Unit of Advanced Study
University of Pretoria, South Africa
This document describes procedures for installing USB Flash Memory devices and the way of formatting them for various file systems such as vfat (the way they usually come) and ext2 (the way they fit in better with Linux). It is also described how to partition the device in two sections with different file types.
The purpose of this document is to describe procedures for implementing USB flash memory devices (memory sticks) on Linux. This entails the following:
The procedures described in this document should work on any USB memory stick device which can be operated as a read-write device. Read-only devices do not fall in this category. The procedures will not work if the write facility of the device is disabled (i.e. if the write protect switch is on).
The procedures described for formatting and maintenance in particular filesystems may (in principle) be adapted for other filesystems such as ext3 and reiserfs. These have not been tried and tested. Neither has any attention been paid to adaptation of the procedures to other related operating systems.
Flash memory devices come mainly in the form of small, portable devices often referred to as memory sticks or keychains, and as part of digital cameras. They are non-volatile devices which operate on the principle that electric charges are used to represent data in binary format. These charges can remain unchanged almost indefinitely, but changes (such as writing to the device) limit the life span of the device (100000 writes of 8MB each).
Memory sticks plug directly into a USB-port at the back of your computer. The power it needs is supplied by the USB-port. It is sometimes convenient to use a USB-extension cable to bring the device within easy reach.
Memory sticks usually have write protect switches which should be turned off if you want to use it as read-write device. If the switch is in the "on" position, the device is read-only. The devices are usually equipped with a led (light emitting diode), which indicates that the device is operational. The led also flickers while data is being transferred to or from the device.
Digital cameras use flash memory for image storage. These are powered by the camera and connected to the computer's USB-port via an electronic interface. To be able to connect to the computer, the camera's power must be turned on for the duration of the interfacing, and should only be turned off after the device had been unmounted from the computer. It is important to remember that the device takes power from the camera's battery. The operation should therefore be terminated as soon as possible.
Important considerations when buying a memory stick are the following:
Users of this document might want to acquaint themselves with the following: 
This is a description of the salient features of the setup used to develop the procedures described below. All the procedures have been tried and tested, also for Linux-2.6. The screen-like displays are precise copies of what appeared on my screen.
It is uncertain if USB-support is sufficient in kernels earlier than 2.4.xx. The following support, relevant to this document, was compiled into the kernel. A modular approach may also be followed.
In the lists above the entities in uppercase refer to the variable names to be found in the .config file in the upper level directory of the kernel source (/usr/src/linux/). The entities xxx.o refer to the modules created when a modular approach is followed. When there is no reference to a module, the option can only be hard-compiled into the kernel.
Different kernel versions may have different indications of options when, for example, make menuconfig or make xconfig are run. Variables such as CONFIG_USB, which can be gleaned from the various help options, may be a more reliable indication.
Very recent Linux distributions such as RedHat and SuSE probably have the appropriate kernel options compiled in.
Under USB-support, options for a number of digital cameras are available.
Please consult the relevant texts as set out in Section 4 if you consider (re)compiling your kernel.
In this section we discuss in a very brief manner, some basic Linux concepts which should contribute to an understanding of the procedures described below.
The /proc filesystem serves as a window through which we can see the workings of a Linux setup. The objects of most interest for this document are the directories/proc/bus/usb/ and /proc/scsi/. These will be used to verify that the kernel had been set up properly for our purpose (Section 7.2).
Your memory stick will be considered as a USB mass storage device posing as a removable SCSI disk (sd). SCSI disk devices are mapped to the /dev (devices) directory under /dev/sda , /dev/sdb , ... When different disk devices are present, they will be mapped to /dev/sda, /dev/sdb, etc. If, for example, a memory stick and a digital camera are plugged in, the one would be mapped to /dev/sda and the other to /dev/sdb. Tests have indicated that the first device to be detected is mapped to sda and represented in the directory /proc/scsi/usb-storage-0. The second will be mapped to sdb and represented in /proc/scsi/usb-storage-0. What will happen when both devices are present at boot-up, is not known. An overview of the partitions relevant to this document, the file /proc/partitions may be consulted. Look for entries like sda, sda1, sdb. Under the Linux-2.6 series of kernels this problem may be very neatly resolved (Section 12).
In the descriptions further on, it will be assumed that only one memory stick is present, and therefore mapped to /dev/sda.
In most distributions these points exist in abundance. You should check that this is the case in your setup (ls /dev/sda*. If there are none you can create them by doing (as root, indicated by the hash).
etc, up to /dev/sda15 if necessary. The device as a whole is mapped to /dev/sda, while /dev/sdax (x = 1 ... 15) represent the various partitions of the device. We shall describe, later on, how to create different partitions for different file systems (see Section 8 and Section 9 ). If you intend to use your memory stick with a single partition (filesystem), /dev/sda1 is sufficient.
This dynamically generated filesystem should be mounted at /proc/bus/usb/. It is therefore essential that this mount point exists. Once it is mounted, there should be more to be seen in /proc/bus/usb/ and in /proc/scsi/ (Section 7.3). To find out which USB devices are present, read the file /proc/bus/usb/devices ( less /proc/bus/usb/devices). It takes a small effort to interpret the writing on the screen, but it is not difficult. A memory stick is indicated by Mass Storage Device.
The ext2 (second extended) filesystem still is the most prevalent in Linux setups. It is most versatile and sophisticated, carrying with it permissions (read-write-execute, who is permitted to do what), ownership (user, group, others), a timestamp (when last modified), etc. Moreover, tools for the proper maintenance of a device endowed with ext2, exist (Section 11). If a flash memory device is to be used exclusively within Linux setups, it is therefore desirable to format it in ext2 (see Section 8).
Enabling vfat in the kernel makes it possible to mount filesystems created under Dos/Windows to be mounted in a Linux system. Most memory sticks are formatted for direct use on Windows and should therefore be considered as vfat-formatted. The vfat filesystem is less sophisticated than ext2 and will therefore be more economic as far as memory space is concerned. It does not carry permissions in the same way as ext2 with the result that an ext2-file saved in vfat will re-appear with its permissions changed. Nonetheless, if a memory stick is to be used for the purpose of carrying data between Linux and Windows machines, it is best to keep it in the vfat format. A compromise is to partition the memory stick in two parts: one in vfat and the other in ext2. At least the Linux machine should be able to handle both. This is described in Section 9
Here is a short list of command options that occur in this document:
It is possible that your system is already set up for handling a USB-flash memory device. To check this, simply issue the command $ mount from an X-terminal. If something like
appears among the output, you are ready to go to Section 7.4 and proceed from there. It may be virtuous, however, to glance through the intervening sections. If the test is unsuccessful, all is not lost. Please read on.
A few things in the /proc directory can be checked so see if the kernel options have been properly included or the appropriate modules properly inserted. As a first check, see if the directory /proc/bus/usb exists. If it does, your kernel supports the USB-subsystem correctly. If this is not the case the kernel has to be re-compiled with appropriate USB support (see Section 5.2.3) or the kernel has to be updated. Next, check if the directory /proc/scsi exists. If it does, you're well-away. If it doesn't, SCSI support has not been compiled into your kernel (see Section 5.2.1).
In /proc there should be provision for a mount point for the USB filesystem. That point is /proc/bus/usb. If it exists, the kernel is correctly set up.
If the verifications of the previous sections were positive, the next step is to mount the USB filesystem. This is issued from root as follows:
If there are no complaints, do some more tests. The first is the quick test (Section 7.1). A deeper probe would be to issue the command # ls -l /proc/bus/usb . This should give something similar to
The path /proc/scsi/usb-storage-0/ should now exist and one of the files at the end will show this. In my setup the command $ less /proc/scsi/usb-storage-0/1 gives
If the flashdrive is present, the last line will have the entry "Yes". If it is not plugged in, the entry will be "No".
You are now ready to find out if the memory stick is working. Let us suppose that you bought it over the counter. Its instruction leaflet will most likely tell you how to set it up for Windows. That is an indication that it is formatted in vfat. Before you try to mount it, create a mount point for it. Do something like # mkdir -m 777 /mnt/memstick. The mounting command would most likely be
If everything went smoothly, you should be able to see the device: # ls /mnt/memstick.
Now try some standard things like making a directory on the device and copying a favourite text file to it:
List again (# ls -l /mnt/memstick ) and pay attention to the permissions.
Unmount the device (# umount /dev/sda1 ) and mount it again as above. List again and check the permissions. Most likely your favourite text file will now have an x-permission. It became executable. That is normal in the vfat filesystem. If you are happy with that, unmount the device and skip to Section 10 .
The procedures described here will give the entire flash memory device a single partition endowed with the ext2 filesystem. This will make the device more suitable for use between Linux machines. Don't do this if the device is supposed to operate between Linux and Windows machines.
It is assumed that the flash memory is mapped to /dev/sda. In this section we treat the situation where the whole flash memory device will become devoted to a single ext2 partition. In the example procedure shown below, a 128MB flash memory was formatted for ext2. That is why figures like 131MB and 888 cylinders appear. The fdisk tool, which is used, simply reads that from the device.
All operations are done by root. The single partition will be created on /dev/sda (please note: not /dev/sda1 ) The procedure is described in a series of steps with comments. The standard prompt of fdisk is Command (m for help): and you can, at any stage enter m to see the available commands. If you do that the result would be
Your memory stick must be plugged in, but not mounted. Take care that write protect is off.
Now is a good time to verify that all went well by printing out the partition details and table.
Make sure that the partition is not set up as bootable. There should not be an asterisk under the heading "Boot" in the partition table. Furthermore, the headings "Id" and "System" should be as they are in the printout shown above. That indicates that you can format the device in ext2 (next section). These are the defaults. If they are not, they can be changed by
If (or when) the partition table is correct, you can conclude the procedure with
Having completed the partition part, we go straight on to "formatting" the device in ext2. For this we use the tool mke2fs. This is straightforward.
You might like to watch the led on your memory stick flicker while this is happening. When it stops, the job is done.
The second exception is that the permissions on your favourite text file should not change anymore.
The procedures described here will partition the flash memory device in two pieces -- one of the partitions will be formatted in ext2 and the other in vfat. This may be handy when the device is to be used to transport data between two (or more) Linux machines (ext2) and between Linux and Windows machines (vfat)
In this section we describe how to set up a dual partition for the memory stick: One half of the device will be devoted to the vfat filesystem, and the other will be ext2. In this way your device will be more versatile, but also a little shattered. The procedure will be given in somewhat less detailed than in Section 8.1 . Note that the procedure of the previous section is repeated and that the hex code for each file type is explicitly entered (lines following the entry "t": Win95 = b; Linux = 83). When the partition will only be for Linux, this is unnecessary -- Linux is the default. Also note that the vfat partition goes on first -- that is the way Windows would have it.
The choice of splitting the device right in the middle (1--444; 445--888 for a 128 MB memory stick) is arbitrary. Any other well-conceived splitting would be fine.
If you had more than one partition to begin with, the "delete" part of the procedure would have to be repeated.
This step is a repetition of Section 8.2. The first step makes the dos partition, the second makes the Linux ext2 partition.
to mount both partitions on your device.
The Linux machine you did the partitioning on, will be configured for the dual partition. Evidently you need to carry the keychain to another Linux machine. There you will find that it only sees /dev/sda1. This is, perhaps, a shortcoming of the kernel. The workaround is as follows: On the new host machine do (keychain plugged in but not mounted)
Your device with its partitions should be there. Data on your keychain should be untouched. For a single partition this is not necessary.
What we have described so far, are simply the procedures of setting up the system and formatting memory sticks. We have to make the mounting of the USB filesystem more permanent, and we have to make mounting of such a device easier for non-root users. There are only a few things to be done.
The procedure described below is for the more complex situation of a memory stick with a dual partition (Section 9). If the partition is a single ext2, the procedure should be simplified by leaving out the line referring to the vfat part and changing sda2 to sda1.
Suppose that the mount points /mnt/memstick and /mnt/fatstick have been created. You do not have to use those words and they don't even have to be sub-directories of /mnt
To make remembering easier, create the following symbolic links:
For smoother and easier mounting add the following lines to /etc/fstab : ( See NOTE at the end of Section 7.3 )
The middle line enables mounting either in ext2 or vfat. The last two lines make it possible for any user to mount the device by the command
It is possible to mount both partitions simultaneously. See Section 9.
In the case of an ext2-formatted device it does not seem possible for a user (other than root) to mount in read-write mode. The solution is to create a directory on the memory stick owned by the particular user with complete permissions. Root should do that:
If the device is in the vfat format, this step is unnecessary.
In conclusion, here is a small script I use to mount and unmount my (ext2) flash drive:
I call this from an FVWM menu by Exec exec xterm -geometry 43x2+1250+0 -e /home/nikos/bin/flashdrive which places it on a tiny x-term in one corner of the screen. The horizontal displacement of 1250 may have to be modified according to the horizontal resolution of your screen.
It is a virtuous habit to inspect your ext2 filesystem on the flash memory regularly. To do this, the tool dumpe2fs may be used in the following way: # dumpe2fs -h /dev/sda1 (you must be root and the device should not be mounted). The result should be similar to the following:
A few important aspects need to be pointed out.
You might get warnings about these things when you mount the device or when you try the read files from the device.
When the usage allocation has been spent, or there is evidence of file corruption, the thing to do is to run # fsck.ext2 /dev/sda1 with the device unmounted. After that, usage parameters will be freshly allocated and bad blocks will be gone.
Setting up your system as described above for the Linux-2.4.x kernels will be perfectly good, but with the Linux-2.6.x kernels (x = 0, 1, 2, 3 at the time of this writing) much more is possible:
Kernel configuration options are very much the same as described in Section 5.2. The configuration menu (# make menuconfig) has been organized differently and more systematically. Here are the relevant choices:
It may be in your interest to consult Migrating to Linux Kernel 2.6, but it is not necessary to follow steps that cannot be followed.
In Section 6.2 the situation of two USB devices plugged in at the same time was discussed. The difficulty is that it becomes hard for the user to distinguish the two devices. Progress recently made in the Linux Hotplug Project (search for hotplug) and implemented in the 2.6-series of kernels, has made it possible to address this problem efficiently. For implementation some utilities are needed:
The sysfs filesystem.
The sysfs filesystem is intrinsic to the Linux-2.6 kernel series. It is used by udev and sysfstools to obtain information about kernel objects (devices, etc.), their attributes and mutual linkages. Activation is according to the following procedure:
If you do ls -F /sys, a whole directory structure with entries like block/, bus/, ... will be shown. You were successful. The mounting should now be made permanent by adding the following line to /etc/fstab:
This is installed in one of the usual ways without having to make special configuration options. Instructions in the README file are very clear. You should, however, first make sure that the entry /usr/local/lib is entered in /etc/ld.so.conf so that the libsysfs.so libraries will be recognisable by the system. After installation, check that the utilities /usr/local/bin/lsbus and /usr/local/bin/systool have been installed. Also check that the libraries mentioned above have been properly linked in (ldconfig -p | grep libsysfs).
Next install the hotplug scripts. The first thing to do is to check that hotplugging will work on your system. Do: less /proc/sys/kernel/hotplug. The result should be /sbin/hotplug. If it is not so, you should add the line echo "/sbin/hotplug" > /proc/sys/kernel/hotplug to an appropriate initialization script, e.g. /etc/rc.d/rc.local. This command could also be given (as root) before checking that everything works. Installation is extremely simple. Do: # make install. Check afterwards that the script /sbin/hotplug exists.
This is the last installation. Study the README file and follow the instructions necessary to be able to run make USE_KLIBC=true. If you have compiled the Linux-2.6.x kernel, this should not be too difficult. A quick check would be to find out if /sbin/udev and /sbin/udevinfo exist. There are other occurences as well, a very important one being /etc/udev/udev.rules. If you have a working setup for memory sticks (as described above) and a memory stick plugged in you could try ls /udev, you should see the device (sda, sda1, ...). If it is there, unplug the device and list again. It should be gone!
With the udev utility in place, it is possible to identify a memory stick on the basis of some built-in attribute. One that works, is the vendor identity. To find this for an attached device, do systool -vb scsi | grep vendor. This should give a list of attached SCSI devices (remember that your memory stick poses as a SCSI device Section 6.2). Let us suppose that you have two memory sticks plugged in at the same time and found (as I did) two vendor Id's, namely "UFD" and "STORAGE". Then add to the file /etc/udev/udev.rules the following lines (at the top if you like):
The vendor-identities may have trailing blank spaces, and therefore the wildcard asterisks have a purpose. Do not ignore them. Remove and replace the memory sticks, then do ls /udev. Instead of the impersonal sda, sda1, etc. there should now be kalahari, kalahari1, namib and namib1. The devices have been named according to the "vendor" information in them. Thus we can identify the device, no matter where it was mapped. All that remains to be done is to creat appropriate mount points like /mnt/namib and kalahari and make entries like
in /etc/fstab to enable mounting from other places.
This is great fun ... and remember, there are other deserts on the planet.
This document is copyrighted (c) 2003 Niko Sauer and is distributed under the terms of the Linux Documentation Project (LDP) license, stated below.
Unless otherwise stated, Linux HOWTO documents are copyrighted by their respective authors. Linux HOWTO documents may be reproduced and distributed in whole or in part, in any medium physical or electronic, as long as this copyright notice is retained on all copies. Commercial redistribution is allowed and encouraged; however, the author would like to be notified of any such distributions.
All translations, derivative works, or aggregate works incorporating any Linux HOWTO documents must be covered under this copyright notice. That is, you may not produce a derivative work from a HOWTO and impose additional restrictions on its distribution. Exceptions to these rules may be granted under certain conditions; please contact the Linux HOWTO coordinator at the address given below.
In short, we wish to promote dissemination of this information through as many channels as possible. However, we do wish to retain copyright on the HOWTO documents, and would like to be notified of any plans to redistribute the HOWTOs.
If you have any questions, please contact <firstname.lastname@example.org >
No liability for the contents of this documents can be accepted. Use the concepts, examples and other content at your own risk. As this is a new document, there may be errors and inaccuracies, that may of course be damaging to your system. Proceed with caution, and although this is highly unlikely, the author do not take any responsibility for that.
All copyrights are held by their by their respective owners, unless specifically noted otherwise. Use of a term in this document should not be regarded as affecting the validity of any trademark or service mark.
Naming of particular products or brands should not be seen as endorsements.
You are strongly recommended to take a backup of your system before major installation and backups at regular intervals.
I wish to express my deep gratitude towards my sons Paul and Philip who in 1995 persuaded me to move to Linux. Philip gave me profound advice and insights which made this project a pleasant learning experience.
Emma Jane Hogbin, in her review, raised a number of important questions which served to improve this document immensely. My sincerest thanks to her.