Messages in this thread Patch in this message | | | Date | Tue, 22 Jul 2008 09:50:00 +0200 | From | Ingo Molnar <> | Subject | Re: [bug, netconsole, SLUB] BUG skbuff_head_cache: Poison overwritten |
| |
* David Miller <davem@davemloft.net> wrote:
> From: Evgeniy Polyakov <johnpol@2ka.mipt.ru> > Date: Tue, 22 Jul 2008 01:24:48 +0400 > > > On Mon, Jul 21, 2008 at 09:21:38PM +0200, Ingo Molnar (mingo@elte.hu) wrote: > > > So it's now a strong likelyhood that this crash is a combination of > > > e1000e+netconsole. > > > > e1000_clean_tx_irq() call looks particulary suspicious: it is called > > without adapter->tx_queue_lock in poll controller (netconsole callback) > > and with that lock in NAPI handler. > > > > Can you check kind of this patch: > > The call even seems pointless, since the caller will call ->poll() > (which is e1000_clean) as the very next action, and that will invoke > e1000_clean_tx_irq() properly. > > I would just delete this call from e1000_netpoll() entirely.
ok, i've added the patch below to tip/out-of-tree.
Overnight test had about 100 successful bootups on this testbox. (until it stopped on a drivers/net/hp.c build error - which is unrelated to this problem)
So testing with netconsole disabled is conclusive enough to implicate netconsole strongly. I've now re-enabled netconsole on the testbox and will continue the test with the fix below. Previously it would crash within 10-40 iterations.
Ingo
-----------------> commit bf89280dea6d97671aa5f75f2591ae7e8e3e6699 Author: Ingo Molnar <mingo@elte.hu> Date: Tue Jul 22 09:44:32 2008 +0200
e1000e: fix e1000_netpoll(), remove extraneous e1000_clean_tx_irq() call Evgeniy Polyakov noticed that drivers/net/e1000e/netdev.c:e1000_netpoll() was calling e1000_clean_tx_irq() without taking the TX lock. David Miller suggested to remove the call altogether: since in this callpah there's periodic calls to ->poll() anyway which will do e1000_clean_tx_irq() and will garbage-collect any finished TX ring descriptors. This might solve the e1000e+netconsole crashes i've been seeing: ============================================================================= BUG skbuff_head_cache: Poison overwritten ----------------------------------------------------------------------------- INFO: 0xf658ae9c-0xf658ae9c. First byte 0x6a instead of 0x6b INFO: Allocated in __alloc_skb+0x2c/0x110 age=0 cpu=0 pid=5098 INFO: Freed in __kfree_skb+0x31/0x80 age=0 cpu=1 pid=4440 INFO: Slab 0xc16cc140 objects=16 used=1 fp=0xf658ae00 flags=0x400000c3 INFO: Object 0xf658ae00 @offset=3584 fp=0xf658af00 Signed-off-by: Ingo Molnar <mingo@elte.hu> --- drivers/net/e1000e/netdev.c | 2 -- 1 files changed, 0 insertions(+), 2 deletions(-)
diff --git a/drivers/net/e1000e/netdev.c b/drivers/net/e1000e/netdev.c index 869544b..9c0f56b 100644 --- a/drivers/net/e1000e/netdev.c +++ b/drivers/net/e1000e/netdev.c @@ -4067,8 +4067,6 @@ static void e1000_netpoll(struct net_device *netdev) disable_irq(adapter->pdev->irq); e1000_intr(adapter->pdev->irq, netdev); - e1000_clean_tx_irq(adapter); - enable_irq(adapter->pdev->irq); } #endif
| |