Mailing Lists: Apple Mailing Lists

Image of Mac OS face in stamp
 
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: exactly how unsupported is com.apple.kpi.unsupported?



Dean Reece writes:
 > On Jun 2, 2005, at 12:59, Andrew Gallatin wrote:
 > > I'd like to use a symbol (m_get_packets) which I can find only in  
 > > com.apple.kpi.unsupported.  However, the name makes me a little  
 > > nervous, and I was wondering what was meant by "unsupported"
 > >
 > > Granted the name sounds less than encouraging, but at least   
 > > com.apple.kpi.unsupported does not seem to bump its   
 > > OSBundleCompatibleVersion on each minor release.   This makes me  
 > > think that its at least a little more stable the com.apple.kernel.
 > >
 > > What's the policy for changes to things in com.apple.kpi.unsupported?
 > 
 > It roughly translates to "no lifeguard on duty".  We don't make any  
 > promises about the release-to-release compatibility of anything in  
 > the "unsupported" kext.  From a practical standpoint, we tend to  
 > leave things in there alone (I don't think it changed from 10.4.0 to  
 > 10.4.1).  When we do make changes there, it will usually be to remove  
 > symbols.
 > 
 > So:
 > 1) Don't use anything in com.apple.kpi.unsupported.
 > 2) If you do use it, don't expect it to work in the next release
 > 3) If it does work in the next release, goto 2

;)

Thanks for making me look harder.  I think I may be able to get away with using
mbuf_allocpacket().

I'm trying to convert my ethernet driver from allocating a big cluster
for every slot as its used on each interrupt to allocating a large
number of big clusters after some low-water mark is reached.  It looks
like mbuf_allocpacket(how, num_wanted * mbuf_stat.bigmclbytes, ....)
should return a chain of num_wanted big clusters with M_PKTHDR set in
just the first one.  In some sense, this is ideal because for every
cluster but the first in a received jumbo frame, I currently end up
having clear the M_PKTHDR flag.  So this should save me about 3
function calls per 9000 byte packet (remove 2 calls to mbuf_flags,
remove 2 calls to mbuf_setflags, and add one call to mbuf_setflags).

Now if I could just do the same thing for plain mbufs, I'd be a really
happy camper..  In fact, it would be really nice if I could control
the actual allocation size, since I can see wanting to use 2KB
clusters if somebody reduces my MTU to 1500 bytes.

Drew


 _______________________________________________
Do not post admin requests to the list. They will be ignored.
Darwin-drivers mailing list      (email@hidden)
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/darwin-drivers/email@hidden

This email sent to email@hidden

References: 
 >exactly how unsupported is com.apple.kpi.unsupported? (From: Andrew Gallatin <email@hidden>)
 >Re: exactly how unsupported is com.apple.kpi.unsupported? (From: Dean Reece <email@hidden>)



Visit the Apple Store online or at retail locations.
1-800-MY-APPLE

Contact Apple | Terms of Use | Privacy Policy

Copyright © 2007 Apple Inc. All rights reserved.