--3V7upXqbjpZ4EhLz
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline

Dear folks,

as some of you might know i'm currently implementing an UDF filingsystem 
for NetBSD's userland and eventually kernel. Since i want to make the 
userland as close as possible to a preliminary kernel implementation i'm 
stuck with the following questions: 

a) If all user-data and meta-data (not nessisary including nodes) are 
managed by the NetBSD kernel, what part decides to shedule reads and writes 
i.e. push out dirty buffers and on what criterea?

b) Or are all bufs created by read and write operations are passed to a 
filingsystem's vop_strategy() anyway and its up to the filingsystem to 
order the reads and writes at will and maintain whatever bufs it want? If 
memory gets short the kernel calls a fsync() on a few nodes maybe but thats 
it?

Esp. with the introduction of the buffercache i'm a bit at a loss on how 
its done nowadays... and i dont like to reverse-engineer that much possibly 
leading to wrong conclusions...

Anyone got some pointers? (the BSD book i dont have...yet)

Cheers,
Reinoud


--3V7upXqbjpZ4EhLz
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.6 (NetBSD)

iQEVAwUBQrh8poKcNwBDyKpoAQIzvAf/U4iO7hm57f7qrJJl1UkI+9k7SuouFh2Y
trYnQBg1mtAvlZzkF5Jv0LSASb2hq0fw5oK6uFCowM8RVD4rPvNjTgHRwBt5/XmU
AkHqT/4x3pmsH4ORpu4yD0T9Gnp44wFGUPjzQBEMakZMEgvY+xLDKFBTSfHTxTt5
Dcw6uzS9VOnPhajrDfiJJi0VO2oKd/egP+jchdAPFLAuOXWDSWCRBygnSOHGVd+r
suSuXtUTcNLvEYMPy7rcI32EEJOsVAihd/iIp0v/COsADc5qp5Fm4ysCQB7iR9ma
UUZ2SBA+e6HZ35nUbwP90uvX23HSBuYxpLuN9605M0aE7KkH+RgYXQ==
=8ArU
-----END PGP SIGNATURE-----

--3V7upXqbjpZ4EhLz--