Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 27 Aug 1999 16:51:20 +0930
From:      Greg Lehey <grog@lemis.com>
To:        Poul-Henning Kamp <phk@critter.freebsd.dk>
Cc:        Matthew Dillon <dillon@apollo.backplane.com>, Christopher Masto <chris@netmonger.net>, "John W. DeBoskey" <jwd@unx.sas.com>, julian@whistle.com, freebsd-current@FreeBSD.ORG
Subject:   Re: -current kernel problems (spec_getpages & vm_fault)
Message-ID:  <19990827165120.Y483@freebie.lemis.com>
In-Reply-To: <3839.935736904@critter.freebsd.dk>; from Poul-Henning Kamp on Fri, Aug 27, 1999 at 08:55:04AM %2B0200
References:  <19990827160950.X483@freebie.lemis.com> <3839.935736904@critter.freebsd.dk>

next in thread | previous in thread | raw e-mail | index | archive | help
On Friday, 27 August 1999 at  8:55:04 +0200, Poul-Henning Kamp wrote:
> In message <19990827160950.X483@freebie.lemis.com>, Greg Lehey writes:
>> On Thursday, 26 August 1999 at 16:25:14 -0700, Matthew Dillon wrote:
>>>>>>      int devminor;					    /* minor number */
>>>>>>
>>>>>>      devminor = minor(dev);
>>>>>> +    dev->si_bsize_phys = DEV_BSIZE;
>>>>>> +    dev->si_bsize_best = BLKDEV_IOSIZE;
>>>>>> +    dev->si_bsize_max = MAXBSIZE;
>>>>>
>>>>> Bingo!  Thank you.
>>>>
>>>> Cool, I expect grog will commit it soon.
>>
>> So did I until I read this message.
>>
>>>     The patch for ccd is not quite right.  Here is the patch from my
>>>     big fat patch at http://www.backplane.com/FreeBSD4/
>>>
>>>     The problem is that you cannot simply set the physical sector
>>>     size to DEV_BSIZE if the underlying device has a larger sector
>>>     size.  If you do, specfs's blocksize alignment (another fix in
>>>     my big fat patch) will be incorrect and result in an I/O error
>>>     on the physical media.
>>>
>>>     For example, swap-backed VN devices have a sector size of one page,
>>>     i.e. 4K.
>>
>> Vinum currently doesn't track the sector size of the drives.  I'll
>> have to put some code in there to DTRT.  In the meantime, I'll set
>> si_bsize_best to 4K in the hope that this will be OK.  Are there any
>> larger block sizes in current use?
>
> 4K is probably too conservative, it only applies if you have a
> swap-backed vn device at the bottom, which would be a truly
> weird thing to do with vinum.

I think I've heard of people doing things like this.  Cybernet were
thinking of it a long time ago, but I think I've talked them out of
it.

> Don't spend too much time on tracking the sector size yet.  I think
> the si_bsize stuff needs some thinking before we do more with it.

OK, I'll set it to 4 kB at the moment and make a note to myself to fix
it when dev_t stops changing.

Greg
--
See complete headers for address, home page and phone numbers
finger grog@lemis.com for PGP public key


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19990827165120.Y483>