Date: Sat, 18 Jul 1998 14:18:57 -0700 (PDT) From: Julian Elischer <julian@whistle.com> To: Spidey <beaupran@JSP.UMontreal.CA> Cc: Greg Lehey <grog@lemis.com>, Question=answer <freebsd-questions@FreeBSD.ORG> Subject: Re: Howto: UNUSED data on HD, Melting 2 partitions into 1 Message-ID: <Pine.BSF.3.95.980718141150.21408A-100000@current1.whistle.com> In-Reply-To: <Pine.BSF.3.95.980718111821.17992A-100000@current1.whistle.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, 18 Jul 1998, Julian Elischer wrote:
> Sorry for taking so long to get back to you..
>
> Some of this should go into the FAQ I think..
>
> Greg is both right and wrong..
> It's dangerous and trick, but it can be done..
>
> There are some rather extreme shortcuts that can be made..
>
> I have done far worse stuff than this..
> I once shifted the root partition while I was running on it :-)
> but it's been a couple of years since it tried this so beware..
>
>
> firstly terminology..
> I will ALWAYS refer to the partitions defined by block0 (the MBR) as
> slices.
> I will always refer to the partitions defined in a BSD disklabel as
> partitions..
>
> Slices are defined by a table in the MBR. the program 'fdisk' prints out
> the contents of this table. (or lets you edit it). There are rows in the
^ 4
> table. Each row is capable of defining a portion of disk. This portion
> becomes the slice of that number.. e.g. the first entry in the table
> becomes slice 1, the 2nd becomes slice 2 etc.
>
> The position of each slice on the disk is totally independent of the order
> in which they appear in the table. So the last entry in the table may
> point to the first section of disk and the first may point to a portion of
> disk beyond that pointed to by the last. If an entry in the table points
> to nothing (i.e. is all zeroes) it is shown as <UNUSED>. any portion of
> the disk can be pointed to by 0, 1 ,2 ,3 or 4 entries in the table. it is
> usually considered an error to have a block included in 2 or more entries,
> but it can be done..
>
> If the 'type' of a slice is '386BSD' then the system will further look
> inside that slice to see if there is a 'disklabel' (on block 1 (the second
> block) of that slice). If there is, it will read that disklabel and
> extract the "partition table" from it. This is similar in concept to the
> table I mentioned before. There are 8 rows to the table, each of which
> defines a partition. You may leave a row empty in which case it wil not
> appear in the disklabel when printed. the first entry in the table defines
> partition 'a', the second defines partition 'b', etc.
>
> One quirk of the present disklabels is that the 'on disk' format of the
> disklabels defines those partitions with block numbers relative to the
> BEGINNING OF THE DISK, rather than relative to the beginning of the slice
> in which the disklabel resides. (in the future this may change) This is
> a historical bogosity, which is hidden from you when you look at the
> disklabel, by a huge programming kludge.. Another thing to remember is
> that by definition the 3rd row of the table (partition 'c') is magic, and
> you should always make it equal to the whole slice on which you have the
> disklabel. In some cases if this is not so, the system might even correct
> it before showing it to you. I try to discourage people from using the 'c'
> partiton, because the 'magic' may change some day in which case they would
> be screwed..
>
> Once again, any block in the slice can be pointed to by 0 , 1, 2 .... 8
> tables. The numbers 1, or 2 being legal. A patch of disk bay be pointed to
^ entries
> by 0 if it is at the end of the slice and beyond the 'c' part. This should
> only happen in odd cases. (like there is a bad block out there you want to
> avoid). If a block is not being used in any real partition, it should
> still be within the range pointed to by the 'c' part.. This is the '1'
> case. The normal case is if it is pointed to by a single 'normal' part,
> and of course is also within the range of the 'c' part. This is the '2'
> case.
>
>
>
> Now having said that.. lets look at your questions..
>
> On Sat, 18 Jul 1998, Spidey wrote:
> >
> > First, I must ask another question prior to the others. Could somebody
> > explain these outputs of fdisk and df?
> > >>>>>>>>>>>>>>prompt>df
> > Filesystem 1K-blocks Used Avail Capacity Mounted on
> > /dev/wd0s2a 31775 16238 12995 56% /
> > /dev/wd0s3 1023856 942944 80912 92% /dos
> > /dev/wd0s2f 1132103 989859 51676 95% /usr
> > /dev/wd0s2e 29727 24393 2956 89% /var
> > procfs 4 4 0 100% /proc
> > >>>>>>>>>>>>>>>prompt>fdisk
> > ******* Working on device /dev/rwd0 *******
> > parameters extracted from
> > in-core disklabel are:
> > cylinders=839 heads=128 sectors/track=63 (8064 blks/cyl)
> >
> > parameters to be used for BIOS calculations are:
> > cylinders=839 heads=128 sectors/track=63 (8064 blks/cyl)
> >
> > Media sector size is 512
> > Warning: BIOS sector numbering starts with sector 1
> > Information from DOS bootblock is:
> > The data for partition 1 is:
> > <UNUSED>
> > The data for partition 2 is:
> > sysid 165,(FreeBSD/NetBSD/386BSD)
> > start 3072384, size 3690312 (1801 Meg), flag 80
> > beg: cyl 381/ sector 1/ head 0;
> > end: cyl 838/ sector 24/ head 80
> > The data for partition 3 is:
> > sysid 6,(Primary 'big' DOS (> 32MB))
> > start 8064, size 2048256 (1000 Meg), flag 0
> > beg: cyl 1/ sector 1/ head 0;
> > end: cyl 254/ sector 63/ head 127
> > The data for partition 4 is:
> > <UNUSED>
> > >>>>>>>>>>>>>>>>>>>>>>>>>
> >
> > Few mysterieies:
> > 1) It is said clearly that my FBSD partition is 1801Mb in size. How come
> > that the sum of my df numbers is only 1193609Kb (?=1.2Gb)???? I miss
> > about 700Mb here!!!!
>
> you didn't show the disklabel output.. You have defined a BSD slice but we
> can't se what you defined the internal structure of that space to be
> like..
>
>
> >
> > 2) Why isn't fdisk giving the size of <UNUSED> partitions?
>
> fdisk only shows the contents of the table..
> the unused regions of the disk are unused exactly because they are not in
> the table..
> You can only discover the unused portions by inspection..
> they are:
> sectors 1->8063 (4 MB) Microsoft wastage.
> sectors 2056320 -> 3072384 (~500MB)
> sectors 6762696 -> ??? (possibly 0 MB, as I don't know the number
> of sectors in your drive.)
>
> There are also sections of the BSD slice that are not pointed to by any
> real disklabel entry (at least none that you have mounted.) As you
> haven't show the output of disklabel I can't give you exact help, but you
> shoud do 'disklabel -e -r /dev/rwd0s2' and add a table entry on the end
> that points to the 500MB that you haven't used in that slice. Make sure
> you also update the 'number of partitions' entry as well or the new entry
> in the table will not be used.. (maybe this is hte mistake you already
> made?)
>
> >
> > 2.2) Therefore, how can I identify What are these <UNUSED> partitions? I
> > know one is 500Mb of free space, but the other????
>
> No you are confused.. these are just empty entries in the table.
>
>
> >
> > Now back to my reply...
> >
> > On Sat, 18 Jul 1998, Greg Lehey wrote:
> >
> > > On Friday, 17 July 1998 at 8:00:44 -0400, Spidey wrote:
> > > > Hi!
> >
> > > > 1 GB DOS
> > > > 500 MB Free
> > > > 1750 MB FBSD
> > > >
> > > > Now I wish to fit in the 500 MB that is before the FBSD. Is it a different
> > > > procedure?
> > >
> > > >From what?
> >
>
> yes it is..
>
> first:
> do fdisk and disklabel and PRINT THEM OUT
> you'll need them to recover if we screw up..
>
> TO add space to the FRONT of a FreeBSD slice we need to make use of the
> fact that the Partition table entries in a disklabel are **based from
> block0 of the DISK**. (remember I mentioned this before?)
> This means that no matter where the disklabel is in the disk, it still
> refers to the same portions of disk. This is actually bad in all cases
> except this single case..
>
> what you should do is get the 'fixit' floppy and the boot floppy
> from the 'floppies' directory in the distribution.
>
> Boot from them
> (boot from the boot floppy and select the 'fixit' option
> and then follow directions.)
>
> You could also do this from single-user mode but if you get it wrong you
> need to be able to fix it :-)
>
> in either case the procedure is the same
>
> ########## WARNING.. GROSS HACK!! ##############################
> fdisk -u /dev/rwd0
>
> edit the LAST table entry (presently <UNUSED>)
> give it the following values:
>
> start: 2056320
> size: 1016064
> ending: 3072383
> type 165
>
> this will define a new BSD partition that points to the unused space
> below the BSD partition.
> but it is LATER in the table so it won't be the default BSD part..
> it will show up after a reboot as /dev/rwd0s4
>
> -reboot- in the same way you just did..
> copy the bootblocks and disklabel from the existing BSD partition to
the
> new one.
> dd if=/dev/rwd0s2 of=/dev/rwd0s4 count=16
> this copies the first 8k. you now have a disklabel in that slice.
> and becasue the blocks are refered to in ABSOLUTE TERMS the table entries
> for the partitions still point to the starts of the partitions in the
> other slice.
>
> reboot: (may not be needed but......)
>
> disklabel -r /dev/rwd0s4
> we should get some valid info but it will probably complain mightily.
>
It occurs to me that this may be enough for you..
The new space doesn't need to be part of the existing BSD slice..
we can simply add another.
Using disklabel -e
you can correct the partitions here so that only partition c and d exist,
and both cover the entire SLICE. then newfs the new 'd' partition.
newfs /dev/rwd0s4d
this would be all you ned to do..
enter it into the /etc/fstab as wd0s4d
you should be done..
> If code has been added to munge the disklabel when we read it from the
> other partition (since I tried this last), all the offsets would be wrong
> and the 'a' part would start at 0 instead of 500MB in. CHECK IF THIS IS
> SO. If so we need to fix the offsets. let me know and we'll ammend the
> instructions..
>
> TRY use the disklabel -e command below and fix the fields mentionned
> below, if it won't let you, do the next boot and continue..
>
> Up until this point you should still have a running system.......
>
> Now come the part when we pull the carpet out from under our own feet..
> do fdisk -u /dev/rwd0
> and expand the last part to size 4706376
> and set the type, size and start of the old BSD part to 0
>
> reboot again. (if it doesn't boot into single user mode, use the fixit
> disk/boot disk combo.)
>
> disklabel -er /dev/rwd0s4
> fix the size of the 'c' partition
> fix the 'sectors per unit' entry.
>
> reboot
>
> THEORETICALLY you should now have a single bsd slice, with the existing
> partitions in it with a bunch of free space before them
>
> use dikslabel -e -r /dev/rwd0s4
> to add a new partition to use the new 500MB space
>
> use newfs to put filesystems on both new partitions (teh noew one at teh
> front and the new one at the end.
> add entries in /etc/fstab to use them.
>
>
> IF this fails you can recover by:
> booting the fixit way,
> doing
> fdisk -u /dev/rwd0
> and typing in all the entries that were there before..
>
>
> let me know how it goes!
>
> julian
>
>
>
>
>
>
> To Unsubscribe: send mail to majordomo@FreeBSD.org
> with "unsubscribe freebsd-questions" in the body of the message
>
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-questions" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.3.95.980718141150.21408A-100000>