Re: prism54 vs netfilter problem? Badness in local_bh_enable at kernel/softirq.c:140

[Patrick McHardy <kaber@trash.net>]

[-- Attachment #1: Type: text/plain, Size: 2375 bytes --]

Daniel Drake wrote:
> Hi,
> 
> A gentoo user reported these messages on boot:
> 
> eth0: resetting device...
> eth0: uploading firmware...
> eth0: firmware version: 1.0.4.3
> eth0: firmware upload complete
> eth0: interface reset complete
> spurious 8259A interrupt: IRQ7.
> eth0: islpci_close ()
> eth0: resetting device...
> eth0: uploading firmware...
> eth0: firmware version: 1.0.4.3
> eth0: firmware upload complete
> eth0: interface reset complete
> Badness in local_bh_enable at kernel/softirq.c:140
>  [<c0125f9f>] local_bh_enable+0x8f/0xa0
>  [<c03a71b9>] destroy_conntrack+0xd9/0xf0
>  [<c0357d8a>] __kfree_skb+0x7a/0xf0
>  [<c79181fa>] islpci_eth_transmit+0x15a/0x380 [prism54]
>  [<c035d692>] dev_queue_xmit_nit+0xd2/0x140
>  [<c036b04e>] qdisc_restart+0x15e/0x200
>  [<c035dabf>] dev_queue_xmit+0x1cf/0x270
>  [<c0363b23>] neigh_resolve_output+0xd3/0x1d0
>  [<c03636f6>] neigh_update+0x2a6/0x350
>  [<c0397963>] arp_process+0x1c3/0x570
>  [<c036902e>] nf_hook_slow+0x7e/0x140
>  [<c03977a0>] arp_process+0x0/0x570
>  [<c0397e12>] arp_rcv+0x102/0x180
>  [<c03977a0>] arp_process+0x0/0x570
>  [<c035e0a8>] netif_receive_skb+0x148/0x1d0
>  [<c035e1c9>] process_backlog+0x99/0x130
>  [<c035e2e1>] net_rx_action+0x81/0x110
>  [<c0125eb6>] __do_softirq+0xd6/0xf0
>  [<c0125f05>] do_softirq+0x35/0x40
>  [<c0125ff5>] irq_exit+0x45/0x50
>  [<c0105aee>] do_IRQ+0x1e/0x30
>  [<c0103d72>] common_interrupt+0x1a/0x20
>  [<c0277a26>] acpi_processor_idle+0x126/0x270
>  [<c0101110>] cpu_idle+0x70/0x80
>  [<c04f6959>] start_kernel+0x169/0x190
>  [<c04f6370>] unknown_bootoption+0x0/0x1d0
> 
> This isn't a reproducible problem and apparently everything worked fine
> even after these messages had appeared. Just thought I should point out
> that something isn't quite right somewhere.
> 
> The Gentoo bug report is here:
> 
>     http://bugs.gentoo.org/show_bug.cgi?id=103803
> 
> Something similar has been reported previously:
> 
>     http://prism54.org/pipermail/prism54-users/2003-November/000083.html
>     http://prism54.org/pipermail/prism54-users/2003-November/000047.html
> 
> Hope this is useful.

The dev_kfree_skb in islpci_eth_transmit happens while irqs are still
disabled, so either dev_kfree_skb_irq needs to be used or the skb
needs to be freed after irqs have been enabled again. This patch
should fix it.

Signed-off-by: Patrick McHardy <kaber@trash.net>

[-- Attachment #2: x --]
[-- Type: text/plain, Size: 512 bytes --]

diff --git a/drivers/net/wireless/prism54/islpci_eth.c b/drivers/net/wireless/prism54/islpci_eth.c
--- a/drivers/net/wireless/prism54/islpci_eth.c
+++ b/drivers/net/wireless/prism54/islpci_eth.c
@@ -246,12 +246,10 @@ islpci_eth_transmit(struct sk_buff *skb,
 	return 0;
 
       drop_free:
-	/* free the skbuf structure before aborting */
-	dev_kfree_skb(skb);
-	skb = NULL;
-
 	priv->statistics.tx_dropped++;
 	spin_unlock_irqrestore(&priv->slock, flags);
+	dev_kfree_skb(skb);
+	skb = NULL;
 	return err;
 }