Date: Sun, 2 Apr 2000 16:26:37 -0700 (PDT) From: Matthew Dillon <dillon@apollo.backplane.com> To: Poul-Henning Kamp <phk@critter.freebsd.dk> Cc: cvs-committers@FreeBSD.ORG, cvs-all@FreeBSD.ORG Subject: Re: cvs commit: src/sys/kern subr_devstat.c src/sys/ufs/ufs ufs_disksubr.c src/sys/sys buf.h devicestat.h disklabel.h Message-ID: <200004022326.QAA51447@apollo.backplane.com> References: <16382.954714834@critter.freebsd.dk>
next in thread | previous in thread | raw e-mail | index | archive | help
:
:In message <200004022223.PAA51020@apollo.backplane.com>, Matthew Dillon writes:
:
:> b_data is integral to buffer cache KVA management & operation and
:> should be in the main struct buf structure, not the bio structure.
:
:No, b_data is (currently) also a property of the I/O request.
No it isn't. b_data is the property of the buffer cache. If you
screw with it, you blow up the buffer cache. Any device driver that
screws with it must restore it prior to reentering the buffer cache
code.
Several devices drivers mess with b_data ... but they really shouldn't.
Just because they do is no justification to move b_data into the bio.
If you want you can duplicate b_data ... that is, have the buffer cache
managed KVA be a field in struct buf and copy it into a new field in the
bio which the device drivers use (and get rid of the save field that
the device drivers currently use to save/restore b_data). But that isn't
what you've done.
-Matt
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe cvs-all" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200004022326.QAA51447>