From patchwork Tue Mar 31 18:43:02 2009 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Gregory Haskins X-Patchwork-Id: 15455 Received: from vger.kernel.org (vger.kernel.org [209.132.176.167]) by demeter.kernel.org (8.14.2/8.14.2) with ESMTP id n2VIf3XS031684 for ; Tue, 31 Mar 2009 18:41:03 GMT Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753886AbZCaSlD (ORCPT ); Tue, 31 Mar 2009 14:41:03 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1761565AbZCaSk7 (ORCPT ); Tue, 31 Mar 2009 14:40:59 -0400 Received: from victor.provo.novell.com ([137.65.250.26]:52022 "EHLO victor.provo.novell.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754884AbZCaSkq (ORCPT ); Tue, 31 Mar 2009 14:40:46 -0400 Received: from dev.haskins.net (prv-ext-foundry1.gns.novell.com [137.65.251.240]) by victor.provo.novell.com with ESMTP (TLS encrypted); Tue, 31 Mar 2009 12:40:40 -0600 Received: from dev.haskins.net (localhost [127.0.0.1]) by dev.haskins.net (Postfix) with ESMTP id C5BEE46422A; Tue, 31 Mar 2009 14:43:02 -0400 (EDT) From: Gregory Haskins Subject: [RFC PATCH 03/17] vbus: add connection-client helper infrastructure To: linux-kernel@vger.kernel.org Cc: agraf@suse.de, pmullaney@novell.com, pmorreale@novell.com, anthony@codemonkey.ws, rusty@rustcorp.com.au, netdev@vger.kernel.org, kvm@vger.kernel.org Date: Tue, 31 Mar 2009 14:43:02 -0400 Message-ID: <20090331184302.28333.18745.stgit@dev.haskins.net> In-Reply-To: <20090331184057.28333.77287.stgit@dev.haskins.net> References: <20090331184057.28333.77287.stgit@dev.haskins.net> User-Agent: StGIT/0.14.3 MIME-Version: 1.0 Sender: kvm-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: kvm@vger.kernel.org We expect to have various types of connection-clients (e.g. userspace, kvm, etc), each of which is likely to have common access patterns and marshalling duties. Therefore we create a "client" API to simplify client development by helping with mundane tasks such as handle-2-pointer translation, etc. Special thanks to Pat Mullaney for suggesting the optimization to pass a cookie object down during DEVICESHM operations to save lookup overhead on the event channel. Signed-off-by: Gregory Haskins --- include/linux/vbus_client.h | 115 +++++++++ kernel/vbus/Makefile | 2 kernel/vbus/client.c | 527 +++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 643 insertions(+), 1 deletions(-) create mode 100644 include/linux/vbus_client.h create mode 100644 kernel/vbus/client.c -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html diff --git a/include/linux/vbus_client.h b/include/linux/vbus_client.h new file mode 100644 index 0000000..62dab78 --- /dev/null +++ b/include/linux/vbus_client.h @@ -0,0 +1,115 @@ +/* + * Copyright 2009 Novell. All Rights Reserved. + * + * Virtual-Bus - Client interface + * + * We expect to have various types of connection-clients (e.g. userspace, + * kvm, etc). Each client will be connecting from some environment outside + * of the kernel, and therefore will not have direct access to the API as + * presented in ./linux/vbus.h. There will undoubtedly be some parameter + * marshalling that must occur, as well as common patterns for the handling + * of those marshalled parameters (e.g. translating a handle into a pointer, + * etc). + * + * Therefore this "client" API is provided to simplify the development + * of any clients. Of course, a client is free to bypass this API entirely + * and communicate with the direct VBUS API if desired. + * + * Author: + * Gregory Haskins + * + * This file is free software; you can redistribute it and/or modify + * it under the terms of version 2 of the GNU General Public License + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software Foundation, + * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA. + */ + +#ifndef _LINUX_VBUS_CLIENT_H +#define _LINUX_VBUS_CLIENT_H + +#include +#include + +struct vbus_deviceopen { + __u32 devid; + __u32 version; /* device ABI version */ + __u64 handle; /* return value for devh */ +}; + +struct vbus_devicecall { + __u64 devh; /* device-handle (returned from DEVICEOPEN */ + __u32 func; + __u32 len; + __u32 flags; + __u64 datap; +}; + +struct vbus_deviceshm { + __u64 devh; /* device-handle (returned from DEVICEOPEN */ + __u32 id; + __u32 len; + __u32 flags; + struct { + __u32 offset; + __u32 prio; + __u64 cookie; /* token to pass back when signaling client */ + } signal; + __u64 datap; + __u64 handle; /* return value for signaling from client to kernel */ +}; + +#ifdef __KERNEL__ + +#include +#include +#include + +struct vbus_client; + +struct vbus_client_ops { + int (*deviceopen)(struct vbus_client *client, struct vbus_memctx *ctx, + __u32 devid, __u32 version, __u64 *devh); + int (*deviceclose)(struct vbus_client *client, __u64 devh); + int (*devicecall)(struct vbus_client *client, + __u64 devh, __u32 func, + void *data, __u32 len, __u32 flags); + int (*deviceshm)(struct vbus_client *client, + __u64 devh, __u32 id, + struct vbus_shm *shm, struct shm_signal *signal, + __u32 flags, __u64 *handle); + int (*shmsignal)(struct vbus_client *client, __u64 handle); + void (*release)(struct vbus_client *client); +}; + +struct vbus_client { + atomic_t refs; + struct vbus_client_ops *ops; +}; + +static inline void vbus_client_get(struct vbus_client *client) +{ + atomic_inc(&client->refs); +} + +static inline void vbus_client_put(struct vbus_client *client) +{ + if (atomic_dec_and_test(&client->refs)) + client->ops->release(client); +} + +struct vbus_client *vbus_client_attach(struct vbus *bus); + +extern struct vbus_memctx *current_memctx; +struct vbus_memctx *task_memctx_alloc(struct task_struct *task); + +#endif /* __KERNEL__ */ + +#endif /* _LINUX_VBUS_CLIENT_H */ diff --git a/kernel/vbus/Makefile b/kernel/vbus/Makefile index 367f65b..4d440e5 100644 --- a/kernel/vbus/Makefile +++ b/kernel/vbus/Makefile @@ -1 +1 @@ -obj-$(CONFIG_VBUS) += core.o devclass.o config.o attribute.o map.o +obj-$(CONFIG_VBUS) += core.o devclass.o config.o attribute.o map.o client.o diff --git a/kernel/vbus/client.c b/kernel/vbus/client.c new file mode 100644 index 0000000..ff8d6df --- /dev/null +++ b/kernel/vbus/client.c @@ -0,0 +1,527 @@ +#include +#include +#include +#include +#include +#include +#include +#include "vbus.h" + +static int +nodeptr_item_compare(struct rb_node *lhs, struct rb_node *rhs) +{ + unsigned long l = (unsigned long)lhs; + unsigned long r = (unsigned long)rhs; + + return l - r; +} + +static int +nodeptr_key_compare(const void *key, struct rb_node *node) +{ + unsigned long item = (unsigned long)node; + unsigned long _key = *(unsigned long *)key; + + return _key - item; +} + +static struct map_ops nodeptr_map_ops = { + .key_compare = &nodeptr_key_compare, + .item_compare = &nodeptr_item_compare, +}; + +struct _signal { + atomic_t refs; + struct rb_node node; + struct list_head list; + struct shm_signal *signal; +}; + +struct _connection { + atomic_t refs; + struct rb_node node; + struct list_head signals; + struct vbus_connection *conn; +}; + +static inline void _signal_get(struct _signal *_signal) +{ + atomic_inc(&_signal->refs); +} + +static inline void _signal_put(struct _signal *_signal) +{ + if (atomic_dec_and_test(&_signal->refs)) { + shm_signal_put(_signal->signal); + kfree(_signal); + } +} + +static inline void conn_get(struct _connection *_conn) +{ + atomic_inc(&_conn->refs); +} + +static inline void conn_put(struct _connection *_conn) +{ + if (atomic_dec_and_test(&_conn->refs)) { + struct _signal *_signal, *tmp; + + list_for_each_entry_safe(_signal, tmp, &_conn->signals, + list) { + list_del(&_signal->list); + _signal_put(_signal); + } + + vbus_connection_put(_conn->conn); + kfree(_conn); + } +} + +struct _client { + struct mutex lock; + struct map conn_map; + struct map signal_map; + struct vbus *vbus; + struct vbus_client client; +}; + +struct _connection *to_conn(struct rb_node *node) +{ + return node ? container_of(node, struct _connection, node) : NULL; +} + +static struct _signal *to_signal(struct rb_node *node) +{ + return node ? container_of(node, struct _signal, node) : NULL; +} + +static struct _client *to_client(struct vbus_client *client) +{ + return container_of(client, struct _client, client); +} + +static struct _connection * +connection_find(struct _client *c, unsigned long devid) +{ + struct _connection *_conn; + + /* + * We could, in theory, cast devid to _conn->node, but this would + * be pretty stupid to trust. Therefore, we must validate that + * the pointer is legit by seeing if it exists in our conn_map + */ + + mutex_lock(&c->lock); + + _conn = to_conn(map_find(&c->conn_map, &devid)); + if (likely(_conn)) + conn_get(_conn); + + mutex_unlock(&c->lock); + + return _conn; +} + +static int +_deviceopen(struct vbus_client *client, struct vbus_memctx *ctx, + __u32 devid, __u32 version, __u64 *devh) +{ + struct _client *c = to_client(client); + struct vbus_connection *conn; + struct _connection *_conn; + struct vbus_device_interface *intf = NULL; + int ret; + + /* + * We only get here if the device has never been opened before, + * so we need to create a new connection + */ + ret = vbus_interface_find(c->vbus, devid, &intf); + if (ret < 0) + return ret; + + ret = intf->ops->open(intf, ctx, version, &conn); + kobject_put(&intf->kobj); + if (ret < 0) + return ret; + + _conn = kzalloc(sizeof(*_conn), GFP_KERNEL); + if (!_conn) { + vbus_connection_put(conn); + return -ENOMEM; + } + + atomic_set(&_conn->refs, 1); + _conn->conn = conn; + + INIT_LIST_HEAD(&_conn->signals); + + mutex_lock(&c->lock); + ret = map_add(&c->conn_map, &_conn->node); + mutex_unlock(&c->lock); + + if (ret < 0) { + conn_put(_conn); + return ret; + } + + /* in theory, &_conn->node should be unique */ + *devh = (__u64)&_conn->node; + + return 0; + +} + +/* + * Assumes client->lock is held (or we are releasing and dont need to lock) + */ +static void +conn_del(struct _client *c, struct _connection *_conn) +{ + struct _signal *_signal, *tmp; + + /* Delete and release each opened queue */ + list_for_each_entry_safe(_signal, tmp, &_conn->signals, list) { + map_del(&c->signal_map, &_signal->node); + _signal_put(_signal); + } + + map_del(&c->conn_map, &_conn->node); +} + +static int +_deviceclose(struct vbus_client *client, __u64 devh) +{ + struct _client *c = to_client(client); + struct _connection *_conn; + + mutex_lock(&c->lock); + + _conn = to_conn(map_find(&c->conn_map, &devh)); + if (likely(_conn)) + conn_del(c, _conn); + + mutex_unlock(&c->lock); + + if (unlikely(!_conn)) + return -ENOENT; + + /* this _put is the compliment to the _get performed at _deviceopen */ + conn_put(_conn); + + return 0; +} + +static int +_devicecall(struct vbus_client *client, + __u64 devh, __u32 func, void *data, __u32 len, __u32 flags) +{ + struct _client *c = to_client(client); + struct _connection *_conn; + struct vbus_connection *conn; + int ret; + + _conn = connection_find(c, devh); + if (!_conn) + return -ENOENT; + + conn = _conn->conn; + + ret = conn->ops->call(conn, func, data, len, flags); + + conn_put(_conn); + + return ret; +} + +static int +_deviceshm(struct vbus_client *client, + __u64 devh, + __u32 id, + struct vbus_shm *shm, + struct shm_signal *signal, + __u32 flags, + __u64 *handle) +{ + struct _client *c = to_client(client); + struct _signal *_signal = NULL; + struct _connection *_conn; + struct vbus_connection *conn; + int ret; + + *handle = 0; + + _conn = connection_find(c, devh); + if (!_conn) + return -ENOENT; + + conn = _conn->conn; + + ret = conn->ops->shm(conn, id, shm, signal, flags); + if (ret < 0) { + conn_put(_conn); + return ret; + } + + if (signal) { + _signal = kzalloc(sizeof(*_signal), GFP_KERNEL); + if (!_signal) { + conn_put(_conn); + return -ENOMEM; + } + + /* one for map-ref, one for list-ref */ + atomic_set(&_signal->refs, 2); + _signal->signal = signal; + shm_signal_get(signal); + + mutex_lock(&c->lock); + ret = map_add(&c->signal_map, &_signal->node); + list_add_tail(&_signal->list, &_conn->signals); + mutex_unlock(&c->lock); + + if (!ret) + *handle = (__u64)&_signal->node; + } + + conn_put(_conn); + + return 0; +} + +static int +_shmsignal(struct vbus_client *client, __u64 handle) +{ + struct _client *c = to_client(client); + struct _signal *_signal; + + mutex_lock(&c->lock); + + _signal = to_signal(map_find(&c->signal_map, &handle)); + if (likely(_signal)) + _signal_get(_signal); + + mutex_unlock(&c->lock); + + if (!_signal) + return -ENOENT; + + _shm_signal_wakeup(_signal->signal); + + _signal_put(_signal); + + return 0; +} + +static void +_release(struct vbus_client *client) +{ + struct _client *c = to_client(client); + struct rb_node *node; + + /* Drop all of our open connections */ + while ((node = rb_first(&c->conn_map.root))) { + struct _connection *_conn = to_conn(node); + + conn_del(c, _conn); + conn_put(_conn); + } + + vbus_put(c->vbus); + kfree(c); +} + +struct vbus_client_ops _client_ops = { + .deviceopen = _deviceopen, + .deviceclose = _deviceclose, + .devicecall = _devicecall, + .deviceshm = _deviceshm, + .shmsignal = _shmsignal, + .release = _release, +}; + +struct vbus_client *vbus_client_attach(struct vbus *vbus) +{ + struct _client *c; + + BUG_ON(!vbus); + + c = kzalloc(sizeof(*c), GFP_KERNEL); + if (!c) + return NULL; + + atomic_set(&c->client.refs, 1); + c->client.ops = &_client_ops; + + mutex_init(&c->lock); + map_init(&c->conn_map, &nodeptr_map_ops); + map_init(&c->signal_map, &nodeptr_map_ops); + c->vbus = vbus_get(vbus); + + return &c->client; +} +EXPORT_SYMBOL_GPL(vbus_client_attach); + +/* + * memory context helpers + */ + +static unsigned long +current_memctx_copy_to(struct vbus_memctx *ctx, void *dst, const void *src, + unsigned long len) +{ + return copy_to_user(dst, src, len); +} + +static unsigned long +current_memctx_copy_from(struct vbus_memctx *ctx, void *dst, const void *src, + unsigned long len) +{ + return copy_from_user(dst, src, len); +} + +static void +current_memctx_release(struct vbus_memctx *ctx) +{ + panic("dropped last reference to current_memctx"); +} + +static struct vbus_memctx_ops current_memctx_ops = { + .copy_to = ¤t_memctx_copy_to, + .copy_from = ¤t_memctx_copy_from, + .release = ¤t_memctx_release, +}; + +static struct vbus_memctx _current_memctx = + VBUS_MEMCTX_INIT((¤t_memctx_ops)); + +struct vbus_memctx *current_memctx = &_current_memctx; + +/* + * task_mem allows you to have a copy_from_user/copy_to_user like + * environment, except that it supports copying to tasks other + * than "current" as ctu/cfu() do + */ +struct task_memctx { + struct task_struct *task; + struct vbus_memctx ctx; +}; + +static struct task_memctx *to_task_memctx(struct vbus_memctx *ctx) +{ + return container_of(ctx, struct task_memctx, ctx); +} + +static unsigned long +task_memctx_copy_to(struct vbus_memctx *ctx, void *dst, const void *src, + unsigned long n) +{ + struct task_memctx *tm = to_task_memctx(ctx); + struct task_struct *p = tm->task; + + while (n) { + unsigned long offset = ((unsigned long)dst)%PAGE_SIZE; + unsigned long len = PAGE_SIZE - offset; + int ret; + struct page *pg; + void *maddr; + + if (len > n) + len = n; + + down_read(&p->mm->mmap_sem); + ret = get_user_pages(p, p->mm, + (unsigned long)dst, 1, 1, 0, &pg, NULL); + + if (ret != 1) { + up_read(&p->mm->mmap_sem); + break; + } + + maddr = kmap_atomic(pg, KM_USER0); + memcpy(maddr + offset, src, len); + kunmap_atomic(maddr, KM_USER0); + set_page_dirty_lock(pg); + put_page(pg); + up_read(&p->mm->mmap_sem); + + src += len; + dst += len; + n -= len; + } + + return n; +} + +static unsigned long +task_memctx_copy_from(struct vbus_memctx *ctx, void *dst, const void *src, + unsigned long n) +{ + struct task_memctx *tm = to_task_memctx(ctx); + struct task_struct *p = tm->task; + + while (n) { + unsigned long offset = ((unsigned long)src)%PAGE_SIZE; + unsigned long len = PAGE_SIZE - offset; + int ret; + struct page *pg; + void *maddr; + + if (len > n) + len = n; + + down_read(&p->mm->mmap_sem); + ret = get_user_pages(p, p->mm, + (unsigned long)src, 1, 1, 0, &pg, NULL); + + if (ret != 1) { + up_read(&p->mm->mmap_sem); + break; + } + + maddr = kmap_atomic(pg, KM_USER0); + memcpy(dst, maddr + offset, len); + kunmap_atomic(maddr, KM_USER0); + put_page(pg); + up_read(&p->mm->mmap_sem); + + src += len; + dst += len; + n -= len; + } + + return n; +} + +static void +task_memctx_release(struct vbus_memctx *ctx) +{ + struct task_memctx *tm = to_task_memctx(ctx); + + put_task_struct(tm->task); + kfree(tm); +} + +static struct vbus_memctx_ops task_memctx_ops = { + .copy_to = &task_memctx_copy_to, + .copy_from = &task_memctx_copy_from, + .release = &task_memctx_release, +}; + +struct vbus_memctx *task_memctx_alloc(struct task_struct *task) +{ + struct task_memctx *tm; + + tm = kzalloc(sizeof(*tm), GFP_KERNEL); + if (!tm) + return NULL; + + get_task_struct(task); + + tm->task = task; + vbus_memctx_init(&tm->ctx, &task_memctx_ops); + + return &tm->ctx; +} +EXPORT_SYMBOL_GPL(task_memctx_alloc);