Hi Roland,
Pretty late indeed, but here are two SRIOV locking related fixes for issues
spotted by lockdep. Basically, as we're late in the cycle, I guess there are
two options here, either push to 3.8 and later to -stable once 3.7 is released,
or push to 3.7 this week. This way or another, would be nice if you can place
them on a branch at your tree which will cause them to spend the night in
linux-next.
Or.
Jack Morgenstein (2):
IB/mlx4: Fix spinlock order to avoid lockdep warnings
NET/mlx4_core: Fix potential deadlock in mlx4_eq_int
drivers/infiniband/hw/mlx4/cm.c | 4 ++--
drivers/net/ethernet/mellanox/mlx4/cmd.c | 9 +++++----
drivers/net/ethernet/mellanox/mlx4/eq.c | 10 ++++++----
3 files changed, 13 insertions(+), 10 deletions(-)
These are the lockdep warnings, initially those that relate to patch #1
and later those relating to patch #2. Personally, I don't see the added value
in placing them in the change-log, so they are brought up here for the purpose
of reporting on the problems.
(1) those that relate to patch #1
======================================================
[ INFO: HARDIRQ-safe -> HARDIRQ-unsafe lock order detected ]
3.7.0-rc6+ #68 Not tainted
------------------------------------------------------
kworker/u:3/1547 [HC0[0]:SC0[0]:HE0:SE1] is trying to acquire:
(&(&dev->sriov.id_map_lock)->rlock){+.+...}, at: [<ffffffffa03d5387>]
schedule_delayed+0x38/0x77 [mlx4_ib]
and this task is already holding:
(&(&dev->sriov.going_down_lock)->rlock){-.-...}, at: [<ffffffffa03d537c>]
schedule_delayed+0x2d/0x77 [mlx4_ib]
which would create a new lock dependency:
(&(&dev->sriov.going_down_lock)->rlock){-.-...} ->
(&(&dev->sriov.id_map_lock)->rlock){+.+...}
but this new dependency connects a HARDIRQ-irq-safe lock:
(&(&dev->sriov.going_down_lock)->rlock){-.-...}
... which became HARDIRQ-irq-safe at:
[<ffffffff81070ea6>] __lock_acquire+0x5e9/0x9ae
[<ffffffff8107135b>] lock_acquire+0xf0/0x116
[<ffffffff813845e8>] _raw_spin_lock_irqsave+0x46/0x58
[<ffffffffa03c8ad9>] mlx4_ib_tunnel_comp_handler+0x22/0x57 [mlx4_ib]
[<ffffffffa03c63ba>] mlx4_ib_cq_comp+0x12/0x14 [mlx4_ib]
[<ffffffffa03e9b22>] mlx4_cq_completion+0x5c/0x62 [mlx4_core]
[<ffffffffa03eb1fb>] mlx4_eq_int+0x89/0x865 [mlx4_core]
[<ffffffffa03eb9e6>] mlx4_msi_x_interrupt+0xf/0x16 [mlx4_core]
[<ffffffff8108b0d7>] handle_irq_event_percpu+0x93/0x1d7
[<ffffffff8108b257>] handle_irq_event+0x3c/0x5c
[<ffffffff8108ddad>] handle_edge_irq+0xcc/0xf3
[<ffffffff810032e7>] handle_irq+0x1f/0x28
[<ffffffff81002a94>] do_IRQ+0x48/0xaf
[<ffffffff81384fef>] ret_from_intr+0x0/0x13
[<ffffffff81008aa9>] cpu_idle+0x6e/0xab
[<ffffffff8137d16a>] start_secondary+0x1af/0x1b3
to a HARDIRQ-irq-unsafe lock:
(&(&dev->sriov.id_map_lock)->rlock){+.+...}
... which became HARDIRQ-irq-unsafe at:
... [<ffffffff81070f1f>] __lock_acquire+0x662/0x9ae
[<ffffffff8107135b>] lock_acquire+0xf0/0x116
[<ffffffff813844e5>] _raw_spin_lock+0x3b/0x4a
[<ffffffffa03d55bd>] mlx4_ib_multiplex_cm_handler+0x128/0x2ff [mlx4_ib]
[<ffffffffa03c900f>] mlx4_ib_multiplex_mad+0x1a4/0x2bd [mlx4_ib]
[<ffffffffa03c9196>] mlx4_ib_tunnel_comp_worker+0x6e/0x140 [mlx4_ib]
[<ffffffff8104604c>] process_one_work+0x2e1/0x498
[<ffffffff8104669d>] worker_thread+0x225/0x35d
[<ffffffff8104c887>] kthread+0xc2/0xca
[<ffffffff8138b96c>] ret_from_fork+0x7c/0xb0
other info that might help us debug this:
Possible interrupt unsafe locking scenario:
CPU0 CPU1
---- ----
lock(&(&dev->sriov.id_map_lock)->rlock);
local_irq_disable();
lock(&(&dev->sriov.going_down_lock)->rlock);
lock(&(&dev->sriov.id_map_lock)->rlock);
<Interrupt>
lock(&(&dev->sriov.going_down_lock)->rlock);
*** DEADLOCK ***
3 locks held by kworker/u:3/1547:
#0: ((name)#2){.+.+.+}, at: [<ffffffff81045fa3>] process_one_work+0x238/0x498
#1: ((&ctx->work)#2){+.+.+.}, at: [<ffffffff81045fa3>]
process_one_work+0x238/0x498
#2: (&(&dev->sriov.going_down_lock)->rlock){-.-...}, at: [<ffffffffa03d537c>]
schedule_delayed+0x2d/0x77 [mlx4_ib]
the dependencies between HARDIRQ-irq-safe lock and the holding lock:
-> (&(&dev->sriov.going_down_lock)->rlock){-.-...} ops: 1222 {
IN-HARDIRQ-W at:
[<ffffffff81070ea6>] __lock_acquire+0x5e9/0x9ae
[<ffffffff8107135b>] lock_acquire+0xf0/0x116
[<ffffffff813845e8>] _raw_spin_lock_irqsave+0x46/0x58
[<ffffffffa03c8ad9>] mlx4_ib_tunnel_comp_handler+0x22/0x57
[mlx4_ib]
[<ffffffffa03c63ba>] mlx4_ib_cq_comp+0x12/0x14 [mlx4_ib]
[<ffffffffa03e9b22>] mlx4_cq_completion+0x5c/0x62
[mlx4_core]
[<ffffffffa03eb1fb>] mlx4_eq_int+0x89/0x865 [mlx4_core]
[<ffffffffa03eb9e6>] mlx4_msi_x_interrupt+0xf/0x16
[mlx4_core]
[<ffffffff8108b0d7>] handle_irq_event_percpu+0x93/0x1d7
[<ffffffff8108b257>] handle_irq_event+0x3c/0x5c
[<ffffffff8108ddad>] handle_edge_irq+0xcc/0xf3
[<ffffffff810032e7>] handle_irq+0x1f/0x28
[<ffffffff81002a94>] do_IRQ+0x48/0xaf
[<ffffffff81384fef>] ret_from_intr+0x0/0x13
[<ffffffff81008aa9>] cpu_idle+0x6e/0xab
[<ffffffff8137d16a>] start_secondary+0x1af/0x1b3
IN-SOFTIRQ-W at:
[<ffffffff81070ec9>] __lock_acquire+0x60c/0x9ae
[<ffffffff8107135b>] lock_acquire+0xf0/0x116
[<ffffffff813845e8>] _raw_spin_lock_irqsave+0x46/0x58
[<ffffffffa03c8ad9>] mlx4_ib_tunnel_comp_handler+0x22/0x57
[mlx4_ib]
[<ffffffffa03c63ba>] mlx4_ib_cq_comp+0x12/0x14 [mlx4_ib]
[<ffffffffa03e9b22>] mlx4_cq_completion+0x5c/0x62
[mlx4_core]
[<ffffffffa03eb1fb>] mlx4_eq_int+0x89/0x865 [mlx4_core]
[<ffffffffa03eb9e6>] mlx4_msi_x_interrupt+0xf/0x16
[mlx4_core]
[<ffffffff8108b0d7>] handle_irq_event_percpu+0x93/0x1d7
[<ffffffff8108b257>] handle_irq_event+0x3c/0x5c
[<ffffffff8108ddad>] handle_edge_irq+0xcc/0xf3
[<ffffffff810032e7>] handle_irq+0x1f/0x28
[<ffffffff81002a94>] do_IRQ+0x48/0xaf
[<ffffffff81384fef>] ret_from_intr+0x0/0x13
[<ffffffff812cd807>] flush_unmaps_timeout+0x2c/0x31
[<ffffffff8103be30>] call_timer_fn+0xb2/0x178
[<ffffffff8103c0ed>] run_timer_softirq+0x1f7/0x24d
[<ffffffff81035d86>] __do_softirq+0x10c/0x1fe
[<ffffffff8138cbcc>] call_softirq+0x1c/0x30
[<ffffffff81003280>] do_softirq+0x38/0x80
[<ffffffff81035a3c>] irq_exit+0x4e/0x83
[<ffffffff8101e8d7>] smp_apic_timer_interrupt+0x86/0x94
[<ffffffff8138c52f>] apic_timer_interrupt+0x6f/0x80
[<ffffffff81008aa9>] cpu_idle+0x6e/0xab
[<ffffffff8137d16a>] start_secondary+0x1af/0x1b3
INITIAL USE at:
[<ffffffff81070f9a>] __lock_acquire+0x6dd/0x9ae
[<ffffffff8107135b>] lock_acquire+0xf0/0x116
[<ffffffff813845e8>] _raw_spin_lock_irqsave+0x46/0x58
[<ffffffffa03caeb9>] do_slave_init+0x164/0x1be [mlx4_ib]
[<ffffffffa03cbd85>] mlx4_ib_add+0x966/0xa43 [mlx4_ib]
[<ffffffffa03eefed>] mlx4_add_device+0x47/0x9a [mlx4_core]
[<ffffffffa03ef127>] mlx4_register_interface+0x5a/0x93
[mlx4_core]
[<ffffffffa03df053>] 0xffffffffa03df053
[<ffffffff810001fa>] do_one_initcall+0x7a/0x12e
[<ffffffff8107beed>] sys_init_module+0x7a/0x1bd
[<ffffffff8138ba12>] system_call_fastpath+0x16/0x1b
}
... key at: [<ffffffffa03da250>] __key.29772+0x0/0xffffffffffffd7cf
[mlx4_ib]
... acquired at:
[<ffffffff8106f0f6>] check_irq_usage+0x5d/0xbe
[<ffffffff8106fe59>] validate_chain+0x890/0xe5a
[<ffffffff810711f7>] __lock_acquire+0x93a/0x9ae
[<ffffffff8107135b>] lock_acquire+0xf0/0x116
[<ffffffff813844e5>] _raw_spin_lock+0x3b/0x4a
[<ffffffffa03d5387>] schedule_delayed+0x38/0x77 [mlx4_ib]
[<ffffffffa03d576f>] mlx4_ib_multiplex_cm_handler+0x2da/0x2ff [mlx4_ib]
[<ffffffffa03c900f>] mlx4_ib_multiplex_mad+0x1a4/0x2bd [mlx4_ib]
[<ffffffffa03c9196>] mlx4_ib_tunnel_comp_worker+0x6e/0x140 [mlx4_ib]
[<ffffffff8104604c>] process_one_work+0x2e1/0x498
[<ffffffff8104669d>] worker_thread+0x225/0x35d
[<ffffffff8104c887>] kthread+0xc2/0xca
[<ffffffff8138b96c>] ret_from_fork+0x7c/0xb0
the dependencies between the lock to be acquired and HARDIRQ-irq-unsafe lock:
-> (&(&dev->sriov.id_map_lock)->rlock){+.+...} ops: 5 {
HARDIRQ-ON-W at:
[<ffffffff81070f1f>] __lock_acquire+0x662/0x9ae
[<ffffffff8107135b>] lock_acquire+0xf0/0x116
[<ffffffff813844e5>] _raw_spin_lock+0x3b/0x4a
[<ffffffffa03d55bd>]
mlx4_ib_multiplex_cm_handler+0x128/0x2ff [mlx4_ib]
[<ffffffffa03c900f>] mlx4_ib_multiplex_mad+0x1a4/0x2bd
[mlx4_ib]
[<ffffffffa03c9196>] mlx4_ib_tunnel_comp_worker+0x6e/0x140
[mlx4_ib]
[<ffffffff8104604c>] process_one_work+0x2e1/0x498
[<ffffffff8104669d>] worker_thread+0x225/0x35d
[<ffffffff8104c887>] kthread+0xc2/0xca
[<ffffffff8138b96c>] ret_from_fork+0x7c/0xb0
SOFTIRQ-ON-W at:
[<ffffffff81070f42>] __lock_acquire+0x685/0x9ae
[<ffffffff8107135b>] lock_acquire+0xf0/0x116
[<ffffffff813844e5>] _raw_spin_lock+0x3b/0x4a
[<ffffffffa03d55bd>]
mlx4_ib_multiplex_cm_handler+0x128/0x2ff [mlx4_ib]
[<ffffffffa03c900f>] mlx4_ib_multiplex_mad+0x1a4/0x2bd
[mlx4_ib]
[<ffffffffa03c9196>] mlx4_ib_tunnel_comp_worker+0x6e/0x140
[mlx4_ib]
[<ffffffff8104604c>] process_one_work+0x2e1/0x498
[<ffffffff8104669d>] worker_thread+0x225/0x35d
[<ffffffff8104c887>] kthread+0xc2/0xca
[<ffffffff8138b96c>] ret_from_fork+0x7c/0xb0
INITIAL USE at:
[<ffffffff81070f9a>] __lock_acquire+0x6dd/0x9ae
[<ffffffff8107135b>] lock_acquire+0xf0/0x116
[<ffffffff813844e5>] _raw_spin_lock+0x3b/0x4a
[<ffffffffa03d55bd>]
mlx4_ib_multiplex_cm_handler+0x128/0x2ff [mlx4_ib]
[<ffffffffa03c900f>] mlx4_ib_multiplex_mad+0x1a4/0x2bd
[mlx4_ib]
[<ffffffffa03c9196>] mlx4_ib_tunnel_comp_worker+0x6e/0x140
[mlx4_ib]
[<ffffffff8104604c>] process_one_work+0x2e1/0x498
[<ffffffff8104669d>] worker_thread+0x225/0x35d
[<ffffffff8104c887>] kthread+0xc2/0xca
[<ffffffff8138b96c>] ret_from_fork+0x7c/0xb0
}
... key at: [<ffffffffa03da378>] __key.27645+0x0/0xffffffffffffd6a7
[mlx4_ib]
... acquired at:
[<ffffffff8106f0f6>] check_irq_usage+0x5d/0xbe
[<ffffffff8106fe59>] validate_chain+0x890/0xe5a
[<ffffffff810711f7>] __lock_acquire+0x93a/0x9ae
[<ffffffff8107135b>] lock_acquire+0xf0/0x116
[<ffffffff813844e5>] _raw_spin_lock+0x3b/0x4a
[<ffffffffa03d5387>] schedule_delayed+0x38/0x77 [mlx4_ib]
[<ffffffffa03d576f>] mlx4_ib_multiplex_cm_handler+0x2da/0x2ff [mlx4_ib]
[<ffffffffa03c900f>] mlx4_ib_multiplex_mad+0x1a4/0x2bd [mlx4_ib]
[<ffffffffa03c9196>] mlx4_ib_tunnel_comp_worker+0x6e/0x140 [mlx4_ib]
[<ffffffff8104604c>] process_one_work+0x2e1/0x498
[<ffffffff8104669d>] worker_thread+0x225/0x35d
[<ffffffff8104c887>] kthread+0xc2/0xca
[<ffffffff8138b96c>] ret_from_fork+0x7c/0xb0
stack backtrace:
Pid: 1547, comm: kworker/u:3 Not tainted 3.7.0-rc6+ #68
Call Trace:
[<ffffffff81030325>] ? console_unlock+0x358/0x37e
[<ffffffff8106f085>] check_usage+0x525/0x539
[<ffffffff8106f0f6>] check_irq_usage+0x5d/0xbe
[<ffffffff8106fe59>] validate_chain+0x890/0xe5a
[<ffffffff8106e463>] ? trace_hardirqs_on+0xd/0xf
[<ffffffff810711f7>] __lock_acquire+0x93a/0x9ae
[<ffffffff8107135b>] lock_acquire+0xf0/0x116
[<ffffffffa03d5387>] ? schedule_delayed+0x38/0x77 [mlx4_ib]
[<ffffffff813844e5>] _raw_spin_lock+0x3b/0x4a
[<ffffffffa03d5387>] ? schedule_delayed+0x38/0x77 [mlx4_ib]
[<ffffffffa03d5387>] schedule_delayed+0x38/0x77 [mlx4_ib]
[<ffffffffa03d576f>] mlx4_ib_multiplex_cm_handler+0x2da/0x2ff [mlx4_ib]
[<ffffffffa03c900f>] mlx4_ib_multiplex_mad+0x1a4/0x2bd [mlx4_ib]
[<ffffffff8106e41f>] ? trace_hardirqs_on_caller+0x11e/0x155
[<ffffffffa03c6c0e>] ? mlx4_ib_poll_cq+0x609/0x62d [mlx4_ib]
[<ffffffffa03c9196>] mlx4_ib_tunnel_comp_worker+0x6e/0x140 [mlx4_ib]
[<ffffffff8104604c>] process_one_work+0x2e1/0x498
[<ffffffff81045fa3>] ? process_one_work+0x238/0x498
[<ffffffff81384c23>] ? _raw_spin_unlock_irq+0x2b/0x40
[<ffffffffa03c9128>] ? mlx4_ib_multiplex_mad+0x2bd/0x2bd [mlx4_ib]
[<ffffffff8104669d>] worker_thread+0x225/0x35d
[<ffffffff81046478>] ? manage_workers+0x275/0x275
[<ffffffff8104c887>] kthread+0xc2/0xca
[<ffffffff8104c7c5>] ? __init_kthread_worker+0x56/0x56
[<ffffffff8138b96c>] ret_from_fork+0x7c/0xb0
[<ffffffff8104c7c5>] ? __init_kthread_worker+0x56/0x56
[sched_delayed] sched: RT throttling activated
(2) those that relate to patch #2
inconsistent {HARDIRQ-ON-W} -> {IN-HARDIRQ-W} usage.
swapper/7/0 [HC1[1]:SC0[0]:HE0:SE1] takes:
(&(&priv->mfunc.master.slave_state_lock)->rlock){?.+...}, at:
[<ffffffffa042e786>] mlx4_eq_int+0x5e7/0x84b [mlx4_core]
{HARDIRQ-ON-W} state was registered at:
[<ffffffff8107143f>] __lock_acquire+0x662/0x9ae
[<ffffffff8107187b>] lock_acquire+0xf0/0x116
[<ffffffff813848ad>] _raw_spin_lock+0x3b/0x4a
[<ffffffffa042c1d4>] mlx4_master_comm_channel+0x3e3/0x4d5 [mlx4_core]
[<ffffffff81045f9c>] process_one_work+0x2e1/0x498
[<ffffffff810465ed>] worker_thread+0x225/0x35d
[<ffffffff8104c7cb>] kthread+0xc2/0xca
[<ffffffff8138bdec>] ret_from_fork+0x7c/0xb0
irq event stamp: 1140398
hardirqs last enabled at (1140395): [<ffffffff810086db>] mwait_idle+0x133/0x208
hardirqs last disabled at (1140396): [<ffffffff813853aa>]
common_interrupt+0x6a/0x6f
softirqs last enabled at (1140398): [<ffffffff810359f7>]
_local_bh_enable+0xe/0x10
softirqs last disabled at (1140397): [<ffffffff81035bc0>] irq_enter+0x44/0x76
other info that might help us debug this:
Possible unsafe locking scenario:
CPU0
----
lock(&(&priv->mfunc.master.slave_state_lock)->rlock);
<Interrupt>
lock(&(&priv->mfunc.master.slave_state_lock)->rlock);
*** DEADLOCK ***
no locks held by swapper/7/0.
stack backtrace:
Pid: 0, comm: swapper/7 Not tainted 3.7.0-rc1+ #2
Call Trace:
<IRQ> [<ffffffff8103021a>] ? console_unlock+0x2d5/0x37e
[<ffffffff8106e086>] print_usage_bug+0x297/0x2a8
[<ffffffff8100cc92>] ? save_stack_trace+0x2a/0x47
[<ffffffff8106eee8>] ? print_irq_inversion_bug+0x1d7/0x1d7
[<ffffffff8106e38a>] mark_lock+0x2f3/0x52b
[<ffffffff810713c6>] __lock_acquire+0x5e9/0x9ae
[<ffffffff8107187b>] lock_acquire+0xf0/0x116
[<ffffffffa042e786>] ? mlx4_eq_int+0x5e7/0x84b [mlx4_core]
[<ffffffff813848ad>] _raw_spin_lock+0x3b/0x4a
[<ffffffffa042e786>] ? mlx4_eq_int+0x5e7/0x84b [mlx4_core]
[<ffffffff8106e925>] ? trace_hardirqs_on_caller+0x104/0x155
[<ffffffffa042e786>] mlx4_eq_int+0x5e7/0x84b [mlx4_core]
[<ffffffff8106e983>] ? trace_hardirqs_on+0xd/0xf
[<ffffffffa042e9f9>] mlx4_msi_x_interrupt+0xf/0x16 [mlx4_core]
[<ffffffff8108b5f3>] handle_irq_event_percpu+0x93/0x1d7
[<ffffffff8108b773>] handle_irq_event+0x3c/0x5c
[<ffffffff8108e2c9>] handle_edge_irq+0xcc/0xf3
[<ffffffff81003407>] handle_irq+0x1f/0x28
[<ffffffff81002bb4>] do_IRQ+0x48/0xaf
[<ffffffff813853af>] common_interrupt+0x6f/0x6f
<EOI> [<ffffffff810086e4>] ? mwait_idle+0x13c/0x208
[<ffffffff810086db>] ? mwait_idle+0x133/0x208
[<ffffffff81008bc9>] cpu_idle+0x6e/0xab
[<ffffffff8137d512>] start_secondary+0x1af/0x1b3
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
