Re: [mmotm 2008-12-30-16-05] Warning at MPT driver initialization.

["Eric Paris" <eparis@parisplace.org>]

[-- Attachment #1: Type: text/plain, Size: 3363 bytes --]

On Wed, Dec 31, 2008 at 9:26 PM, Andrew Morton
<akpm@linux-foundation.org> wrote:
> On Thu, 1 Jan 2009 10:51:12 +0900 Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> wrote:
>
>> Hello.
>>
>> I got below warning.
>>
>> Regards.
>> ----------------------------------------
>> Fusion MPT base driver 3.04.07
>> Copyright (c) 1999-2008 LSI Corporation
>> Fusion MPT SPI Host driver 3.04.07
>> ACPI: PCI Interrupt Link [LNKB] enabled at IRQ 11
>> PCI: setting IRQ 11 as level-triggered
>> mptspi 0000:00:10.0: PCI INT A -> Link[LNKB] -> GSI 11 (level, low) -> IRQ 11
>> mptbase: ioc0: Initiating bringup
>> ioc0: LSI53C1030 B0: Capabilities={Initiator}
>> ODEBUG: object is on stack, but not annotated
>> ------------[ cut here ]------------
>> WARNING: at lib/debugobjects.c:253 __debug_object_init+0x1f3/0x276()
>> Hardware name: VMware Virtual Platform
>> Modules linked in: mptspi(+) mptscsih mptbase scsi_transport_spi ext3 jbd mbcache
>> Pid: 540, comm: insmod Not tainted 2.6.28-mm1 #2
>> Call Trace:
>>  [<c042c51c>] warn_slowpath+0x74/0x8a
>>  [<c0469600>] ? start_critical_timing+0x96/0xb7
>>  [<c060c8ea>] ? _spin_unlock_irqrestore+0x2f/0x3c
>>  [<c0446fad>] ? trace_hardirqs_off_caller+0x18/0xaf
>>  [<c044704f>] ? trace_hardirqs_off+0xb/0xd
>>  [<c060c8ea>] ? _spin_unlock_irqrestore+0x2f/0x3c
>>  [<c042cb84>] ? release_console_sem+0x1a5/0x1ad
>>  [<c05013e6>] __debug_object_init+0x1f3/0x276
>>  [<c0501494>] debug_object_init+0x13/0x17
>>  [<c0433c56>] init_timer+0x10/0x1a
>>  [<e08e5b54>] mpt_config+0x1c1/0x2b7 [mptbase]
>>  [<e08e3b82>] ? kmalloc+0x8/0xa [mptbase]
>>  [<e08e3b82>] ? kmalloc+0x8/0xa [mptbase]
>>  [<e08e6fa2>] mpt_do_ioc_recovery+0x950/0x1212 [mptbase]
>>  [<c04496c2>] ? __lock_acquire+0xa69/0xacc
>>  [<c060c8f1>] ? _spin_unlock_irqrestore+0x36/0x3c
>>  [<c060c3af>] ? _spin_unlock_irq+0x22/0x26
>>  [<c04f2d8b>] ? string+0x2b/0x76
>>  [<c04f310e>] ? vsnprintf+0x338/0x7b3
>>  [<c04496c2>] ? __lock_acquire+0xa69/0xacc
>>  [<c060c8ea>] ? _spin_unlock_irqrestore+0x2f/0x3c
>>  [<c04496c2>] ? __lock_acquire+0xa69/0xacc
>>  [<c044897d>] ? debug_check_no_locks_freed+0xeb/0x105
>>  [<c060c8f1>] ? _spin_unlock_irqrestore+0x36/0x3c
>>  [<c04488bc>] ? debug_check_no_locks_freed+0x2a/0x105
>>  [<c0446b8c>] ? lock_release_holdtime+0x43/0x48
>>  [<c043f742>] ? up_read+0x16/0x29
>>  [<c05076f8>] ? pci_get_slot+0x66/0x72
>>  [<e08e89ca>] mpt_attach+0x881/0x9b1 [mptbase]
>>  [<e091c8e5>] mptspi_probe+0x11/0x354 [mptspi]
>
> Well it's the debugobjects stuff complaining about
> init_timer(&pCfg->timer) in mpt_config().
>
> I spent a minute trying to work out what the heck
> debug_object_is_on_stack() is trying to tell me, but it seems that code
> was designed to only be used by Thomas, so let's cc him and ask.
>
> Thomas, could we have some nice code comments please, so that random
> kernel developers don't need to go and reverse engineer the
> debugobjects design before they can work out what they did wrong?

Noticing that every caller of mpt_config has its CONFIGPARMS struct
declared on the stack and thus the &pCfg->timer is always on the stack
I changed init_timer() to init_timer_on_stack() and it seems to have
shut up.....

Admittedly I don't know either the debugobject code or the mpt code,
but when the timer is on the stack and we have an "init when the timer
is on the stack" it seemed like an easy enough fix.

[-- Attachment #2: tmp.patch --]
[-- Type: application/octet-stream, Size: 561 bytes --]

 drivers/message/fusion/mptbase.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/drivers/message/fusion/mptbase.c b/drivers/message/fusion/mptbase.c
index c4e8b9a..e188cc9 100644
--- a/drivers/message/fusion/mptbase.c
+++ b/drivers/message/fusion/mptbase.c
@@ -5902,7 +5902,7 @@ mpt_config(MPT_ADAPTER *ioc, CONFIGPARMS *pCfg)
 
 	/* Initalize the timer
 	 */
-	init_timer(&pCfg->timer);
+	init_timer_on_stack(&pCfg->timer);
 	pCfg->timer.data = (unsigned long) ioc;
 	pCfg->timer.function = mpt_timer_expired;
 	pCfg->wait_done = 0;