[email protected] wrote:
On (07/01/09 17:41), Garrett D'Amore wrote:
Like it or not, the fact is that there *are* devices that need unique
tunables, either for performance or (in some cases) even functionality
reasons.
I think we all agree that driver-private tunables are a necessar/unavoidable
evil, but the question is whether dladm should print these by default,
or even by some extensions. Printing the values by default doesn't sound
like a good idea - the idea behind dladm was to provide uniformity
and predictability, and this would be counter to that goal.
whether we should support something like "ndd -get /dev/<..> \?" is
a different question. I concur with Meem that, since the private properties
are really driver specific, information about their syntax and semantics
should be provided with the driver's man page or other documentation
Of course they *should* be documented in the documentation!
But that's not sufficient. For example, imagine a driver that supports
two revs of a chip. One has a tunable for the rx fifo, and the other
does not. How does a sysadmin figure out whether the tunable is present
or not?
I agree that these shouldn't be displayed by default. But to suppose
that nobody should ever want to list the device private tunables out
suggests living in a mythical utopia where device private tunables are
*never* needed. That's not the real world we live in.
If you want ndd to go away, then you need to provide *full* replacements
for all of ndd's features, including \?.
-- Garrett
--Sowmini
_______________________________________________
networking-discuss mailing list
[email protected]
