Resizing SD card via command line

[lumpynose]

Re-imaged my sd card (I'm doing this on a Raspberry Pi) and tried the above; the second part, the resizing with resize2fs fails because the filesystem is mounted read only and it complains about that. The man page for resize2fs says

If the filesystem is mounted, it can be used to expand the size of the mounted filesystem, assuming the kernel supports on-line resizing. (As of this writing, the Linux 2.6 kernel supports on-line resize for filesystems mounted using ext3 and ext4).

Question: (I'm a bit of an old fart and used to administer SunOS and Solaris systems; I don't have a lot of Linux experience so my terminology may not be correct.) Would it be possible to specify the character device for the disk instead of the block device? My experience is that programs that complain about systems being mounted (or not) may work with the character device if the block device had been specified.

[pepedog]

I was dubious this could be done on a mounted system, but it works.
There may be complaints that e2fsck has to be run before or after resize, that works too

[lumpynose]

Here are my latest notes to myself. I've added some reboots and going to single user mode because I'm paranoid:

Before running fdisk I brought the system down to single user and remounted the root partition read-only. In the following the # is root’s prompt.

# systemctl rescue
# mount -o ro,remount /
# fdisk /dev/mmcblk0

Delete the second partition /dev/mmcblk0p2.

d
2

Create a new primary partition and use default sizes prompted. This will then create a partiton that fills the disk.

n
p
2
enter
enter

Save and exit fdisk:

w

Reboot and run resize2fs:

# shutdown -r now
# systemctl rescue
# resize2fs /dev/mmcblk0p2

Your main / partition should be the full size of the disk now. Don’t remount the root partition read-only before you run resize2fs; resize2fs fails if it’s mounted read-only.

After running resize2fs reboot, go to single user, remount root read-only, and run e2fsck:

# shutdown -r now
# systemctl rescue
# mount -o ro,remount /
# e2fsck /dev/mmcblk0p2
# shutdown -r now

[pepedog]

One step I would encourage is before # shutdown -r now, or reboot or poweroff, use sync command

[lumpynose]

One step I would encourage is before # shutdown -r now, or reboot or poweroff, use sync command

Very true. That's why I used shutdown; it does the sync for you.

Edit: or at least shutdown called sync on the systems I'm familiar with, which was bsd, sunos, and solaris.

[lumpynose]

One step I would encourage is before # shutdown -r now, or reboot or poweroff, use sync command

Very true. That's why I used shutdown; it does the sync for you.

Edit: or at least shutdown called sync on the systems I'm familiar with, which was bsd, sunos, and solaris.

/sbin/shutdown is a sym link to /usr/bin/systemctl. I'm guessing that systemctl looks at how it was invoked and when it's invoked as reboot it runs its reboot command. My interpretation is that it effectively does a sync before it reboots. The man page for systemctl doesn't say what happens when it's invoked as reboot. For the reboot command to systemctl the man page for systemctl says

Shut down and reboot the system. This is mostly equivalent to start reboot.target but also prints a wall message to all users. If combined with --force shutdown of all running services is skipped, however all processes are killed and all file systems are unmounted or mounted read-only, immediately followed by the reboot. If --force is specified twice the operation is immediately executed without terminating any processes or unmounting any file systems. This may result in data loss.

I suppose it's possible that systemctl uses --force twice in order to not do the read-only remounting but I would think that very unlikely.

[pepedog]

Which of all those options, including poweroff, turns off most LEDs?
I think sync is flaky with all shutdown methods (all distro), especially where fat is concerned.
Might seem paranoid but I have seen instructions for untarring rootfs and it ends with

Code: Select all

sync
sync
sync
sleep 3
halt

So you make up your mind on that
I have had a pi for 16 months, never had corruption, and always synced. Search posts in Raspberry Pi forum and see how many point to sync as "Solved" soloution

[lumpynose]

It actually depends on the phase of the moon, and sometimes the date. When the moon is waxing you only need 2 syncs. When it's waning you need the 3 syncs. On February 29th you need another sync after the sleep.

[jawbonegroove]

excellent post. can we get this sticky'd...?

[paranoidandroid]

I have tried this on the latest release (archlinux-hf-2013-07-22.img) and the geometry has changed, the disk now has an extended partition with a logical one inside that…

fdisk reports (after imaging onto 8GB card)

Code: Select all

        Device Boot      Start         End      Blocks   Id  System
/dev/mmcblk0p1            2048      186367       92160    c  W95 FAT32 (LBA)
/dev/mmcblk0p2          186368     3667967     1740800    5  Extended
/dev/mmcblk0p5          188416     3667967     1739776   83  Linux


It looks like both the extended & logical partitions need to be deleted (p2 and p5)

Something like…

Code: Select all

fdisk "/dev/mmcblk0"
p
d
2
n
e
(return) # accept default partition no
(return) # accept default start
(return) # accept default end
n
l
(return) # accept default start
(return) # accept default end
p
w

sync; reboot
...
resize2fs /dev/mmcblk0p5
fsck etc


I get the following geometry from fdisk…

Code: Select all

        Device Boot      Start         End      Blocks   Id  System
/dev/mmcblk0p1            2048      186367       92160    c  W95 FAT32 (LBA)
/dev/mmcblk0p2          186368    15652863     7733248    5  Extended
/dev/mmcblk0p5          188416    15652863     7732224   83  Linux


I tried only deleting the last partition but it is limited by the size of the enclosing extended partition, so you can't expand the partition until the extended one is also expanded.

Thanks for all the tips so far, this is handy - hope this helps too