Message ID | alpine.DEB.2.00.1207190213001.16299@artax.karlin.mff.cuni.cz |
---|---|
State | Rejected, archived |
Delegated to: | David Miller |
Headers | show |
On Thu, 2012-07-19 at 03:12 +0200, Mikulas Patocka wrote: > Hi > > I want to report a crash when using the tun driver. The crash can be > reproduced by starting and stopping miredo in the loop: > > while ! dmesg | grep -q BUG; do > /etc/init.d/miredo start;/etc/init.d/miredo stop > done > > The crash happens in iput_final because inode->i_sb->s_op is NULL. > > The crash happens in 3.4, 3.4.5, 3.5-rc7. The crash does not happen in > 3.3. When I attempted to bisect, unrelated changes in completely different > driver regarding memory allocation triggered the crash --- so it is > likely buggy even in 3.3 and before, it just didn't show up. > > > What is obviously wrong: > in the tun driver "struct socket" is embedded in "struct tun_struct". > > The backtrace goes through: > netdev_run_todo -> tun_free_netdev -> sk_release_kernel -> sock_release. > > sock_release calls iput(SOCK_INODE(sock)). SOCK_INODE assumes that struct > socket is embedded in "struct socket_alloc" (which is not true, it is > embedded in "struct tun_struct"), gets a pointer to non-existing inode --- > and there goes the crash. > > > The crash can be fixed by writing any non-NULL value to tun->socket.file > to prevent sock_release from calling iput(SOCK_INODE(sock)). Or maybe you > come up with a better fix. > > > Note another bug - when you are repeatedly starting and stopping miredo, > even if it doesn't crash, the value "sockets: used" in > /proc/*/net/sockstat keeps on decreasing. That's because sock_release > decrements sockets_in_use, but there was no sock_alloc to increment it. > > > Mikulas > > --- > > Signed-off-by: Mikulas Patocka <mikulas@artax.karlin.mff.cuni.cz> > > --- > drivers/net/tun.c | 2 ++ > net/socket.c | 4 +++- > 2 files changed, 5 insertions(+), 1 deletion(-) > > Index: linux-3.4.5-fast/drivers/net/tun.c > =================================================================== > --- linux-3.4.5-fast.orig/drivers/net/tun.c 2012-07-19 02:42:56.000000000 +0200 > +++ linux-3.4.5-fast/drivers/net/tun.c 2012-07-19 02:50:13.000000000 +0200 > @@ -358,6 +358,8 @@ static void tun_free_netdev(struct net_d > { > struct tun_struct *tun = netdev_priv(dev); > > + /* Prevent the code in sock_release from calling iput. */ > + tun->socket.file = (void *)1; > sk_release_kernel(tun->socket.sk); > } > > Index: linux-3.4.5-fast/net/socket.c > =================================================================== > --- linux-3.4.5-fast.orig/net/socket.c 2012-07-19 03:00:30.000000000 +0200 > +++ linux-3.4.5-fast/net/socket.c 2012-07-19 03:05:36.000000000 +0200 > @@ -522,7 +522,9 @@ void sock_release(struct socket *sock) > if (rcu_dereference_protected(sock->wq, 1)->fasync_list) > printk(KERN_ERR "sock_release: fasync list not empty!\n"); > > - percpu_sub(sockets_in_use, 1); > + /* a hack - sockets_in_use should not be decremented when tun calls this */ > + if (sock->file != (void *)1) > + percpu_sub(sockets_in_use, 1); > if (!sock->file) { > iput(SOCK_INODE(sock)); > return; > > --- > > BUG: unable to handle kernel NULL pointer dereference at 0000000000000020 > IP: [<ffffffff81113246>] iput+0x76/0x230 > PGD 43e5e9067 PUD 4468d5067 PMD 0 > Oops: 0000 [#1] PREEMPT SMP > CPU 1 > Modules linked in: ip6table_filter ip6_tables iptable_filter ip_tables > ebtable_nat ebtables x_tables kvm_amd kvm tun cpufreq_userspace > cpufreq_stats cpufreq_powersave cpufreq_ondemand cpufreq_conservative ipv6 > fuse raid0 md_mod lm85 hwmon_vid snd_usb_audio snd_pcm_oss snd_mixer_oss > snd_pcm snd_timer snd_page_alloc snd_hwdep snd_usbmidi_lib snd_rawmidi snd > soundcore ide_cd_mod cdrom ohci_hcd sata_svw libata serverworks ide_core > powernow_k8 ehci_hcd usbcore tg3 usb_common floppy freq_table e100 skge > mii i2c_piix4 libphy mperf k10temp rtc_cmos processor button hwmon > microcode unix > > Pid: 10826, comm: miredo Not tainted 3.4.5 #85 empty empty/S3992-E > RIP: 0010:[<ffffffff81113246>] [<ffffffff81113246>] iput+0x76/0x230 > RSP: 0018:ffff88043dc73e48 EFLAGS: 00010246 > RAX: 0000000000000001 RBX: ffff88023eba3f20 RCX: ffff880447d80000 > RDX: 0000000000000001 RSI: ffff88023eba3f98 RDI: ffff88023eba3f98 > RBP: ffff88023eba3f98 R08: ffffffff814804d8 R09: 0000000000000000 > R10: 0000000000000096 R11: dead000000100100 R12: ffff88023eba3f48 > R13: 0000000000000000 R14: ffff88043dc73e88 R15: 00000000ffffad3a > FS: 00007fe71bbb9700(0000) GS:ffff880247c80000(0000) > knlGS:0000000000000000 > CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b > CR2: 0000000000000020 CR3: 000000043dcf0000 CR4: 00000000000007f0 > DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 > DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 > Process miredo (pid: 10826, threadinfo ffff88043dc72000, task > ffff88043dc1b8d0) Stack: > ffff88043dc73e88 ffff88023efef000 ffff88023eba3800 ffff88023eba3c00 > ffff88024749ecc0 ffffffff81281d93 ffff88023e8fa500 ffffffff81295438 > ffff88043dc73e88 ffff88043dc73e88 ffff88023eba38f8 ffff88023e8fa500 > Call Trace: > [<ffffffff81281d93>] ? sk_release_kernel+0x23/0x40 > [<ffffffff81295438>] ? netdev_run_todo+0x1a8/0x260 > [<ffffffffa02522a3>] ? tun_chr_close+0x93/0xb0 [tun] > [<ffffffff810fc3ed>] ? fput+0xdd/0x260 > [<ffffffff810f8aff>] ? filp_close+0x5f/0x90 > [<ffffffff810f8bc7>] ? sys_close+0x97/0x100 > [<ffffffff81311f22>] ? system_call_fastpath+0x16/0x1b > Code: 6c 24 10 4c 8b 64 24 18 4c 8b 6c 24 20 48 83 c4 28 c3 0f 1f 00 f6 83 > 90 00 00 00 08 4c 8b 63 28 4d 8b 6c 24 30 0f 85 7d 01 00 00 <49> 8b 45 20 > 48 85 c0 0f 84 9d 00 00 00 48 89 df ff d0 85 c0 0f > RIP [<ffffffff81113246>] iput+0x76/0x230 > RSP <ffff88043dc73e48> > CR2: 0000000000000020 > ---[ end trace 6bd5160ffd3ba7a2 ]--- > note: miredo[10826] exited with preempt_count 1 Hi Mikulas A fix for this problem is : http://patchwork.ozlabs.org/patch/170440/ -- 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: Eric Dumazet <eric.dumazet@gmail.com> Date: Thu, 19 Jul 2012 08:09:35 +0200 > A fix for this problem is : http://patchwork.ozlabs.org/patch/170440/ But it was submitted as an RFC so much be resubmitted formally as a non-RFC 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
Index: linux-3.4.5-fast/drivers/net/tun.c =================================================================== --- linux-3.4.5-fast.orig/drivers/net/tun.c 2012-07-19 02:42:56.000000000 +0200 +++ linux-3.4.5-fast/drivers/net/tun.c 2012-07-19 02:50:13.000000000 +0200 @@ -358,6 +358,8 @@ static void tun_free_netdev(struct net_d { struct tun_struct *tun = netdev_priv(dev); + /* Prevent the code in sock_release from calling iput. */ + tun->socket.file = (void *)1; sk_release_kernel(tun->socket.sk); } Index: linux-3.4.5-fast/net/socket.c =================================================================== --- linux-3.4.5-fast.orig/net/socket.c 2012-07-19 03:00:30.000000000 +0200 +++ linux-3.4.5-fast/net/socket.c 2012-07-19 03:05:36.000000000 +0200 @@ -522,7 +522,9 @@ void sock_release(struct socket *sock) if (rcu_dereference_protected(sock->wq, 1)->fasync_list) printk(KERN_ERR "sock_release: fasync list not empty!\n"); - percpu_sub(sockets_in_use, 1); + /* a hack - sockets_in_use should not be decremented when tun calls this */ + if (sock->file != (void *)1) + percpu_sub(sockets_in_use, 1); if (!sock->file) { iput(SOCK_INODE(sock)); return;
Hi I want to report a crash when using the tun driver. The crash can be reproduced by starting and stopping miredo in the loop: while ! dmesg | grep -q BUG; do /etc/init.d/miredo start;/etc/init.d/miredo stop done The crash happens in iput_final because inode->i_sb->s_op is NULL. The crash happens in 3.4, 3.4.5, 3.5-rc7. The crash does not happen in 3.3. When I attempted to bisect, unrelated changes in completely different driver regarding memory allocation triggered the crash --- so it is likely buggy even in 3.3 and before, it just didn't show up. What is obviously wrong: in the tun driver "struct socket" is embedded in "struct tun_struct". The backtrace goes through: netdev_run_todo -> tun_free_netdev -> sk_release_kernel -> sock_release. sock_release calls iput(SOCK_INODE(sock)). SOCK_INODE assumes that struct socket is embedded in "struct socket_alloc" (which is not true, it is embedded in "struct tun_struct"), gets a pointer to non-existing inode --- and there goes the crash. The crash can be fixed by writing any non-NULL value to tun->socket.file to prevent sock_release from calling iput(SOCK_INODE(sock)). Or maybe you come up with a better fix. Note another bug - when you are repeatedly starting and stopping miredo, even if it doesn't crash, the value "sockets: used" in /proc/*/net/sockstat keeps on decreasing. That's because sock_release decrements sockets_in_use, but there was no sock_alloc to increment it. Mikulas --- Signed-off-by: Mikulas Patocka <mikulas@artax.karlin.mff.cuni.cz> --- drivers/net/tun.c | 2 ++ net/socket.c | 4 +++- 2 files changed, 5 insertions(+), 1 deletion(-) --- BUG: unable to handle kernel NULL pointer dereference at 0000000000000020 IP: [<ffffffff81113246>] iput+0x76/0x230 PGD 43e5e9067 PUD 4468d5067 PMD 0 Oops: 0000 [#1] PREEMPT SMP CPU 1 Modules linked in: ip6table_filter ip6_tables iptable_filter ip_tables ebtable_nat ebtables x_tables kvm_amd kvm tun cpufreq_userspace cpufreq_stats cpufreq_powersave cpufreq_ondemand cpufreq_conservative ipv6 fuse raid0 md_mod lm85 hwmon_vid snd_usb_audio snd_pcm_oss snd_mixer_oss snd_pcm snd_timer snd_page_alloc snd_hwdep snd_usbmidi_lib snd_rawmidi snd soundcore ide_cd_mod cdrom ohci_hcd sata_svw libata serverworks ide_core powernow_k8 ehci_hcd usbcore tg3 usb_common floppy freq_table e100 skge mii i2c_piix4 libphy mperf k10temp rtc_cmos processor button hwmon microcode unix Pid: 10826, comm: miredo Not tainted 3.4.5 #85 empty empty/S3992-E RIP: 0010:[<ffffffff81113246>] [<ffffffff81113246>] iput+0x76/0x230 RSP: 0018:ffff88043dc73e48 EFLAGS: 00010246 RAX: 0000000000000001 RBX: ffff88023eba3f20 RCX: ffff880447d80000 RDX: 0000000000000001 RSI: ffff88023eba3f98 RDI: ffff88023eba3f98 RBP: ffff88023eba3f98 R08: ffffffff814804d8 R09: 0000000000000000 R10: 0000000000000096 R11: dead000000100100 R12: ffff88023eba3f48 R13: 0000000000000000 R14: ffff88043dc73e88 R15: 00000000ffffad3a FS: 00007fe71bbb9700(0000) GS:ffff880247c80000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: 0000000000000020 CR3: 000000043dcf0000 CR4: 00000000000007f0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process miredo (pid: 10826, threadinfo ffff88043dc72000, task ffff88043dc1b8d0) Stack: ffff88043dc73e88 ffff88023efef000 ffff88023eba3800 ffff88023eba3c00 ffff88024749ecc0 ffffffff81281d93 ffff88023e8fa500 ffffffff81295438 ffff88043dc73e88 ffff88043dc73e88 ffff88023eba38f8 ffff88023e8fa500 Call Trace: [<ffffffff81281d93>] ? sk_release_kernel+0x23/0x40 [<ffffffff81295438>] ? netdev_run_todo+0x1a8/0x260 [<ffffffffa02522a3>] ? tun_chr_close+0x93/0xb0 [tun] [<ffffffff810fc3ed>] ? fput+0xdd/0x260 [<ffffffff810f8aff>] ? filp_close+0x5f/0x90 [<ffffffff810f8bc7>] ? sys_close+0x97/0x100 [<ffffffff81311f22>] ? system_call_fastpath+0x16/0x1b Code: 6c 24 10 4c 8b 64 24 18 4c 8b 6c 24 20 48 83 c4 28 c3 0f 1f 00 f6 83 90 00 00 00 08 4c 8b 63 28 4d 8b 6c 24 30 0f 85 7d 01 00 00 <49> 8b 45 20 48 85 c0 0f 84 9d 00 00 00 48 89 df ff d0 85 c0 0f RIP [<ffffffff81113246>] iput+0x76/0x230 RSP <ffff88043dc73e48> CR2: 0000000000000020 ---[ end trace 6bd5160ffd3ba7a2 ]--- note: miredo[10826] exited with preempt_count 1 -- 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