[2150] in SIPB-AFS-requests
Re: afsd: priorities/threads
daemon@ATHENA.MIT.EDU (Bill Sommerfeld)
Tue Sep 19 23:22:37 1995
To: "Richard Basch" <basch@MIT.EDU>
Cc: sipb-afsreq@MIT.EDU, jhawk@MIT.EDU, warlord@MIT.EDU, afsdev@MIT.EDU,
mhbraun@MIT.EDU, jweiss@MIT.EDU
In-Reply-To: Your message of "Tue, 19 Sep 1995 17:47:49 -0400 ."
<9509192147.AA03556@tardis.MIT.EDU>
Date: Tue, 19 Sep 1995 23:12:10 -0400
From: Bill Sommerfeld <sommerfeld@orchard.medford.ma.us>
-----BEGIN PGP SIGNED MESSAGE-----
> It may even help other performance
> since on most systems AFS replaces the UDP handler, and calls the old
> UDP handler only if it is not AFS data
This isn't likely, since (at least in Berkeley kernels..) the upcall
into the UDP handler happens at interrupt level (from a software
network interrupt) and not from the process context of any of the afsd
processes.
> (sick, isn't it...).
The scary thing about that was that Kazar at one point claimed that
doing that *helped* AFS portability, since that interface was more
stable than the internal socket interfaces..
- Bill
-----BEGIN PGP SIGNATURE-----
Version: 2.6.1
iQCVAwUBMF+Gh7T+rHlVUGpxAQET6gP/XM1w041neSzzOM9YCTDvTpnpQDuNR4MW
q/B5/Y08iaTB0I0rcLm1BO4Hxgv7V04kuUeTCz8zjwCbJLRlt/lmDtf4nf5uHFRl
0TQwf6qfMp/e2Q87Odao6sRfYUJUaeqebJQd+72NoR1K/np6/QY1tT0mCR71fBjO
GqZAnhAfNxw=
=ybjD
-----END PGP SIGNATURE-----
