me and my friend are trying to develop custom memory allocator in linux ubuntu 16.04.
We got stuck because of an error, btw its our first time
that we are trying to code something like that so we are not the best debuggers the error is : Segmentation fault (core dumped)
and here is the code.
can anybody help us understand whats wrong ?
Thank you!
#include <unistd.h>
#include <string.h>
#include <pthread.h>
#include <stdio.h>
struct header_t {
size_t size;
unsigned is_free;
struct header_t *next; };
struct header_t *head = NULL, *tail = NULL;
pthread_mutex_t global_malloc_lock;
struct header_t *get_free_block(size_t size)
{
struct header_t *curr = head;
while(curr) {
/* see if there's a free block that can accomodate requested size */
if (curr->is_free && curr->size >= size)
return curr;
curr = curr->next;
}
return NULL;
}
void free(void *block)
{
struct header_t *header, *tmp;
/* program break is the end of the
process's data segment */
void *programbreak;
if (!block)
return;
pthread_mutex_lock(&global_malloc_lock);
header = (struct header_t*)block - 1;
/* sbrk(0) gives the current program break address */
programbreak = sbrk(0);
/*
Check if the block to be freed is the last one in the
linked list. If it is, then we could shrink the size of the
heap and release memory to OS. Else, we will keep the block
but mark it as free. */
if ((char*)block + header->size == programbreak) {
if (head == tail) {
head = tail = NULL;
} else {
tmp = head;
while (tmp) {
if(tmp->next == tail) {
tmp->next = NULL;
tail = tmp;
}
tmp = tmp->next;
}
}
/* sbrk() with a negative argument decrements the program break.
So memory is released by the program to OS. */
sbrk(0 - header->size - sizeof(struct header_t));
/* Note: This lock does not really assure thread
safety, because sbrk() itself is not really
thread safe. Suppose there occurs a foregin sbrk(N)
after we find the program break and before we decrement
it, then we end up realeasing the memory obtained by
the foreign sbrk(). */
pthread_mutex_unlock(&global_malloc_lock);
return;
}
header->is_free = 1;
pthread_mutex_unlock(&global_malloc_lock);
}
void *malloc(size_t size)
{
size_t total_size;
void *block;
struct header_t *header;
if (!size)
return NULL;
pthread_mutex_lock(&global_malloc_lock);
header = get_free_block(size);
if (header) {
/* Woah, found a free block to accomodate requested memory. */
header->is_free = 0;
pthread_mutex_unlock(&global_malloc_lock);
return (void*)(header + 1);
}
/* We need to get memory to fit in the requested block and header
from OS. */
total_size = sizeof(struct header_t) + size;
block = sbrk(total_size);
if (block == (void*) -1) {
pthread_mutex_unlock(&global_malloc_lock);
return NULL;
}
header = block;
header->size = size;
header->is_free = 0;
header->next = NULL;
if (!head)
head = header;
if (tail)
tail->next = header;
tail = header;
pthread_mutex_unlock(&global_malloc_lock);
return (void*)(header + 1);
}
void *calloc(size_t num, size_t nsize)
{
size_t size;
void *block;
if (!num || !nsize)
return NULL;
size = num * nsize;
/* check mul overflow */
if (nsize != size / num)
return NULL;
block = malloc(size);
if (!block)
return NULL;
memset(block, 0, size);
return block;
}
void *realloc(void *block, size_t size)
{
struct header_t *header;
void *ret;
if (!block || !size)
return malloc(size);
header = (struct header_t*)block - 1;
if (header->size >= size)
return block;
ret = malloc(size);
if (ret) {
/* Relocate contents to the new bigger block */
memcpy(ret, block, header->size);
/* Free the old memory block */
free(block);
}
return ret;
}
The problem occurred because the functions were not prototyped [decalred].
Once I added functions prototype. The code worked.
For more information about prototyping: http://www.trytoprogram.com/c-programming/function-prototype-in-c/
mutex variable should be initialized before using it for applying lock. your global_malloc_lock is not initialized.
you can't initialize mutex variable as of normal variable.
pthread_mutex_t global_malloc_lock = 0 ;// invalid .. you may thinking since it's it declared as global it's initialized with 0 which is wrong
Initialize the mutex variable by calling pthread_mutex_init() or using PTHREAD_MUTEX_INITIALIZER ;
for your code add this
pthread_mutex_t global_malloc_lock = pthread_mutex_t global_malloc_lock;
Related
I am learning how to write Linux Device Driver.
I wrote a dummy character device driver, implemented open, release, write, read in fops;
When I read from device , everything was ok;
When I wrote to device by "echo xx > ", the OS was hang.
Even I comment out all codes in write function except pr_alert and return statements, It still hangs;
Could anybody help me figure it out?
#include <linux/init.h>
#include <linux/types.h>
#include <linux/module.h>
#include <linux/mutex.h>
#include <linux/cdev.h>
#include <linux/fs.h>
#include <linux/slab.h>
#include <linux/string.h>
struct hello_dev
{
char *buffer;
int length;
dev_t dev;
struct mutex lock;
struct cdev *pcdev;
};
struct hello_dev *pHelloDev;
int open_device(struct inode *pinode, struct file *filp)
{
filp->private_data = pHelloDev;
return 0;
}
int close_device(struct inode *pinode, struct file *filp)
{
struct hello_dev *pDev = filp->private_data;
if (pDev->buffer != NULL)
kfree(pDev->buffer);
pDev->buffer = NULL;
return 0;
}
ssize_t read_device(struct file *filp, char __user *buffer, size_t len, loff_t *loff)
{
pr_alert("read\n");
struct hello_dev *pDev = filp->private_data;
mutex_lock(&pDev->lock);
if (pDev->buffer == NULL)
{
mutex_unlock(&pDev->lock);
return 0;
}
int length = strlen(pDev->buffer);
// offset max than strlen in buffer, return
if (*loff > (length - 1))
{
mutex_unlock(&pDev->lock);
return 0;
} else {
// available to read
int len2read = length - *loff;
if (len < len2read)
{// buffer length less than available data
len2read = len;
}
int read = copy_to_user(buffer, pDev->buffer + *loff, len2read);
if (read)
{
*loff = *loff + read;
mutex_unlock(&pDev->lock);
return read;
} else {
*loff = *loff + len2read;
mutex_unlock(&pDev->lock);
return len2read;
}
}
}
ssize_t write_device(struct file *filp , const char __user *buffer, size_t len, loff_t* loff) {
pr_alert("write %s\n", buffer);
// struct hello_dev *pDev = filp->private_data;
// mutex_lock(&pDev->lock);
// if(pDev->buffer == NULL) {
// pDev->buffer = kmalloc(100, GFP_KERNEL);
// pDev->length = 100;
// }
// copy_from_user(pDev->buffer, buffer, len);
// *loff = *loff + len;
// mutex_unlock(&pDev->lock);
return len;
}
struct file_operations fops = {
.open = open_device,
.release = close_device,
.read = read_device,
.write = write_device
};
int init_device(void)
{
pr_alert("init device\n");
pHelloDev = kmalloc(sizeof(struct hello_dev), GFP_KERNEL);
pHelloDev->buffer = NULL;
pHelloDev->length = 0;
int ret = alloc_chrdev_region(&pHelloDev->dev, 0, 1, "hello");
if (ret)
goto alloc_error;
if (pHelloDev == NULL)
goto kmalloc_error;
pHelloDev->pcdev = cdev_alloc();
pHelloDev->pcdev->ops = &fops;
mutex_init(&pHelloDev->lock);
ret = cdev_add(pHelloDev->pcdev, pHelloDev->dev, 1);
if (ret)
goto cdev_add_error;
return 0;
alloc_error:
pr_alert("alloc_chrdev_region error, %d\n", ret);
return ret;
kmalloc_error:
pr_alert("alloc struct hello_dev error");
return -ENOMEM;
cdev_add_error:
pr_alert("cdev_add error, %d\n", ret);
return ret;
}
void cleanup_device(void)
{
pr_alert("unload ko\n");
cdev_del(pHelloDev->pcdev);
unregister_chrdev_region(pHelloDev->dev, 1);
}
MODULE_LICENSE("GPL");
module_init(init_device);
module_exit(cleanup_device);
I found why write to device hangs.
//this statements has problem
//maybe there is no \0 in buffer
//so I print it out, it will hang
//I wrote a program to write something to device
//and used strace to trace system call made by this program
//and found it hangs at write(...) system call
//and there was nothing printed out
//so, it must be this statement causing the problem
//when I removed this statement, everything was ok
pr_alert("write %s\n", buffer);
I'm trying to implement the malloc function and it looks like that gdb is giving me some weird values from this struct:
struct MemoryBlock {
struct MemoryBlock * next;
size_t size;
signed char is_free;
} startBlock;
And that's the function where I'm debugging it with gdb:
struct MemoryBlock * create_new_block(size_t size)
{
struct MemoryBlock * ret_block;
// add some space for the struct block
size += sizeof(struct MemoryBlock);
ret_block = (void *) sbrk(size);
// test first, if we can allocate that much of ram
if (ret_block == (void *) -1)
return NULL;
ret_block->size = size - sizeof(struct MemoryBlock);
ret_block->is_free = 0;
ret_block->next = NULL;
return ret_block; // HERE'S the breakpoint
}
So here's the issue (I'm at the breakpoint return ret_block):
If I want to see what kind of values are inside of the ret_block pointer, than I'm getting this:
(gdb) p (struct MemoryBlock) ret_block
$26 = {next = 0x555555559000, size = 140737488347680, is_free = -53 '\313'}
size is fine, because if I convert it into the decimal system than I'm getting 3 as expected. (the argument size from the function is currently 3)
But I'm surprised that next and is_free aren't 0 since the last three lines should set both to 0.
So I looked up what is in the memory:
As you can see each value is correctly stored in my heap. But why am I getting these values if I do p (struct MemoryBlock) ret_block?
If you need the whole code
#include <unistd.h>
#include <stdio.h>
/* ============
* Structs
* ============ */
struct MemoryBlock {
struct MemoryBlock * next;
size_t size;
signed char is_free;
} startBlock;
/* ==============
* Functions
* ============== */
struct MemoryBlock * create_new_block(size_t size);
void * malloc(size_t size);
/* ==================
* Main Programm
* ================== */
int main()
{
char * buffer;
char * b2;
unsigned short index;
// The start of my heap :D
startBlock.is_free = 0;
startBlock.size = 0;
buffer = malloc(3);
b2 = malloc(3);
// ----- ERROR -----
if (buffer == NULL || b2 == NULL)
return 1;
// ----- ERROR -----
// fill the buffers with random stuff
for (index=0; index<2; index++) {
buffer[index] = 'a';
b2[index] = 'b';
}
buffer[index] = '\0';
b2[index] = '\0';
puts(buffer);
puts(b2);
return 0;
}
struct MemoryBlock * create_new_block(size_t size)
{
struct MemoryBlock * ret_block;
// add some space for the struct block
size += sizeof(struct MemoryBlock);
ret_block = (void *) sbrk(size);
// test first, if we can allocate that much of ram
if (ret_block == (void *) -1)
return NULL;
ret_block->size = size - sizeof(struct MemoryBlock);
ret_block->is_free = 0;
ret_block->next = NULL;
return ret_block;
}
void * malloc (size_t size)
{
struct MemoryBlock * ret_block;
struct MemoryBlock * prev_block;
prev_block = &startBlock;
ret_block = startBlock.next;
// go through the linked lists and look if you can find a suitable block
while (ret_block != NULL && (ret_block->size < size || !ret_block->is_free))
{
prev_block = ret_block;
ret_block = ret_block->next;
}
// couldn't find a suitable block => create a new one
if (ret_block == NULL) {
ret_block = create_new_block(size);
if (ret_block == NULL)
return NULL;
}
prev_block->next = ret_block;
ret_block->is_free = 0;
return ret_block;
}
Ok, one of my friends told me my issue... The casting was wrong! Here's the solution:
(gdb) p * ret_block
$57 = {next = 0x0, size = 3, is_free = 0 '\000'}
A star was enough to get the desired output...
How do I create my own bio request to read a sector from the disk drive ?
I am trying the following but it freezes the system.
static void read_bio()
{
struct bio *b;
struct page *p;
b = bio_alloc(GFP_KERNEL, 1);
if (!b) {
printk(KERN_INFO "bio allocation failed\n");
}
bio_init(b);
b->bi_sector = 10000;
b->bi_bdev = bd; /* "/dev/sda1" */
b->bi_end_io = bio_end_clone;
p = alloc_page(GFP_KERNEL);
if (!p) {
printk(KERN_INFO "page allocation failed\n");
}
bio_add_page(b, p, PAGE_SIZE, 0);
b->bi_private = p;
bio_get(b);
submit_bio(READ, b);
printk(KERN_DEBUG "submit read request\n");
}
It is an old question, but anyway here is the code for reading, I hope it will help someone:
int readPage(struct block_device *device, sector_t sector, int size,
struct page *page)
{
int ret;
struct completion event;
struct bio *bio = bio_alloc(device, 1, REQ_OP_READ, GFP_NOIO);
bio->bi_iter.bi_sector = sector;
bio_add_page(bio, page, size, 0);
init_completion(&event);
bio->bi_private = &event;
bio->bi_end_io = readComplete;
submit_bio(bio);
wait_for_completion(&event);
ret = test_bit(BIO_UPTODATE, &bio->bi_flags);
bio_put(bio);
return ret;
}
And for writing:
void writePage(struct block_device *device,
sector_t sector, int size, struct page *page)
{
struct bio *bio = bio_alloc(device, 1, REQ_OP_WRITE, GFP_NOIO);
bio->bi_iter.bi_sector = sector;
bio_add_page(bio, page, size, 0);
bio->bi_end_io = writeComplete;
submit_bio(bio);
}
page can be allocated with alloc_page(GFP_KERNEL). Also for changing data in page use page_address(page). It returns void* so you can interpret that pointer as whatever you want.
Let's say the buffer is allocated using a page based scheme. One way to implement mmap would be to use remap_pfn_range but LDD3 says this does not work for conventional memory. It appears we can work around this by marking the page(s) reserved using SetPageReserved so that it gets locked in memory. But isn't all kernel memory already non-swappable i.e. already reserved? Why the need to set the reserved bit explicitly?
Does this have something to do with pages allocated from HIGH_MEM?
The simplest way to map a set of pages from the kernel in your mmap method is to use the fault handler to map the pages. Basically you end up with something like:
static int my_mmap(struct file *filp, struct vm_area_struct *vma)
{
vma->vm_ops = &my_vm_ops;
return 0;
}
static const struct file_operations my_fops = {
.owner = THIS_MODULE,
.open = nonseekable_open,
.mmap = my_mmap,
.llseek = no_llseek,
};
(where the other file operations are whatever your module needs). Also in my_mmap you do whatever range checking etc. is needed to validate the mmap parameters.
Then the vm_ops look like:
static int my_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
{
vmf->page = my_page_at_index(vmf->pgoff);
get_page(vmf->page);
return 0;
}
static const struct vm_operations_struct my_vm_ops = {
.fault = my_fault
}
where you just need to figure out for a given vma / vmf passed to your fault function which page to map into userspace. This depends on exactly how your module works. For example, if you did
my_buf = vmalloc_user(MY_BUF_SIZE);
then the page you use would be something like
vmalloc_to_page(my_buf + (vmf->pgoff << PAGE_SHIFT));
But you could easily create an array and allocate a page for each entry, use kmalloc, whatever.
[just noticed that my_fault is a slightly amusing name for a function]
Minimal runnable example and userland test
Kernel module:
#include <linux/fs.h>
#include <linux/init.h>
#include <linux/kernel.h> /* min */
#include <linux/mm.h>
#include <linux/module.h>
#include <linux/proc_fs.h>
#include <linux/uaccess.h> /* copy_from_user, copy_to_user */
#include <linux/slab.h>
static const char *filename = "lkmc_mmap";
enum { BUFFER_SIZE = 4 };
struct mmap_info {
char *data;
};
/* After unmap. */
static void vm_close(struct vm_area_struct *vma)
{
pr_info("vm_close\n");
}
/* First page access. */
static vm_fault_t vm_fault(struct vm_fault *vmf)
{
struct page *page;
struct mmap_info *info;
pr_info("vm_fault\n");
info = (struct mmap_info *)vmf->vma->vm_private_data;
if (info->data) {
page = virt_to_page(info->data);
get_page(page);
vmf->page = page;
}
return 0;
}
/* After mmap. TODO vs mmap, when can this happen at a different time than mmap? */
static void vm_open(struct vm_area_struct *vma)
{
pr_info("vm_open\n");
}
static struct vm_operations_struct vm_ops =
{
.close = vm_close,
.fault = vm_fault,
.open = vm_open,
};
static int mmap(struct file *filp, struct vm_area_struct *vma)
{
pr_info("mmap\n");
vma->vm_ops = &vm_ops;
vma->vm_flags |= VM_DONTEXPAND | VM_DONTDUMP;
vma->vm_private_data = filp->private_data;
vm_open(vma);
return 0;
}
static int open(struct inode *inode, struct file *filp)
{
struct mmap_info *info;
pr_info("open\n");
info = kmalloc(sizeof(struct mmap_info), GFP_KERNEL);
pr_info("virt_to_phys = 0x%llx\n", (unsigned long long)virt_to_phys((void *)info));
info->data = (char *)get_zeroed_page(GFP_KERNEL);
memcpy(info->data, "asdf", BUFFER_SIZE);
filp->private_data = info;
return 0;
}
static ssize_t read(struct file *filp, char __user *buf, size_t len, loff_t *off)
{
struct mmap_info *info;
ssize_t ret;
pr_info("read\n");
if ((size_t)BUFFER_SIZE <= *off) {
ret = 0;
} else {
info = filp->private_data;
ret = min(len, (size_t)BUFFER_SIZE - (size_t)*off);
if (copy_to_user(buf, info->data + *off, ret)) {
ret = -EFAULT;
} else {
*off += ret;
}
}
return ret;
}
static ssize_t write(struct file *filp, const char __user *buf, size_t len, loff_t *off)
{
struct mmap_info *info;
pr_info("write\n");
info = filp->private_data;
if (copy_from_user(info->data, buf, min(len, (size_t)BUFFER_SIZE))) {
return -EFAULT;
} else {
return len;
}
}
static int release(struct inode *inode, struct file *filp)
{
struct mmap_info *info;
pr_info("release\n");
info = filp->private_data;
free_page((unsigned long)info->data);
kfree(info);
filp->private_data = NULL;
return 0;
}
static const struct file_operations fops = {
.mmap = mmap,
.open = open,
.release = release,
.read = read,
.write = write,
};
static int myinit(void)
{
proc_create(filename, 0, NULL, &fops);
return 0;
}
static void myexit(void)
{
remove_proc_entry(filename, NULL);
}
module_init(myinit)
module_exit(myexit)
MODULE_LICENSE("GPL");
GitHub upstream.
Userland test:
#define _XOPEN_SOURCE 700
#include <assert.h>
#include <fcntl.h>
#include <stdio.h>
#include <stdlib.h>
#include <stdint.h> /* uintmax_t */
#include <string.h>
#include <sys/mman.h>
#include <unistd.h> /* sysconf */
/* Format documented at:
* https://github.com/torvalds/linux/blob/v4.9/Documentation/vm/pagemap.txt
*/
typedef struct {
uint64_t pfn : 54;
unsigned int soft_dirty : 1;
unsigned int file_page : 1;
unsigned int swapped : 1;
unsigned int present : 1;
} PagemapEntry;
/* Parse the pagemap entry for the given virtual address.
*
* #param[out] entry the parsed entry
* #param[in] pagemap_fd file descriptor to an open /proc/pid/pagemap file
* #param[in] vaddr virtual address to get entry for
* #return 0 for success, 1 for failure
*/
int pagemap_get_entry(PagemapEntry *entry, int pagemap_fd, uintptr_t vaddr)
{
size_t nread;
ssize_t ret;
uint64_t data;
nread = 0;
while (nread < sizeof(data)) {
ret = pread(pagemap_fd, ((uint8_t*)&data) + nread, sizeof(data),
(vaddr / sysconf(_SC_PAGE_SIZE)) * sizeof(data) + nread);
nread += ret;
if (ret <= 0) {
return 1;
}
}
entry->pfn = data & (((uint64_t)1 << 54) - 1);
entry->soft_dirty = (data >> 54) & 1;
entry->file_page = (data >> 61) & 1;
entry->swapped = (data >> 62) & 1;
entry->present = (data >> 63) & 1;
return 0;
}
/* Convert the given virtual address to physical using /proc/PID/pagemap.
*
* #param[out] paddr physical address
* #param[in] pid process to convert for
* #param[in] vaddr virtual address to get entry for
* #return 0 for success, 1 for failure
*/
int virt_to_phys_user(uintptr_t *paddr, pid_t pid, uintptr_t vaddr)
{
char pagemap_file[BUFSIZ];
int pagemap_fd;
snprintf(pagemap_file, sizeof(pagemap_file), "/proc/%ju/pagemap", (uintmax_t)pid);
pagemap_fd = open(pagemap_file, O_RDONLY);
if (pagemap_fd < 0) {
return 1;
}
PagemapEntry entry;
if (pagemap_get_entry(&entry, pagemap_fd, vaddr)) {
return 1;
}
close(pagemap_fd);
*paddr = (entry.pfn * sysconf(_SC_PAGE_SIZE)) + (vaddr % sysconf(_SC_PAGE_SIZE));
return 0;
}
enum { BUFFER_SIZE = 4 };
int main(int argc, char **argv)
{
int fd;
long page_size;
char *address1, *address2;
char buf[BUFFER_SIZE];
uintptr_t paddr;
if (argc < 2) {
printf("Usage: %s <mmap_file>\n", argv[0]);
return EXIT_FAILURE;
}
page_size = sysconf(_SC_PAGE_SIZE);
printf("open pathname = %s\n", argv[1]);
fd = open(argv[1], O_RDWR | O_SYNC);
if (fd < 0) {
perror("open");
assert(0);
}
printf("fd = %d\n", fd);
/* mmap twice for double fun. */
puts("mmap 1");
address1 = mmap(NULL, page_size, PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0);
if (address1 == MAP_FAILED) {
perror("mmap");
assert(0);
}
puts("mmap 2");
address2 = mmap(NULL, page_size, PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0);
if (address2 == MAP_FAILED) {
perror("mmap");
return EXIT_FAILURE;
}
assert(address1 != address2);
/* Read and modify memory. */
puts("access 1");
assert(!strcmp(address1, "asdf"));
/* vm_fault */
puts("access 2");
assert(!strcmp(address2, "asdf"));
/* vm_fault */
strcpy(address1, "qwer");
/* Also modified. So both virtual addresses point to the same physical address. */
assert(!strcmp(address2, "qwer"));
/* Check that the physical addresses are the same.
* They are, but TODO why virt_to_phys on kernel gives a different value? */
assert(!virt_to_phys_user(&paddr, getpid(), (uintptr_t)address1));
printf("paddr1 = 0x%jx\n", (uintmax_t)paddr);
assert(!virt_to_phys_user(&paddr, getpid(), (uintptr_t)address2));
printf("paddr2 = 0x%jx\n", (uintmax_t)paddr);
/* Check that modifications made from userland are also visible from the kernel. */
read(fd, buf, BUFFER_SIZE);
assert(!memcmp(buf, "qwer", BUFFER_SIZE));
/* Modify the data from the kernel, and check that the change is visible from userland. */
write(fd, "zxcv", 4);
assert(!strcmp(address1, "zxcv"));
assert(!strcmp(address2, "zxcv"));
/* Cleanup. */
puts("munmap 1");
if (munmap(address1, page_size)) {
perror("munmap");
assert(0);
}
puts("munmap 2");
if (munmap(address2, page_size)) {
perror("munmap");
assert(0);
}
puts("close");
close(fd);
return EXIT_SUCCESS;
}
GitHub upstream.
Tested on kernel 5.4.3.
Though the pages are reserved via a kernel driver, it is meant to be accessed via user space. As a result, the PTE (page table entries) do not know if the pfn belongs to user space or kernel space (even though they are allocated via kernel driver).
This is why they are marked with SetPageReserved.
I am writing a module for the Linux kernel, and I want to create some device nodes in the init() function:
int init_module(void)
{
Major = register_chrdev(0, DEVICE_NAME, &fops);
// Now I want to create device nodes
// with the returned major number
}
I also want the kernel to assign a minor number for my first node, and then I will assign the other nodes' minor numbers by myself.
How can I do this in the code? I don’t want to create devices from the shell using mknod().
To have more control over the device numbers and the device creation, you could do the following steps (instead of register_chrdev()):
Call alloc_chrdev_region() to get a major number and a range of minor numbers to work with.
Create a device class for your devices with class_create().
For each device, call cdev_init() and cdev_add() to add the character device to the system.
For each device, call device_create(). As a result, among other things, Udev will create device nodes for your devices. There isn’t any need for mknod() or the like. device_create() also allows you to control the names of the devices.
There are probably many examples of this on the Internet, and one of them is here.
static int __init ofcd_init(void) /* Constructor */
{
printk(KERN_INFO "Welcome!");
if (alloc_chrdev_region(&first, 0, 1, "char_dev") < 0) //$cat /proc/devices
{
return -1;
}
if ((cl = class_create(THIS_MODULE, "chardrv")) == NULL) //$ls /sys/class
{
unregister_chrdev_region(first, 1);
return -1;
}
if (device_create(cl, NULL, first, NULL, "mynull") == NULL) //$ls /dev/
{
class_destroy(cl);
unregister_chrdev_region(first, 1);
return -1;
}
cdev_init(&c_dev, &fops);
if (cdev_add(&c_dev, first, 1) == -1)
{
device_destroy(cl, first);
class_destroy(cl);
unregister_chrdev_region(first, 1);
return -1;
}
return 0;
}
Minimal runnable example
Minimized from other answers. GitHub upstream with test setup.
#include <linux/cdev.h>
#include <linux/device.h>
#include <linux/fs.h> /* register_chrdev, unregister_chrdev */
#include <linux/module.h>
#include <linux/seq_file.h> /* seq_read, seq_lseek, single_release */
#define NAME "lkmc_character_device_create"
static int major = -1;
static struct cdev mycdev;
static struct class *myclass = NULL;
static int show(struct seq_file *m, void *v)
{
seq_printf(m, "abcd");
return 0;
}
static int open(struct inode *inode, struct file *file)
{
return single_open(file, show, NULL);
}
static const struct file_operations fops = {
.llseek = seq_lseek,
.open = open,
.owner = THIS_MODULE,
.read = seq_read,
.release = single_release,
};
static void cleanup(int device_created)
{
if (device_created) {
device_destroy(myclass, major);
cdev_del(&mycdev);
}
if (myclass)
class_destroy(myclass);
if (major != -1)
unregister_chrdev_region(major, 1);
}
static int myinit(void)
{
int device_created = 0;
/* cat /proc/devices */
if (alloc_chrdev_region(&major, 0, 1, NAME "_proc") < 0)
goto error;
/* ls /sys/class */
if ((myclass = class_create(THIS_MODULE, NAME "_sys")) == NULL)
goto error;
/* ls /dev/ */
if (device_create(myclass, NULL, major, NULL, NAME "_dev") == NULL)
goto error;
device_created = 1;
cdev_init(&mycdev, &fops);
if (cdev_add(&mycdev, major, 1) == -1)
goto error;
return 0;
error:
cleanup(device_created);
return -1;
}
static void myexit(void)
{
cleanup(1);
}
module_init(myinit)
module_exit(myexit)
MODULE_LICENSE("GPL");