Date: Mon, 6 Mar 95 14:33:02 MST From: terry@cs.weber.edu (Terry Lambert) To: davidg@Root.COM Cc: current@FreeBSD.org Subject: Re: "Text file busy" with program not running anymore? Message-ID: <9503062133.AA19806@cs.weber.edu> In-Reply-To: <199503061835.KAA02220@corbin.Root.COM> from "David Greenman" at Mar 6, 95 10:35:12 am
next in thread | previous in thread | raw e-mail | index | archive | help
> The man page says it insures that all disk writes have completed (for the > command) and that modified buffers are written out (for the system call). Yes, > it does indeed still do this (and in fact, has been extended to write out all > modified pages in the system as well). Since the problem mentioned above has > nothing to do with writes completing or modified buffers being synced, I fail > to see how your comment is relevant. Sorry, I was under the impression that the cache was actually flushed as a result. > Once again, the problem is that the pager > of a cached object retains a reference to the vnode while it exists and the > VTEXT flag isn't cleared when the object is put into the object cache. This > shouldn't be a difficult problem to solve. (Although a quick hack I just added > to do this very thing isn't working as expected - which suggests that there > may be a (reference count) problem with objects going into the cache > correctly.) Yes; I think without a split cache, reference count will be a big issue. Like I said in other mail, you can hack this by fetching from the cache, and if you get a hit, consider the reference count one lower than it's actual value and force a cache flush of the object when you unset VTEXT. This is relatively high overhead; I would call it unacceptably high. Terry Lambert terry@cs.weber.edu --- Any opinions in this posting are my own and not those of my present or previous employers.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?9503062133.AA19806>