| Message ID | 49C61AAA.8050507@gmail.com |
|---|---|
| State | Rejected, archived |
| Delegated to: | David Miller |
| Headers | show |
On Sun, 2009-03-22 at 12:02 +0100, Marcin Slusarz wrote: > When ndo_open (eg skge_up) function printks something, netconsole decides > it can use this device because it checks state only (netif_running) which is > set before ndo_open. Check device flags too. That's fairly unfortunate semantics for netif_running. But if Dave agrees that it's reasonable for that to be set to true at this point in time, then I guess we'll go with it.
From: Matt Mackall <mpm@selenic.com> Date: Sun, 22 Mar 2009 20:20:58 -0500 > On Sun, 2009-03-22 at 12:02 +0100, Marcin Slusarz wrote: > > When ndo_open (eg skge_up) function printks something, netconsole decides > > it can use this device because it checks state only (netif_running) which is > > set before ndo_open. Check device flags too. > > That's fairly unfortunate semantics for netif_running. But if Dave > agrees that it's reasonable for that to be set to true at this point in > time, then I guess we'll go with it. These kind of printk's simply are not allowed, we've removed such printk's from other driver ->open() methods to fix this problem and that's what should be done here. I'm rejecting this patch. -- 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
On 23-03-2009 05:21, David Miller wrote: > From: Matt Mackall <mpm@selenic.com> > Date: Sun, 22 Mar 2009 20:20:58 -0500 > >> On Sun, 2009-03-22 at 12:02 +0100, Marcin Slusarz wrote: >>> When ndo_open (eg skge_up) function printks something, netconsole decides >>> it can use this device because it checks state only (netif_running) which is >>> set before ndo_open. Check device flags too. >> That's fairly unfortunate semantics for netif_running. But if Dave >> agrees that it's reasonable for that to be set to true at this point in >> time, then I guess we'll go with it. > > These kind of printk's simply are not allowed, we've removed such > printk's from other driver ->open() methods to fix this problem and > that's what should be done here. What is the rationale of this decision? printk is a basic tool, especially designed to work in as many places as possible, and netconsole is rather something secondary (sorry Matt)?! Jarek P. -- 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
From: Jarek Poplawski <jarkao2@gmail.com> Date: Mon, 23 Mar 2009 08:04:55 +0000 > What is the rationale of this decision? printk is a basic tool, > especially designed to work in as many places as possible, and > netconsole is rather something secondary (sorry Matt)?! And this basic tool cannot work from the drivers ->open() method. -- 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
On Mon, Mar 23, 2009 at 01:05:41AM -0700, David Miller wrote: > From: Jarek Poplawski <jarkao2@gmail.com> > Date: Mon, 23 Mar 2009 08:04:55 +0000 > > > What is the rationale of this decision? printk is a basic tool, > > especially designed to work in as many places as possible, and > > netconsole is rather something secondary (sorry Matt)?! > > And this basic tool cannot work from the drivers ->open() method. And in any function used in the drivers ->open(). BTW, with Marcin's patch it can... Jarek P. -- 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
From: Jarek Poplawski <jarkao2@gmail.com> Date: Mon, 23 Mar 2009 08:11:58 +0000 > On Mon, Mar 23, 2009 at 01:05:41AM -0700, David Miller wrote: > > From: Jarek Poplawski <jarkao2@gmail.com> > > Date: Mon, 23 Mar 2009 08:04:55 +0000 > > > > > What is the rationale of this decision? printk is a basic tool, > > > especially designed to work in as many places as possible, and > > > netconsole is rather something secondary (sorry Matt)?! > > > > And this basic tool cannot work from the drivers ->open() method. > > And in any function used in the drivers ->open(). BTW, with Marcin's > patch it can... This issue came up before, and after we added the netif_running() check we hit this IIF_UP one and at the time we looked into it and the result we came up with is that you just can't do it in a network driver's ->open() Look up the thread, I'm too lazy... -- 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
On Mon, Mar 23, 2009 at 01:15:08AM -0700, David Miller wrote: > From: Jarek Poplawski <jarkao2@gmail.com> > Date: Mon, 23 Mar 2009 08:11:58 +0000 > > > On Mon, Mar 23, 2009 at 01:05:41AM -0700, David Miller wrote: > > > From: Jarek Poplawski <jarkao2@gmail.com> > > > Date: Mon, 23 Mar 2009 08:04:55 +0000 > > > > > > > What is the rationale of this decision? printk is a basic tool, > > > > especially designed to work in as many places as possible, and > > > > netconsole is rather something secondary (sorry Matt)?! > > > > > > And this basic tool cannot work from the drivers ->open() method. > > > > And in any function used in the drivers ->open(). BTW, with Marcin's > > patch it can... > > This issue came up before, and after we added the netif_running() > check we hit this IIF_UP one and at the time we looked into it > and the result we came up with is that you just can't do it in > a network driver's ->open() > > Look up the thread, I'm too lazy... > So I try to make appear I'm less lazy, and read this one thread only, but can't see this IIF_UP being mentioned: http://marc.info/?t=123306255900001&r=1&w=2 Jarek P. -- 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
On Mon, Mar 23, 2009 at 08:11:58AM +0000, Jarek Poplawski wrote: > On Mon, Mar 23, 2009 at 01:05:41AM -0700, David Miller wrote: > > From: Jarek Poplawski <jarkao2@gmail.com> > > Date: Mon, 23 Mar 2009 08:04:55 +0000 > > > > > What is the rationale of this decision? printk is a basic tool, > > > especially designed to work in as many places as possible, and > > > netconsole is rather something secondary (sorry Matt)?! > > > > And this basic tool cannot work from the drivers ->open() method. > > And in any function used in the drivers ->open(). BTW, with Marcin's > patch it can... And from any function called anywhere on another cpu while driver's ->open() is running. BTW, I've had a look at this and it seems the main problem is netif_tx_stopped() isn't handled properly by the driver(s). Jarek P. -- 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 --git a/drivers/net/netconsole.c b/drivers/net/netconsole.c index d304d38..97e30b0 100644 --- a/drivers/net/netconsole.c +++ b/drivers/net/netconsole.c @@ -705,7 +705,7 @@ static void write_msg(struct console *con, const char *msg, unsigned int len) spin_lock_irqsave(&target_list_lock, flags); list_for_each_entry(nt, &target_list, list) { netconsole_target_get(nt); - if (nt->enabled && netif_running(nt->np.dev)) { + if (nt->enabled && netif_running(nt->np.dev) && (nt->np.dev->flags & IFF_UP)) { /* * We nest this inside the for-each-target loop above * so that we're able to get as much logging out to
When ndo_open (eg skge_up) function printks something, netconsole decides it can use this device because it checks state only (netif_running) which is set before ndo_open. Check device flags too. [35437.623580] skge eth1: enabling interface [35437.623601] ------------[ cut here ]------------ [35437.623603] kernel BUG at drivers/net/skge.c:2767! [35437.623606] invalid opcode: 0000 [#1] PREEMPT [35437.623608] last sysfs file: /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor [35437.623611] CPU 0 [35437.623613] Modules linked in: [35437.623617] Pid: 12711, comm: ip Not tainted 2.6.29-rc6-idle #82 To Be Filled By O.E.M. [35437.623619] RIP: 0010:[<ffffffff803f3c30>] [<ffffffff803f3c30>] skge_xmit_frame+0xbe/0x3ba [35437.623628] RSP: 0018:ffff88003cc0f8b8 EFLAGS: 00010086 [35437.623630] RAX: 000000000000007f RBX: ffff88003e850000 RCX: 0000000000000001 [35437.623632] RDX: 0000000000000001 RSI: ffff88003f188720 RDI: ffff88002e568900 [35437.623635] RBP: ffff88003cc0f918 R08: 0000000000000002 R09: 0000000000000000 [35437.623637] R10: 0000000000000006 R11: 0000000000000000 R12: ffff88002e568900 [35437.623639] R13: ffff88003e850000 R14: ffffffff807180c0 R15: 0000000000000001 [35437.623642] FS: 00007f46b39086f0(0000) GS:ffffffff807dc020(0000) knlGS:00000000f6577b90 [35437.623644] CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b [35437.623646] CR2: 00007f46b3282110 CR3: 000000002ab18000 CR4: 00000000000006e0 [35437.623648] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [35437.623651] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 [35437.623653] Process ip (pid: 12711, threadinfo ffff88003cc0e000, task ffff88002a8b5af0) [35437.623655] Stack: [35437.623657] ffff88003eb43000 000000008077cc20 ffff88003f188000 ffff88003f188720 [35437.623660] ffff88003efba800 ffff88003f188000 0000000000000000 ffff88003efd78a0 [35437.623664] 0000000000000086 ffff88003f188000 0000000000000000 0000000000000001 [35437.623667] Call Trace: [35437.623670] [<ffffffff804fac57>] netpoll_send_skb+0xd8/0x1a1 [35437.623675] [<ffffffff804fb228>] netpoll_send_udp+0x214/0x220 [35437.623678] [<ffffffff803fdf39>] write_msg+0x80/0xbf [35437.623682] [<ffffffff80233413>] __call_console_drivers+0x58/0x69 [35437.623687] [<ffffffff80233485>] _call_console_drivers+0x61/0x66 [35437.623691] [<ffffffff802335bb>] release_console_sem+0x131/0x1d4 [35437.623694] [<ffffffff80233c0c>] vprintk+0x389/0x3b8 [35437.623698] [<ffffffff802556c3>] ? __lock_acquire+0x73b/0x797 [35437.623703] [<ffffffff80233ca2>] printk+0x67/0x69 [35437.623706] [<ffffffff802556c3>] ? __lock_acquire+0x73b/0x797 [35437.623709] [<ffffffff802539e1>] ? mark_held_locks+0x52/0x72 [35437.623712] [<ffffffff80237609>] ? local_bh_enable_ip+0xbe/0xda [35437.623716] [<ffffffff803f0a0f>] skge_up+0x7c/0x88e [35437.623719] [<ffffffff804ebf62>] ? dev_set_rx_mode+0x29/0x2e [35437.623723] [<ffffffff80237609>] ? local_bh_enable_ip+0xbe/0xda [35437.623726] [<ffffffff804f0164>] dev_open+0x73/0xa8 [35437.623729] [<ffffffff804edf99>] dev_change_flags+0xa8/0x167 [35437.623732] [<ffffffff8052e0d5>] devinet_ioctl+0x26a/0x5e3 [35437.623736] [<ffffffff8052efed>] inet_ioctl+0x92/0xaa [35437.623739] [<ffffffff804e1eec>] sock_ioctl+0x1e2/0x20e [35437.623742] [<ffffffff802a3ed0>] vfs_ioctl+0x2a/0x77 [35437.623745] [<ffffffff802a4375>] do_vfs_ioctl+0x458/0x4b0 [35437.623747] [<ffffffff802491bd>] ? up_read+0x26/0x2b [35437.623751] [<ffffffff8020b4cc>] ? sysret_check+0x27/0x62 [35437.623754] [<ffffffff802a440f>] sys_ioctl+0x42/0x65 [35437.623757] [<ffffffff8020b49b>] system_call_fastpath+0x16/0x1b [35437.623760] Code: 52 04 69 c0 cd cc cc cc 8d 44 30 ff ff c2 39 d0 0f 8c 00 03 00 00 48 8b 75 b8 4c 8b ae a8 00 00 00 4d 8b 75 08 41 83 3e 00 79 04 <0f> 0b eb fe 4d 89 65 10 41 8b 44 24 68 45 31 ff 41 8b 54 24 6c [35437.623787] RIP [<ffffffff803f3c30>] skge_xmit_frame+0xbe/0x3ba [35437.623790] RSP <ffff88003cc0f8b8> [35437.623793] ---[ end trace 4dbaa362038903db ]--- [35437.623796] note: ip[12711] exited with preempt_count 3 I could reliably trigger it by: ifconfig eth0 down; while [ true ]; do ifconfig eth1 down; ifconfig eth1 up; done Netconsole oopsed that way since at least 2.6.22 (oldest kernel I tried). Fixes bug 12160. Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com> Cc: David S. Miller <davem@davemloft.net> Cc: Keiichi Kii <k-keiichi@bx.jp.nec.com> Cc: Matt Mackall <mpm@selenic.com> Cc: stable <stable@kernel.org> ? --- drivers/net/netconsole.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)