2 * Copyright (c) 2010 Isilon Systems, Inc.
3 * Copyright (c) 2010 iX Systems, Inc.
4 * Copyright (c) 2010 Panasas, Inc.
5 * Copyright (c) 2013-2018 Mellanox Technologies, Ltd.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice unmodified, this list of conditions, and the following
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
18 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
19 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
20 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
21 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
22 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
23 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
24 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
25 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
26 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
27 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
30 #include <sys/cdefs.h>
31 __FBSDID("$FreeBSD$");
33 #include "opt_stack.h"
35 #include <sys/param.h>
36 #include <sys/systm.h>
37 #include <sys/malloc.h>
38 #include <sys/kernel.h>
39 #include <sys/sysctl.h>
41 #include <sys/sglist.h>
42 #include <sys/sleepqueue.h>
43 #include <sys/refcount.h>
45 #include <sys/mutex.h>
47 #include <sys/fcntl.h>
49 #include <sys/filio.h>
50 #include <sys/rwlock.h>
52 #include <sys/stack.h>
57 #include <vm/vm_object.h>
58 #include <vm/vm_page.h>
59 #include <vm/vm_pager.h>
61 #include <machine/stdarg.h>
63 #if defined(__i386__) || defined(__amd64__)
64 #include <machine/md_var.h>
67 #include <linux/kobject.h>
68 #include <linux/device.h>
69 #include <linux/slab.h>
70 #include <linux/module.h>
71 #include <linux/moduleparam.h>
72 #include <linux/cdev.h>
73 #include <linux/file.h>
74 #include <linux/sysfs.h>
77 #include <linux/vmalloc.h>
78 #include <linux/netdevice.h>
79 #include <linux/timer.h>
80 #include <linux/interrupt.h>
81 #include <linux/uaccess.h>
82 #include <linux/list.h>
83 #include <linux/kthread.h>
84 #include <linux/kernel.h>
85 #include <linux/compat.h>
86 #include <linux/poll.h>
87 #include <linux/smp.h>
89 #if defined(__i386__) || defined(__amd64__)
93 SYSCTL_NODE(_compat, OID_AUTO, linuxkpi, CTLFLAG_RW, 0, "LinuxKPI parameters");
95 MALLOC_DEFINE(M_KMALLOC, "linux", "Linux kmalloc compat");
97 #include <linux/rbtree.h>
98 /* Undo Linux compat changes. */
102 #define RB_ROOT(head) (head)->rbh_root
104 static void linux_cdev_deref(struct linux_cdev *ldev);
105 static struct vm_area_struct *linux_cdev_handle_find(void *handle);
107 struct kobject linux_class_root;
108 struct device linux_root_device;
109 struct class linux_class_misc;
110 struct list_head pci_drivers;
111 struct list_head pci_devices;
114 unsigned long linux_timer_hz_mask;
117 panic_cmp(struct rb_node *one, struct rb_node *two)
122 RB_GENERATE(linux_root, rb_node, __entry, panic_cmp);
125 kobject_set_name_vargs(struct kobject *kobj, const char *fmt, va_list args)
135 if (old && fmt == NULL)
138 /* compute length of string */
139 va_copy(tmp_va, args);
140 len = vsnprintf(&dummy, 0, fmt, tmp_va);
143 /* account for zero termination */
146 /* check for error */
150 /* allocate memory for string */
151 name = kzalloc(len, GFP_KERNEL);
154 vsnprintf(name, len, fmt, args);
157 /* free old string */
160 /* filter new string */
161 for (; *name != '\0'; name++)
168 kobject_set_name(struct kobject *kobj, const char *fmt, ...)
174 error = kobject_set_name_vargs(kobj, fmt, args);
181 kobject_add_complete(struct kobject *kobj, struct kobject *parent)
183 const struct kobj_type *t;
186 kobj->parent = parent;
187 error = sysfs_create_dir(kobj);
188 if (error == 0 && kobj->ktype && kobj->ktype->default_attrs) {
189 struct attribute **attr;
192 for (attr = t->default_attrs; *attr != NULL; attr++) {
193 error = sysfs_create_file(kobj, *attr);
198 sysfs_remove_dir(kobj);
205 kobject_add(struct kobject *kobj, struct kobject *parent, const char *fmt, ...)
211 error = kobject_set_name_vargs(kobj, fmt, args);
216 return kobject_add_complete(kobj, parent);
220 linux_kobject_release(struct kref *kref)
222 struct kobject *kobj;
225 kobj = container_of(kref, struct kobject, kref);
226 sysfs_remove_dir(kobj);
228 if (kobj->ktype && kobj->ktype->release)
229 kobj->ktype->release(kobj);
234 linux_kobject_kfree(struct kobject *kobj)
240 linux_kobject_kfree_name(struct kobject *kobj)
247 const struct kobj_type linux_kfree_type = {
248 .release = linux_kobject_kfree
252 linux_device_release(struct device *dev)
254 pr_debug("linux_device_release: %s\n", dev_name(dev));
259 linux_class_show(struct kobject *kobj, struct attribute *attr, char *buf)
261 struct class_attribute *dattr;
264 dattr = container_of(attr, struct class_attribute, attr);
267 error = dattr->show(container_of(kobj, struct class, kobj),
273 linux_class_store(struct kobject *kobj, struct attribute *attr, const char *buf,
276 struct class_attribute *dattr;
279 dattr = container_of(attr, struct class_attribute, attr);
282 error = dattr->store(container_of(kobj, struct class, kobj),
288 linux_class_release(struct kobject *kobj)
292 class = container_of(kobj, struct class, kobj);
293 if (class->class_release)
294 class->class_release(class);
297 static const struct sysfs_ops linux_class_sysfs = {
298 .show = linux_class_show,
299 .store = linux_class_store,
302 const struct kobj_type linux_class_ktype = {
303 .release = linux_class_release,
304 .sysfs_ops = &linux_class_sysfs
308 linux_dev_release(struct kobject *kobj)
312 dev = container_of(kobj, struct device, kobj);
313 /* This is the precedence defined by linux. */
316 else if (dev->class && dev->class->dev_release)
317 dev->class->dev_release(dev);
321 linux_dev_show(struct kobject *kobj, struct attribute *attr, char *buf)
323 struct device_attribute *dattr;
326 dattr = container_of(attr, struct device_attribute, attr);
329 error = dattr->show(container_of(kobj, struct device, kobj),
335 linux_dev_store(struct kobject *kobj, struct attribute *attr, const char *buf,
338 struct device_attribute *dattr;
341 dattr = container_of(attr, struct device_attribute, attr);
344 error = dattr->store(container_of(kobj, struct device, kobj),
349 static const struct sysfs_ops linux_dev_sysfs = {
350 .show = linux_dev_show,
351 .store = linux_dev_store,
354 const struct kobj_type linux_dev_ktype = {
355 .release = linux_dev_release,
356 .sysfs_ops = &linux_dev_sysfs
360 device_create(struct class *class, struct device *parent, dev_t devt,
361 void *drvdata, const char *fmt, ...)
366 dev = kzalloc(sizeof(*dev), M_WAITOK);
367 dev->parent = parent;
370 dev->driver_data = drvdata;
371 dev->release = linux_device_release;
373 kobject_set_name_vargs(&dev->kobj, fmt, args);
375 device_register(dev);
381 kobject_init_and_add(struct kobject *kobj, const struct kobj_type *ktype,
382 struct kobject *parent, const char *fmt, ...)
387 kobject_init(kobj, ktype);
389 kobj->parent = parent;
393 error = kobject_set_name_vargs(kobj, fmt, args);
397 return kobject_add_complete(kobj, parent);
401 linux_kq_lock(void *arg)
408 linux_kq_unlock(void *arg)
416 linux_kq_lock_owned(void *arg)
421 mtx_assert(&s->m, MA_OWNED);
426 linux_kq_lock_unowned(void *arg)
431 mtx_assert(&s->m, MA_NOTOWNED);
436 linux_file_kqfilter_poll(struct linux_file *, int);
439 linux_file_alloc(void)
441 struct linux_file *filp;
443 filp = kzalloc(sizeof(*filp), GFP_KERNEL);
445 /* set initial refcount */
448 /* setup fields needed by kqueue support */
449 spin_lock_init(&filp->f_kqlock);
450 knlist_init(&filp->f_selinfo.si_note, &filp->f_kqlock,
451 linux_kq_lock, linux_kq_unlock,
452 linux_kq_lock_owned, linux_kq_lock_unowned);
458 linux_file_free(struct linux_file *filp)
460 if (filp->_file == NULL) {
461 if (filp->f_shmem != NULL)
462 vm_object_deallocate(filp->f_shmem);
466 * The close method of the character device or file
467 * will free the linux_file structure:
469 _fdrop(filp->_file, curthread);
474 linux_cdev_pager_fault(vm_object_t vm_obj, vm_ooffset_t offset, int prot,
477 struct vm_area_struct *vmap;
479 vmap = linux_cdev_handle_find(vm_obj->handle);
482 MPASS(vmap->vm_private_data == vm_obj->handle);
484 if (likely(vmap->vm_ops != NULL && offset < vmap->vm_len)) {
485 vm_paddr_t paddr = IDX_TO_OFF(vmap->vm_pfn) + offset;
488 if (((*mres)->flags & PG_FICTITIOUS) != 0) {
490 * If the passed in result page is a fake
491 * page, update it with the new physical
495 vm_page_updatefake(page, paddr, vm_obj->memattr);
498 * Replace the passed in "mres" page with our
499 * own fake page and free up the all of the
502 VM_OBJECT_WUNLOCK(vm_obj);
503 page = vm_page_getfake(paddr, vm_obj->memattr);
504 VM_OBJECT_WLOCK(vm_obj);
506 vm_page_replace_checked(page, vm_obj,
507 (*mres)->pindex, *mres);
511 vm_page_unlock(*mres);
514 page->valid = VM_PAGE_BITS_ALL;
515 return (VM_PAGER_OK);
517 return (VM_PAGER_FAIL);
521 linux_cdev_pager_populate(vm_object_t vm_obj, vm_pindex_t pidx, int fault_type,
522 vm_prot_t max_prot, vm_pindex_t *first, vm_pindex_t *last)
524 struct vm_area_struct *vmap;
527 linux_set_current(curthread);
529 /* get VM area structure */
530 vmap = linux_cdev_handle_find(vm_obj->handle);
532 MPASS(vmap->vm_private_data == vm_obj->handle);
534 VM_OBJECT_WUNLOCK(vm_obj);
536 down_write(&vmap->vm_mm->mmap_sem);
537 if (unlikely(vmap->vm_ops == NULL)) {
538 err = VM_FAULT_SIGBUS;
542 /* fill out VM fault structure */
543 vmf.virtual_address = (void *)(uintptr_t)IDX_TO_OFF(pidx);
544 vmf.flags = (fault_type & VM_PROT_WRITE) ? FAULT_FLAG_WRITE : 0;
549 vmap->vm_pfn_count = 0;
550 vmap->vm_pfn_pcount = &vmap->vm_pfn_count;
551 vmap->vm_obj = vm_obj;
553 err = vmap->vm_ops->fault(vmap, &vmf);
555 while (vmap->vm_pfn_count == 0 && err == VM_FAULT_NOPAGE) {
556 kern_yield(PRI_USER);
557 err = vmap->vm_ops->fault(vmap, &vmf);
561 /* translate return code */
564 err = VM_PAGER_AGAIN;
566 case VM_FAULT_SIGBUS:
569 case VM_FAULT_NOPAGE:
571 * By contract the fault handler will return having
572 * busied all the pages itself. If pidx is already
573 * found in the object, it will simply xbusy the first
574 * page and return with vm_pfn_count set to 1.
576 *first = vmap->vm_pfn_first;
577 *last = *first + vmap->vm_pfn_count - 1;
581 err = VM_PAGER_ERROR;
584 up_write(&vmap->vm_mm->mmap_sem);
585 VM_OBJECT_WLOCK(vm_obj);
589 static struct rwlock linux_vma_lock;
590 static TAILQ_HEAD(, vm_area_struct) linux_vma_head =
591 TAILQ_HEAD_INITIALIZER(linux_vma_head);
594 linux_cdev_handle_free(struct vm_area_struct *vmap)
596 /* Drop reference on vm_file */
597 if (vmap->vm_file != NULL)
600 /* Drop reference on mm_struct */
607 linux_cdev_handle_remove(struct vm_area_struct *vmap)
609 rw_wlock(&linux_vma_lock);
610 TAILQ_REMOVE(&linux_vma_head, vmap, vm_entry);
611 rw_wunlock(&linux_vma_lock);
614 static struct vm_area_struct *
615 linux_cdev_handle_find(void *handle)
617 struct vm_area_struct *vmap;
619 rw_rlock(&linux_vma_lock);
620 TAILQ_FOREACH(vmap, &linux_vma_head, vm_entry) {
621 if (vmap->vm_private_data == handle)
624 rw_runlock(&linux_vma_lock);
629 linux_cdev_pager_ctor(void *handle, vm_ooffset_t size, vm_prot_t prot,
630 vm_ooffset_t foff, struct ucred *cred, u_short *color)
633 MPASS(linux_cdev_handle_find(handle) != NULL);
639 linux_cdev_pager_dtor(void *handle)
641 const struct vm_operations_struct *vm_ops;
642 struct vm_area_struct *vmap;
644 vmap = linux_cdev_handle_find(handle);
648 * Remove handle before calling close operation to prevent
649 * other threads from reusing the handle pointer.
651 linux_cdev_handle_remove(vmap);
653 down_write(&vmap->vm_mm->mmap_sem);
654 vm_ops = vmap->vm_ops;
655 if (likely(vm_ops != NULL))
657 up_write(&vmap->vm_mm->mmap_sem);
659 linux_cdev_handle_free(vmap);
662 static struct cdev_pager_ops linux_cdev_pager_ops[2] = {
665 .cdev_pg_populate = linux_cdev_pager_populate,
666 .cdev_pg_ctor = linux_cdev_pager_ctor,
667 .cdev_pg_dtor = linux_cdev_pager_dtor
671 .cdev_pg_fault = linux_cdev_pager_fault,
672 .cdev_pg_ctor = linux_cdev_pager_ctor,
673 .cdev_pg_dtor = linux_cdev_pager_dtor
678 zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
685 if (obj == NULL || (obj->flags & OBJ_UNMANAGED) != 0)
687 VM_OBJECT_RLOCK(obj);
688 for (m = vm_page_find_least(obj, OFF_TO_IDX(address));
689 m != NULL && m->pindex < OFF_TO_IDX(address + size);
690 m = TAILQ_NEXT(m, listq))
692 VM_OBJECT_RUNLOCK(obj);
696 static struct file_operations dummy_ldev_ops = {
700 static struct linux_cdev dummy_ldev = {
701 .ops = &dummy_ldev_ops,
704 #define LDEV_SI_DTR 0x0001
705 #define LDEV_SI_REF 0x0002
708 linux_get_fop(struct linux_file *filp, const struct file_operations **fop,
709 struct linux_cdev **dev)
711 struct linux_cdev *ldev;
717 for (siref = ldev->siref;;) {
718 if ((siref & LDEV_SI_DTR) != 0) {
722 MPASS((ldev->siref & LDEV_SI_DTR) == 0);
723 } else if (atomic_fcmpset_int(&ldev->siref, &siref,
724 siref + LDEV_SI_REF)) {
733 linux_drop_fop(struct linux_cdev *ldev)
738 MPASS((ldev->siref & ~LDEV_SI_DTR) != 0);
739 atomic_subtract_int(&ldev->siref, LDEV_SI_REF);
742 #define OPW(fp,td,code) ({ \
743 struct file *__fpop; \
744 __typeof(code) __retval; \
746 __fpop = (td)->td_fpop; \
747 (td)->td_fpop = (fp); \
749 (td)->td_fpop = __fpop; \
754 linux_dev_fdopen(struct cdev *dev, int fflags, struct thread *td,
757 struct linux_cdev *ldev;
758 struct linux_file *filp;
759 const struct file_operations *fop;
764 filp = linux_file_alloc();
765 filp->f_dentry = &filp->f_dentry_store;
766 filp->f_op = ldev->ops;
767 filp->f_mode = file->f_flag;
768 filp->f_flags = file->f_flag;
769 filp->f_vnode = file->f_vnode;
771 refcount_acquire(&ldev->refs);
774 linux_set_current(td);
775 linux_get_fop(filp, &fop, &ldev);
777 if (fop->open != NULL) {
778 error = -fop->open(file->f_vnode, filp);
780 linux_drop_fop(ldev);
781 linux_cdev_deref(filp->f_cdev);
787 /* hold on to the vnode - used for fstat() */
788 vhold(filp->f_vnode);
790 /* release the file from devfs */
791 finit(file, filp->f_mode, DTYPE_DEV, filp, &linuxfileops);
792 linux_drop_fop(ldev);
796 #define LINUX_IOCTL_MIN_PTR 0x10000UL
797 #define LINUX_IOCTL_MAX_PTR (LINUX_IOCTL_MIN_PTR + IOCPARM_MAX)
800 linux_remap_address(void **uaddr, size_t len)
802 uintptr_t uaddr_val = (uintptr_t)(*uaddr);
804 if (unlikely(uaddr_val >= LINUX_IOCTL_MIN_PTR &&
805 uaddr_val < LINUX_IOCTL_MAX_PTR)) {
806 struct task_struct *pts = current;
812 /* compute data offset */
813 uaddr_val -= LINUX_IOCTL_MIN_PTR;
815 /* check that length is within bounds */
816 if ((len > IOCPARM_MAX) ||
817 (uaddr_val + len) > pts->bsd_ioctl_len) {
822 /* re-add kernel buffer address */
823 uaddr_val += (uintptr_t)pts->bsd_ioctl_data;
825 /* update address location */
826 *uaddr = (void *)uaddr_val;
833 linux_copyin(const void *uaddr, void *kaddr, size_t len)
835 if (linux_remap_address(__DECONST(void **, &uaddr), len)) {
838 memcpy(kaddr, uaddr, len);
841 return (-copyin(uaddr, kaddr, len));
845 linux_copyout(const void *kaddr, void *uaddr, size_t len)
847 if (linux_remap_address(&uaddr, len)) {
850 memcpy(uaddr, kaddr, len);
853 return (-copyout(kaddr, uaddr, len));
857 linux_clear_user(void *_uaddr, size_t _len)
859 uint8_t *uaddr = _uaddr;
862 /* make sure uaddr is aligned before going into the fast loop */
863 while (((uintptr_t)uaddr & 7) != 0 && len > 7) {
864 if (subyte(uaddr, 0))
870 /* zero 8 bytes at a time */
873 if (suword64(uaddr, 0))
876 if (suword32(uaddr, 0))
878 if (suword32(uaddr + 4, 0))
885 /* zero fill end, if any */
887 if (subyte(uaddr, 0))
896 linux_access_ok(int rw, const void *uaddr, size_t len)
901 /* get start and end address */
902 saddr = (uintptr_t)uaddr;
903 eaddr = (uintptr_t)uaddr + len;
905 /* verify addresses are valid for userspace */
906 return ((saddr == eaddr) ||
907 (eaddr > saddr && eaddr <= VM_MAXUSER_ADDRESS));
911 * This function should return either EINTR or ERESTART depending on
912 * the signal type sent to this thread:
915 linux_get_error(struct task_struct *task, int error)
917 /* check for signal type interrupt code */
918 if (error == EINTR || error == ERESTARTSYS || error == ERESTART) {
919 error = -linux_schedule_get_interrupt_value(task);
927 linux_file_ioctl_sub(struct file *fp, struct linux_file *filp,
928 const struct file_operations *fop, u_long cmd, caddr_t data,
931 struct task_struct *task = current;
935 size = IOCPARM_LEN(cmd);
936 /* refer to logic in sys_ioctl() */
939 * Setup hint for linux_copyin() and linux_copyout().
941 * Background: Linux code expects a user-space address
942 * while FreeBSD supplies a kernel-space address.
944 task->bsd_ioctl_data = data;
945 task->bsd_ioctl_len = size;
946 data = (void *)LINUX_IOCTL_MIN_PTR;
948 /* fetch user-space pointer */
949 data = *(void **)data;
951 #if defined(__amd64__)
952 if (td->td_proc->p_elf_machine == EM_386) {
953 /* try the compat IOCTL handler first */
954 if (fop->compat_ioctl != NULL) {
955 error = -OPW(fp, td, fop->compat_ioctl(filp,
961 /* fallback to the regular IOCTL handler, if any */
962 if (error == ENOTTY && fop->unlocked_ioctl != NULL) {
963 error = -OPW(fp, td, fop->unlocked_ioctl(filp,
969 if (fop->unlocked_ioctl != NULL) {
970 error = -OPW(fp, td, fop->unlocked_ioctl(filp,
977 task->bsd_ioctl_data = NULL;
978 task->bsd_ioctl_len = 0;
981 if (error == EWOULDBLOCK) {
982 /* update kqfilter status, if any */
983 linux_file_kqfilter_poll(filp,
984 LINUX_KQ_FLAG_HAS_READ | LINUX_KQ_FLAG_HAS_WRITE);
986 error = linux_get_error(task, error);
991 #define LINUX_POLL_TABLE_NORMAL ((poll_table *)1)
994 * This function atomically updates the poll wakeup state and returns
995 * the previous state at the time of update.
998 linux_poll_wakeup_state(atomic_t *v, const uint8_t *pstate)
1004 while ((old = atomic_cmpxchg(v, c, pstate[c])) != c)
1012 linux_poll_wakeup_callback(wait_queue_t *wq, unsigned int wq_state, int flags, void *key)
1014 static const uint8_t state[LINUX_FWQ_STATE_MAX] = {
1015 [LINUX_FWQ_STATE_INIT] = LINUX_FWQ_STATE_INIT, /* NOP */
1016 [LINUX_FWQ_STATE_NOT_READY] = LINUX_FWQ_STATE_NOT_READY, /* NOP */
1017 [LINUX_FWQ_STATE_QUEUED] = LINUX_FWQ_STATE_READY,
1018 [LINUX_FWQ_STATE_READY] = LINUX_FWQ_STATE_READY, /* NOP */
1020 struct linux_file *filp = container_of(wq, struct linux_file, f_wait_queue.wq);
1022 switch (linux_poll_wakeup_state(&filp->f_wait_queue.state, state)) {
1023 case LINUX_FWQ_STATE_QUEUED:
1024 linux_poll_wakeup(filp);
1032 linux_poll_wait(struct linux_file *filp, wait_queue_head_t *wqh, poll_table *p)
1034 static const uint8_t state[LINUX_FWQ_STATE_MAX] = {
1035 [LINUX_FWQ_STATE_INIT] = LINUX_FWQ_STATE_NOT_READY,
1036 [LINUX_FWQ_STATE_NOT_READY] = LINUX_FWQ_STATE_NOT_READY, /* NOP */
1037 [LINUX_FWQ_STATE_QUEUED] = LINUX_FWQ_STATE_QUEUED, /* NOP */
1038 [LINUX_FWQ_STATE_READY] = LINUX_FWQ_STATE_QUEUED,
1041 /* check if we are called inside the select system call */
1042 if (p == LINUX_POLL_TABLE_NORMAL)
1043 selrecord(curthread, &filp->f_selinfo);
1045 switch (linux_poll_wakeup_state(&filp->f_wait_queue.state, state)) {
1046 case LINUX_FWQ_STATE_INIT:
1047 /* NOTE: file handles can only belong to one wait-queue */
1048 filp->f_wait_queue.wqh = wqh;
1049 filp->f_wait_queue.wq.func = &linux_poll_wakeup_callback;
1050 add_wait_queue(wqh, &filp->f_wait_queue.wq);
1051 atomic_set(&filp->f_wait_queue.state, LINUX_FWQ_STATE_QUEUED);
1059 linux_poll_wait_dequeue(struct linux_file *filp)
1061 static const uint8_t state[LINUX_FWQ_STATE_MAX] = {
1062 [LINUX_FWQ_STATE_INIT] = LINUX_FWQ_STATE_INIT, /* NOP */
1063 [LINUX_FWQ_STATE_NOT_READY] = LINUX_FWQ_STATE_INIT,
1064 [LINUX_FWQ_STATE_QUEUED] = LINUX_FWQ_STATE_INIT,
1065 [LINUX_FWQ_STATE_READY] = LINUX_FWQ_STATE_INIT,
1068 seldrain(&filp->f_selinfo);
1070 switch (linux_poll_wakeup_state(&filp->f_wait_queue.state, state)) {
1071 case LINUX_FWQ_STATE_NOT_READY:
1072 case LINUX_FWQ_STATE_QUEUED:
1073 case LINUX_FWQ_STATE_READY:
1074 remove_wait_queue(filp->f_wait_queue.wqh, &filp->f_wait_queue.wq);
1082 linux_poll_wakeup(struct linux_file *filp)
1084 /* this function should be NULL-safe */
1088 selwakeup(&filp->f_selinfo);
1090 spin_lock(&filp->f_kqlock);
1091 filp->f_kqflags |= LINUX_KQ_FLAG_NEED_READ |
1092 LINUX_KQ_FLAG_NEED_WRITE;
1094 /* make sure the "knote" gets woken up */
1095 KNOTE_LOCKED(&filp->f_selinfo.si_note, 1);
1096 spin_unlock(&filp->f_kqlock);
1100 linux_file_kqfilter_detach(struct knote *kn)
1102 struct linux_file *filp = kn->kn_hook;
1104 spin_lock(&filp->f_kqlock);
1105 knlist_remove(&filp->f_selinfo.si_note, kn, 1);
1106 spin_unlock(&filp->f_kqlock);
1110 linux_file_kqfilter_read_event(struct knote *kn, long hint)
1112 struct linux_file *filp = kn->kn_hook;
1114 mtx_assert(&filp->f_kqlock.m, MA_OWNED);
1116 return ((filp->f_kqflags & LINUX_KQ_FLAG_NEED_READ) ? 1 : 0);
1120 linux_file_kqfilter_write_event(struct knote *kn, long hint)
1122 struct linux_file *filp = kn->kn_hook;
1124 mtx_assert(&filp->f_kqlock.m, MA_OWNED);
1126 return ((filp->f_kqflags & LINUX_KQ_FLAG_NEED_WRITE) ? 1 : 0);
1129 static struct filterops linux_dev_kqfiltops_read = {
1131 .f_detach = linux_file_kqfilter_detach,
1132 .f_event = linux_file_kqfilter_read_event,
1135 static struct filterops linux_dev_kqfiltops_write = {
1137 .f_detach = linux_file_kqfilter_detach,
1138 .f_event = linux_file_kqfilter_write_event,
1142 linux_file_kqfilter_poll(struct linux_file *filp, int kqflags)
1145 const struct file_operations *fop;
1146 struct linux_cdev *ldev;
1149 if ((filp->f_kqflags & kqflags) == 0)
1154 linux_get_fop(filp, &fop, &ldev);
1155 /* get the latest polling state */
1156 temp = OPW(filp->_file, td, fop->poll(filp, NULL));
1157 linux_drop_fop(ldev);
1159 spin_lock(&filp->f_kqlock);
1161 filp->f_kqflags &= ~(LINUX_KQ_FLAG_NEED_READ |
1162 LINUX_KQ_FLAG_NEED_WRITE);
1163 /* update kqflags */
1164 if ((temp & (POLLIN | POLLOUT)) != 0) {
1165 if ((temp & POLLIN) != 0)
1166 filp->f_kqflags |= LINUX_KQ_FLAG_NEED_READ;
1167 if ((temp & POLLOUT) != 0)
1168 filp->f_kqflags |= LINUX_KQ_FLAG_NEED_WRITE;
1170 /* make sure the "knote" gets woken up */
1171 KNOTE_LOCKED(&filp->f_selinfo.si_note, 0);
1173 spin_unlock(&filp->f_kqlock);
1177 linux_file_kqfilter(struct file *file, struct knote *kn)
1179 struct linux_file *filp;
1184 filp = (struct linux_file *)file->f_data;
1185 filp->f_flags = file->f_flag;
1186 if (filp->f_op->poll == NULL)
1189 spin_lock(&filp->f_kqlock);
1190 switch (kn->kn_filter) {
1192 filp->f_kqflags |= LINUX_KQ_FLAG_HAS_READ;
1193 kn->kn_fop = &linux_dev_kqfiltops_read;
1195 knlist_add(&filp->f_selinfo.si_note, kn, 1);
1199 filp->f_kqflags |= LINUX_KQ_FLAG_HAS_WRITE;
1200 kn->kn_fop = &linux_dev_kqfiltops_write;
1202 knlist_add(&filp->f_selinfo.si_note, kn, 1);
1209 spin_unlock(&filp->f_kqlock);
1212 linux_set_current(td);
1214 /* update kqfilter status, if any */
1215 linux_file_kqfilter_poll(filp,
1216 LINUX_KQ_FLAG_HAS_READ | LINUX_KQ_FLAG_HAS_WRITE);
1222 linux_file_mmap_single(struct file *fp, const struct file_operations *fop,
1223 vm_ooffset_t *offset, vm_size_t size, struct vm_object **object,
1224 int nprot, struct thread *td)
1226 struct task_struct *task;
1227 struct vm_area_struct *vmap;
1228 struct mm_struct *mm;
1229 struct linux_file *filp;
1233 filp = (struct linux_file *)fp->f_data;
1234 filp->f_flags = fp->f_flag;
1236 if (fop->mmap == NULL)
1237 return (EOPNOTSUPP);
1239 linux_set_current(td);
1242 * The same VM object might be shared by multiple processes
1243 * and the mm_struct is usually freed when a process exits.
1245 * The atomic reference below makes sure the mm_struct is
1246 * available as long as the vmap is in the linux_vma_head.
1250 if (atomic_inc_not_zero(&mm->mm_users) == 0)
1253 vmap = kzalloc(sizeof(*vmap), GFP_KERNEL);
1255 vmap->vm_end = size;
1256 vmap->vm_pgoff = *offset / PAGE_SIZE;
1258 vmap->vm_flags = vmap->vm_page_prot = (nprot & VM_PROT_ALL);
1259 vmap->vm_ops = NULL;
1260 vmap->vm_file = get_file(filp);
1263 if (unlikely(down_write_killable(&vmap->vm_mm->mmap_sem))) {
1264 error = linux_get_error(task, EINTR);
1266 error = -OPW(fp, td, fop->mmap(filp, vmap));
1267 error = linux_get_error(task, error);
1268 up_write(&vmap->vm_mm->mmap_sem);
1272 linux_cdev_handle_free(vmap);
1276 attr = pgprot2cachemode(vmap->vm_page_prot);
1278 if (vmap->vm_ops != NULL) {
1279 struct vm_area_struct *ptr;
1280 void *vm_private_data;
1283 if (vmap->vm_ops->open == NULL ||
1284 vmap->vm_ops->close == NULL ||
1285 vmap->vm_private_data == NULL) {
1286 /* free allocated VM area struct */
1287 linux_cdev_handle_free(vmap);
1291 vm_private_data = vmap->vm_private_data;
1293 rw_wlock(&linux_vma_lock);
1294 TAILQ_FOREACH(ptr, &linux_vma_head, vm_entry) {
1295 if (ptr->vm_private_data == vm_private_data)
1298 /* check if there is an existing VM area struct */
1300 /* check if the VM area structure is invalid */
1301 if (ptr->vm_ops == NULL ||
1302 ptr->vm_ops->open == NULL ||
1303 ptr->vm_ops->close == NULL) {
1308 vm_no_fault = (ptr->vm_ops->fault == NULL);
1311 /* insert VM area structure into list */
1312 TAILQ_INSERT_TAIL(&linux_vma_head, vmap, vm_entry);
1314 vm_no_fault = (vmap->vm_ops->fault == NULL);
1316 rw_wunlock(&linux_vma_lock);
1319 /* free allocated VM area struct */
1320 linux_cdev_handle_free(vmap);
1321 /* check for stale VM area struct */
1322 if (error != EEXIST)
1326 /* check if there is no fault handler */
1328 *object = cdev_pager_allocate(vm_private_data, OBJT_DEVICE,
1329 &linux_cdev_pager_ops[1], size, nprot, *offset,
1332 *object = cdev_pager_allocate(vm_private_data, OBJT_MGTDEVICE,
1333 &linux_cdev_pager_ops[0], size, nprot, *offset,
1337 /* check if allocating the VM object failed */
1338 if (*object == NULL) {
1340 /* remove VM area struct from list */
1341 linux_cdev_handle_remove(vmap);
1342 /* free allocated VM area struct */
1343 linux_cdev_handle_free(vmap);
1350 sg = sglist_alloc(1, M_WAITOK);
1351 sglist_append_phys(sg,
1352 (vm_paddr_t)vmap->vm_pfn << PAGE_SHIFT, vmap->vm_len);
1354 *object = vm_pager_allocate(OBJT_SG, sg, vmap->vm_len,
1355 nprot, 0, td->td_ucred);
1357 linux_cdev_handle_free(vmap);
1359 if (*object == NULL) {
1365 if (attr != VM_MEMATTR_DEFAULT) {
1366 VM_OBJECT_WLOCK(*object);
1367 vm_object_set_memattr(*object, attr);
1368 VM_OBJECT_WUNLOCK(*object);
1374 struct cdevsw linuxcdevsw = {
1375 .d_version = D_VERSION,
1376 .d_fdopen = linux_dev_fdopen,
1377 .d_name = "lkpidev",
1381 linux_file_read(struct file *file, struct uio *uio, struct ucred *active_cred,
1382 int flags, struct thread *td)
1384 struct linux_file *filp;
1385 const struct file_operations *fop;
1386 struct linux_cdev *ldev;
1391 filp = (struct linux_file *)file->f_data;
1392 filp->f_flags = file->f_flag;
1393 /* XXX no support for I/O vectors currently */
1394 if (uio->uio_iovcnt != 1)
1395 return (EOPNOTSUPP);
1396 if (uio->uio_resid > DEVFS_IOSIZE_MAX)
1398 linux_set_current(td);
1399 linux_get_fop(filp, &fop, &ldev);
1400 if (fop->read != NULL) {
1401 bytes = OPW(file, td, fop->read(filp,
1402 uio->uio_iov->iov_base,
1403 uio->uio_iov->iov_len, &uio->uio_offset));
1405 uio->uio_iov->iov_base =
1406 ((uint8_t *)uio->uio_iov->iov_base) + bytes;
1407 uio->uio_iov->iov_len -= bytes;
1408 uio->uio_resid -= bytes;
1410 error = linux_get_error(current, -bytes);
1415 /* update kqfilter status, if any */
1416 linux_file_kqfilter_poll(filp, LINUX_KQ_FLAG_HAS_READ);
1417 linux_drop_fop(ldev);
1423 linux_file_write(struct file *file, struct uio *uio, struct ucred *active_cred,
1424 int flags, struct thread *td)
1426 struct linux_file *filp;
1427 const struct file_operations *fop;
1428 struct linux_cdev *ldev;
1432 filp = (struct linux_file *)file->f_data;
1433 filp->f_flags = file->f_flag;
1434 /* XXX no support for I/O vectors currently */
1435 if (uio->uio_iovcnt != 1)
1436 return (EOPNOTSUPP);
1437 if (uio->uio_resid > DEVFS_IOSIZE_MAX)
1439 linux_set_current(td);
1440 linux_get_fop(filp, &fop, &ldev);
1441 if (fop->write != NULL) {
1442 bytes = OPW(file, td, fop->write(filp,
1443 uio->uio_iov->iov_base,
1444 uio->uio_iov->iov_len, &uio->uio_offset));
1446 uio->uio_iov->iov_base =
1447 ((uint8_t *)uio->uio_iov->iov_base) + bytes;
1448 uio->uio_iov->iov_len -= bytes;
1449 uio->uio_resid -= bytes;
1452 error = linux_get_error(current, -bytes);
1457 /* update kqfilter status, if any */
1458 linux_file_kqfilter_poll(filp, LINUX_KQ_FLAG_HAS_WRITE);
1460 linux_drop_fop(ldev);
1466 linux_file_poll(struct file *file, int events, struct ucred *active_cred,
1469 struct linux_file *filp;
1470 const struct file_operations *fop;
1471 struct linux_cdev *ldev;
1474 filp = (struct linux_file *)file->f_data;
1475 filp->f_flags = file->f_flag;
1476 linux_set_current(td);
1477 linux_get_fop(filp, &fop, &ldev);
1478 if (fop->poll != NULL) {
1479 revents = OPW(file, td, fop->poll(filp,
1480 LINUX_POLL_TABLE_NORMAL)) & events;
1484 linux_drop_fop(ldev);
1489 linux_file_close(struct file *file, struct thread *td)
1491 struct linux_file *filp;
1492 const struct file_operations *fop;
1493 struct linux_cdev *ldev;
1496 filp = (struct linux_file *)file->f_data;
1498 KASSERT(file_count(filp) == 0,
1499 ("File refcount(%d) is not zero", file_count(filp)));
1502 filp->f_flags = file->f_flag;
1503 linux_set_current(td);
1504 linux_poll_wait_dequeue(filp);
1505 linux_get_fop(filp, &fop, &ldev);
1506 if (fop->release != NULL)
1507 error = -OPW(file, td, fop->release(filp->f_vnode, filp));
1508 funsetown(&filp->f_sigio);
1509 if (filp->f_vnode != NULL)
1510 vdrop(filp->f_vnode);
1511 linux_drop_fop(ldev);
1512 if (filp->f_cdev != NULL)
1513 linux_cdev_deref(filp->f_cdev);
1520 linux_file_ioctl(struct file *fp, u_long cmd, void *data, struct ucred *cred,
1523 struct linux_file *filp;
1524 const struct file_operations *fop;
1525 struct linux_cdev *ldev;
1529 filp = (struct linux_file *)fp->f_data;
1530 filp->f_flags = fp->f_flag;
1531 linux_get_fop(filp, &fop, &ldev);
1533 linux_set_current(td);
1538 if (fop->fasync == NULL)
1540 error = -OPW(fp, td, fop->fasync(0, filp, fp->f_flag & FASYNC));
1543 error = fsetown(*(int *)data, &filp->f_sigio);
1545 if (fop->fasync == NULL)
1547 error = -OPW(fp, td, fop->fasync(0, filp,
1548 fp->f_flag & FASYNC));
1552 *(int *)data = fgetown(&filp->f_sigio);
1555 error = linux_file_ioctl_sub(fp, filp, fop, cmd, data, td);
1558 linux_drop_fop(ldev);
1563 linux_file_mmap_sub(struct thread *td, vm_size_t objsize, vm_prot_t prot,
1564 vm_prot_t *maxprotp, int *flagsp, struct file *fp,
1565 vm_ooffset_t *foff, const struct file_operations *fop, vm_object_t *objp)
1568 * Character devices do not provide private mappings
1571 if ((*maxprotp & VM_PROT_WRITE) == 0 &&
1572 (prot & VM_PROT_WRITE) != 0)
1574 if ((*flagsp & (MAP_PRIVATE | MAP_COPY)) != 0)
1577 return (linux_file_mmap_single(fp, fop, foff, objsize, objp,
1582 linux_file_mmap(struct file *fp, vm_map_t map, vm_offset_t *addr, vm_size_t size,
1583 vm_prot_t prot, vm_prot_t cap_maxprot, int flags, vm_ooffset_t foff,
1586 struct linux_file *filp;
1587 const struct file_operations *fop;
1588 struct linux_cdev *ldev;
1595 filp = (struct linux_file *)fp->f_data;
1599 return (EOPNOTSUPP);
1602 * Ensure that file and memory protections are
1606 if (mp != NULL && (mp->mnt_flag & MNT_NOEXEC) != 0) {
1607 maxprot = VM_PROT_NONE;
1608 if ((prot & VM_PROT_EXECUTE) != 0)
1611 maxprot = VM_PROT_EXECUTE;
1612 if ((fp->f_flag & FREAD) != 0)
1613 maxprot |= VM_PROT_READ;
1614 else if ((prot & VM_PROT_READ) != 0)
1618 * If we are sharing potential changes via MAP_SHARED and we
1619 * are trying to get write permission although we opened it
1620 * without asking for it, bail out.
1622 * Note that most character devices always share mappings.
1624 * Rely on linux_file_mmap_sub() to fail invalid MAP_PRIVATE
1625 * requests rather than doing it here.
1627 if ((flags & MAP_SHARED) != 0) {
1628 if ((fp->f_flag & FWRITE) != 0)
1629 maxprot |= VM_PROT_WRITE;
1630 else if ((prot & VM_PROT_WRITE) != 0)
1633 maxprot &= cap_maxprot;
1635 linux_get_fop(filp, &fop, &ldev);
1636 error = linux_file_mmap_sub(td, size, prot, &maxprot, &flags, fp,
1637 &foff, fop, &object);
1641 error = vm_mmap_object(map, addr, size, prot, maxprot, flags, object,
1644 vm_object_deallocate(object);
1646 linux_drop_fop(ldev);
1651 linux_file_stat(struct file *fp, struct stat *sb, struct ucred *active_cred,
1654 struct linux_file *filp;
1658 filp = (struct linux_file *)fp->f_data;
1659 if (filp->f_vnode == NULL)
1660 return (EOPNOTSUPP);
1664 vn_lock(vp, LK_SHARED | LK_RETRY);
1665 error = vn_stat(vp, sb, td->td_ucred, NOCRED, td);
1672 linux_file_fill_kinfo(struct file *fp, struct kinfo_file *kif,
1673 struct filedesc *fdp)
1675 struct linux_file *filp;
1683 kif->kf_type = KF_TYPE_DEV;
1686 FILEDESC_SUNLOCK(fdp);
1687 error = vn_fill_kinfo_vnode(vp, kif);
1689 kif->kf_type = KF_TYPE_VNODE;
1690 FILEDESC_SLOCK(fdp);
1696 linux_iminor(struct inode *inode)
1698 struct linux_cdev *ldev;
1700 if (inode == NULL || inode->v_rdev == NULL ||
1701 inode->v_rdev->si_devsw != &linuxcdevsw)
1703 ldev = inode->v_rdev->si_drv1;
1707 return (minor(ldev->dev));
1710 struct fileops linuxfileops = {
1711 .fo_read = linux_file_read,
1712 .fo_write = linux_file_write,
1713 .fo_truncate = invfo_truncate,
1714 .fo_kqfilter = linux_file_kqfilter,
1715 .fo_stat = linux_file_stat,
1716 .fo_fill_kinfo = linux_file_fill_kinfo,
1717 .fo_poll = linux_file_poll,
1718 .fo_close = linux_file_close,
1719 .fo_ioctl = linux_file_ioctl,
1720 .fo_mmap = linux_file_mmap,
1721 .fo_chmod = invfo_chmod,
1722 .fo_chown = invfo_chown,
1723 .fo_sendfile = invfo_sendfile,
1724 .fo_flags = DFLAG_PASSABLE,
1728 * Hash of vmmap addresses. This is infrequently accessed and does not
1729 * need to be particularly large. This is done because we must store the
1730 * caller's idea of the map size to properly unmap.
1733 LIST_ENTRY(vmmap) vm_next;
1735 unsigned long vm_size;
1739 struct vmmap *lh_first;
1741 #define VMMAP_HASH_SIZE 64
1742 #define VMMAP_HASH_MASK (VMMAP_HASH_SIZE - 1)
1743 #define VM_HASH(addr) ((uintptr_t)(addr) >> PAGE_SHIFT) & VMMAP_HASH_MASK
1744 static struct vmmaphd vmmaphead[VMMAP_HASH_SIZE];
1745 static struct mtx vmmaplock;
1748 vmmap_add(void *addr, unsigned long size)
1750 struct vmmap *vmmap;
1752 vmmap = kmalloc(sizeof(*vmmap), GFP_KERNEL);
1753 mtx_lock(&vmmaplock);
1754 vmmap->vm_size = size;
1755 vmmap->vm_addr = addr;
1756 LIST_INSERT_HEAD(&vmmaphead[VM_HASH(addr)], vmmap, vm_next);
1757 mtx_unlock(&vmmaplock);
1760 static struct vmmap *
1761 vmmap_remove(void *addr)
1763 struct vmmap *vmmap;
1765 mtx_lock(&vmmaplock);
1766 LIST_FOREACH(vmmap, &vmmaphead[VM_HASH(addr)], vm_next)
1767 if (vmmap->vm_addr == addr)
1770 LIST_REMOVE(vmmap, vm_next);
1771 mtx_unlock(&vmmaplock);
1776 #if defined(__i386__) || defined(__amd64__) || defined(__powerpc__)
1778 _ioremap_attr(vm_paddr_t phys_addr, unsigned long size, int attr)
1782 addr = pmap_mapdev_attr(phys_addr, size, attr);
1785 vmmap_add(addr, size);
1794 struct vmmap *vmmap;
1796 vmmap = vmmap_remove(addr);
1799 #if defined(__i386__) || defined(__amd64__) || defined(__powerpc__)
1800 pmap_unmapdev((vm_offset_t)addr, vmmap->vm_size);
1807 vmap(struct page **pages, unsigned int count, unsigned long flags, int prot)
1812 size = count * PAGE_SIZE;
1813 off = kva_alloc(size);
1816 vmmap_add((void *)off, size);
1817 pmap_qenter(off, pages, count);
1819 return ((void *)off);
1825 struct vmmap *vmmap;
1827 vmmap = vmmap_remove(addr);
1830 pmap_qremove((vm_offset_t)addr, vmmap->vm_size / PAGE_SIZE);
1831 kva_free((vm_offset_t)addr, vmmap->vm_size);
1836 kvasprintf(gfp_t gfp, const char *fmt, va_list ap)
1843 len = vsnprintf(NULL, 0, fmt, aq);
1846 p = kmalloc(len + 1, gfp);
1848 vsnprintf(p, len + 1, fmt, ap);
1854 kasprintf(gfp_t gfp, const char *fmt, ...)
1860 p = kvasprintf(gfp, fmt, ap);
1867 linux_timer_callback_wrapper(void *context)
1869 struct timer_list *timer;
1871 linux_set_current(curthread);
1874 timer->function(timer->data);
1878 mod_timer(struct timer_list *timer, int expires)
1881 timer->expires = expires;
1882 callout_reset(&timer->callout,
1883 linux_timer_jiffies_until(expires),
1884 &linux_timer_callback_wrapper, timer);
1888 add_timer(struct timer_list *timer)
1891 callout_reset(&timer->callout,
1892 linux_timer_jiffies_until(timer->expires),
1893 &linux_timer_callback_wrapper, timer);
1897 add_timer_on(struct timer_list *timer, int cpu)
1900 callout_reset_on(&timer->callout,
1901 linux_timer_jiffies_until(timer->expires),
1902 &linux_timer_callback_wrapper, timer, cpu);
1906 linux_timer_init(void *arg)
1910 * Compute an internal HZ value which can divide 2**32 to
1911 * avoid timer rounding problems when the tick value wraps
1914 linux_timer_hz_mask = 1;
1915 while (linux_timer_hz_mask < (unsigned long)hz)
1916 linux_timer_hz_mask *= 2;
1917 linux_timer_hz_mask--;
1919 SYSINIT(linux_timer, SI_SUB_DRIVERS, SI_ORDER_FIRST, linux_timer_init, NULL);
1922 linux_complete_common(struct completion *c, int all)
1929 wakeup_swapper = sleepq_broadcast(c, SLEEPQ_SLEEP, 0, 0);
1931 if (c->done != UINT_MAX)
1933 wakeup_swapper = sleepq_signal(c, SLEEPQ_SLEEP, 0, 0);
1941 * Indefinite wait for done != 0 with or without signals.
1944 linux_wait_for_common(struct completion *c, int flags)
1946 struct task_struct *task;
1949 if (SCHEDULER_STOPPED())
1955 flags = SLEEPQ_INTERRUPTIBLE | SLEEPQ_SLEEP;
1957 flags = SLEEPQ_SLEEP;
1963 sleepq_add(c, NULL, "completion", flags, 0);
1964 if (flags & SLEEPQ_INTERRUPTIBLE) {
1966 error = -sleepq_wait_sig(c, 0);
1969 linux_schedule_save_interrupt_value(task, error);
1970 error = -ERESTARTSYS;
1979 if (c->done != UINT_MAX)
1988 * Time limited wait for done != 0 with or without signals.
1991 linux_wait_for_timeout_common(struct completion *c, int timeout, int flags)
1993 struct task_struct *task;
1994 int end = jiffies + timeout;
1997 if (SCHEDULER_STOPPED())
2003 flags = SLEEPQ_INTERRUPTIBLE | SLEEPQ_SLEEP;
2005 flags = SLEEPQ_SLEEP;
2011 sleepq_add(c, NULL, "completion", flags, 0);
2012 sleepq_set_timeout(c, linux_timer_jiffies_until(end));
2015 if (flags & SLEEPQ_INTERRUPTIBLE)
2016 error = -sleepq_timedwait_sig(c, 0);
2018 error = -sleepq_timedwait(c, 0);
2022 /* check for timeout */
2023 if (error == -EWOULDBLOCK) {
2024 error = 0; /* timeout */
2026 /* signal happened */
2027 linux_schedule_save_interrupt_value(task, error);
2028 error = -ERESTARTSYS;
2033 if (c->done != UINT_MAX)
2037 /* return how many jiffies are left */
2038 error = linux_timer_jiffies_until(end);
2044 linux_try_wait_for_completion(struct completion *c)
2049 isdone = (c->done != 0);
2050 if (c->done != 0 && c->done != UINT_MAX)
2057 linux_completion_done(struct completion *c)
2062 isdone = (c->done != 0);
2068 linux_cdev_deref(struct linux_cdev *ldev)
2071 if (refcount_release(&ldev->refs))
2076 linux_cdev_release(struct kobject *kobj)
2078 struct linux_cdev *cdev;
2079 struct kobject *parent;
2081 cdev = container_of(kobj, struct linux_cdev, kobj);
2082 parent = kobj->parent;
2083 linux_destroy_dev(cdev);
2084 linux_cdev_deref(cdev);
2085 kobject_put(parent);
2089 linux_cdev_static_release(struct kobject *kobj)
2091 struct linux_cdev *cdev;
2092 struct kobject *parent;
2094 cdev = container_of(kobj, struct linux_cdev, kobj);
2095 parent = kobj->parent;
2096 linux_destroy_dev(cdev);
2097 kobject_put(parent);
2101 linux_destroy_dev(struct linux_cdev *ldev)
2104 if (ldev->cdev == NULL)
2107 MPASS((ldev->siref & LDEV_SI_DTR) == 0);
2108 atomic_set_int(&ldev->siref, LDEV_SI_DTR);
2109 while ((atomic_load_int(&ldev->siref) & ~LDEV_SI_DTR) != 0)
2110 pause("ldevdtr", hz / 4);
2112 destroy_dev(ldev->cdev);
2116 const struct kobj_type linux_cdev_ktype = {
2117 .release = linux_cdev_release,
2120 const struct kobj_type linux_cdev_static_ktype = {
2121 .release = linux_cdev_static_release,
2125 linux_handle_ifnet_link_event(void *arg, struct ifnet *ifp, int linkstate)
2127 struct notifier_block *nb;
2130 if (linkstate == LINK_STATE_UP)
2131 nb->notifier_call(nb, NETDEV_UP, ifp);
2133 nb->notifier_call(nb, NETDEV_DOWN, ifp);
2137 linux_handle_ifnet_arrival_event(void *arg, struct ifnet *ifp)
2139 struct notifier_block *nb;
2142 nb->notifier_call(nb, NETDEV_REGISTER, ifp);
2146 linux_handle_ifnet_departure_event(void *arg, struct ifnet *ifp)
2148 struct notifier_block *nb;
2151 nb->notifier_call(nb, NETDEV_UNREGISTER, ifp);
2155 linux_handle_iflladdr_event(void *arg, struct ifnet *ifp)
2157 struct notifier_block *nb;
2160 nb->notifier_call(nb, NETDEV_CHANGEADDR, ifp);
2164 linux_handle_ifaddr_event(void *arg, struct ifnet *ifp)
2166 struct notifier_block *nb;
2169 nb->notifier_call(nb, NETDEV_CHANGEIFADDR, ifp);
2173 register_netdevice_notifier(struct notifier_block *nb)
2176 nb->tags[NETDEV_UP] = EVENTHANDLER_REGISTER(
2177 ifnet_link_event, linux_handle_ifnet_link_event, nb, 0);
2178 nb->tags[NETDEV_REGISTER] = EVENTHANDLER_REGISTER(
2179 ifnet_arrival_event, linux_handle_ifnet_arrival_event, nb, 0);
2180 nb->tags[NETDEV_UNREGISTER] = EVENTHANDLER_REGISTER(
2181 ifnet_departure_event, linux_handle_ifnet_departure_event, nb, 0);
2182 nb->tags[NETDEV_CHANGEADDR] = EVENTHANDLER_REGISTER(
2183 iflladdr_event, linux_handle_iflladdr_event, nb, 0);
2189 register_inetaddr_notifier(struct notifier_block *nb)
2192 nb->tags[NETDEV_CHANGEIFADDR] = EVENTHANDLER_REGISTER(
2193 ifaddr_event, linux_handle_ifaddr_event, nb, 0);
2198 unregister_netdevice_notifier(struct notifier_block *nb)
2201 EVENTHANDLER_DEREGISTER(ifnet_link_event,
2202 nb->tags[NETDEV_UP]);
2203 EVENTHANDLER_DEREGISTER(ifnet_arrival_event,
2204 nb->tags[NETDEV_REGISTER]);
2205 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
2206 nb->tags[NETDEV_UNREGISTER]);
2207 EVENTHANDLER_DEREGISTER(iflladdr_event,
2208 nb->tags[NETDEV_CHANGEADDR]);
2214 unregister_inetaddr_notifier(struct notifier_block *nb)
2217 EVENTHANDLER_DEREGISTER(ifaddr_event,
2218 nb->tags[NETDEV_CHANGEIFADDR]);
2223 struct list_sort_thunk {
2224 int (*cmp)(void *, struct list_head *, struct list_head *);
2229 linux_le_cmp(void *priv, const void *d1, const void *d2)
2231 struct list_head *le1, *le2;
2232 struct list_sort_thunk *thunk;
2235 le1 = *(__DECONST(struct list_head **, d1));
2236 le2 = *(__DECONST(struct list_head **, d2));
2237 return ((thunk->cmp)(thunk->priv, le1, le2));
2241 list_sort(void *priv, struct list_head *head, int (*cmp)(void *priv,
2242 struct list_head *a, struct list_head *b))
2244 struct list_sort_thunk thunk;
2245 struct list_head **ar, *le;
2249 list_for_each(le, head)
2251 ar = malloc(sizeof(struct list_head *) * count, M_KMALLOC, M_WAITOK);
2253 list_for_each(le, head)
2257 qsort_r(ar, count, sizeof(struct list_head *), &thunk, linux_le_cmp);
2258 INIT_LIST_HEAD(head);
2259 for (i = 0; i < count; i++)
2260 list_add_tail(ar[i], head);
2261 free(ar, M_KMALLOC);
2265 linux_irq_handler(void *ent)
2267 struct irq_ent *irqe;
2269 linux_set_current(curthread);
2272 irqe->handler(irqe->irq, irqe->arg);
2275 #if defined(__i386__) || defined(__amd64__)
2277 linux_wbinvd_on_all_cpus(void)
2280 pmap_invalidate_cache();
2286 linux_on_each_cpu(void callback(void *), void *data)
2289 smp_rendezvous(smp_no_rendezvous_barrier, callback,
2290 smp_no_rendezvous_barrier, data);
2295 linux_in_atomic(void)
2298 return ((curthread->td_pflags & TDP_NOFAULTING) != 0);
2302 linux_find_cdev(const char *name, unsigned major, unsigned minor)
2304 dev_t dev = MKDEV(major, minor);
2308 LIST_FOREACH(cdev, &linuxcdevsw.d_devs, si_list) {
2309 struct linux_cdev *ldev = cdev->si_drv1;
2310 if (ldev->dev == dev &&
2311 strcmp(kobject_name(&ldev->kobj), name) == 0) {
2317 return (cdev != NULL ? cdev->si_drv1 : NULL);
2321 __register_chrdev(unsigned int major, unsigned int baseminor,
2322 unsigned int count, const char *name,
2323 const struct file_operations *fops)
2325 struct linux_cdev *cdev;
2329 for (i = baseminor; i < baseminor + count; i++) {
2330 cdev = cdev_alloc();
2332 kobject_set_name(&cdev->kobj, name);
2334 ret = cdev_add(cdev, makedev(major, i), 1);
2342 __register_chrdev_p(unsigned int major, unsigned int baseminor,
2343 unsigned int count, const char *name,
2344 const struct file_operations *fops, uid_t uid,
2345 gid_t gid, int mode)
2347 struct linux_cdev *cdev;
2351 for (i = baseminor; i < baseminor + count; i++) {
2352 cdev = cdev_alloc();
2354 kobject_set_name(&cdev->kobj, name);
2356 ret = cdev_add_ext(cdev, makedev(major, i), uid, gid, mode);
2364 __unregister_chrdev(unsigned int major, unsigned int baseminor,
2365 unsigned int count, const char *name)
2367 struct linux_cdev *cdevp;
2370 for (i = baseminor; i < baseminor + count; i++) {
2371 cdevp = linux_find_cdev(name, major, i);
2378 linux_dump_stack(void)
2389 #if defined(__i386__) || defined(__amd64__)
2390 bool linux_cpu_has_clflush;
2394 linux_compat_init(void *arg)
2396 struct sysctl_oid *rootoid;
2399 #if defined(__i386__) || defined(__amd64__)
2400 linux_cpu_has_clflush = (cpu_feature & CPUID_CLFSH);
2402 rw_init(&linux_vma_lock, "lkpi-vma-lock");
2404 rootoid = SYSCTL_ADD_ROOT_NODE(NULL,
2405 OID_AUTO, "sys", CTLFLAG_RD|CTLFLAG_MPSAFE, NULL, "sys");
2406 kobject_init(&linux_class_root, &linux_class_ktype);
2407 kobject_set_name(&linux_class_root, "class");
2408 linux_class_root.oidp = SYSCTL_ADD_NODE(NULL, SYSCTL_CHILDREN(rootoid),
2409 OID_AUTO, "class", CTLFLAG_RD|CTLFLAG_MPSAFE, NULL, "class");
2410 kobject_init(&linux_root_device.kobj, &linux_dev_ktype);
2411 kobject_set_name(&linux_root_device.kobj, "device");
2412 linux_root_device.kobj.oidp = SYSCTL_ADD_NODE(NULL,
2413 SYSCTL_CHILDREN(rootoid), OID_AUTO, "device", CTLFLAG_RD, NULL,
2415 linux_root_device.bsddev = root_bus;
2416 linux_class_misc.name = "misc";
2417 class_register(&linux_class_misc);
2418 INIT_LIST_HEAD(&pci_drivers);
2419 INIT_LIST_HEAD(&pci_devices);
2420 spin_lock_init(&pci_lock);
2421 mtx_init(&vmmaplock, "IO Map lock", NULL, MTX_DEF);
2422 for (i = 0; i < VMMAP_HASH_SIZE; i++)
2423 LIST_INIT(&vmmaphead[i]);
2425 SYSINIT(linux_compat, SI_SUB_DRIVERS, SI_ORDER_SECOND, linux_compat_init, NULL);
2428 linux_compat_uninit(void *arg)
2430 linux_kobject_kfree_name(&linux_class_root);
2431 linux_kobject_kfree_name(&linux_root_device.kobj);
2432 linux_kobject_kfree_name(&linux_class_misc.kobj);
2434 mtx_destroy(&vmmaplock);
2435 spin_lock_destroy(&pci_lock);
2436 rw_destroy(&linux_vma_lock);
2438 SYSUNINIT(linux_compat, SI_SUB_DRIVERS, SI_ORDER_SECOND, linux_compat_uninit, NULL);
2441 * NOTE: Linux frequently uses "unsigned long" for pointer to integer
2442 * conversion and vice versa, where in FreeBSD "uintptr_t" would be
2443 * used. Assert these types have the same size, else some parts of the
2444 * LinuxKPI may not work like expected:
2446 CTASSERT(sizeof(unsigned long) == sizeof(uintptr_t));