Skip to contentSkip to conversation

2.6.17-mm3 -- BUG: illegal lock usage -- illegal {softirq-on-W} -> {in-softirq-R} usage.

PATCH In review Miles Lane 5 participants 22 messages Jun 29, 2006 – Jul 1, 2006

Subsystems (5)
lore
Share & tools
lore

Miles Lane

author
#0 lore raw

[ BUG: illegal lock usage! ]
----------------------------
illegal {softirq-on-W} -> {in-softirq-R} usage. java_vm/4418 [HC0[0]:SC1[1]:HE1:SE0] takes:
(&sk->sk_dst_lock){---?}, at: [<c119d0a9>] sk_dst_check+0x1b/0xe6 {softirq-on-W} state was registered at:

 [<c102d1c8>] lock_acquire+0x60/0x80
 [<c12012d7>] _write_lock+0x23/0x32
 [<c11ddbe7>] inet_bind+0x16c/0x1cc
 [<c119ae58>] sys_bind+0x61/0x80
 [<c119b465>] sys_socketcall+0x7d/0x186
 [<c1002d6d>] sysenter_past_esp+0x56/0x8d
irq event stamp: 11052
hardirqs last  enabled at (11052): [<c105d454>] kmem_cache_alloc+0x89/0xa6
hardirqs last disabled at (11051): [<c105d405>] kmem_cache_alloc+0x3a/0xa6
softirqs last  enabled at (11040): [<c11a506d>] dev_queue_xmit+0x224/0x24b
softirqs last disabled at (11041): [<c1004a64>] do_softirq+0x58/0xbd

other info that might help us debug this:
1 lock held by java_vm/4418:
#0: (af_family_keys + (sk)->sk_family#4){-+..}, at: [<f93c9281>]

tcp_v6_rcv+0x308/0x7b7 [ipv6]

stack backtrace:
[<c1003502>] show_trace_log_lvl+0x54/0xfd
[<c1003b6a>] show_trace+0xd/0x10
[<c1003c0e>] dump_stack+0x19/0x1b
[<c102b833>] print_usage_bug+0x1cc/0x1d9
[<c102bd34>] mark_lock+0x193/0x360
[<c102c94a>] __lock_acquire+0x3b7/0x970
[<c102d1c8>] lock_acquire+0x60/0x80
[<c12013eb>] _read_lock+0x23/0x32
[<c119d0a9>] sk_dst_check+0x1b/0xe6
[<f93ae479>] ip6_dst_lookup+0x31/0x172 [ipv6]
[<f93c7065>] tcp_v6_send_synack+0x10f/0x238 [ipv6]
[<f93c7dc5>] tcp_v6_conn_request+0x281/0x2c7 [ipv6]
[<c11cca33>] tcp_rcv_state_process+0x5d/0xbde
[<f93c7414>] tcp_v6_do_rcv+0x26d/0x384 [ipv6]
[<f93c96d6>] tcp_v6_rcv+0x75d/0x7b7 [ipv6]
[<f93afadd>] ip6_input+0x201/0x2d1 [ipv6]
[<f93b002d>] ipv6_rcv+0x190/0x1bf [ipv6]
[<c11a3200>] netif_receive_skb+0x2e6/0x37f
[<c11a4b81>] process_backlog+0x80/0x112
[<c11a4c9e>] net_rx_action+0x8b/0x1e8
[<c101a711>] __do_softirq+0x55/0xb0
[<c1004a64>] do_softirq+0x58/0xbd
[<c101a978>] local_bh_enable+0xd0/0x107
[<c11a506d>] dev_queue_xmit+0x224/0x24b
[<c11a9bb8>] neigh_resolve_output+0x1e2/0x20e
[<f93add64>] ip6_output2+0x1de/0x1fc [ipv6]
[<f93ae41e>] ip6_output+0x69c/0x6c6 [ipv6]
[<f93aeb58>] ip6_xmit+0x22b/0x295 [ipv6]
[<f93cc893>] inet6_csk_xmit+0x200/0x20e [ipv6]
[<c11cea04>] tcp_transmit_skb+0x5de/0x60c
[<c11d0cd7>] tcp_connect+0x2bb/0x31a
[<f93c894a>] tcp_v6_connect+0x520/0x655 [ipv6]
[<c11dd889>] inet_stream_connect+0x83/0x20f
[<c119adda>] sys_connect+0x67/0x84
[<c119b474>] sys_socketcall+0x8c/0x186
[<c1002d6d>] sysenter_past_esp+0x56/0x8d
#1 lore raw
Quoted 2 lines · On Thu, 29 Jun 2006 12:01:06 -0700 "Miles Lane" <[email protected]> wrote:

[ BUG: illegal lock usage! ]
----------------------------

This is claiming that we're taking sk->sk_dst_lock in a deadlockable manner.

Quoted 1 line

illegal {softirq-on-W} -> {in-softirq-R} usage.

It found someone doing write_lock(sk_dst_lock) with softirqs enabled, but someone else takes read_lock(dst_lock) inside softirqs.

Quoted 9 lines

java_vm/4418 [HC0[0]:SC1[1]:HE1:SE0] takes:
(&sk->sk_dst_lock){---?}, at: [<c119d0a9>] sk_dst_check+0x1b/0xe6
{softirq-on-W} state was registered at:

  [<c102d1c8>] lock_acquire+0x60/0x80
  [<c12012d7>] _write_lock+0x23/0x32
  [<c11ddbe7>] inet_bind+0x16c/0x1cc
  [<c119ae58>] sys_bind+0x61/0x80
  [<c119b465>] sys_socketcall+0x7d/0x186
  [<c1002d6d>] sysenter_past_esp+0x56/0x8d
	inet_bind()
	->sk_dst_get
	  ->read_lock(&sk->sk_dst_lock)
Quoted 10 lines
irq event stamp: 11052
hardirqs last  enabled at (11052): [<c105d454>] kmem_cache_alloc+0x89/0xa6
hardirqs last disabled at (11051): [<c105d405>] kmem_cache_alloc+0x3a/0xa6
softirqs last  enabled at (11040): [<c11a506d>] dev_queue_xmit+0x224/0x24b
softirqs last disabled at (11041): [<c1004a64>] do_softirq+0x58/0xbd

other info that might help us debug this:
1 lock held by java_vm/4418:
#0: (af_family_keys + (sk)->sk_family#4){-+..}, at: [<f93c9281>]
tcp_v6_rcv+0x308/0x7b7 [ipv6]

	softirq
	->ip6_dst_lookup
	  ->sk_dst_check
	    ->sk_dst_reset
	      ->write_lock(&sk->sk_dst_lock);
Quoted 29 lines
stack backtrace:
 [<c1003502>] show_trace_log_lvl+0x54/0xfd
 [<c1003b6a>] show_trace+0xd/0x10
 [<c1003c0e>] dump_stack+0x19/0x1b
 [<c102b833>] print_usage_bug+0x1cc/0x1d9
 [<c102bd34>] mark_lock+0x193/0x360
 [<c102c94a>] __lock_acquire+0x3b7/0x970
 [<c102d1c8>] lock_acquire+0x60/0x80
 [<c12013eb>] _read_lock+0x23/0x32
 [<c119d0a9>] sk_dst_check+0x1b/0xe6
 [<f93ae479>] ip6_dst_lookup+0x31/0x172 [ipv6]
 [<f93c7065>] tcp_v6_send_synack+0x10f/0x238 [ipv6]
 [<f93c7dc5>] tcp_v6_conn_request+0x281/0x2c7 [ipv6]
 [<c11cca33>] tcp_rcv_state_process+0x5d/0xbde
 [<f93c7414>] tcp_v6_do_rcv+0x26d/0x384 [ipv6]
 [<f93c96d6>] tcp_v6_rcv+0x75d/0x7b7 [ipv6]
 [<f93afadd>] ip6_input+0x201/0x2d1 [ipv6]
 [<f93b002d>] ipv6_rcv+0x190/0x1bf [ipv6]
 [<c11a3200>] netif_receive_skb+0x2e6/0x37f
 [<c11a4b81>] process_backlog+0x80/0x112
 [<c11a4c9e>] net_rx_action+0x8b/0x1e8
 [<c101a711>] __do_softirq+0x55/0xb0
 [<c1004a64>] do_softirq+0x58/0xbd
 [<c101a978>] local_bh_enable+0xd0/0x107
 [<c11a506d>] dev_queue_xmit+0x224/0x24b
 [<c11a9bb8>] neigh_resolve_output+0x1e2/0x20e
 [<f93add64>] ip6_output2+0x1de/0x1fc [ipv6]
 [<f93ae41e>] ip6_output+0x69c/0x6c6 [ipv6]
 [<f93aeb58>] ip6_xmit+0x22b/0x295 [ipv6]
Quoted 8 lines
 [<f93cc893>] inet6_csk_xmit+0x200/0x20e [ipv6]
 [<c11cea04>] tcp_transmit_skb+0x5de/0x60c
 [<c11d0cd7>] tcp_connect+0x2bb/0x31a
 [<f93c894a>] tcp_v6_connect+0x520/0x655 [ipv6]
 [<c11dd889>] inet_stream_connect+0x83/0x20f
 [<c119adda>] sys_connect+0x67/0x84
 [<c119b474>] sys_socketcall+0x8c/0x186
 [<c1002d6d>] sysenter_past_esp+0x56/0x8d

So the allegation is that if a softirq runs sk_dst_reset() while process-context code is running sk_dst_set(), we'll do write_lock() while holding read_lock().

#2 lore raw
Quoted 23 lines · Andrew Morton
Quoted 2 lines · On Thu, 29 Jun 2006 12:01:06 -0700 "Miles Lane" <[email protected]> wrote:

[ BUG: illegal lock usage! ]
----------------------------

This is claiming that we're taking sk->sk_dst_lock in a deadlockable manner.

Quoted 1 line

illegal {softirq-on-W} -> {in-softirq-R} usage.

It found someone doing write_lock(sk_dst_lock) with softirqs enabled, but someone else takes read_lock(dst_lock) inside softirqs.

Quoted 9 lines

java_vm/4418 [HC0[0]:SC1[1]:HE1:SE0] takes:
(&sk->sk_dst_lock){---?}, at: [<c119d0a9>] sk_dst_check+0x1b/0xe6
{softirq-on-W} state was registered at:

  [<c102d1c8>] lock_acquire+0x60/0x80
  [<c12012d7>] _write_lock+0x23/0x32
  [<c11ddbe7>] inet_bind+0x16c/0x1cc
  [<c119ae58>] sys_bind+0x61/0x80
  [<c119b465>] sys_socketcall+0x7d/0x186
  [<c1002d6d>] sysenter_past_esp+0x56/0x8d
Quoted 3 lines
	inet_bind()
	->sk_dst_get
	  ->read_lock(&sk->sk_dst_lock)

actually write_lock() not read_lock()

Quoted 12 lines
Quoted 11 lines
irq event stamp: 11052
hardirqs last  enabled at (11052): [<c105d454>] kmem_cache_alloc+0x89/0xa6
hardirqs last disabled at (11051): [<c105d405>] kmem_cache_alloc+0x3a/0xa6
softirqs last  enabled at (11040): [<c11a506d>] dev_queue_xmit+0x224/0x24b
softirqs last disabled at (11041): [<c1004a64>] do_softirq+0x58/0xbd

other info that might help us debug this:
1 lock held by java_vm/4418:
#0: (af_family_keys + (sk)->sk_family#4){-+..}, at: [<f93c9281>]
tcp_v6_rcv+0x308/0x7b7 [ipv6]

Quoted 5 lines
	softirq
	->ip6_dst_lookup
	  ->sk_dst_check
	    ->sk_dst_reset
	      ->write_lock(&sk->sk_dst_lock);

write_lock.. or read_lock() ?

Quoted 10 lines
Quoted 10 lines
stack backtrace:
 [<c1003502>] show_trace_log_lvl+0x54/0xfd
 [<c1003b6a>] show_trace+0xd/0x10
 [<c1003c0e>] dump_stack+0x19/0x1b
 [<c102b833>] print_usage_bug+0x1cc/0x1d9
 [<c102bd34>] mark_lock+0x193/0x360
 [<c102c94a>] __lock_acquire+0x3b7/0x970
 [<c102d1c8>] lock_acquire+0x60/0x80
 [<c12013eb>] _read_lock+0x23/0x32

backtrace says read lock to me ...

Quoted 10 lines
Quoted 5 lines
 [<c119d0a9>] sk_dst_check+0x1b/0xe6
 [<f93ae479>] ip6_dst_lookup+0x31/0x172 [ipv6]
 [<f93c7065>] tcp_v6_send_synack+0x10f/0x238 [ipv6]
 [<f93c7dc5>] tcp_v6_conn_request+0x281/0x2c7 [ipv6]
 [<c11cca33>] tcp_rcv_state_process+0x5d/0xbde

So the allegation is that if a softirq runs sk_dst_reset() while process-context code is running sk_dst_set(), we'll do write_lock() while holding read_lock().

hmm or...

we're doing a write_lock(), then an interrupt can happen that triggers the softirq that triggers the read_lock(), which will deadlock because we interrupted the writer...

#3 lore raw
Quoted 24 lines · On Thu, 29 Jun 2006 21:42:34 +0200 Arjan van de Ven <[email protected]> wrote:
Quoted 23 lines · Andrew Morton
Quoted 2 lines · On Thu, 29 Jun 2006 12:01:06 -0700 "Miles Lane" <[email protected]> wrote:

[ BUG: illegal lock usage! ]
----------------------------

This is claiming that we're taking sk->sk_dst_lock in a deadlockable manner.

Quoted 1 line

illegal {softirq-on-W} -> {in-softirq-R} usage.

It found someone doing write_lock(sk_dst_lock) with softirqs enabled, but someone else takes read_lock(dst_lock) inside softirqs.

Quoted 9 lines

java_vm/4418 [HC0[0]:SC1[1]:HE1:SE0] takes:
(&sk->sk_dst_lock){---?}, at: [<c119d0a9>] sk_dst_check+0x1b/0xe6
{softirq-on-W} state was registered at:

  [<c102d1c8>] lock_acquire+0x60/0x80
  [<c12012d7>] _write_lock+0x23/0x32
  [<c11ddbe7>] inet_bind+0x16c/0x1cc
  [<c119ae58>] sys_bind+0x61/0x80
  [<c119b465>] sys_socketcall+0x7d/0x186
  [<c1002d6d>] sysenter_past_esp+0x56/0x8d
Quoted 6 lines
Quoted 3 lines
	inet_bind()
	->sk_dst_get
	  ->read_lock(&sk->sk_dst_lock)

actually write_lock() not read_lock()

static inline struct dst_entry *
sk_dst_get(struct sock *sk)

{
	struct dst_entry *dst;

	read_lock(&sk->sk_dst_lock);
Quoted 12 lines
Quoted 12 lines
Quoted 11 lines
irq event stamp: 11052
hardirqs last  enabled at (11052): [<c105d454>] kmem_cache_alloc+0x89/0xa6
hardirqs last disabled at (11051): [<c105d405>] kmem_cache_alloc+0x3a/0xa6
softirqs last  enabled at (11040): [<c11a506d>] dev_queue_xmit+0x224/0x24b
softirqs last disabled at (11041): [<c1004a64>] do_softirq+0x58/0xbd

other info that might help us debug this:
1 lock held by java_vm/4418:
#0: (af_family_keys + (sk)->sk_family#4){-+..}, at: [<f93c9281>]
tcp_v6_rcv+0x308/0x7b7 [ipv6]

Quoted 7 lines
Quoted 5 lines
	softirq
	->ip6_dst_lookup
	  ->sk_dst_check
	    ->sk_dst_reset
	      ->write_lock(&sk->sk_dst_lock);

write_lock.. or read_lock() ?

static inline void
sk_dst_reset(struct sock *sk)

{
	write_lock(&sk->sk_dst_lock);
Quoted 23 lines
Quoted 10 lines
Quoted 10 lines
stack backtrace:
 [<c1003502>] show_trace_log_lvl+0x54/0xfd
 [<c1003b6a>] show_trace+0xd/0x10
 [<c1003c0e>] dump_stack+0x19/0x1b
 [<c102b833>] print_usage_bug+0x1cc/0x1d9
 [<c102bd34>] mark_lock+0x193/0x360
 [<c102c94a>] __lock_acquire+0x3b7/0x970
 [<c102d1c8>] lock_acquire+0x60/0x80
 [<c12013eb>] _read_lock+0x23/0x32

backtrace says read lock to me ...

Quoted 10 lines
Quoted 5 lines
 [<c119d0a9>] sk_dst_check+0x1b/0xe6
 [<f93ae479>] ip6_dst_lookup+0x31/0x172 [ipv6]
 [<f93c7065>] tcp_v6_send_synack+0x10f/0x238 [ipv6]
 [<f93c7dc5>] tcp_v6_conn_request+0x281/0x2c7 [ipv6]
 [<c11cca33>] tcp_rcv_state_process+0x5d/0xbde

So the allegation is that if a softirq runs sk_dst_reset() while process-context code is running sk_dst_set(), we'll do write_lock() while holding read_lock().

Quoted 6 lines

hmm or...

we're doing a write_lock(), then an interrupt can happen that triggers the softirq that triggers the read_lock(), which will deadlock because we interrupted the writer...

either way ain't good ;)

#4 lore raw
Quoted 4 lines · Andrew Morton
Quoted 4 lines
Quoted 4 lines
    inet_bind()
    ->sk_dst_get
      ->read_lock(&sk->sk_dst_lock)

We are still holding the sock lock when doing sk_dst_get.

Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
Quoted 3 lines

1 lock held by java_vm/4418:
#0: (af_family_keys + (sk)->sk_family#4){-+..}, at: [<f93c9281>]
tcp_v6_rcv+0x308/0x7b7 [ipv6]

    softirq
    ->ip6_dst_lookup
      ->sk_dst_check
        ->sk_dst_reset
          ->write_lock(&sk->sk_dst_lock);

The sock lock prevents this path from being entered. Instead the received TCP packet is queued and replayed when the sock lock is released.

Cheers,

Signature · 4 lines
Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
#5 lore raw
In reply to Miles Lane
Quoted 52 lines · Miles Lane

[ BUG: illegal lock usage! ]
----------------------------
illegal {softirq-on-W} -> {in-softirq-R} usage. java_vm/4418 [HC0[0]:SC1[1]:HE1:SE0] takes:
(&sk->sk_dst_lock){---?}, at: [<c119d0a9>] sk_dst_check+0x1b/0xe6 {softirq-on-W} state was registered at:

 [<c102d1c8>] lock_acquire+0x60/0x80
 [<c12012d7>] _write_lock+0x23/0x32
 [<c11ddbe7>] inet_bind+0x16c/0x1cc
 [<c119ae58>] sys_bind+0x61/0x80
 [<c119b465>] sys_socketcall+0x7d/0x186
 [<c1002d6d>] sysenter_past_esp+0x56/0x8d
irq event stamp: 11052
hardirqs last  enabled at (11052): [<c105d454>] kmem_cache_alloc+0x89/0xa6
hardirqs last disabled at (11051): [<c105d405>] kmem_cache_alloc+0x3a/0xa6
softirqs last  enabled at (11040): [<c11a506d>] dev_queue_xmit+0x224/0x24b
softirqs last disabled at (11041): [<c1004a64>] do_softirq+0x58/0xbd

other info that might help us debug this:
1 lock held by java_vm/4418:
#0: (af_family_keys + (sk)->sk_family#4){-+..}, at: [<f93c9281>]

tcp_v6_rcv+0x308/0x7b7 [ipv6]

stack backtrace:
[<c1003502>] show_trace_log_lvl+0x54/0xfd
[<c1003b6a>] show_trace+0xd/0x10
[<c1003c0e>] dump_stack+0x19/0x1b
[<c102b833>] print_usage_bug+0x1cc/0x1d9
[<c102bd34>] mark_lock+0x193/0x360
[<c102c94a>] __lock_acquire+0x3b7/0x970
[<c102d1c8>] lock_acquire+0x60/0x80
[<c12013eb>] _read_lock+0x23/0x32
[<c119d0a9>] sk_dst_check+0x1b/0xe6
[<f93ae479>] ip6_dst_lookup+0x31/0x172 [ipv6]
[<f93c7065>] tcp_v6_send_synack+0x10f/0x238 [ipv6]
[<f93c7dc5>] tcp_v6_conn_request+0x281/0x2c7 [ipv6]
[<c11cca33>] tcp_rcv_state_process+0x5d/0xbde
[<f93c7414>] tcp_v6_do_rcv+0x26d/0x384 [ipv6]
[<f93c96d6>] tcp_v6_rcv+0x75d/0x7b7 [ipv6]
[<f93afadd>] ip6_input+0x201/0x2d1 [ipv6]
[<f93b002d>] ipv6_rcv+0x190/0x1bf [ipv6]
[<c11a3200>] netif_receive_skb+0x2e6/0x37f
[<c11a4b81>] process_backlog+0x80/0x112
[<c11a4c9e>] net_rx_action+0x8b/0x1e8
[<c101a711>] __do_softirq+0x55/0xb0
[<c1004a64>] do_softirq+0x58/0xbd
[<c101a978>] local_bh_enable+0xd0/0x107
[<c11a506d>] dev_queue_xmit+0x224/0x24b
[<c11a9bb8>] neigh_resolve_output+0x1e2/0x20e
[<f93add64>] ip6_output2+0x1de/0x1fc [ipv6]
[<f93ae41e>] ip6_output+0x69c/0x6c6 [ipv6]
[<f93aeb58>] ip6_xmit+0x22b/0x295 [ipv6]
Quoted 8 lines
[<f93cc893>] inet6_csk_xmit+0x200/0x20e [ipv6]
[<c11cea04>] tcp_transmit_skb+0x5de/0x60c
[<c11d0cd7>] tcp_connect+0x2bb/0x31a
[<f93c894a>] tcp_v6_connect+0x520/0x655 [ipv6]
[<c11dd889>] inet_stream_connect+0x83/0x20f
[<c119adda>] sys_connect+0x67/0x84
[<c119b474>] sys_socketcall+0x8c/0x186
[<c1002d6d>] sysenter_past_esp+0x56/0x8d

hm, is this all the log output you got? In particular the printouts of where various lock-class state changing events happened that are important too. Could you send me the whole dmesg?

Ingo

#6 lore raw
Quoted 3 lines · Ingo Molnar

hm, is this all the log output you got? In particular the printouts of where various lock-class state changing events happened that are important too. Could you send me the whole dmesg?

sorry, i take this back - the messages are complete.

Ingo

#7 lore raw

PATCH 1/1 lockdep, annotate slocks: turn lockdep off for them

In reply to Ingo Molnar

Miles, does the patch below make the message go away?

Ingo

----------------
Subject: lockdep, annotate slocks: turn lockdep off for them
From: Ingo Molnar <[email protected]>

temporary solution to turn off slock related false positives:
the slock is pretty much the only lock type in the kernel that is half spinlock, half waitqueue. "Process level" and "softirq level" uses of slock are excluded - albeit the spinlock itself is not permanently held in process context.

The right solution will be to annotate slock uses with acquire()/release(). (i.e. to treat sock_owned_by_user() flagged areas as an exclusion area too)

(this temporary solution is not as bad as it might sound, because it does not eliminate the various ->sk_backlog_rcv() related dependencies from the validator's dependency graph - what it does is that it doesnt record them relative to slock. [the callbacks will still be executed and covered when the backlog is processed.])

---
 net/core/sock.c |   12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)
net/core/sock.c +10 -2
Index: linux/net/core/sock.c
===================================================================
@@ -250,9 +250,17 @@ int sk_receive_skb(struct sock *sk, stru
 	skb->dev = NULL;
 
 	bh_lock_sock(sk);
-	if (!sock_owned_by_user(sk))
+	if (!sock_owned_by_user(sk)) {
+		/*
+		 * FIXME: teach the validator about slocks.
+		 *
+		 * For now we dont record dependencies in
+		 * this codepath.
+		 */
+		lockdep_off();
 		rc = sk->sk_backlog_rcv(sk, skb);
-	else
+		lockdep_on();
+	} else
 		sk_add_backlog(sk, skb);
 	bh_unlock_sock(sk);
 out:
#8 lore raw

Re: [patch] lockdep, annotate slocks: turn lockdep off for them

Quoted 9 lines · Ingo Molnar

temporary solution to turn off slock related false positives: the slock is pretty much the only lock type in the kernel that is half spinlock, half waitqueue. "Process level" and "softirq level" uses of slock are excluded - albeit the spinlock itself is not permanently held in process context.

The right solution will be to annotate slock uses with acquire()/release(). (i.e. to treat sock_owned_by_user() flagged areas as an exclusion area too)

i've got the right solution coded up too - it was a fairly straightforward process of teaching a new lock type (sk_lock) to the validator.

Miles, do the two attached patches (ontop of -mm4) fix your problem too? [you should unapply the previous lockdep_off()/on() patch i sent)

Andrew, i'd suggest to not apply the two attached patches yet - while they work fine for me we first should see whether they fix Miles' problem too.

Herbert, do the acquire/release semantics as expressed in the lockdep-annotate-slock.patch match sk_lock semantics?

the patch also makes the slock names more readable:

neptune:/home/mingo> cat /proc/lockdep | grep AF

 c07f7088 OPS:    4125 FD:    1 BD:    2 -...: slock-AF_UNIX
 c07f7188 OPS:    1375 FD:    2 BD:    1 --..: sk_lock-AF_UNIX
 c07f7100 OPS:     252 FD:    1 BD:    2 -...: slock-AF_NETLINK
 c07f7200 OPS:      84 FD:    2 BD:    1 --..: sk_lock-AF_NETLINK
 c07f7090 OPS:    2641 FD:   20 BD:    4 -+..: slock-AF_INET
 c07f7190 OPS:     851 FD:   64 BD:    1 --..: sk_lock-AF_INET
 c07f7108 OPS:      27 FD:    1 BD:    2 -...: slock-AF_PACKET
 c07f7208 OPS:       9 FD:    4 BD:    1 --..: sk_lock-AF_PACKET
 c07f7091 OPS:     285 FD:   44 BD:    2 -+..: slock-AF_INET/1

(that should make it easier to tell which sk_lock class is mentioned in a validator message, instead of the current slock#2/slock#3 enumeration.)

Ingo

Subject: lockdep: core, add set_class_and_name()
From: Ingo Molnar <[email protected]>

add set_class_and_name() API, to allow lock initialization places to beautify the naming of their lock classes.

---
 include/linux/lockdep.h |    4 ++++
 1 file changed, 4 insertions(+)
include/linux/lockdep.h +4 -0
Index: linux/include/linux/lockdep.h
===================================================================
@@ -212,6 +212,8 @@ extern void lockdep_init_map(struct lock
  */
 #define lockdep_set_class(lock, key) \
 		lockdep_init_map(&(lock)->dep_map, #key, key)
+#define lockdep_set_class_and_name(lock, key, name) \
+		lockdep_init_map(&(lock)->dep_map, name, key)
 
 /*
  * Acquire a lock.
@@ -257,6 +259,8 @@ static inline int lockdep_internal(void)
 # define lockdep_info()				do { } while (0)
 # define lockdep_init_map(lock, name, key)	do { } while (0)
 # define lockdep_set_class(lock, key)		do { (void)(key); } while (0)
+# define lockdep_set_class_and_name(lock, key, name) \
+		do { (void)(key); (void)(name); } while (0)
 # define INIT_LOCKDEP
 # define lockdep_reset()		do { debug_locks = 1; } while (0)
 # define lockdep_free_key_range(start, size)	do { } while (0)

Subject: lockdep, annotate sk_locks
From: Ingo Molnar <[email protected]>

Teach sk_lock semantics to the lock validator. In the softirq path the slock has mutex_trylock()+mutex_unlock() semantics, in the process context sock_lock() case it has mutex_lock()/mutex_unlock() semantics.

Thus we treat sock_owned_by_user() flagged areas as an exclusion area too, not just those areas covered by a held sk_lock.slock.

Effect on non-lockdep kernels: minimal, sk_lock_sock_init() has been turned into an inline function.

---
 include/net/sock.h |   20 +++++------
 net/core/sock.c    |   93 +++++++++++++++++++++++++++++++++++++++++++++++------
 2 files changed, 94 insertions(+), 19 deletions(-)
include/net/sock.h +10 -10
Index: linux/include/net/sock.h
===================================================================
@@ -44,6 +44,7 @@
 #include <linux/timer.h>
 #include <linux/cache.h>
 #include <linux/module.h>
+#include <linux/lockdep.h>
 #include <linux/netdevice.h>
 #include <linux/skbuff.h>	/* struct sk_buff */
 #include <linux/security.h>
@@ -78,18 +79,17 @@ typedef struct {
 	spinlock_t		slock;
 	struct sock_iocb	*owner;
 	wait_queue_head_t	wq;
+	/*
+	 * We express the mutex-alike socket_lock semantics
+	 * to the lock validator by explicitly managing
+	 * the slock as a lock variant (in addition to
+	 * the slock itself):
+	 */
+#ifdef CONFIG_DEBUG_LOCK_ALLOC
+	struct lockdep_map dep_map;
+#endif
 } socket_lock_t;
 
-extern struct lock_class_key af_family_keys[AF_MAX];
-
-#define sock_lock_init(__sk) \
-do {	spin_lock_init(&((__sk)->sk_lock.slock)); \
-	lockdep_set_class(&(__sk)->sk_lock.slock, \
-			  af_family_keys + (__sk)->sk_family); \
-	(__sk)->sk_lock.owner = NULL; \
-	init_waitqueue_head(&((__sk)->sk_lock.wq)); \
-} while(0)
-
 struct sock;
 struct proto;
 
net/core/sock.c +84 -9

130 lines not on this page · raw message

#9 lore raw

Re: [patch] lockdep, annotate slocks: turn lockdep off for them

Quoted 3 lines · Ingo Molnar

Herbert, do the acquire/release semantics as expressed in the lockdep-annotate-slock.patch match sk_lock semantics?

I think it should be fine.

net/core/sock.c int sk_receive_skb(struct sock *sk, stru · line 291

Quoted 11 lines
net/core/sock.c +6 -1
@@ -250,9 +283,18 @@ int sk_receive_skb(struct sock *sk, stru
 	skb->dev = NULL;
 
 	bh_lock_sock(sk);
-	if (!sock_owned_by_user(sk))
+	if (!sock_owned_by_user(sk)) {
+		/*
+		 * trylock + unlock semantics:
+		 */
+		spin_release(&sk->sk_lock.slock.dep_map, 1, _RET_IP_);
+		mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);

Although it would seem that keeping the spin lock would fit the actual semantics better. I suppose there must be a technical reason why this wouldn't work.

Cheers,

Signature · 4 lines
Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
#10 lore raw

Re: [patch] lockdep, annotate slocks: turn lockdep off for them

Quoted 4 lines · Herbert Xu
Quoted 4 lines
(quoted diff) +2 -1
 	bh_lock_sock(sk);
-	if (!sock_owned_by_user(sk))
+	if (!sock_owned_by_user(sk)) {
+		/*
Quoted 8 lines

Review

Quoted 4 lines
(quoted diff) +4 -0
+		 * trylock + unlock semantics:
+		 */
+		spin_release(&sk->sk_lock.slock.dep_map, 1, _RET_IP_);
+		mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);

Although it would seem that keeping the spin lock would fit the actual semantics better. I suppose there must be a technical reason why this wouldn't work.

good point. The basic issue is the 'virtual lock inversion' that occurs in the lock vs. release paths. [between taking the slock and taking the new sk_lock type]

The situation is like this: we construct 'complex' lock types [mutex, rwsem, sk_lock] out of 'primitive' lock types [spinlock, rwlock]. Both the complex type and the primite types exist separately, and might have lock-validator acquire/release operations. These locks can interact and if we do the complex lock acquire/release while holding the primitive lock, the validator sees inverse ordering between them.

For the mutex code i solved the inversion problem by using a raw_spinlock for the primitive type (which has no lockdep operations), hence the complex lock type.

But in this particular sk_lock case we can do it even more cleanly i think and can preserve the lockdep awareness of the primitive type too:
by releasing the complex lock before taking the primitive lock in the release_sock() unlock path. The updated patch below does this - and thus i was able to remove the dropping of the primitive spinlock type.

it is not a problem that the release of the complex lock type does not happen inside the critical section: from the point where we release the complex lock-type _this_ context cannot take any other locks, so there are no dependencies missed.

As you can see, the lock validator can easily cover completely new lock types like sk_lock too, as long as the new lock type has some minimalistic "works like a lock" properties. (such as owner-does-unlock)

later on i'll try the same cleanup for the mutex code too - it should be possible. (that way the implementation of complex lock types can be lock-validator checked too)

Ingo

--------------->
Subject: lockdep, annotate sk_locks
From: Ingo Molnar <[email protected]>

Teach sk_lock semantics to the lock validator. In the softirq path the slock has mutex_trylock()+mutex_unlock() semantics, in the process context sock_lock() case it has mutex_lock()/mutex_unlock() semantics.

Thus we treat sock_owned_by_user() flagged areas as an exclusion area too, not just those areas covered by a held sk_lock.slock.

Effect on non-lockdep kernels: minimal, sk_lock_sock_init() has been turned into an inline function.

---
 include/net/sock.h |   20 +++++------
 net/core/sock.c    |   92 +++++++++++++++++++++++++++++++++++++++++++++++------
 2 files changed, 93 insertions(+), 19 deletions(-)
include/net/sock.h +10 -10
Index: linux/include/net/sock.h
===================================================================
@@ -44,6 +44,7 @@
 #include <linux/timer.h>
 #include <linux/cache.h>
 #include <linux/module.h>
+#include <linux/lockdep.h>
 #include <linux/netdevice.h>
 #include <linux/skbuff.h>	/* struct sk_buff */
 #include <linux/security.h>
@@ -78,18 +79,17 @@ typedef struct {
 	spinlock_t		slock;
 	struct sock_iocb	*owner;
 	wait_queue_head_t	wq;
+	/*
+	 * We express the mutex-alike socket_lock semantics
+	 * to the lock validator by explicitly managing
+	 * the slock as a lock variant (in addition to
+	 * the slock itself):
+	 */
+#ifdef CONFIG_DEBUG_LOCK_ALLOC
+	struct lockdep_map dep_map;
+#endif
 } socket_lock_t;
 
-extern struct lock_class_key af_family_keys[AF_MAX];
-
-#define sock_lock_init(__sk) \
-do {	spin_lock_init(&((__sk)->sk_lock.slock)); \
-	lockdep_set_class(&(__sk)->sk_lock.slock, \
-			  af_family_keys + (__sk)->sk_family); \
-	(__sk)->sk_lock.owner = NULL; \
-	init_waitqueue_head(&((__sk)->sk_lock.wq)); \
-} while(0)
-
 struct sock;
 struct proto;
 
net/core/sock.c +83 -9

129 lines not on this page · raw message

#11 lore raw

Re: [patch] lockdep, annotate slocks: turn lockdep off for them

Quoted 8 lines · Ingo Molnar

As you can see, the lock validator can easily cover completely new lock types like sk_lock too, as long as the new lock type has some minimalistic "works like a lock" properties. (such as owner-does-unlock)

later on i'll try the same cleanup for the mutex code too - it should be possible. (that way the implementation of complex lock types can be lock-validator checked too)

Yes, this looks very nice. Thanks!

Signature · 4 lines
Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
author
#12 lore raw

Re: [patch] lockdep, annotate slocks: turn lockdep off for them

In reply to Ingo Molnar
Quoted 240 lines · Ingo Molnar
Quoted 9 lines · Herbert Xu

net/core/sock.c int sk_receive_skb(struct sock *sk, stru · line 290

Quoted 8 lines
    bh_lock_sock(sk);
net/core/sock.c +6 -1
-   if (!sock_owned_by_user(sk))
+   if (!sock_owned_by_user(sk)) {
+           /*
+            * trylock + unlock semantics:
+            */
+           spin_release(&sk->sk_lock.slock.dep_map, 1, _RET_IP_);
+           mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
Quoted 3 lines

Although it would seem that keeping the spin lock would fit the actual semantics better. I suppose there must be a technical reason why this wouldn't work.

good point. The basic issue is the 'virtual lock inversion' that occurs in the lock vs. release paths. [between taking the slock and taking the new sk_lock type]

The situation is like this: we construct 'complex' lock types [mutex, rwsem, sk_lock] out of 'primitive' lock types [spinlock, rwlock]. Both the complex type and the primite types exist separately, and might have lock-validator acquire/release operations. These locks can interact and if we do the complex lock acquire/release while holding the primitive lock, the validator sees inverse ordering between them.

For the mutex code i solved the inversion problem by using a raw_spinlock for the primitive type (which has no lockdep operations), hence the complex lock type.

But in this particular sk_lock case we can do it even more cleanly i think and can preserve the lockdep awareness of the primitive type too:
by releasing the complex lock before taking the primitive lock in the release_sock() unlock path. The updated patch below does this - and thus i was able to remove the dropping of the primitive spinlock type.

it is not a problem that the release of the complex lock type does not happen inside the critical section: from the point where we release the complex lock-type _this_ context cannot take any other locks, so there are no dependencies missed.

As you can see, the lock validator can easily cover completely new lock types like sk_lock too, as long as the new lock type has some minimalistic "works like a lock" properties. (such as owner-does-unlock)

later on i'll try the same cleanup for the mutex code too - it should be possible. (that way the implementation of complex lock types can be lock-validator checked too)

Ingo

--------------->
Subject: lockdep, annotate sk_locks
From: Ingo Molnar <[email protected]>

Teach sk_lock semantics to the lock validator. In the softirq path the slock has mutex_trylock()+mutex_unlock() semantics, in the process context sock_lock() case it has mutex_lock()/mutex_unlock() semantics.

Thus we treat sock_owned_by_user() flagged areas as an exclusion area too, not just those areas covered by a held sk_lock.slock.

Effect on non-lockdep kernels: minimal, sk_lock_sock_init() has been turned into an inline function.

---
 include/net/sock.h |   20 +++++------
 net/core/sock.c    |   92 +++++++++++++++++++++++++++++++++++++++++++++++------
 2 files changed, 93 insertions(+), 19 deletions(-)

Index: linux/include/net/sock.h
===================================================================

include/net/sock.h +10 -10
@@ -44,6 +44,7 @@
 #include <linux/timer.h>
 #include <linux/cache.h>
 #include <linux/module.h>
+#include <linux/lockdep.h>
 #include <linux/netdevice.h>
 #include <linux/skbuff.h>      /* struct sk_buff */
 #include <linux/security.h>
@@ -78,18 +79,17 @@ typedef struct {
        spinlock_t              slock;
        struct sock_iocb        *owner;
        wait_queue_head_t       wq;
+       /*
+        * We express the mutex-alike socket_lock semantics
+        * to the lock validator by explicitly managing
+        * the slock as a lock variant (in addition to
+        * the slock itself):
+        */
+#ifdef CONFIG_DEBUG_LOCK_ALLOC
+       struct lockdep_map dep_map;
+#endif
 } socket_lock_t;
 
-extern struct lock_class_key af_family_keys[AF_MAX];
-
-#define sock_lock_init(__sk) \
-do {   spin_lock_init(&((__sk)->sk_lock.slock)); \
-       lockdep_set_class(&(__sk)->sk_lock.slock, \
-                         af_family_keys + (__sk)->sk_family); \
-       (__sk)->sk_lock.owner = NULL; \
-       init_waitqueue_head(&((__sk)->sk_lock.wq)); \
-} while(0)
-
 struct sock;
 struct proto;
 

Index: linux/net/core/sock.c
===================================================================

net/core/sock.c +80 -6
@@ -134,7 +134,40 @@
  * Each address family might have different locking rules, so we have
  * one slock key per address family:
  */
-struct lock_class_key af_family_keys[AF_MAX];
+static struct lock_class_key af_family_keys[AF_MAX];
+static struct lock_class_key af_family_slock_keys[AF_MAX];
+
+#ifdef CONFIG_LOCKDEP
+/*
+ * Make lock validator output more readable:
+ */
+static const char *af_family_key_strings[AF_MAX+1] = {
+  "sk_lock-AF_UNSPEC", "sk_lock-AF_UNIX"     , "sk_lock-AF_INET"     ,
+  "sk_lock-AF_AX25"  , "sk_lock-AF_IPX"      , "sk_lock-AF_APPLETALK",
+  "sk_lock-AF_NETROM", "sk_lock-AF_BRIDGE"   , "sk_lock-AF_ATMPVC"   ,
+  "sk_lock-AF_X25"   , "sk_lock-AF_INET6"    , "sk_lock-AF_ROSE"     ,
+  "sk_lock-AF_DECnet", "sk_lock-AF_NETBEUI"  , "sk_lock-AF_SECURITY" ,
+  "sk_lock-AF_KEY"   , "sk_lock-AF_NETLINK"  , "sk_lock-AF_PACKET"   ,
+  "sk_lock-AF_ASH"   , "sk_lock-AF_ECONET"   , "sk_lock-AF_ATMSVC"   ,
+  "sk_lock-21"       , "sk_lock-AF_SNA"      , "sk_lock-AF_IRDA"     ,
+  "sk_lock-AF_PPPOX" , "sk_lock-AF_WANPIPE"  , "sk_lock-AF_LLC"      ,
+  "sk_lock-27"       , "sk_lock-28"          , "sk_lock-29"          ,
+  "sk_lock-AF_TIPC"  , "sk_lock-AF_BLUETOOTH", "sk_lock-AF_MAX"
+};
+static const char *af_family_slock_key_strings[AF_MAX+1] = {
+  "slock-AF_UNSPEC", "slock-AF_UNIX"     , "slock-AF_INET"     ,
+  "slock-AF_AX25"  , "slock-AF_IPX"      , "slock-AF_APPLETALK",
+  "slock-AF_NETROM", "slock-AF_BRIDGE"   , "slock-AF_ATMPVC"   ,
+  "slock-AF_X25"   , "slock-AF_INET6"    , "slock-AF_ROSE"     ,
+  "slock-AF_DECnet", "slock-AF_NETBEUI"  , "slock-AF_SECURITY" ,
+  "slock-AF_KEY"   , "slock-AF_NETLINK"  , "slock-AF_PACKET"   ,
+  "slock-AF_ASH"   , "slock-AF_ECONET"   , "slock-AF_ATMSVC"   ,
+  "slock-21"       , "slock-AF_SNA"      , "slock-AF_IRDA"     ,
+  "slock-AF_PPPOX" , "slock-AF_WANPIPE"  , "slock-AF_LLC"      ,
+  "slock-27"       , "slock-28"          , "slock-29"          ,
+  "slock-AF_TIPC"  , "slock-AF_BLUETOOTH", "slock-AF_MAX"
+};
+#endif
 
 /*
  * sk_callback_lock locking rules are per-address-family,
@@ -250,9 +283,16 @@ int sk_receive_skb(struct sock *sk, stru
        skb->dev = NULL;
 
        bh_lock_sock(sk);
-       if (!sock_owned_by_user(sk))
+       if (!sock_owned_by_user(sk)) {
+               /*
+                * trylock + unlock semantics:
+                */
+               mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
+
                rc = sk->sk_backlog_rcv(sk, skb);
-       else
+
+               mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
+       } else
                sk_add_backlog(sk, skb);
        bh_unlock_sock(sk);
 out:
@@ -762,6 +802,30 @@ lenout:
        return 0;
 }
 
+/*
+ * Initialize an sk_lock.
+ *
+ * (We also register the sk_lock with the lock validator.)
+ */
+static void inline sock_lock_init(struct sock *sk)
+{
+       spin_lock_init(&sk->sk_lock.slock);
+       lockdep_set_class_and_name(&sk->sk_lock.slock,
+                                  af_family_slock_keys + sk->sk_family,
+                                  af_family_slock_key_strings[sk->sk_family]);
+       sk->sk_lock.owner = NULL;
+       init_waitqueue_head(&sk->sk_lock.wq);
+#ifdef CONFIG_DEBUG_LOCK_ALLOC
+       /*
+        * Make sure we are not reinitializing a held lock:
+        */
+       debug_check_no_locks_freed((void *)&sk->sk_lock, sizeof(sk->sk_lock));
+       lockdep_init_map(&sk->sk_lock.dep_map,
+                        af_family_key_strings[sk->sk_family],
+                        af_family_keys + sk->sk_family);
+#endif
+}
+
 /**
  *     sk_alloc - All socket objects are allocated here
  *     @family: protocol family
@@ -1466,24 +1530,34 @@ void sock_init_data(struct socket *sock,
 void fastcall lock_sock(struct sock *sk)
 {
        might_sleep();
-       spin_lock_bh(&(sk->sk_lock.slock));
+       spin_lock_bh(&sk->sk_lock.slock);
        if (sk->sk_lock.owner)
                __lock_sock(sk);
        sk->sk_lock.owner = (void *)1;
-       spin_unlock_bh(&(sk->sk_lock.slock));
+       spin_unlock(&sk->sk_lock.slock);
+       /*
+        * The sk_lock has mutex_lock() semantics here:
+        */
+       mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
+       local_bh_enable();
 }
 
 EXPORT_SYMBOL(lock_sock);
 
 void fastcall release_sock(struct sock *sk)
 {
-       spin_lock_bh(&(sk->sk_lock.slock));
+       /*
+        * The sk_lock has mutex_unlock() semantics:
+        */
+       mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
+
+       spin_lock_bh(&sk->sk_lock.slock);
        if (sk->sk_backlog.tail)
                __release_sock(sk);

net/core/sock.c void sock_init_data(struct socket *sock, · line 1563

Quoted 10 lines
net/core/sock.c +3 -3
        sk->sk_lock.owner = NULL;
-        if (waitqueue_active(&(sk->sk_lock.wq)))
-               wake_up(&(sk->sk_lock.wq));
-       spin_unlock_bh(&(sk->sk_lock.slock));
+       if (waitqueue_active(&sk->sk_lock.wq))
+               wake_up(&sk->sk_lock.wq);
+       spin_unlock_bh(&sk->sk_lock.slock);
 }
 EXPORT_SYMBOL(release_sock);
 

I cannot get this patch to apply cleanly to 2.6.17-mm4. Since the patch listed in this message covers the same files as your previous lockdep-annotate-slock.patch, I am assuming this is supposed to replace it. I should also still apply lockdep-core-add-set-class-and-name.patch, correct?

patch -p1 -l --dry-run < ../molnar-latest.patch
patching file include/net/sock.h
patching file net/core/sock.c
Hunk #1 FAILED at 134.
Hunk #3 succeeded at 802 with fuzz 2.
1 out of 4 hunks FAILED -- saving rejects to file net/core/sock.c.rej

#13 lore raw

Re: [patch] lockdep, annotate slocks: turn lockdep off for them

Quoted 5 lines · Miles Lane

I cannot get this patch to apply cleanly to 2.6.17-mm4. Since the patch listed in this message covers the same files as your previous lockdep-annotate-slock.patch, I am assuming this is supposed to replace it. I should also still apply lockdep-core-add-set-class-and-name.patch, correct?

correct. My latest combo patch has it all included, it's at:

http://redhat.com/~mingo/lockdep-patches/lockdep-combo-2.6.17-mm4.patch

could you try that one?

Ingo

author
#14 lore raw

Re: [patch] lockdep, annotate slocks: turn lockdep off for them

Okay, I rebuilt my kernel with your combo patch applied. Then, I inserted my US Robotics USR2210 PCMCIA wifi card, ran "pccardutil eject", popped out the card and then inserted a Compaq iPaq wifi card. This triggered the following.

[ INFO: possible circular locking dependency detected ]
-------------------------------------------------------
syslogd/1886 is trying to acquire lock:
(&dev->queue_lock){-+..}, at: [<c11a50b5>] dev_queue_xmit+0x120/0x24b

but task is already holding lock:
(&dev->_xmit_lock){-+..}, at: [<c11a5118>] dev_queue_xmit+0x183/0x24b

which lock already depends on the new lock.

the existing dependency chain (in reverse order) is:

-> #1 (&dev->_xmit_lock){-+..}:

      [<c102d1b6>] lock_acquire+0x60/0x80
      [<c12013ed>] _spin_lock_bh+0x28/0x37
      [<c11b0045>] dev_activate+0xce/0x100
      [<c11a42e7>] dev_open+0x4c/0x64
      [<c11a2e44>] dev_change_flags+0x51/0xf1
      [<c11ab07d>] do_setlink+0x73/0x312
      [<c11aadcc>] rtnetlink_rcv_msg+0x1ae/0x1d1
      [<c11b5c91>] netlink_run_queue+0x69/0x100
      [<c11aabd4>] rtnetlink_rcv+0x29/0x42
      [<c11b6104>] netlink_data_ready+0x12/0x4f
      [<c11b4f51>] netlink_sendskb+0x1f/0x50
      [<c11b5a42>] netlink_unicast+0x1b4/0x1ce
      [<c11b60e5>] netlink_sendmsg+0x259/0x266
      [<c1199c7a>] sock_sendmsg+0xe8/0x103
      [<c119a2a5>] sys_sendmsg+0x14f/0x1aa
      [<c119b5e3>] sys_socketcall+0x16b/0x186
      [<c1002d6d>] sysenter_past_esp+0x56/0x8d

-> #0 (&dev->queue_lock){-+..}:

      [<c102d1b6>] lock_acquire+0x60/0x80
      [<c12013b6>] _spin_lock+0x23/0x32
      [<c11a50b5>] dev_queue_xmit+0x120/0x24b
      [<f948379e>] hostap_data_start_xmit+0x610/0x61a [hostap]
      [<c11a371d>] dev_hard_start_xmit+0x206/0x212
      [<c11a5134>] dev_queue_xmit+0x19f/0x24b
      [<f939c7c9>] mld_sendpack+0x1a0/0x26a [ipv6]
      [<f939d3a3>] mld_ifc_timer_expire+0x1d6/0x1fd [ipv6]
      [<c101db26>] run_timer_softirq+0xf2/0x14a
      [<c101a709>] __do_softirq+0x55/0xb0
      [<c1004a64>] do_softirq+0x58/0xbd

other info that might help us debug this:

4 locks held by syslogd/1886:
#0: (&inode->i_mutex){--..}, at: [<c120028e>] mutex_lock+0x1c/0x1f #1: (&journal->j_state_lock){--..}, at: [<c10aee62>]

journal_stop+0x109/0x1f8 #2:  (&transaction->t_handle_lock){--..}, at:
[<c10aee6c>] journal_stop+0x113/0x1f8
#3:  (&dev->_xmit_lock){-+..}, at: [<c11a5118>] dev_queue_xmit+0x183/0x24b

stack backtrace:
[<c1003502>] show_trace_log_lvl+0x54/0xfd
[<c1003b6a>] show_trace+0xd/0x10
[<c1003c0e>] dump_stack+0x19/0x1b
[<c102c576>] print_circular_bug_tail+0x59/0x64
[<c102cd5c>] __lock_acquire+0x7db/0x970
[<c102d1b6>] lock_acquire+0x60/0x80
[<c12013b6>] _spin_lock+0x23/0x32
[<c11a50b5>] dev_queue_xmit+0x120/0x24b
[<f948379e>] hostap_data_start_xmit+0x610/0x61a [hostap]
[<c11a371d>] dev_hard_start_xmit+0x206/0x212
[<c11a5134>] dev_queue_xmit+0x19f/0x24b
[<f939c7c9>] mld_sendpack+0x1a0/0x26a [ipv6]
[<f939d3a3>] mld_ifc_timer_expire+0x1d6/0x1fd [ipv6]
[<c101db26>] run_timer_softirq+0xf2/0x14a
[<c101a709>] __do_softirq+0x55/0xb0
[<c1004a64>] do_softirq+0x58/0xbd
[<c101a6a8>] irq_exit+0x3f/0x4b
[<c1004b89>] do_IRQ+0xc0/0xcf
[<c1002fd9>] common_interrupt+0x25/0x2c
Quoted 17 lines · Ingo Molnar
Quoted 5 lines · Miles Lane

I cannot get this patch to apply cleanly to 2.6.17-mm4. Since the patch listed in this message covers the same files as your previous lockdep-annotate-slock.patch, I am assuming this is supposed to replace it. I should also still apply lockdep-core-add-set-class-and-name.patch, correct?

correct. My latest combo patch has it all included, it's at:

http://redhat.com/~mingo/lockdep-patches/lockdep-combo-2.6.17-mm4.patch

could you try that one?

Ingo

author
#15 lore raw

Re: [patch] lockdep, annotate slocks: turn lockdep off for them

i just hit this again. The stacktrace is identical except for this bit:

other info that might help us debug this:

2 locks held by usplash_write/2182:
#0: (&tty->atomic_write_lock){--..}, at: [<c11fff5e>]

mutex_lock_interruptible+0x1c/0x1f
#1:  (&dev->_xmit_lock){-+..}, at: [<c11a5118>] dev_queue_xmit+0x183/0x24b
#16 lore raw

Re: [patch] lockdep, annotate slocks: turn lockdep off for them

In reply to Miles Lane
Quoted 4 lines · Miles Lane

Okay, I rebuilt my kernel with your combo patch applied. Then, I inserted my US Robotics USR2210 PCMCIA wifi card, ran "pccardutil eject", popped out the card and then inserted a Compaq iPaq wifi card. This triggered the following.

ok - but the ipv6 one went away, which is good. Will have a look at this new one.

Ingo

#17 lore raw

Re: [patch] lockdep, annotate slocks: turn lockdep off for them

In reply to Miles Lane
Quoted 10 lines · Miles Lane

Okay, I rebuilt my kernel with your combo patch applied. Then, I inserted my US Robotics USR2210 PCMCIA wifi card, ran "pccardutil eject", popped out the card and then inserted a Compaq iPaq wifi card. This triggered the following.

[ INFO: possible circular locking dependency detected ]
-------------------------------------------------------
syslogd/1886 is trying to acquire lock:
(&dev->queue_lock){-+..}, at: [<c11a50b5>] dev_queue_xmit+0x120/0x24b

Quoted 4 lines

but task is already holding lock:
(&dev->_xmit_lock){-+..}, at: [<c11a5118>] dev_queue_xmit+0x183/0x24b

which lock already depends on the new lock.

ok this appears to be hostap playing games... it has 2 network devices for one piece of hardware and one calls the other via the networking layer; there is thankfully a natural ordering between the two, so just making the slave one a separate type ought to make this work.

Can you test the patch below?

---
 drivers/net/wireless/hostap/hostap_hw.c |   10 ++++++++++
 1 file changed, 10 insertions(+)
drivers/net/wireless/hostap/hostap_hw.c +10 -0
Index: linux-2.6.17-mm4/drivers/net/wireless/hostap/hostap_hw.c
===================================================================
@@ -3096,6 +3096,14 @@ static void prism2_clear_set_tim_queue(l
 }
 
 
+/*
+ * HostAP uses two layers of net devices, where the inner
+ * layer gets called all the time from the outer layer.
+ * This is a natural nesting, which needs a split lock type.
+ */
+static struct lock_class_key hostap_netdev_xmit_lock_key;
+
+
 static struct net_device *
 prism2_init_local_data(struct prism2_helper_functions *funcs, int card_idx,
 		       struct device *sdev)
@@ -3260,6 +3268,8 @@ while (0)
 	SET_NETDEV_DEV(dev, sdev);
 	if (ret >= 0)
 		ret = register_netdevice(dev);
+
+	lockdep_set_class(&dev->_xmit_lock, &hostap_netdev_xmit_lock_key);
 	rtnl_unlock();
 	if (ret < 0) {
 		printk(KERN_WARNING "%s: register netdevice failed!\n",
author
#18 lore raw

Re: [patch] lockdep, annotate slocks: turn lockdep off for them

Quoted 48 lines · Arjan van de Ven
Quoted 10 lines · Miles Lane

Okay, I rebuilt my kernel with your combo patch applied. Then, I inserted my US Robotics USR2210 PCMCIA wifi card, ran "pccardutil eject", popped out the card and then inserted a Compaq iPaq wifi card. This triggered the following.

[ INFO: possible circular locking dependency detected ]
-------------------------------------------------------
syslogd/1886 is trying to acquire lock:
(&dev->queue_lock){-+..}, at: [<c11a50b5>] dev_queue_xmit+0x120/0x24b

Quoted 4 lines

but task is already holding lock:
(&dev->_xmit_lock){-+..}, at: [<c11a5118>] dev_queue_xmit+0x183/0x24b

which lock already depends on the new lock.

ok this appears to be hostap playing games... it has 2 network devices for one piece of hardware and one calls the other via the networking layer; there is thankfully a natural ordering between the two, so just making the slave one a separate type ought to make this work.

Can you test the patch below?

---
 drivers/net/wireless/hostap/hostap_hw.c |   10 ++++++++++
 1 file changed, 10 insertions(+)

Index: linux-2.6.17-mm4/drivers/net/wireless/hostap/hostap_hw.c
===================================================================

drivers/net/wireless/hostap/hostap_hw.c +8 -0
@@ -3096,6 +3096,14 @@ static void prism2_clear_set_tim_queue(l
 }
 
 
+/*
+ * HostAP uses two layers of net devices, where the inner
+ * layer gets called all the time from the outer layer.
+ * This is a natural nesting, which needs a split lock type.
+ */
+static struct lock_class_key hostap_netdev_xmit_lock_key;
+
+
 static struct net_device *
 prism2_init_local_data(struct prism2_helper_functions *funcs, int card_idx,
                       struct device *sdev)

drivers/net/wireless/hostap/hostap_hw.c while (0) · line 3275

Quoted 8 lines
drivers/net/wireless/hostap/hostap_hw.c +2 -0
@@ -3260,6 +3268,8 @@ while (0)
        SET_NETDEV_DEV(dev, sdev);
        if (ret >= 0)
                ret = register_netdevice(dev);
+
+       lockdep_set_class(&dev->_xmit_lock, &hostap_netdev_xmit_lock_key);
        rtnl_unlock();
        if (ret < 0) {
                printk(KERN_WARNING "%s: register netdevice failed!\n",

(Sorry for the duplicate message Arjan, I forgot to Reply All the first time.)

After rebuilding with this patch, I still get the lockdep message. Everything is the same except now the "other info" section reads:

other info that might help us debug this:

1 lock held by swapper/0:
#0: (&dev->_xmit_lock){-+..}, at: [<c11a5118>] dev_queue_xmit+0x183/0x24b

#19 lore raw

Re: [patch] lockdep, annotate slocks: turn lockdep off for them

Quoted 13 lines · Miles Lane

drivers/net/wireless/hostap/hostap_hw.c while (0) · line 3275

Quoted 10 lines
drivers/net/wireless/hostap/hostap_hw.c +2 -0
@@ -3260,6 +3268,8 @@ while (0)
       SET_NETDEV_DEV(dev, sdev);
       if (ret >= 0)
               ret = register_netdevice(dev);
+
+       lockdep_set_class(&dev->_xmit_lock, &hostap_netdev_xmit_lock_key);
       rtnl_unlock();
       if (ret < 0) {
               printk(KERN_WARNING "%s: register netdevice failed!\n",

After rebuilding with this patch, I still get the lockdep message. Everything is the same except now the "other info" section reads:

Perhaps the same trick needs to be applied to the queue lock?

Cheers,

Signature · 4 lines
Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
author
#20 lore raw

Re: [patch] lockdep, annotate slocks: turn lockdep off for them

Doh! User error. I did a dry-run to make sure it'd apply and forgot to run again without it. I'll fix and restest.

Sorry,
Miles

author
#21 lore raw

Re: [patch] lockdep, annotate slocks: turn lockdep off for them

Okay, the patch worked! Thanks.

BTW, I noticed this unrelated issue when recompiling:
WARNING: "__stack_chk_fail" [drivers/net/wireless/hostap/hostap_plx.ko] undefined!
WARNING: "__stack_chk_fail" [drivers/net/wireless/hostap/hostap_pci.ko] undefined!
WARNING: "__stack_chk_fail" [drivers/net/wireless/hostap/hostap_cs.ko] undefined!

Cheers,
Miles