In Installing Basic System Software, we installed the Udev package. Before we go into the details regarding how this works, a brief history of previous methods of handling devices is in order.
Linux systems in general traditionally use a static device creation method, whereby a great many device nodes are created under /dev (sometimes literally thousands of nodes), regardless of whether the corresponding hardware devices actually exist. This is typically done via a MAKEDEV script, which contains a number of calls to the mknod program with the relevant major and minor device numbers for every possible device that might exist in the world.
Using the Udev method, only those devices which are detected by the kernel get device nodes created for them. Because these device nodes will be created each time the system boots, they will be stored on a tmpfs file system (a virtual file system that resides entirely in system memory). Device nodes do not require much space, so the memory that is used is negligible.
In February 2000, a new filesystem called devfs was merged into the 2.3.46 kernel and was made available during the 2.4 series of stable kernels. Although it was present in the kernel source itself, this method of creating devices dynamically never received overwhelming support from the core kernel developers.
The main problem with the approach adopted by devfs was the way it handled device detection, creation, and naming. The latter issue, that of device node naming, was perhaps the most critical. It is generally accepted that if device names are allowed to be configurable, then the device naming policy should be up to a system administrator, not imposed on them by any particular developer(s). The devfs file system also suffers from race conditions that are inherent in its design and cannot be fixed without a substantial revision to the kernel. It has also been marked as deprecated due to a lack of recent maintenance.
With the development of the unstable 2.5 kernel tree, later released as the 2.6 series of stable kernels, a new virtual filesystem called sysfs came to be. The job of sysfs is to export a view of the system's hardware configuration to userspace processes. With this userspace-visible representation, the possibility of seeing a userspace replacement for devfs became much more realistic.
The sysfs filesystem was mentioned briefly above. One may wonder how sysfs knows about the devices present on a system and what device numbers should be used for them. Drivers that have been compiled into the kernel directly register their objects with sysfs as they are detected by the kernel. For drivers compiled as modules, this registration will happen when the module is loaded. Once the sysfs filesystem is mounted (on /sys), data which the built-in drivers registered with sysfs are available to userspace processes and to udevd for device node creation.
The S10udev initscript takes care of creating device nodes when Linux is booted. The script unsets the uevent handler from the default of /sbin/hotplug. This is done because the kernel no longer needs to call out to an external binary. Instead udevd will listen on a netlink socket for uevents that the kernel raises. Next, the bootscript copies any static device nodes that exist in /lib/udev/devices to /dev. This is necessary because some devices, directories, and symlinks are needed before the dynamic device handling processes are available during the early stages of booting a system. Creating static device nodes in /lib/udev/devices also provides an easy workaround for devices that are not supported by the dynamic device handling infrastructure. The bootscript then starts the Udev daemon, udevd, which will act on any uevents it receives. Finally, the bootscript forces the kernel to replay uevents for any devices that have already been registered and then waits for udevd to handle them.
To obtain the right major and minor number for a device, Udev relies on the information provided by sysfs in /sys. For example, /sys/class/tty/vcs/dev contains the string “7:0”. This string is used by udevd to create a device node with major number 7 and minor 0. The names and permissions of the nodes created under the /dev directory are determined by rules specified in the files within the /etc/udev/rules.d/ directory. These are numbered in a similar fashion to the CLFS-Bootscripts package. If udevd can't find a rule for the device it is creating, it will default permissions to 660 and ownership to root:root. Documentation on the syntax of the Udev rules configuration files are available in /usr/share/doc/udev-103/index.html
Device drivers compiled as modules may have aliases built into them. Aliases are visible in the output of the modinfo program and are usually related to the bus-specific identifiers of devices supported by a module. For example, the snd-fm801 driver supports PCI devices with vendor ID 0x1319 and device ID 0x0801, and has an alias of “pci:v00001319d00000801sv*sd*bc04sc01i*”. For most devices, the bus driver exports the alias of the driver that would handle the device via sysfs. E.g., the /sys/bus/pci/devices/0000:00:0d.0/modalias file might contain the string “pci:v00001319d00000801sv00001319sd00001319bc04sc01i00”. The rules that CLFS installs will cause udevd to call out to /sbin/modprobe with the contents of the MODALIAS uevent environment variable (that should be the same as the contents of the modalias file in sysfs), thus loading all modules whose aliases match this string after wildcard expansion.
In this example, this means that, in addition to snd-fm801, the obsolete (and unwanted) forte driver will be loaded if it is available. See below for ways in which the loading of unwanted drivers can be prevented.
The kernel itself is also able to load modules for network protocols, filesystems and NLS support on demand.
There are a few possible problems when it comes to automatically creating device nodes.
Udev will only load a module if it has a bus-specific alias and the bus driver properly exports the necessary aliases to sysfs. In other cases, one should arrange module loading by other means. With Linux-2.6.19, Udev is known to load properly-written drivers for INPUT, IDE, PCI, USB, SCSI, SERIO and FireWire devices.
To determine if the device driver you require has the necessary support for Udev, run modinfo with the module name as the argument. Now try locating the device directory under /sys/bus and check whether there is a modalias file there.
If the modalias file exists in sysfs, the driver supports the device and can talk to it directly, but doesn't have the alias, it is a bug in the driver. Load the driver without the help from Udev and expect the issue to be fixed later.
If there is no modalias file in the relevant directory under /sys/bus, this means that the kernel developers have not yet added modalias support to this bus type. With Linux-2.6.19, this is the case with ISA busses. Expect this issue to be fixed in later kernel versions.
Udev is not intended to load “wrapper” drivers such as snd-pcm-oss and non-hardware drivers such as loop at all.
If the “wrapper” module only enhances the functionality provided by some other module (e.g., snd-pcm-oss enhances the functionality of snd-pcm by making the sound cards available to OSS applications), configure modprobe to load the wrapper after Udev loads the wrapped module. To do this, add an “install” line in /etc/modprobe.conf. For example:
install snd-pcm /sbin/modprobe -i snd-pcm ; \ /sbin/modprobe snd-pcm-oss ; true
If the module in question is not a wrapper and is useful by itself, configure the S05modules bootscript to load this module on system boot. To do this, add the module name to the /etc/sysconfig/modules file on a separate line. This works for wrapper modules too, but is suboptimal in that case.
Either don't build the module, or blacklist it in /etc/modprobe.conf file as done with the forte module in the example below:
Blacklisted modules can still be loaded manually with the explicit modprobe command.
This usually happens if a rule unexpectedly matches a device. For example, a poorly-writen rule can match both a SCSI disk (as desired) and the corresponding SCSI generic device (incorrectly) by vendor. Find the offending rule and make it more specific.
This may be another manifestation of the previous problem. If not, and your rule uses sysfs attributes, it may be a kernel timing issue, to be fixed in later kernels. For now, you can work around it by creating a rule that waits for the used sysfs attribute and appending it to the /etc/udev/rules.d/10-wait_for_sysfs.rules file. Please notify the CLFS Development list if you do so and it helps.
Further text assumes that the driver is built statically into the kernel or already loaded as a module, and that you have already checked that Udev doesn't create a misnamed device.
Udev has no information needed to create a device node if a kernel driver does not export its data to sysfs. This is most common with third party drivers from outside the kernel tree. Create a static device node in /lib/udev/devices with the appropriate major/minor numbers (see the file devices.txt inside the kernel documentation or the documentation provided by the third party driver vendor). The static device node will be copied to /dev by the S10udev bootscript.
This is due to the fact that Udev, by design, handles uevents and loads modules in parallel, and thus in an unpredictable order. This will never be “fixed”. You should not rely upon the kernel device names being stable. Instead, create your own rules that make symlinks with stable names based on some stable attributes of the device, such as a serial number or the output of various *_id utilities installed by Udev. See Section 7.13, “Creating custom symlinks to devices” and Section 7.14, “Configuring the network Script” for examples.
Additional helpful documentation is available at the following sites:
A Userspace Implementation of devfs http://www.kroah.com/linux/talks/ols_2003_udev_paper/Reprint-Kroah-Hartman-OLS2003.pdf
The sysfs Filesystem http://www.kernel.org/pub/linux/kernel/people/mochel/doc/papers/ols-2005/mochel.pdf