On Mon, Sep 18, 2006 at 08:07:52PM -0400, Garry T. Williams wrote:
> The dmesg log although verbose shows nothing remarkable except for
> this (after starting KDE and kmail):
>
> =============================================
> [ INFO: possible recursive locking detected ]
> 2.6.17-1.3001.fc5 #1
> ---------------------------------------------
> kio_pop3/2453 is trying to acquire lock:
> (slock-AF_INET6){-+..}, at: [<c05a0110>] sk_clone+0xb9/0x2c0
>
> but task is already holding lock:
> (slock-AF_INET6){-+..}, at: [<f8bea96d>] tcp_v6_rcv+0x308/0x6c3 [ipv6]
>
> other info that might help us debug this:
> 1 lock held by kio_pop3/2453:
> #0: (slock-AF_INET6){-+..}, at: [<f8bea96d>] tcp_v6_rcv+0x308/0x6c3 [ipv6]
>
> stack backtrace:
> [<c0403f8e>] show_trace_log_lvl+0x58/0x171
> [<c040456d>] show_trace+0xd/0x10
> [<c040460a>] dump_stack+0x19/0x1b
> [<c0431410>] __lock_acquire+0x75f/0x986
> [<c0431ba8>] lock_acquire+0x4b/0x6b
> [<c05fb928>] _spin_lock+0x19/0x28
> [<c05a0110>] sk_clone+0xb9/0x2c0
> [<c05c6fcd>] inet_csk_clone+0xd/0x5e
> [<c05d75c8>] tcp_create_openreq_child+0x1b/0x391
> [<f8be9691>] tcp_v6_syn_recv_sock+0x248/0x55c [ipv6]
> [<c05d7aee>] tcp_check_req+0x1b0/0x2b1
> [<f8be89c6>] tcp_v6_do_rcv+0x142/0x334 [ipv6]
> [<f8beacd9>] tcp_v6_rcv+0x674/0x6c3 [ipv6]
> [<f8bd1991>] ip6_input+0x1bd/0x249 [ipv6]
> [<f8bd1e2a>] ipv6_rcv+0x19f/0x1cd [ipv6]
> [<c05a613b>] netif_receive_skb+0x202/0x25e
> [<c05a78fb>] process_backlog+0x80/0xdf
> [<c05a7abb>] net_rx_action+0x80/0x162
> [<c041fcf1>] __do_softirq+0x46/0x9c
> [<c040517c>] do_softirq+0x4d/0xb1
> [<c041fbc4>] local_bh_enable_ip+0x9b/0xc1
> [<c05fb8c6>] _spin_unlock_bh+0x25/0x28
> [<c059ef4f>] release_sock+0xb9/0xc1
> [<c05df513>] inet_stream_connect+0x205/0x20f
> [<c059df1e>] sys_connect+0x67/0x84
> [<c059e580>] sys_socketcall+0x8c/0x186
> [<c0402dbf>] syscall_call+0x7/0xb
> DWARF2 unwinder stuck at syscall_call+0x7/0xb
> Leftover inexact backtrace:
> ------------------- End of log -----------------------
>
> This looks like an application error, though. If I can't find the
> same in my older kernel logs, I'll open a bug.
It's a kernel locking bug found by the shiny new lock-validator.
Mark the bugzilla as blocking FCMETA_LOCKDEP
> But much more interesting to me is S3 wake is working on this kernel!
> See bug https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=198998 .
Yeah, I think this kernel might fix up a number of regressions in this area.
I wish I knew which changesets in 2.6.18rc were responsible, as I'd have
backported them to 2.6.17.x. Sadly with ~6000 changes, it gets a little
tricky to isolate fixes.
Dave
