Date: Thu, 26 Jul 2001 10:15:12 -0700 From: Terry Lambert <tlambert2@mindspring.com> To: Julian Elischer <julian@elischer.org> Cc: Zhihui Zhang <zzhang@cs.binghamton.edu>, Bosko Milekic <bmilekic@technokratis.com>, vishwanath pargaonkar <vishubp@yahoo.com>, freebsd-hackers@FreeBSD.ORG Subject: Re: cluster size Message-ID: <3B605020.9D2B1DB2@mindspring.com> References: <Pine.BSF.4.21.0107251404390.4973-100000@InterJet.elischer.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Julian Elischer wrote: > > no.. it has to do with the fact that it would be unwise > to make a cluster > 1 page size since we have no guarantee that > all drivers could handle breaking up a DMA if a cluster spanned 2 > physical address ranges. (they can handle a chain of discontinuous > mbufs but may assume that a single mbuf will have physically > contiguous data. Now since we cannot span a page boundary, > we should fit in exacly to get as much room as possible > and since (pagesize/3) is too small, the next possibility is (pagesize/2). FWIW: the way clusters work quarantees contiguity, based on the zone allocator being backed by contiguous regions in single pages. This is what Julian said, but it's less verbose... 8-). -- Terry To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3B605020.9D2B1DB2>