Re: [patch] kernel events layer

[Matt Mackall <mpm@selenic.com>]

On Sun, Jul 25, 2004 at 11:09:51PM -0700, Andrew Morton wrote:
> "Perez-Gonzalez, Inaky" <inaky.perez-gonzalez@intel.com> wrote:
> >
> > If you guys are up to it, I volunteer to write/port such a tool to scan 
> >  out the send_kevent{_atomic,}()s and make a catalog out of it.
> 
> I must say that my gut feeling here is that bolting an arbitrary new
> namespace into the kernel in this manner is not the way to proceed.

An uncontrolled namespace is no better than the existing printk info,
IMO. And I think it's next to impossible to control the kevent
namespace if it's scattered across the tree as strings, having tried
to do something analogous for another large project.
 
> I hope we'll hear more from Greg on this next week - see if we can come up
> with some way to use the kobject/sysfs namespace for this.

An API that looks like sysfs + dnotify to userspace is almost what you
want. While the sysfs namespace has some of the problems above, we're
already stuck with it.
 
> Although heaven knows how "tmpfs just ran out of space" would map onto
> kobject/sysfs.

Per mountpoint sysfs trees? I'm sure there are lifetime issues there.

Btw, we probably already have potential issues with kevents being
stale by the time userspace picks them up - eth0 up, eth0 down, eth1
renamed eth0, userspace notices eth0 up, tries to config downed eth1.

-- 
Mathematics is the supreme nostalgia of our time.