Date: Mon, 11 Jan 1999 16:29:38 -0800 (PST) From: Matthew Dillon <dillon@apollo.backplane.com> To: Matthew Jacob <mjacob@feral.com> Cc: Julian Elischer <julian@whistle.com>, freebsd-hackers@FreeBSD.ORG Subject: Re: what is the meaning of this useless code? Message-ID: <199901120029.QAA98242@apollo.backplane.com>
next in thread | raw e-mail | index | archive | help
:> On Mon, 11 Jan 1999, Matthew Jacob wrote:
:> > if (resid > maxio)
:> > resid = maxio;
:> >
:> > return getpbuf();
:> > }
:> >
:> > why is resid being truyncaated and maxio being set when this has no
:> > effect?
:>
:> I think resid was originally an agument to getpbuf
:> but mechanical edits have turned it all into a NOP.
:> Good catch..
:
:Okay.
This is definitely a junk assignment.
:> However there is a related problem..
:>
:> the maximum IO size allowed for a device is kept in the
:> cdevsw entry, which is the WRONG place, because
:> it cannot correctly reflect the case when there are two differnt devices
:> with different limits on the same controller (e.g. two different drives
:> or even two differend controllers on one driver).
:>
:
:Well, there are a number of limits- some are device specific, some are bus
:specific and some are VM/OS specific. It's a bit confusing which applies
:here.
:
:I'm going to ponder this as time permits. I'm trying to track down a hang
:on alpha with larget I/Os and this led me to try and follow the raw I/O
:path which led to this...
The (struct buf *)->b_pages[] array is limited to MAXPHYS bytes worth
of pages, where MAXPHYS is typically 128 KBytes. If you try to stuff
a struct buf with more then that, you will overrun the array and bad
things will happen.
Make sure the devices properly limit the size of the I/O operation they
try to stuff into a struct buf. This is just one of several limits, as
you know.
d_maxio could probably be removed from the kernel entirely with a little
work. I wouldn't worry about it too much, all that really matters is
that the value is large enough such that the transaction rate does not
overburden the device. For a hard disk, anything over 8KBytes tends
to accomplish this.
-Matt
Matthew Dillon Engineering, HiWay Technologies, Inc. & BEST Internet
Communications & God knows what else.
<dillon@backplane.com> (Please include original email in any response)
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?199901120029.QAA98242>