* /dev/poll patch for 2.4
@ 2001-03-08 15:22 Mathieu Dube
0 siblings, 0 replies; only message in thread
From: Mathieu Dube @ 2001-03-08 15:22 UTC (permalink / raw)
To: linux-kernel
[-- Attachment #1: Type: text/plain, Size: 156 bytes --]
Hi,
Somebody asked me for that patch it might need some tweaking, adding
#include <linux/poll.h> somewhere
--
Mathieu Dube
Mondo-Live
www.flipr.com
[-- Attachment #2: devpoll.2.4.0.diff --]
[-- Type: text/plain, Size: 34379 bytes --]
diff -urP linux-2.4.0/drivers/char/Config.in linux-2.4.0-devpoll/drivers/char/Config.in
--- linux-2.4.0/drivers/char/Config.in Fri Dec 29 14:07:21 2000
+++ linux-2.4.0-devpoll/drivers/char/Config.in Mon Jan 29 16:24:42 2001
@@ -156,6 +156,7 @@
dep_tristate 'Intel i8x0 Random Number Generator support' CONFIG_INTEL_RNG $CONFIG_PCI
tristate '/dev/nvram support' CONFIG_NVRAM
+tristate '/dev/poll support' CONFIG_DEVPOLL
tristate 'Enhanced Real Time Clock Support' CONFIG_RTC
if [ "$CONFIG_IA64" = "y" ]; then
bool 'EFI Real Time Clock Services' CONFIG_EFI_RTC
diff -urP linux-2.4.0/drivers/char/Makefile linux-2.4.0-devpoll/drivers/char/Makefile
--- linux-2.4.0/drivers/char/Makefile Thu Jan 4 13:00:55 2001
+++ linux-2.4.0-devpoll/drivers/char/Makefile Mon Jan 29 16:24:42 2001
@@ -181,6 +181,7 @@
obj-$(CONFIG_60XX_WDT) += sbc60xxwdt.o
obj-$(CONFIG_WDT) += wdt.o
obj-$(CONFIG_WDTPCI) += wdt_pci.o
+obj-$(CONFIG_DEVPOLL) += devpoll.o
obj-$(CONFIG_21285_WATCHDOG) += wdt285.o
obj-$(CONFIG_977_WATCHDOG) += wdt977.o
obj-$(CONFIG_I810_TCO) += i810-tco.o
diff -urP linux-2.4.0/drivers/char/devpoll.c linux-2.4.0-devpoll/drivers/char/devpoll.c
--- linux-2.4.0/drivers/char/devpoll.c Wed Dec 31 16:00:00 1969
+++ linux-2.4.0-devpoll/drivers/char/devpoll.c Tue Jan 30 05:40:19 2001
@@ -0,0 +1,754 @@
+/*
+ * /dev/poll
+ * by Niels Provos <provos@citi.umich.edu>
+ *
+ * provides poll() support via /dev/poll as in Solaris.
+ *
+ * Linux 2.3/2.4 port by Michal Ostrowski
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/file.h>
+#include <linux/signal.h>
+#include <linux/errno.h>
+#include <linux/mm.h>
+#include <linux/malloc.h>
+#include <linux/vmalloc.h>
+#include <linux/poll.h>
+#include <linux/miscdevice.h>
+#include <linux/random.h>
+#include <linux/smp_lock.h>
+#include <linux/wrapper.h>
+
+#include <linux/devpoll.h>
+
+#include <asm/uaccess.h>
+#include <asm/system.h>
+#include <asm/io.h>
+
+/*#define DEBUG 1 */
+#ifdef DEBUG
+#define DPRINTK(x) printk x
+#define DNPRINTK(n,x) if (n <= DEBUG) printk x
+#else
+#define DPRINTK(x)
+#define DNPRINTK(n,x)
+#endif
+
+/* Various utility functions */
+
+#define DEFAULT_POLLMASK (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM)
+
+/* Do dynamic hashing */
+
+#define INITIAL_BUCKET_BITS 6
+#define MAX_BUCKET_BITS 16
+#define RESIZE_LENGTH 2
+
+
+static void free_pg_vec(struct devpoll *dp);
+
+
+/* Initalize the hash table */
+
+int
+dp_init(struct devpoll *dp)
+{
+ int i;
+ int num_buckets;
+ DNPRINTK(3,(KERN_INFO "/dev/poll: dp_init\n"));
+
+ dp->dp_lock = RW_LOCK_UNLOCKED;
+ dp->dp_entries = 0;
+ dp->dp_max = 0;
+ dp->dp_avg = dp->dp_count = 0;
+ dp->dp_cached = dp->dp_calls = 0;
+ dp->dp_bucket_bits = INITIAL_BUCKET_BITS;
+ dp->dp_bucket_mask = (1 << INITIAL_BUCKET_BITS) - 1;
+
+ num_buckets = (dp->dp_bucket_mask + 1);
+ dp->dp_tab = kmalloc(num_buckets * sizeof(struct list_head),
+ GFP_KERNEL);
+
+ if (!dp->dp_tab)
+ return -ENOMEM;
+
+ for (i = 0; i < num_buckets ; i++) {
+ INIT_LIST_HEAD(&dp->dp_tab[i]);
+ }
+
+ return (0);
+}
+
+int
+dp_resize(struct devpoll *dp)
+{
+ u_int16_t new_mask, old_mask;
+ int i;
+ struct list_head *new_tab, *old_tab;
+ struct dp_fd *dpfd;
+ unsigned long flags;
+ int num_buckets;
+
+ old_mask = dp->dp_bucket_mask;
+ new_mask = (old_mask + 1) * 2 - 1;
+ num_buckets = new_mask + 1;
+
+ DPRINTK((KERN_INFO "/dev/poll: resize %d -> %d\n",
+ old_mask, new_mask));
+
+ new_tab = kmalloc( num_buckets * sizeof(struct list_head), GFP_KERNEL);
+ if (!new_tab)
+ return -ENOMEM;
+
+ for (i = 0; i < num_buckets; i++) {
+ INIT_LIST_HEAD(&new_tab[i]);
+ }
+
+ old_tab = dp->dp_tab;
+
+ /* Rehash all entries */
+ write_lock_irqsave(&dp->dp_lock, flags);
+ for (i = 0; i <= old_mask; i++) {
+ while(!list_empty(&old_tab[i])){
+ dpfd = list_entry(old_tab[i].next,struct dp_fd,next);
+ list_del(&dpfd->next);
+ INIT_LIST_HEAD(&dpfd->next);
+ list_add(&dpfd->next,&new_tab[dpfd->pfd.fd & new_mask]);
+ }
+ }
+
+ dp->dp_tab = new_tab;
+ dp->dp_bucket_bits++;
+ dp->dp_bucket_mask = new_mask;
+ write_unlock_irqrestore(&dp->dp_lock, flags);
+
+ kfree (old_tab);
+
+ return (0);
+}
+
+int
+dp_insert(struct devpoll *dp, struct pollfd *pfd)
+{
+ struct dp_fd *dpfd;
+ u_int16_t bucket = pfd->fd & dp->dp_bucket_mask;
+ unsigned long flags;
+ struct file *file;
+
+ dpfd = kmalloc(sizeof(struct dp_fd), GFP_KERNEL);
+ if (!dpfd)
+ return -ENOMEM;
+
+ dpfd->flags = 0;
+ set_bit(DPH_DIRTY, &dpfd->flags);
+ dpfd->pfd = *pfd;
+ dpfd->pfd.revents = 0;
+ INIT_LIST_HEAD(&dpfd->next);
+
+ write_lock_irqsave(&dp->dp_lock, flags);
+
+ list_add(&dpfd->next,&dp->dp_tab[bucket]);
+
+ file = fcheck(pfd->fd);
+ if (file != NULL) {
+ write_lock(&(file)->f_dplock);
+ poll_backmap(pfd->fd, dpfd, &(file)->f_backmap);
+ write_unlock(&(file)->f_dplock);
+ set_bit(DPH_BACKMAP, &(dpfd)->flags);
+ }
+ write_unlock_irqrestore(&dp->dp_lock, flags);
+
+ dp->dp_entries++;
+ /* Check if we need to resize the hash table */
+ if ((dp->dp_entries >> dp->dp_bucket_bits) > RESIZE_LENGTH &&
+ dp->dp_bucket_bits < MAX_BUCKET_BITS)
+ dp_resize(dp);
+
+ return (0);
+}
+
+struct dp_fd *
+dp_find(struct devpoll *dp, int fd)
+{
+ struct dp_fd *dpfd=NULL;
+ struct list_head *lh;
+ u_int16_t bucket = fd & dp->dp_bucket_mask;
+
+ read_lock(&dp->dp_lock);
+ list_for_each(lh,&dp->dp_tab[bucket]){
+ dpfd = list_entry(lh,struct dp_fd,next);
+ if(dpfd->pfd.fd == fd ) break;
+ dpfd = NULL;
+ }
+
+
+ read_unlock(&dp->dp_lock);
+ DNPRINTK(2, (KERN_INFO "dp_find: %d -> %p\n", fd, dpfd));
+
+ return dpfd;
+}
+
+void
+dp_delete(struct devpoll *dp, struct dp_fd *dpfd)
+{
+ unsigned long flags;
+ int fd;
+ struct file *filp;
+
+ write_lock_irqsave(&dp->dp_lock, flags);
+ list_del(&dpfd->next);
+
+ INIT_LIST_HEAD(&dpfd->next);
+
+ /* Remove backmaps if necessary */
+ if (current->files) {
+ fd = dpfd->pfd.fd;
+ filp = fcheck(fd);
+
+ if (test_bit(DPH_BACKMAP, &dpfd->flags) &&
+ filp && filp->f_backmap){
+ write_lock(&filp->f_dplock);
+ poll_remove_backmap(&filp->f_backmap, fd,
+ current->files);
+ write_unlock(&filp->f_dplock);
+ }
+ }
+ write_unlock_irqrestore(&dp->dp_lock, flags);
+
+ kfree (dpfd);
+
+ dp->dp_entries--;
+}
+
+void
+dp_free(struct devpoll *dp)
+{
+ int i;
+ struct dp_fd *dpfd = NULL;
+
+ lock_kernel();
+ for (i = 0; i <= dp->dp_bucket_mask; i++) {
+ while(!list_empty(&dp->dp_tab[i])){
+ dpfd = list_entry(dp->dp_tab[i].next,struct dp_fd,next);
+ dp_delete(dp, dpfd);
+ }
+ }
+ unlock_kernel();
+
+ kfree (dp->dp_tab);
+}
+
+
+/*
+ * poll the fds that we keep in our state, return after we reached
+ * max changed fds or are done.
+ * XXX - I do not like how the wait table stuff is done.
+ */
+
+int
+dp_poll(struct devpoll *dp, int max, poll_table *wait,
+ long timeout, struct pollfd *rfds, int usemmap)
+{
+ int count = 0;
+ lock_kernel();
+ read_lock(&dp->dp_lock);
+ for (;;) {
+ unsigned int j=0;
+ struct dp_fd *dpfd = NULL;
+ struct pollfd *fdpnt, pfd;
+ struct file *file;
+
+ set_current_state(TASK_INTERRUPTIBLE);
+ for (j = 0; (j <= dp->dp_bucket_mask) && count < max; j++) {
+ struct list_head *lh;
+ list_for_each(lh, &dp->dp_tab[j]){
+
+ int fd;
+ unsigned int mask = 0;
+ unsigned int rm =0;
+ dpfd = list_entry(lh,struct dp_fd,next);
+
+ if(count>=max){
+ break;
+ }
+
+ fdpnt = &dpfd->pfd;
+ fd = fdpnt->fd;
+
+ /* poll_wait increments f_count if needed */
+ file = fcheck(fd);
+ if (file == NULL) {
+ /* Got to move backward first;
+ * dp_delete will remove lh from
+ * the list otherwise
+ */
+ lh = lh->prev;
+ dp_delete(dp, dpfd);
+ dpfd = NULL;
+ continue;
+ }
+
+ mask = fdpnt->revents;
+ if (test_and_clear_bit(DPH_DIRTY,
+ &dpfd->flags) ||
+ wait != NULL ||
+ (mask & fdpnt->events)) {
+
+ mask = DEFAULT_POLLMASK;
+ if (file->f_op && file->f_op->poll)
+ mask = file->f_op->poll(file, wait);
+ /* if POLLHINT not supported by file
+ * then set bit to dirty ---
+ * must poll this file every time,
+ * otherwise bit will be set by
+ * calls to dp_add_hint
+ */
+ if (!(mask & POLLHINT))
+ set_bit(DPH_DIRTY, &dpfd->flags);
+ fdpnt->revents = mask;
+ }else
+ dp->dp_cached++;
+
+
+ dp->dp_calls++;
+
+ mask &= fdpnt->events | POLLERR | POLLHUP;
+ if (mask) {
+ wait = NULL;
+ count++;
+
+ if (usemmap) {
+ *rfds = *fdpnt;
+ rfds->revents = mask;
+ } else {
+ pfd = *fdpnt;
+ pfd.revents = mask;
+ __copy_to_user(rfds, &pfd,
+ sizeof(struct pollfd));
+ }
+
+ rfds++;
+ }
+ }
+ }
+
+ wait = NULL;
+ if (count || !timeout || signal_pending(current))
+ break;
+ read_unlock(&dp->dp_lock);
+ timeout = schedule_timeout(timeout);
+ read_lock(&dp->dp_lock);
+ }
+ set_current_state(TASK_RUNNING);
+ read_unlock(&dp->dp_lock);
+ unlock_kernel();
+
+ if( !count && signal_pending(current) )
+ return -EINTR;
+
+ return count;
+}
+
+/*
+ * close a /dev/poll
+ */
+
+static int
+close_devpoll(struct inode * inode, struct file * file)
+{
+ struct devpoll *dp = file->private_data;
+
+ DNPRINTK(1, (KERN_INFO "close /dev/poll, max: %d, avg: %d(%d/%d) %d/%d\n",
+ dp->dp_max, dp->dp_avg/dp->dp_count,
+ dp->dp_avg, dp->dp_count,
+ dp->dp_cached, dp->dp_calls));
+
+ /* free allocated memory */
+ if (dp->dp_memvec)
+ free_pg_vec(dp);
+
+ /* Free the hash table */
+ dp_free(dp);
+
+ kfree(dp);
+
+ MOD_DEC_USE_COUNT;
+ return 0;
+}
+
+/*
+ * open a /dev/poll
+ */
+
+static int
+open_devpoll(struct inode * inode, struct file * file)
+{
+ struct devpoll *dp;
+ int r;
+
+ /* allocated state */
+ dp = kmalloc(sizeof(struct devpoll), GFP_KERNEL);
+ if (dp == NULL)
+ return -ENOMEM;
+
+ memset( dp, 0, sizeof(struct devpoll));
+ if ((r = dp_init(dp))) {
+ kfree (dp);
+ return r;
+ }
+
+ file->private_data = dp;
+
+ MOD_INC_USE_COUNT;
+
+ DNPRINTK(3, (KERN_INFO "open /dev/poll\n"));
+
+ return 0;
+}
+
+/*
+ * write to /dev/poll:
+ * a user writes struct pollfds and we add them to our list, or remove
+ * them if (events & POLLREMOVE) is true
+ */
+
+static int
+write_devpoll(struct file *file, const char *buffer, size_t count,
+ loff_t *ppos)
+{
+ int r,rcount;
+ struct devpoll *dp = file->private_data;
+ struct pollfd pfd;
+ struct dp_fd *dpfd;
+#ifdef DEBUG
+ int add = 0, delete = 0, change = 0;
+#endif
+
+ DNPRINTK(3, (KERN_INFO "write /dev/poll %i\n",count));
+
+ if (count % sizeof(struct pollfd))
+ return -EINVAL;
+
+ if ((r = verify_area(VERIFY_READ, buffer, count)))
+ return r;
+
+ rcount = count;
+
+ lock_kernel();
+
+ while (count > 0) {
+ __copy_from_user(&pfd, buffer, sizeof(pfd)); /* no check */
+
+ dpfd = dp_find(dp, pfd.fd);
+
+ if (pfd.fd >= current->files->max_fds ||
+ current->files->fd[pfd.fd] == NULL) {
+ /* Be tolerant, maybe the close happened already */
+ pfd.events = POLLREMOVE;
+ }
+ /* See if we need to remove the file descriptor. If it
+ * already exists OR the event fields, otherwise insert
+ */
+ if (pfd.events & POLLREMOVE) {
+ if (dpfd)
+ dp_delete(dp, dpfd);
+#ifdef DEBUG
+ delete++;
+#endif
+ } else if (dpfd) {
+ /* XXX dpfd->pfd.events |= pfd.events; */
+ dpfd->pfd.events = pfd.events;
+#ifdef DEBUG
+ change++;
+#endif
+ } else {
+ dp_insert(dp, &pfd);
+#ifdef DEBUG
+ add++;
+#endif
+ }
+
+ buffer += sizeof(pfd);
+ count -= sizeof(pfd);
+ }
+
+ unlock_kernel();
+
+ if (dp->dp_max < dp->dp_entries) {
+ dp->dp_max = dp->dp_entries;
+ DNPRINTK(2, (KERN_INFO "/dev/poll: new max %d\n", dp->dp_max));
+ }
+
+ DNPRINTK(3, (KERN_INFO "write /dev/poll: %d entries (%d/%d/%d)\n",
+ dp->dp_entries, add, delete, change));
+
+ return (rcount);
+}
+
+static int
+ioctl_devpoll(struct inode *inode, struct file *file,
+ unsigned int cmd, unsigned long arg)
+{
+ struct devpoll *dp = file->private_data;
+ unsigned mapsize=0;
+ unsigned num_pages=0;
+ int i=0;
+ switch (cmd) {
+ case DP_ALLOC:
+ if (arg > current->rlim[RLIMIT_NOFILE].rlim_cur)
+ return -EINVAL;
+ if (dp->dp_mmap)
+ return -EPERM;
+
+ mapsize = DP_MMAP_SIZE(arg);
+
+ num_pages = ( PAGE_ALIGN(mapsize) >> PAGE_SHIFT);
+
+ dp->dp_memvec = kmalloc( num_pages * sizeof(unsigned long*),
+ GFP_KERNEL);
+
+ if( dp->dp_memvec == NULL )
+ return -EINVAL;
+
+ memset(dp->dp_memvec, 0, num_pages * sizeof(unsigned long*));
+
+ for( i = 0 ; i < num_pages ; ++i){
+ unsigned long map, mapend;
+ dp->dp_memvec[i] = (u_char*)__get_free_pages(GFP_KERNEL,0);
+ if(!dp->dp_memvec[i]){
+ free_pg_vec(dp);
+ return -ENOMEM;
+ }
+
+ mapend = virt_to_page(dp->dp_memvec[i] + PAGE_SIZE - 1);
+ for (map = virt_to_page(dp->dp_memvec[i]);
+ map <= mapend;
+ map++)
+ set_bit(PG_reserved, &mem_map[map].flags);
+
+ ++dp->dp_numvec;
+ }
+
+ dp->dp_nfds = arg;
+
+ DPRINTK((KERN_INFO "allocated %d pollfds\n", dp->dp_nfds));
+
+ return 0;
+ case DP_FREE:
+ if( atomic_read(&dp->dp_mmapped) )
+ return -EBUSY;
+
+ if(dp->dp_memvec[i]){
+ free_pg_vec( dp );
+ }
+
+
+ DPRINTK((KERN_INFO "freed %d pollfds\n", dp->dp_nfds));
+ dp->dp_nfds = 0;
+
+ return 0;
+ case DP_ISPOLLED: {
+ struct pollfd pfd;
+ struct dp_fd *dpfd;
+
+ if (copy_from_user(&pfd, (void *)arg, sizeof(pfd)))
+ return -EFAULT;
+ dpfd = dp_find(dp, pfd.fd);
+ if (dpfd == NULL)
+ return (0);
+
+ /* We poll this fd, return the evens we poll on */
+ pfd.events = dpfd->pfd.events;
+ pfd.revents = 0;
+
+ if (copy_to_user((void *)arg, &pfd, sizeof(pfd)))
+ return -EFAULT;
+ return (1);
+ }
+ case DP_POLL: {
+ struct dvpoll dopoll;
+ int nfds, usemmap = 0;
+ unsigned long timeout;
+ poll_table wait;
+ struct pollfd *rpfds = NULL;
+
+ if (copy_from_user(&dopoll, (void *)arg, sizeof(dopoll)))
+ return -EFAULT;
+
+ /* We do not need to check this value, its user space */
+ nfds = dopoll.dp_nfds;
+ if (nfds <= 0)
+ return -EINVAL;
+
+ if (dopoll.dp_fds == NULL) {
+ if (dp->dp_mmap == NULL )
+ return -EINVAL;
+ rpfds = (struct pollfd*)dp->dp_mmap;
+ usemmap = 1;
+ } else {
+ rpfds = dopoll.dp_fds;
+ if (verify_area(VERIFY_WRITE, rpfds,
+ nfds * sizeof(struct pollfd)))
+ return -EFAULT;
+ usemmap = 0;
+ }
+
+ timeout = dopoll.dp_timeout;
+ if (timeout) {
+ /* Careful about overflow in the intermediate values */
+ if ((unsigned long)timeout < MAX_SCHEDULE_TIMEOUT / HZ)
+ timeout = (timeout*HZ+999)/1000+1;
+ else /* Negative or overflow */
+ timeout = MAX_SCHEDULE_TIMEOUT;
+ }
+
+ /* Initalize wait table */
+ poll_initwait(&wait);
+
+
+
+ nfds = dp_poll(dp, nfds, &wait, timeout, rpfds, usemmap);
+
+ DNPRINTK(2, (KERN_INFO "poll time %ld -> %d\n", timeout, nfds));
+
+
+ poll_freewait(&wait);
+
+ dp->dp_avg += dp->dp_entries;
+ dp->dp_count++;
+
+ return nfds;
+ }
+ default:
+ DPRINTK((KERN_INFO "ioctl(%x) /dev/poll\n", cmd));
+ break;
+ }
+
+ return -EINVAL;
+}
+
+
+static void free_pg_vec(struct devpoll *dp)
+{
+ int i;
+
+ for (i=0; i< dp->dp_numvec; i++) {
+ if (dp->dp_memvec[i]) {
+ unsigned long map, mapend;
+
+ mapend = virt_to_page(dp->dp_memvec[i] + PAGE_SIZE - 1);
+ for (map = virt_to_page(dp->dp_memvec[i]);
+ map <= mapend;
+ map++)
+ clear_bit(PG_reserved, &mem_map[map].flags);
+ free_pages( (unsigned)dp->dp_memvec[i], 0);
+ }
+ }
+ kfree(dp->dp_memvec);
+ dp->dp_numvec = 0 ;
+}
+
+
+static void devpoll_mm_open( struct vm_area_struct * vma){
+ struct file *file = vma->vm_file;
+ struct devpoll *dp = file->private_data;
+ if(dp)
+ atomic_inc(&dp->dp_mmapped);
+}
+
+static void devpoll_mm_close( struct vm_area_struct * vma){
+ struct file *file = vma->vm_file;
+ struct devpoll *dp = file->private_data;
+ if(dp)
+ atomic_dec(&dp->dp_mmapped);
+}
+
+static struct vm_operations_struct devpoll_mmap_ops = {
+ open: devpoll_mm_open,
+ close: devpoll_mm_close,
+};
+
+/*
+ * mmap shared memory. the first half is an array of struct pollfd,
+ * followed by an array of ints to indicate which file descriptors
+ * changed status.
+ */
+
+static int
+mmap_devpoll(struct file *file, struct vm_area_struct *vma)
+{
+ struct devpoll *dp = file->private_data;
+ unsigned long start; /* Evil type to remap_page_range */
+ int i=0;
+ int num_pages = 0;
+ size_t size, mapsize;
+
+ DPRINTK((KERN_INFO "mmap /dev/poll: %lx %lx\n",
+ vma->vm_start, vma->vm_pgoff<<PAGE_SHIFT));
+
+ if ( (vma->vm_pgoff<<PAGE_SHIFT) != 0)
+ return -EINVAL;
+
+ /* Calculate how much memory we can map */
+ size = PAGE_ALIGN(DP_MMAP_SIZE(dp->dp_nfds));
+ mapsize = PAGE_ALIGN(vma->vm_end - vma->vm_start);
+ num_pages = mapsize >> PAGE_SHIFT;
+
+ /* Check if the requested size is within our size */
+ if (mapsize > dp->dp_numvec<<PAGE_SHIFT)
+ return -EINVAL;
+
+
+ start = vma->vm_start;
+ atomic_set(&dp->dp_mmapped,1);
+ for( i = 0 ; i < num_pages ; ++i){
+ if( remap_page_range(start, __pa(dp->dp_memvec[i]),
+ PAGE_SIZE,
+ vma->vm_page_prot) )
+ return -EINVAL;
+ start += PAGE_SIZE;
+ }
+ dp->dp_mmap = (u_char*)vma->vm_start;
+ vma->vm_ops = &devpoll_mmap_ops;
+
+ DPRINTK((KERN_INFO "mmap /dev/poll: %lx %x\n", page, mapsize));
+ return 0;
+}
+
+
+
+struct file_operations devpoll_fops = {
+ write:write_devpoll,
+ ioctl: ioctl_devpoll,
+ mmap: mmap_devpoll,
+ open: open_devpoll,
+ release:close_devpoll
+};
+
+static struct miscdevice devpoll = {
+ DEVPOLL_MINOR, "devpoll", &devpoll_fops
+};
+
+int __init devpoll_init(void)
+{
+ printk(KERN_INFO "/dev/poll driver installed.\n");
+ misc_register(&devpoll);
+
+
+ return 0;
+}
+
+module_init(devpoll_init);
+#ifdef MODULE
+
+void cleanup_module(void)
+{
+ misc_deregister(&devpoll);
+}
+#endif
+
diff -urP linux-2.4.0/fs/file_table.c linux-2.4.0-devpoll/fs/file_table.c
--- linux-2.4.0/fs/file_table.c Tue Dec 5 10:27:31 2000
+++ linux-2.4.0-devpoll/fs/file_table.c Mon Jan 29 16:24:42 2001
@@ -11,6 +11,7 @@
#include <linux/init.h>
#include <linux/module.h>
#include <linux/smp_lock.h>
+#include <linux/spinlock.h>
/* sysctl tunables... */
struct files_stat_struct files_stat = {0, 0, NR_FILE};
@@ -45,6 +46,7 @@
f->f_version = ++event;
f->f_uid = current->fsuid;
f->f_gid = current->fsgid;
+ rwlock_init(&f->f_dplock);
list_add(&f->f_list, &anon_list);
file_list_unlock();
return f;
diff -urP linux-2.4.0/fs/open.c linux-2.4.0-devpoll/fs/open.c
--- linux-2.4.0/fs/open.c Thu Oct 26 08:11:21 2000
+++ linux-2.4.0-devpoll/fs/open.c Mon Jan 29 16:24:42 2001
@@ -13,6 +13,8 @@
#include <linux/dnotify.h>
#include <linux/module.h>
#include <linux/slab.h>
+#include <linux/poll.h>
+#include <linux/devpoll.h>
#include <asm/uaccess.h>
@@ -802,6 +804,14 @@
unlock_kernel();
}
fcntl_dirnotify(0, filp, 0);
+
+ if (filp->f_backmap) {
+ unsigned long flags;
+ write_lock_irqsave(&filp->f_dplock, flags);
+ poll_clean_backmap(&filp->f_backmap);
+ write_unlock_irqrestore(&filp->f_dplock, flags);
+ }
+
locks_remove_posix(filp, id);
fput(filp);
return retval;
@@ -827,6 +837,14 @@
FD_CLR(fd, files->close_on_exec);
__put_unused_fd(files, fd);
write_unlock(&files->file_lock);
+
+ if (filp->f_backmap) {
+ unsigned long flags;
+ write_lock_irqsave(&filp->f_dplock, flags);
+ poll_remove_backmap(&filp->f_backmap, fd, files);
+ write_unlock_irqrestore(&filp->f_dplock, flags);
+ }
+
return filp_close(filp, files);
out_unlock:
diff -urP linux-2.4.0/include/asm-i386/devpoll.h linux-2.4.0-devpoll/include/asm-i386/devpoll.h
--- linux-2.4.0/include/asm-i386/devpoll.h Wed Dec 31 16:00:00 1969
+++ linux-2.4.0-devpoll/include/asm-i386/devpoll.h Mon Jan 29 16:24:42 2001
@@ -0,0 +1,86 @@
+/*
+ *
+ * /dev/poll
+ * by Niels Provos <provos@citi.umich.edu>
+ *
+ * Linux 2.3/2.4 port by Michal Ostrowski
+ */
+
+#ifndef _LINUX_DEVPOLL_H
+#define _LINUX_DEVPOLL_H
+
+#include <asm/bitops.h>
+#include <linux/list.h>
+#include <asm/atomic.h>
+
+#define DPH_DIRTY 0 /* entry is dirty - bit */
+#define DPH_BACKMAP 1 /* file has an fd back map - bit*/
+#ifdef __KERNEL__
+struct dp_fd {
+ struct list_head next;
+ struct pollfd pfd;
+ int flags; /* for hinting */
+};
+
+
+struct devpoll {
+ struct list_head *dp_tab;
+ int dp_entries; /* Entries in hash table */
+ int dp_max; /* statistics */
+ int dp_avg; /* more */
+ int dp_count;
+ int dp_cached;
+ int dp_calls;
+ int dp_bucket_bits;
+ int dp_bucket_mask;
+ int dp_nfds; /* Number of poll fds */
+ u_char* dp_mmap; /* vaddr of mapped region */
+ atomic_t dp_mmapped; /* Are we mmapped */
+ rwlock_t dp_lock;
+ u_char** dp_memvec; /* Pointer to pages allocated for mmap */
+ int dp_numvec; /* Size of above array */
+};
+#endif
+/* Match solaris */
+
+struct dvpoll {
+ struct pollfd * dp_fds; /* Leave this ZERO for mmap */
+ int dp_nfds;
+ int dp_timeout;
+};
+
+
+#define DEVPOLL_MINOR 125 /* Minor device # for /dev/poll */
+
+
+#define DP_MMAP_SIZE(x) ((x) * sizeof(struct pollfd))
+
+#define DP_ALLOC _IOR('P', 1, int)
+#define DP_POLL _IOWR('P', 2, struct dvpoll)
+#define DP_FREE _IO('P', 3)
+#define DP_ISPOLLED _IOWR('P', 4, struct pollfd)
+
+#ifdef __KERNEL__
+extern rwlock_t devpoll_lock;
+/* Function Prototypes */
+
+extern inline void
+dp_add_hint(struct poll_backmap ** map, rwlock_t *lock)
+{
+ struct poll_backmap *entry;
+ struct dp_fd *dpfd;
+ if (!map)
+ return;
+
+ read_lock(lock);
+ entry = *map;
+ while (entry) {
+ dpfd = entry->arg;
+ set_bit(DPH_DIRTY, &dpfd->flags); /* atomic */
+ entry = entry->next;
+ }
+ read_unlock(lock);
+}
+#endif /* __KERNEL__ */
+
+#endif
diff -urP linux-2.4.0/include/asm-i386/poll.h linux-2.4.0-devpoll/include/asm-i386/poll.h
--- linux-2.4.0/include/asm-i386/poll.h Thu Jan 23 11:01:28 1997
+++ linux-2.4.0-devpoll/include/asm-i386/poll.h Mon Jan 29 16:24:42 2001
@@ -15,6 +15,8 @@
#define POLLWRNORM 0x0100
#define POLLWRBAND 0x0200
#define POLLMSG 0x0400
+#define POLLREMOVE 0x1000
+#define POLLHINT 0x2000
struct pollfd {
int fd;
diff -urP linux-2.4.0/include/linux/dcache.h linux-2.4.0-devpoll/include/linux/dcache.h
--- linux-2.4.0/include/linux/dcache.h Thu Jan 4 14:50:46 2001
+++ linux-2.4.0-devpoll/include/linux/dcache.h Mon Jan 29 16:24:42 2001
@@ -6,6 +6,8 @@
#include <asm/atomic.h>
#include <linux/mount.h>
+#include <asm/system.h>
+
/*
* linux/include/linux/dcache.h
*
diff -urP linux-2.4.0/include/linux/devpoll.h linux-2.4.0-devpoll/include/linux/devpoll.h
--- linux-2.4.0/include/linux/devpoll.h Wed Dec 31 16:00:00 1969
+++ linux-2.4.0-devpoll/include/linux/devpoll.h Mon Jan 29 16:24:42 2001
@@ -0,0 +1,86 @@
+/*
+ *
+ * /dev/poll
+ * by Niels Provos <provos@citi.umich.edu>
+ *
+ * Linux 2.3/2.4 port by Michal Ostrowski
+ */
+
+#ifndef _LINUX_DEVPOLL_H
+#define _LINUX_DEVPOLL_H
+
+#include <asm/bitops.h>
+#include <linux/list.h>
+#include <asm/atomic.h>
+
+#define DPH_DIRTY 0 /* entry is dirty - bit */
+#define DPH_BACKMAP 1 /* file has an fd back map - bit*/
+#ifdef __KERNEL__
+struct dp_fd {
+ struct list_head next;
+ struct pollfd pfd;
+ int flags; /* for hinting */
+};
+
+
+struct devpoll {
+ struct list_head *dp_tab;
+ int dp_entries; /* Entries in hash table */
+ int dp_max; /* statistics */
+ int dp_avg; /* more */
+ int dp_count;
+ int dp_cached;
+ int dp_calls;
+ int dp_bucket_bits;
+ int dp_bucket_mask;
+ int dp_nfds; /* Number of poll fds */
+ u_char* dp_mmap; /* vaddr of mapped region */
+ atomic_t dp_mmapped; /* Are we mmapped */
+ rwlock_t dp_lock;
+ u_char** dp_memvec; /* Pointer to pages allocated for mmap */
+ int dp_numvec; /* Size of above array */
+};
+#endif
+/* Match solaris */
+
+struct dvpoll {
+ struct pollfd * dp_fds; /* Leave this ZERO for mmap */
+ int dp_nfds;
+ int dp_timeout;
+};
+
+
+#define DEVPOLL_MINOR 125 /* Minor device # for /dev/poll */
+
+
+#define DP_MMAP_SIZE(x) ((x) * sizeof(struct pollfd))
+
+#define DP_ALLOC _IOR('P', 1, int)
+#define DP_POLL _IOWR('P', 2, struct dvpoll)
+#define DP_FREE _IO('P', 3)
+#define DP_ISPOLLED _IOWR('P', 4, struct pollfd)
+
+#ifdef __KERNEL__
+extern rwlock_t devpoll_lock;
+/* Function Prototypes */
+
+extern inline void
+dp_add_hint(struct poll_backmap ** map, rwlock_t *lock)
+{
+ struct poll_backmap *entry;
+ struct dp_fd *dpfd;
+ if (!map)
+ return;
+
+ read_lock(lock);
+ entry = *map;
+ while (entry) {
+ dpfd = entry->arg;
+ set_bit(DPH_DIRTY, &dpfd->flags); /* atomic */
+ entry = entry->next;
+ }
+ read_unlock(lock);
+}
+#endif /* __KERNEL__ */
+
+#endif
diff -urP linux-2.4.0/include/linux/fs.h linux-2.4.0-devpoll/include/linux/fs.h
--- linux-2.4.0/include/linux/fs.h Thu Jan 4 14:50:47 2001
+++ linux-2.4.0-devpoll/include/linux/fs.h Mon Jan 29 16:24:42 2001
@@ -511,6 +511,11 @@
unsigned long f_version;
+
+ /* used by /dev/poll hinting */
+ struct poll_backmap *f_backmap;
+ rwlock_t f_dplock;
+
/* needed for tty driver, and maybe others */
void *private_data;
};
diff -urP linux-2.4.0/include/linux/poll.h linux-2.4.0-devpoll/include/linux/poll.h
--- linux-2.4.0/include/linux/poll.h Thu Jan 4 14:51:10 2001
+++ linux-2.4.0-devpoll/include/linux/poll.h Mon Jan 29 18:31:12 2001
@@ -8,10 +8,18 @@
#include <linux/wait.h>
#include <linux/string.h>
#include <linux/mm.h>
+#include <linux/malloc.h>
#include <asm/uaccess.h>
struct poll_table_page;
+struct poll_backmap {
+ struct poll_backmap *next;
+ void *arg; /* pointer to devpoll */
+ struct files_struct *files; /* files which has this file as */
+ int fd; /* file descriptor number fd */
+};
+
typedef struct poll_table_struct {
int error;
struct poll_table_page * table;
@@ -83,7 +91,89 @@
memset(fdset, 0, FDS_BYTES(nr));
}
+extern inline void
+poll_backmap(int fd, void *arg, struct poll_backmap ** entry)
+{
+ struct poll_backmap *tmp;
+
+ if (!entry)
+ return;
+
+ /*
+ * See if we have an entry in the backmap already, in general
+ * we expect this linked list to be very short.
+ */
+ tmp = *entry;
+ while (tmp != NULL) {
+ if (tmp->files == current->files && tmp->fd == fd &&
+ arg==tmp->arg)
+ return;
+ tmp = tmp->next;
+ }
+
+ tmp = (struct poll_backmap *) kmalloc(sizeof(*entry), GFP_KERNEL);
+ if (tmp == NULL)
+ return;
+
+ tmp->arg = arg;
+ tmp->files = current->files;
+ tmp->fd = fd;
+ tmp->next = *entry;
+
+ *entry = tmp;
+}
+
+extern inline void poll_remove_backmap(struct poll_backmap **map, int fd,
+ struct files_struct *files)
+{
+ struct poll_backmap *tmp = *map, *old = NULL;
+
+ while (tmp != NULL) {
+ if (tmp->files == files && tmp->fd == fd){
+ struct poll_backmap *next = tmp->next;
+ if( old==NULL )
+ *map = next;
+ else
+ old->next = next;
+ kfree(tmp);
+ tmp = next;
+ }else{
+ old = tmp;
+ tmp = tmp->next;
+ }
+ }
+
+ if (!tmp)
+ return;
+
+ if (old == NULL)
+ *map = tmp->next;
+ else
+ old->next = tmp->next;
+
+ kfree (tmp);
+}
+
+extern inline void poll_clean_backmap(struct poll_backmap **map)
+{
+ struct poll_backmap *tmp = *map, *old;
+
+ printk("poll_clean_backmap: map %p\n", map);
+ printk("poll_clean_backmap: *map %p\n", *map);
+
+ while (tmp) {
+ printk("poll_clean_backmap: tmp %p\n", tmp);
+ old = tmp;
+ tmp = tmp->next;
+ kfree (old);
+ }
+
+ *map = NULL;
+}
+
extern int do_select(int n, fd_set_bits *fds, long *timeout);
+
+extern void poll_freewait(poll_table *p);
#endif /* KERNEL */
diff -urP linux-2.4.0/include/net/sock.h linux-2.4.0-devpoll/include/net/sock.h
--- linux-2.4.0/include/net/sock.h Thu Jan 4 14:51:21 2001
+++ linux-2.4.0-devpoll/include/net/sock.h Mon Jan 29 18:31:12 2001
@@ -655,6 +655,10 @@
/* Identd and reporting IO signals */
struct socket *socket;
+ /* For Poll hinting */
+ void *backmap;
+ void *dplock;
+
/* RPC layer private data */
void *user_data;
diff -urP linux-2.4.0/lib/dec_and_lock.c linux-2.4.0-devpoll/lib/dec_and_lock.c
--- linux-2.4.0/lib/dec_and_lock.c Fri Jul 7 16:22:48 2000
+++ linux-2.4.0-devpoll/lib/dec_and_lock.c Mon Jan 29 16:24:42 2001
@@ -1,5 +1,6 @@
#include <linux/spinlock.h>
#include <asm/atomic.h>
+#include <asm/system.h>
/*
* This is an architecture-neutral, but slow,
diff -urP linux-2.4.0/net/atm/common.c linux-2.4.0-devpoll/net/atm/common.c
--- linux-2.4.0/net/atm/common.c Mon Jan 1 09:54:07 2001
+++ linux-2.4.0-devpoll/net/atm/common.c Mon Jan 29 16:24:42 2001
@@ -481,7 +481,10 @@
vcc = ATM_SD(sock);
poll_wait(file,&vcc->sleep,wait);
- mask = 0;
+ sk->backmap = &file->f_backmap;
+ sk->dplock = &file->f_dplock;
+ mask = POLLHINT;
+
if (skb_peek(&vcc->recvq) || skb_peek(&vcc->listenq))
mask |= POLLIN | POLLRDNORM;
if (test_bit(ATM_VF_RELEASED,&vcc->flags) ||
diff -urP linux-2.4.0/net/core/sock.c linux-2.4.0-devpoll/net/core/sock.c
--- linux-2.4.0/net/core/sock.c Fri Dec 29 14:07:24 2000
+++ linux-2.4.0-devpoll/net/core/sock.c Mon Jan 29 16:24:42 2001
@@ -1128,6 +1128,8 @@
sk->state = TCP_CLOSE;
sk->zapped = 1;
sk->socket = sock;
+ sk->backmap = NULL;
+ sk->dplock = NULL;
if(sock)
{
diff -urP linux-2.4.0/net/ipv4/af_inet.c linux-2.4.0-devpoll/net/ipv4/af_inet.c
--- linux-2.4.0/net/ipv4/af_inet.c Fri Dec 29 14:07:24 2000
+++ linux-2.4.0-devpoll/net/ipv4/af_inet.c Mon Jan 29 16:24:42 2001
@@ -442,6 +442,7 @@
if (sk->linger && !(current->flags & PF_EXITING))
timeout = sk->lingertime;
sock->sk = NULL;
+ sk->backmap = NULL;
sk->prot->close(sk, timeout);
}
return(0);
diff -urP linux-2.4.0/net/ipv4/tcp.c linux-2.4.0-devpoll/net/ipv4/tcp.c
--- linux-2.4.0/net/ipv4/tcp.c Tue Nov 28 21:53:45 2000
+++ linux-2.4.0-devpoll/net/ipv4/tcp.c Mon Jan 29 16:24:42 2001
@@ -418,6 +418,7 @@
#include <linux/types.h>
#include <linux/fcntl.h>
#include <linux/poll.h>
+#include <linux/devpoll.h>
#include <linux/init.h>
#include <linux/smp_lock.h>
@@ -548,15 +549,17 @@
struct tcp_opt *tp = &(sk->tp_pinfo.af_tcp);
poll_wait(file, sk->sleep, wait);
+ sk->backmap = &file->f_backmap;
+ sk->dplock = &file->f_dplock;
if (sk->state == TCP_LISTEN)
- return tcp_listen_poll(sk, wait);
+ return tcp_listen_poll(sk, wait) | POLLHINT;
/* Socket is not locked. We are protected from async events
by poll logic and correct handling of state changes
made by another threads is impossible in any case.
*/
- mask = 0;
+ mask = POLLHINT;
if (sk->err)
mask = POLLERR;
@@ -622,7 +625,7 @@
if (tp->urg_data & TCP_URG_VALID)
mask |= POLLPRI;
}
- return mask;
+ return mask | POLLHINT;
}
/*
diff -urP linux-2.4.0/net/ipv4/tcp_input.c linux-2.4.0-devpoll/net/ipv4/tcp_input.c
--- linux-2.4.0/net/ipv4/tcp_input.c Fri Dec 29 14:07:24 2000
+++ linux-2.4.0-devpoll/net/ipv4/tcp_input.c Mon Jan 29 18:36:12 2001
@@ -68,6 +68,7 @@
#include <net/tcp.h>
#include <net/inet_common.h>
#include <linux/ipsec.h>
+#include <linux/devpoll.h>
/* These are on by default so the code paths get tested.
@@ -2912,8 +2913,10 @@
clear_bit(SOCK_NOSPACE, &sock->flags);
- if (sk->sleep && waitqueue_active(sk->sleep))
+ if (sk->sleep && waitqueue_active(sk->sleep)) {
+ dp_add_hint(sk->backmap, sk->dplock);
wake_up_interruptible(sk->sleep);
+ }
if (sock->fasync_list && !(sk->shutdown&SEND_SHUTDOWN))
sock_wake_async(sock, 2, POLL_OUT);
diff -urP linux-2.4.0/net/ipx/af_spx.c linux-2.4.0-devpoll/net/ipx/af_spx.c
--- linux-2.4.0/net/ipx/af_spx.c Sun Aug 6 22:20:09 2000
+++ linux-2.4.0-devpoll/net/ipx/af_spx.c Mon Jan 29 16:24:42 2001
@@ -66,7 +66,10 @@
unsigned int mask;
poll_wait(file, sk->sleep, wait);
- mask = 0;
+ sk->backmap = &file->f_backmap;
+ sk->dplock = &file->f_dplock;
+
+ mask = POLLHINT;
/* exceptional events? */
if (sk->err || !skb_queue_empty(&sk->error_queue))
diff -urP linux-2.4.0/net/irda/af_irda.c linux-2.4.0-devpoll/net/irda/af_irda.c
--- linux-2.4.0/net/irda/af_irda.c Sat Nov 11 18:11:23 2000
+++ linux-2.4.0-devpoll/net/irda/af_irda.c Mon Jan 29 16:24:42 2001
@@ -1637,7 +1637,9 @@
IRDA_DEBUG(4, __FUNCTION__ "()\n");
poll_wait(file, sk->sleep, wait);
- mask = 0;
+ sk->backmap = &file->f_backmap;
+ sk->dplock = &file->f_dplock;
+ mask = POLLHINT;
/* exceptional events? */
if (sk->err)
diff -urP linux-2.4.0/net/unix/af_unix.c linux-2.4.0-devpoll/net/unix/af_unix.c
--- linux-2.4.0/net/unix/af_unix.c Sat Nov 11 19:02:41 2000
+++ linux-2.4.0-devpoll/net/unix/af_unix.c Mon Jan 29 16:24:42 2001
@@ -107,6 +107,7 @@
#include <net/scm.h>
#include <linux/init.h>
#include <linux/poll.h>
+#include <linux/devpoll.h>
#include <linux/smp_lock.h>
#include <asm/checksum.h>
@@ -300,8 +301,10 @@
{
read_lock(&sk->callback_lock);
if (unix_writable(sk)) {
- if (sk->sleep && waitqueue_active(sk->sleep))
+ if (sk->sleep && waitqueue_active(sk->sleep)) {
+ dp_add_hint(sk->backmap, sk->dplock);
wake_up_interruptible(sk->sleep);
+ }
sk_wake_async(sk, 2, POLL_OUT);
}
read_unlock(&sk->callback_lock);
@@ -1708,7 +1711,10 @@
unsigned int mask;
poll_wait(file, sk->sleep, wait);
- mask = 0;
+
+ sk->backmap = &file->f_backmap;
+ sk->dplock = &file->f_dplock;
+ mask = POLLHINT;
/* exceptional events? */
if (sk->err)
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2001-03-08 15:18 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2001-03-08 15:22 /dev/poll patch for 2.4 Mathieu Dube
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®