Linux Kernel Architecture

Chapter 10: Filesystems without Persistent Storage

Implementationofproc_file_write

Writing toprocfiles is also a simple matter — at least from the perspective of the filesystem. The code of

proc_file_writeis very compact and thus is reproduced in full below.

fs/proc/generic.c

static ssize_t

proc_file_write(struct file * file, const char __user *buffer,

size_t count, loff_t *ppos)

{

struct inode *inode = file->f_dentry->d_inode;

struct proc_dir_entry * dp;

dp = PDE(inode);

if (!dp->write_proc)

return -EIO;

return dp->write_proc(file, buffer, count, dp->data);

}

ThePDEfunction needed to obtain the requiredproc_dir_entryinstance from the VFS inode using

the container mechanism is very simple. All it does is executePROC_I(inode)->pde. As discussed in

Section 10.1.2,PROC_Ifinds theproc_inodeinstance associated with aninode(in the case ofprocinodes,

the inode data always immediately precede the VFS inode).

Once theproc_dir_entryinstance has been found, the routine registered for write purposes must be

invoked with suitable parameters — assuming, of course, that the routine exists and is not assigned a

null pointer.

How does the kernel implement a write routine forprocentries? This question is answered using

proc_write_foobar, which is included as an example for a write handler in the kernel sources:

kernel/Documentation/DocBook/procfs_example.c

static int proc_write_foobar(struct file *file,

const char *buffer,

unsigned long count,

void *data)

{

int len;

struct fb_data_t *fb_data = (struct fb_data_t *)data;

if(count > FOOBAR_LEN)

len = FOOBAR_LEN;

else

len = count;

if(copy_from_user(fb_data->value, buffer, len))

return -EFAULT;

fb_data->value[len] = ’\0’;