Block filesystems

Storage devices are classified in two main types: block devices and flash devices
- They are handled by different subsystems and different filesystems
Block devices can be read and written to on a per-block basis, in random order, without erasing.
- Hard disks, floppy disks, RAM disks
- USB keys, SSD, Compact Flash, SD card, eMMC: these are based on flash storage, but have an integrated controller that emulates a block device, managing the flash in a transparent way.
Raw flash devices are driven by a controller on the SoC. They can be read, but writing requires erasing, and often occurs on a larger size than the “block” size.
- NOR flash, NAND flash

Block device list

The list of all block devices available in the system can be found in /proc/partitions $ cat /proc/partitions major minor #blocks name
And also in /sys/block/

Block devices can be partitioned to store diﬀerent parts of a system
The partition table is stored inside the device itself, and is read and analyzed automatically by the Linux kernel
- mmcblk0 is the entire device
- mmcblk0p2 is the second partition of mmcblk0
Two partition table formats:
- MBR, the legacy format
- GPT, the new format, not yet used everywhere, but becoming more and more common
Numerous tools to create and modify the partitions on a block device: fdisk, gdisk, cfdisk, sfdisk, parted, etc.

It is often necessary to transfer data to or from a block device in a raw way
- Especially to write a ﬁlesystem image to a block device
This directly writes to the block device itself, bypassing any ﬁlesystem layer.
The block devices in /dev/ allow such raw access
dd is the tool of choice for such transfers:
dd if=/dev/mmcblk0p1 of=testfile bs=1M count=16 Transfers 16 blocks of 1 MB from /dev/mmcblk0p1 to testfile
dd if=testfile of=/dev/sda2 bs=1M seek=4 Transfers the complete contents of testfile to /dev/sda2, by blocks of 1 MB, but starting at oﬀset 4 MB in /dev/sda2

The standard ﬁlesystem used on Linux systems is the series of ext{2,3,4} ﬁlesystems
- ext2
- ext3, brought journaling compared to ext2
- ext4, mainly brought performance improvements and support for even larger ﬁlesystems
ext4 is now the default ﬁlesystem used on most Linux distributions
It supports all features Linux needs from a ﬁlesystem: permissions, ownership, device ﬁles, symbolic links, etc.

Designed to stay in a coherent state even after system crashes or a sudden poweroﬀ

Writes are ﬁrst described in the journal before being committed to ﬁles (can be all writes, or only metadata writes depending on the conﬁguration)
Allows to skip a full disk check at boot time after an unclean shutdown

Thanks to the journal, the recovery at boot time is quick, since the operations in progress at the moment of the unclean shutdown are clearly identiﬁed
Does not mean that the latest writes made it to the storage: this depends on syncing the changes to the ﬁlesystem.

btrfs, intended to become the next standard ﬁlesystem for Linux. Integrates numerous features: data checksuming, integrated volume management, snapshots, etc.
XFS, high-performance ﬁlesystem inherited from SGI IRIX, still actively developed.
JFS, inherited from IBM AIX. No longer actively developed, provided mainly for compatibility.
reiserFS, used to be a popular ﬁlesystem, but its latest version Reiser4 was never merged upstream. All those ﬁlesystems provide the necessary functionalities for Linux systems: symbolic links, permissions, ownership, device ﬁles, etc.

Filesystem that takes into account the characteristics of ﬂash-based storage: eMMC, SD cards, SSD, etc.
Developed and contributed by Samsung
Available in the mainline Linux kernel
For optimal results, need a number of details about the storage internal behavior which may not easy to get
Benchmarks: best performer on ﬂash devices most of the time: See http://lwn.net/Articles/520003/
Technical details: http://lwn.net/Articles/518988/
Not as widely used as ext3,4, even on ﬂash-based storage.

Read-only, compressed ﬁlesystem for block devices. Fine for parts of a ﬁlesystem which can be read-only (kernel, binaries...)
Great compression rate, which generally brings improved read performance
Used in most live CDs and live USB distributions
Supports several compression algorithm (LZO, XZ, etc.)
Benchmarks: roughly 3 times smaller than ext3, and 2-4 times faster (http://elinux.org/Squash_Fs_Comparisons)
Details: http://squashfs.sourceforge.net/

Linux also supports several other ﬁlesystem formats, mainly to be interopable with other operating systems:

vfat for compatibility with the FAT ﬁlesystem used in the Windows world and on numerous removable devices
- This ﬁlesystem does not support features like permissions, ownership, symbolic links, etc. Cannot be used for a Linux root ﬁlesystem.
ntfs for compatibility with the NTFS ﬁlesystem used on Windows
hfs for compatibility with the HFS ﬁlesystem used on Mac OS
iso9660, the ﬁlesystem format used on CD-ROMs, obviously a read-only ﬁlesystem

Not a block ﬁlesystem of course!
Perfect to store temporary data in RAM: system log ﬁles, connection data, temporary ﬁles...
More space-eﬃcient than ramdisks: ﬁles are directly in the ﬁle cache, grows and shrinks to accommodate stored ﬁles
How to use: choose a name to distinguish the various tmpfs instances you could have. Examples: mount -t tmpfs varrun /var/run mount -t tmpfs udev /dev
See Documentation/filesystems/tmpfs.txt in kernel sources.