Re: [PATCH, RFC] panic-note: Annotation from user space for panics

[Simon Kagstrom <simon.kagstrom@netinsight.net>]

(Also fix David Woodhouses address and add Atal)

On Fri, 13 Nov 2009 13:45:48 +0200
Artem Bityutskiy <dedekind1@gmail.com> wrote:

> So we really need to print some user-space supplied information during
> the panic, and then we store it on flash with mtdoops, and the later,
> when the device has network access we send whole bunch of oopses via the
> network.

Yes, I see that your case would have to be handled differently. A
complication (which I believe was discussed before) is that kmsg_dump()
is done before the panic notifiers are called.

The reason I put it there is to have it before crash_kexec(), so I
guess we'll have to take up the discussion on what to do with it. For
me it now seems like it would be OK to move kmsg_dump() down below the
panic notifiers.

If you have a kdump kernel to load, then you will most likely not need
the kmsg dumped data anyway.

// Simon