X-Git-Url: http://ftp.safe.ca/?a=blobdiff_plain;f=drivers%2Finfiniband%2Fcore%2Fuverbs_main.c;h=eb36a81dd09bff2d544675de4c37c447614d3262;hb=132380a06b24704fd6c9be55c44d4ef3972cead2;hp=fbbe03d8c901260227ae4d32ed72c160c61d7ff9;hpb=bc38a6abdd5a50e007d0fcd9b9b6280132b79e62;p=safe%2Fjmp%2Flinux-2.6 diff --git a/drivers/infiniband/core/uverbs_main.c b/drivers/infiniband/core/uverbs_main.c index fbbe03d..eb36a81 100644 --- a/drivers/infiniband/core/uverbs_main.c +++ b/drivers/infiniband/core/uverbs_main.c @@ -1,6 +1,9 @@ /* * Copyright (c) 2005 Topspin Communications. All rights reserved. - * Copyright (c) 2005 Cisco Systems. All rights reserved. + * Copyright (c) 2005, 2006 Cisco Systems. All rights reserved. + * Copyright (c) 2005 Mellanox Technologies. All rights reserved. + * Copyright (c) 2005 Voltaire, Inc. All rights reserved. + * Copyright (c) 2005 PathScale, Inc. All rights reserved. * * This software is available to you under a choice of one of two * licenses. You may choose to be licensed under the terms of the GNU @@ -29,8 +32,6 @@ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. - * - * $Id: uverbs_main.c 2733 2005-06-28 19:14:34Z roland $ */ #include @@ -41,6 +42,7 @@ #include #include #include +#include #include @@ -60,37 +62,52 @@ enum { #define IB_UVERBS_BASE_DEV MKDEV(IB_UVERBS_MAJOR, IB_UVERBS_BASE_MINOR) -DECLARE_MUTEX(ib_uverbs_idr_mutex); +static struct class *uverbs_class; + +DEFINE_SPINLOCK(ib_uverbs_idr_lock); DEFINE_IDR(ib_uverbs_pd_idr); DEFINE_IDR(ib_uverbs_mr_idr); DEFINE_IDR(ib_uverbs_mw_idr); DEFINE_IDR(ib_uverbs_ah_idr); DEFINE_IDR(ib_uverbs_cq_idr); DEFINE_IDR(ib_uverbs_qp_idr); +DEFINE_IDR(ib_uverbs_srq_idr); static spinlock_t map_lock; +static struct ib_uverbs_device *dev_table[IB_UVERBS_MAX_DEVICES]; static DECLARE_BITMAP(dev_map, IB_UVERBS_MAX_DEVICES); static ssize_t (*uverbs_cmd_table[])(struct ib_uverbs_file *file, const char __user *buf, int in_len, int out_len) = { - [IB_USER_VERBS_CMD_QUERY_PARAMS] = ib_uverbs_query_params, - [IB_USER_VERBS_CMD_GET_CONTEXT] = ib_uverbs_get_context, - [IB_USER_VERBS_CMD_QUERY_DEVICE] = ib_uverbs_query_device, - [IB_USER_VERBS_CMD_QUERY_PORT] = ib_uverbs_query_port, - [IB_USER_VERBS_CMD_QUERY_GID] = ib_uverbs_query_gid, - [IB_USER_VERBS_CMD_QUERY_PKEY] = ib_uverbs_query_pkey, - [IB_USER_VERBS_CMD_ALLOC_PD] = ib_uverbs_alloc_pd, - [IB_USER_VERBS_CMD_DEALLOC_PD] = ib_uverbs_dealloc_pd, - [IB_USER_VERBS_CMD_REG_MR] = ib_uverbs_reg_mr, - [IB_USER_VERBS_CMD_DEREG_MR] = ib_uverbs_dereg_mr, - [IB_USER_VERBS_CMD_CREATE_CQ] = ib_uverbs_create_cq, - [IB_USER_VERBS_CMD_DESTROY_CQ] = ib_uverbs_destroy_cq, - [IB_USER_VERBS_CMD_CREATE_QP] = ib_uverbs_create_qp, - [IB_USER_VERBS_CMD_MODIFY_QP] = ib_uverbs_modify_qp, - [IB_USER_VERBS_CMD_DESTROY_QP] = ib_uverbs_destroy_qp, - [IB_USER_VERBS_CMD_ATTACH_MCAST] = ib_uverbs_attach_mcast, - [IB_USER_VERBS_CMD_DETACH_MCAST] = ib_uverbs_detach_mcast, + [IB_USER_VERBS_CMD_GET_CONTEXT] = ib_uverbs_get_context, + [IB_USER_VERBS_CMD_QUERY_DEVICE] = ib_uverbs_query_device, + [IB_USER_VERBS_CMD_QUERY_PORT] = ib_uverbs_query_port, + [IB_USER_VERBS_CMD_ALLOC_PD] = ib_uverbs_alloc_pd, + [IB_USER_VERBS_CMD_DEALLOC_PD] = ib_uverbs_dealloc_pd, + [IB_USER_VERBS_CMD_REG_MR] = ib_uverbs_reg_mr, + [IB_USER_VERBS_CMD_DEREG_MR] = ib_uverbs_dereg_mr, + [IB_USER_VERBS_CMD_CREATE_COMP_CHANNEL] = ib_uverbs_create_comp_channel, + [IB_USER_VERBS_CMD_CREATE_CQ] = ib_uverbs_create_cq, + [IB_USER_VERBS_CMD_RESIZE_CQ] = ib_uverbs_resize_cq, + [IB_USER_VERBS_CMD_POLL_CQ] = ib_uverbs_poll_cq, + [IB_USER_VERBS_CMD_REQ_NOTIFY_CQ] = ib_uverbs_req_notify_cq, + [IB_USER_VERBS_CMD_DESTROY_CQ] = ib_uverbs_destroy_cq, + [IB_USER_VERBS_CMD_CREATE_QP] = ib_uverbs_create_qp, + [IB_USER_VERBS_CMD_QUERY_QP] = ib_uverbs_query_qp, + [IB_USER_VERBS_CMD_MODIFY_QP] = ib_uverbs_modify_qp, + [IB_USER_VERBS_CMD_DESTROY_QP] = ib_uverbs_destroy_qp, + [IB_USER_VERBS_CMD_POST_SEND] = ib_uverbs_post_send, + [IB_USER_VERBS_CMD_POST_RECV] = ib_uverbs_post_recv, + [IB_USER_VERBS_CMD_POST_SRQ_RECV] = ib_uverbs_post_srq_recv, + [IB_USER_VERBS_CMD_CREATE_AH] = ib_uverbs_create_ah, + [IB_USER_VERBS_CMD_DESTROY_AH] = ib_uverbs_destroy_ah, + [IB_USER_VERBS_CMD_ATTACH_MCAST] = ib_uverbs_attach_mcast, + [IB_USER_VERBS_CMD_DETACH_MCAST] = ib_uverbs_detach_mcast, + [IB_USER_VERBS_CMD_CREATE_SRQ] = ib_uverbs_create_srq, + [IB_USER_VERBS_CMD_MODIFY_SRQ] = ib_uverbs_modify_srq, + [IB_USER_VERBS_CMD_QUERY_SRQ] = ib_uverbs_query_srq, + [IB_USER_VERBS_CMD_DESTROY_SRQ] = ib_uverbs_destroy_srq, }; static struct vfsmount *uverbs_event_mnt; @@ -98,60 +115,143 @@ static struct vfsmount *uverbs_event_mnt; static void ib_uverbs_add_one(struct ib_device *device); static void ib_uverbs_remove_one(struct ib_device *device); -static int ib_dealloc_ucontext(struct ib_ucontext *context) +static void ib_uverbs_release_dev(struct kref *ref) +{ + struct ib_uverbs_device *dev = + container_of(ref, struct ib_uverbs_device, ref); + + complete(&dev->comp); +} + +static void ib_uverbs_release_event_file(struct kref *ref) +{ + struct ib_uverbs_event_file *file = + container_of(ref, struct ib_uverbs_event_file, ref); + + kfree(file); +} + +void ib_uverbs_release_ucq(struct ib_uverbs_file *file, + struct ib_uverbs_event_file *ev_file, + struct ib_ucq_object *uobj) +{ + struct ib_uverbs_event *evt, *tmp; + + if (ev_file) { + spin_lock_irq(&ev_file->lock); + list_for_each_entry_safe(evt, tmp, &uobj->comp_list, obj_list) { + list_del(&evt->list); + kfree(evt); + } + spin_unlock_irq(&ev_file->lock); + + kref_put(&ev_file->ref, ib_uverbs_release_event_file); + } + + spin_lock_irq(&file->async_file->lock); + list_for_each_entry_safe(evt, tmp, &uobj->async_list, obj_list) { + list_del(&evt->list); + kfree(evt); + } + spin_unlock_irq(&file->async_file->lock); +} + +void ib_uverbs_release_uevent(struct ib_uverbs_file *file, + struct ib_uevent_object *uobj) +{ + struct ib_uverbs_event *evt, *tmp; + + spin_lock_irq(&file->async_file->lock); + list_for_each_entry_safe(evt, tmp, &uobj->event_list, obj_list) { + list_del(&evt->list); + kfree(evt); + } + spin_unlock_irq(&file->async_file->lock); +} + +static void ib_uverbs_detach_umcast(struct ib_qp *qp, + struct ib_uqp_object *uobj) +{ + struct ib_uverbs_mcast_entry *mcast, *tmp; + + list_for_each_entry_safe(mcast, tmp, &uobj->mcast_list, list) { + ib_detach_mcast(qp, &mcast->gid, mcast->lid); + list_del(&mcast->list); + kfree(mcast); + } +} + +static int ib_uverbs_cleanup_ucontext(struct ib_uverbs_file *file, + struct ib_ucontext *context) { struct ib_uobject *uobj, *tmp; if (!context) return 0; - down(&ib_uverbs_idr_mutex); + context->closing = 1; - /* XXX Free AHs */ + list_for_each_entry_safe(uobj, tmp, &context->ah_list, list) { + struct ib_ah *ah = uobj->object; + + idr_remove_uobj(&ib_uverbs_ah_idr, uobj); + ib_destroy_ah(ah); + kfree(uobj); + } list_for_each_entry_safe(uobj, tmp, &context->qp_list, list) { - struct ib_qp *qp = idr_find(&ib_uverbs_qp_idr, uobj->id); - idr_remove(&ib_uverbs_qp_idr, uobj->id); + struct ib_qp *qp = uobj->object; + struct ib_uqp_object *uqp = + container_of(uobj, struct ib_uqp_object, uevent.uobject); + + idr_remove_uobj(&ib_uverbs_qp_idr, uobj); + ib_uverbs_detach_umcast(qp, uqp); ib_destroy_qp(qp); - list_del(&uobj->list); - kfree(uobj); + ib_uverbs_release_uevent(file, &uqp->uevent); + kfree(uqp); } list_for_each_entry_safe(uobj, tmp, &context->cq_list, list) { - struct ib_cq *cq = idr_find(&ib_uverbs_cq_idr, uobj->id); - idr_remove(&ib_uverbs_cq_idr, uobj->id); + struct ib_cq *cq = uobj->object; + struct ib_uverbs_event_file *ev_file = cq->cq_context; + struct ib_ucq_object *ucq = + container_of(uobj, struct ib_ucq_object, uobject); + + idr_remove_uobj(&ib_uverbs_cq_idr, uobj); ib_destroy_cq(cq); - list_del(&uobj->list); - kfree(uobj); + ib_uverbs_release_ucq(file, ev_file, ucq); + kfree(ucq); + } + + list_for_each_entry_safe(uobj, tmp, &context->srq_list, list) { + struct ib_srq *srq = uobj->object; + struct ib_uevent_object *uevent = + container_of(uobj, struct ib_uevent_object, uobject); + + idr_remove_uobj(&ib_uverbs_srq_idr, uobj); + ib_destroy_srq(srq); + ib_uverbs_release_uevent(file, uevent); + kfree(uevent); } - /* XXX Free SRQs */ /* XXX Free MWs */ list_for_each_entry_safe(uobj, tmp, &context->mr_list, list) { - struct ib_mr *mr = idr_find(&ib_uverbs_mr_idr, uobj->id); - struct ib_umem_object *memobj; + struct ib_mr *mr = uobj->object; - idr_remove(&ib_uverbs_mr_idr, uobj->id); + idr_remove_uobj(&ib_uverbs_mr_idr, uobj); ib_dereg_mr(mr); - - memobj = container_of(uobj, struct ib_umem_object, uobject); - ib_umem_release_on_close(mr->device, &memobj->umem); - - list_del(&uobj->list); - kfree(memobj); + kfree(uobj); } list_for_each_entry_safe(uobj, tmp, &context->pd_list, list) { - struct ib_pd *pd = idr_find(&ib_uverbs_pd_idr, uobj->id); - idr_remove(&ib_uverbs_pd_idr, uobj->id); + struct ib_pd *pd = uobj->object; + + idr_remove_uobj(&ib_uverbs_pd_idr, uobj); ib_dealloc_pd(pd); - list_del(&uobj->list); kfree(uobj); } - up(&ib_uverbs_idr_mutex); - return context->device->dealloc_ucontext(context); } @@ -161,6 +261,8 @@ static void ib_uverbs_release_file(struct kref *ref) container_of(ref, struct ib_uverbs_file, ref); module_put(file->device->ib_dev->owner); + kref_put(&file->device->ref, ib_uverbs_release_dev); + kfree(file); } @@ -168,46 +270,42 @@ static ssize_t ib_uverbs_event_read(struct file *filp, char __user *buf, size_t count, loff_t *pos) { struct ib_uverbs_event_file *file = filp->private_data; - void *event; + struct ib_uverbs_event *event; int eventsz; int ret = 0; spin_lock_irq(&file->lock); - while (list_empty(&file->event_list) && file->fd >= 0) { + while (list_empty(&file->event_list)) { spin_unlock_irq(&file->lock); if (filp->f_flags & O_NONBLOCK) return -EAGAIN; if (wait_event_interruptible(file->poll_wait, - !list_empty(&file->event_list) || - file->fd < 0)) + !list_empty(&file->event_list))) return -ERESTARTSYS; spin_lock_irq(&file->lock); } - if (file->fd < 0) { - spin_unlock_irq(&file->lock); - return -ENODEV; - } + event = list_entry(file->event_list.next, struct ib_uverbs_event, list); - if (file->is_async) { - event = list_entry(file->event_list.next, - struct ib_uverbs_async_event, list); + if (file->is_async) eventsz = sizeof (struct ib_uverbs_async_event_desc); - } else { - event = list_entry(file->event_list.next, - struct ib_uverbs_comp_event, list); + else eventsz = sizeof (struct ib_uverbs_comp_event_desc); - } if (eventsz > count) { ret = -EINVAL; event = NULL; - } else + } else { list_del(file->event_list.next); + if (event->counter) { + ++(*event->counter); + list_del(&event->obj_list); + } + } spin_unlock_irq(&file->lock); @@ -232,146 +330,241 @@ static unsigned int ib_uverbs_event_poll(struct file *filp, poll_wait(filp, &file->poll_wait, wait); spin_lock_irq(&file->lock); - if (file->fd < 0) - pollflags = POLLERR; - else if (!list_empty(&file->event_list)) + if (!list_empty(&file->event_list)) pollflags = POLLIN | POLLRDNORM; spin_unlock_irq(&file->lock); return pollflags; } -static void ib_uverbs_event_release(struct ib_uverbs_event_file *file) +static int ib_uverbs_event_fasync(int fd, struct file *filp, int on) { - struct list_head *entry, *tmp; + struct ib_uverbs_event_file *file = filp->private_data; - spin_lock_irq(&file->lock); - if (file->fd != -1) { - file->fd = -1; - list_for_each_safe(entry, tmp, &file->event_list) - if (file->is_async) - kfree(list_entry(entry, struct ib_uverbs_async_event, list)); - else - kfree(list_entry(entry, struct ib_uverbs_comp_event, list)); - } - spin_unlock_irq(&file->lock); + return fasync_helper(fd, filp, on, &file->async_queue); } static int ib_uverbs_event_close(struct inode *inode, struct file *filp) { struct ib_uverbs_event_file *file = filp->private_data; + struct ib_uverbs_event *entry, *tmp; + + spin_lock_irq(&file->lock); + file->is_closed = 1; + list_for_each_entry_safe(entry, tmp, &file->event_list, list) { + if (entry->counter) + list_del(&entry->obj_list); + kfree(entry); + } + spin_unlock_irq(&file->lock); - ib_uverbs_event_release(file); - kref_put(&file->uverbs_file->ref, ib_uverbs_release_file); + if (file->is_async) { + ib_unregister_event_handler(&file->uverbs_file->event_handler); + kref_put(&file->uverbs_file->ref, ib_uverbs_release_file); + } + kref_put(&file->ref, ib_uverbs_release_event_file); return 0; } -static struct file_operations uverbs_event_fops = { - /* - * No .owner field since we artificially create event files, - * so there is no increment to the module reference count in - * the open path. All event files come from a uverbs command - * file, which already takes a module reference, so this is OK. - */ +static const struct file_operations uverbs_event_fops = { + .owner = THIS_MODULE, .read = ib_uverbs_event_read, .poll = ib_uverbs_event_poll, - .release = ib_uverbs_event_close + .release = ib_uverbs_event_close, + .fasync = ib_uverbs_event_fasync }; void ib_uverbs_comp_handler(struct ib_cq *cq, void *cq_context) { - struct ib_uverbs_file *file = cq_context; - struct ib_uverbs_comp_event *entry; - unsigned long flags; + struct ib_uverbs_event_file *file = cq_context; + struct ib_ucq_object *uobj; + struct ib_uverbs_event *entry; + unsigned long flags; + + if (!file) + return; + + spin_lock_irqsave(&file->lock, flags); + if (file->is_closed) { + spin_unlock_irqrestore(&file->lock, flags); + return; + } entry = kmalloc(sizeof *entry, GFP_ATOMIC); - if (!entry) + if (!entry) { + spin_unlock_irqrestore(&file->lock, flags); return; + } + + uobj = container_of(cq->uobject, struct ib_ucq_object, uobject); - entry->desc.cq_handle = cq->uobject->user_handle; + entry->desc.comp.cq_handle = cq->uobject->user_handle; + entry->counter = &uobj->comp_events_reported; - spin_lock_irqsave(&file->comp_file[0].lock, flags); - list_add_tail(&entry->list, &file->comp_file[0].event_list); - spin_unlock_irqrestore(&file->comp_file[0].lock, flags); + list_add_tail(&entry->list, &file->event_list); + list_add_tail(&entry->obj_list, &uobj->comp_list); + spin_unlock_irqrestore(&file->lock, flags); - wake_up_interruptible(&file->comp_file[0].poll_wait); + wake_up_interruptible(&file->poll_wait); + kill_fasync(&file->async_queue, SIGIO, POLL_IN); } static void ib_uverbs_async_handler(struct ib_uverbs_file *file, - __u64 element, __u64 event) + __u64 element, __u64 event, + struct list_head *obj_list, + u32 *counter) { - struct ib_uverbs_async_event *entry; + struct ib_uverbs_event *entry; unsigned long flags; + spin_lock_irqsave(&file->async_file->lock, flags); + if (file->async_file->is_closed) { + spin_unlock_irqrestore(&file->async_file->lock, flags); + return; + } + entry = kmalloc(sizeof *entry, GFP_ATOMIC); - if (!entry) + if (!entry) { + spin_unlock_irqrestore(&file->async_file->lock, flags); return; + } - entry->desc.element = element; - entry->desc.event_type = event; + entry->desc.async.element = element; + entry->desc.async.event_type = event; + entry->counter = counter; - spin_lock_irqsave(&file->async_file.lock, flags); - list_add_tail(&entry->list, &file->async_file.event_list); - spin_unlock_irqrestore(&file->async_file.lock, flags); + list_add_tail(&entry->list, &file->async_file->event_list); + if (obj_list) + list_add_tail(&entry->obj_list, obj_list); + spin_unlock_irqrestore(&file->async_file->lock, flags); - wake_up_interruptible(&file->async_file.poll_wait); + wake_up_interruptible(&file->async_file->poll_wait); + kill_fasync(&file->async_file->async_queue, SIGIO, POLL_IN); } void ib_uverbs_cq_event_handler(struct ib_event *event, void *context_ptr) { - ib_uverbs_async_handler(context_ptr, - event->element.cq->uobject->user_handle, - event->event); + struct ib_ucq_object *uobj = container_of(event->element.cq->uobject, + struct ib_ucq_object, uobject); + + ib_uverbs_async_handler(uobj->uverbs_file, uobj->uobject.user_handle, + event->event, &uobj->async_list, + &uobj->async_events_reported); } void ib_uverbs_qp_event_handler(struct ib_event *event, void *context_ptr) { - ib_uverbs_async_handler(context_ptr, - event->element.qp->uobject->user_handle, - event->event); + struct ib_uevent_object *uobj; + + uobj = container_of(event->element.qp->uobject, + struct ib_uevent_object, uobject); + + ib_uverbs_async_handler(context_ptr, uobj->uobject.user_handle, + event->event, &uobj->event_list, + &uobj->events_reported); +} + +void ib_uverbs_srq_event_handler(struct ib_event *event, void *context_ptr) +{ + struct ib_uevent_object *uobj; + + uobj = container_of(event->element.srq->uobject, + struct ib_uevent_object, uobject); + + ib_uverbs_async_handler(context_ptr, uobj->uobject.user_handle, + event->event, &uobj->event_list, + &uobj->events_reported); } -static void ib_uverbs_event_handler(struct ib_event_handler *handler, - struct ib_event *event) +void ib_uverbs_event_handler(struct ib_event_handler *handler, + struct ib_event *event) { struct ib_uverbs_file *file = container_of(handler, struct ib_uverbs_file, event_handler); - ib_uverbs_async_handler(file, event->element.port_num, event->event); + ib_uverbs_async_handler(file, event->element.port_num, event->event, + NULL, NULL); } -static int ib_uverbs_event_init(struct ib_uverbs_event_file *file, - struct ib_uverbs_file *uverbs_file) +struct file *ib_uverbs_alloc_event_file(struct ib_uverbs_file *uverbs_file, + int is_async, int *fd) { + struct ib_uverbs_event_file *ev_file; struct file *filp; + int ret; - spin_lock_init(&file->lock); - INIT_LIST_HEAD(&file->event_list); - init_waitqueue_head(&file->poll_wait); - file->uverbs_file = uverbs_file; - - file->fd = get_unused_fd(); - if (file->fd < 0) - return file->fd; + ev_file = kmalloc(sizeof *ev_file, GFP_KERNEL); + if (!ev_file) + return ERR_PTR(-ENOMEM); + + kref_init(&ev_file->ref); + spin_lock_init(&ev_file->lock); + INIT_LIST_HEAD(&ev_file->event_list); + init_waitqueue_head(&ev_file->poll_wait); + ev_file->uverbs_file = uverbs_file; + ev_file->async_queue = NULL; + ev_file->is_async = is_async; + ev_file->is_closed = 0; + + *fd = get_unused_fd(); + if (*fd < 0) { + ret = *fd; + goto err; + } - filp = get_empty_filp(); + /* + * fops_get() can't fail here, because we're coming from a + * system call on a uverbs file, which will already have a + * module reference. + */ + filp = alloc_file(uverbs_event_mnt, dget(uverbs_event_mnt->mnt_root), + FMODE_READ, fops_get(&uverbs_event_fops)); if (!filp) { - put_unused_fd(file->fd); - return -ENFILE; + ret = -ENFILE; + goto err_fd; } - filp->f_op = &uverbs_event_fops; - filp->f_vfsmnt = mntget(uverbs_event_mnt); - filp->f_dentry = dget(uverbs_event_mnt->mnt_root); - filp->f_mapping = filp->f_dentry->d_inode->i_mapping; - filp->f_flags = O_RDONLY; - filp->f_mode = FMODE_READ; - filp->private_data = file; + filp->private_data = ev_file; - fd_install(file->fd, filp); + return filp; - return 0; +err_fd: + put_unused_fd(*fd); + +err: + kfree(ev_file); + return ERR_PTR(ret); +} + +/* + * Look up a completion event file by FD. If lookup is successful, + * takes a ref to the event file struct that it returns; if + * unsuccessful, returns NULL. + */ +struct ib_uverbs_event_file *ib_uverbs_lookup_comp_file(int fd) +{ + struct ib_uverbs_event_file *ev_file = NULL; + struct file *filp; + + filp = fget(fd); + if (!filp) + return NULL; + + if (filp->f_op != &uverbs_event_fops) + goto out; + + ev_file = filp->private_data; + if (ev_file->is_async) { + ev_file = NULL; + goto out; + } + + kref_get(&ev_file->ref); + +out: + fput(filp); + return ev_file; } static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf, @@ -389,11 +582,13 @@ static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf, if (hdr.in_words * 4 != count) return -EINVAL; - if (hdr.command < 0 || hdr.command >= ARRAY_SIZE(uverbs_cmd_table)) + if (hdr.command < 0 || + hdr.command >= ARRAY_SIZE(uverbs_cmd_table) || + !uverbs_cmd_table[hdr.command] || + !(file->device->ib_dev->uverbs_cmd_mask & (1ull << hdr.command))) return -EINVAL; - if (!file->ucontext && - hdr.command != IB_USER_VERBS_CMD_QUERY_PARAMS && + if (!file->ucontext && hdr.command != IB_USER_VERBS_CMD_GET_CONTEXT) return -EINVAL; @@ -411,91 +606,84 @@ static int ib_uverbs_mmap(struct file *filp, struct vm_area_struct *vma) return file->device->ib_dev->mmap(file->ucontext, vma); } +/* + * ib_uverbs_open() does not need the BKL: + * + * - dev_table[] accesses are protected by map_lock, the + * ib_uverbs_device structures are properly reference counted, and + * everything else is purely local to the file being created, so + * races against other open calls are not a problem; + * - there is no ioctl method to race against; + * - the device is added to dev_table[] as the last part of module + * initialization, the open method will either immediately run + * -ENXIO, or all required initialization will be done. + */ static int ib_uverbs_open(struct inode *inode, struct file *filp) { - struct ib_uverbs_device *dev = - container_of(inode->i_cdev, struct ib_uverbs_device, dev); + struct ib_uverbs_device *dev; struct ib_uverbs_file *file; - int i = 0; int ret; - if (!try_module_get(dev->ib_dev->owner)) - return -ENODEV; - - file = kmalloc(sizeof *file + - (dev->num_comp - 1) * sizeof (struct ib_uverbs_event_file), - GFP_KERNEL); - if (!file) - return -ENOMEM; - - file->device = dev; - kref_init(&file->ref); + spin_lock(&map_lock); + dev = dev_table[iminor(inode) - IB_UVERBS_BASE_MINOR]; + if (dev) + kref_get(&dev->ref); + spin_unlock(&map_lock); - file->ucontext = NULL; + if (!dev) + return -ENXIO; - ret = ib_uverbs_event_init(&file->async_file, file); - if (ret) + if (!try_module_get(dev->ib_dev->owner)) { + ret = -ENODEV; goto err; + } - file->async_file.is_async = 1; - - kref_get(&file->ref); - - for (i = 0; i < dev->num_comp; ++i) { - ret = ib_uverbs_event_init(&file->comp_file[i], file); - if (ret) - goto err_async; - kref_get(&file->ref); - file->comp_file[i].is_async = 0; + file = kmalloc(sizeof *file, GFP_KERNEL); + if (!file) { + ret = -ENOMEM; + goto err_module; } + file->device = dev; + file->ucontext = NULL; + file->async_file = NULL; + kref_init(&file->ref); + mutex_init(&file->mutex); filp->private_data = file; - INIT_IB_EVENT_HANDLER(&file->event_handler, dev->ib_dev, - ib_uverbs_event_handler); - if (ib_register_event_handler(&file->event_handler)) - goto err_async; - return 0; -err_async: - while (i--) - ib_uverbs_event_release(&file->comp_file[i]); - - ib_uverbs_event_release(&file->async_file); +err_module: + module_put(dev->ib_dev->owner); err: - kref_put(&file->ref, ib_uverbs_release_file); - + kref_put(&dev->ref, ib_uverbs_release_dev); return ret; } static int ib_uverbs_close(struct inode *inode, struct file *filp) { struct ib_uverbs_file *file = filp->private_data; - int i; - ib_unregister_event_handler(&file->event_handler); - ib_uverbs_event_release(&file->async_file); - ib_dealloc_ucontext(file->ucontext); + ib_uverbs_cleanup_ucontext(file, file->ucontext); - for (i = 0; i < file->device->num_comp; ++i) - ib_uverbs_event_release(&file->comp_file[i]); + if (file->async_file) + kref_put(&file->async_file->ref, ib_uverbs_release_event_file); kref_put(&file->ref, ib_uverbs_release_file); return 0; } -static struct file_operations uverbs_fops = { +static const struct file_operations uverbs_fops = { .owner = THIS_MODULE, .write = ib_uverbs_write, .open = ib_uverbs_open, .release = ib_uverbs_close }; -static struct file_operations uverbs_mmap_fops = { +static const struct file_operations uverbs_mmap_fops = { .owner = THIS_MODULE, .write = ib_uverbs_write, .mmap = ib_uverbs_mmap, @@ -509,29 +697,29 @@ static struct ib_client uverbs_client = { .remove = ib_uverbs_remove_one }; -static ssize_t show_ibdev(struct class_device *class_dev, char *buf) +static ssize_t show_ibdev(struct device *device, struct device_attribute *attr, + char *buf) { - struct ib_uverbs_device *dev = - container_of(class_dev, struct ib_uverbs_device, class_dev); + struct ib_uverbs_device *dev = dev_get_drvdata(device); + + if (!dev) + return -ENODEV; return sprintf(buf, "%s\n", dev->ib_dev->name); } -static CLASS_DEVICE_ATTR(ibdev, S_IRUGO, show_ibdev, NULL); +static DEVICE_ATTR(ibdev, S_IRUGO, show_ibdev, NULL); -static void ib_uverbs_release_class_dev(struct class_device *class_dev) +static ssize_t show_dev_abi_version(struct device *device, + struct device_attribute *attr, char *buf) { - struct ib_uverbs_device *dev = - container_of(class_dev, struct ib_uverbs_device, class_dev); + struct ib_uverbs_device *dev = dev_get_drvdata(device); - cdev_del(&dev->dev); - clear_bit(dev->devnum, dev_map); - kfree(dev); -} + if (!dev) + return -ENODEV; -static struct class uverbs_class = { - .name = "infiniband_verbs", - .release = ib_uverbs_release_class_dev -}; + return sprintf(buf, "%d\n", dev->ib_dev->uverbs_abi_ver); +} +static DEVICE_ATTR(abi_version, S_IRUGO, show_dev_abi_version, NULL); static ssize_t show_abi_version(struct class *class, char *buf) { @@ -546,11 +734,12 @@ static void ib_uverbs_add_one(struct ib_device *device) if (!device->alloc_ucontext) return; - uverbs_dev = kmalloc(sizeof *uverbs_dev, GFP_KERNEL); + uverbs_dev = kzalloc(sizeof *uverbs_dev, GFP_KERNEL); if (!uverbs_dev) return; - memset(uverbs_dev, 0, sizeof *uverbs_dev); + kref_init(&uverbs_dev->ref); + init_completion(&uverbs_dev->comp); spin_lock(&map_lock); uverbs_dev->devnum = find_first_zero_bit(dev_map, IB_UVERBS_MAX_DEVICES); @@ -561,40 +750,47 @@ static void ib_uverbs_add_one(struct ib_device *device) set_bit(uverbs_dev->devnum, dev_map); spin_unlock(&map_lock); - uverbs_dev->ib_dev = device; - uverbs_dev->num_comp = 1; + uverbs_dev->ib_dev = device; + uverbs_dev->num_comp_vectors = device->num_comp_vectors; - if (device->mmap) - cdev_init(&uverbs_dev->dev, &uverbs_mmap_fops); - else - cdev_init(&uverbs_dev->dev, &uverbs_fops); - uverbs_dev->dev.owner = THIS_MODULE; - kobject_set_name(&uverbs_dev->dev.kobj, "uverbs%d", uverbs_dev->devnum); - if (cdev_add(&uverbs_dev->dev, IB_UVERBS_BASE_DEV + uverbs_dev->devnum, 1)) + uverbs_dev->cdev = cdev_alloc(); + if (!uverbs_dev->cdev) goto err; + uverbs_dev->cdev->owner = THIS_MODULE; + uverbs_dev->cdev->ops = device->mmap ? &uverbs_mmap_fops : &uverbs_fops; + kobject_set_name(&uverbs_dev->cdev->kobj, "uverbs%d", uverbs_dev->devnum); + if (cdev_add(uverbs_dev->cdev, IB_UVERBS_BASE_DEV + uverbs_dev->devnum, 1)) + goto err_cdev; - uverbs_dev->class_dev.class = &uverbs_class; - uverbs_dev->class_dev.dev = device->dma_device; - uverbs_dev->class_dev.devt = uverbs_dev->dev.dev; - snprintf(uverbs_dev->class_dev.class_id, BUS_ID_SIZE, "uverbs%d", uverbs_dev->devnum); - if (class_device_register(&uverbs_dev->class_dev)) + uverbs_dev->dev = device_create(uverbs_class, device->dma_device, + uverbs_dev->cdev->dev, uverbs_dev, + "uverbs%d", uverbs_dev->devnum); + if (IS_ERR(uverbs_dev->dev)) goto err_cdev; - if (class_device_create_file(&uverbs_dev->class_dev, &class_device_attr_ibdev)) + if (device_create_file(uverbs_dev->dev, &dev_attr_ibdev)) + goto err_class; + if (device_create_file(uverbs_dev->dev, &dev_attr_abi_version)) goto err_class; + spin_lock(&map_lock); + dev_table[uverbs_dev->devnum] = uverbs_dev; + spin_unlock(&map_lock); + ib_set_client_data(device, &uverbs_client, uverbs_dev); return; err_class: - class_device_unregister(&uverbs_dev->class_dev); + device_destroy(uverbs_class, uverbs_dev->cdev->dev); err_cdev: - cdev_del(&uverbs_dev->dev); + cdev_del(uverbs_dev->cdev); clear_bit(uverbs_dev->devnum, dev_map); err: + kref_put(&uverbs_dev->ref, ib_uverbs_release_dev); + wait_for_completion(&uverbs_dev->comp); kfree(uverbs_dev); return; } @@ -606,14 +802,27 @@ static void ib_uverbs_remove_one(struct ib_device *device) if (!uverbs_dev) return; - class_device_unregister(&uverbs_dev->class_dev); + dev_set_drvdata(uverbs_dev->dev, NULL); + device_destroy(uverbs_class, uverbs_dev->cdev->dev); + cdev_del(uverbs_dev->cdev); + + spin_lock(&map_lock); + dev_table[uverbs_dev->devnum] = NULL; + spin_unlock(&map_lock); + + clear_bit(uverbs_dev->devnum, dev_map); + + kref_put(&uverbs_dev->ref, ib_uverbs_release_dev); + wait_for_completion(&uverbs_dev->comp); + kfree(uverbs_dev); } -static struct super_block *uverbs_event_get_sb(struct file_system_type *fs_type, int flags, - const char *dev_name, void *data) +static int uverbs_event_get_sb(struct file_system_type *fs_type, int flags, + const char *dev_name, void *data, + struct vfsmount *mnt) { return get_sb_pseudo(fs_type, "infinibandevent:", NULL, - INFINIBANDEVENTFS_MAGIC); + INFINIBANDEVENTFS_MAGIC, mnt); } static struct file_system_type uverbs_event_fs = { @@ -636,13 +845,14 @@ static int __init ib_uverbs_init(void) goto out; } - ret = class_register(&uverbs_class); - if (ret) { + uverbs_class = class_create(THIS_MODULE, "infiniband_verbs"); + if (IS_ERR(uverbs_class)) { + ret = PTR_ERR(uverbs_class); printk(KERN_ERR "user_verbs: couldn't create class infiniband_verbs\n"); goto out_chrdev; } - ret = class_create_file(&uverbs_class, &class_attr_abi_version); + ret = class_create_file(uverbs_class, &class_attr_abi_version); if (ret) { printk(KERN_ERR "user_verbs: couldn't create abi_version attribute\n"); goto out_class; @@ -676,7 +886,7 @@ out_fs: unregister_filesystem(&uverbs_event_fs); out_class: - class_unregister(&uverbs_class); + class_destroy(uverbs_class); out_chrdev: unregister_chrdev_region(IB_UVERBS_BASE_DEV, IB_UVERBS_MAX_DEVICES); @@ -690,8 +900,15 @@ static void __exit ib_uverbs_cleanup(void) ib_unregister_client(&uverbs_client); mntput(uverbs_event_mnt); unregister_filesystem(&uverbs_event_fs); - class_unregister(&uverbs_class); + class_destroy(uverbs_class); unregister_chrdev_region(IB_UVERBS_BASE_DEV, IB_UVERBS_MAX_DEVICES); + idr_destroy(&ib_uverbs_pd_idr); + idr_destroy(&ib_uverbs_mr_idr); + idr_destroy(&ib_uverbs_mw_idr); + idr_destroy(&ib_uverbs_ah_idr); + idr_destroy(&ib_uverbs_cq_idr); + idr_destroy(&ib_uverbs_qp_idr); + idr_destroy(&ib_uverbs_srq_idr); } module_init(ib_uverbs_init);