i830_dma.c

linux-2.6.15.6

/* i830_dma.c -- DMA support for the I830 -*- linux-c -*-
 * Created: Mon Dec 13 01:50:01 1999 by [e-mail]
 *
 * Copyright 1999 Precision Insight, Inc., Cedar Park, Texas.
 * Copyright 2000 VA Linux Systems, Inc., Sunnyvale, California.
 * All Rights Reserved.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice (including the next
 * paragraph) shall be included in all copies or substantial portions of the
 * Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
 * DEALINGS IN THE SOFTWARE.
 *
 * Authors: Rickard E. (Rik) Faith <[e-mail]>
 *	    Jeff Hartmann <[e-mail]>
 *	    Keith Whitwell <[e-mail]>
 *	    Abraham vd Merwe <[e-mail]>
 *
 */

#include "drmP.h"
#include "drm.h"
#include "i830_drm.h"
#include "i830_drv.h"
#include <linux/interrupt.h>	/* For task queue support */
#include <linux/pagemap.h>	/* For FASTCALL on unlock_page() */
#include <linux/delay.h>
#include <asm/uaccess.h>

#define I830_BUF_FREE		2
#define I830_BUF_CLIENT		1
#define I830_BUF_HARDWARE      	0

#define I830_BUF_UNMAPPED 0
#define I830_BUF_MAPPED   1

static drm_buf_t *i830_freelist_get(drm_device_t * dev)
{
	drm_device_dma_t *dma = dev->dma;
	int i;
	int used;

	/* Linear search might not be the best solution */

	for (i = 0; i < dma->buf_count; i++) {
		drm_buf_t *buf = dma->buflist[i];
		drm_i830_buf_priv_t *buf_priv = buf->dev_private;
		/* In use is already a pointer */
		used = cmpxchg(buf_priv->in_use, I830_BUF_FREE,
			       I830_BUF_CLIENT);
		if (used == I830_BUF_FREE) {
			return buf;
		}
	}
	return NULL;
}

/* This should only be called if the buffer is not sent to the hardware
 * yet, the hardware updates in use for us once its on the ring buffer.
 */

static int i830_freelist_put(drm_device_t * dev, drm_buf_t * buf)
{
	drm_i830_buf_priv_t *buf_priv = buf->dev_private;
	int used;

	/* In use is already a pointer */
	used = cmpxchg(buf_priv->in_use, I830_BUF_CLIENT, I830_BUF_FREE);
	if (used != I830_BUF_CLIENT) {
		DRM_ERROR("Freeing buffer thats not in use : %d\n", buf->idx);
		return -EINVAL;
	}

	return 0;
}

static int i830_mmap_buffers(struct file *filp, struct vm_area_struct *vma)
{
	drm_file_t *priv = filp->private_data;
	drm_device_t *dev;
	drm_i830_private_t *dev_priv;
	drm_buf_t *buf;
	drm_i830_buf_priv_t *buf_priv;

	lock_kernel();
	dev = priv->head->dev;
	dev_priv = dev->dev_private;
	buf = dev_priv->mmap_buffer;
	buf_priv = buf->dev_private;

	vma->vm_flags |= (VM_IO | VM_DONTCOPY);
	vma->vm_file = filp;

	buf_priv->currently_mapped = I830_BUF_MAPPED;
	unlock_kernel();

	if (io_remap_pfn_range(vma, vma->vm_start,
			       VM_OFFSET(vma) >> PAGE_SHIFT,
			       vma->vm_end - vma->vm_start, vma->vm_page_prot))
		return -EAGAIN;
	return 0;
}

static struct file_operations i830_buffer_fops = {
	.open = drm_open,
	.flush = drm_flush,
	.release = drm_release,
	.ioctl = drm_ioctl,
	.mmap = i830_mmap_buffers,
	.fasync = drm_fasync,
};

static int i830_map_buffer(drm_buf_t * buf, struct file *filp)
{
	drm_file_t *priv = filp->private_data;
	drm_device_t *dev = priv->head->dev;
	drm_i830_buf_priv_t *buf_priv = buf->dev_private;
	drm_i830_private_t *dev_priv = dev->dev_private;
	struct file_operations *old_fops;
	unsigned long virtual;
	int retcode = 0;

	if (buf_priv->currently_mapped == I830_BUF_MAPPED)
		return -EINVAL;

	down_write(&current->mm->mmap_sem);
	old_fops = filp->f_op;
	filp->f_op = &i830_buffer_fops;
	dev_priv->mmap_buffer = buf;
	virtual = do_mmap(filp, 0, buf->total, PROT_READ | PROT_WRITE,
			  MAP_SHARED, buf->bus_address);
	dev_priv->mmap_buffer = NULL;
	filp->f_op = old_fops;
	if (IS_ERR((void *)virtual)) {	/* ugh */
		/* Real error */
		DRM_ERROR("mmap error\n");
		retcode = virtual;
		buf_priv->virtual = NULL;
	} else {
		buf_priv->virtual = (void __user *)virtual;
	}
	up_write(&current->mm->mmap_sem);

	return retcode;
}

static int i830_unmap_buffer(drm_buf_t * buf)
{
	drm_i830_buf_priv_t *buf_priv = buf->dev_private;
	int retcode = 0;

	if (buf_priv->currently_mapped != I830_BUF_MAPPED)
		return -EINVAL;

	down_write(&current->mm->mmap_sem);
	retcode = do_munmap(current->mm,
			    (unsigned long)buf_priv->virtual,
			    (size_t) buf->total);
	up_write(&current->mm->mmap_sem);

	buf_priv->currently_mapped = I830_BUF_UNMAPPED;
	buf_priv->virtual = NULL;

	return retcode;
}

static int i830_dma_get_buffer(drm_device_t * dev, drm_i830_dma_t * d,
			       struct file *filp)
{
	drm_buf_t *buf;
	drm_i830_buf_priv_t *buf_priv;
	int retcode = 0;

	buf = i830_freelist_get(dev);
	if (!buf) {
		retcode = -ENOMEM;
		DRM_DEBUG("retcode=%d\n", retcode);
		return retcode;
	}

	retcode = i830_map_buffer(buf, filp);
	if (retcode) {
		i830_freelist_put(dev, buf);
		DRM_ERROR("mapbuf failed, retcode %d\n", retcode);
		return retcode;
	}
	buf->filp = filp;
	buf_priv = buf->dev_private;
	d->granted = 1;
	d->request_idx = buf->idx;
	d->request_size = buf->total;
	d->virtual = buf_priv->virtual;

	return retcode;
}

static int i830_dma_cleanup(drm_device_t * dev)
{
	drm_device_dma_t *dma = dev->dma;

	/* Make sure interrupts are disabled here because the uninstall ioctl
	 * may not have been called from userspace and after dev_private
	 * is freed, it's too late.
	 */
	if (dev->irq_enabled)
		drm_irq_uninstall(dev);

	if (dev->dev_private) {
		int i;
		drm_i830_private_t *dev_priv =
		    (drm_i830_private_t *) dev->dev_private;

		if (dev_priv->ring.virtual_start) {
			drm_ioremapfree((void *)dev_priv->ring.virtual_start,
					dev_priv->ring.Size, dev);
		}
		if (dev_priv->hw_status_page) {
			pci_free_consistent(dev->pdev, PAGE_SIZE,
					    dev_priv->hw_status_page,
					    dev_priv->dma_status_page);
			/* Need to rewrite hardware status page */
			I830_WRITE(0x02080, 0x1ffff000);
		}

		drm_free(dev->dev_private, sizeof(drm_i830_private_t),
			 DRM_MEM_DRIVER);
		dev->dev_private = NULL;

		for (i = 0; i < dma->buf_count; i++) {
			drm_buf_t *buf = dma->buflist[i];
			drm_i830_buf_priv_t *buf_priv = buf->dev_private;
			if (buf_priv->kernel_virtual && buf->total)
				drm_ioremapfree(buf_priv->kernel_virtual,
						buf->total, dev);
		}
	}
	return 0;
}

int i830_wait_ring(drm_device_t * dev, int n, const char *caller)
{
	drm_i830_private_t *dev_priv = dev->dev_private;
	drm_i830_ring_buffer_t *ring = &(dev_priv->ring);
	int iters = 0;
	unsigned long end;
	unsigned int last_head = I830_READ(LP_RING + RING_HEAD) & HEAD_ADDR;

	end = jiffies + (HZ * 3);
	while (ring->space < n) {
		ring->head = I830_READ(LP_RING + RING_HEAD) & HEAD_ADDR;
		ring->space = ring->head - (ring->tail + 8);
		if (ring->space < 0)
			ring->space += ring->Size;

		if (ring->head != last_head) {
			end = jiffies + (HZ * 3);
			last_head = ring->head;
		}

		iters++;
		if (time_before(end, jiffies)) {
			DRM_ERROR("space: %d wanted %d\n", ring->space, n);
			DRM_ERROR("lockup\n");
			goto out_wait_ring;
		}
		udelay(1);
		dev_priv->sarea_priv->perf_boxes |= I830_BOX_WAIT;
	}

      out_wait_ring:
	return iters;
}

static void i830_kernel_lost_context(drm_device_t * dev)
{
	drm_i830_private_t *dev_priv = dev->dev_private;
	drm_i830_ring_buffer_t *ring = &(dev_priv->ring);

	ring->head = I830_READ(LP_RING + RING_HEAD) & HEAD_ADDR;
	ring->tail = I830_READ(LP_RING + RING_TAIL) & TAIL_ADDR;
	ring->space = ring->head - (ring->tail + 8);
	if (ring->space < 0)
		ring->space += ring->Size;

	if (ring->head == ring->tail)
		dev_priv->sarea_priv->perf_boxes |= I830_BOX_RING_EMPTY;
}

static int i830_freelist_init(drm_device_t * dev, drm_i830_private_t * dev_priv)
{
	drm_device_dma_t *dma = dev->dma;
	int my_idx = 36;
	u32 *hw_status = (u32 *) (dev_priv->hw_status_page + my_idx);
	int i;

	if (dma->buf_count > 1019) {
		/* Not enough space in the status page for the freelist */
		return -EINVAL;
	}

	for (i = 0; i < dma->buf_count; i++) {
		drm_buf_t *buf = dma->buflist[i];
		drm_i830_buf_priv_t *buf_priv = buf->dev_private;

		buf_priv->in_use = hw_status++;
		buf_priv->my_use_idx = my_idx;
		my_idx += 4;

		*buf_priv->in_use = I830_BUF_FREE;

		buf_priv->kernel_virtual = drm_ioremap(buf->bus_address,
						       buf->total, dev);
	}
	return 0;
}

static int i830_dma_initialize(drm_device_t * dev,
			       drm_i830_private_t * dev_priv,
			       drm_i830_init_t * init)
{
	struct list_head *list;

	memset(dev_priv, 0, sizeof(drm_i830_private_t));

	list_for_each(list, &dev->maplist->head) {
		drm_map_list_t *r_list = list_entry(list, drm_map_list_t, head);
		if (r_list->map &&
		    r_list->map->type == _DRM_SHM &&
		    r_list->map->flags & _DRM_CONTAINS_LOCK) {
			dev_priv->sarea_map = r_list->map;
			break;
		}
	}

	if (!dev_priv->sarea_map) {
		dev->dev_private = (void *)dev_priv;
		i830_dma_cleanup(dev);
		DRM_ERROR("can not find sarea!\n");
		return -EINVAL;
	}
	dev_priv->mmio_map = drm_core_findmap(dev, init->mmio_offset);
	if (!dev_priv->mmio_map) {
		dev->dev_private = (void *)dev_priv;
		i830_dma_cleanup(dev);
		DRM_ERROR("can not find mmio map!\n");
		return -EINVAL;
	}
	dev->agp_buffer_token = init->buffers_offset;
	dev->agp_buffer_map = drm_core_findmap(dev, init->buffers_offset);
	if (!dev->agp_buffer_map) {
		dev->dev_private = (void *)dev_priv;
		i830_dma_cleanup(dev);
		DRM_ERROR("can not find dma buffer map!\n");
		return -EINVAL;
	}

	dev_priv->sarea_priv = (drm_i830_sarea_t *)
	    ((u8 *) dev_priv->sarea_map->handle + init->sarea_priv_offset);

	dev_priv->ring.Start = init->ring_start;
	dev_priv->ring.End = init->ring_end;
	dev_priv->ring.Size = init->ring_size;

	dev_priv->ring.virtual_start = drm_ioremap(dev->agp->base +
						   init->ring_start,
						   init->ring_size, dev);

	if (dev_priv->ring.virtual_start == NULL) {
		dev->dev_private = (void *)dev_priv;
		i830_dma_cleanup(dev);
		DRM_ERROR("can not ioremap virtual address for"
			  " ring buffer\n");
		return -ENOMEM;
	}

	dev_priv->ring.tail_mask = dev_priv->ring.Size - 1;

	dev_priv->w = init->w;
	dev_priv->h = init->h;
	dev_priv->pitch = init->pitch;
	dev_priv->back_offset = init->back_offset;
	dev_priv->depth_offset = init->depth_offset;
	dev_priv->front_offset = init->front_offset;

	dev_priv->front_di1 = init->front_offset | init->pitch_bits;
	dev_priv->back_di1 = init->back_offset | init->pitch_bits;
	dev_priv->zi1 = init->depth_offset | init->pitch_bits;

	DRM_DEBUG("front_di1 %x\n", dev_priv->front_di1);
	DRM_DEBUG("back_offset %x\n", dev_priv->back_offset);
	DRM_DEBUG("back_di1 %x\n", dev_priv->back_di1);
	DRM_DEBUG("pitch_bits %x\n", init->pitch_bits);

	dev_priv->cpp = init->cpp;
	/* We are using separate values as placeholders for mechanisms for
	 * private backbuffer/depthbuffer usage.
	 */

	dev_priv->back_pitch = init->back_pitch;
	dev_priv->depth_pitch = init->depth_pitch;
	dev_priv->do_boxes = 0;
	dev_priv->use_mi_batchbuffer_start = 0;

	/* Program Hardware Status Page */
	dev_priv->hw_status_page =
	    pci_alloc_consistent(dev->pdev, PAGE_SIZE,
				 &dev_priv->dma_status_page);
	if (!dev_priv->hw_status_page) {
		dev->dev_private = (void *)dev_priv;
		i830_dma_cleanup(dev);
		DRM_ERROR("Can not allocate hardware status page\n");
		return -ENOMEM;
	}
	memset(dev_priv->hw_status_page, 0, PAGE_SIZE);
	DRM_DEBUG("hw status page @ %p\n", dev_priv->hw_status_page);

	I830_WRITE(0x02080, dev_priv->dma_status_page);
	DRM_DEBUG("Enabled hardware status page\n");

	/* Now we need to init our freelist */
	if (i830_freelist_init(dev, dev_priv) != 0) {
		dev->dev_private = (void *)dev_priv;
		i830_dma_cleanup(dev);
		DRM_ERROR("Not enough space in the status page for"
			  " the freelist\n");
		return -ENOMEM;
	}
	dev->dev_private = (void *)dev_priv;

	return 0;
}

static int i830_dma_init(struct inode *inode, struct file *filp,
			 unsigned int cmd, unsigned long arg)
{
	drm_file_t *priv = filp->private_data;
	drm_device_t *dev = priv->head->dev;
	drm_i830_private_t *dev_priv;
	drm_i830_init_t init;
	int retcode = 0;

	if (copy_from_user(&init, (void *__user)arg, sizeof(init)))
		return -EFAULT;

	switch (init.func) {
	case I830_INIT_DMA:
		dev_priv = drm_alloc(sizeof(drm_i830_private_t),
				     DRM_MEM_DRIVER);
		if (dev_priv == NULL)
			return -ENOMEM;
		retcode = i830_dma_initialize(dev, dev_priv, &init);
		break;
	case I830_CLEANUP_DMA:
		retcode = i830_dma_cleanup(dev);
		break;
	default:
		retcode = -EINVAL;
		break;
	}

	return retcode;
}

#define GFX_OP_STIPPLE           ((0x3<<29)|(0x1d<<24)|(0x83<<16))
#define ST1_ENABLE               (1<<16)
#define ST1_MASK                 (0xffff)

/* Most efficient way to verify state for the i830 is as it is
 * emitted.  Non-conformant state is silently dropped.
 */
static void i830EmitContextVerified(drm_device_t * dev, unsigned int *code)
{
	drm_i830_private_t *dev_priv = dev->dev_private;
	int i, j = 0;
	unsigned int tmp;
	RING_LOCALS;

	BEGIN_LP_RING(I830_CTX_SETUP_SIZE + 4);

	for (i = 0; i < I830_CTXREG_BLENDCOLR0; i++) {
		tmp = code[i];
		if ((tmp & (7 << 29)) == CMD_3D &&
		    (tmp & (0x1f << 24)) < (0x1d << 24)) {
			OUT_RING(tmp);
			j++;
		} else {
			DRM_ERROR("Skipping %d\n", i);
		}
	}

	OUT_RING(STATE3D_CONST_BLEND_COLOR_CMD);
	OUT_RING(code[I830_CTXREG_BLENDCOLR]);
	j += 2;

	for (i = I830_CTXREG_VF; i < I830_CTXREG_MCSB0; i++) {
		tmp = code[i];
		if ((tmp & (7 << 29)) == CMD_3D &&
		    (tmp & (0x1f << 24)) < (0x1d << 24)) {
			OUT_RING(tmp);
			j++;
		} else {
			DRM_ERROR("Skipping %d\n", i);
		}
	}

	OUT_RING(STATE3D_MAP_COORD_SETBIND_CMD);
	OUT_RING(code[I830_CTXREG_MCSB1]);
	j += 2;

	if (j & 1)
		OUT_RING(0);

	ADVANCE_LP_RING();
}