is there any way to capture and log kernel panics on disk or ...?
shabanip wrote:
>is there any way to capture and log kernel panics on disk or ...?
>
>
My guess would be, at the very least, it depends on what part of the
kernel is causing the panic. Most likely I would say no, although
here's another question: if running a second kernel under user-mode
Linux, can this be done?
James
James Colannino <[email protected]> writes:
> shabanip wrote:
>
>>is there any way to capture and log kernel panics on disk or ...?
>
> My guess would be, at the very least, it depends on what part of the
> kernel is causing the panic.
A kernel panic means that the kernel no longer knows what it's doing,
and therefore stops doing anything immediately. Hence it won't use the
filesystems and cannot log the panic to anything but the console. I
would think the best solution to your problem is to set up a serial
console to another machine, and log everything to disk on that
machine. See Documentation/serial-console.txt for how to set up a
serial console.
Regards,
--
Trond Hasle Amundsen <[email protected]>
Center for Information Technology Services, University of Oslo
shabanip wrote:
> is there any way to capture and log kernel panics on disk or ...?
In former times, the Linux kernel tried to sync in the panic function. (If
the panic did not happen in interrupt context) Unfortunately this had
severe side effects in cases where the panic was triggered by file system
block device code or any other part which is necessary for syncing. In most
cases the call trace never made it onto disk anyway. So currently the
kernel does not support saving a panic.
Apart from using a serial console, you might have a look at several
kexec/kdump/lkcd tools where people are working on being able to dump the
memory of a paniced kernel.
cheers
Christian
==> Regarding Re: how to capture kernel panics; Christian Borntraeger <[email protected]> adds:
linux-kernel> shabanip wrote:
>> is there any way to capture and log kernel panics on disk or ...?
linux-kernel> In former times, the Linux kernel tried to sync in the panic
linux-kernel> function. (If the panic did not happen in interrupt context)
linux-kernel> Unfortunately this had severe side effects in cases where the
linux-kernel> panic was triggered by file system block device code or any
linux-kernel> other part which is necessary for syncing. In most cases the
linux-kernel> call trace never made it onto disk anyway. So currently the
linux-kernel> kernel does not support saving a panic.
linux-kernel> Apart from using a serial console, you might have a look at
linux-kernel> several kexec/kdump/lkcd tools where people are working on
linux-kernel> being able to dump the memory of a paniced kernel.
Or netconsole, which will dump printk's do the server:port of your
choosing.
-Jeff
as i see netconsole is a kernel module.
so i just need to load netconsole module with server:port parameters.
am i right?
Payam Shabanian
shabanip -at- avapajoohesh.com
> ==> Regarding Re: how to capture kernel panics; Christian Borntraeger
> <[email protected]> adds:
>
> linux-kernel> shabanip wrote:
>>> is there any way to capture and log kernel panics on disk or ...?
>
> linux-kernel> In former times, the Linux kernel tried to sync in the panic
> linux-kernel> function. (If the panic did not happen in interrupt context)
> linux-kernel> Unfortunately this had severe side effects in cases where
> the
> linux-kernel> panic was triggered by file system block device code or any
> linux-kernel> other part which is necessary for syncing. In most cases the
> linux-kernel> call trace never made it onto disk anyway. So currently the
> linux-kernel> kernel does not support saving a panic.
>
> linux-kernel> Apart from using a serial console, you might have a look at
> linux-kernel> several kexec/kdump/lkcd tools where people are working on
> linux-kernel> being able to dump the memory of a paniced kernel.
>
> Or netconsole, which will dump printk's do the server:port of your
> choosing.
>
> -Jeff
>
==> Regarding Re: how to capture kernel panics; "shabanip" <[email protected]> adds:
shabanip> as i see netconsole is a kernel module. so i just need to load
shabanip> netconsole module with server:port parameters. am i right?
MODULE_PARM_DESC(netconsole, " netconsole=[src-port]@[src-ip]/[dev],[tgt-port]@<tgt-ip>/[tgt-macaddr]\n");
So, for example:
modprobe netconsole [email protected]/eth0,[email protected]/00:40:95:9A:12:34
-Jeff
shabanip> Payam Shabanian shabanip -at- avapajoohesh.com
>> ==> Regarding Re: how to capture kernel panics; Christian Borntraeger
>> <[email protected]> adds:
>>
linux-kernel> shabanip wrote:
>>>> is there any way to capture and log kernel panics on disk or ...?
>>
linux-kernel> In former times, the Linux kernel tried to sync in the panic
linux-kernel> function. (If the panic did not happen in interrupt context)
linux-kernel> Unfortunately this had severe side effects in cases where
>> the
linux-kernel> panic was triggered by file system block device code or any
linux-kernel> other part which is necessary for syncing. In most cases the
linux-kernel> call trace never made it onto disk anyway. So currently the
linux-kernel> kernel does not support saving a panic.
>>
linux-kernel> Apart from using a serial console, you might have a look at
linux-kernel> several kexec/kdump/lkcd tools where people are working on
linux-kernel> being able to dump the memory of a paniced kernel.
>> Or netconsole, which will dump printk's do the server:port of your
>> choosing.
>>
>> -Jeff
>>
shabanip> - To unsubscribe from this list: send the line "unsubscribe
shabanip> linux-kernel" in the body of a message to
shabanip> [email protected] More majordomo info at
shabanip> http://vger.kernel.org/majordomo-info.html Please read the FAQ at
shabanip> http://www.tux.org/lkml/
thanks for the help.
does anything else left to configure?
Payam Shabanian
shabanip -at- avapajoohesh.com
> ==> Regarding Re: how to capture kernel panics; "shabanip"
> <[email protected]> adds:
>
> shabanip> as i see netconsole is a kernel module. so i just need to load
> shabanip> netconsole module with server:port parameters. am i right?
>
> MODULE_PARM_DESC(netconsole, "
> netconsole=[src-port]@[src-ip]/[dev],[tgt-port]@<tgt-ip>/[tgt-macaddr]\n");
>
> So, for example:
>
> modprobe netconsole
> [email protected]/eth0,[email protected]/00:40:95:9A:12:34
>
> -Jeff
>
>
> shabanip> Payam Shabanian shabanip -at- avapajoohesh.com
>
>>> ==> Regarding Re: how to capture kernel panics; Christian Borntraeger
>>> <[email protected]> adds:
>>>
> linux-kernel> shabanip wrote:
>>>>> is there any way to capture and log kernel panics on disk or ...?
>>>
> linux-kernel> In former times, the Linux kernel tried to sync in the panic
> linux-kernel> function. (If the panic did not happen in interrupt context)
> linux-kernel> Unfortunately this had severe side effects in cases where
>>> the
> linux-kernel> panic was triggered by file system block device code or any
> linux-kernel> other part which is necessary for syncing. In most cases the
> linux-kernel> call trace never made it onto disk anyway. So currently the
> linux-kernel> kernel does not support saving a panic.
>>>
> linux-kernel> Apart from using a serial console, you might have a look at
> linux-kernel> several kexec/kdump/lkcd tools where people are working on
> linux-kernel> being able to dump the memory of a paniced kernel.
>>> Or netconsole, which will dump printk's do the server:port of your
>>> choosing.
>>>
>>> -Jeff
>>>
>
> shabanip> - To unsubscribe from this list: send the line "unsubscribe
> shabanip> linux-kernel" in the body of a message to
> shabanip> [email protected] More majordomo info at
> shabanip> http://vger.kernel.org/majordomo-info.html Please read the FAQ
> at
> shabanip> http://www.tux.org/lkml/
>
==> Regarding Re: how to capture kernel panics; "shabanip" <[email protected]> adds:
shabanip> thanks for the help. does anything else left to configure?
On the log target system, you will probably want to run something similar
to netcat to capture the logs. To simply output to a terminal, you could
do something like this:
nc -l -p 6666 -u
-Jeff
shabanip> Payam Shabanian shabanip -at- avapajoohesh.com
>> ==> Regarding Re: how to capture kernel panics; "shabanip"
>> <[email protected]> adds:
>>
shabanip> as i see netconsole is a kernel module. so i just need to load
shabanip> netconsole module with server:port parameters. am i right?
>> MODULE_PARM_DESC(netconsole, "
>> netconsole=[src-port]@[src-ip]/[dev],[tgt-port]@<tgt-ip>/[tgt-macaddr]\n");
>>
>> So, for example:
>>
>> modprobe netconsole
>> [email protected]/eth0,[email protected]/00:40:95:9A:12:34
>>
>> -Jeff
>>
>>
shabanip> Payam Shabanian shabanip -at- avapajoohesh.com
>>
>>>> ==> Regarding Re: how to capture kernel panics; Christian Borntraeger
>>>> <[email protected]> adds:
>>>>
linux-kernel> shabanip wrote:
>>>>>> is there any way to capture and log kernel panics on disk or ...?
>>>>
linux-kernel> In former times, the Linux kernel tried to sync in the panic
linux-kernel> function. (If the panic did not happen in interrupt context)
linux-kernel> Unfortunately this had severe side effects in cases where
>>>> the
linux-kernel> panic was triggered by file system block device code or any
linux-kernel> other part which is necessary for syncing. In most cases the
linux-kernel> call trace never made it onto disk anyway. So currently the
linux-kernel> kernel does not support saving a panic.
>>>>
linux-kernel> Apart from using a serial console, you might have a look at
linux-kernel> several kexec/kdump/lkcd tools where people are working on
linux-kernel> being able to dump the memory of a paniced kernel.
>>>> Or netconsole, which will dump printk's do the server:port of your
>>>> choosing.
>>>>
>>>> -Jeff
>>>>
>>
shabanip> - To unsubscribe from this list: send the line "unsubscribe
shabanip> linux-kernel" in the body of a message to
shabanip> [email protected] More majordomo info at
shabanip> http://vger.kernel.org/majordomo-info.html Please read the FAQ
>> at
shabanip> http://www.tux.org/lkml/
>>