[Crash-utility] PATCH 00/10] teach crash to work with "live" ramdump

Dave Anderson anderson at redhat.com
Mon Apr 25 18:30:55 UTC 2016



----- Original Message -----
> Dave, sorry I need to run away. I'll reply tomorrow. And I need to
> actually read your emails first, perhaps I misunderstood you...
> 
> Just one note,
> 
> On 04/25, Dave Anderson wrote:
> >
> > As I see it, this facility is simply another LIVE_SYSTEM memory source,
> > of which there currently are /dev/mem, /proc/kcore and /dev/crash.  The
> > essential difference between them is the pc->readmem plugin:
> 
> Well yes, to some degree...
> 
> But let me repeat, I believe that crash needs something like 1-7 (and probably
> more) anyway. Otherwise it can't work with remote LIVE_SYSTEM correctly.

But the ACTIVE() macro has been the standard since forever, external extension modules
depend upon it, etc., and so I'd strongly prefer to not change it to LOCAL_ACTIVE().
I'd rather keep the handling of this new facility segregated, maybe create something 
like an ACTIVE_QEMU macro/define that can be plugged in wherever you've modified the
ACTIVE() callers where ACTIVE() alone is not enough.

> And while I think that 10/10 can be useful by itself (and least for me), this
> is mostly POC which which allows to test/justify these LOCAL_ACTIVE changes
> in 1-7.

OK good, so you can keep your stuff completely outside of ramdump.c, and not use
is_ramdump(), etc., and then place as many of your changes as possible in a new
file, say something like qemu-live.c.  There may be redundancies between the new
file and ramdump.c, but so be it.  

> 
> Oleg.
> 
> 




More information about the Crash-utility mailing list