Changeset View
Changeset View
Standalone View
Standalone View
sys/ofed/drivers/infiniband/core/ib_uverbs_main.c
Show First 20 Lines • Show All 43 Lines • ▼ Show 20 Lines | |||||
#include <linux/err.h> | #include <linux/err.h> | ||||
#include <linux/fs.h> | #include <linux/fs.h> | ||||
#include <linux/poll.h> | #include <linux/poll.h> | ||||
#include <linux/sched.h> | #include <linux/sched.h> | ||||
#include <linux/file.h> | #include <linux/file.h> | ||||
#include <linux/cdev.h> | #include <linux/cdev.h> | ||||
#include <linux/slab.h> | #include <linux/slab.h> | ||||
#include <linux/pci.h> | #include <linux/pci.h> | ||||
#include <linux/lockdep.h> | |||||
#include <asm/uaccess.h> | #include <asm/uaccess.h> | ||||
#include <rdma/ib.h> | #include <rdma/ib.h> | ||||
#include <rdma/uverbs_std_types.h> | |||||
#include "uverbs.h" | #include "uverbs.h" | ||||
#include "core_priv.h" | |||||
#include "rdma_core.h" | |||||
MODULE_AUTHOR("Roland Dreier"); | MODULE_AUTHOR("Roland Dreier"); | ||||
MODULE_DESCRIPTION("InfiniBand userspace verbs access"); | MODULE_DESCRIPTION("InfiniBand userspace verbs access"); | ||||
MODULE_LICENSE("Dual BSD/GPL"); | MODULE_LICENSE("Dual BSD/GPL"); | ||||
enum { | enum { | ||||
IB_UVERBS_MAJOR = 231, | IB_UVERBS_MAJOR = 231, | ||||
IB_UVERBS_BASE_MINOR = 192, | IB_UVERBS_BASE_MINOR = 192, | ||||
IB_UVERBS_MAX_DEVICES = 32 | IB_UVERBS_MAX_DEVICES = RDMA_MAX_PORTS, | ||||
IB_UVERBS_NUM_FIXED_MINOR = 32, | |||||
IB_UVERBS_NUM_DYNAMIC_MINOR = IB_UVERBS_MAX_DEVICES - IB_UVERBS_NUM_FIXED_MINOR, | |||||
}; | }; | ||||
#define IB_UVERBS_BASE_DEV MKDEV(IB_UVERBS_MAJOR, IB_UVERBS_BASE_MINOR) | #define IB_UVERBS_BASE_DEV MKDEV(IB_UVERBS_MAJOR, IB_UVERBS_BASE_MINOR) | ||||
static dev_t dynamic_uverbs_dev; | |||||
static struct class *uverbs_class; | static struct class *uverbs_class; | ||||
DEFINE_SPINLOCK(ib_uverbs_idr_lock); | static DEFINE_IDA(uverbs_ida); | ||||
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); | |||||
DEFINE_IDR(ib_uverbs_xrcd_idr); | |||||
DEFINE_IDR(ib_uverbs_rule_idr); | |||||
DEFINE_IDR(ib_uverbs_wq_idr); | |||||
DEFINE_IDR(ib_uverbs_rwq_ind_tbl_idr); | |||||
static DEFINE_SPINLOCK(map_lock); | |||||
static DECLARE_BITMAP(dev_map, IB_UVERBS_MAX_DEVICES); | |||||
static ssize_t (*uverbs_cmd_table[])(struct ib_uverbs_file *file, | |||||
struct ib_device *ib_dev, | |||||
const char __user *buf, int in_len, | |||||
int out_len) = { | |||||
[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_REREG_MR] = ib_uverbs_rereg_mr, | |||||
[IB_USER_VERBS_CMD_DEREG_MR] = ib_uverbs_dereg_mr, | |||||
[IB_USER_VERBS_CMD_ALLOC_MW] = ib_uverbs_alloc_mw, | |||||
[IB_USER_VERBS_CMD_DEALLOC_MW] = ib_uverbs_dealloc_mw, | |||||
[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, | |||||
[IB_USER_VERBS_CMD_OPEN_XRCD] = ib_uverbs_open_xrcd, | |||||
[IB_USER_VERBS_CMD_CLOSE_XRCD] = ib_uverbs_close_xrcd, | |||||
[IB_USER_VERBS_CMD_CREATE_XSRQ] = ib_uverbs_create_xsrq, | |||||
[IB_USER_VERBS_CMD_OPEN_QP] = ib_uverbs_open_qp, | |||||
}; | |||||
static int (*uverbs_ex_cmd_table[])(struct ib_uverbs_file *file, | |||||
struct ib_device *ib_dev, | |||||
struct ib_udata *ucore, | |||||
struct ib_udata *uhw) = { | |||||
[IB_USER_VERBS_EX_CMD_CREATE_FLOW] = ib_uverbs_ex_create_flow, | |||||
[IB_USER_VERBS_EX_CMD_DESTROY_FLOW] = ib_uverbs_ex_destroy_flow, | |||||
[IB_USER_VERBS_EX_CMD_QUERY_DEVICE] = ib_uverbs_ex_query_device, | |||||
[IB_USER_VERBS_EX_CMD_CREATE_CQ] = ib_uverbs_ex_create_cq, | |||||
[IB_USER_VERBS_EX_CMD_CREATE_QP] = ib_uverbs_ex_create_qp, | |||||
[IB_USER_VERBS_EX_CMD_CREATE_WQ] = ib_uverbs_ex_create_wq, | |||||
[IB_USER_VERBS_EX_CMD_MODIFY_WQ] = ib_uverbs_ex_modify_wq, | |||||
[IB_USER_VERBS_EX_CMD_DESTROY_WQ] = ib_uverbs_ex_destroy_wq, | |||||
[IB_USER_VERBS_EX_CMD_CREATE_RWQ_IND_TBL] = ib_uverbs_ex_create_rwq_ind_table, | |||||
[IB_USER_VERBS_EX_CMD_DESTROY_RWQ_IND_TBL] = ib_uverbs_ex_destroy_rwq_ind_table, | |||||
}; | |||||
static void ib_uverbs_add_one(struct ib_device *device); | static void ib_uverbs_add_one(struct ib_device *device); | ||||
static void ib_uverbs_remove_one(struct ib_device *device, void *client_data); | static void ib_uverbs_remove_one(struct ib_device *device, void *client_data); | ||||
/* | /* | ||||
* Must be called with the ufile->device->disassociate_srcu held, and the lock | * Must be called with the ufile->device->disassociate_srcu held, and the lock | ||||
* must be held until use of the ucontext is finished. | * must be held until use of the ucontext is finished. | ||||
*/ | */ | ||||
struct ib_ucontext *ib_uverbs_get_ucontext_file(struct ib_uverbs_file *ufile) | struct ib_ucontext *ib_uverbs_get_ucontext_file(struct ib_uverbs_file *ufile) | ||||
Show All 22 Lines | int uverbs_dealloc_mw(struct ib_mw *mw) | ||||
int ret; | int ret; | ||||
ret = mw->device->dealloc_mw(mw); | ret = mw->device->dealloc_mw(mw); | ||||
if (!ret) | if (!ret) | ||||
atomic_dec(&pd->usecnt); | atomic_dec(&pd->usecnt); | ||||
return ret; | return ret; | ||||
} | } | ||||
static void ib_uverbs_release_dev(struct kobject *kobj) | static void ib_uverbs_release_dev(struct device *device) | ||||
{ | { | ||||
struct ib_uverbs_device *dev = | struct ib_uverbs_device *dev = | ||||
container_of(kobj, struct ib_uverbs_device, kobj); | container_of(device, struct ib_uverbs_device, dev); | ||||
uverbs_destroy_api(dev->uapi); | |||||
cleanup_srcu_struct(&dev->disassociate_srcu); | cleanup_srcu_struct(&dev->disassociate_srcu); | ||||
mutex_destroy(&dev->lists_mutex); | |||||
mutex_destroy(&dev->xrcd_tree_mutex); | |||||
kfree(dev); | kfree(dev); | ||||
} | } | ||||
static struct kobj_type ib_uverbs_dev_ktype = { | void ib_uverbs_release_ucq(struct ib_uverbs_completion_event_file *ev_file, | ||||
.release = ib_uverbs_release_dev, | |||||
}; | |||||
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_ucq_object *uobj) | ||||
{ | { | ||||
struct ib_uverbs_event *evt, *tmp; | struct ib_uverbs_event *evt, *tmp; | ||||
if (ev_file) { | if (ev_file) { | ||||
spin_lock_irq(&ev_file->lock); | spin_lock_irq(&ev_file->ev_queue.lock); | ||||
list_for_each_entry_safe(evt, tmp, &uobj->comp_list, obj_list) { | list_for_each_entry_safe(evt, tmp, &uobj->comp_list, obj_list) { | ||||
list_del(&evt->list); | list_del(&evt->list); | ||||
kfree(evt); | kfree(evt); | ||||
} | } | ||||
spin_unlock_irq(&ev_file->lock); | spin_unlock_irq(&ev_file->ev_queue.lock); | ||||
kref_put(&ev_file->ref, ib_uverbs_release_event_file); | uverbs_uobject_put(&ev_file->uobj); | ||||
} | } | ||||
spin_lock_irq(&file->async_file->lock); | ib_uverbs_release_uevent(&uobj->uevent); | ||||
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, | void ib_uverbs_release_uevent(struct ib_uevent_object *uobj) | ||||
struct ib_uevent_object *uobj) | |||||
{ | { | ||||
struct ib_uverbs_async_event_file *async_file = | |||||
READ_ONCE(uobj->uobject.ufile->async_file); | |||||
struct ib_uverbs_event *evt, *tmp; | struct ib_uverbs_event *evt, *tmp; | ||||
spin_lock_irq(&file->async_file->lock); | if (!async_file) | ||||
return; | |||||
spin_lock_irq(&async_file->ev_queue.lock); | |||||
list_for_each_entry_safe(evt, tmp, &uobj->event_list, obj_list) { | list_for_each_entry_safe(evt, tmp, &uobj->event_list, obj_list) { | ||||
list_del(&evt->list); | list_del(&evt->list); | ||||
kfree(evt); | kfree(evt); | ||||
} | } | ||||
spin_unlock_irq(&file->async_file->lock); | spin_unlock_irq(&async_file->ev_queue.lock); | ||||
} | } | ||||
static void ib_uverbs_detach_umcast(struct ib_qp *qp, | void ib_uverbs_detach_umcast(struct ib_qp *qp, | ||||
struct ib_uqp_object *uobj) | struct ib_uqp_object *uobj) | ||||
{ | { | ||||
struct ib_uverbs_mcast_entry *mcast, *tmp; | struct ib_uverbs_mcast_entry *mcast, *tmp; | ||||
list_for_each_entry_safe(mcast, tmp, &uobj->mcast_list, list) { | list_for_each_entry_safe(mcast, tmp, &uobj->mcast_list, list) { | ||||
ib_detach_mcast(qp, &mcast->gid, mcast->lid); | ib_detach_mcast(qp, &mcast->gid, mcast->lid); | ||||
list_del(&mcast->list); | list_del(&mcast->list); | ||||
kfree(mcast); | kfree(mcast); | ||||
} | } | ||||
} | } | ||||
static int ib_uverbs_cleanup_ucontext(struct ib_uverbs_file *file, | |||||
struct ib_ucontext *context) | |||||
{ | |||||
struct ib_uobject *uobj, *tmp; | |||||
context->closing = 1; | |||||
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); | |||||
} | |||||
/* Remove MWs before QPs, in order to support type 2A MWs. */ | |||||
list_for_each_entry_safe(uobj, tmp, &context->mw_list, list) { | |||||
struct ib_mw *mw = uobj->object; | |||||
idr_remove_uobj(&ib_uverbs_mw_idr, uobj); | |||||
uverbs_dealloc_mw(mw); | |||||
kfree(uobj); | |||||
} | |||||
list_for_each_entry_safe(uobj, tmp, &context->rule_list, list) { | |||||
struct ib_flow *flow_id = uobj->object; | |||||
idr_remove_uobj(&ib_uverbs_rule_idr, uobj); | |||||
ib_destroy_flow(flow_id); | |||||
kfree(uobj); | |||||
} | |||||
list_for_each_entry_safe(uobj, tmp, &context->qp_list, list) { | |||||
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); | |||||
if (qp == qp->real_qp) | |||||
ib_uverbs_detach_umcast(qp, uqp); | |||||
ib_destroy_qp(qp); | |||||
ib_uverbs_release_uevent(file, &uqp->uevent); | |||||
kfree(uqp); | |||||
} | |||||
list_for_each_entry_safe(uobj, tmp, &context->rwq_ind_tbl_list, list) { | |||||
struct ib_rwq_ind_table *rwq_ind_tbl = uobj->object; | |||||
struct ib_wq **ind_tbl = rwq_ind_tbl->ind_tbl; | |||||
idr_remove_uobj(&ib_uverbs_rwq_ind_tbl_idr, uobj); | |||||
ib_destroy_rwq_ind_table(rwq_ind_tbl); | |||||
kfree(ind_tbl); | |||||
kfree(uobj); | |||||
} | |||||
list_for_each_entry_safe(uobj, tmp, &context->wq_list, list) { | |||||
struct ib_wq *wq = uobj->object; | |||||
struct ib_uwq_object *uwq = | |||||
container_of(uobj, struct ib_uwq_object, uevent.uobject); | |||||
idr_remove_uobj(&ib_uverbs_wq_idr, uobj); | |||||
ib_destroy_wq(wq); | |||||
ib_uverbs_release_uevent(file, &uwq->uevent); | |||||
kfree(uwq); | |||||
} | |||||
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); | |||||
} | |||||
list_for_each_entry_safe(uobj, tmp, &context->cq_list, list) { | |||||
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); | |||||
ib_uverbs_release_ucq(file, ev_file, ucq); | |||||
kfree(ucq); | |||||
} | |||||
list_for_each_entry_safe(uobj, tmp, &context->mr_list, list) { | |||||
struct ib_mr *mr = uobj->object; | |||||
idr_remove_uobj(&ib_uverbs_mr_idr, uobj); | |||||
ib_dereg_mr(mr); | |||||
kfree(uobj); | |||||
} | |||||
mutex_lock(&file->device->xrcd_tree_mutex); | |||||
list_for_each_entry_safe(uobj, tmp, &context->xrcd_list, list) { | |||||
struct ib_xrcd *xrcd = uobj->object; | |||||
struct ib_uxrcd_object *uxrcd = | |||||
container_of(uobj, struct ib_uxrcd_object, uobject); | |||||
idr_remove_uobj(&ib_uverbs_xrcd_idr, uobj); | |||||
ib_uverbs_dealloc_xrcd(file->device, xrcd); | |||||
kfree(uxrcd); | |||||
} | |||||
mutex_unlock(&file->device->xrcd_tree_mutex); | |||||
list_for_each_entry_safe(uobj, tmp, &context->pd_list, list) { | |||||
struct ib_pd *pd = uobj->object; | |||||
idr_remove_uobj(&ib_uverbs_pd_idr, uobj); | |||||
ib_dealloc_pd(pd); | |||||
kfree(uobj); | |||||
} | |||||
put_pid(context->tgid); | |||||
return context->device->dealloc_ucontext(context); | |||||
} | |||||
static void ib_uverbs_comp_dev(struct ib_uverbs_device *dev) | static void ib_uverbs_comp_dev(struct ib_uverbs_device *dev) | ||||
{ | { | ||||
complete(&dev->comp); | complete(&dev->comp); | ||||
} | } | ||||
static void ib_uverbs_release_file(struct kref *ref) | void ib_uverbs_release_file(struct kref *ref) | ||||
{ | { | ||||
struct ib_uverbs_file *file = | struct ib_uverbs_file *file = | ||||
container_of(ref, struct ib_uverbs_file, ref); | container_of(ref, struct ib_uverbs_file, ref); | ||||
struct ib_device *ib_dev; | struct ib_device *ib_dev; | ||||
int srcu_key; | int srcu_key; | ||||
release_ufile_idr_uobject(file); | |||||
srcu_key = srcu_read_lock(&file->device->disassociate_srcu); | srcu_key = srcu_read_lock(&file->device->disassociate_srcu); | ||||
ib_dev = srcu_dereference(file->device->ib_dev, | ib_dev = srcu_dereference(file->device->ib_dev, | ||||
&file->device->disassociate_srcu); | &file->device->disassociate_srcu); | ||||
if (ib_dev && !ib_dev->disassociate_ucontext) | if (ib_dev && !ib_dev->disassociate_ucontext) | ||||
module_put(ib_dev->owner); | module_put(ib_dev->owner); | ||||
srcu_read_unlock(&file->device->disassociate_srcu, srcu_key); | srcu_read_unlock(&file->device->disassociate_srcu, srcu_key); | ||||
if (atomic_dec_and_test(&file->device->refcount)) | if (atomic_dec_and_test(&file->device->refcount)) | ||||
ib_uverbs_comp_dev(file->device); | ib_uverbs_comp_dev(file->device); | ||||
if (file->async_file) | |||||
uverbs_uobject_put(&file->async_file->uobj); | |||||
put_device(&file->device->dev); | |||||
mutex_destroy(&file->umap_lock); | |||||
mutex_destroy(&file->ucontext_lock); | |||||
kfree(file); | kfree(file); | ||||
} | } | ||||
static ssize_t ib_uverbs_event_read(struct file *filp, char __user *buf, | static ssize_t ib_uverbs_event_read(struct ib_uverbs_event_queue *ev_queue, | ||||
size_t count, loff_t *pos) | struct file *filp, char __user *buf, | ||||
size_t count, loff_t *pos, | |||||
size_t eventsz) | |||||
{ | { | ||||
struct ib_uverbs_event_file *file = filp->private_data; | |||||
struct ib_uverbs_event *event; | struct ib_uverbs_event *event; | ||||
int eventsz; | |||||
int ret = 0; | int ret = 0; | ||||
spin_lock_irq(&file->lock); | spin_lock_irq(&ev_queue->lock); | ||||
while (list_empty(&file->event_list)) { | while (list_empty(&ev_queue->event_list)) { | ||||
spin_unlock_irq(&file->lock); | spin_unlock_irq(&ev_queue->lock); | ||||
if (filp->f_flags & O_NONBLOCK) | if (filp->f_flags & O_NONBLOCK) | ||||
return -EAGAIN; | return -EAGAIN; | ||||
if (wait_event_interruptible(file->poll_wait, | if (wait_event_interruptible(ev_queue->poll_wait, | ||||
(!list_empty(&file->event_list) || | (!list_empty(&ev_queue->event_list) || | ||||
/* The barriers built into wait_event_interruptible() | ev_queue->is_closed))) | ||||
* and wake_up() guarentee this will see the null set | |||||
* without using RCU | |||||
*/ | |||||
!file->uverbs_file->device->ib_dev))) | |||||
return -ERESTARTSYS; | return -ERESTARTSYS; | ||||
spin_lock_irq(&ev_queue->lock); | |||||
/* If device was disassociated and no event exists set an error */ | /* If device was disassociated and no event exists set an error */ | ||||
if (list_empty(&file->event_list) && | if (list_empty(&ev_queue->event_list) && ev_queue->is_closed) { | ||||
!file->uverbs_file->device->ib_dev) | spin_unlock_irq(&ev_queue->lock); | ||||
return -EIO; | return -EIO; | ||||
spin_lock_irq(&file->lock); | |||||
} | } | ||||
} | |||||
event = list_entry(file->event_list.next, struct ib_uverbs_event, list); | event = list_entry(ev_queue->event_list.next, struct ib_uverbs_event, list); | ||||
if (file->is_async) | |||||
eventsz = sizeof (struct ib_uverbs_async_event_desc); | |||||
else | |||||
eventsz = sizeof (struct ib_uverbs_comp_event_desc); | |||||
if (eventsz > count) { | if (eventsz > count) { | ||||
ret = -EINVAL; | ret = -EINVAL; | ||||
event = NULL; | event = NULL; | ||||
} else { | } else { | ||||
list_del(file->event_list.next); | list_del(ev_queue->event_list.next); | ||||
if (event->counter) { | if (event->counter) { | ||||
++(*event->counter); | ++(*event->counter); | ||||
list_del(&event->obj_list); | list_del(&event->obj_list); | ||||
} | } | ||||
} | } | ||||
spin_unlock_irq(&file->lock); | spin_unlock_irq(&ev_queue->lock); | ||||
if (event) { | if (event) { | ||||
if (copy_to_user(buf, event, eventsz)) | if (copy_to_user(buf, event, eventsz)) | ||||
ret = -EFAULT; | ret = -EFAULT; | ||||
else | else | ||||
ret = eventsz; | ret = eventsz; | ||||
} | } | ||||
kfree(event); | kfree(event); | ||||
return ret; | return ret; | ||||
} | } | ||||
static unsigned int ib_uverbs_event_poll(struct file *filp, | static ssize_t ib_uverbs_async_event_read(struct file *filp, char __user *buf, | ||||
size_t count, loff_t *pos) | |||||
{ | |||||
struct ib_uverbs_async_event_file *file = filp->private_data; | |||||
return ib_uverbs_event_read(&file->ev_queue, filp, buf, count, pos, | |||||
sizeof(struct ib_uverbs_async_event_desc)); | |||||
} | |||||
static ssize_t ib_uverbs_comp_event_read(struct file *filp, char __user *buf, | |||||
size_t count, loff_t *pos) | |||||
{ | |||||
struct ib_uverbs_completion_event_file *comp_ev_file = | |||||
filp->private_data; | |||||
return ib_uverbs_event_read(&comp_ev_file->ev_queue, filp, buf, count, | |||||
pos, | |||||
sizeof(struct ib_uverbs_comp_event_desc)); | |||||
} | |||||
static __poll_t ib_uverbs_event_poll(struct ib_uverbs_event_queue *ev_queue, | |||||
struct file *filp, | |||||
struct poll_table_struct *wait) | struct poll_table_struct *wait) | ||||
{ | { | ||||
unsigned int pollflags = 0; | __poll_t pollflags = 0; | ||||
struct ib_uverbs_event_file *file = filp->private_data; | |||||
poll_wait(filp, &file->poll_wait, wait); | poll_wait(filp, &ev_queue->poll_wait, wait); | ||||
spin_lock_irq(&file->lock); | spin_lock_irq(&ev_queue->lock); | ||||
if (!list_empty(&file->event_list)) | if (!list_empty(&ev_queue->event_list)) | ||||
pollflags = POLLIN | POLLRDNORM; | pollflags = POLLIN | POLLRDNORM; | ||||
spin_unlock_irq(&file->lock); | spin_unlock_irq(&ev_queue->lock); | ||||
return pollflags; | return pollflags; | ||||
} | } | ||||
static int ib_uverbs_event_fasync(int fd, struct file *filp, int on) | static __poll_t ib_uverbs_async_event_poll(struct file *filp, | ||||
struct poll_table_struct *wait) | |||||
{ | { | ||||
struct ib_uverbs_event_file *file = filp->private_data; | struct ib_uverbs_async_event_file *file = filp->private_data; | ||||
return fasync_helper(fd, filp, on, &file->async_queue); | return ib_uverbs_event_poll(&file->ev_queue, filp, wait); | ||||
} | } | ||||
static int ib_uverbs_event_close(struct inode *inode, struct file *filp) | static __poll_t ib_uverbs_comp_event_poll(struct file *filp, | ||||
struct poll_table_struct *wait) | |||||
{ | { | ||||
struct ib_uverbs_event_file *file = filp->private_data; | struct ib_uverbs_completion_event_file *comp_ev_file = | ||||
struct ib_uverbs_event *entry, *tmp; | filp->private_data; | ||||
int closed_already = 0; | |||||
mutex_lock(&file->uverbs_file->device->lists_mutex); | return ib_uverbs_event_poll(&comp_ev_file->ev_queue, filp, wait); | ||||
spin_lock_irq(&file->lock); | |||||
closed_already = file->is_closed; | |||||
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); | |||||
if (!closed_already) { | static int ib_uverbs_async_event_fasync(int fd, struct file *filp, int on) | ||||
list_del(&file->list); | { | ||||
if (file->is_async) | struct ib_uverbs_async_event_file *file = filp->private_data; | ||||
ib_unregister_event_handler(&file->uverbs_file-> | |||||
event_handler); | return fasync_helper(fd, filp, on, &file->ev_queue.async_queue); | ||||
} | } | ||||
mutex_unlock(&file->uverbs_file->device->lists_mutex); | |||||
kref_put(&file->uverbs_file->ref, ib_uverbs_release_file); | static int ib_uverbs_comp_event_fasync(int fd, struct file *filp, int on) | ||||
kref_put(&file->ref, ib_uverbs_release_event_file); | { | ||||
struct ib_uverbs_completion_event_file *comp_ev_file = | |||||
filp->private_data; | |||||
return 0; | return fasync_helper(fd, filp, on, &comp_ev_file->ev_queue.async_queue); | ||||
} | } | ||||
static const struct file_operations uverbs_event_fops = { | const struct file_operations uverbs_event_fops = { | ||||
.owner = THIS_MODULE, | .owner = THIS_MODULE, | ||||
.read = ib_uverbs_event_read, | .read = ib_uverbs_comp_event_read, | ||||
.poll = ib_uverbs_event_poll, | .poll = ib_uverbs_comp_event_poll, | ||||
.release = ib_uverbs_event_close, | .release = uverbs_uobject_fd_release, | ||||
.fasync = ib_uverbs_event_fasync, | .fasync = ib_uverbs_comp_event_fasync, | ||||
.llseek = no_llseek, | .llseek = no_llseek, | ||||
}; | }; | ||||
const struct file_operations uverbs_async_event_fops = { | |||||
.owner = THIS_MODULE, | |||||
.read = ib_uverbs_async_event_read, | |||||
.poll = ib_uverbs_async_event_poll, | |||||
.release = uverbs_uobject_fd_release, | |||||
.fasync = ib_uverbs_async_event_fasync, | |||||
.llseek = no_llseek, | |||||
}; | |||||
void ib_uverbs_comp_handler(struct ib_cq *cq, void *cq_context) | void ib_uverbs_comp_handler(struct ib_cq *cq, void *cq_context) | ||||
{ | { | ||||
struct ib_uverbs_event_file *file = cq_context; | struct ib_uverbs_event_queue *ev_queue = cq_context; | ||||
struct ib_ucq_object *uobj; | struct ib_ucq_object *uobj; | ||||
struct ib_uverbs_event *entry; | struct ib_uverbs_event *entry; | ||||
unsigned long flags; | unsigned long flags; | ||||
if (!file) | if (!ev_queue) | ||||
return; | return; | ||||
spin_lock_irqsave(&file->lock, flags); | spin_lock_irqsave(&ev_queue->lock, flags); | ||||
if (file->is_closed) { | if (ev_queue->is_closed) { | ||||
spin_unlock_irqrestore(&file->lock, flags); | spin_unlock_irqrestore(&ev_queue->lock, flags); | ||||
return; | return; | ||||
} | } | ||||
entry = kmalloc(sizeof *entry, GFP_ATOMIC); | entry = kmalloc(sizeof(*entry), GFP_ATOMIC); | ||||
if (!entry) { | if (!entry) { | ||||
spin_unlock_irqrestore(&file->lock, flags); | spin_unlock_irqrestore(&ev_queue->lock, flags); | ||||
return; | return; | ||||
} | } | ||||
uobj = container_of(cq->uobject, struct ib_ucq_object, uobject); | uobj = cq->uobject; | ||||
entry->desc.comp.cq_handle = cq->uobject->user_handle; | entry->desc.comp.cq_handle = cq->uobject->uevent.uobject.user_handle; | ||||
entry->counter = &uobj->comp_events_reported; | entry->counter = &uobj->comp_events_reported; | ||||
list_add_tail(&entry->list, &file->event_list); | list_add_tail(&entry->list, &ev_queue->event_list); | ||||
list_add_tail(&entry->obj_list, &uobj->comp_list); | list_add_tail(&entry->obj_list, &uobj->comp_list); | ||||
spin_unlock_irqrestore(&file->lock, flags); | spin_unlock_irqrestore(&ev_queue->lock, flags); | ||||
wake_up_interruptible(&file->poll_wait); | wake_up_interruptible(&ev_queue->poll_wait); | ||||
kill_fasync(&file->async_queue, SIGIO, POLL_IN); | kill_fasync(&ev_queue->async_queue, SIGIO, POLL_IN); | ||||
} | } | ||||
static void ib_uverbs_async_handler(struct ib_uverbs_file *file, | static void | ||||
__u64 element, __u64 event, | ib_uverbs_async_handler(struct ib_uverbs_async_event_file *async_file, | ||||
struct list_head *obj_list, | __u64 element, __u64 event, struct list_head *obj_list, | ||||
u32 *counter) | u32 *counter) | ||||
{ | { | ||||
struct ib_uverbs_event *entry; | struct ib_uverbs_event *entry; | ||||
unsigned long flags; | unsigned long flags; | ||||
spin_lock_irqsave(&file->async_file->lock, flags); | if (!async_file) | ||||
if (file->async_file->is_closed) { | |||||
spin_unlock_irqrestore(&file->async_file->lock, flags); | |||||
return; | return; | ||||
spin_lock_irqsave(&async_file->ev_queue.lock, flags); | |||||
if (async_file->ev_queue.is_closed) { | |||||
spin_unlock_irqrestore(&async_file->ev_queue.lock, flags); | |||||
return; | |||||
} | } | ||||
entry = kmalloc(sizeof *entry, GFP_ATOMIC); | entry = kmalloc(sizeof(*entry), GFP_ATOMIC); | ||||
if (!entry) { | if (!entry) { | ||||
spin_unlock_irqrestore(&file->async_file->lock, flags); | spin_unlock_irqrestore(&async_file->ev_queue.lock, flags); | ||||
return; | return; | ||||
} | } | ||||
entry->desc.async.element = element; | entry->desc.async.element = element; | ||||
entry->desc.async.event_type = event; | entry->desc.async.event_type = event; | ||||
entry->desc.async.reserved = 0; | entry->desc.async.reserved = 0; | ||||
entry->counter = counter; | entry->counter = counter; | ||||
list_add_tail(&entry->list, &file->async_file->event_list); | list_add_tail(&entry->list, &async_file->ev_queue.event_list); | ||||
if (obj_list) | if (obj_list) | ||||
list_add_tail(&entry->obj_list, obj_list); | list_add_tail(&entry->obj_list, obj_list); | ||||
spin_unlock_irqrestore(&file->async_file->lock, flags); | spin_unlock_irqrestore(&async_file->ev_queue.lock, flags); | ||||
wake_up_interruptible(&file->async_file->poll_wait); | wake_up_interruptible(&async_file->ev_queue.poll_wait); | ||||
kill_fasync(&file->async_file->async_queue, SIGIO, POLL_IN); | kill_fasync(&async_file->ev_queue.async_queue, SIGIO, POLL_IN); | ||||
} | } | ||||
void ib_uverbs_cq_event_handler(struct ib_event *event, void *context_ptr) | static void uverbs_uobj_event(struct ib_uevent_object *eobj, | ||||
struct ib_event *event) | |||||
{ | { | ||||
struct ib_ucq_object *uobj = container_of(event->element.cq->uobject, | ib_uverbs_async_handler(READ_ONCE(eobj->uobject.ufile->async_file), | ||||
struct ib_ucq_object, uobject); | eobj->uobject.user_handle, event->event, | ||||
&eobj->event_list, &eobj->events_reported); | |||||
} | |||||
ib_uverbs_async_handler(uobj->uverbs_file, uobj->uobject.user_handle, | void ib_uverbs_cq_event_handler(struct ib_event *event, void *context_ptr) | ||||
event->event, &uobj->async_list, | { | ||||
&uobj->async_events_reported); | uverbs_uobj_event(&event->element.cq->uobject->uevent, event); | ||||
} | } | ||||
void ib_uverbs_qp_event_handler(struct ib_event *event, void *context_ptr) | void ib_uverbs_qp_event_handler(struct ib_event *event, void *context_ptr) | ||||
{ | { | ||||
struct ib_uevent_object *uobj; | |||||
/* for XRC target qp's, check that qp is live */ | /* for XRC target qp's, check that qp is live */ | ||||
if (!event->element.qp->uobject || !event->element.qp->uobject->live) | if (!event->element.qp->uobject) | ||||
return; | return; | ||||
uobj = container_of(event->element.qp->uobject, | uverbs_uobj_event(&event->element.qp->uobject->uevent, event); | ||||
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_wq_event_handler(struct ib_event *event, void *context_ptr) | void ib_uverbs_wq_event_handler(struct ib_event *event, void *context_ptr) | ||||
{ | { | ||||
struct ib_uevent_object *uobj = container_of(event->element.wq->uobject, | uverbs_uobj_event(&event->element.wq->uobject->uevent, event); | ||||
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) | void ib_uverbs_srq_event_handler(struct ib_event *event, void *context_ptr) | ||||
{ | { | ||||
struct ib_uevent_object *uobj; | uverbs_uobj_event(&event->element.srq->uobject->uevent, event); | ||||
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); | |||||
} | } | ||||
void ib_uverbs_event_handler(struct ib_event_handler *handler, | static void ib_uverbs_event_handler(struct ib_event_handler *handler, | ||||
struct ib_event *event) | struct ib_event *event) | ||||
{ | { | ||||
struct ib_uverbs_file *file = | ib_uverbs_async_handler( | ||||
container_of(handler, struct ib_uverbs_file, event_handler); | container_of(handler, struct ib_uverbs_async_event_file, | ||||
event_handler), | |||||
ib_uverbs_async_handler(file, event->element.port_num, event->event, | event->element.port_num, event->event, NULL, NULL); | ||||
NULL, NULL); | |||||
} | } | ||||
void ib_uverbs_free_async_event_file(struct ib_uverbs_file *file) | void ib_uverbs_init_event_queue(struct ib_uverbs_event_queue *ev_queue) | ||||
{ | { | ||||
kref_put(&file->async_file->ref, ib_uverbs_release_event_file); | spin_lock_init(&ev_queue->lock); | ||||
file->async_file = NULL; | INIT_LIST_HEAD(&ev_queue->event_list); | ||||
init_waitqueue_head(&ev_queue->poll_wait); | |||||
ev_queue->is_closed = 0; | |||||
ev_queue->async_queue = NULL; | |||||
} | } | ||||
struct file *ib_uverbs_alloc_event_file(struct ib_uverbs_file *uverbs_file, | void ib_uverbs_init_async_event_file( | ||||
struct ib_device *ib_dev, | struct ib_uverbs_async_event_file *async_file) | ||||
int is_async) | |||||
{ | { | ||||
struct ib_uverbs_event_file *ev_file; | struct ib_uverbs_file *uverbs_file = async_file->uobj.ufile; | ||||
struct file *filp; | struct ib_device *ib_dev = async_file->uobj.context->device; | ||||
int ret; | |||||
ev_file = kzalloc(sizeof(*ev_file), GFP_KERNEL); | ib_uverbs_init_event_queue(&async_file->ev_queue); | ||||
if (!ev_file) | |||||
return ERR_PTR(-ENOMEM); | |||||
kref_init(&ev_file->ref); | /* The first async_event_file becomes the default one for the file. */ | ||||
spin_lock_init(&ev_file->lock); | mutex_lock(&uverbs_file->ucontext_lock); | ||||
INIT_LIST_HEAD(&ev_file->event_list); | if (!uverbs_file->async_file) { | ||||
init_waitqueue_head(&ev_file->poll_wait); | /* Pairs with the put in ib_uverbs_release_file */ | ||||
ev_file->uverbs_file = uverbs_file; | uverbs_uobject_get(&async_file->uobj); | ||||
kref_get(&ev_file->uverbs_file->ref); | atomic_store_rel_ptr((uintptr_t *)&uverbs_file->async_file, (uintptr_t)async_file); | ||||
ev_file->async_queue = NULL; | } | ||||
ev_file->is_closed = 0; | mutex_unlock(&uverbs_file->ucontext_lock); | ||||
/* | INIT_IB_EVENT_HANDLER(&async_file->event_handler, ib_dev, | ||||
* 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(FMODE_READ, fops_get(&uverbs_event_fops)); | |||||
if (IS_ERR(filp)) | |||||
goto err_put_refs; | |||||
filp->private_data = ev_file; | |||||
mutex_lock(&uverbs_file->device->lists_mutex); | |||||
list_add_tail(&ev_file->list, | |||||
&uverbs_file->device->uverbs_events_file_list); | |||||
mutex_unlock(&uverbs_file->device->lists_mutex); | |||||
if (is_async) { | |||||
WARN_ON(uverbs_file->async_file); | |||||
uverbs_file->async_file = ev_file; | |||||
kref_get(&uverbs_file->async_file->ref); | |||||
INIT_IB_EVENT_HANDLER(&uverbs_file->event_handler, | |||||
ib_dev, | |||||
ib_uverbs_event_handler); | ib_uverbs_event_handler); | ||||
ret = ib_register_event_handler(&uverbs_file->event_handler); | ib_register_event_handler(&async_file->event_handler); | ||||
if (ret) | |||||
goto err_put_file; | |||||
/* At that point async file stuff was fully set */ | |||||
ev_file->is_async = 1; | |||||
} | } | ||||
return filp; | static ssize_t verify_hdr(struct ib_uverbs_cmd_hdr *hdr, | ||||
struct ib_uverbs_ex_cmd_hdr *ex_hdr, size_t count, | |||||
const struct uverbs_api_write_method *method_elm) | |||||
{ | |||||
if (method_elm->is_ex) { | |||||
count -= sizeof(*hdr) + sizeof(*ex_hdr); | |||||
err_put_file: | if ((hdr->in_words + ex_hdr->provider_in_words) * 8 != count) | ||||
fput(filp); | return -EINVAL; | ||||
kref_put(&uverbs_file->async_file->ref, ib_uverbs_release_event_file); | |||||
uverbs_file->async_file = NULL; | |||||
return ERR_PTR(ret); | |||||
err_put_refs: | if (hdr->in_words * 8 < method_elm->req_size) | ||||
kref_put(&ev_file->uverbs_file->ref, ib_uverbs_release_file); | return -ENOSPC; | ||||
kref_put(&ev_file->ref, ib_uverbs_release_event_file); | |||||
return filp; | |||||
} | |||||
/* | if (ex_hdr->cmd_hdr_reserved) | ||||
* Look up a completion event file by FD. If lookup is successful, | return -EINVAL; | ||||
* 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 fd f = fdget(fd); | |||||
if (!f.file) | if (ex_hdr->response) { | ||||
return NULL; | if (!hdr->out_words && !ex_hdr->provider_out_words) | ||||
return -EINVAL; | |||||
if (f.file->f_op != &uverbs_event_fops) | if (hdr->out_words * 8 < method_elm->resp_size) | ||||
goto out; | return -ENOSPC; | ||||
ev_file = f.file->private_data; | if (!access_ok(u64_to_user_ptr(ex_hdr->response), | ||||
if (ev_file->is_async) { | (hdr->out_words + ex_hdr->provider_out_words) * 8)) | ||||
ev_file = NULL; | return -EFAULT; | ||||
goto out; | } else { | ||||
if (hdr->out_words || ex_hdr->provider_out_words) | |||||
return -EINVAL; | |||||
} | } | ||||
kref_get(&ev_file->ref); | return 0; | ||||
out: | |||||
fdput(f); | |||||
return ev_file; | |||||
} | } | ||||
static int verify_command_mask(struct ib_device *ib_dev, __u32 command) | /* not extended command */ | ||||
{ | if (hdr->in_words * 4 != count) | ||||
u64 mask; | return -EINVAL; | ||||
if (command <= IB_USER_VERBS_CMD_OPEN_QP) | if (count < method_elm->req_size + sizeof(hdr)) { | ||||
mask = ib_dev->uverbs_cmd_mask; | /* | ||||
else | * rdma-core v18 and v19 have a bug where they send DESTROY_CQ | ||||
mask = ib_dev->uverbs_ex_cmd_mask; | * with a 16 byte write instead of 24. Old kernels didn't | ||||
* check the size so they allowed this. Now that the size is | |||||
if (mask & ((u64)1 << command)) | * checked provide a compatibility work around to not break | ||||
* those userspaces. | |||||
*/ | |||||
if (hdr->command == IB_USER_VERBS_CMD_DESTROY_CQ && | |||||
count == 16) { | |||||
hdr->in_words = 6; | |||||
return 0; | return 0; | ||||
} | |||||
return -ENOSPC; | |||||
} | |||||
if (hdr->out_words * 4 < method_elm->resp_size) | |||||
return -ENOSPC; | |||||
return -1; | return 0; | ||||
} | } | ||||
static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf, | static ssize_t ib_uverbs_write(struct file *filp, const char __user *buf, | ||||
size_t count, loff_t *pos) | size_t count, loff_t *pos) | ||||
{ | { | ||||
struct ib_uverbs_file *file = filp->private_data; | struct ib_uverbs_file *file = filp->private_data; | ||||
struct ib_device *ib_dev; | const struct uverbs_api_write_method *method_elm; | ||||
struct uverbs_api *uapi = file->device->uapi; | |||||
struct ib_uverbs_ex_cmd_hdr ex_hdr; | |||||
struct ib_uverbs_cmd_hdr hdr; | struct ib_uverbs_cmd_hdr hdr; | ||||
__u32 command; | struct uverbs_attr_bundle bundle; | ||||
__u32 flags; | |||||
int srcu_key; | int srcu_key; | ||||
ssize_t ret; | ssize_t ret; | ||||
if (WARN_ON_ONCE(!ib_safe_file_access(filp))) | if (!ib_safe_file_access(filp)) { | ||||
pr_warn_once("uverbs_write: process %d (%s) changed security contexts after opening file descriptor, this is not allowed.\n", | |||||
current->pid, current->comm); | |||||
return -EACCES; | return -EACCES; | ||||
} | |||||
if (count < sizeof hdr) | if (count < sizeof(hdr)) | ||||
return -EINVAL; | return -EINVAL; | ||||
if (copy_from_user(&hdr, buf, sizeof hdr)) | if (copy_from_user(&hdr, buf, sizeof(hdr))) | ||||
return -EFAULT; | return -EFAULT; | ||||
srcu_key = srcu_read_lock(&file->device->disassociate_srcu); | method_elm = uapi_get_method(uapi, hdr.command); | ||||
ib_dev = srcu_dereference(file->device->ib_dev, | if (IS_ERR(method_elm)) | ||||
&file->device->disassociate_srcu); | return PTR_ERR(method_elm); | ||||
if (!ib_dev) { | |||||
ret = -EIO; | |||||
goto out; | |||||
} | |||||
if (hdr.command & ~(__u32)(IB_USER_VERBS_CMD_FLAGS_MASK | | if (method_elm->is_ex) { | ||||
IB_USER_VERBS_CMD_COMMAND_MASK)) { | if (count < (sizeof(hdr) + sizeof(ex_hdr))) | ||||
ret = -EINVAL; | return -EINVAL; | ||||
goto out; | if (copy_from_user(&ex_hdr, buf + sizeof(hdr), sizeof(ex_hdr))) | ||||
return -EFAULT; | |||||
} | } | ||||
command = hdr.command & IB_USER_VERBS_CMD_COMMAND_MASK; | ret = verify_hdr(&hdr, &ex_hdr, count, method_elm); | ||||
if (verify_command_mask(ib_dev, command)) { | if (ret) | ||||
ret = -EOPNOTSUPP; | return ret; | ||||
goto out; | |||||
} | |||||
if (!file->ucontext && | srcu_key = srcu_read_lock(&file->device->disassociate_srcu); | ||||
command != IB_USER_VERBS_CMD_GET_CONTEXT) { | |||||
ret = -EINVAL; | |||||
goto out; | |||||
} | |||||
flags = (hdr.command & | buf += sizeof(hdr); | ||||
IB_USER_VERBS_CMD_FLAGS_MASK) >> IB_USER_VERBS_CMD_FLAGS_SHIFT; | |||||
if (!flags) { | memset(bundle.attr_present, 0, sizeof(bundle.attr_present)); | ||||
if (command >= ARRAY_SIZE(uverbs_cmd_table) || | bundle.ufile = file; | ||||
!uverbs_cmd_table[command]) { | bundle.context = NULL; /* only valid if bundle has uobject */ | ||||
ret = -EINVAL; | if (!method_elm->is_ex) { | ||||
goto out; | size_t in_len = hdr.in_words * 4 - sizeof(hdr); | ||||
} | size_t out_len = hdr.out_words * 4; | ||||
u64 response = 0; | |||||
if (hdr.in_words * 4 != count) { | if (method_elm->has_udata) { | ||||
ret = -EINVAL; | bundle.driver_udata.inlen = | ||||
goto out; | in_len - method_elm->req_size; | ||||
in_len = method_elm->req_size; | |||||
if (bundle.driver_udata.inlen) | |||||
bundle.driver_udata.inbuf = buf + in_len; | |||||
else | |||||
bundle.driver_udata.inbuf = NULL; | |||||
} else { | |||||
memset(&bundle.driver_udata, 0, | |||||
sizeof(bundle.driver_udata)); | |||||
} | } | ||||
ret = uverbs_cmd_table[command](file, ib_dev, | if (method_elm->has_resp) { | ||||
buf + sizeof(hdr), | /* | ||||
hdr.in_words * 4, | * The macros check that if has_resp is set | ||||
hdr.out_words * 4); | * then the command request structure starts | ||||
* with a '__aligned u64 response' member. | |||||
*/ | |||||
ret = get_user(response, (const u64 __user *)buf); | |||||
if (ret) | |||||
goto out_unlock; | |||||
} else if (flags == IB_USER_VERBS_CMD_FLAG_EXTENDED) { | if (method_elm->has_udata) { | ||||
struct ib_uverbs_ex_cmd_hdr ex_hdr; | bundle.driver_udata.outlen = | ||||
struct ib_udata ucore; | out_len - method_elm->resp_size; | ||||
struct ib_udata uhw; | out_len = method_elm->resp_size; | ||||
size_t written_count = count; | if (bundle.driver_udata.outlen) | ||||
bundle.driver_udata.outbuf = | |||||
if (command >= ARRAY_SIZE(uverbs_ex_cmd_table) || | u64_to_user_ptr(response + | ||||
!uverbs_ex_cmd_table[command]) { | out_len); | ||||
ret = -ENOSYS; | else | ||||
goto out; | bundle.driver_udata.outbuf = NULL; | ||||
} | } | ||||
} else { | |||||
if (!file->ucontext) { | bundle.driver_udata.outlen = 0; | ||||
ret = -EINVAL; | bundle.driver_udata.outbuf = NULL; | ||||
goto out; | |||||
} | } | ||||
if (count < (sizeof(hdr) + sizeof(ex_hdr))) { | ib_uverbs_init_udata_buf_or_null( | ||||
ret = -EINVAL; | &bundle.ucore, buf, u64_to_user_ptr(response), | ||||
goto out; | in_len, out_len); | ||||
} | |||||
if (copy_from_user(&ex_hdr, buf + sizeof(hdr), sizeof(ex_hdr))) { | |||||
ret = -EFAULT; | |||||
goto out; | |||||
} | |||||
count -= sizeof(hdr) + sizeof(ex_hdr); | |||||
buf += sizeof(hdr) + sizeof(ex_hdr); | |||||
if ((hdr.in_words + ex_hdr.provider_in_words) * 8 != count) { | |||||
ret = -EINVAL; | |||||
goto out; | |||||
} | |||||
if (ex_hdr.cmd_hdr_reserved) { | |||||
ret = -EINVAL; | |||||
goto out; | |||||
} | |||||
if (ex_hdr.response) { | |||||
if (!hdr.out_words && !ex_hdr.provider_out_words) { | |||||
ret = -EINVAL; | |||||
goto out; | |||||
} | |||||
if (!access_ok((void __user *) (unsigned long) ex_hdr.response, | |||||
(hdr.out_words + ex_hdr.provider_out_words) * 8)) { | |||||
ret = -EFAULT; | |||||
goto out; | |||||
} | |||||
} else { | } else { | ||||
if (hdr.out_words || ex_hdr.provider_out_words) { | buf += sizeof(ex_hdr); | ||||
ret = -EINVAL; | |||||
goto out; | |||||
} | |||||
} | |||||
ib_uverbs_init_udata_buf_or_null(&ucore, buf, | ib_uverbs_init_udata_buf_or_null(&bundle.ucore, buf, | ||||
u64_to_user_ptr(ex_hdr.response), | u64_to_user_ptr(ex_hdr.response), | ||||
hdr.in_words * 8, hdr.out_words * 8); | hdr.in_words * 8, hdr.out_words * 8); | ||||
ib_uverbs_init_udata_buf_or_null(&uhw, | ib_uverbs_init_udata_buf_or_null( | ||||
buf + ucore.inlen, | &bundle.driver_udata, buf + bundle.ucore.inlen, | ||||
u64_to_user_ptr(ex_hdr.response + ucore.outlen), | u64_to_user_ptr(ex_hdr.response + bundle.ucore.outlen), | ||||
ex_hdr.provider_in_words * 8, | ex_hdr.provider_in_words * 8, | ||||
ex_hdr.provider_out_words * 8); | ex_hdr.provider_out_words * 8); | ||||
ret = uverbs_ex_cmd_table[command](file, | |||||
ib_dev, | |||||
&ucore, | |||||
&uhw); | |||||
if (!ret) | |||||
ret = written_count; | |||||
} else { | |||||
ret = -ENOSYS; | |||||
} | } | ||||
out: | ret = method_elm->handler(&bundle); | ||||
out_unlock: | |||||
srcu_read_unlock(&file->device->disassociate_srcu, srcu_key); | srcu_read_unlock(&file->device->disassociate_srcu, srcu_key); | ||||
return ret; | return (ret) ? : count; | ||||
} | } | ||||
static const struct vm_operations_struct rdma_umap_ops; | |||||
static int ib_uverbs_mmap(struct file *filp, struct vm_area_struct *vma) | static int ib_uverbs_mmap(struct file *filp, struct vm_area_struct *vma) | ||||
{ | { | ||||
struct ib_uverbs_file *file = filp->private_data; | struct ib_uverbs_file *file = filp->private_data; | ||||
struct ib_device *ib_dev; | struct ib_ucontext *ucontext; | ||||
int ret = 0; | int ret = 0; | ||||
int srcu_key; | int srcu_key; | ||||
srcu_key = srcu_read_lock(&file->device->disassociate_srcu); | srcu_key = srcu_read_lock(&file->device->disassociate_srcu); | ||||
ib_dev = srcu_dereference(file->device->ib_dev, | ucontext = ib_uverbs_get_ucontext_file(file); | ||||
&file->device->disassociate_srcu); | if (IS_ERR(ucontext)) { | ||||
if (!ib_dev) { | ret = PTR_ERR(ucontext); | ||||
ret = -EIO; | |||||
goto out; | goto out; | ||||
} | } | ||||
vma->vm_ops = &rdma_umap_ops; | |||||
if (!file->ucontext) | ret = ucontext->device->mmap(ucontext, vma); | ||||
ret = -ENODEV; | |||||
else | |||||
ret = ib_dev->mmap(file->ucontext, vma); | |||||
out: | out: | ||||
srcu_read_unlock(&file->device->disassociate_srcu, srcu_key); | srcu_read_unlock(&file->device->disassociate_srcu, srcu_key); | ||||
return ret; | return ret; | ||||
} | } | ||||
/* | /* | ||||
* The VMA has been dup'd, initialize the vm_private_data with a new tracking | |||||
* struct | |||||
*/ | |||||
static void rdma_umap_open(struct vm_area_struct *vma) | |||||
{ | |||||
struct ib_uverbs_file *ufile = vma->vm_file->private_data; | |||||
struct rdma_umap_priv *opriv = vma->vm_private_data; | |||||
struct rdma_umap_priv *priv; | |||||
if (!opriv) | |||||
return; | |||||
/* We are racing with disassociation */ | |||||
if (!down_read_trylock(&ufile->hw_destroy_rwsem)) | |||||
goto out_zap; | |||||
/* | |||||
* Disassociation already completed, the VMA should already be zapped. | |||||
*/ | |||||
if (!ufile->ucontext) | |||||
goto out_unlock; | |||||
priv = kzalloc(sizeof(*priv), GFP_KERNEL); | |||||
if (!priv) | |||||
goto out_unlock; | |||||
rdma_umap_priv_init(priv, vma, opriv->entry); | |||||
up_read(&ufile->hw_destroy_rwsem); | |||||
return; | |||||
out_unlock: | |||||
up_read(&ufile->hw_destroy_rwsem); | |||||
out_zap: | |||||
/* | |||||
* We can't allow the VMA to be created with the actual IO pages, that | |||||
* would break our API contract, and it can't be stopped at this | |||||
* point, so zap it. | |||||
*/ | |||||
vma->vm_private_data = NULL; | |||||
zap_vma_ptes(vma, vma->vm_start, vma->vm_end - vma->vm_start); | |||||
} | |||||
static void rdma_umap_close(struct vm_area_struct *vma) | |||||
{ | |||||
struct ib_uverbs_file *ufile = vma->vm_file->private_data; | |||||
struct rdma_umap_priv *priv = vma->vm_private_data; | |||||
if (!priv) | |||||
return; | |||||
/* | |||||
* The vma holds a reference on the struct file that created it, which | |||||
* in turn means that the ib_uverbs_file is guaranteed to exist at | |||||
* this point. | |||||
*/ | |||||
mutex_lock(&ufile->umap_lock); | |||||
list_del(&priv->list); | |||||
mutex_unlock(&ufile->umap_lock); | |||||
kfree(priv); | |||||
} | |||||
static const struct vm_operations_struct rdma_umap_ops = { | |||||
.open = rdma_umap_open, | |||||
.close = rdma_umap_close, | |||||
}; | |||||
void uverbs_user_mmap_disassociate(struct ib_uverbs_file *ufile) | |||||
{ | |||||
struct rdma_umap_priv *priv, *next_priv; | |||||
lockdep_assert_held(&ufile->hw_destroy_rwsem); | |||||
while (1) { | |||||
struct mm_struct *mm = NULL; | |||||
/* Get an arbitrary mm pointer that hasn't been cleaned yet */ | |||||
mutex_lock(&ufile->umap_lock); | |||||
while (!list_empty(&ufile->umaps)) { | |||||
int ret; | |||||
priv = list_first_entry(&ufile->umaps, | |||||
struct rdma_umap_priv, list); | |||||
mm = priv->vma->vm_mm; | |||||
ret = mmget_not_zero(mm); | |||||
if (!ret) { | |||||
list_del_init(&priv->list); | |||||
mm = NULL; | |||||
continue; | |||||
} | |||||
break; | |||||
} | |||||
mutex_unlock(&ufile->umap_lock); | |||||
if (!mm) | |||||
return; | |||||
/* | |||||
* The umap_lock is nested under mmap_sem since it used within | |||||
* the vma_ops callbacks, so we have to clean the list one mm | |||||
* at a time to get the lock ordering right. Typically there | |||||
* will only be one mm, so no big deal. | |||||
*/ | |||||
down_read(&mm->mmap_sem); | |||||
mutex_lock(&ufile->umap_lock); | |||||
list_for_each_entry_safe (priv, next_priv, &ufile->umaps, | |||||
list) { | |||||
struct vm_area_struct *vma = priv->vma; | |||||
if (vma->vm_mm != mm) | |||||
continue; | |||||
list_del_init(&priv->list); | |||||
zap_vma_ptes(vma, vma->vm_start, | |||||
vma->vm_end - vma->vm_start); | |||||
} | |||||
mutex_unlock(&ufile->umap_lock); | |||||
up_read(&mm->mmap_sem); | |||||
mmput(mm); | |||||
} | |||||
} | |||||
/* | |||||
* ib_uverbs_open() does not need the BKL: | * ib_uverbs_open() does not need the BKL: | ||||
* | * | ||||
* - the ib_uverbs_device structures are properly reference counted and | * - the ib_uverbs_device structures are properly reference counted and | ||||
* everything else is purely local to the file being created, so | * everything else is purely local to the file being created, so | ||||
* races against other open calls are not a problem; | * races against other open calls are not a problem; | ||||
* - there is no ioctl method to race against; | * - there is no ioctl method to race against; | ||||
* - the open method will either immediately run -ENXIO, or all | * - the open method will either immediately run -ENXIO, or all | ||||
* required initialization will be done. | * required initialization will be done. | ||||
*/ | */ | ||||
static int ib_uverbs_open(struct inode *inode, struct file *filp) | static int ib_uverbs_open(struct inode *inode, struct file *filp) | ||||
{ | { | ||||
struct ib_uverbs_device *dev; | struct ib_uverbs_device *dev; | ||||
struct ib_uverbs_file *file; | struct ib_uverbs_file *file; | ||||
struct ib_device *ib_dev; | struct ib_device *ib_dev; | ||||
int ret; | int ret; | ||||
int module_dependent; | int module_dependent; | ||||
int srcu_key; | int srcu_key; | ||||
dev = container_of(inode->i_cdev->si_drv1, struct ib_uverbs_device, cdev); | dev = container_of(inode->i_cdev->si_drv1, struct ib_uverbs_device, cdev); | ||||
if (!atomic_inc_not_zero(&dev->refcount)) | if (!atomic_inc_not_zero(&dev->refcount)) | ||||
return -ENXIO; | return -ENXIO; | ||||
get_device(&dev->dev); | |||||
srcu_key = srcu_read_lock(&dev->disassociate_srcu); | srcu_key = srcu_read_lock(&dev->disassociate_srcu); | ||||
mutex_lock(&dev->lists_mutex); | mutex_lock(&dev->lists_mutex); | ||||
ib_dev = srcu_dereference(dev->ib_dev, | ib_dev = srcu_dereference(dev->ib_dev, | ||||
&dev->disassociate_srcu); | &dev->disassociate_srcu); | ||||
if (!ib_dev) { | if (!ib_dev) { | ||||
ret = -EIO; | ret = -EIO; | ||||
goto err; | goto err; | ||||
} | } | ||||
Show All 15 Lines | if (!file) { | ||||
ret = -ENOMEM; | ret = -ENOMEM; | ||||
if (module_dependent) | if (module_dependent) | ||||
goto err_module; | goto err_module; | ||||
goto err; | goto err; | ||||
} | } | ||||
file->device = dev; | file->device = dev; | ||||
file->ucontext = NULL; | |||||
file->async_file = NULL; | |||||
kref_init(&file->ref); | kref_init(&file->ref); | ||||
mutex_init(&file->mutex); | mutex_init(&file->ucontext_lock); | ||||
mutex_init(&file->cleanup_mutex); | |||||
spin_lock_init(&file->uobjects_lock); | |||||
INIT_LIST_HEAD(&file->uobjects); | |||||
init_rwsem(&file->hw_destroy_rwsem); | |||||
mutex_init(&file->umap_lock); | |||||
INIT_LIST_HEAD(&file->umaps); | |||||
filp->private_data = file; | filp->private_data = file; | ||||
kobject_get(&dev->kobj); | |||||
list_add_tail(&file->list, &dev->uverbs_file_list); | list_add_tail(&file->list, &dev->uverbs_file_list); | ||||
mutex_unlock(&dev->lists_mutex); | mutex_unlock(&dev->lists_mutex); | ||||
srcu_read_unlock(&dev->disassociate_srcu, srcu_key); | srcu_read_unlock(&dev->disassociate_srcu, srcu_key); | ||||
setup_ufile_idr_uobject(file); | |||||
return nonseekable_open(inode, filp); | return nonseekable_open(inode, filp); | ||||
err_module: | err_module: | ||||
module_put(ib_dev->owner); | module_put(ib_dev->owner); | ||||
err: | err: | ||||
mutex_unlock(&dev->lists_mutex); | mutex_unlock(&dev->lists_mutex); | ||||
srcu_read_unlock(&dev->disassociate_srcu, srcu_key); | srcu_read_unlock(&dev->disassociate_srcu, srcu_key); | ||||
if (atomic_dec_and_test(&dev->refcount)) | if (atomic_dec_and_test(&dev->refcount)) | ||||
ib_uverbs_comp_dev(dev); | ib_uverbs_comp_dev(dev); | ||||
put_device(&dev->dev); | |||||
return ret; | return ret; | ||||
} | } | ||||
static int ib_uverbs_close(struct inode *inode, struct file *filp) | static int ib_uverbs_close(struct inode *inode, struct file *filp) | ||||
{ | { | ||||
struct ib_uverbs_file *file = filp->private_data; | struct ib_uverbs_file *file = filp->private_data; | ||||
struct ib_uverbs_device *dev = file->device; | |||||
mutex_lock(&file->cleanup_mutex); | uverbs_destroy_ufile_hw(file, RDMA_REMOVE_CLOSE); | ||||
if (file->ucontext) { | |||||
ib_uverbs_cleanup_ucontext(file, file->ucontext); | |||||
file->ucontext = NULL; | |||||
} | |||||
mutex_unlock(&file->cleanup_mutex); | |||||
mutex_lock(&file->device->lists_mutex); | mutex_lock(&file->device->lists_mutex); | ||||
if (!file->is_closed) { | list_del_init(&file->list); | ||||
list_del(&file->list); | |||||
file->is_closed = 1; | |||||
} | |||||
mutex_unlock(&file->device->lists_mutex); | mutex_unlock(&file->device->lists_mutex); | ||||
if (file->async_file) | |||||
kref_put(&file->async_file->ref, ib_uverbs_release_event_file); | |||||
kref_put(&file->ref, ib_uverbs_release_file); | kref_put(&file->ref, ib_uverbs_release_file); | ||||
kobject_put(&dev->kobj); | |||||
return 0; | return 0; | ||||
} | } | ||||
static const struct file_operations uverbs_fops = { | static const struct file_operations uverbs_fops = { | ||||
.owner = THIS_MODULE, | .owner = THIS_MODULE, | ||||
.write = ib_uverbs_write, | .write = ib_uverbs_write, | ||||
.open = ib_uverbs_open, | .open = ib_uverbs_open, | ||||
.release = ib_uverbs_close, | .release = ib_uverbs_close, | ||||
.llseek = no_llseek, | .llseek = no_llseek, | ||||
.unlocked_ioctl = ib_uverbs_ioctl, | |||||
.compat_ioctl = NULL, | |||||
}; | }; | ||||
static const struct file_operations uverbs_mmap_fops = { | static const struct file_operations uverbs_mmap_fops = { | ||||
.owner = THIS_MODULE, | .owner = THIS_MODULE, | ||||
.write = ib_uverbs_write, | .write = ib_uverbs_write, | ||||
.mmap = ib_uverbs_mmap, | .mmap = ib_uverbs_mmap, | ||||
.open = ib_uverbs_open, | .open = ib_uverbs_open, | ||||
.release = ib_uverbs_close, | .release = ib_uverbs_close, | ||||
.llseek = no_llseek, | .llseek = no_llseek, | ||||
.unlocked_ioctl = ib_uverbs_ioctl, | |||||
.compat_ioctl = NULL, | |||||
}; | }; | ||||
static struct ib_client uverbs_client = { | static struct ib_client uverbs_client = { | ||||
.name = "uverbs", | .name = "uverbs", | ||||
.add = ib_uverbs_add_one, | .add = ib_uverbs_add_one, | ||||
.remove = ib_uverbs_remove_one | .remove = ib_uverbs_remove_one | ||||
}; | }; | ||||
static ssize_t show_ibdev(struct device *device, struct device_attribute *attr, | static ssize_t ibdev_show(struct device *device, struct device_attribute *attr, | ||||
char *buf) | char *buf) | ||||
{ | { | ||||
struct ib_uverbs_device *dev = | |||||
container_of(device, struct ib_uverbs_device, dev); | |||||
int ret = -ENODEV; | int ret = -ENODEV; | ||||
int srcu_key; | int srcu_key; | ||||
struct ib_uverbs_device *dev = dev_get_drvdata(device); | |||||
struct ib_device *ib_dev; | struct ib_device *ib_dev; | ||||
if (!dev) | |||||
return -ENODEV; | |||||
srcu_key = srcu_read_lock(&dev->disassociate_srcu); | srcu_key = srcu_read_lock(&dev->disassociate_srcu); | ||||
ib_dev = srcu_dereference(dev->ib_dev, &dev->disassociate_srcu); | ib_dev = srcu_dereference(dev->ib_dev, &dev->disassociate_srcu); | ||||
if (ib_dev) | if (ib_dev) | ||||
ret = sprintf(buf, "%s\n", ib_dev->name); | ret = sprintf(buf, "%s\n", ib_dev->name); | ||||
srcu_read_unlock(&dev->disassociate_srcu, srcu_key); | srcu_read_unlock(&dev->disassociate_srcu, srcu_key); | ||||
return ret; | return ret; | ||||
} | } | ||||
static DEVICE_ATTR(ibdev, S_IRUGO, show_ibdev, NULL); | static DEVICE_ATTR(ibdev, S_IRUGO, ibdev_show, NULL); | ||||
static ssize_t show_dev_abi_version(struct device *device, | static ssize_t abi_version_show(struct device *device, | ||||
struct device_attribute *attr, char *buf) | struct device_attribute *attr, char *buf) | ||||
{ | { | ||||
struct ib_uverbs_device *dev = dev_get_drvdata(device); | struct ib_uverbs_device *dev = | ||||
container_of(device, struct ib_uverbs_device, dev); | |||||
int ret = -ENODEV; | int ret = -ENODEV; | ||||
int srcu_key; | int srcu_key; | ||||
struct ib_device *ib_dev; | struct ib_device *ib_dev; | ||||
if (!dev) | |||||
return -ENODEV; | |||||
srcu_key = srcu_read_lock(&dev->disassociate_srcu); | srcu_key = srcu_read_lock(&dev->disassociate_srcu); | ||||
ib_dev = srcu_dereference(dev->ib_dev, &dev->disassociate_srcu); | ib_dev = srcu_dereference(dev->ib_dev, &dev->disassociate_srcu); | ||||
if (ib_dev) | if (ib_dev) | ||||
ret = sprintf(buf, "%d\n", ib_dev->uverbs_abi_ver); | ret = sprintf(buf, "%d\n", ib_dev->uverbs_abi_ver); | ||||
srcu_read_unlock(&dev->disassociate_srcu, srcu_key); | srcu_read_unlock(&dev->disassociate_srcu, srcu_key); | ||||
return ret; | return ret; | ||||
} | } | ||||
static DEVICE_ATTR(abi_version, S_IRUGO, show_dev_abi_version, NULL); | static DEVICE_ATTR(abi_version, S_IRUGO, abi_version_show, NULL); | ||||
static CLASS_ATTR_STRING(abi_version, S_IRUGO, | static CLASS_ATTR_STRING(abi_version, S_IRUGO, | ||||
__stringify(IB_USER_VERBS_ABI_VERSION)); | __stringify(IB_USER_VERBS_ABI_VERSION)); | ||||
static dev_t overflow_maj; | static int ib_uverbs_create_uapi(struct ib_device *device, | ||||
static DECLARE_BITMAP(overflow_map, IB_UVERBS_MAX_DEVICES); | struct ib_uverbs_device *uverbs_dev) | ||||
/* | |||||
* If we have more than IB_UVERBS_MAX_DEVICES, dynamically overflow by | |||||
* requesting a new major number and doubling the number of max devices we | |||||
* support. It's stupid, but simple. | |||||
*/ | |||||
static int find_overflow_devnum(void) | |||||
{ | { | ||||
int ret; | struct uverbs_api *uapi; | ||||
if (!overflow_maj) { | uapi = uverbs_alloc_api(device); | ||||
ret = alloc_chrdev_region(&overflow_maj, 0, IB_UVERBS_MAX_DEVICES, | if (IS_ERR(uapi)) | ||||
"infiniband_verbs"); | return PTR_ERR(uapi); | ||||
if (ret) { | |||||
pr_err("user_verbs: couldn't register dynamic device number\n"); | |||||
return ret; | |||||
} | |||||
} | |||||
ret = find_first_zero_bit(overflow_map, IB_UVERBS_MAX_DEVICES); | uverbs_dev->uapi = uapi; | ||||
if (ret >= IB_UVERBS_MAX_DEVICES) | return 0; | ||||
return -1; | |||||
return ret; | |||||
} | } | ||||
static ssize_t | static ssize_t | ||||
show_dev_device(struct device *device, struct device_attribute *attr, char *buf) | show_dev_device(struct device *device, struct device_attribute *attr, char *buf) | ||||
{ | { | ||||
struct ib_uverbs_device *dev = dev_get_drvdata(device); | struct ib_uverbs_device *dev = | ||||
container_of(device, struct ib_uverbs_device, dev); | |||||
if (!dev || !dev->ib_dev->dma_device) | if (!dev->ib_dev->dma_device) | ||||
return -ENODEV; | return -ENODEV; | ||||
return sprintf(buf, "0x%04x\n", | return sprintf(buf, "0x%04x\n", | ||||
((struct pci_dev *)dev->ib_dev->dma_device)->device); | ((struct pci_dev *)dev->ib_dev->dma_device)->device); | ||||
} | } | ||||
static DEVICE_ATTR(device, S_IRUGO, show_dev_device, NULL); | static DEVICE_ATTR(device, S_IRUGO, show_dev_device, NULL); | ||||
static ssize_t | static ssize_t | ||||
show_dev_vendor(struct device *device, struct device_attribute *attr, char *buf) | show_dev_vendor(struct device *device, struct device_attribute *attr, char *buf) | ||||
{ | { | ||||
struct ib_uverbs_device *dev = dev_get_drvdata(device); | struct ib_uverbs_device *dev = | ||||
container_of(device, struct ib_uverbs_device, dev); | |||||
if (!dev || !dev->ib_dev->dma_device) | if (!dev->ib_dev->dma_device) | ||||
return -ENODEV; | return -ENODEV; | ||||
return sprintf(buf, "0x%04x\n", | return sprintf(buf, "0x%04x\n", | ||||
((struct pci_dev *)dev->ib_dev->dma_device)->vendor); | ((struct pci_dev *)dev->ib_dev->dma_device)->vendor); | ||||
} | } | ||||
static DEVICE_ATTR(vendor, S_IRUGO, show_dev_vendor, NULL); | static DEVICE_ATTR(vendor, S_IRUGO, show_dev_vendor, NULL); | ||||
struct attribute *device_attrs[] = | static struct attribute *ib_dev_attrs[] = { | ||||
{ | |||||
&dev_attr_device.attr, | &dev_attr_device.attr, | ||||
&dev_attr_vendor.attr, | &dev_attr_vendor.attr, | ||||
NULL | NULL, | ||||
}; | }; | ||||
static struct attribute_group device_group = { | static const struct attribute_group dev_attr_group = { | ||||
.name = "device", | .name = "device", | ||||
.attrs = device_attrs | .attrs = ib_dev_attrs, | ||||
}; | }; | ||||
static void ib_uverbs_add_one(struct ib_device *device) | static void ib_uverbs_add_one(struct ib_device *device) | ||||
{ | { | ||||
int devnum; | int devnum; | ||||
dev_t base; | dev_t base; | ||||
struct ib_uverbs_device *uverbs_dev; | struct ib_uverbs_device *uverbs_dev; | ||||
int ret; | int ret; | ||||
if (!device->alloc_ucontext) | if (!device->alloc_ucontext) | ||||
return; | return; | ||||
uverbs_dev = kzalloc(sizeof *uverbs_dev, GFP_KERNEL); | uverbs_dev = kzalloc(sizeof(*uverbs_dev), GFP_KERNEL); | ||||
if (!uverbs_dev) | if (!uverbs_dev) | ||||
return; | return; | ||||
ret = init_srcu_struct(&uverbs_dev->disassociate_srcu); | ret = init_srcu_struct(&uverbs_dev->disassociate_srcu); | ||||
if (ret) { | if (ret) { | ||||
kfree(uverbs_dev); | kfree(uverbs_dev); | ||||
return; | return; | ||||
} | } | ||||
uverbs_dev->dev.class = uverbs_class; | |||||
uverbs_dev->dev.parent = device->dev.parent; | |||||
uverbs_dev->dev.release = ib_uverbs_release_dev; | |||||
device_initialize(&uverbs_dev->dev); | |||||
atomic_set(&uverbs_dev->refcount, 1); | atomic_set(&uverbs_dev->refcount, 1); | ||||
init_completion(&uverbs_dev->comp); | init_completion(&uverbs_dev->comp); | ||||
uverbs_dev->xrcd_tree = RB_ROOT; | uverbs_dev->xrcd_tree = RB_ROOT; | ||||
mutex_init(&uverbs_dev->xrcd_tree_mutex); | mutex_init(&uverbs_dev->xrcd_tree_mutex); | ||||
kobject_init(&uverbs_dev->kobj, &ib_uverbs_dev_ktype); | |||||
mutex_init(&uverbs_dev->lists_mutex); | mutex_init(&uverbs_dev->lists_mutex); | ||||
INIT_LIST_HEAD(&uverbs_dev->uverbs_file_list); | INIT_LIST_HEAD(&uverbs_dev->uverbs_file_list); | ||||
INIT_LIST_HEAD(&uverbs_dev->uverbs_events_file_list); | rcu_assign_pointer(uverbs_dev->ib_dev, device); | ||||
uverbs_dev->num_comp_vectors = device->num_comp_vectors; | |||||
spin_lock(&map_lock); | devnum = ida_alloc_max(&uverbs_ida, IB_UVERBS_MAX_DEVICES - 1, | ||||
devnum = find_first_zero_bit(dev_map, IB_UVERBS_MAX_DEVICES); | GFP_KERNEL); | ||||
if (devnum >= IB_UVERBS_MAX_DEVICES) { | |||||
spin_unlock(&map_lock); | |||||
devnum = find_overflow_devnum(); | |||||
if (devnum < 0) | if (devnum < 0) | ||||
goto err; | goto err; | ||||
spin_lock(&map_lock); | |||||
uverbs_dev->devnum = devnum + IB_UVERBS_MAX_DEVICES; | |||||
base = devnum + overflow_maj; | |||||
set_bit(devnum, overflow_map); | |||||
} else { | |||||
uverbs_dev->devnum = devnum; | uverbs_dev->devnum = devnum; | ||||
base = devnum + IB_UVERBS_BASE_DEV; | if (devnum >= IB_UVERBS_NUM_FIXED_MINOR) | ||||
set_bit(devnum, dev_map); | base = dynamic_uverbs_dev + devnum - IB_UVERBS_NUM_FIXED_MINOR; | ||||
} | else | ||||
spin_unlock(&map_lock); | base = IB_UVERBS_BASE_DEV + devnum; | ||||
rcu_assign_pointer(uverbs_dev->ib_dev, device); | if (ib_uverbs_create_uapi(device, uverbs_dev)) | ||||
uverbs_dev->num_comp_vectors = device->num_comp_vectors; | goto err_uapi; | ||||
cdev_init(&uverbs_dev->cdev, NULL); | uverbs_dev->dev.devt = base; | ||||
dev_set_name(&uverbs_dev->dev, "uverbs%d", uverbs_dev->devnum); | |||||
cdev_init(&uverbs_dev->cdev, | |||||
device->mmap ? &uverbs_mmap_fops : &uverbs_fops); | |||||
uverbs_dev->cdev.owner = THIS_MODULE; | uverbs_dev->cdev.owner = THIS_MODULE; | ||||
uverbs_dev->cdev.ops = device->mmap ? &uverbs_mmap_fops : &uverbs_fops; | |||||
uverbs_dev->cdev.kobj.parent = &uverbs_dev->kobj; | |||||
kobject_set_name(&uverbs_dev->cdev.kobj, "uverbs%d", uverbs_dev->devnum); | kobject_set_name(&uverbs_dev->cdev.kobj, "uverbs%d", uverbs_dev->devnum); | ||||
if (cdev_add(&uverbs_dev->cdev, base, 1)) | ret = cdev_device_add(&uverbs_dev->cdev, &uverbs_dev->dev); | ||||
goto err_cdev; | if (ret) | ||||
goto err_uapi; | |||||
uverbs_dev->dev = device_create(uverbs_class, device->dma_device, | if (device_create_file(&uverbs_dev->dev, &dev_attr_ibdev)) | ||||
uverbs_dev->cdev.dev, uverbs_dev, | |||||
"uverbs%d", uverbs_dev->devnum); | |||||
if (IS_ERR(uverbs_dev->dev)) | |||||
goto err_cdev; | goto err_cdev; | ||||
if (device_create_file(&uverbs_dev->dev, &dev_attr_abi_version)) | |||||
goto err_cdev; | |||||
if (sysfs_create_group(&uverbs_dev->dev.kobj, &dev_attr_group)) | |||||
goto err_cdev; | |||||
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; | |||||
if (sysfs_create_group(&uverbs_dev->dev->kobj, &device_group)) | |||||
goto err_class; | |||||
ib_set_client_data(device, &uverbs_client, uverbs_dev); | ib_set_client_data(device, &uverbs_client, uverbs_dev); | ||||
return; | return; | ||||
err_class: | |||||
device_destroy(uverbs_class, uverbs_dev->cdev.dev); | |||||
err_cdev: | err_cdev: | ||||
cdev_del(&uverbs_dev->cdev); | cdev_device_del(&uverbs_dev->cdev, &uverbs_dev->dev); | ||||
if (uverbs_dev->devnum < IB_UVERBS_MAX_DEVICES) | err_uapi: | ||||
clear_bit(devnum, dev_map); | ida_free(&uverbs_ida, devnum); | ||||
else | |||||
clear_bit(devnum, overflow_map); | |||||
err: | err: | ||||
if (atomic_dec_and_test(&uverbs_dev->refcount)) | if (atomic_dec_and_test(&uverbs_dev->refcount)) | ||||
ib_uverbs_comp_dev(uverbs_dev); | ib_uverbs_comp_dev(uverbs_dev); | ||||
wait_for_completion(&uverbs_dev->comp); | wait_for_completion(&uverbs_dev->comp); | ||||
kobject_put(&uverbs_dev->kobj); | put_device(&uverbs_dev->dev); | ||||
return; | return; | ||||
} | } | ||||
static void ib_uverbs_free_hw_resources(struct ib_uverbs_device *uverbs_dev, | static void ib_uverbs_free_hw_resources(struct ib_uverbs_device *uverbs_dev, | ||||
struct ib_device *ib_dev) | struct ib_device *ib_dev) | ||||
{ | { | ||||
struct ib_uverbs_file *file; | struct ib_uverbs_file *file; | ||||
struct ib_uverbs_event_file *event_file; | |||||
struct ib_event event; | |||||
/* Pending running commands to terminate */ | /* Pending running commands to terminate */ | ||||
synchronize_srcu(&uverbs_dev->disassociate_srcu); | uverbs_disassociate_api_pre(uverbs_dev); | ||||
event.event = IB_EVENT_DEVICE_FATAL; | |||||
event.element.port_num = 0; | |||||
event.device = ib_dev; | |||||
mutex_lock(&uverbs_dev->lists_mutex); | mutex_lock(&uverbs_dev->lists_mutex); | ||||
while (!list_empty(&uverbs_dev->uverbs_file_list)) { | while (!list_empty(&uverbs_dev->uverbs_file_list)) { | ||||
struct ib_ucontext *ucontext; | |||||
file = list_first_entry(&uverbs_dev->uverbs_file_list, | file = list_first_entry(&uverbs_dev->uverbs_file_list, | ||||
struct ib_uverbs_file, list); | struct ib_uverbs_file, list); | ||||
file->is_closed = 1; | list_del_init(&file->list); | ||||
list_del(&file->list); | |||||
kref_get(&file->ref); | kref_get(&file->ref); | ||||
/* We must release the mutex before going ahead and calling | |||||
* uverbs_cleanup_ufile, as it might end up indirectly calling | |||||
* uverbs_close, for example due to freeing the resources (e.g | |||||
* mmput). | |||||
*/ | |||||
mutex_unlock(&uverbs_dev->lists_mutex); | mutex_unlock(&uverbs_dev->lists_mutex); | ||||
ib_uverbs_async_handler(READ_ONCE(file->async_file), 0, | |||||
IB_EVENT_DEVICE_FATAL, NULL, NULL); | |||||
mutex_lock(&file->cleanup_mutex); | uverbs_destroy_ufile_hw(file, RDMA_REMOVE_DRIVER_REMOVE); | ||||
ucontext = file->ucontext; | kref_put(&file->ref, ib_uverbs_release_file); | ||||
file->ucontext = NULL; | |||||
mutex_unlock(&file->cleanup_mutex); | |||||
/* At this point ib_uverbs_close cannot be running | |||||
* ib_uverbs_cleanup_ucontext | |||||
*/ | |||||
if (ucontext) { | |||||
/* We must release the mutex before going ahead and | |||||
* calling disassociate_ucontext. disassociate_ucontext | |||||
* might end up indirectly calling uverbs_close, | |||||
* for example due to freeing the resources | |||||
* (e.g mmput). | |||||
*/ | |||||
ib_uverbs_event_handler(&file->event_handler, &event); | |||||
ib_dev->disassociate_ucontext(ucontext); | |||||
ib_uverbs_cleanup_ucontext(file, ucontext); | |||||
} | |||||
mutex_lock(&uverbs_dev->lists_mutex); | mutex_lock(&uverbs_dev->lists_mutex); | ||||
kref_put(&file->ref, ib_uverbs_release_file); | |||||
} | } | ||||
mutex_unlock(&uverbs_dev->lists_mutex); | |||||
while (!list_empty(&uverbs_dev->uverbs_events_file_list)) { | uverbs_disassociate_api(uverbs_dev->uapi); | ||||
event_file = list_first_entry(&uverbs_dev-> | |||||
uverbs_events_file_list, | |||||
struct ib_uverbs_event_file, | |||||
list); | |||||
spin_lock_irq(&event_file->lock); | |||||
event_file->is_closed = 1; | |||||
spin_unlock_irq(&event_file->lock); | |||||
list_del(&event_file->list); | |||||
if (event_file->is_async) { | |||||
ib_unregister_event_handler(&event_file->uverbs_file-> | |||||
event_handler); | |||||
event_file->uverbs_file->event_handler.device = NULL; | |||||
} | } | ||||
wake_up_interruptible(&event_file->poll_wait); | |||||
kill_fasync(&event_file->async_queue, SIGIO, POLL_IN); | |||||
} | |||||
mutex_unlock(&uverbs_dev->lists_mutex); | |||||
} | |||||
static void ib_uverbs_remove_one(struct ib_device *device, void *client_data) | static void ib_uverbs_remove_one(struct ib_device *device, void *client_data) | ||||
{ | { | ||||
struct ib_uverbs_device *uverbs_dev = client_data; | struct ib_uverbs_device *uverbs_dev = client_data; | ||||
int wait_clients = 1; | int wait_clients = 1; | ||||
if (!uverbs_dev) | if (!uverbs_dev) | ||||
return; | return; | ||||
sysfs_remove_group(&uverbs_dev->dev->kobj, &device_group); | cdev_device_del(&uverbs_dev->cdev, &uverbs_dev->dev); | ||||
dev_set_drvdata(uverbs_dev->dev, NULL); | ida_free(&uverbs_ida, uverbs_dev->devnum); | ||||
device_destroy(uverbs_class, uverbs_dev->cdev.dev); | |||||
cdev_del(&uverbs_dev->cdev); | |||||
if (uverbs_dev->devnum < IB_UVERBS_MAX_DEVICES) | |||||
clear_bit(uverbs_dev->devnum, dev_map); | |||||
else | |||||
clear_bit(uverbs_dev->devnum - IB_UVERBS_MAX_DEVICES, overflow_map); | |||||
if (device->disassociate_ucontext) { | if (device->disassociate_ucontext) { | ||||
/* We disassociate HW resources and immediately return. | /* We disassociate HW resources and immediately return. | ||||
* Userspace will see a EIO errno for all future access. | * Userspace will see a EIO errno for all future access. | ||||
* Upon returning, ib_device may be freed internally and is not | * Upon returning, ib_device may be freed internally and is not | ||||
* valid any more. | * valid any more. | ||||
* uverbs_device is still available until all clients close | * uverbs_device is still available until all clients close | ||||
* their files, then the uverbs device ref count will be zero | * their files, then the uverbs device ref count will be zero | ||||
* and its resources will be freed. | * and its resources will be freed. | ||||
* Note: At this point no more files can be opened since the | * Note: At this point no more files can be opened since the | ||||
* cdev was deleted, however active clients can still issue | * cdev was deleted, however active clients can still issue | ||||
* commands and close their open files. | * commands and close their open files. | ||||
*/ | */ | ||||
rcu_assign_pointer(uverbs_dev->ib_dev, NULL); | |||||
ib_uverbs_free_hw_resources(uverbs_dev, device); | ib_uverbs_free_hw_resources(uverbs_dev, device); | ||||
wait_clients = 0; | wait_clients = 0; | ||||
} | } | ||||
if (atomic_dec_and_test(&uverbs_dev->refcount)) | if (atomic_dec_and_test(&uverbs_dev->refcount)) | ||||
ib_uverbs_comp_dev(uverbs_dev); | ib_uverbs_comp_dev(uverbs_dev); | ||||
if (wait_clients) | if (wait_clients) | ||||
wait_for_completion(&uverbs_dev->comp); | wait_for_completion(&uverbs_dev->comp); | ||||
kobject_put(&uverbs_dev->kobj); | |||||
put_device(&uverbs_dev->dev); | |||||
} | } | ||||
static char *uverbs_devnode(struct device *dev, umode_t *mode) | static char *uverbs_devnode(struct device *dev, umode_t *mode) | ||||
{ | { | ||||
if (mode) | if (mode) | ||||
*mode = 0666; | *mode = 0666; | ||||
return kasprintf(GFP_KERNEL, "infiniband/%s", dev_name(dev)); | return kasprintf(GFP_KERNEL, "infiniband/%s", dev_name(dev)); | ||||
} | } | ||||
static int __init ib_uverbs_init(void) | static int __init ib_uverbs_init(void) | ||||
{ | { | ||||
int ret; | int ret; | ||||
ret = register_chrdev_region(IB_UVERBS_BASE_DEV, IB_UVERBS_MAX_DEVICES, | ret = register_chrdev_region(IB_UVERBS_BASE_DEV, | ||||
IB_UVERBS_NUM_FIXED_MINOR, | |||||
"infiniband_verbs"); | "infiniband_verbs"); | ||||
if (ret) { | if (ret) { | ||||
pr_err("user_verbs: couldn't register device number\n"); | pr_err("user_verbs: couldn't register device number\n"); | ||||
goto out; | goto out; | ||||
} | } | ||||
ret = alloc_chrdev_region(&dynamic_uverbs_dev, 0, | |||||
IB_UVERBS_NUM_DYNAMIC_MINOR, | |||||
"infiniband_verbs"); | |||||
if (ret) { | |||||
pr_err("couldn't register dynamic device number\n"); | |||||
goto out_alloc; | |||||
} | |||||
uverbs_class = class_create(THIS_MODULE, "infiniband_verbs"); | uverbs_class = class_create(THIS_MODULE, "infiniband_verbs"); | ||||
if (IS_ERR(uverbs_class)) { | if (IS_ERR(uverbs_class)) { | ||||
ret = PTR_ERR(uverbs_class); | ret = PTR_ERR(uverbs_class); | ||||
pr_err("user_verbs: couldn't create class infiniband_verbs\n"); | pr_err("user_verbs: couldn't create class infiniband_verbs\n"); | ||||
goto out_chrdev; | goto out_chrdev; | ||||
} | } | ||||
uverbs_class->devnode = uverbs_devnode; | uverbs_class->devnode = uverbs_devnode; | ||||
Show All 11 Lines | static int __init ib_uverbs_init(void) | ||||
} | } | ||||
return 0; | return 0; | ||||
out_class: | out_class: | ||||
class_destroy(uverbs_class); | class_destroy(uverbs_class); | ||||
out_chrdev: | out_chrdev: | ||||
unregister_chrdev_region(IB_UVERBS_BASE_DEV, IB_UVERBS_MAX_DEVICES); | unregister_chrdev_region(dynamic_uverbs_dev, | ||||
IB_UVERBS_NUM_DYNAMIC_MINOR); | |||||
out_alloc: | |||||
unregister_chrdev_region(IB_UVERBS_BASE_DEV, | |||||
IB_UVERBS_NUM_FIXED_MINOR); | |||||
out: | out: | ||||
return ret; | return ret; | ||||
} | } | ||||
static void __exit ib_uverbs_cleanup(void) | static void __exit ib_uverbs_cleanup(void) | ||||
{ | { | ||||
ib_unregister_client(&uverbs_client); | ib_unregister_client(&uverbs_client); | ||||
class_destroy(uverbs_class); | class_destroy(uverbs_class); | ||||
unregister_chrdev_region(IB_UVERBS_BASE_DEV, IB_UVERBS_MAX_DEVICES); | unregister_chrdev_region(IB_UVERBS_BASE_DEV, | ||||
if (overflow_maj) | IB_UVERBS_NUM_FIXED_MINOR); | ||||
unregister_chrdev_region(overflow_maj, IB_UVERBS_MAX_DEVICES); | unregister_chrdev_region(dynamic_uverbs_dev, | ||||
idr_destroy(&ib_uverbs_pd_idr); | IB_UVERBS_NUM_DYNAMIC_MINOR); | ||||
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_order(ib_uverbs_init, SI_ORDER_FIFTH); | module_init_order(ib_uverbs_init, SI_ORDER_FIFTH); | ||||
module_exit_order(ib_uverbs_cleanup, SI_ORDER_FIFTH); | module_exit_order(ib_uverbs_cleanup, SI_ORDER_FIFTH); |