| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <4B8EFA85.4010003@xenotime.net>
Date: Wed, 03 Mar 2010 16:10:45 -0800
From: Randy Dunlap <rdunlap@...otime.net>
To: Andrew Morton <akpm@...ux-foundation.org>
CC: M G Berberich <berberic@....uni-passau.de>,
Américo Wang
<xiyou.wangcong@...il.com>, linux-kernel@...r.kernel.org,
Linux Kernel Network Developers <netdev@...r.kernel.org>,
Jesse Barnes <jbarnes@...tuousgeek.org>,
Greg KH <greg@...ah.com>, Kay Sievers <kay.sievers@...y.org>
Subject: Re: 2.6.33 dies on modprobe
On 03/03/10 16:05, Andrew Morton wrote:
> On Wed, 3 Mar 2010 14:24:20 -0800
> Andrew Morton <akpm@...ux-foundation.org> wrote:
>
>> On Wed, 3 Mar 2010 23:16:02 +0100
>> M G Berberich <berberic@....uni-passau.de> wrote:
>>
>>> Hello,
>>>
>>> Am Dienstag, den 02. M__rz schrieb Andrew Morton:
>>>
>>>> It could be that some kobject on that list has become invalid (memory
>>>> was freed, module was unloaded, etc) and later code stumbled across the
>>>> now-invalid object on that list and then crashed.
>>>>
>>>> What we can do to find this is to add a diagnostic each time an object
>>>> is registered, and a diagnostic each time kset_find_obj() looks at the
>>>> objects. Then we'll see which kobject caused the crash, then we can
>>>> look back and see where that kobject was registered from.
>>>
>>> [...]
>>>
>>>> This will generate a lot of output and we don't want to lose any of it.
>>>> I'd suggest setting up netconsole so all the output can be reliably
>>>> saved: Documentation/networking/netconsole.txt
>>>
>>> I have a serial connection to a netbook. Log attached.
>>
>> drat, my patch didn't work. Can you try this one please?
>
> He did.
>
>> --- a/lib/kobject.c~a
>> +++ a/lib/kobject.c
>> @@ -126,6 +126,8 @@ static void kobj_kset_join(struct kobjec
>>
>> kset_get(kobj->kset);
>> spin_lock(&kobj->kset->list_lock);
>> + printk("kobj_kset_join:%p\n", kobj);
>> + dump_stack();
>> list_add_tail(&kobj->entry, &kobj->kset->list);
>> spin_unlock(&kobj->kset->list_lock);
>> }
>> @@ -751,9 +753,12 @@ struct kobject *kset_find_obj(struct kse
>>
>> spin_lock(&kset->list_lock);
>> list_for_each_entry(k, &kset->list, entry) {
>> - if (kobject_name(k) && !strcmp(kobject_name(k), name)) {
>> - ret = kobject_get(k);
>> - break;
>> + if (kobject_name(k)) {
>> + printk("kset_find_obj:%p\n", k);
>> + if (!strcmp(kobject_name(k), name)) {
>> + ret = kobject_get(k);
>> + break;
>> + }
>> }
>> }
>> spin_unlock(&kset->list_lock);
>
> And here's what we have:
>
> kobj_kset_join:ffff88012fa80e40
> Pid: 1, comm: swapper Not tainted 2.6.33-bmg #2
> Call Trace:
> [<ffffffff811853ce>] kobject_add_internal+0x8e/0x210
> [<ffffffff81185668>] kobject_add_varg+0x38/0x60
> [<ffffffff811856e3>] kobject_init_and_add+0x53/0x70
> [<ffffffff810ae754>] ? kmem_cache_alloc+0x74/0xc0
> [<ffffffff81235fa4>] bus_add_driver+0x94/0x260
> [<ffffffff81236ce8>] driver_register+0x78/0x140
> [<ffffffff8119b7a1>] __pci_register_driver+0x51/0xd0
> [<ffffffff81513f50>] ? pcie_portdrv_init+0x0/0x4c
> [<ffffffff81513f50>] ? pcie_portdrv_init+0x0/0x4c
> [<ffffffff81513f8b>] pcie_portdrv_init+0x3b/0x4c
> [<ffffffff810001d7>] do_one_initcall+0x37/0x190
> [<ffffffff814fa6a4>] kernel_init+0x14d/0x1a3
> [<ffffffff81003194>] kernel_thread_helper+0x4/0x10
> [<ffffffff814fa557>] ? kernel_init+0x0/0x1a3
> [<ffffffff81003190>] ? kernel_thread_helper+0x0/0x10
>
> ...
>
> kset_find_obj:ffff88012fa80e40
> BUG: unable to handle kernel paging request at ffffffffa005c57f
> IP: [<ffffffff8118960b>] strcmp+0xb/0x30
> PGD 1498067 PUD 149c063 PMD 12d72c067 PTE 0
> Oops: 0000 [#1] SMP
> last sysfs file: /sys/devices/pci0000:00/0000:00:05.0/host0/target0:0:0/0:0:0:0/block/sda/uevent
> CPU 1
> Pid: 1263, comm: modprobe Not tainted 2.6.33-bmg #2 M55S-S3/
> RIP: 0010:[<ffffffff8118960b>] [<ffffffff8118960b>] strcmp+0xb/0x30
> RSP: 0018:ffff88012ef83e58 EFLAGS: 00010292
> RAX: 0000000000000070 RBX: ffff88012fa80e40 RCX: 00000000000005b0
> RDX: 0000000000000000 RSI: ffffffffa005c57f RDI: ffff88012f99adb0
> RBP: ffff88012ef83e58 R08: 0000000000000000 R09: ffff88012ef83e08
> R10: 0000000000000000 R11: 000000000000000f R12: ffff88012f8842a0
> R13: ffffffffa005c57f R14: 0000000001a35970 R15: 0000000001a33050
> FS: 00007f3b044766f0(0000) GS:ffff880028280000(0000) knlGS:0000000000000000
> CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b
> CR2: ffffffffa005c57f CR3: 000000012ea48000 CR4: 00000000000006e0
> DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
> Process modprobe (pid: 1263, threadinfo ffff88012ef82000, task ffff88012e4d34e0)
> Stack:
> ffff88012ef83e88 ffffffff81185320 ffff88012ef83e88 ffffffffa00540a0
> <0> ffffffffa0054100 0000000001a35970 ffff88012ef83e98 ffffffff81236bb7
> <0> ffff88012ef83ed8 ffffffff81236cd7 ffff88012ef83f08 ffffffff81034855
> Call Trace:
> [<ffffffff81185320>] kset_find_obj+0x70/0x90
> [<ffffffff81236bb7>] driver_find+0x17/0x30
> [<ffffffff81236cd7>] driver_register+0x67/0x140
> [<ffffffff81034855>] ? try_to_wake_up+0x215/0x2f0
> [<ffffffff8119b7a1>] __pci_register_driver+0x51/0xd0
> [<ffffffffa0058000>] ? init_nic+0x0/0x20 [forcedeth]
> [<ffffffffa005801e>] init_nic+0x1e/0x20 [forcedeth]
> [<ffffffff810001d7>] do_one_initcall+0x37/0x190
> [<ffffffff81067296>] sys_init_module+0xd6/0x250
> [<ffffffff8100246b>] system_call_fastpath+0x16/0x1b
> Code: 0f b6 3e 48 ff c6 40 84 ff 40 88 39 74 0b 48 ff c1 48 ff ca 75 ea c6 01 00 c9 c3 0f 1f 44 00 00 55 48 89 e5 0f 1f 40 00 0f b6 07 <0f> b6 16 48 ff c7 48 ff c6 38 d0 75 08 84 c0 75 ec 31 c0 c9 c3
> RIP [<ffffffff8118960b>] strcmp+0xb/0x30
> RSP <ffff88012ef83e58>
> CR2: ffffffffa005c57f
>
>
> So a kobject which was created under pcie_portdrv_init() caused an oops
> much later when kset_find_obj() did strcmp(kobject_name(k), name)) on
> that object. Which tends to imply that someone freed that memory or
> trashed kobj->name while that pcie kobject was on the list.
>
> Greg, Jesse, Kay, could you take a look please?
>
> I guess one thing we could do is to change that debug patch to print
> kobj->name as well, see whether it changes.
>
> M G, do you have all debug options enabled, especially the
> memory-management ones? Perhaps CONFIG_DEBUG_PAGEALLOC will pick
> something up.
or using SLUB MM and slub_debug possibly.
--
~Randy
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists