diff mbox

[1/1] net: fec: fix miss init spinlock

Message ID 1359708986-23634-1-git-send-email-Frank.Li@freescale.com (mailing list archive)
State New, archived
Headers show

Commit Message

Frank Li Feb. 1, 2013, 8:56 a.m. UTC
BUG: spinlock bad magic on CPU#1, swapper/0/1
lock: 0xbfae0f8c, .magic: 00000000, .owner: <none>/-1, .owner_cpu: 0
Backtrace:
 [<80011d54>] (dump_backtrace+0x0/0x10c) from [<804e7800>] (dump_stack+0x18/0x1c)
 r6:bfae0000 r5:bfae0f8c r4:00000000 r3:806c1310
 [<804e77e8>] (dump_stack+0x0/0x1c) from [<804e9f20>] (spin_dump+0x80/0x94)
 [<804e9ea0>] (spin_dump+0x0/0x94) from [<804e9f60>] (spin_bug+0x2c/0x30)
 r5:805f6f8c r4:bfae0f8c
 [<804e9f34>] (spin_bug+0x0/0x30) from [<80257984>] (do_raw_spin_lock+0x170/0x1b0                                         )
 r5:806b4950 r4:bfae0f8c
 [<80257814>] (do_raw_spin_lock+0x0/0x1b0) from [<804ed15c>] (_raw_spin_lock_irqs                                         ave+0x18/0x20)
 [<804ed144>] (_raw_spin_lock_irqsave+0x0/0x20) from [<8033c694>] (fec_ptp_start_                                         cyclecounter+0x3c/0x120)
 r4:bfae0f8c r3:00000002
 [<8033c658>] (fec_ptp_start_cyclecounter+0x0/0x120) from [<80339e08>] (fec_resta                                         rt+0x56c/0x5f8)
 r8:00000000 r7:806e6f48 r6:00000112 r5:806b4950 r4:bfae0000
 [<8033989c>] (fec_restart+0x0/0x5f8) from [<8033b9e4>] (fec_probe+0x508/0xa48)

Signed-off-by: Frank Li <Frank.Li@freescale.com>
---
 drivers/net/ethernet/freescale/fec.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

Comments

David Miller Feb. 3, 2013, 4:08 a.m. UTC | #1
From: Frank Li <Frank.Li@freescale.com>
Date: Fri, 1 Feb 2013 16:56:26 +0800

> @@ -1607,6 +1607,7 @@ static int fec_enet_init(struct net_device *ndev)
>  	}
>  
>  	spin_lock_init(&fep->hw_lock);
> +	spin_lock_init(&fep->tmreg_lock);

This breaks the build, tmreg_lock is only present in certain
configurations.
Zhi Li Feb. 4, 2013, 2:22 a.m. UTC | #2
2013/2/3 David Miller <davem@davemloft.net>:
> From: Frank Li <Frank.Li@freescale.com>
> Date: Fri, 1 Feb 2013 16:56:26 +0800
>
>> @@ -1607,6 +1607,7 @@ static int fec_enet_init(struct net_device *ndev)
>>       }
>>
>>       spin_lock_init(&fep->hw_lock);
>> +     spin_lock_init(&fep->tmreg_lock);
>
> This breaks the build, tmreg_lock is only present in certain
> configurations.

No, FEC have changed to check dramatically instead of static config.
You can look fec.h.  tmreg_lock is always defined.

        struct  napi_struct napi;

        struct ptp_clock *ptp_clock;
        struct ptp_clock_info ptp_caps;
        unsigned long last_overflow_check;
        spinlock_t tmreg_lock;
        struct cyclecounter cc;
        struct timecounter tc;
        int rx_hwtstamp_filter;
        u32 base_incval;
        u32 cycle_speed;
        int hwts_rx_en;
        int hwts_tx_en;
        struct timer_list time_keep;

best regards
Frank Li
David Miller Feb. 4, 2013, 5:24 a.m. UTC | #3
From: Frank Li <lznuaa@gmail.com>
Date: Mon, 4 Feb 2013 10:22:23 +0800

> 2013/2/3 David Miller <davem@davemloft.net>:
>> From: Frank Li <Frank.Li@freescale.com>
>> Date: Fri, 1 Feb 2013 16:56:26 +0800
>>
>>> @@ -1607,6 +1607,7 @@ static int fec_enet_init(struct net_device *ndev)
>>>       }
>>>
>>>       spin_lock_init(&fep->hw_lock);
>>> +     spin_lock_init(&fep->tmreg_lock);
>>
>> This breaks the build, tmreg_lock is only present in certain
>> configurations.
> 
> No, FEC have changed to check dramatically instead of static config.
> You can look fec.h.  tmreg_lock is always defined.

Not in the 'net' tree or you don't want this bug fixed there at all?
Zhi Li Feb. 4, 2013, 6:31 a.m. UTC | #4
>>> This breaks the build, tmreg_lock is only present in certain
>>> configurations.
>>
>> No, FEC have changed to check dramatically instead of static config.
>> You can look fec.h.  tmreg_lock is always defined.
>
> Not in the 'net' tree or you don't want this bug fixed there at all?

Sorry, I forget mask tree.
This patch is for net-next.

best regards
Frank Li
David Miller Feb. 4, 2013, 8:03 p.m. UTC | #5
From: Frank Li <lznuaa@gmail.com>
Date: Mon, 4 Feb 2013 14:31:52 +0800

>>>> This breaks the build, tmreg_lock is only present in certain
>>>> configurations.
>>>
>>> No, FEC have changed to check dramatically instead of static config.
>>> You can look fec.h.  tmreg_lock is always defined.
>>
>> Not in the 'net' tree or you don't want this bug fixed there at all?
> 
> Sorry, I forget mask tree.
> This patch is for net-next.

Ok, applied to net-next.
Zhi Li Feb. 5, 2013, 2:40 a.m. UTC | #6
2013/2/5 Dmitry Eremin-Solenikov <dbaryshkov@gmail.com>:
> On Fri, 01 Feb 2013 16:56:26 +0800, Frank Li wrote:
>
>> BUG: spinlock bad magic on CPU#1, swapper/0/1 lock: 0xbfae0f8c, .magic:
>> 00000000, .owner: <none>/-1, .owner_cpu: 0 Backtrace:
>>  [<80011d54>] (dump_backtrace+0x0/0x10c) from [<804e7800>]
>>  (dump_stack+0x18/0x1c)
>>  r6:bfae0000 r5:bfae0f8c r4:00000000 r3:806c1310 [<804e77e8>]
>>  (dump_stack+0x0/0x1c) from [<804e9f20>] (spin_dump+0x80/0x94)
>>  [<804e9ea0>] (spin_dump+0x0/0x94) from [<804e9f60>]
>>  (spin_bug+0x2c/0x30)
>>  r5:805f6f8c r4:bfae0f8c [<804e9f34>] (spin_bug+0x0/0x30) from
>>  [<80257984>] (do_raw_spin_lock+0x170/0x1b0
>>             )
>>  r5:806b4950 r4:bfae0f8c [<80257814>] (do_raw_spin_lock+0x0/0x1b0) from
>>  [<804ed15c>] (_raw_spin_lock_irqs
>>    ave+0x18/0x20) [<804ed144>] (_raw_spin_lock_irqsave+0x0/0x20) from
>>  [<8033c694>] (fec_ptp_start_
>>  cyclecounter+0x3c/0x120)
>>  r4:bfae0f8c r3:00000002 [<8033c658>]
>>  (fec_ptp_start_cyclecounter+0x0/0x120) from [<80339e08>] (fec_resta
>>                                      rt+0x56c/0x5f8) r8:00000000
>>  r7:806e6f48 r6:00000112 r5:806b4950 r4:bfae0000 [<8033989c>]
>>  (fec_restart+0x0/0x5f8) from [<8033b9e4>] (fec_probe+0x508/0xa48)
>>
>> Signed-off-by: Frank Li <Frank.Li@freescale.com>
>> ---
>>  drivers/net/ethernet/freescale/fec.c |    1 +
>>  1 files changed, 1 insertions(+), 0 deletions(-)
>>
>> diff --git a/drivers/net/ethernet/freescale/fec.c
>> b/drivers/net/ethernet/freescale/fec.c index 69b16b9..f900ae4 100644 ---
>> a/drivers/net/ethernet/freescale/fec.c +++
>> b/drivers/net/ethernet/freescale/fec.c @@ -1607,6 +1607,7 @@ static int
>> fec_enet_init(struct net_device *ndev)
>>       }
>>
>>       spin_lock_init(&fep->hw_lock);
>> +     spin_lock_init(&fep->tmreg_lock);
>>
>>       fep->netdev = ndev;
>
> Please excuse me, if I'm wrong. I would consider this as a very wrong
> solution:
>
> First, then the spin lock is initialised twice (first time in
> fec_enet_init(), second time in fec_ptp_init()).
>
> Then this patch actually hides the fact that PTP part of fec
> is accessed from fec_ptp_start_cyclecounter() way before
> than fec_ptp_init() is called to initialize PTP).
>
> In my opinion the right patch should either move fec_ptp_init() call
> before fec_restart(), or make fec_restart intelelctual about calling
> fec_ptp_init()/fec_ptp_startcyclecounter() at proper time.

Yes, you are right.

>
> --
> With best wishes
> Dmitry
>
>
> --
> To unsubscribe from this list: send the line "unsubscribe netdev" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
diff mbox

Patch

diff --git a/drivers/net/ethernet/freescale/fec.c b/drivers/net/ethernet/freescale/fec.c
index 69b16b9..f900ae4 100644
--- a/drivers/net/ethernet/freescale/fec.c
+++ b/drivers/net/ethernet/freescale/fec.c
@@ -1607,6 +1607,7 @@  static int fec_enet_init(struct net_device *ndev)
 	}
 
 	spin_lock_init(&fep->hw_lock);
+	spin_lock_init(&fep->tmreg_lock);
 
 	fep->netdev = ndev;