linux.git/kernel/irq/spurious.c, branch v2.6.32

genirq: try_one_irq() must be called with irq disabled

2009-11-07T20:44:45+00:00

Prarit reported:
=================================
[ INFO: inconsistent lock state ]
2.6.32-rc5 #1
---------------------------------
inconsistent {IN-HARDIRQ-W} -> {HARDIRQ-ON-W} usage.
swapper/0 [HC0[0]:SC1[1]:HE1:SE0] takes:
 (&irq_desc_lock_class){?.-...}, at: [] try_one_irq+0x32/0x138
{IN-HARDIRQ-W} state was registered at:
 [] __lock_acquire+0x2fc/0xd5d
 [] lock_acquire+0xf3/0x12d
 [] _spin_lock+0x40/0x89
 [] handle_level_irq+0x30/0x105
 [] handle_irq+0x95/0xb7
 [] do_IRQ+0x6a/0xe0
 [] ret_from_intr+0x0/0x16
irq event stamp: 195096
hardirqs last  enabled at (195096): [] _spin_unlock_irq+0x3a/0x5c
hardirqs last disabled at (195095): [] _spin_lock_irq+0x29/0x95
softirqs last  enabled at (195088): [] __do_softirq+0x1c1/0x1ef
softirqs last disabled at (195093): [] call_softirq+0x1c/0x30

other info that might help us debug this:
1 lock held by swapper/0:
 #0:  (kernel/irq/spurious.c:21){+.-...}, at: []
run_timer_softirq+0x1a9/0x315

stack backtrace:
Pid: 0, comm: swapper Not tainted 2.6.32-rc5 #1
Call Trace:
   [] valid_state+0x187/0x1ae
 [] mark_lock+0x129/0x253
 [] __lock_acquire+0x370/0xd5d
 [] lock_acquire+0xf3/0x12d
 [] _spin_lock+0x40/0x89
 [] try_one_irq+0x32/0x138
 [] poll_all_shared_irqs+0x41/0x6d
 [] poll_spurious_irqs+0x1c/0x49
 [] run_timer_softirq+0x239/0x315
 [] __do_softirq+0x102/0x1ef
 [] call_softirq+0x1c/0x30
 [] do_softirq+0x59/0xca
 [] irq_exit+0x58/0xae
 [] smp_apic_timer_interrupt+0x94/0xba
 [] apic_timer_interrupt+0x13/0x20

The reason is that try_one_irq() is called from hardirq context with
interrupts disabled and from softirq context (poll_all_shared_irqs())
with interrupts enabled.

Disable interrupts before calling it from poll_all_shared_irqs().

Reported-and-tested-by: Prarit Bhargava 
Signed-off-by: Yong Zhang 
LKML-Reference: <1257563773-4620-1-git-send-email-yong.zhang0@gmail.com>
Signed-off-by: Thomas Gleixner

irq: Clean up by removing irqfixup MODULE_PARM_DESC()

2009-08-04T15:00:11+00:00

It's wrong (the parm takes no arguments) and compile-time wiped
away anyway (due to !MODULE).

Signed-off-by: Jiri Slaby 
LKML-Reference: <1248991326-26267-1-git-send-email-jirislaby@gmail.com>
Signed-off-by: Ingo Molnar

irq: provide debug_poll_all_shared_irqs() method under CONFIG_DEBUG_SHIRQ

2009-01-16T16:46:49+00:00

Provide a shared interrupt debug facility under CONFIG_DEBUG_SHIRQ:
it uses the existing irqpoll facilities to iterate through all
registered interrupt handlers and call those which can handle shared
IRQ lines.

This can be handy for suspend/resume debugging: if we call this function
early during resume we can trigger crashes in those drivers which have
incorrect assumptions about when exactly their ISRs will be called
during suspend/resume.

Signed-off-by: Ingo Molnar

irq: simplify for_each_irq_desc() usage

2008-12-26T08:48:18+00:00

Impact: cleanup

all for_each_irq_desc() usage point have !desc check.
then its check can move into for_each_irq_desc() macro.

Signed-off-by: KOSAKI Motohiro 
Acked-by: Yinghai Lu 
Signed-off-by: Ingo Molnar

sparse irq_desc[] array: core kernel and x86 changes

2008-12-08T13:31:51+00:00

Impact: new feature

Problem on distro kernels: irq_desc[NR_IRQS] takes megabytes of RAM with
NR_CPUS set to large values. The goal is to be able to scale up to much
larger NR_IRQS value without impacting the (important) common case.

To solve this, we generalize irq_desc[NR_IRQS] to an (optional) array of
irq_desc pointers.

When CONFIG_SPARSE_IRQ=y is used, we use kzalloc_node to get irq_desc,
this also makes the IRQ descriptors NUMA-local (to the site that calls
request_irq()).

This gets rid of the irq_cfg[] static array on x86 as well: irq_cfg now
uses desc->chip_data for x86 to store irq_cfg.

Signed-off-by: Yinghai Lu 
Signed-off-by: Ingo Molnar

genirq: cleanup the sparseirq modifications

2008-10-16T14:53:29+00:00

Signed-off-by: Thomas Gleixner

irq: fix irqpoll && sparseirq

2008-10-16T14:53:10+00:00

Steven Noonan reported a boot hang when using irqpoll and
CONFIG_HAVE_SPARSE_IRQ=y.

The irqpoll loop needs to be updated to not iterate from 1 to nr_irqs
but to iterate via for_each_irq_desc(). (in the former case desc can
be NULL which crashes the box)

Reported-by: Steven Noonan 
Tested-by: Steven Noonan 
Signed-off-by: Ingo Molnar

generic: sparse irqs: use irq_desc() together with dyn_array, instead of irq_desc[]

2008-10-16T14:52:29+00:00

add CONFIG_HAVE_SPARSE_IRQ to for use condensed array.
Get rid of irq_desc[] array assumptions.

Preallocate 32 irq_desc, and irq_desc() will try to get more.

( No change in functionality is expected anywhere, except the odd build
  failure where we missed a code site or where a crossing commit itroduces
  new irq_desc[] usage. )

v2: according to Eric, change get_irq_desc() to irq_desc()

Signed-off-by: Yinghai Lu 
Signed-off-by: Ingo Molnar

irq: introduce nr_irqs

2008-10-16T14:52:05+00:00

at this point nr_irqs is equal NR_IRQS

convert a few easy users from NR_IRQS to dynamic nr_irqs.

v2: according to Eric, we need to take care of arch without generic_hardirqs

Signed-off-by: Yinghai Lu 
Signed-off-by: Ingo Molnar

genirq: enable polling for disabled screaming irqs

2008-07-18T17:21:13+00:00

When we disable a screaming irq we never see it again.  If the irq
line is shared or if the driver half works this is a real pain.  So
periodically poll the handlers for screaming interrupts.

I use a timer instead of the classic irq poll technique of working off
the timer interrupt because when we use the local apic timers
note_interrupt is never called (bug?).  Further on a system with
dynamic ticks the timer interrupt might not even fire unless there is
a timer telling it it needs to.

I forced this case on my test system with an e1000 nic and my ssh
session remained responsive despite the interrupt handler only being
called every 10th of a second.

Signed-off-by: Eric W. Biederman 
Signed-off-by: Ingo Molnar