[src/riastradh-drm2]: src/sys/external/bsd/drm2/drm Add a local drm_lock.c, s...

[riastradh <riastradh%NetBSD.org@localhost>]

details:   https://anonhg.NetBSD.org/src/rev/2fab173c6672
branches:  riastradh-drm2
changeset: 788171:2fab173c6672
user:      riastradh <riastradh%NetBSD.org@localhost>
date:      Wed Jul 24 02:33:17 2013 +0000

description:
Add a local drm_lock.c, since I can't make sense of Linux's.

This leaves drm_idlelock_take and drm_idlelock_release unimplemented
until I can figure out the intended semantics.  Linux's code doesn't
guarantee that the lock is taken when drm_idlelock_take returns, and
provides no way to check whether the current lwp/thread/whatever owns
the lock as intended, so there's no way to add diagnostic checks for
the correct use of the API.

diffstat:

 sys/external/bsd/drm2/drm/drm_lock.c |  285 +++++++++++++++++++++++++++++++++++
 1 files changed, 285 insertions(+), 0 deletions(-)

diffs (289 lines):

diff -r 5973cf57e72a -r 2fab173c6672 sys/external/bsd/drm2/drm/drm_lock.c
--- /dev/null   Thu Jan 01 00:00:00 1970 +0000
+++ b/sys/external/bsd/drm2/drm/drm_lock.c      Wed Jul 24 02:33:17 2013 +0000
@@ -0,0 +1,285 @@
+/*     $NetBSD: drm_lock.c,v 1.1.2.1 2013/07/24 02:33:17 riastradh Exp $       */
+
+/*-
+ * Copyright (c) 2013 The NetBSD Foundation, Inc.
+ * All rights reserved.
+ *
+ * This code is derived from software contributed to The NetBSD Foundation
+ * by Taylor R. Campbell.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ *    notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ *    notice, this list of conditions and the following disclaimer in the
+ *    documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
+ * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
+ * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+ * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
+ * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/*
+ * DRM lock.  Each drm master has a lock to provide mutual exclusion
+ * for access to something about the hardware (XXX what?).  The lock
+ * can be held by the kernel or by a drm file (XXX not by a process!
+ * and multiple processes may share a common drm file).  (XXX What
+ * excludes different kthreads?)  The DRM_LOCK ioctl grants the lock to
+ * the userland.  The kernel may need to steal this lock in order to
+ * close a drm file; I believe drm_global_mutex excludes different
+ * kernel threads from doing this simultaneously.
+ *
+ * XXX This description is incoherent and incomplete.
+ */
+
+#include <sys/cdefs.h>
+__KERNEL_RCSID(0, "$NetBSD: drm_lock.c,v 1.1.2.1 2013/07/24 02:33:17 riastradh Exp $");
+
+#include <sys/types.h>
+#include <sys/errno.h>
+#include <sys/systm.h>
+
+#include <drm/drmP.h>
+
+static bool    drm_lock_acquire(struct drm_lock_data *, int);
+static void    drm_lock_release(struct drm_lock_data *, int);
+static int     drm_lock_block_signals(struct drm_device *, struct drm_lock *,
+                   struct drm_file *);
+static void    drm_lock_unblock_signals(struct drm_device *,
+                   struct drm_lock *, struct drm_file *);
+
+/*
+ * Take the lock on behalf of userland.
+ */
+int
+drm_lock(struct drm_device *dev, void *data, struct drm_file *file)
+{
+       struct drm_lock *lock_request = data;
+       struct drm_master *master = file->master;
+       bool acquired;
+       int error;
+
+       KASSERT(mutex_is_locked(&drm_global_mutex));
+
+       /* Refuse to lock on behalf of the kernel.  */
+       if (lock_request->context == DRM_KERNEL_CONTEXT)
+               return -EINVAL;
+
+       /* Count it in the file and device statistics (XXX why here?).  */
+       file->lock_count++;
+       atomic_inc(&dev->counts[_DRM_STAT_LOCKS]);
+
+       /* Wait until the hardware lock is gone or we can acquire it.   */
+       spin_lock(&master->lock.spinlock);
+       DRM_SPIN_WAIT_UNTIL(error, &master->lock.lock_queue,
+           &master->lock.spinlock,
+           ((master->lock.hw_lock == NULL) ||
+               (acquired = drm_lock_acquire(&master->lock,
+                   lock_request->context))));
+       if (error)
+               goto out;
+
+       /* If the lock is gone, give up.  */
+       if (master->lock.hw_lock == NULL) {
+#if 0                          /* XXX Linux sends SIGTERM, but why?  */
+               mutex_enter(proc_lock);
+               psignal(curproc, SIGTERM);
+               mutex_exit(proc_lock);
+               error = -EINTR;
+#else
+               error = -ENXIO;
+#endif
+               goto out;
+       }
+
+       /* Mark the lock as owned by file.  */
+       master->lock.file_priv = file;
+#ifndef __NetBSD__
+       master->lock.lock_time = jiffies; /* XXX Unused?  */
+#endif
+
+       /* Block signals while the lock is held.  */
+       error = drm_lock_block_signals(dev, lock_request, file);
+       if (error)
+               goto fail0;
+
+       /* Enter the DMA quiescent state if requested and available.  */
+       if (ISSET(lock_request->flags, _DRM_LOCK_QUIESCENT) &&
+           (dev->driver->dma_quiescent != NULL)) {
+               error = (*dev->driver->dma_quiescent)(dev);
+               if (error)
+                       goto fail1;
+       }
+
+       /* Success!  */
+       error = 0;
+       goto out;
+
+fail1: drm_lock_unblock_signals(dev, lock_request, file);
+fail0: drm_lock_release(&master->lock, lock_request->context);
+       master->lock.file_priv = NULL;
+out:   spin_unlock(&master->lock.spinlock);
+       return error;
+}
+
+/*
+ * Try to relinquish a lock that userland thinks it holds, per
+ * userland's request.  Fail if it doesn't actually hold the lock.
+ */
+int
+drm_unlock(struct drm_device *dev, void *data, struct drm_file *file)
+{
+       struct drm_lock *lock_request = data;
+       struct drm_master *master = file->master;
+       int error;
+
+       KASSERT(mutex_is_locked(&drm_global_mutex));
+
+       /* Refuse to unlock on behalf of the kernel.  */
+       if (lock_request->context == DRM_KERNEL_CONTEXT)
+               return -EINVAL;
+
+       /* Count it in the device statistics.  */
+       atomic_inc(&dev->counts[_DRM_STAT_UNLOCKS]);
+
+       /* Lock the internal spin lock to make changes.  */
+       spin_lock(&master->lock.spinlock);
+
+       /* Make sure it's actually locked.  */
+       if (!_DRM_LOCK_IS_HELD(master->lock.hw_lock->lock)) {
+               error = -EINVAL;        /* XXX Right error?  */
+               goto out;
+       }
+
+       /* Make sure it's locked in the right context.  */
+       if (_DRM_LOCKING_CONTEXT(master->lock.hw_lock->lock) !=
+           lock_request->context) {
+               error = -EACCES;        /* XXX Right error?  */
+               goto out;
+       }
+
+       /* Make sure it's locked by us.  */
+       if (master->lock.file_priv != file) {
+               error = -EACCES;        /* XXX Right error?  */
+               goto out;
+       }
+
+       /* Actually release the lock.  */
+       drm_lock_release(&master->lock, lock_request->context);
+
+       /* Clear the lock's file pointer, just in case.  */
+       master->lock.file_priv = NULL;
+
+       /* Unblock the signals we blocked in drm_lock.  */
+       drm_lock_unblock_signals(dev, lock_request, file);
+
+       /* Success!  */
+       error = 0;
+
+out:   spin_unlock(&master->lock.spinlock);
+       return error;
+}
+
+/*
+ * Take the lock for the kernel's use.
+ */
+void
+drm_idlelock_take(struct drm_lock_data *lock_data __unused)
+{
+       KASSERT(mutex_is_locked(&drm_global_mutex));
+       panic("drm_idlelock_take is not yet implemented"); /* XXX */
+}
+
+/*
+ * Release the lock from the kernel.
+ */
+void
+drm_idlelock_release(struct drm_lock_data *lock_data __unused)
+{
+       KASSERT(mutex_is_locked(&drm_global_mutex));
+       panic("drm_idlelock_release is not yet implemented"); /* XXX */
+}
+
+/*
+ * Try to acquire the lock.  Return true if successful, false if not.
+ *
+ * Lock's spinlock must be held.
+ */
+static bool
+drm_lock_acquire(struct drm_lock_data *lock_data, int context)
+{
+
+       KASSERT(spin_is_locked(&lock_data->spinlock));
+
+       /* Test and set.  */
+       if (_DRM_LOCK_IS_HELD(lock_data->hw_lock->lock)) {
+               return false;
+       } else {
+               lock_data->hw_lock->lock = (context | _DRM_LOCK_HELD);
+               return true;
+       }
+}
+
+/*
+ * Release the lock held in the given context.  Wake any waiters,
+ * preferring kernel waiters over userland waiters.
+ *
+ * Lock's spinlock must be held and lock must be held in this context.
+ */
+static void
+drm_lock_release(struct drm_lock_data *lock_data, int context)
+{
+
+       (void)context;          /* ignore */
+       KASSERT(spin_is_locked(&lock_data->spinlock));
+       KASSERT(_DRM_LOCK_IS_HELD(lock_data->hw_lock->lock));
+       KASSERT(_DRM_LOCKING_CONTEXT(lock_data->hw_lock->lock) == context);
+
+       /* Are there any kernel waiters?  */
+       if (lock_data->n_kernel_waiters > 0) {
+               /* If there's a kernel waiter, grant it ownership.  */
+               lock_data->hw_lock->lock = (DRM_KERNEL_CONTEXT |
+                   _DRM_LOCK_HELD);
+               lock_data->n_kernel_waiters--;
+               DRM_SPIN_WAKEUP_ONE(&lock_data->kernel_lock_queue,
+                   &lock_data->spinlock);
+       } else {
+               /* Otherwise, free it up and wake someone else.  */
+               lock_data->hw_lock->lock = 0;
+               DRM_SPIN_WAKEUP_ONE(&lock_data->lock_queue,
+                   &lock_data->spinlock);
+       }
+}
+
+/*
+ * Block signals for a process that holds a drm lock.
+ *
+ * XXX It's not processes but files that hold drm locks, so blocking
+ * signals in a process seems wrong, and it's not clear that blocking
+ * signals automatically is remotely sensible anyway.
+ */
+static int
+drm_lock_block_signals(struct drm_device *dev __unused,
+    struct drm_lock *lock_request __unused, struct drm_file *file __unused)
+{
+       return 0;
+}
+
+/*
+ * Unblock the signals that drm_lock_block_signals blocked.
+ */
+static void
+drm_lock_unblock_signals(struct drm_device *dev __unused,
+    struct drm_lock *lock_request __unused, struct drm_file *file __unused)
+{
+}