Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 25 Jul 2003 01:41:33 -0600
From:      Scott Long <scottl@freebsd.org>
To:        Ruslan Ermilov <ru@FreeBSD.org>
Cc:        cvs-all@FreeBSD.org
Subject:   Re: cvs commit: src/release/i386 drivers.conf
Message-ID:  <3F20DF2D.8030209@freebsd.org>
In-Reply-To: <20030725053806.GA36982@sunbay.com>
References:  <200307250010.h6P0AYKW061085@repoman.freebsd.org> <20030725053806.GA36982@sunbay.com>

next in thread | previous in thread | raw e-mail | index | archive | help
Ruslan Ermilov wrote:
> On Thu, Jul 24, 2003 at 05:10:33PM -0700, Scott Long wrote:
> 
>>scottl      2003/07/24 17:10:33 PDT
>>
>>  FreeBSD src repository
>>
>>  Modified files:
>>    release/i386         drivers.conf 
>>  Log:
>>  Move the mlx driver back into the kernel so that the drivers floppy doesn't
>>  overflow.
>>  
>>  Revision  Changes    Path
>>  1.29      +0 -1      src/release/i386/drivers.conf
>>
> 
> Which reminds me.  The floppy wasn't overflowing when building
> 5.x snapshot on 4.x, due to the differences in file system
> layout between 4.x and 5.x.  Please see the attached if you're
> interested.
> 
> 
> Cheers,
> 
> 

Now that -O2 is the default for newfs in 5-current, does the 'Oflag == 
2' trick still work?  Even if it does, it feels like a hack to me.  I'd
rather that newfs have an option to directly override the number of
c/g's, rather than rely on indirect calculations and magic constants.
Please don't commit this until we have discussed it more.

Scott


> ------------------------------------------------------------------------
> 
> Subject:
> Re: MINCYLGRPS (was: 4-stable releases on -current?)
> From:
> Ruslan Ermilov <ru@FreeBSD.org>
> Date:
> Wed, 12 Mar 2003 00:32:21 +0200
> To:
> Poul-Henning Kamp <phk@FreeBSD.org>, Kirk McKusick <mckusick@FreeBSD.org>
> 
> 
> Hey, is anybody ever going to answer this email?
> 
> On Wed, Feb 26, 2003 at 02:48:10PM +0200, Ruslan Ermilov wrote:
> 
>>On Tue, Feb 25, 2003 at 09:52:02PM +0200, Ruslan Ermilov wrote:
>>
>>>On Tue, Feb 25, 2003 at 08:02:19PM +0200, John Hay wrote:
>>
>>[...]
>>
>>>>Ok, with the patches below I can get to where it tries to build the
>>>>fixit floppy in release.10. It breaks because the floppy is too small.
>>>>I must still check to make sure it is true.
>>>>
>>>
>>>We'll know this in less than 1:30 -- I've just launched the
>>>snapshot build for 4.x/i386 on my fast -stable box.
>>>
>>>While I was writing it, it's already finished (successfully).
>>>
>>>ftp://ftp.sunbay.net/pub/FreeBSD/snapshots/i386/4.x-20030225-STABLE/
>>>
>>
>>OK, I've tracked it down to the differences in newfs(8) between
>>4.x and 5.x.  In 4.x, newfs'ing a 1.44MB floppy results in a
>>single cylinder group, but in 5.x there's a thing called
>>MINCYLGRPS, which results in fewer free space on a floppy:
>>
>>: # uname -r
>>: 4.7-STABLE
>>: # ./x
>>: Warning: Block size restricts cylinders per group to 6.
>>: Warning: 1216 sector(s) in last cylinder unallocated
>>: /dev/vn0c:	2880 sectors in 1 cylinders of 1 tracks, 4096 sectors
>>: 	1.4MB in 1 cyl groups (6 c/g, 12.00MB/g, 384 i/g)
>>: super-block backups (for fsck -b #) at:
>>:  32
>>: Filesystem 1K-blocks    Used   Avail Capacity iused   ifree %iused  Mounted on
>>: /dev/vn0        1363       0    1363     0%       1     381    0%   
>>
>>: # uname -r
>>: 5.0-CURRENT
>>: # ./x
>>: /dev/md0c: 1.4MB (2880 sectors) block size 4096, fragment size 512
>>: 	using 4 cylinder groups of 0.36MB, 91 blks, 128 inodes.
>>: super-block backups (for fsck -b #) at:
>>:  32, 760, 1488, 2216
>>: Filesystem 1K-blocks    Used   Avail Capacity iused  ifree %iused  Mounted on
>>: /dev/md0        1311       0    1311     0%       1    509    0%   
>>
>>With this patch to newfs(8),
>>
>>: Index: mkfs.c
>>: ===================================================================
>>: RCS file: /home/ncvs/src/sbin/newfs/mkfs.c,v
>>: retrieving revision 1.74
>>: diff -u -p -r1.74 mkfs.c
>>: --- mkfs.c	22 Feb 2003 23:26:11 -0000	1.74
>>: +++ mkfs.c	26 Feb 2003 11:20:57 -0000
>>: @@ -317,7 +317,7 @@ mkfs(struct partition *pp, char *fsys)
>>:  	for ( ; sblock.fs_fpg < maxblkspercg; sblock.fs_fpg += sblock.fs_frag) {
>>:  		sblock.fs_ipg = roundup(howmany(sblock.fs_fpg, fragsperinode),
>>:  		    INOPB(&sblock));
>>: -		if (sblock.fs_size / sblock.fs_fpg < MINCYLGRPS)
>>: +		if (sblock.fs_size / sblock.fs_fpg < (Oflag == 2 ? MINCYLGRPS : 1))
>>:  			break;
>>:  		if (CGSIZE(&sblock) < (unsigned long)sblock.fs_bsize)
>>:  			continue;
>>
>>I still don't get the same picture as on 4.x, but it's now at least
>>sufficient to make release.10 happy.
>>
>>: # ./x
>>: /dev/md0c: 1.4MB (2880 sectors) block size 4096, fragment size 512
>>: 	using 1 cylinder groups of 1.41MB, 361 blks, 416 inodes.
>>: super-block backups (for fsck -b #) at:
>>:  32
>>: Filesystem 1K-blocks    Used   Avail Capacity iused  ifree %iused  Mounted on
>>: /dev/md0        1359       0    1359     0%       1    413    0%   
>>
>>John, I'm sending you the complete patchset in another email.
>>
>>
>>Cheers,
>>-- 
>>Ruslan Ermilov		Sysadmin and DBA,
>>ru@sunbay.com		Sunbay Software AG,
>>ru@FreeBSD.org		FreeBSD committer,
>>+380.652.512.251	Simferopol, Ukraine
>>
>>http://www.FreeBSD.org	The Power To Serve
>>http://www.oracle.com	Enabling The Information Age
> 
> 
>>FSIMG=fixit.flp
>>FSSIZE=1440
>>FSLABEL=fd1440
>>FSINODE=4000
>>
>>dd of=${FSIMG} if=/dev/zero count=${FSSIZE} bs=1k 2>/dev/null
>>
>>case `uname -r` in
>>4.*)
>>	DEVICE=vn0
>>	vnconfig -s labels -c /dev/${DEVICE} ${FSIMG}
>>	;;
>>5.*)
>>	DEVICE=`mdconfig -a -t vnode -f ${FSIMG}`
>>	;;
>>esac
>>
>>disklabel -w -B ${DEVICE} ${FSLABEL}
>>newfs -i ${FSINODE} -o space -m 0 /dev/${DEVICE}c
>>
>>#disklabel ${DEVICE}
>>df -i /dev/${DEVICE}
>>
>>case `uname -r` in
>>4.*)
>>	vnconfig -u ${DEVICE}
>>	;;
>>5.*)
>>	mdconfig -d -u ${DEVICE}
>>	;;
>>esac
> 
> 
> 
> 
> 




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3F20DF2D.8030209>