240 likes | 552 Views
How to make a pseudo-file. As a follow-up to our first lab, we examine the steps needed to create our own ‘/proc’ file . The ‘extensibility’ imperative. A modern OS needs the ability to evolve It will need to support new devices It will need to allow ‘bugs’ to be repaired
E N D
How to make a pseudo-file As a follow-up to our first lab, we examine the steps needed to create our own ‘/proc’ file
The ‘extensibility’ imperative • A modern OS needs the ability to evolve • It will need to support new devices • It will need to allow ‘bugs’ to be repaired • It will need to permit performance gains • Otherwise: suffer early obsolescence!
Two Extensibility Mechanisms • ‘Open Source’ programming • ‘Installable’ kernel modules
‘Open Source’ • Source text for the Linux kernel is on disk (usual directory-location is ‘/usr/src/linux’) • Majority of the sources are written in ‘C’ (so ‘portable’ across CPU architectures) • Some are written in assembly languages (for nearly twenty different architectures) • Files are grouped into major categories (kernel, mm, fs, net, ipc, lib, drivers, init) • You could edit and recompile your kernel
Installable kernel modules • Great mechanism for kernel ‘extensibility’ • Neat tool for studying how kernel works • Kernel can be modified while it’s running • Unnecessary to recompile and then reboot • But inherently unsafe: programming bugs in the kernel can cause system crashes!
Some ‘superuser’ privileges • Since modifying a running kernel is ‘risky’, only authorized ‘system administrators’ are normally allowed to install kernel modules • But our systems are specially configured to permit you to install or remove modules • This is purely for ‘educational’ purposes (so you should use this privilege wisely)
Linux module structure • Two ‘module administration’ functions are mandatory components in every module plus • Appropriate ‘module service’ functions and their supporting kernel data-structures are optional components in particular modules also • Recent kernels require a Module License!
A minimal module-template #include <linux/module.h> int init_module( void ) { // code here gets called during module installation } void cleanup_module( void ) { // code here gets called during module removal } MODULE_LICENSE(“GPL”);
How to compile a module • You could directly invoke the C-compiler: $ gcc –c –O –D__KERNEL__ -DMODULE –I/lib/modules/2.4.26/build/include mod.c • OR: you can use the ‘make’ utility: $ make mod.o
The ‘printk()’ function • Kernel modules cannot call any functions in the C runtime library (e.g., ‘printf()’) • But similar kernel versions of important functions are provided (e.g., ‘printk()’) • Syntax and semantics are slightly different (e.g., priority and message-destination) • Capabilities may be somewhat restricted (e.g., printk() can’t show floating-point)
Simple module example #include <linux/module.h> int init_module( void ) { printk( “<1>Hello, world!\n” ); return 0; // SUCCESS } void cleanup_module( void ) { printk( “<1>Goodbye everyone\n” ); } MODULE_LICENSE(“GPL”);
How to install and remove root# /sbin/insmod hello.o root# /sbin/rmmod hello
A non-trivial module-example • Let’s see how we can use kernel functions to create our own ‘/proc’ file • Easy if we use ‘create_proc_read_entry()’ during module-initialization (and then use ‘remove_proc_entry()’ during cleanup) • Prototypes in the <linux/proc_fs.h> header • We’ll show the current value of a volatile kernel variable, named ‘jiffies’
jiffies • unsigned long volatile jiffies; • global kernel variable (used by scheduler) • Initialized to zero when system reboots • Gets incremented when timer interrupts • So it counts ‘clock-ticks’ since cpu restart • ‘tick-frequency’ is architecture dependent
Writing the ‘jiffies.c’ module • We need to declare a name for pseudo-file static char modname[ ] = “jiffies”; • We need to define a ‘proc_read’ function (see code on the following slide) • We need to ‘register’ our filename, along with its ‘read’ method, in ‘init_module()’ • We need to ‘unregister’ our pseudo-file in ‘cleanup_module()’
Our module’s ‘read’ method static int my_proc_read( char *buf, char **start, off_t off, int count, int *eof, void *data ) { return sprintf( buf, “jiffies=%lu\n”, jiffies ); }
Our ‘initialization’ function int init_module( void ) { create_proc_read_entry( modname, 0, NULL, my_proc_read, NULL ); return 0; }
Our ‘cleanup’ function void cleanup_module( void ) { remove_proc_entry( modname, NULL ); }
In-class exercise #1 • Use an editor (e.g., ‘vi’) to create a source file in C (named ‘jiffies.c’) for your module • Use the ‘make’ command to compile your module’s source-file into an object-file • Use the ‘insmod’ command to install your module object-file into the running kernel • Use the ‘cat’ command to display ‘jiffies’ • Then use ‘rmmod’ to remove your module
Makefile • We need it to automate module compiles • Otherwise we type a VERY long command • To compile ‘jiffies.c’ type: $ make jiffies.o • Our Makefile defines some ‘implicit rules’ • ‘make’ works by doing pattern-matching
Rule syntax targets … : dependencies … commands ...
Pattern Rule example CC = gcc INCLUDE = /lib/modules/2.4.26/build/include CFLAGS = -c –O CFLAGS += -D__KERNEL__ -DMODULE # primary pattern rule %.o : %.c %.h $(CC) -I$(INCLUDE) $CFLAGS $< # fallback pattern rule %.o : %.c $(CC) -I$(INCLUDE) $CFLAGS $<
‘Makefile’ is on class website • You can download the ‘Makefile’ from our website: http://cs.usfca.edu/~cruse/cs326/ • Put the ‘Makefile’ in your current working directory (along with your module source) • Then you can compile by typing this short command: $ make jiffies.o
In-class exercise #2 • Use your knowledge of standard C library functions (e.g., open(), read(), close() ) to write an application-program (name it ‘showjifs.cpp’) which reads the information from your ‘proc/jiffies’ pseudo-file and then prints it on the screen • Use a program-loop to re-read the pseudo file multiple times • How rapidly does ‘jiffies’ value increase?