clock () - start; +u64_stats_inc (&stats->cnt); +u64_stats_add (&stats->nsecs,sched_clock () - start);u64_stats_update_end_irqrestore (&stats->syncp, flags); } }
Signature · 1 line
2.33.0.1079.g6e70778dc9-goog
Alexei Starovoitov
AppliedQuoted 8 lines · Eric Dumazet
From: Eric Dumazet <[email protected]>
Two first patches fix bugs added in 5.1 and 5.5
Third patch replaces the u64 fields in struct
bpf_prog_statswithu64_stats_tones to avoid possible sampling errors, in case of load/store stearing.
Applied to bpf-next. Thanks!