2 * pxa3xx-gcu.c - Linux kernel module for PXA3xx graphics controllers
4 * This driver needs a DirectFB counterpart in user space, communication
5 * is handled via mmap()ed memory areas and an ioctl.
7 * Copyright (c) 2009 Daniel Mack <daniel@caiaq.de>
8 * Copyright (c) 2009 Janine Kropp <nin@directfb.org>
9 * Copyright (c) 2009 Denis Oliver Kropp <dok@directfb.org>
11 * This program is free software; you can redistribute it and/or modify
12 * it under the terms of the GNU General Public License as published by
13 * the Free Software Foundation; either version 2 of the License, or
14 * (at your option) any later version.
16 * This program is distributed in the hope that it will be useful,
17 * but WITHOUT ANY WARRANTY; without even the implied warranty of
18 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19 * GNU General Public License for more details.
21 * You should have received a copy of the GNU General Public License
22 * along with this program; if not, write to the Free Software
23 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
27 * WARNING: This controller is attached to System Bus 2 of the PXA which
28 * needs its arbiter to be enabled explicitly (CKENB & 1<<9).
29 * There is currently no way to do this from Linux, so you need to teach
30 * your bootloader for now.
33 #include <linux/module.h>
34 #include <linux/platform_device.h>
35 #include <linux/dma-mapping.h>
36 #include <linux/miscdevice.h>
37 #include <linux/interrupt.h>
38 #include <linux/spinlock.h>
39 #include <linux/uaccess.h>
40 #include <linux/ioctl.h>
41 #include <linux/delay.h>
42 #include <linux/sched.h>
43 #include <linux/slab.h>
44 #include <linux/clk.h>
49 #include "pxa3xx-gcu.h"
51 #define DRV_NAME "pxa3xx-gcu"
52 #define MISCDEV_MINOR 197
55 #define GCCR_SYNC_CLR (1 << 9)
56 #define GCCR_BP_RST (1 << 8)
57 #define GCCR_ABORT (1 << 6)
58 #define GCCR_STOP (1 << 4)
60 #define REG_GCISCR 0x04
61 #define REG_GCIECR 0x08
62 #define REG_GCRBBR 0x20
63 #define REG_GCRBLR 0x24
64 #define REG_GCRBHR 0x28
65 #define REG_GCRBTR 0x2C
66 #define REG_GCRBEXHR 0x30
68 #define IE_EOB (1 << 0)
69 #define IE_EEOB (1 << 5)
72 #define SHARED_SIZE PAGE_ALIGN(sizeof(struct pxa3xx_gcu_shared))
74 /* #define PXA3XX_GCU_DEBUG */
75 /* #define PXA3XX_GCU_DEBUG_TIMER */
77 #ifdef PXA3XX_GCU_DEBUG
80 QPRINT(priv, KERN_DEBUG, msg); \
83 #define QDUMP(msg) do {} while (0)
88 QPRINT(priv, KERN_ERR, msg); \
91 struct pxa3xx_gcu_batch {
92 struct pxa3xx_gcu_batch *next;
98 struct pxa3xx_gcu_priv {
100 void __iomem *mmio_base;
102 struct pxa3xx_gcu_shared *shared;
103 dma_addr_t shared_phys;
104 struct resource *resource_mem;
105 struct miscdevice misc_dev;
106 wait_queue_head_t wait_idle;
107 wait_queue_head_t wait_free;
109 struct timespec64 base_time;
111 struct pxa3xx_gcu_batch *free;
112 struct pxa3xx_gcu_batch *ready;
113 struct pxa3xx_gcu_batch *ready_last;
114 struct pxa3xx_gcu_batch *running;
117 static inline unsigned long
118 gc_readl(struct pxa3xx_gcu_priv *priv, unsigned int off)
120 return __raw_readl(priv->mmio_base + off);
124 gc_writel(struct pxa3xx_gcu_priv *priv, unsigned int off, unsigned long val)
126 __raw_writel(val, priv->mmio_base + off);
129 #define QPRINT(priv, level, msg) \
131 struct timespec64 ts; \
132 struct pxa3xx_gcu_shared *shared = priv->shared; \
133 u32 base = gc_readl(priv, REG_GCRBBR); \
135 ktime_get_ts64(&ts); \
136 ts = timespec64_sub(ts, priv->base_time); \
138 printk(level "%lld.%03ld.%03ld - %-17s: %-21s (%s, " \
140 "0x%02lx, B 0x%08lx [%ld], E %5ld, H %5ld, " \
143 ts.tv_nsec / NSEC_PER_MSEC, \
144 (ts.tv_nsec % NSEC_PER_MSEC) / USEC_PER_MSEC, \
146 shared->hw_running ? "running" : " idle", \
147 gc_readl(priv, REG_GCISCR), \
148 gc_readl(priv, REG_GCRBBR), \
149 gc_readl(priv, REG_GCRBLR), \
150 (gc_readl(priv, REG_GCRBEXHR) - base) / 4, \
151 (gc_readl(priv, REG_GCRBHR) - base) / 4, \
152 (gc_readl(priv, REG_GCRBTR) - base) / 4); \
156 pxa3xx_gcu_reset(struct pxa3xx_gcu_priv *priv)
160 /* disable interrupts */
161 gc_writel(priv, REG_GCIECR, 0);
164 gc_writel(priv, REG_GCCR, GCCR_ABORT);
165 gc_writel(priv, REG_GCCR, 0);
167 memset(priv->shared, 0, SHARED_SIZE);
168 priv->shared->buffer_phys = priv->shared_phys;
169 priv->shared->magic = PXA3XX_GCU_SHARED_MAGIC;
171 ktime_get_ts64(&priv->base_time);
173 /* set up the ring buffer pointers */
174 gc_writel(priv, REG_GCRBLR, 0);
175 gc_writel(priv, REG_GCRBBR, priv->shared_phys);
176 gc_writel(priv, REG_GCRBTR, priv->shared_phys);
178 /* enable all IRQs except EOB */
179 gc_writel(priv, REG_GCIECR, IE_ALL & ~IE_EOB);
183 dump_whole_state(struct pxa3xx_gcu_priv *priv)
185 struct pxa3xx_gcu_shared *sh = priv->shared;
186 u32 base = gc_readl(priv, REG_GCRBBR);
190 printk(KERN_DEBUG "== PXA3XX-GCU DUMP ==\n"
191 "%s, STATUS 0x%02lx, B 0x%08lx [%ld], E %5ld, H %5ld, T %5ld\n",
192 sh->hw_running ? "running" : "idle ",
193 gc_readl(priv, REG_GCISCR),
194 gc_readl(priv, REG_GCRBBR),
195 gc_readl(priv, REG_GCRBLR),
196 (gc_readl(priv, REG_GCRBEXHR) - base) / 4,
197 (gc_readl(priv, REG_GCRBHR) - base) / 4,
198 (gc_readl(priv, REG_GCRBTR) - base) / 4);
202 flush_running(struct pxa3xx_gcu_priv *priv)
204 struct pxa3xx_gcu_batch *running = priv->running;
205 struct pxa3xx_gcu_batch *next;
208 next = running->next;
209 running->next = priv->free;
210 priv->free = running;
214 priv->running = NULL;
218 run_ready(struct pxa3xx_gcu_priv *priv)
220 unsigned int num = 0;
221 struct pxa3xx_gcu_shared *shared = priv->shared;
222 struct pxa3xx_gcu_batch *ready = priv->ready;
228 shared->buffer[num++] = 0x05000000;
231 shared->buffer[num++] = 0x00000001;
232 shared->buffer[num++] = ready->phys;
236 shared->buffer[num++] = 0x05000000;
237 priv->running = priv->ready;
238 priv->ready = priv->ready_last = NULL;
239 gc_writel(priv, REG_GCRBLR, 0);
240 shared->hw_running = 1;
242 /* ring base address */
243 gc_writel(priv, REG_GCRBBR, shared->buffer_phys);
245 /* ring tail address */
246 gc_writel(priv, REG_GCRBTR, shared->buffer_phys + num * 4);
249 gc_writel(priv, REG_GCRBLR, ((num + 63) & ~63) * 4);
253 pxa3xx_gcu_handle_irq(int irq, void *ctx)
255 struct pxa3xx_gcu_priv *priv = ctx;
256 struct pxa3xx_gcu_shared *shared = priv->shared;
257 u32 status = gc_readl(priv, REG_GCISCR) & IE_ALL;
264 spin_lock(&priv->spinlock);
265 shared->num_interrupts++;
267 if (status & IE_EEOB) {
271 wake_up_all(&priv->wait_free);
276 /* There is no more data prepared by the userspace.
277 * Set hw_running = 0 and wait for the next userspace
280 shared->hw_running = 0;
284 /* set ring buffer length to zero */
285 gc_writel(priv, REG_GCRBLR, 0);
287 wake_up_all(&priv->wait_idle);
293 dump_whole_state(priv);
296 /* Clear the interrupt */
297 gc_writel(priv, REG_GCISCR, status);
298 spin_unlock(&priv->spinlock);
304 pxa3xx_gcu_wait_idle(struct pxa3xx_gcu_priv *priv)
308 QDUMP("Waiting for idle...");
310 /* Does not need to be atomic. There's a lock in user space,
311 * but anyhow, this is just for statistics. */
312 priv->shared->num_wait_idle++;
314 while (priv->shared->hw_running) {
315 int num = priv->shared->num_interrupts;
316 u32 rbexhr = gc_readl(priv, REG_GCRBEXHR);
318 ret = wait_event_interruptible_timeout(priv->wait_idle,
319 !priv->shared->hw_running, HZ*4);
324 if (gc_readl(priv, REG_GCRBEXHR) == rbexhr &&
325 priv->shared->num_interrupts == num) {
338 pxa3xx_gcu_wait_free(struct pxa3xx_gcu_priv *priv)
342 QDUMP("Waiting for free...");
344 /* Does not need to be atomic. There's a lock in user space,
345 * but anyhow, this is just for statistics. */
346 priv->shared->num_wait_free++;
348 while (!priv->free) {
349 u32 rbexhr = gc_readl(priv, REG_GCRBEXHR);
351 ret = wait_event_interruptible_timeout(priv->wait_free,
360 if (gc_readl(priv, REG_GCRBEXHR) == rbexhr) {
372 /* Misc device layer */
374 static inline struct pxa3xx_gcu_priv *to_pxa3xx_gcu_priv(struct file *file)
376 struct miscdevice *dev = file->private_data;
377 return container_of(dev, struct pxa3xx_gcu_priv, misc_dev);
381 * provide an empty .open callback, so the core sets file->private_data
384 static int pxa3xx_gcu_open(struct inode *inode, struct file *file)
390 pxa3xx_gcu_write(struct file *file, const char *buff,
391 size_t count, loff_t *offp)
395 struct pxa3xx_gcu_batch *buffer;
396 struct pxa3xx_gcu_priv *priv = to_pxa3xx_gcu_priv(file);
398 int words = count / 4;
400 /* Does not need to be atomic. There's a lock in user space,
401 * but anyhow, this is just for statistics. */
402 priv->shared->num_writes++;
403 priv->shared->num_words += words;
405 /* Last word reserved for batch buffer end command */
406 if (words >= PXA3XX_GCU_BATCH_WORDS)
409 /* Wait for a free buffer */
411 ret = pxa3xx_gcu_wait_free(priv);
417 * Get buffer from free list
419 spin_lock_irqsave(&priv->spinlock, flags);
421 priv->free = buffer->next;
422 spin_unlock_irqrestore(&priv->spinlock, flags);
425 /* Copy data from user into buffer */
426 ret = copy_from_user(buffer->ptr, buff, words * 4);
428 spin_lock_irqsave(&priv->spinlock, flags);
429 buffer->next = priv->free;
431 spin_unlock_irqrestore(&priv->spinlock, flags);
435 buffer->length = words;
437 /* Append batch buffer end command */
438 buffer->ptr[words] = 0x01000000;
441 * Add buffer to ready list
443 spin_lock_irqsave(&priv->spinlock, flags);
448 BUG_ON(priv->ready_last == NULL);
450 priv->ready_last->next = buffer;
452 priv->ready = buffer;
454 priv->ready_last = buffer;
456 if (!priv->shared->hw_running)
459 spin_unlock_irqrestore(&priv->spinlock, flags);
466 pxa3xx_gcu_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
469 struct pxa3xx_gcu_priv *priv = to_pxa3xx_gcu_priv(file);
472 case PXA3XX_GCU_IOCTL_RESET:
473 spin_lock_irqsave(&priv->spinlock, flags);
474 pxa3xx_gcu_reset(priv);
475 spin_unlock_irqrestore(&priv->spinlock, flags);
478 case PXA3XX_GCU_IOCTL_WAIT_IDLE:
479 return pxa3xx_gcu_wait_idle(priv);
486 pxa3xx_gcu_mmap(struct file *file, struct vm_area_struct *vma)
488 unsigned int size = vma->vm_end - vma->vm_start;
489 struct pxa3xx_gcu_priv *priv = to_pxa3xx_gcu_priv(file);
491 switch (vma->vm_pgoff) {
493 /* hand out the shared data area */
494 if (size != SHARED_SIZE)
497 return dma_mmap_coherent(priv->dev, vma,
498 priv->shared, priv->shared_phys, size);
500 case SHARED_SIZE >> PAGE_SHIFT:
501 /* hand out the MMIO base for direct register access
503 if (size != resource_size(priv->resource_mem))
506 vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
508 return io_remap_pfn_range(vma, vma->vm_start,
509 priv->resource_mem->start >> PAGE_SHIFT,
510 size, vma->vm_page_prot);
517 #ifdef PXA3XX_GCU_DEBUG_TIMER
518 static struct timer_list pxa3xx_gcu_debug_timer;
519 static struct pxa3xx_gcu_priv *debug_timer_priv;
521 static void pxa3xx_gcu_debug_timedout(struct timer_list *unused)
523 struct pxa3xx_gcu_priv *priv = debug_timer_priv;
525 QERROR("Timer DUMP");
527 mod_timer(&pxa3xx_gcu_debug_timer, jiffies + 5 * HZ);
530 static void pxa3xx_gcu_init_debug_timer(struct pxa3xx_gcu_priv *priv)
532 /* init the timer structure */
533 debug_timer_priv = priv;
534 timer_setup(&pxa3xx_gcu_debug_timer, pxa3xx_gcu_debug_timedout, 0);
535 pxa3xx_gcu_debug_timedout(NULL);
538 static inline void pxa3xx_gcu_init_debug_timer(struct pxa3xx_gcu_priv *priv) {}
542 pxa3xx_gcu_add_buffer(struct device *dev,
543 struct pxa3xx_gcu_priv *priv)
545 struct pxa3xx_gcu_batch *buffer;
547 buffer = kzalloc(sizeof(struct pxa3xx_gcu_batch), GFP_KERNEL);
551 buffer->ptr = dma_alloc_coherent(dev, PXA3XX_GCU_BATCH_WORDS * 4,
552 &buffer->phys, GFP_KERNEL);
558 buffer->next = priv->free;
565 pxa3xx_gcu_free_buffers(struct device *dev,
566 struct pxa3xx_gcu_priv *priv)
568 struct pxa3xx_gcu_batch *next, *buffer = priv->free;
573 dma_free_coherent(dev, PXA3XX_GCU_BATCH_WORDS * 4,
574 buffer->ptr, buffer->phys);
583 static const struct file_operations pxa3xx_gcu_miscdev_fops = {
584 .owner = THIS_MODULE,
585 .open = pxa3xx_gcu_open,
586 .write = pxa3xx_gcu_write,
587 .unlocked_ioctl = pxa3xx_gcu_ioctl,
588 .mmap = pxa3xx_gcu_mmap,
591 static int pxa3xx_gcu_probe(struct platform_device *pdev)
595 struct pxa3xx_gcu_priv *priv;
596 struct device *dev = &pdev->dev;
598 priv = devm_kzalloc(dev, sizeof(struct pxa3xx_gcu_priv), GFP_KERNEL);
602 init_waitqueue_head(&priv->wait_idle);
603 init_waitqueue_head(&priv->wait_free);
604 spin_lock_init(&priv->spinlock);
606 /* we allocate the misc device structure as part of our own allocation,
607 * so we can get a pointer to our priv structure later on with
608 * container_of(). This isn't really necessary as we have a fixed minor
609 * number anyway, but this is to avoid statics. */
611 priv->misc_dev.minor = MISCDEV_MINOR,
612 priv->misc_dev.name = DRV_NAME,
613 priv->misc_dev.fops = &pxa3xx_gcu_miscdev_fops;
615 /* handle IO resources */
616 r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
617 priv->mmio_base = devm_ioremap_resource(dev, r);
618 if (IS_ERR(priv->mmio_base))
619 return PTR_ERR(priv->mmio_base);
621 /* enable the clock */
622 priv->clk = devm_clk_get(dev, NULL);
623 if (IS_ERR(priv->clk)) {
624 dev_err(dev, "failed to get clock\n");
625 return PTR_ERR(priv->clk);
628 /* request the IRQ */
629 irq = platform_get_irq(pdev, 0);
631 dev_err(dev, "no IRQ defined: %d\n", irq);
635 ret = devm_request_irq(dev, irq, pxa3xx_gcu_handle_irq,
638 dev_err(dev, "request_irq failed\n");
642 /* allocate dma memory */
643 priv->shared = dma_alloc_coherent(dev, SHARED_SIZE,
644 &priv->shared_phys, GFP_KERNEL);
646 dev_err(dev, "failed to allocate DMA memory\n");
650 /* register misc device */
651 ret = misc_register(&priv->misc_dev);
653 dev_err(dev, "misc_register() for minor %d failed\n",
658 ret = clk_prepare_enable(priv->clk);
660 dev_err(dev, "failed to enable clock\n");
661 goto err_misc_deregister;
664 for (i = 0; i < 8; i++) {
665 ret = pxa3xx_gcu_add_buffer(dev, priv);
667 dev_err(dev, "failed to allocate DMA memory\n");
668 goto err_disable_clk;
672 platform_set_drvdata(pdev, priv);
673 priv->resource_mem = r;
675 pxa3xx_gcu_reset(priv);
676 pxa3xx_gcu_init_debug_timer(priv);
678 dev_info(dev, "registered @0x%p, DMA 0x%p (%d bytes), IRQ %d\n",
679 (void *) r->start, (void *) priv->shared_phys,
684 dma_free_coherent(dev, SHARED_SIZE,
685 priv->shared, priv->shared_phys);
688 misc_deregister(&priv->misc_dev);
691 clk_disable_unprepare(priv->clk);
696 static int pxa3xx_gcu_remove(struct platform_device *pdev)
698 struct pxa3xx_gcu_priv *priv = platform_get_drvdata(pdev);
699 struct device *dev = &pdev->dev;
701 pxa3xx_gcu_wait_idle(priv);
702 misc_deregister(&priv->misc_dev);
703 dma_free_coherent(dev, SHARED_SIZE, priv->shared, priv->shared_phys);
704 pxa3xx_gcu_free_buffers(dev, priv);
710 static const struct of_device_id pxa3xx_gcu_of_match[] = {
711 { .compatible = "marvell,pxa300-gcu", },
714 MODULE_DEVICE_TABLE(of, pxa3xx_gcu_of_match);
717 static struct platform_driver pxa3xx_gcu_driver = {
718 .probe = pxa3xx_gcu_probe,
719 .remove = pxa3xx_gcu_remove,
722 .of_match_table = of_match_ptr(pxa3xx_gcu_of_match),
726 module_platform_driver(pxa3xx_gcu_driver);
728 MODULE_DESCRIPTION("PXA3xx graphics controller unit driver");
729 MODULE_LICENSE("GPL");
730 MODULE_ALIAS_MISCDEV(MISCDEV_MINOR);
731 MODULE_AUTHOR("Janine Kropp <nin@directfb.org>, "
732 "Denis Oliver Kropp <dok@directfb.org>, "
733 "Daniel Mack <daniel@caiaq.de>");