Brian Utterback writes: > > > [EMAIL PROTECTED] wrote: > > > > > I suspect that Prakash is suggesting that there is a perception > > that 'ndd -set /dev/ip ip_debug 10' is somehow safer than > > "echo 'ip_debug/W 0xa'|mdb -kw". > > It is clearly safer to use ndd than "mdb -kw" in general.
I'm not so sure. Some ndd commands will just trash a driver or even the system itself. Others will enable debug modes so horrendous as to bring a system to its knees -- and render it useless. Still others change the way locking or threading or function calls are done in some particular subsystem, with unpredictable results. Yes, mdb -kw has the guards off the business end of the saw. ndd, though, has an undeserved reputation for being "safe." > The ndd > command gives you a limited set of variables to change with type > checking and sanity checking. The mdb form allows you to write to any > variable, with no sanity checking of the range and type. Getting the > case wrong in the command could be a disaster. And I think that there > is a presumption that the set variables provided by ndd are at least > safe to change on a running system. You'd think. :-/ Most of them effectively do little more than what the equivalent mdb command would do. In any event, the higher level issues are that (a) tuning shouldn't be needed on any well-designed system and (b) to the extent that it *is* needed anyway, passing plain text strings between user space and the kernel is an awful "framework." I realize we've forced customers to become wedded to ndd, but that alone doesn't really argue too strongly against killing it. -- James Carlson, Solaris Networking <[EMAIL PROTECTED]> Sun Microsystems / 35 Network Drive 71.232W Vox +1 781 442 2084 MS UBUR02-212 / Burlington MA 01803-2757 42.496N Fax +1 781 442 1677 _______________________________________________ networking-discuss mailing list [email protected]
