On Fri, Jan 27, 2006 at 11:22:42PM +1100, Herbert Xu wrote:
> OK, although we can't rule out sky2/netfilter from the enquiry, I've
> identified two bugs in ppp/pppoe that may be responsible for what you
> are seeing. So please try the following patch and let us know if the
> problem still exists (or deteriorates/improves).
Borrowing this thread for a related problem, I'm getting lots of those on a
bridge device (this one running skge, rmmod skge; modprobe sk98lin actually
seemed to do it too, I've disabled rx checksums with ethtool for now).
Kernel is a 2.6.15.1-ish Fedora one.
skge:
bridge-cd: hw csum failure.
[<c02c8fd8>] __skb_checksum_complete+0x56/0x5c [<f8a5f833>]
icmp_error+0xbf/0x1af [ip_conntrack]
[<c012073a>] __wake_up+0x32/0x43 [<f8a5f774>] icmp_error+0x0/0x1af
[ip_conntrack]
[<f8a5cb57>] ip_conntrack_in+0x95/0x2d6 [ip_conntrack] [<c01393c7>]
__wake_up_bit+0x2e/0x33
[<c016b6d9>] end_buffer_async_write+0xbf/0x12a [<c02e26fd>]
nf_iterate+0x60/0x84
[<f8a8208d>] br_nf_pre_routing_finish+0x0/0x320 [bridge] [<c02e276e>]
nf_hook_slow+0x4d/0xf9
[<f8a8208d>] br_nf_pre_routing_finish+0x0/0x320 [bridge] [<f8a829ec>]
br_nf_pre_routing+0x2f5/0x431 [bridge]
[<f8a8208d>] br_nf_pre_routing_finish+0x0/0x320 [bridge] [<c02e26fd>]
nf_iterate+0x60/0x84
[<f8a7e8f9>] br_handle_frame_finish+0x0/0xe9 [bridge] [<c02e276e>]
nf_hook_slow+0x4d/0xf9
[<f8a7e8f9>] br_handle_frame_finish+0x0/0xe9 [bridge] [<f8a7eb46>]
br_handle_frame+0x164/0x23e [bridge]
[<f8a7e8f9>] br_handle_frame_finish+0x0/0xe9 [bridge] [<c02cbbe5>]
netif_receive_skb+0x1ac/0x325
[<f88d5975>] skge_poll+0x3b6/0x4be [skge] [<c012dca4>]
__mod_timer+0x85/0xa0
[<c02cbf3e>] net_rx_action+0xb7/0x1bb [<c012a2f2>]
__do_softirq+0x72/0xdc
[<c0106393>] do_softirq+0x4b/0x4f
=======================
[<c0106275>] do_IRQ+0x55/0x86 [<c0119d81>]
smp_apic_timer_interrupt+0xc1/0xca
[<c0104a8e>] common_interrupt+0x1a/0x20 [<c0102287>]
mwait_idle+0x2a/0x34
[<c01020ef>] cpu_idle+0x6c/0xa7 [<c040187f>] start_kernel+0x173/0x1ca
[<c0401304>] unknown_bootoption+0x0/0x1b6
and sk98lin
bridge-cd: hw csum failure.
[<c02c8fd8>] __skb_checksum_complete+0x56/0x5c [<f8a5f833>]
icmp_error+0xbf/0x1af [ip_conntrack]
[<c01393c7>] __wake_up_bit+0x2e/0x33 [<f8a5f774>] icmp_error+0x0/0x1af
[ip_conntrack]
[<f8a5cb57>] ip_conntrack_in+0x95/0x2d6 [ip_conntrack] [<c014d4a2>]
mempool_free+0x3a/0x73
[<c016dfad>] end_bio_bh_io_sync+0x0/0x4f [<c016dfad>]
end_bio_bh_io_sync+0x0/0x4f
[<c02e26fd>] nf_iterate+0x60/0x84 [<f8a8208d>]
br_nf_pre_routing_finish+0x0/0x320 [bridge]
[<c02e276e>] nf_hook_slow+0x4d/0xf9 [<f8a8208d>]
br_nf_pre_routing_finish+0x0/0x320 [bridge]
[<f8a829ec>] br_nf_pre_routing+0x2f5/0x431 [bridge] [<f8a8208d>]
br_nf_pre_routing_finish+0x0/0x320 [bridge]
[<c02e26fd>] nf_iterate+0x60/0x84 [<f8a7e8f9>]
br_handle_frame_finish+0x0/0xe9 [bridge]
[<c02e276e>] nf_hook_slow+0x4d/0xf9 [<f8a7e8f9>]
br_handle_frame_finish+0x0/0xe9 [bridge]
[<f8a7eb46>] br_handle_frame+0x164/0x23e [bridge] [<f8a7e8f9>]
br_handle_frame_finish+0x0/0xe9 [bridge]
[<c02cbbe5>] netif_receive_skb+0x1ac/0x325 [<c02cbde1>]
process_backlog+0x83/0x129
[<c02cbf3e>] net_rx_action+0xb7/0x1bb [<c012a2f2>]
__do_softirq+0x72/0xdc
[<c0106393>] do_softirq+0x4b/0x4f
=======================
[<c0106275>] do_IRQ+0x55/0x86 [<c0104a8e>] common_interrupt+0x1a/0x20
[<c014a162>] page_waitqueue+0x5/0x32 [<c014a217>] unlock_page+0x1d/0x27
[<c016c7e8>] __block_write_full_page+0x1e7/0x354 [<f8985176>]
ext3_get_block+0x0/0x90 [ext3]
[<c016df49>] block_write_full_page+0xe3/0x109 [<f8985176>]
ext3_get_block+0x0/0x90 [ext3]
[<f8985cea>] ext3_ordered_writepage+0xe5/0x183 [ext3] [<f8985be5>]
bget_one+0x0/0x7 [ext3]
[<c018cfdb>] mpage_writepages+0x222/0x3ee [<f8985c05>]
ext3_ordered_writepage+0x0/0x183 [ext3]
[<c0149c78>] __filemap_fdatawrite_range+0x66/0x72 [<c0149ca7>]
filemap_fdatawrite+0x23/0x27
[<c016b2e9>] do_fsync+0x55/0xc8 [<c0104049>] syscall_call+0x7/0xb
iptables forward chain is just ACCEPT...
On Tue, 31 Jan 2006 01:16:59 +0200
Pekka Pietikainen <[email protected]> wrote:
> On Fri, Jan 27, 2006 at 11:22:42PM +1100, Herbert Xu wrote:
> > OK, although we can't rule out sky2/netfilter from the enquiry, I've
> > identified two bugs in ppp/pppoe that may be responsible for what you
> > are seeing. So please try the following patch and let us know if the
> > problem still exists (or deteriorates/improves).
> Borrowing this thread for a related problem, I'm getting lots of those on a
> bridge device (this one running skge, rmmod skge; modprobe sk98lin actually
> seemed to do it too, I've disabled rx checksums with ethtool for now).
> Kernel is a 2.6.15.1-ish Fedora one.
Okay, what is the hardware version:
dmesg | grep skge
Maybe that chip rev is no good.
On Mon, Jan 30, 2006 at 04:15:51PM -0800, Stephen Hemminger wrote:
> Okay, what is the hardware version:
> dmesg | grep skge
> Maybe that chip rev is no good.
skge 1.3 addr 0xd042c000 irq 169 chip Yukon-Lite rev 7
skge eth0: addr 00:10:f3:06:83:d4
skge 1.3 addr 0xd0420000 irq 193 chip Yukon-Lite rev 7
skge eth1: addr 00:10:f3:06:83:d5
skge 1.3 addr 0xd0424000 irq 185 chip Yukon-Lite rev 7
skge eth2: addr 00:10:f3:06:83:d6
skge 1.3 addr 0xd0428000 irq 177 chip Yukon-Lite rev 7
skge eth3: addr 00:10:f3:06:83:d7
In related news, the box has been running with a 2.6.16-rc1-git4 based
kernel for 8 hours now without any errors. First kernel+driver version
that does that (apart from vendor-sk98lin on a kernel without slab
debugging ;) ;) ;) ).