tree ff205a42ca1598a2038352eb2159115969b778d1
parent bccc2f7b4c1a7fd3d51e34f9dc3397312afb030b
author Mike Galbraith <efault@gmx.de> 1278491341 +0200
committer Ingo Molnar <mingo@elte.hu> 1315905132 +0200

locking, drivers/dca: Annotate dca_lock as raw

The dca_lock can be taken in atomic context:

[   25.607536] Call Trace:
[   25.607557]  [<ffffffff820078a1>] try_stack_unwind+0x151/0x1a0
[   25.607566]  [<ffffffff820062c2>] dump_trace+0x92/0x370
[   25.607573]  [<ffffffff8200731c>] show_trace_log_lvl+0x5c/0x80
[   25.607578]  [<ffffffff82007355>] show_trace+0x15/0x20
[   25.607587]  [<ffffffff823f4588>] dump_stack+0x77/0x8f
[   25.607595]  [<ffffffff82043f2a>] __might_sleep+0x11a/0x130
[   25.607602]  [<ffffffff823f7b93>] rt_spin_lock+0x83/0x90
[   25.607611]  [<ffffffffa0209138>] dca_common_get_tag+0x28/0x80 [dca]
[   25.607622]  [<ffffffffa02091c8>] dca3_get_tag+0x18/0x20 [dca]
[   25.607634]  [<ffffffffa0244e71>] igb_update_dca+0xb1/0x1d0 [igb]
[   25.607649]  [<ffffffffa0244ff5>] igb_setup_dca+0x65/0x80 [igb]
[   25.607663]  [<ffffffffa02535a6>] igb_probe+0x946/0xe4d [igb]
[   25.607678]  [<ffffffff82247517>] local_pci_probe+0x17/0x20
[   25.607686]  [<ffffffff82248661>] pci_device_probe+0x121/0x130
[   25.607699]  [<ffffffff822e4832>] driver_probe_device+0xd2/0x2e0
[   25.607707]  [<ffffffff822e4adb>] __driver_attach+0x9b/0xa0
[   25.607714]  [<ffffffff822e3d1b>] bus_for_each_dev+0x6b/0xa0
[   25.607720]  [<ffffffff822e4591>] driver_attach+0x21/0x30
[   25.607727]  [<ffffffff822e3425>] bus_add_driver+0x1e5/0x350
[   25.607734]  [<ffffffff822e4e41>] driver_register+0x81/0x160
[   25.607742]  [<ffffffff8224890f>] __pci_register_driver+0x6f/0xf0
[   25.607752]  [<ffffffffa011505b>] igb_init_module+0x5b/0x5d [igb]
[   25.607769]  [<ffffffff820001dd>] do_one_initcall+0x3d/0x1a0
[   25.607778]  [<ffffffff820961f6>] sys_init_module+0xe6/0x270
[   25.607786]  [<ffffffff82003232>] system_call_fastpath+0x16/0x1b
[   25.607794]  [<00007f84d6783f4a>] 0x7f84d6783f4a

and thus must not be preempted on -rt.

In mainline this change documents the low level nature of
the lock - otherwise there's no functional difference. Lockdep
and Sparse checking will work as usual.

Signed-off-by: Mike Galbraith <efault@gmx.de>
[ Fixed the domain allocation which was calling kzalloc from the irq disabled section ]
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>