Date: Tue, 5 Aug 1997 08:35:04 +0200 From: j@uriah.heep.sax.de (J Wunsch) To: scsi@FreeBSD.ORG Cc: dkelly@hiwaay.net Subject: Re: Maximum Tape Block Size? Message-ID: <19970805083504.NT06658@uriah.heep.sax.de> In-Reply-To: <199708042257.RAA12906@nexgen.hiwaay.net>; from dkelly@hiwaay.net on Aug 4, 1997 17:57:39 -0500 References: <199708042257.RAA12906@nexgen.hiwaay.net>
next in thread | previous in thread | raw e-mail | index | archive | help
As dkelly@hiwaay.net wrote: > A year or so ago I recall Jordan had an SGI DAT tape with 256k blocksize > which wouldn't read on FreeBSD. Am wondering how this problem was > eventually solved? No, it hasn't. The problem is the way requests are being `sliced' in physio(9). Right now, physio() uses a fixed 64 KB slicing, which is the smallest common denominator for the SCSI controller with the smallest scatter-gather list (AHA1540 with 16 segments per 4 KB in worst case, one page each). At least, that's how i understood it. It should be made more flexible so that controllers with larger scatter-gather lists could be handled efficiently. Those with a small list could perhaps optionally use a large physically contiguous region as a bounce buffer, but that's quite wasteful. Feel free to implement a better scheme. ;-) -- cheers, J"org joerg_wunsch@uriah.heep.sax.de -- http://www.sax.de/~joerg/ -- NIC: JW11-RIPE Never trust an operating system you don't have sources for. ;-)
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19970805083504.NT06658>