From patchwork Tue Jun 5 23:45:46 2012 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-Patchwork-Submitter: Erik Gilling X-Patchwork-Id: 9124 Return-Path: X-Original-To: patchwork@peony.canonical.com Delivered-To: patchwork@peony.canonical.com Received: from fiordland.canonical.com (fiordland.canonical.com [91.189.94.145]) by peony.canonical.com (Postfix) with ESMTP id C7A4823E5A for ; Tue, 5 Jun 2012 23:46:18 +0000 (UTC) Received: from mail-yw0-f52.google.com (mail-yw0-f52.google.com [209.85.213.52]) by fiordland.canonical.com (Postfix) with ESMTP id 65E2CA189DF for ; Tue, 5 Jun 2012 23:46:18 +0000 (UTC) Received: by yhpp61 with SMTP id p61so4963687yhp.11 for ; Tue, 05 Jun 2012 16:46:18 -0700 (PDT) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=20120113; h=x-forwarded-to:x-forwarded-for:delivered-to:received-spf:from:to :date:message-id:x-mailer:in-reply-to:references:mime-version :subject:x-beenthere:x-mailman-version:precedence:list-id :list-unsubscribe:list-archive:list-post:list-help:list-subscribe :content-type:content-transfer-encoding:sender:errors-to :x-gm-message-state; bh=q6EqWfYQEH7zavNXuPr/IV6lF2lIlotSo5jyRecdvak=; b=l/7p0ApsC1djlkZ67/6tuBHdW2u3YOas1kLi6o0G44C468RgbavcL/vgn6b5jghvT0 d7Jja2cxagLe+xnH9ccazL6n+Mp0e6GZL68MVsMT0nTvMoD4Mds4qMoWizOdy9bWLjvU 0bflmpLRQGnU9PH3TH+zHqGIrAa8lRD73CtM7k0l2F6go6oqERCqQerc8rnCIxpC7E9y iHpul0NtnzH28VBy+jLyyBxOWQbBEpkiAMk6UVuCxsUHKOqZ0tFQy7663Ua+jjtts0Cv rxiBtdafPYsglXWR2zB2wkm5u8h3GBpOGxRoYqySqrmia8YyWFGwn1KMhGXelesuEOk0 Jyow== Received: by 10.50.160.198 with SMTP id xm6mr4937412igb.0.1338939977387; Tue, 05 Jun 2012 16:46:17 -0700 (PDT) X-Forwarded-To: linaro-patchwork@canonical.com X-Forwarded-For: patch@linaro.org linaro-patchwork@canonical.com Delivered-To: patches@linaro.org Received: by 10.231.24.148 with SMTP id v20csp52387ibb; Tue, 5 Jun 2012 16:46:16 -0700 (PDT) Received: by 10.236.46.132 with SMTP id r4mr13722284yhb.29.1338939974353; Tue, 05 Jun 2012 16:46:14 -0700 (PDT) Received: from mombin.canonical.com (mombin.canonical.com. [91.189.95.16]) by mx.google.com with ESMTP id n15si1292281ano.19.2012.06.05.16.46.12; Tue, 05 Jun 2012 16:46:14 -0700 (PDT) Received-SPF: neutral (google.com: 91.189.95.16 is neither permitted nor denied by best guess record for domain of linaro-mm-sig-bounces@lists.linaro.org) client-ip=91.189.95.16; Authentication-Results: mx.google.com; spf=neutral (google.com: 91.189.95.16 is neither permitted nor denied by best guess record for domain of linaro-mm-sig-bounces@lists.linaro.org) smtp.mail=linaro-mm-sig-bounces@lists.linaro.org Received: from localhost ([127.0.0.1] helo=mombin.canonical.com) by mombin.canonical.com with esmtp (Exim 4.71) (envelope-from ) id 1Sc3Rw-0005hd-TO; Tue, 05 Jun 2012 23:46:09 +0000 Received: from mail-ee0-f74.google.com ([74.125.83.74]) by mombin.canonical.com with esmtp (Exim 4.71) (envelope-from ) id 1Sc3Ru-0005hC-Lg for linaro-mm-sig@lists.linaro.org; Tue, 05 Jun 2012 23:46:06 +0000 Received: by eeit10 with SMTP id t10so364602eei.1 for ; Tue, 05 Jun 2012 16:46:06 -0700 (PDT) Received: by 10.14.94.207 with SMTP id n55mr5595231eef.16.1338939966467; Tue, 05 Jun 2012 16:46:06 -0700 (PDT) Received: by 10.14.94.207 with SMTP id n55mr5595228eef.16.1338939966389; Tue, 05 Jun 2012 16:46:06 -0700 (PDT) Received: from hpza10.eem.corp.google.com ([74.125.121.33]) by gmr-mx.google.com with ESMTPS id b16si7689eeg.3.2012.06.05.16.46.06 (version=TLSv1/SSLv3 cipher=AES128-SHA); Tue, 05 Jun 2012 16:46:06 -0700 (PDT) Received: from dirtysouth.mtv.corp.google.com (dirtysouth.mtv.corp.google.com [172.18.104.93]) by hpza10.eem.corp.google.com (Postfix) with ESMTP id B8ED4200059; Tue, 5 Jun 2012 16:46:05 -0700 (PDT) Received: by dirtysouth.mtv.corp.google.com (Postfix, from userid 94083) id 27D1140DBA; Tue, 5 Jun 2012 16:46:05 -0700 (PDT) From: Erik Gilling To: linaro-mm-sig@lists.linaro.org Date: Tue, 5 Jun 2012 16:45:46 -0700 Message-Id: <1338939956-17312-2-git-send-email-konkers@android.com> X-Mailer: git-send-email 1.7.7.3 In-Reply-To: <1338939956-17312-1-git-send-email-konkers@android.com> References: <1338939956-17312-1-git-send-email-konkers@android.com> MIME-Version: 1.0 Subject: [Linaro-mm-sig] =?utf-8?q?=5BRFC_01/11=5D_sync=3A_Add_synchroniza?= =?utf-8?q?tion_framework?= X-BeenThere: linaro-mm-sig@lists.linaro.org X-Mailman-Version: 2.1.13 Precedence: list List-Id: "Unified memory management interest group." List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: linaro-mm-sig-bounces@lists.linaro.org Errors-To: linaro-mm-sig-bounces@lists.linaro.org X-Gm-Message-State: ALoCoQk8p8u2CSKQsQHNHtbhbezPHHohvgqmA41gtsWoUlkrtWnGvIDq+wggaAJjrkDqdSybOn5C In complicated DMA pipelines such as graphics (multimedia, camera, gpu, display) a consumer of a buffer needs to know when the producer has finished producing it. Likewise the producer needs to know when the consumer is finished with the buffer so it can reuse it. A particular buffer may be consumed by multiple consumers which will retain the buffer for different amounts of time. In addition, a consumer may consume multiple buffers atomically. The sync framework adds an API which allows synchronization between the producers and consumers in a generic way while also allowing platforms which have shared hardware synchronization primitives to exploit them. See Documentation/sync.txt for more information Signed-off-by: Erik Gilling --- Documentation/sync.txt | 75 +++++++ drivers/base/Kconfig | 8 + drivers/base/Makefile | 2 + drivers/base/sync.c | 525 ++++++++++++++++++++++++++++++++++++++++++++++++ include/linux/sync.h | 316 +++++++++++++++++++++++++++++ 5 files changed, 926 insertions(+), 0 deletions(-) create mode 100644 Documentation/sync.txt create mode 100644 drivers/base/sync.c create mode 100644 include/linux/sync.h diff --git a/Documentation/sync.txt b/Documentation/sync.txt new file mode 100644 index 0000000..a2d05e7 --- /dev/null +++ b/Documentation/sync.txt @@ -0,0 +1,75 @@ +Motivation: + +In complicated DMA pipelines such as graphics (multimedia, camera, gpu, display) +a consumer of a buffer needs to know when the producer has finished producing +it. Likewise the producer needs to know when the consumer is finished with the +buffer so it can reuse it. A particular buffer may be consumed by multiple +consumers which will retain the buffer for different amounts of time. In +addition, a consumer may consume multiple buffers atomically. +The sync framework adds an API which allows synchronization between the +producers and consumers in a generic way while also allowing platforms which +have shared hardware synchronization primitives to exploit them. + +Goals: + * provide a generic API for expressing synchronization dependencies + * allow drivers to exploit hardware synchronization between hardware + blocks + * provide a userspace API that allows a compositor to manage + dependencies. + * provide rich telemetry data to allow debugging slowdowns and stalls of + the graphics pipeline. + +Objects: + * sync_timeline + * sync_pt + * sync_fence + +sync_timeline: + +A sync_timeline is an abstract monotonically increasing counter. In general, +each driver/hardware block context will have one of these. They can be backed +by the appropriate hardware or rely on the generic sw_sync implementation. +Timelines are only ever created through their specific implementations +(i.e. sw_sync.) + +sync_pt: + +A sync_pt is an abstract value which marks a point on a sync_timeline. Sync_pts +have a single timeline parent. They have 3 states: active, signaled, and error. +They start in active state and transition, once, to either signaled (when the +timeline counter advances beyond the sync_pt’s value) or error state. + +sync_fence: + +Sync_fences are the primary primitives used by drivers to coordinate +synchronization of their buffers. They are a collection of sync_pts which may +or may not have the same timeline parent. A sync_pt can only exist in one fence +and the fence's list of sync_pts is immutable once created. Fences can be +waited on synchronously or asynchronously. Two fences can also be merged to +create a third fence containing a copy of the two fences’ sync_pts. Fences are +backed by file descriptors to allow userspace to coordinate the display pipeline +dependencies. + +Use: + +A driver implementing sync support should have a work submission function which: + * takes a fence argument specifying when to begin work + * asynchronously queues that work to kick off when the fence is signaled + * returns a fence to indicate when its work will be done. + * signals the returned fence once the work is completed. + +Consider an imaginary display driver that has the following API: +/* + * assumes buf is ready to be displayed. + * blocks until the buffer is on screen. + */ + void display_buffer(struct dma_buf *buf); + +The new API will become: +/* + * will display buf when fence is signaled. + * returns immediately with a fence that will signal when buf + * is no longer displayed. + */ +struct sync_fence* display_buffer(struct dma_buf *buf, + struct sync_fence *fence); diff --git a/drivers/base/Kconfig b/drivers/base/Kconfig index 9aa618a..784d08b 100644 --- a/drivers/base/Kconfig +++ b/drivers/base/Kconfig @@ -192,4 +192,12 @@ config DMA_SHARED_BUFFER APIs extension; the file's descriptor can then be passed on to other driver. +config SYNC + bool "Synchronization framework" + default n + select ANON_INODES + help + This option enables the framework for synchronization between multiple + drivers. Sync implementations can take advantage of hardware + synchronization built into devices like GPUs. endmenu diff --git a/drivers/base/Makefile b/drivers/base/Makefile index b6d1b9c..a43e369 100644 --- a/drivers/base/Makefile +++ b/drivers/base/Makefile @@ -21,5 +21,7 @@ obj-$(CONFIG_SYS_HYPERVISOR) += hypervisor.o obj-$(CONFIG_REGMAP) += regmap/ obj-$(CONFIG_SOC_BUS) += soc.o +obj-$(CONFIG_SYNC) += sync.o + ccflags-$(CONFIG_DEBUG_DRIVER) := -DDEBUG diff --git a/drivers/base/sync.c b/drivers/base/sync.c new file mode 100644 index 0000000..4a8604b --- /dev/null +++ b/drivers/base/sync.c @@ -0,0 +1,525 @@ +/* + * drivers/base/sync.c + * + * Copyright (C) 2012 Google, Inc. + * + * This software is licensed under the terms of the GNU General Public + * License version 2, as published by the Free Software Foundation, and + * may be copied, distributed, and modified under those terms. + * + * 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. + * + */ + +#include +#include +#include +#include +#include +#include +#include + +#include + +static void sync_fence_signal_pt(struct sync_pt *pt); +static int _sync_pt_has_signaled(struct sync_pt *pt); + +struct sync_timeline *sync_timeline_create(const struct sync_timeline_ops *ops, + int size, const char *name) +{ + struct sync_timeline *obj; + + if (size < sizeof(struct sync_timeline)) + return NULL; + + obj = kzalloc(size, GFP_KERNEL); + if (obj == NULL) + return NULL; + + obj->ops = ops; + strlcpy(obj->name, name, sizeof(obj->name)); + + INIT_LIST_HEAD(&obj->child_list_head); + spin_lock_init(&obj->child_list_lock); + + INIT_LIST_HEAD(&obj->active_list_head); + spin_lock_init(&obj->active_list_lock); + + return obj; +} + +void sync_timeline_destroy(struct sync_timeline *obj) +{ + unsigned long flags; + bool needs_freeing; + + spin_lock_irqsave(&obj->child_list_lock, flags); + obj->destroyed = true; + needs_freeing = list_empty(&obj->child_list_head); + spin_unlock_irqrestore(&obj->child_list_lock, flags); + + if (needs_freeing) + kfree(obj); + else + sync_timeline_signal(obj); +} + +static void sync_timeline_add_pt(struct sync_timeline *obj, struct sync_pt *pt) +{ + unsigned long flags; + + pt->parent = obj; + + spin_lock_irqsave(&obj->child_list_lock, flags); + list_add_tail(&pt->child_list, &obj->child_list_head); + spin_unlock_irqrestore(&obj->child_list_lock, flags); +} + +static void sync_timeline_remove_pt(struct sync_pt *pt) +{ + struct sync_timeline *obj = pt->parent; + unsigned long flags; + bool needs_freeing; + + spin_lock_irqsave(&obj->active_list_lock, flags); + if (!list_empty(&pt->active_list)) + list_del_init(&pt->active_list); + spin_unlock_irqrestore(&obj->active_list_lock, flags); + + spin_lock_irqsave(&obj->child_list_lock, flags); + list_del(&pt->child_list); + needs_freeing = obj->destroyed && list_empty(&obj->child_list_head); + spin_unlock_irqrestore(&obj->child_list_lock, flags); + + if (needs_freeing) + kfree(obj); +} + +void sync_timeline_signal(struct sync_timeline *obj) +{ + unsigned long flags; + LIST_HEAD(signaled_pts); + struct list_head *pos, *n; + + spin_lock_irqsave(&obj->active_list_lock, flags); + + list_for_each_safe(pos, n, &obj->active_list_head) { + struct sync_pt *pt = + container_of(pos, struct sync_pt, active_list); + + if (_sync_pt_has_signaled(pt)) + list_move(pos, &signaled_pts); + } + + spin_unlock_irqrestore(&obj->active_list_lock, flags); + + list_for_each_safe(pos, n, &signaled_pts) { + struct sync_pt *pt = + container_of(pos, struct sync_pt, active_list); + + list_del_init(pos); + sync_fence_signal_pt(pt); + } +} + +struct sync_pt *sync_pt_create(struct sync_timeline *parent, int size) +{ + struct sync_pt *pt; + + if (size < sizeof(struct sync_pt)) + return NULL; + + pt = kzalloc(size, GFP_KERNEL); + if (pt == NULL) + return NULL; + + INIT_LIST_HEAD(&pt->active_list); + sync_timeline_add_pt(parent, pt); + + return pt; +} + +void sync_pt_free(struct sync_pt *pt) +{ + if (pt->parent->ops->free_pt) + pt->parent->ops->free_pt(pt); + + sync_timeline_remove_pt(pt); + + kfree(pt); +} + +/* call with pt->parent->active_list_lock held */ +static int _sync_pt_has_signaled(struct sync_pt *pt) +{ + if (!pt->status) + pt->status = pt->parent->ops->has_signaled(pt); + + if (!pt->status && pt->parent->destroyed) + pt->status = -ENOENT; + + return pt->status; +} + +static struct sync_pt *sync_pt_dup(struct sync_pt *pt) +{ + return pt->parent->ops->dup(pt); +} + +/* Adds a sync pt to the active queue. Called when added to a fence */ +static void sync_pt_activate(struct sync_pt *pt) +{ + struct sync_timeline *obj = pt->parent; + unsigned long flags; + int err; + + spin_lock_irqsave(&obj->active_list_lock, flags); + + err = _sync_pt_has_signaled(pt); + if (err != 0) + goto out; + + list_add_tail(&pt->active_list, &obj->active_list_head); + +out: + spin_unlock_irqrestore(&obj->active_list_lock, flags); +} + +static int sync_fence_release(struct inode *inode, struct file *file); +static long sync_fence_ioctl(struct file *file, unsigned int cmd, + unsigned long arg); + + +static const struct file_operations sync_fence_fops = { + .release = sync_fence_release, + .unlocked_ioctl = sync_fence_ioctl, +}; + +static struct sync_fence *sync_fence_alloc(const char *name) +{ + struct sync_fence *fence; + + fence = kzalloc(sizeof(struct sync_fence), GFP_KERNEL); + if (fence == NULL) + return NULL; + + fence->file = anon_inode_getfile("sync_fence", &sync_fence_fops, + fence, 0); + if (fence->file == NULL) + goto err; + + strlcpy(fence->name, name, sizeof(fence->name)); + + INIT_LIST_HEAD(&fence->pt_list_head); + INIT_LIST_HEAD(&fence->waiter_list_head); + spin_lock_init(&fence->waiter_list_lock); + + init_waitqueue_head(&fence->wq); + return fence; + +err: + kfree(fence); + return NULL; +} + +/* TODO: implement a create which takes more that one sync_pt */ +struct sync_fence *sync_fence_create(const char *name, struct sync_pt *pt) +{ + struct sync_fence *fence; + + if (pt->fence) + return NULL; + + fence = sync_fence_alloc(name); + if (fence == NULL) + return NULL; + + pt->fence = fence; + list_add(&pt->pt_list, &fence->pt_list_head); + sync_pt_activate(pt); + + return fence; +} + +static int sync_fence_copy_pts(struct sync_fence *dst, struct sync_fence *src) +{ + struct list_head *pos; + + list_for_each(pos, &src->pt_list_head) { + struct sync_pt *orig_pt = + container_of(pos, struct sync_pt, pt_list); + struct sync_pt *new_pt = sync_pt_dup(orig_pt); + + if (new_pt == NULL) + return -ENOMEM; + + new_pt->fence = dst; + list_add(&new_pt->pt_list, &dst->pt_list_head); + sync_pt_activate(new_pt); + } + + return 0; +} + +static void sync_fence_free_pts(struct sync_fence *fence) +{ + struct list_head *pos, *n; + + list_for_each_safe(pos, n, &fence->pt_list_head) { + struct sync_pt *pt = container_of(pos, struct sync_pt, pt_list); + sync_pt_free(pt); + } +} + +struct sync_fence *sync_fence_fdget(int fd) +{ + struct file *file = fget(fd); + + if (file == NULL) + return NULL; + + if (file->f_op != &sync_fence_fops) + goto err; + + return file->private_data; + +err: + fput(file); + return NULL; +} + +void sync_fence_put(struct sync_fence *fence) +{ + fput(fence->file); +} + +void sync_fence_install(struct sync_fence *fence, int fd) +{ + fd_install(fd, fence->file); +} + +static int sync_fence_get_status(struct sync_fence *fence) +{ + struct list_head *pos; + int status = 1; + + list_for_each(pos, &fence->pt_list_head) { + struct sync_pt *pt = container_of(pos, struct sync_pt, pt_list); + int pt_status = pt->status; + + if (pt_status < 0) { + status = pt_status; + break; + } else if (status == 1) { + status = pt_status; + } + } + + return status; +} + +struct sync_fence *sync_fence_merge(const char *name, + struct sync_fence *a, struct sync_fence *b) +{ + struct sync_fence *fence; + int err; + + fence = sync_fence_alloc(name); + if (fence == NULL) + return NULL; + + err = sync_fence_copy_pts(fence, a); + if (err < 0) + goto err; + + err = sync_fence_copy_pts(fence, b); + if (err < 0) + goto err; + + fence->status = sync_fence_get_status(fence); + + return fence; +err: + sync_fence_free_pts(fence); + kfree(fence); + return NULL; +} + +static void sync_fence_signal_pt(struct sync_pt *pt) +{ + LIST_HEAD(signaled_waiters); + struct sync_fence *fence = pt->fence; + struct list_head *pos; + struct list_head *n; + unsigned long flags; + int status; + + status = sync_fence_get_status(fence); + + spin_lock_irqsave(&fence->waiter_list_lock, flags); + /* + * this should protect against two threads racing on the signaled + * false -> true transition + */ + if (status && !fence->status) { + list_for_each_safe(pos, n, &fence->waiter_list_head) + list_move(pos, &signaled_waiters); + + fence->status = status; + } else { + status = 0; + } + spin_unlock_irqrestore(&fence->waiter_list_lock, flags); + + if (status) { + list_for_each_safe(pos, n, &signaled_waiters) { + struct sync_fence_waiter *waiter = + container_of(pos, struct sync_fence_waiter, + waiter_list); + + waiter->callback(fence, waiter->callback_data); + list_del(pos); + kfree(waiter); + } + wake_up(&fence->wq); + } +} + +int sync_fence_wait_async(struct sync_fence *fence, + void (*callback)(struct sync_fence *, void *data), + void *callback_data) +{ + struct sync_fence_waiter *waiter; + unsigned long flags; + int err = 0; + + waiter = kzalloc(sizeof(struct sync_fence_waiter), GFP_KERNEL); + if (waiter == NULL) + return -ENOMEM; + + waiter->callback = callback; + waiter->callback_data = callback_data; + + spin_lock_irqsave(&fence->waiter_list_lock, flags); + + if (fence->status) { + kfree(waiter); + err = fence->status; + goto out; + } + + list_add_tail(&waiter->waiter_list, &fence->waiter_list_head); +out: + spin_unlock_irqrestore(&fence->waiter_list_lock, flags); + + return err; +} + +int sync_fence_wait(struct sync_fence *fence, long timeout) +{ + int err; + + if (timeout) { + timeout = msecs_to_jiffies(timeout); + err = wait_event_interruptible_timeout(fence->wq, + fence->status != 0, + timeout); + } else { + err = wait_event_interruptible(fence->wq, fence->status != 0); + } + + if (err < 0) + return err; + + if (fence->status < 0) + return fence->status; + + if (fence->status == 0) + return -ETIME; + + return 0; +} + +static int sync_fence_release(struct inode *inode, struct file *file) +{ + struct sync_fence *fence = file->private_data; + + sync_fence_free_pts(fence); + kfree(fence); + + return 0; +} + +static long sync_fence_ioctl_wait(struct sync_fence *fence, unsigned long arg) +{ + __u32 value; + + if (copy_from_user(&value, (void __user *)arg, sizeof(value))) + return -EFAULT; + + return sync_fence_wait(fence, value); +} + +static long sync_fence_ioctl_merge(struct sync_fence *fence, unsigned long arg) +{ + int fd = get_unused_fd(); + int err; + struct sync_fence *fence2, *fence3; + struct sync_merge_data data; + + if (copy_from_user(&data, (void __user *)arg, sizeof(data))) + return -EFAULT; + + fence2 = sync_fence_fdget(data.fd2); + if (fence2 == NULL) { + err = -ENOENT; + goto err_put_fd; + } + + data.name[sizeof(data.name) - 1] = '\0'; + fence3 = sync_fence_merge(data.name, fence, fence2); + if (fence3 == NULL) { + err = -ENOMEM; + goto err_put_fence2; + } + + data.fence = fd; + if (copy_to_user((void __user *)arg, &data, sizeof(data))) { + err = -EFAULT; + goto err_put_fence3; + } + + sync_fence_install(fence3, fd); + sync_fence_put(fence2); + return 0; + +err_put_fence3: + sync_fence_put(fence3); + +err_put_fence2: + sync_fence_put(fence2); + +err_put_fd: + put_unused_fd(fd); + return err; +} + + +static long sync_fence_ioctl(struct file *file, unsigned int cmd, + unsigned long arg) +{ + struct sync_fence *fence = file->private_data; + switch (cmd) { + case SYNC_IOC_WAIT: + return sync_fence_ioctl_wait(fence, arg); + + case SYNC_IOC_MERGE: + return sync_fence_ioctl_merge(fence, arg); + default: + return -ENOTTY; + } +} + diff --git a/include/linux/sync.h b/include/linux/sync.h new file mode 100644 index 0000000..2e7ee55 --- /dev/null +++ b/include/linux/sync.h @@ -0,0 +1,316 @@ +/* + * include/linux/sync.h + * + * Copyright (C) 2012 Google, Inc. + * + * 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. + * + */ + +#ifndef _LINUX_SYNC_H +#define _LINUX_SYNC_H + +#include +#ifdef __KERNEL__ + +#include +#include +#include + +struct sync_timeline; +struct sync_pt; +struct sync_fence; + +/** + * struct sync_timeline_ops - sync object implementation ops + * @driver_name: name of the implentation + * @dup: duplicate a sync_pt + * @has_signaled: returns: + * 1 if pt has signaled + * 0 if pt has not signaled + * <0 on error + * must be callable from atomic context + * @compare: returns: + * 1 if b will signal before a + * 0 if a and b will signal at the same time + * -1 if a will signabl before b + * must be callable from atomic context + * @free_pt: called before sync_pt is freed + * @release_obj: called before sync_timeline is freed + */ +struct sync_timeline_ops { + const char *driver_name; + + /* required */ + struct sync_pt *(*dup)(struct sync_pt *pt); + + /* required */ + int (*has_signaled)(struct sync_pt *pt); + + /* required */ + int (*compare)(struct sync_pt *a, struct sync_pt *b); + + /* optional */ + void (*free_pt)(struct sync_pt *sync_pt); + + /* optional */ + void (*release_obj)(struct sync_timeline *sync_timeline); +}; + +/** + * struct sync_timeline - sync object + * @ops: ops that define the implementaiton of the sync_timeline + * @name: name of the sync_timeline. Useful for debugging + * @destoryed: set when sync_timeline is destroyed + * @child_list_head: list of children sync_pts for this sync_timeline + * @child_list_lock: lock protecting @child_list_head, destroyed, and + * sync_pt.status + * @active_list_head: list of active (unsignaled/errored) sync_pts + */ +struct sync_timeline { + const struct sync_timeline_ops *ops; + char name[32]; + + /* protected by child_list_lock */ + bool destroyed; + + struct list_head child_list_head; + spinlock_t child_list_lock; + + struct list_head active_list_head; + spinlock_t active_list_lock; +}; + +/** + * struct sync_pt - sync point + * @parent: sync_timeline to which this sync_pt belongs + * @child_list: membership in sync_timeline.child_list_head + * @active_list: membership in sync_timeline.active_list_head + * @fence: sync_fence to which the sync_pt belongs + * @pt_list: membership in sync_fence.pt_list_head + * @status: 1: signaled, 0:active, <0: error + */ +struct sync_pt { + struct sync_timeline *parent; + struct list_head child_list; + + struct list_head active_list; + + struct sync_fence *fence; + struct list_head pt_list; + + /* protected by parent->active_list_lock */ + int status; +}; + +/** + * struct sync_fence - sync fence + * @file: file representing this fence + * @name: name of sync_fence. Useful for debugging + * @pt_list_head: list of sync_pts in ths fence. immutable once fence + * is created + * @waiter_list_head: list of asynchronous waiters on this fence + * @waiter_list_lock: lock protecting @waiter_list_head and @status + * @status: 1: signaled, 0:active, <0: error + * + * @wq: wait queue for fence signaling + */ +struct sync_fence { + struct file *file; + char name[32]; + + /* this list is immutable once the fence is created */ + struct list_head pt_list_head; + + struct list_head waiter_list_head; + spinlock_t waiter_list_lock; /* also protects status */ + int status; + + wait_queue_head_t wq; +}; + +/** + * struct sync_fence_waiter - metadata for asynchronous waiter on a fence + * @waiter_list: membership in sync_fence.waiter_list_head + * @callback: function pointer to call when fence signals + * @callback_data: pointer to pass to @callback + */ +struct sync_fence_waiter { + struct list_head waiter_list; + + void (*callback)(struct sync_fence *fence, void *data); + void *callback_data; +}; + +/* + * API for sync_timeline implementers + */ + +/** + * sync_timeline_create() - creates a sync object + * @ops: specifies the implemention ops for the object + * @size: size to allocate for this obj + * @name: sync_timeline name + * + * Creates a new sync_timeline which will use the implemetation specified by + * @ops. @size bytes will be allocated allowing for implemntation specific + * data to be kept after the generic sync_timeline stuct. + */ +struct sync_timeline *sync_timeline_create(const struct sync_timeline_ops *ops, + int size, const char *name); + +/** + * sync_timeline_destory() - destorys a sync object + * @obj: sync_timeline to destroy + * + * A sync implemntation should call this when the @obj is going away + * (i.e. module unload.) @obj won't actually be freed until all its childern + * sync_pts are freed. + */ +void sync_timeline_destroy(struct sync_timeline *obj); + +/** + * sync_timeline_signal() - signal a status change on a sync_timeline + * @obj: sync_timeline to signal + * + * A sync implemntation should call this any time one of it's sync_pts + * has signaled or has an error condition. + */ +void sync_timeline_signal(struct sync_timeline *obj); + +/** + * sync_pt_create() - creates a sync pt + * @parent: sync_pt's parent sync_timeline + * @size: size to allocate for this pt + * + * Creates a new sync_pt as a chiled of @parent. @size bytes will be + * allocated allowing for implemntation specific data to be kept after + * the generic sync_timeline struct. + */ +struct sync_pt *sync_pt_create(struct sync_timeline *parent, int size); + +/** + * sync_pt_free() - frees a sync pt + * @pt: sync_pt to free + * + * This should only be called on sync_pts which have been created but + * not added to a fence. + */ +void sync_pt_free(struct sync_pt *pt); + +/** + * sync_fence_create() - creates a sync fence + * @name: name of fence to create + * @pt: sync_pt to add to the fence + * + * Creates a fence containg @pt. Once this is called, the fence takes + * ownership of @pt. + */ +struct sync_fence *sync_fence_create(const char *name, struct sync_pt *pt); + +/* + * API for sync_fence consumers + */ + +/** + * sync_fence_merge() - merge two fences + * @name: name of new fence + * @a: fence a + * @b: fence b + * + * Creates a new fence which contains copies of all the sync_pts in both + * @a and @b. @a and @b remain valid, independent fences. + */ +struct sync_fence *sync_fence_merge(const char *name, + struct sync_fence *a, struct sync_fence *b); + +/** + * sync_fence_fdget() - get a fence from an fd + * @fd: fd referencing a fence + * + * Ensures @fd references a valid fence, increments the refcount of the backing + * file, and returns the fence. + */ +struct sync_fence *sync_fence_fdget(int fd); + +/** + * sync_fence_put() - puts a refernnce of a sync fence + * @fence: fence to put + * + * Puts a reference on @fence. If this is the last reference, the fence and + * all it's sync_pts will be freed + */ +void sync_fence_put(struct sync_fence *fence); + +/** + * sync_fence_install() - installs a fence into a file descriptor + * @fence: fence to instal + * @fd: file descriptor in which to install the fence + * + * Installs @fence into @fd. @fd's should be acquired through get_unused_fd(). + */ +void sync_fence_install(struct sync_fence *fence, int fd); + +/** + * sync_fence_wait_async() - registers and async wait on the fence + * @fence: fence to wait on + * @callback: callback + * @callback_data data to pass to the callback + * + * Returns 1 if @fence has already signaled. + * + * Registers a callback to be called when @fence signals or has an error + */ +int sync_fence_wait_async(struct sync_fence *fence, + void (*callback)(struct sync_fence *, void *data), + void *callback_data); + +/** + * sync_fence_wait() - wait on fence + * @fence: fence to wait on + * @tiemout: timeout in ms + * + * Wait for @fence to be signaled or have an error. Waits indefintly + * if @timeout = 0 + */ +int sync_fence_wait(struct sync_fence *fence, long timeout); + +/* useful for sync driver's debug print handlers */ +const char *sync_status_str(int status); + +#endif /* __KERNEL__ */ + +/** + * struct sync_merge_data - data passed to merge ioctl + * @fd2: file descriptor of second fence + * @name: name of new fence + * @fence: returns the fd of the new fence to userspace + */ +struct sync_merge_data { + __s32 fd2; /* fd of second fence */ + char name[32]; /* name of new fence */ + __s32 fence; /* fd on newly created fence */ +}; + +#define SYNC_IOC_MAGIC '>' + +/** + * DOC: SYNC_IOC_WAIT - wait for a fence to signal + * + * pass timeout in milliseconds. + */ +#define SYNC_IOC_WAIT _IOW(SYNC_IOC_MAGIC, 0, __u32) + +/** + * DOC: SYNC_IOC_MERGE - merge two fences + * + * Takes a struct sync_merge_data. Creates a new fence containing copies of + * the sync_pts in both the calling fd and sync_merge_data.fd2. Returns the + * new fence's fd in sync_merge_data.fence + */ +#define SYNC_IOC_MERGE _IOWR(SYNC_IOC_MAGIC, 1, struct sync_merge_data) + +#endif /* _LINUX_SYNC_H */