lkml.org 
[lkml]   [2012]   [Mar]   [7]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [PATCH 11/13] android: persistent_trace: ftrace into persistent_ram
From
Date
On Wed, 2012-03-07 at 14:10 -0800, Greg KH wrote:
> On Wed, Mar 07, 2012 at 01:58:21PM -0800, John Stultz wrote:
> > From: Colin Cross <ccross@android.com>
> >
> > persistent_trace uses the ftrace infrastructure, but traces
> > into a persistent_ram buffer instead of the regular ftrace
> > ringbuffer. After a reset or panic, the trace can be
> > decoded with cat /sys/kernel/debug/persistent_trace.
>
> Has this patch been run by the tracing kernel developers?

Probably not. If you'd prefer, we can drop this one and come back to it
later.

> And what's the status on merging the persistent_ram stuff in with the
> in-kernel api for this type of thing? That should then get rid of this
> separate tracing module, or at the least, let it be accepted directly by
> the tracing developers and not need to go through staging, right?

From my understanding, ram_console/persistent_ram functionality isn't
really the same as pstore/ramoops.

When asked earlier about this, Colin said:
"pstore is just an interface to write to a persistent storage, with a
tiny driver on it to dump console messages after a panic. With the
number of crashes we see on ARM cpus that are not panics (mostly
watchdog resets), logging only after a panic is not appropriate, so it
can't replace ramconsole. persistent_ram could easily be a pstore
driver, but that would not give any new useful functionality, since
ramconsole is better for us than pstore_dump.

Looking at ramoops, it is exactly the same as pstore on ioremapped
ram, but was originally implemented separately. I see patches now
that convert it to a pstore backend, which leaves us in the same state
as above - only logging panics."

As for the long term merging story, I think Colin's idea of using
persistent_ram a backing for pstore, then adapting pstore as needed to
provide similar continuous console logging like ram_console might be a
way forward, but I don't know of anyone actively working on that atm.

Tony: Do you have any thoughts here?

thanks
-john




\
 
 \ /
  Last update: 2012-03-07 23:43    [W:0.262 / U:0.364 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site