Advanced Char Driver Operations Sarah Diesburg CIS 4930

Resources LDD Chapter 3 LDD module source code for 3.2.x Red font in slides where up-to-date code diverges from book LDD module source code for 3.2.x http://ww2.cs.fsu.edu/~diesburg/courses/dd/code.html

Resources LXR – Cross-referenced Linux Go to http://lxr.linux.no/ Click on Linux 2.6.11 and later Select your kernel version from drop-down menu

Topics Managing ioctl command numbers Block/unblocking a process Seeking on a device Access control

ioctl For operations beyond simple data transfers Alternatives Eject the media Report error information Change hardware settings Self destruct Alternatives Embedded commands in the data stream Driver-specific file systems

ioctl User-level interface int ioctl(int fd, int request, ...); ... Variable number of arguments Problematic for the system call interface In this context, it is meant to pass a single optional argument Traditionally a char *argp Just a way to bypass the type checking For more information, look at man page

ioctl Driver-level interface Ioctl has changed from the LDD3 era int (*unlocked_ioctl) (struct file *filp, unsigned int cmd, unsigned long arg); cmd is passed from the user unchanged arg can be an integer or a pointer Compiler does not type check Ioctl has changed from the LDD3 era Modified to remove the big kernel lock (BKL) http://lwn.net/Articles/119652/

Choosing the ioctl Commands Need a numbering scheme to avoid mistakes E.g., issuing a command to the wrong device (changing the baud rate of an audio device) Check include/linux/ioctl.h and directory Documentation/ioctl/

Choosing the ioctl Commands A command number uses four bitfields Defined in <linux/ioctl.h> < direction, type, number, size> direction: direction of data transfer _IOC_NONE _IOC_READ _IOC_WRITE _IOC_READ | WRITE

Choosing the ioctl Commands type (ioctl device type) 8-bit (_IOC_TYPEBITS) magic number Associated with the device number 8-bit (_IOC_NRBITS) sequential number Unique within device size: size of user data involved The width is either 13 or 14 bits (_IOC_SIZEBITS)

Choosing the ioctl Commands Useful macros to create ioctl command numbers _IO(type, nr) _IOR(type, nr, datatype) _IOW(type, nr, datatype) _IOWR(type, nr, datatype) size = sizeof(datatype)

Choosing the ioctl Commands Useful macros to decode ioctl command numbers _IOC_DIR(nr) _IOC_TYPE(nr) _IOC_NR(nr) _IOC_SIZE(nr)

Choosing the ioctl Commands The scull example /* Use 'k' as magic number */ #define SCULL_IOC_MAGIC 'k‘ /* Please use a different 8-bit number in your code */ #define SCULL_IOCRESET _IO(SCULL_IOC_MAGIC, 0)

Choosing the ioctl Commands The scull example /* * S means "Set" through a ptr, * T means "Tell" directly with the argument value * G means "Get": reply by setting through a pointer * Q means "Query": response is on the return value * X means "eXchange": switch G and S atomically * H means "sHift": switch T and Q atomically */ #define SCULL_IOCSQUANTUM _IOW(SCULL_IOC_MAGIC, 1, int) #define SCULL_IOCSQSET _IOW(SCULL_IOC_MAGIC, 2, int) #define SCULL_IOCTQUANTUM _IO(SCULL_IOC_MAGIC, 3) #define SCULL_IOCTQSET _IO(SCULL_IOC_MAGIC, 4) #define SCULL_IOCGQUANTUM _IOR(SCULL_IOC_MAGIC, 5, int) Set new value and return the old value

Choosing the ioctl Commands The scull example #define SCULL_IOCGQSET _IOR(SCULL_IOC_MAGIC, 6, int) #define SCULL_IOCQQUANTUM _IO(SCULL_IOC_MAGIC, 7) #define SCULL_IOCQQSET _IO(SCULL_IOC_MAGIC, 8) #define SCULL_IOCXQUANTUM _IOWR(SCULL_IOC_MAGIC, 9, int) #define SCULL_IOCXQSET _IOWR(SCULL_IOC_MAGIC,10, int) #define SCULL_IOCHQUANTUM _IO(SCULL_IOC_MAGIC, 11) #define SCULL_IOCHQSET _IO(SCULL_IOC_MAGIC, 12) #define SCULL_IOC_MAXNR 14

The Return Value When the command number is not supported Return –EINVAL Or –ENOTTY (according to the POSIX standard)

The Predefined Commands Handled by the kernel first Will not be passed down to device drivers Three groups For any file (regular, device, FIFO, socket) Magic number: “T.” For regular files only Specific to the file system type

Using the ioctl Argument If it is an integer, just use it directly If it is a pointer Need to check for valid user address int access_ok(int type, const void *addr, unsigned long size); type: either VERIFY_READ or VERIFY_WRITE Returns 1 for success, 0 for failure Driver then results –EFAULT to the caller Defined in <linux/uaccess.h> Mostly called by memory-access routines

Using the ioctl Argument The scull example int scull_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) { int err = 0, tmp; int retval = 0; /* check the magic number and whether the command is defined */ if (_IOC_TYPE(cmd) != SCULL_IOC_MAGIC) { return -ENOTTY; } if (_IOC_NR(cmd) > SCULL_IOC_MAXNR) { …

Using the ioctl Argument The scull example … /* the concept of "read" and "write" is reversed here */ if (_IOC_DIR(cmd) & _IOC_READ) { err = !access_ok(VERIFY_WRITE, (void __user *) arg, _IOC_SIZE(cmd)); } else if (_IOC_DIR(cmd) & _IOC_WRITE) { err = !access_ok(VERIFY_READ, (void __user *) arg, } if (err) return -EFAULT;

Using the ioctl Argument Data transfer functions optimized for most used data sizes (1, 2, 4, and 8 bytes) If the size mismatches Cryptic compiler error message: Conversion to non-scalar type requested Use copy_to_user and copy_from_user #include <linux/uaccess.h> put_user(datum, ptr) Writes to a user-space address Calls access_ok() Returns 0 on success, -EFAULT on error

Using the ioctl Argument __put_user(datum, ptr) Does not check access_ok() Can still fail if the user-space memory is not writable get_user(local, ptr) Reads from a user-space address Calls access_ok() Stores the retrieved value in local Returns 0 on success, -EFAULT on error __get_user(local, ptr) Can still fail if the user-space memory is not readable

Capabilities and Restricted Operations Limit certain ioctl operations to privileged users See <linux/capability.h> for the full set of capabilities To check a certain capability call int capable(int capability); In the scull example if (!capable(CAP_SYS_ADMIN)) { return –EPERM; } A catch-all capability for many system administration operations

The Implementation of the ioctl Commands A giant switch statement … switch(cmd) { case SCULL_IOCRESET: scull_quantum = SCULL_QUANTUM; scull_qset = SCULL_QSET; break; case SCULL_IOCSQUANTUM: /* Set: arg points to the value */ if (!capable(CAP_SYS_ADMIN)) { return -EPERM; } retval = __get_user(scull_quantum, (int __user *)arg);

The Implementation of the ioctl Commands … case SCULL_IOCTQUANTUM: /* Tell: arg is the value */ if (!capable(CAP_SYS_ADMIN)) { return -EPERM; } scull_quantum = arg; break; case SCULL_IOCGQUANTUM: /* Get: arg is pointer to result */ retval = __put_user(scull_quantum, (int __user *) arg); case SCULL_IOCQQUANTUM: /* Query: return it (> 0) */ return scull_quantum;

The Implementation of the ioctl Commands … case SCULL_IOCXQUANTUM: /* eXchange: use arg as pointer */ if (!capable(CAP_SYS_ADMIN)) { return -EPERM; } tmp = scull_quantum; retval = __get_user(scull_quantum, (int __user *) arg); if (retval == 0) { retval = __put_user(tmp, (int __user *) arg); break;

The Implementation of the ioctl Commands … case SCULL_IOCHQUANTUM: /* sHift: like Tell + Query */ if (!capable(CAP_SYS_ADMIN)) { return -EPERM; } tmp = scull_quantum; scull_quantum = arg; return tmp; default: /* redundant, as cmd was checked against MAXNR */ return -ENOTTY; } /* switch */ return retval; } /* scull_ioctl */

The Implementation of the ioctl Commands Six ways to pass and receive arguments from the user space Need to know command number int quantum; ioctl(fd,SCULL_IOCSQUANTUM, &quantum); /* Set by pointer */ ioctl(fd,SCULL_IOCTQUANTUM, quantum); /* Set by value */ ioctl(fd,SCULL_IOCGQUANTUM, &quantum); /* Get by pointer */ quantum = ioctl(fd,SCULL_IOCQQUANTUM); /* Get by return value */ ioctl(fd,SCULL_IOCXQUANTUM, &quantum); /* Exchange by pointer */ /* Exchange by value */ quantum = ioctl(fd,SCULL_IOCHQUANTUM, quantum);

Device Control Without ioctl Writing control sequences into the data stream itself Example: console escape sequences Advantages: No need to implement ioctl methods Disadvantages: Need to make sure that escape sequences do not appear in the normal data stream (e.g., cat a binary file) Need to parse the data stream

Blocking I/O Needed when no data is available for reads When the device is not ready to accept data Output buffer is full

Introduction to Sleeping

Introduction to Sleeping A process is removed from the scheduler’s run queue Certain rules Never sleep when running in an atomic context Multiple steps must be performed without concurrent accesses Not while holding a spinlock, seqlock, or RCU lock Not while disabling interrupts

Introduction to Sleeping Okay to sleep while holding a semaphore Other threads waiting for the semaphore will also sleep Need to keep it short Make sure that it is not blocking the process that will wake it up After waking up Make no assumptions about the state of the system The resource one is waiting for might be gone again Must check the wait condition again

Introduction to Sleeping Wait queue: contains a list of processes waiting for a specific event #include <linux/wait.h> To initialize statically, call DECLARE_WAIT_QUEUE_HEAD(my_queue); To initialize dynamically, call wait_queue_head_t my_queue; init_waitqueue_head(&my_queue);

Simple Sleeping Call variants of wait_event macros wait_event(queue, condition) queue = wait queue head Passed by value Waits until the boolean condition becomes true Puts into an uninterruptible sleep Usually is not what you want wait_event_interruptible(queue, condition) Can be interrupted by signals Returns nonzero if sleep was interrupted Your driver should return -ERESTARTSYS

Simple Sleeping wait_event_timeout(queue, condition, timeout) Wait for a limited time (in jiffies) Returns 0 regardless of condition evaluations wait_event_interruptible_timeout(queue, condition, timeout)

Simple Sleeping To wake up, call variants of wake_up functions void wake_up(wait_queue_head_t *queue); Wakes up all processes waiting on the queue void wake_up_interruptible(wait_queue_head_t *queue); Wakes up processes that perform an interruptible sleep

Multiple threads can wake up at this point Simple Sleeping Example module: sleepy static DECLARE_WAIT_QUEUE_HEAD(wq); static int flag = 0; ssize_t sleepy_read(struct file *filp, char __user *buf, size_t count, loff_t *pos) { printk(KERN_DEBUG "process %i (%s) going to sleep\n", current->pid, current->comm); wait_event_interruptible(wq, flag != 0); flag = 0; printk(KERN_DEBUG "awoken %i (%s)\n", current->pid, current->comm); return 0; /* EOF */ } Multiple threads can wake up at this point

Simple Sleeping Example module: sleepy ssize_t sleepy_write(struct file *filp, const char __user *buf, size_t count, loff_t *pos) { printk(KERN_DEBUG "process %i (%s) awakening the readers...\n", current->pid, current->comm); flag = 1; wake_up_interruptible(&wq); return count; /* succeed, to avoid retrial */ }

Blocking and Nonblocking Operations By default, operations block If no data is available for reads If no space is available for writes Non-blocking I/O is indicated by the O_NONBLOCK flag in filp->f_flags Defined in <linux/fcntl.h> Only open, read, and write calls are affected Returns –EAGAIN immediately instead of block Applications need to distinguish non-blocking returns vs. EOFs

A Blocking I/O Example scullpipe A read process A write process Blocks when no data is available Wakes a blocking write when buffer space becomes available A write process Blocks when no buffer space is available Wakes a blocking read process when data arrives

A Blocking I/O Example scullpipe data structure struct scull_pipe { wait_queue_head_t inq, outq; /* read and write queues */ char *buffer, *end; /* begin of buf, end of buf */ int buffersize; /* used in pointer arithmetic */ char *rp, *wp; /* where to read, where to write */ int nreaders, nwriters; /* number of openings for r/w */ struct fasync_struct *async_queue; /* asynchronous readers */ struct semaphore sem; /* mutual exclusion semaphore */ struct cdev cdev; /* Char device structure */ };

A Blocking I/O Example static ssize_t scull_p_read(struct file *filp, char __user *buf, size_t count, loff_t *f_pos) { struct scull_pipe *dev = filp->private_data; if (down_interruptible(&dev->sem)) return -ERESTARTSYS; while (dev->rp == dev->wp) { /* nothing to read */ up(&dev->sem); /* release the lock */ if (filp->f_flags & O_NONBLOCK) return -EAGAIN; if (wait_event_interruptible(dev->inq, (dev->rp != dev->wp))) return -ERESTARTSYS; }

A Blocking I/O Example if (dev->wp > dev->rp) count = min(count, (size_t)(dev->wp - dev->rp)); else /* the write pointer has wrapped */ count = min(count, (size_t)(dev->end - dev->rp)); if (copy_to_user(buf, dev->rp, count)) { up (&dev->sem); return -EFAULT; } dev->rp += count; if (dev->rp == dev->end) dev->rp = dev->buffer; /* wrapped */ /* finally, awake any writers and return */ wake_up_interruptible(&dev->outq); return count;

Advanced Sleeping

Advanced Sleeping Uses low-level functions to affect a sleep How a process sleeps 1. Allocate and initialize a wait_queue_t structure DEFINE_WAIT(my_wait); Or wait_queue_t my_wait; init_wait(&my_wait); Queue element

Advanced Sleeping 2. Add to the proper wait queue and mark a process as being asleep TASK_RUNNING TASK_INTERRUPTIBLE or TASK_UNINTERRUPTIBLE Call void prepare_to_wait(wait_queue_head_t *queue, wait_queue_t *wait, int state);

Advanced Sleeping 3. Give up the processor Double check the sleeping condition before going to sleep The wakeup thread might have changed the condition between steps 1 and 2 if (/* sleeping condition */) { schedule(); /* yield the CPU */ }

Advanced Sleeping 4. Return from sleep Remove the process from the wait queue if schedule() was not called void finish_wait(wait_queue_head_t *queue, wait_queue_t *wait);

Advanced Sleeping scullpipe write method /* How much space is free? */ static int spacefree(struct scull_pipe *dev) { if (dev->rp == dev->wp) return dev->buffersize - 1; return ((dev->rp + dev->buffersize - dev->wp) % dev->buffersize) - 1; }

Advanced Sleeping static ssize_t scull_p_write(struct file *filp, const char __user *buf, size_t count, loff_t *f_pos) { struct scull_pipe *dev = filp->private_data; int result; if (down_interruptible(&dev->sem)) return -ERESTARTSYS; /* Wait for space for writing */ result = scull_getwritespace(dev, filp); if (result) return result; /* scull_getwritespace called up(&dev->sem) */ /* ok, space is there, accept something */ count = min(count, (size_t)spacefree(dev));

Advanced Sleeping if (dev->wp >= dev->rp) count = min(count, (size_t)(dev->end - dev->wp)); else /* the write pointer has wrapped, fill up to rp - 1 */ count = min(count, (size_t)(dev->rp - dev->wp - 1)); if (copy_from_user(dev->wp, buf, count)) { up (&dev->sem); return -EFAULT; } dev->wp += count; if (dev->wp == dev->end) dev->wp = dev->buffer; /* wrapped */ up(&dev->sem); wake_up_interruptible(&dev->inq); if (dev->async_queue) kill_fasync(&dev->async_queue, SIGIO, POLL_IN); return count;

Advanced Sleeping Queue: full Task state: RUNNING /* Wait for space for writing; caller must hold device semaphore. * On error the semaphore will be released before returning. */ static int scull_getwritespace(struct scull_pipe *dev, struct file *filp) { while (spacefree(dev) == 0) { /* full */ DEFINE_WAIT(wait); up(&dev->sem); if (filp->f_flags & O_NONBLOCK) return -EAGAIN; prepare_to_wait(&dev->outq, &wait, TASK_INTERRUPTIBLE); if (spacefree(dev) == 0) schedule(); finish_wait(&dev->outq, &wait); if (signal_pending(current)) return -ERESTARTSYS; if (down_interruptible(&dev->sem)) return -ERESTARTSYS; } return 0; Queue: full Task state: RUNNING

Advanced Sleeping Queue: full Task state: RUNNING  INTERRUPTIBLE /* Wait for space for writing; caller must hold device semaphore. * On error the semaphore will be released before returning. */ static int scull_getwritespace(struct scull_pipe *dev, struct file *filp) { while (spacefree(dev) == 0) { /* full */ DEFINE_WAIT(wait); up(&dev->sem); if (filp->f_flags & O_NONBLOCK) return -EAGAIN; prepare_to_wait(&dev->outq, &wait, TASK_INTERRUPTIBLE); if (spacefree(dev) == 0) schedule(); finish_wait(&dev->outq, &wait); if (signal_pending(current)) return -ERESTARTSYS; if (down_interruptible(&dev->sem)) return -ERESTARTSYS; } return 0; Queue: full Task state: RUNNING  INTERRUPTIBLE

Advanced Sleeping Queue: full Task state: INTERRUPTIBLE /* sleep */ /* Wait for space for writing; caller must hold device semaphore. * On error the semaphore will be released before returning. */ static int scull_getwritespace(struct scull_pipe *dev, struct file *filp) { while (spacefree(dev) == 0) { /* full */ DEFINE_WAIT(wait); up(&dev->sem); if (filp->f_flags & O_NONBLOCK) return -EAGAIN; prepare_to_wait(&dev->outq, &wait, TASK_INTERRUPTIBLE); if (spacefree(dev) == 0) schedule(); finish_wait(&dev->outq, &wait); if (signal_pending(current)) return -ERESTARTSYS; if (down_interruptible(&dev->sem)) return -ERESTARTSYS; } return 0; Queue: full Task state: INTERRUPTIBLE /* sleep */

Exclusive Waits Avoid waking up all processes waiting on a queue Call Wakes up only one process Call void prepare_to_wait_exclusive(wait_queue_heat_t *queue, wait_queue_t *wait, int state); Set the WQ_FLAG_EXCLUSIVE flag Add the queue entry to the end of the wait queue wake_up stops after waking the first process with the flag set

The Details of Waking Up /* wakes up all processes waiting on the queue */ void wake_up(wait_queue_head_t *queue); /* wakes up processes that perform an interruptible sleep */ void wake_up_interruptible(wait_queue_head_t *queue); /* wake up to nr exclusive waiters */ void wake_up_nr(wait_queue_head_t *queue, int nr); void wake_up_interruptible_nr(wait_queue_head_t *queue, int nr); /* wake up all exclusive waiters */ void wake_up_all(wait_queue_head_t *queue); void wake_up_interruptible_all(wait_queue_head_t *queue); /* do not lose the CPU during this call */ void wake_up_interruptible_sync(wait_queue_head_t *queue);

Testing the scullpipe Driver Window 1 % cat /dev/scullpipe Window2 %

Testing the scullpipe Driver Window 1 % cat /dev/scullpipe Window2 % ls –aF > /dev/scullpipe

Testing the scullpipe Driver Window 1 % cat /dev/scullpipe ./ ../ file1 file2 Window2 % ls –aF > /dev/scullpipe

poll and select Nonblocking I/Os often involve the use of poll, select, and epoll system calls Allow a process to determine whether it can read or write one or more open files without blocking Can block a process until any of a set of file descriptors becomes available for reading and writing select introduced in BSD Linux poll introduced in System V epoll added in 2.5.45 for better scaling

poll and select All three calls supported through the poll method unsigned int (*poll) (struct file *filp, poll_table *wait); 1. Call poll_wait on one or more wait queues that could indicate a change in the poll status If no file descriptors are available, wait 2. Return a bit mask describing the operations that could be immediately performed without blocking

poll and select poll_table defined in <linux/poll.h> To add a wait queue into the poll_table, call void poll_wait(struct file *, wait_queue_head_t *, poll_table *); Bit mask flags defined in <linux/poll.h> POLLIN Set if the device can be read without blocking

poll and select POLLOUT POLLRDNORM POLLWRNORM POLLPRI Set if the device can be written without blocking POLLRDNORM Set if “normal” data is available for reading A readable device returns (POLLIN | POLLRDNORM) POLLWRNORM Same meaning as POLLOUT A writable device returns (POLLOUT | POLLWRNORM) POLLPRI High-priority data can be read without blocking

poll and select POLLHUP POLLERR POLLRDBAND POLLWRBAND Returns when a process reads the end-of-file POLLERR An error condition has occurred POLLRDBAND Out-of-band data is available for reading Associated with sockets POLLWRBAND Data with nonzero priority can be written to the device

poll and select Example static unsigned int scull_p_poll(struct file *filp, poll_table *wait) { struct scull_pipe *dev = filp->private_data; unsigned int mask = 0; down(&dev->sem); poll_wait(filp, &dev->inq, wait); poll_wait(filp, &dev->outq, wait); if (dev->rp != dev->wp) /* circular buffer not empty */ mask |= POLLIN | POLLRDNORM; /* readable */ if (spacefree(dev)) /* circular buffer not full */ mask |= POLLOUT | POLLWRNORM; /* writable */ up(&dev->sem); return mask; }

poll and select No end-of-file support Scull pipe does not implement this If it did… The reader could see an end-of-file when all writers close the file Check dev->nwriters in read and poll Problem when a reader opens the scullpipe before the writer Need blocking within open

Interaction with read and write Reading from the device If there is data in the input buffer, return at least one byte poll returns POLLIN | POLLRDNORM If no data is available If O_NONBLOCK is set, return –EAGAIN poll must report the device unreadable until one byte arrives At the end-of-file, read returns 0, poll returns POLLHUP

Interaction with read and write Writing to the device If there is space in the output buffer, accept at least one byte poll reports that the devices is writable by returning POLLOUT | POLLWRNORM If the output buffer is full, write blocks If O_NONBLOCK is set, write returns –EAGAIN poll reports that the file is not writable If the device is full, write returns -ENOSPC

Interaction with read and write In write, never wait for data transmission before returning Or, select may block To make sure the output buffer is actually transmitted, use fsync call

Interaction with read and write To flush pending output, call fsync int (*fsync) (struct file *file, loff_t, loff_t, int datasync); Should return only when the device has been completely flushed datasync: Used by file systems, ignored by drivers

The Underlying Data Structure

The Underlying Data Structure When the poll call completes, poll_table is deallocated with all wait queue entries removed epoll reduces this overhead of setting up and tearing down the data structure between every I/O

Asynchronous Notification Polling Inefficient for rare events A solution: asynchronous notification Application receives a signal whenever data becomes available Two steps Specify a process as the owner of the file (so that the kernel knows whom to notify) Set the FASYNC flag in the device via fcntl command

Asynchronous Notification Example (user space) /* create a signal handler */ signal(SIGIO, &input_handler); /* set current pid the owner of the stdin */ fcntl(STDIN_FILENO, F_SETOWN, getpid()); /* obtain the current file control flags */ oflags = fcntl(STDIN_FILENO, F_GETFL); /* set the asynchronous flag */ fcntl(STDIN_FILENO, F_SETFL, oflags | FASYNC);

Asynchronous Notification Some catches Not all devices support asynchronous notification Usually available for sockets and ttys Need to know which input file to process Still need to use poll or select

The Driver’s Point of View 1. When F_SETOWN is invoked, a value is assigned to filp->f_owner 2. When F_SETFL is executed to change the status of FASYNC The driver’s fasync method is called static int scull_p_fasync(int fd, struct file *filp, int mode) { struct scull_pipe *dev = filp->private_data; return fasync_helper(fd, filp, mode, &dev->async_queue); }

The Driver’s Point of View fasync_helper adds or removes processes from the asynchronous list void fasync_helper(int fd, struct file *filp, int mode, struct fasync_struct **fa); 3. When data arrives, send a SIGNO signal to all processes registered for asynchronous notification Near the end of write, notify blocked readers if (dev->async_queue) kill_fasync(&dev->async_queue, SIGIO, POLL_IN); Similarly for read, as needed

The Driver’s Point of View 4. When the file is closed, remove the file from the list of asynchronous readers in the release method scull_p_fasync(-1, filp, 0);

The llseek Implementation Implements lseek and llseek system calls Modifies filp->f_pos loff_t scull_llseek(struct file *filp, loff_t off, int whence) { struct scull_dev *dev = filp->private_data; loff_t newpos; switch(whence) { case 0: /* SEEK_SET */ newpos = off; break; case 1: /* SEEK_CUR, relative to the current position */ newpos = filp->f_pos + off;

The llseek Implementation case 2: /* SEEK_END, relative to the end of the file */ newpos = dev->size + off; break; default: /* can't happen */ return -EINVAL; } if (newpos < 0) return -EINVAL; filp->f_pos = newpos; return newpos;

The llseek Implementation Does not make sense for serial ports and keyboard inputs Need to inform the kernel via calling nonseekable_open in the open method int nonseekable_open(struct inode *inode, struct file *filp); Replace llseek method with no_llseek (defined in <linux/fs.h> in your file_operations structure

Access Control on a Device File Prevents unauthorized users from using the device Sometimes permits only one authorized user to open the device at a time

Returns true, if the tested value is 0 Single-Open Devices Example: scullsingle static atomic_t scull_s_available = ATOMIC_INIT(1); static int scull_s_open(struct inode *inode, struct file *filp) { struct scull_dev *dev = &scull_s_device; if (!atomic_dec_and_test(&scull_s_available)) { atomic_inc(&scull_s_available); return -EBUSY; /* already open */ } /* then, everything else is the same as before */ if ((filp->f_flags & O_ACCMODE) == O_WRONLY) scull_trim(dev); filp->private_data = dev; return 0; /* success */ Returns true, if the tested value is 0

Single-Open Devices In the release call, marks the device idle static int scull_s_release(struct inode *inode, struct file *filp) { atomic_inc(&scull_s_available); /* release the device */ return 0; }

Restricting Access to a Single User (with multiple processes) at a Time Example: sculluid Includes the following in the open call spin_lock(&scull_u_lock); if (scull_u_count && /* someone is using the device */ (scull_u_owner != current->uid) && /* not the same user */ (scull_u_owner != current->euid) && /* not the same effective uid (for su) */ !capable(CAP_DAC_OVERRIDE)) { /* not root override */ spin_unlock(&scull_u_lock); return -EBUSY; /* -EPERM would confuse the user */ } if (scull_u_count == 0) scull_u_owner = current->uid; scull_u_count++;

Restricting Access to a Single User (with Multiple Processes) at a Time Includes the following in the release call static int scull_u_release(struct inode *inode, struct file *filp) { spin_lock(&scull_u_lock); scull_u_count--; /* nothing else */ spin_unlock(&scull_u_lock); return 0; }

Blocking open as an Alternative to EBUSY (scullwuid) A user might prefer to wait over getting errors E.g., data communication channel spin_lock(&scull_w_lock); while (!scull_w_available()) { spin_unlock(&scull_w_lock); if (filp->f_flags & O_NONBLOCK) return -EAGAIN; if (wait_event_interruptible(scull_w_wait, scull_w_available())) return -ERESTARTSYS; /* tell the fs layer to handle it */ } if (scull_w_count == 0) scull_w_owner = current->uid; scull_w_count++;

Blocking open as an Alternative to EBUSY (scullwuid) The release method wakes pending processes static int scull_w_release(struct inode *inode, struct file *filp) { int temp; spin_lock(&scull_w_lock); scull_w_count--; temp = scull_w_count; spin_unlock(&scull_w_lock); if (temp == 0) wake_up_interruptible_sync(&scull_w_wait); return 0; }

Blocking open as an Alternative to EBUSY Might not be the right semantics for interactive users Blocking on cp vs. getting a return value –EBUSY or -EPERM Incompatible policies for the same device One solution: one device node per policy

Cloning the Device on open Allows the creation of private, virtual devices E.g., One virtual scull device for each process with different tty device number Example: scullpriv

Cloning the Device on open static int scull_c_open(struct inode *inode, struct file *filp) { struct scull_dev *dev; dev_t key; if (!current->signal->tty) { PDEBUG("Process \"%s\" has no ctl tty\n", current->comm); return -EINVAL; } key = tty_devnum(current->signal->tty); spin_lock(&scull_c_lock); dev = scull_c_lookfor_device(key); spin_unlock(&scull_c_lock); if (!dev) return -ENOMEM; .../* then, everything else is the same as before */

Cloning the Device on open /* The clone-specific data structure includes a key field */ struct scull_listitem { struct scull_dev device; dev_t key; struct list_head list; }; /* The list of devices, and a lock to protect it */ static LIST_HEAD(scull_c_list); static spinlock_t scull_c_lock = SPIN_LOCK_UNLOCKED;

Cloning the Device on open /* Look for a device or create one if missing */ static struct scull_dev *scull_c_lookfor_device(dev_t key) { struct scull_listitem *lptr; list_for_each_entry(lptr, &scull_c_list, list) { if (lptr->key == key) return &(lptr->device); } /* not found */ lptr = kmalloc(sizeof(struct scull_listitem), GFP_KERNEL); if (!lptr) return NULL;

Cloning the Device on open /* initialize the device */ memset(lptr, 0, sizeof(struct scull_listitem)); lptr->key = key; scull_trim(&(lptr->device)); /* initialize it */ init_MUTEX(&(lptr->device.sem)); /* place it in the list */ list_add(&lptr->list, &scull_c_list); return &(lptr->device); }

What’s going on? scull_listitem struct scull_dev device; dev_t key; scull_c_list struct list_head { struct list_head *next; struct list_head *prev; }; struct list_head { struct list_head *next; struct list_head *prev; } list;