diff options
author | Zou Nan hai <nanhai.zou@intel.com> | 2010-05-21 01:08:55 (GMT) |
---|---|---|
committer | Eric Anholt <eric@anholt.net> | 2010-05-26 20:24:49 (GMT) |
commit | 8187a2b70e34c727a06617441f74f202b6fefaf9 (patch) | |
tree | 48622c6f95282dc0a0fa668110aac4efa6e89066 | |
parent | d3301d86b4bf2bcf649982ae464211d8bcf9575a (diff) | |
download | linux-8187a2b70e34c727a06617441f74f202b6fefaf9.tar.xz |
drm/i915: introduce intel_ring_buffer structure (V2)
Introduces a more complete intel_ring_buffer structure with callbacks
for setup and management of a particular ringbuffer, and converts the
render ring buffer consumers to use it.
Signed-off-by: Zou Nan hai <nanhai.zou@intel.com>
Signed-off-by: Xiang Hai hao <haihao.xiang@intel.com>
[anholt: Fixed up whitespace fail and rebased against prep patches]
Signed-off-by: Eric Anholt <eric@anholt.net>
-rw-r--r-- | drivers/gpu/drm/i915/i915_debugfs.c | 6 | ||||
-rw-r--r-- | drivers/gpu/drm/i915/i915_dma.c | 58 | ||||
-rw-r--r-- | drivers/gpu/drm/i915/i915_drv.c | 29 | ||||
-rw-r--r-- | drivers/gpu/drm/i915/i915_drv.h | 80 | ||||
-rw-r--r-- | drivers/gpu/drm/i915/i915_gem.c | 76 | ||||
-rw-r--r-- | drivers/gpu/drm/i915/i915_irq.c | 15 | ||||
-rw-r--r-- | drivers/gpu/drm/i915/intel_display.c | 1 | ||||
-rw-r--r-- | drivers/gpu/drm/i915/intel_overlay.c | 8 | ||||
-rw-r--r-- | drivers/gpu/drm/i915/intel_ringbuffer.c | 582 | ||||
-rw-r--r-- | drivers/gpu/drm/i915/intel_ringbuffer.h | 124 | ||||
-rw-r--r-- | include/drm/i915_drm.h | 4 |
11 files changed, 606 insertions, 377 deletions
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index 488175c..4fddf09 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -317,14 +317,14 @@ static int i915_ringbuffer_data(struct seq_file *m, void *data) u8 *virt; uint32_t *ptr, off; - if (!dev_priv->render_ring.ring_obj) { + if (!dev_priv->render_ring.gem_object) { seq_printf(m, "No ringbuffer setup\n"); return 0; } virt = dev_priv->render_ring.virtual_start; - for (off = 0; off < dev_priv->render_ring.Size; off += 4) { + for (off = 0; off < dev_priv->render_ring.size; off += 4) { ptr = (uint32_t *)(virt + off); seq_printf(m, "%08x : %08x\n", off, *ptr); } @@ -344,7 +344,7 @@ static int i915_ringbuffer_info(struct seq_file *m, void *data) seq_printf(m, "RingHead : %08x\n", head); seq_printf(m, "RingTail : %08x\n", tail); - seq_printf(m, "RingSize : %08lx\n", dev_priv->render_ring.Size); + seq_printf(m, "RingSize : %08lx\n", dev_priv->render_ring.size); seq_printf(m, "Acthd : %08x\n", I915_READ(IS_I965G(dev) ? ACTHD_I965 : ACTHD)); return 0; diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 6de7eac..2541428 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -40,7 +40,6 @@ #include <linux/vga_switcheroo.h> #include <linux/slab.h> - /** * Sets up the hardware status page for devices that need a physical address * in the register. @@ -56,10 +55,11 @@ static int i915_init_phys_hws(struct drm_device *dev) DRM_ERROR("Can not allocate hardware status page\n"); return -ENOMEM; } - dev_priv->hw_status_page = dev_priv->status_page_dmah->vaddr; + dev_priv->render_ring.status_page.page_addr + = dev_priv->status_page_dmah->vaddr; dev_priv->dma_status_page = dev_priv->status_page_dmah->busaddr; - memset(dev_priv->hw_status_page, 0, PAGE_SIZE); + memset(dev_priv->render_ring.status_page.page_addr, 0, PAGE_SIZE); if (IS_I965G(dev)) dev_priv->dma_status_page |= (dev_priv->dma_status_page >> 28) & @@ -95,7 +95,7 @@ void i915_kernel_lost_context(struct drm_device * dev) { drm_i915_private_t *dev_priv = dev->dev_private; struct drm_i915_master_private *master_priv; - drm_i915_ring_buffer_t *ring = &(dev_priv->render_ring); + struct intel_ring_buffer *ring = &dev_priv->render_ring; /* * We should never lose context on the ring with modesetting @@ -108,7 +108,7 @@ void i915_kernel_lost_context(struct drm_device * dev) ring->tail = I915_READ(PRB0_TAIL) & TAIL_ADDR; ring->space = ring->head - (ring->tail + 8); if (ring->space < 0) - ring->space += ring->Size; + ring->space += ring->size; if (!dev->primary->master) return; @@ -128,12 +128,7 @@ static int i915_dma_cleanup(struct drm_device * dev) if (dev->irq_enabled) drm_irq_uninstall(dev); - if (dev_priv->render_ring.virtual_start) { - drm_core_ioremapfree(&dev_priv->render_ring.map, dev); - dev_priv->render_ring.virtual_start = NULL; - dev_priv->render_ring.map.handle = NULL; - dev_priv->render_ring.map.size = 0; - } + intel_cleanup_ring_buffer(dev, &dev_priv->render_ring); /* Clear the HWS virtual address at teardown */ if (I915_NEED_GFX_HWS(dev)) @@ -156,14 +151,14 @@ static int i915_initialize(struct drm_device * dev, drm_i915_init_t * init) } if (init->ring_size != 0) { - if (dev_priv->render_ring.ring_obj != NULL) { + if (dev_priv->render_ring.gem_object != NULL) { i915_dma_cleanup(dev); DRM_ERROR("Client tried to initialize ringbuffer in " "GEM mode\n"); return -EINVAL; } - dev_priv->render_ring.Size = init->ring_size; + dev_priv->render_ring.size = init->ring_size; dev_priv->render_ring.map.offset = init->ring_start; dev_priv->render_ring.map.size = init->ring_size; @@ -201,26 +196,29 @@ static int i915_dma_resume(struct drm_device * dev) { drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; + struct intel_ring_buffer *ring; DRM_DEBUG_DRIVER("%s\n", __func__); - if (dev_priv->render_ring.map.handle == NULL) { + ring = &dev_priv->render_ring; + + if (ring->map.handle == NULL) { DRM_ERROR("can not ioremap virtual address for" " ring buffer\n"); return -ENOMEM; } /* Program Hardware Status Page */ - if (!dev_priv->hw_status_page) { + if (!ring->status_page.page_addr) { DRM_ERROR("Can not find hardware status page\n"); return -EINVAL; } DRM_DEBUG_DRIVER("hw status page @ %p\n", - dev_priv->hw_status_page); - - if (dev_priv->status_gfx_addr != 0) - I915_WRITE(HWS_PGA, dev_priv->status_gfx_addr); + ring->status_page.page_addr); + if (ring->status_page.gfx_addr != 0) + ring->setup_status_page(dev, ring); else I915_WRITE(HWS_PGA, dev_priv->dma_status_page); + DRM_DEBUG_DRIVER("Enabled hardware status page\n"); return 0; @@ -330,9 +328,8 @@ static int i915_emit_cmds(struct drm_device * dev, int *buffer, int dwords) { drm_i915_private_t *dev_priv = dev->dev_private; int i; - RING_LOCALS; - if ((dwords+1) * sizeof(int) >= dev_priv->render_ring.Size - 8) + if ((dwords+1) * sizeof(int) >= dev_priv->render_ring.size - 8) return -EINVAL; BEGIN_LP_RING((dwords+1)&~1); @@ -365,9 +362,7 @@ i915_emit_box(struct drm_device *dev, struct drm_clip_rect *boxes, int i, int DR1, int DR4) { - drm_i915_private_t *dev_priv = dev->dev_private; struct drm_clip_rect box = boxes[i]; - RING_LOCALS; if (box.y2 <= box.y1 || box.x2 <= box.x1 || box.y2 <= 0 || box.x2 <= 0) { DRM_ERROR("Bad box %d,%d..%d,%d\n", @@ -404,7 +399,6 @@ static void i915_emit_breadcrumb(struct drm_device *dev) { drm_i915_private_t *dev_priv = dev->dev_private; struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv; - RING_LOCALS; dev_priv->counter++; if (dev_priv->counter > 0x7FFFFFFFUL) @@ -458,10 +452,8 @@ static int i915_dispatch_batchbuffer(struct drm_device * dev, drm_i915_batchbuffer_t * batch, struct drm_clip_rect *cliprects) { - drm_i915_private_t *dev_priv = dev->dev_private; int nbox = batch->num_cliprects; int i = 0, count; - RING_LOCALS; if ((batch->start | batch->used) & 0x7) { DRM_ERROR("alignment"); @@ -510,7 +502,6 @@ static int i915_dispatch_flip(struct drm_device * dev) drm_i915_private_t *dev_priv = dev->dev_private; struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv; - RING_LOCALS; if (!master_priv->sarea_priv) return -EINVAL; @@ -563,7 +554,8 @@ static int i915_quiescent(struct drm_device * dev) drm_i915_private_t *dev_priv = dev->dev_private; i915_kernel_lost_context(dev); - return i915_wait_ring(dev, dev_priv->render_ring.Size - 8, __func__); + return intel_wait_ring_buffer(dev, &dev_priv->render_ring, + dev_priv->render_ring.size - 8); } static int i915_flush_ioctl(struct drm_device *dev, void *data, @@ -805,6 +797,7 @@ static int i915_set_status_page(struct drm_device *dev, void *data, { drm_i915_private_t *dev_priv = dev->dev_private; drm_i915_hws_addr_t *hws = data; + struct intel_ring_buffer *ring = &dev_priv->render_ring; if (!I915_NEED_GFX_HWS(dev)) return -EINVAL; @@ -821,7 +814,7 @@ static int i915_set_status_page(struct drm_device *dev, void *data, DRM_DEBUG_DRIVER("set status page addr 0x%08x\n", (u32)hws->addr); - dev_priv->status_gfx_addr = hws->addr & (0x1ffff<<12); + ring->status_page.gfx_addr = hws->addr & (0x1ffff<<12); dev_priv->hws_map.offset = dev->agp->base + hws->addr; dev_priv->hws_map.size = 4*1024; @@ -837,10 +830,10 @@ static int i915_set_status_page(struct drm_device *dev, void *data, " G33 hw status page\n"); return -ENOMEM; } - dev_priv->hw_status_page = dev_priv->hws_map.handle; + ring->status_page.page_addr = dev_priv->hws_map.handle; + memset(ring->status_page.page_addr, 0, PAGE_SIZE); + I915_WRITE(HWS_PGA, ring->status_page.gfx_addr); - memset(dev_priv->hw_status_page, 0, PAGE_SIZE); - I915_WRITE(HWS_PGA, dev_priv->status_gfx_addr); DRM_DEBUG_DRIVER("load hws HWS_PGA with gfx mem 0x%x\n", dev_priv->status_gfx_addr); DRM_DEBUG_DRIVER("load hws at %p\n", @@ -1639,7 +1632,6 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags) spin_lock_init(&dev_priv->user_irq_lock); spin_lock_init(&dev_priv->error_lock); - dev_priv->user_irq_refcount = 0; dev_priv->trace_irq_seqno = 0; ret = drm_vblank_init(dev, I915_NUM_PIPE); diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index a1814f6..c57c54f 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -388,33 +388,10 @@ int i965_reset(struct drm_device *dev, u8 flags) * switched away). */ if (drm_core_check_feature(dev, DRIVER_MODESET) || - !dev_priv->mm.suspended) { - drm_i915_ring_buffer_t *ring = &dev_priv->render_ring; - struct drm_gem_object *obj = ring->ring_obj; - struct drm_i915_gem_object *obj_priv = to_intel_bo(obj); + !dev_priv->mm.suspended) { + struct intel_ring_buffer *ring = &dev_priv->render_ring; dev_priv->mm.suspended = 0; - - /* Stop the ring if it's running. */ - I915_WRITE(PRB0_CTL, 0); - I915_WRITE(PRB0_TAIL, 0); - I915_WRITE(PRB0_HEAD, 0); - - /* Initialize the ring. */ - I915_WRITE(PRB0_START, obj_priv->gtt_offset); - I915_WRITE(PRB0_CTL, - ((obj->size - 4096) & RING_NR_PAGES) | - RING_NO_REPORT | - RING_VALID); - if (!drm_core_check_feature(dev, DRIVER_MODESET)) - i915_kernel_lost_context(dev); - else { - ring->head = I915_READ(PRB0_HEAD) & HEAD_ADDR; - ring->tail = I915_READ(PRB0_TAIL) & TAIL_ADDR; - ring->space = ring->head - (ring->tail + 8); - if (ring->space < 0) - ring->space += ring->Size; - } - + ring->init(dev, ring); mutex_unlock(&dev->struct_mutex); drm_irq_uninstall(dev); drm_irq_install(dev); diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index a39440c..6bb7933 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -31,8 +31,8 @@ #define _I915_DRV_H_ #include "i915_reg.h" -#include "i915_drm.h" #include "intel_bios.h" +#include "intel_ringbuffer.h" #include <linux/io-mapping.h> /* General customization: @@ -92,16 +92,6 @@ struct drm_i915_gem_phys_object { struct drm_gem_object *cur_obj; }; -typedef struct _drm_i915_ring_buffer { - unsigned long Size; - u8 *virtual_start; - int head; - int tail; - int space; - drm_local_map_t map; - struct drm_gem_object *ring_obj; -} drm_i915_ring_buffer_t; - struct mem_block { struct mem_block *next; struct mem_block *prev; @@ -244,7 +234,7 @@ typedef struct drm_i915_private { void __iomem *regs; struct pci_dev *bridge_dev; - drm_i915_ring_buffer_t render_ring; + struct intel_ring_buffer render_ring; drm_dma_handle_t *status_page_dmah; void *hw_status_page; @@ -270,8 +260,6 @@ typedef struct drm_i915_private { atomic_t irq_received; /** Protects user_irq_refcount and irq_mask_reg */ spinlock_t user_irq_lock; - /** Refcount for i915_user_irq_get() versus i915_user_irq_put(). */ - int user_irq_refcount; u32 trace_irq_seqno; /** Cached value of IMR to avoid reads in updating the bitfield */ u32 irq_mask_reg; @@ -832,9 +820,7 @@ extern int i915_irq_emit(struct drm_device *dev, void *data, struct drm_file *file_priv); extern int i915_irq_wait(struct drm_device *dev, void *data, struct drm_file *file_priv); -void i915_user_irq_get(struct drm_device *dev); void i915_trace_irq_get(struct drm_device *dev, u32 seqno); -void i915_user_irq_put(struct drm_device *dev); extern void i915_enable_interrupt (struct drm_device *dev); extern irqreturn_t i915_driver_irq_handler(DRM_IRQ_ARGS); @@ -853,8 +839,10 @@ extern int i915_vblank_swap(struct drm_device *dev, void *data, struct drm_file *file_priv); extern void i915_enable_irq(drm_i915_private_t *dev_priv, u32 mask); extern void i915_disable_irq(drm_i915_private_t *dev_priv, u32 mask); -void ironlake_enable_graphics_irq(drm_i915_private_t *dev_priv, u32 mask); -void ironlake_disable_graphics_irq(drm_i915_private_t *dev_priv, u32 mask); +extern void ironlake_enable_graphics_irq(drm_i915_private_t *dev_priv, + u32 mask); +extern void ironlake_disable_graphics_irq(drm_i915_private_t *dev_priv, + u32 mask); void i915_enable_pipestat(drm_i915_private_t *dev_priv, int pipe, u32 mask); @@ -962,8 +950,6 @@ void i915_gem_object_flush_write_domain(struct drm_gem_object *obj); void i915_gem_shrinker_init(void); void i915_gem_shrinker_exit(void); -int i915_gem_init_pipe_control(struct drm_device *dev); -void i915_gem_cleanup_pipe_control(struct drm_device *dev); /* i915_gem_tiling.c */ void i915_gem_detect_bit_6_swizzle(struct drm_device *dev); @@ -1014,16 +1000,6 @@ static inline void ironlake_opregion_gse_intr(struct drm_device *dev) { return; static inline void opregion_enable_asle(struct drm_device *dev) { return; } #endif -/* intel_ringbuffer.c */ -extern void i915_gem_flush(struct drm_device *dev, - uint32_t invalidate_domains, - uint32_t flush_domains); -extern int i915_dispatch_gem_execbuffer(struct drm_device *dev, - struct drm_i915_gem_execbuffer2 *exec, - struct drm_clip_rect *cliprects, - uint64_t exec_offset); -extern uint32_t i915_ring_add_request(struct drm_device *dev); - /* modesetting */ extern void intel_modeset_init(struct drm_device *dev); extern void intel_modeset_cleanup(struct drm_device *dev); @@ -1044,7 +1020,8 @@ extern int intel_trans_dp_port_sel (struct drm_crtc *crtc); * has access to the ring. */ #define RING_LOCK_TEST_WITH_RETURN(dev, file_priv) do { \ - if (((drm_i915_private_t *)dev->dev_private)->render_ring.ring_obj == NULL) \ + if (((drm_i915_private_t *)dev->dev_private)->render_ring.gem_object \ + == NULL) \ LOCK_TEST_WITH_RETURN(dev, file_priv); \ } while (0) @@ -1060,32 +1037,27 @@ extern int intel_trans_dp_port_sel (struct drm_crtc *crtc); #define I915_VERBOSE 0 -#define RING_LOCALS volatile unsigned int *ring_virt__; - -#define BEGIN_LP_RING(n) do { \ - int bytes__ = 4*(n); \ - if (I915_VERBOSE) DRM_DEBUG("BEGIN_LP_RING(%d)\n", (n)); \ - /* a wrap must occur between instructions so pad beforehand */ \ - if (unlikely (dev_priv->render_ring.tail + bytes__ > dev_priv->render_ring.Size)) \ - i915_wrap_ring(dev); \ - if (unlikely (dev_priv->render_ring.space < bytes__)) \ - i915_wait_ring(dev, bytes__, __func__); \ - ring_virt__ = (unsigned int *) \ - (dev_priv->render_ring.virtual_start + dev_priv->render_ring.tail); \ - dev_priv->render_ring.tail += bytes__; \ - dev_priv->render_ring.tail &= dev_priv->render_ring.Size - 1; \ - dev_priv->render_ring.space -= bytes__; \ +#define BEGIN_LP_RING(n) do { \ + drm_i915_private_t *dev_priv = dev->dev_private; \ + if (I915_VERBOSE) \ + DRM_DEBUG(" BEGIN_LP_RING %x\n", (int)(n)); \ + intel_ring_begin(dev, &dev_priv->render_ring, 4*(n)); \ } while (0) -#define OUT_RING(n) do { \ - if (I915_VERBOSE) DRM_DEBUG(" OUT_RING %x\n", (int)(n)); \ - *ring_virt__++ = (n); \ + +#define OUT_RING(x) do { \ + drm_i915_private_t *dev_priv = dev->dev_private; \ + if (I915_VERBOSE) \ + DRM_DEBUG(" OUT_RING %x\n", (int)(x)); \ + intel_ring_emit(dev, &dev_priv->render_ring, x); \ } while (0) #define ADVANCE_LP_RING() do { \ + drm_i915_private_t *dev_priv = dev->dev_private; \ if (I915_VERBOSE) \ - DRM_DEBUG("ADVANCE_LP_RING %x\n", dev_priv->render_ring.tail); \ - I915_WRITE(PRB0_TAIL, dev_priv->render_ring.tail); \ + DRM_DEBUG("ADVANCE_LP_RING %x\n", \ + dev_priv->render_ring.tail); \ + intel_ring_advance(dev, &dev_priv->render_ring); \ } while(0) /** @@ -1103,14 +1075,12 @@ extern int intel_trans_dp_port_sel (struct drm_crtc *crtc); * * The area from dword 0x20 to 0x3ff is available for driver usage. */ -#define READ_HWSP(dev_priv, reg) (((volatile u32*)(dev_priv->hw_status_page))[reg]) +#define READ_HWSP(dev_priv, reg) (((volatile u32 *)\ + (dev_priv->render_ring.status_page.page_addr))[reg]) #define READ_BREADCRUMB(dev_priv) READ_HWSP(dev_priv, I915_BREADCRUMB_INDEX) #define I915_GEM_HWS_INDEX 0x20 #define I915_BREADCRUMB_INDEX 0x21 -extern int i915_wrap_ring(struct drm_device * dev); -extern int i915_wait_ring(struct drm_device * dev, int n, const char *caller); - #define INTEL_INFO(dev) (((struct drm_i915_private *) (dev)->dev_private)->info) #define IS_I830(dev) ((dev)->pci_device == 0x3577) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 95dbe56..58b6e81 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -1590,6 +1590,7 @@ i915_gem_process_flushing_list(struct drm_device *dev, } } } + uint32_t i915_add_request(struct drm_device *dev, struct drm_file *file_priv, uint32_t flush_domains) @@ -1607,7 +1608,8 @@ i915_add_request(struct drm_device *dev, struct drm_file *file_priv, if (request == NULL) return 0; - seqno = i915_ring_add_request(dev); + seqno = dev_priv->render_ring.add_request(dev, &dev_priv->render_ring, + file_priv, flush_domains); DRM_DEBUG_DRIVER("%d\n", seqno); @@ -1645,10 +1647,8 @@ i915_add_request(struct drm_device *dev, struct drm_file *file_priv, static uint32_t i915_retire_commands(struct drm_device *dev) { - drm_i915_private_t *dev_priv = dev->dev_private; uint32_t cmd = MI_FLUSH | MI_NO_WRITE_FLUSH; uint32_t flush_domains = 0; - RING_LOCALS; /* The sampler always gets flushed on i965 (sigh) */ if (IS_I965G(dev)) @@ -1746,7 +1746,9 @@ i915_gem_retire_requests(struct drm_device *dev) drm_i915_private_t *dev_priv = dev->dev_private; uint32_t seqno; - if (!dev_priv->hw_status_page || list_empty(&dev_priv->mm.request_list)) + struct intel_ring_buffer *ring = &(dev_priv->render_ring); + if (!ring->status_page.page_addr + || list_empty(&dev_priv->mm.request_list)) return; seqno = i915_get_gem_seqno(dev); @@ -1773,7 +1775,8 @@ i915_gem_retire_requests(struct drm_device *dev) if (unlikely (dev_priv->trace_irq_seqno && i915_seqno_passed(dev_priv->trace_irq_seqno, seqno))) { - i915_user_irq_put(dev); + + ring->user_irq_put(dev, ring); dev_priv->trace_irq_seqno = 0; } } @@ -1803,6 +1806,7 @@ i915_do_wait_request(struct drm_device *dev, uint32_t seqno, int interruptible) u32 ier; int ret = 0; + struct intel_ring_buffer *ring = &dev_priv->render_ring; BUG_ON(seqno == 0); if (atomic_read(&dev_priv->mm.wedged)) @@ -1823,7 +1827,7 @@ i915_do_wait_request(struct drm_device *dev, uint32_t seqno, int interruptible) trace_i915_gem_request_wait_begin(dev, seqno); dev_priv->mm.waiting_gem_seqno = seqno; - i915_user_irq_get(dev); + ring->user_irq_get(dev, ring); if (interruptible) ret = wait_event_interruptible(dev_priv->irq_queue, i915_seqno_passed(i915_get_gem_seqno(dev), seqno) || @@ -1833,7 +1837,7 @@ i915_do_wait_request(struct drm_device *dev, uint32_t seqno, int interruptible) i915_seqno_passed(i915_get_gem_seqno(dev), seqno) || atomic_read(&dev_priv->mm.wedged)); - i915_user_irq_put(dev); + ring->user_irq_put(dev, ring); dev_priv->mm.waiting_gem_seqno = 0; trace_i915_gem_request_wait_end(dev, seqno); @@ -1867,6 +1871,19 @@ i915_wait_request(struct drm_device *dev, uint32_t seqno) } +static void +i915_gem_flush(struct drm_device *dev, + uint32_t invalidate_domains, + uint32_t flush_domains) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + if (flush_domains & I915_GEM_DOMAIN_CPU) + drm_agp_chipset_flush(dev); + dev_priv->render_ring.flush(dev, &dev_priv->render_ring, + invalidate_domains, + flush_domains); +} + /** * Ensures that all rendering to the object has completed and the object is * safe to unbind from the GTT or access from the CPU. @@ -3820,7 +3837,11 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data, #endif /* Exec the batchbuffer */ - ret = i915_dispatch_gem_execbuffer(dev, args, cliprects, exec_offset); + ret = dev_priv->render_ring.dispatch_gem_execbuffer(dev, + &dev_priv->render_ring, + args, + cliprects, + exec_offset); if (ret) { DRM_ERROR("dispatch failed %d\n", ret); goto err; @@ -4378,7 +4399,8 @@ i915_gem_idle(struct drm_device *dev) mutex_lock(&dev->struct_mutex); - if (dev_priv->mm.suspended || dev_priv->render_ring.ring_obj == NULL) { + if (dev_priv->mm.suspended || + dev_priv->render_ring.gem_object == NULL) { mutex_unlock(&dev->struct_mutex); return 0; } @@ -4420,7 +4442,7 @@ i915_gem_idle(struct drm_device *dev) * 965+ support PIPE_CONTROL commands, which provide finer grained control * over cache flushing. */ -int +static int i915_gem_init_pipe_control(struct drm_device *dev) { drm_i915_private_t *dev_priv = dev->dev_private; @@ -4459,7 +4481,8 @@ err: return ret; } -void + +static void i915_gem_cleanup_pipe_control(struct drm_device *dev) { drm_i915_private_t *dev_priv = dev->dev_private; @@ -4477,6 +4500,37 @@ i915_gem_cleanup_pipe_control(struct drm_device *dev) } int +i915_gem_init_ringbuffer(struct drm_device *dev) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + int ret; + dev_priv->render_ring = render_ring; + if (!I915_NEED_GFX_HWS(dev)) { + dev_priv->render_ring.status_page.page_addr + = dev_priv->status_page_dmah->vaddr; + memset(dev_priv->render_ring.status_page.page_addr, + 0, PAGE_SIZE); + } + if (HAS_PIPE_CONTROL(dev)) { + ret = i915_gem_init_pipe_control(dev); + if (ret) + return ret; + } + ret = intel_init_ring_buffer(dev, &dev_priv->render_ring); + return ret; +} + +void +i915_gem_cleanup_ringbuffer(struct drm_device *dev) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + + intel_cleanup_ring_buffer(dev, &dev_priv->render_ring); + if (HAS_PIPE_CONTROL(dev)) + i915_gem_cleanup_pipe_control(dev); +} + +int i915_gem_entervt_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv) { diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index dd91c97..e07c643 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -545,7 +545,8 @@ i915_ringbuffer_last_batch(struct drm_device *dev) } if (bbaddr == 0) { - ring = (u32 *)(dev_priv->render_ring.virtual_start + dev_priv->render_ring.Size); + ring = (u32 *)(dev_priv->render_ring.virtual_start + + dev_priv->render_ring.size); while (--ring >= (u32 *)dev_priv->render_ring.virtual_start) { bbaddr = i915_get_bbaddr(dev, ring); if (bbaddr) @@ -639,7 +640,8 @@ static void i915_capture_error_state(struct drm_device *dev) error->batchbuffer[1] = i915_error_object_create(dev, batchbuffer[1]); /* Record the ringbuffer */ - error->ringbuffer = i915_error_object_create(dev, dev_priv->render_ring.ring_obj); + error->ringbuffer = i915_error_object_create(dev, + dev_priv->render_ring.gem_object); /* Record buffers on the active list. */ error->active_bo = NULL; @@ -984,7 +986,6 @@ static int i915_emit_irq(struct drm_device * dev) { drm_i915_private_t *dev_priv = dev->dev_private; struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv; - RING_LOCALS; i915_kernel_lost_context(dev); @@ -1009,9 +1010,10 @@ static int i915_emit_irq(struct drm_device * dev) void i915_trace_irq_get(struct drm_device *dev, u32 seqno) { drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; + struct intel_ring_buffer *render_ring = &dev_priv->render_ring; if (dev_priv->trace_irq_seqno == 0) - i915_user_irq_get(dev); + render_ring->user_irq_get(dev, render_ring); dev_priv->trace_irq_seqno = seqno; } @@ -1021,6 +1023,7 @@ static int i915_wait_irq(struct drm_device * dev, int irq_nr) drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; struct drm_i915_master_private *master_priv = dev->primary->master->driver_priv; int ret = 0; + struct intel_ring_buffer *render_ring = &dev_priv->render_ring; DRM_DEBUG_DRIVER("irq_nr=%d breadcrumb=%d\n", irq_nr, READ_BREADCRUMB(dev_priv)); @@ -1034,10 +1037,10 @@ static int i915_wait_irq(struct drm_device * dev, int irq_nr) if (master_priv->sarea_priv) master_priv->sarea_priv->perf_boxes |= I915_BOX_WAIT; - i915_user_irq_get(dev); + render_ring->user_irq_get(dev, render_ring); DRM_WAIT_ON(ret, dev_priv->irq_queue, 3 * DRM_HZ, READ_BREADCRUMB(dev_priv) >= irq_nr); - i915_user_irq_put(dev); + render_ring->user_irq_put(dev, render_ring); if (ret == -EBUSY) { DRM_ERROR("EBUSY -- rec: %d emitted: %d\n", diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index f469a84..b867f3c 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -4629,7 +4629,6 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc, unsigned long flags; int pipesrc_reg = (intel_crtc->pipe == 0) ? PIPEASRC : PIPEBSRC; int ret, pipesrc; - RING_LOCALS; work = kzalloc(sizeof *work, GFP_KERNEL); if (work == NULL) diff --git a/drivers/gpu/drm/i915/intel_overlay.c b/drivers/gpu/drm/i915/intel_overlay.c index b0e17b0..93da837 100644 --- a/drivers/gpu/drm/i915/intel_overlay.c +++ b/drivers/gpu/drm/i915/intel_overlay.c @@ -211,9 +211,7 @@ static void intel_overlay_unmap_regs_atomic(struct intel_overlay *overlay) static int intel_overlay_on(struct intel_overlay *overlay) { struct drm_device *dev = overlay->dev; - drm_i915_private_t *dev_priv = dev->dev_private; int ret; - RING_LOCALS; BUG_ON(overlay->active); @@ -248,7 +246,6 @@ static void intel_overlay_continue(struct intel_overlay *overlay, drm_i915_private_t *dev_priv = dev->dev_private; u32 flip_addr = overlay->flip_addr; u32 tmp; - RING_LOCALS; BUG_ON(!overlay->active); @@ -274,7 +271,6 @@ static int intel_overlay_wait_flip(struct intel_overlay *overlay) drm_i915_private_t *dev_priv = dev->dev_private; int ret; u32 tmp; - RING_LOCALS; if (overlay->last_flip_req != 0) { ret = i915_do_wait_request(dev, overlay->last_flip_req, 1); @@ -314,9 +310,7 @@ static int intel_overlay_off(struct intel_overlay *overlay) { u32 flip_addr = overlay->flip_addr; struct drm_device *dev = overlay->dev; - drm_i915_private_t *dev_priv = dev->dev_private; int ret; - RING_LOCALS; BUG_ON(!overlay->active); @@ -390,11 +384,9 @@ int intel_overlay_recover_from_interrupt(struct intel_overlay *overlay, int interruptible) { struct drm_device *dev = overlay->dev; - drm_i915_private_t *dev_priv = dev->dev_private; struct drm_gem_object *obj; u32 flip_addr; int ret; - RING_LOCALS; if (overlay->hw_wedged == HW_WEDGED) return -EIO; diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index 06058dd..5715c4d8 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -29,30 +29,24 @@ #include "drmP.h" #include "drm.h" -#include "i915_drm.h" #include "i915_drv.h" +#include "i915_drm.h" #include "i915_trace.h" -#include "intel_drv.h" -void -i915_gem_flush(struct drm_device *dev, - uint32_t invalidate_domains, - uint32_t flush_domains) +static void +render_ring_flush(struct drm_device *dev, + struct intel_ring_buffer *ring, + u32 invalidate_domains, + u32 flush_domains) { - drm_i915_private_t *dev_priv = dev->dev_private; - uint32_t cmd; - RING_LOCALS; - #if WATCH_EXEC DRM_INFO("%s: invalidate %08x flush %08x\n", __func__, invalidate_domains, flush_domains); #endif - trace_i915_gem_request_flush(dev, dev_priv->mm.next_gem_seqno, + u32 cmd; + trace_i915_gem_request_flush(dev, ring->next_seqno, invalidate_domains, flush_domains); - if (flush_domains & I915_GEM_DOMAIN_CPU) - drm_agp_chipset_flush(dev); - if ((invalidate_domains | flush_domains) & I915_GEM_GPU_DOMAINS) { /* * read/write caches: @@ -100,19 +94,130 @@ i915_gem_flush(struct drm_device *dev, #if WATCH_EXEC DRM_INFO("%s: queue flush %08x to ring\n", __func__, cmd); #endif - BEGIN_LP_RING(2); - OUT_RING(cmd); - OUT_RING(MI_NOOP); - ADVANCE_LP_RING(); + intel_ring_begin(dev, ring, 8); + intel_ring_emit(dev, ring, cmd); + intel_ring_emit(dev, ring, MI_NOOP); + intel_ring_advance(dev, ring); } +} + +static unsigned int render_ring_get_head(struct drm_device *dev, + struct intel_ring_buffer *ring) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + return I915_READ(PRB0_HEAD) & HEAD_ADDR; +} +static unsigned int render_ring_get_tail(struct drm_device *dev, + struct intel_ring_buffer *ring) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + return I915_READ(PRB0_TAIL) & TAIL_ADDR; } + +static unsigned int render_ring_get_active_head(struct drm_device *dev, + struct intel_ring_buffer *ring) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + u32 acthd_reg = IS_I965G(dev) ? ACTHD_I965 : ACTHD; + + return I915_READ(acthd_reg); +} + +static void render_ring_advance_ring(struct drm_device *dev, + struct intel_ring_buffer *ring) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + I915_WRITE(PRB0_TAIL, ring->tail); +} + +static int init_ring_common(struct drm_device *dev, + struct intel_ring_buffer *ring) +{ + u32 head; + drm_i915_private_t *dev_priv = dev->dev_private; + struct drm_i915_gem_object *obj_priv; + obj_priv = to_intel_bo(ring->gem_object); + + /* Stop the ring if it's running. */ + I915_WRITE(ring->regs.ctl, 0); + I915_WRITE(ring->regs.head, 0); + I915_WRITE(ring->regs.tail, 0); + + /* Initialize the ring. */ + I915_WRITE(ring->regs.start, obj_priv->gtt_offset); + head = ring->get_head(dev, ring); + + /* G45 ring initialization fails to reset head to zero */ + if (head != 0) { + DRM_ERROR("%s head not reset to zero " + "ctl %08x head %08x tail %08x start %08x\n", + ring->name, + I915_READ(ring->regs.ctl), + I915_READ(ring->regs.head), + I915_READ(ring->regs.tail), + I915_READ(ring->regs.start)); + + I915_WRITE(ring->regs.head, 0); + + DRM_ERROR("%s head forced to zero " + "ctl %08x head %08x tail %08x start %08x\n", + ring->name, + I915_READ(ring->regs.ctl), + I915_READ(ring->regs.head), + I915_READ(ring->regs.tail), + I915_READ(ring->regs.start)); + } + + I915_WRITE(ring->regs.ctl, + ((ring->gem_object->size - PAGE_SIZE) & RING_NR_PAGES) + | RING_NO_REPORT | RING_VALID); + + head = I915_READ(ring->regs.head) & HEAD_ADDR; + /* If the head is still not zero, the ring is dead */ + if (head != 0) { + DRM_ERROR("%s initialization failed " + "ctl %08x head %08x tail %08x start %08x\n", + ring->name, + I915_READ(ring->regs.ctl), + I915_READ(ring->regs.head), + I915_READ(ring->regs.tail), + I915_READ(ring->regs.start)); + return -EIO; + } + + if (!drm_core_check_feature(dev, DRIVER_MODESET)) + i915_kernel_lost_context(dev); + else { + ring->head = ring->get_head(dev, ring); + ring->tail = ring->get_tail(dev, ring); + ring->space = ring->head - (ring->tail + 8); + if (ring->space < 0) + ring->space += ring->size; + } + return 0; +} + +static int init_render_ring(struct drm_device *dev, + struct intel_ring_buffer *ring) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + int ret = init_ring_common(dev, ring); + if (IS_I9XX(dev) && !IS_GEN3(dev)) { + I915_WRITE(MI_MODE, + (VS_TIMER_DISPATCH) << 16 | VS_TIMER_DISPATCH); + } + return ret; +} + #define PIPE_CONTROL_FLUSH(addr) \ +do { \ OUT_RING(GFX_OP_PIPE_CONTROL | PIPE_CONTROL_QW_WRITE | \ PIPE_CONTROL_DEPTH_STALL); \ OUT_RING(addr | PIPE_CONTROL_GLOBAL_GTT); \ OUT_RING(0); \ OUT_RING(0); \ +} while (0) /** * Creates a new sequence number, emitting a write of it to the status page @@ -122,21 +227,15 @@ i915_gem_flush(struct drm_device *dev, * * Returned sequence numbers are nonzero on success. */ -uint32_t -i915_ring_add_request(struct drm_device *dev) +static u32 +render_ring_add_request(struct drm_device *dev, + struct intel_ring_buffer *ring, + struct drm_file *file_priv, + u32 flush_domains) { + u32 seqno; drm_i915_private_t *dev_priv = dev->dev_private; - uint32_t seqno; - RING_LOCALS; - - /* Grab the seqno we're going to make this request be, and bump the - * next (skipping 0 so it can be the reserved no-seqno value). - */ - seqno = dev_priv->mm.next_gem_seqno; - dev_priv->mm.next_gem_seqno++; - if (dev_priv->mm.next_gem_seqno == 0) - dev_priv->mm.next_gem_seqno++; - + seqno = intel_ring_get_seqno(dev, ring); if (HAS_PIPE_CONTROL(dev)) { u32 scratch_addr = dev_priv->seqno_gfx_addr + 128; @@ -181,13 +280,26 @@ i915_ring_add_request(struct drm_device *dev) return seqno; } -void i915_user_irq_get(struct drm_device *dev) +static u32 +render_ring_get_gem_seqno(struct drm_device *dev, + struct intel_ring_buffer *ring) +{ + drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; + if (HAS_PIPE_CONTROL(dev)) + return ((volatile u32 *)(dev_priv->seqno_page))[0]; + else + return intel_read_status_page(ring, I915_GEM_HWS_INDEX); +} + +static void +render_ring_get_user_irq(struct drm_device *dev, + struct intel_ring_buffer *ring) { drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; unsigned long irqflags; spin_lock_irqsave(&dev_priv->user_irq_lock, irqflags); - if (dev->irq_enabled && (++dev_priv->user_irq_refcount == 1)) { + if (dev->irq_enabled && (++ring->user_irq_refcount == 1)) { if (HAS_PCH_SPLIT(dev)) ironlake_enable_graphics_irq(dev_priv, GT_PIPE_NOTIFY); else @@ -196,14 +308,16 @@ void i915_user_irq_get(struct drm_device *dev) spin_unlock_irqrestore(&dev_priv->user_irq_lock, irqflags); } -void i915_user_irq_put(struct drm_device *dev) +static void +render_ring_put_user_irq(struct drm_device *dev, + struct intel_ring_buffer *ring) { drm_i915_private_t *dev_priv = (drm_i915_private_t *) dev->dev_private; unsigned long irqflags; spin_lock_irqsave(&dev_priv->user_irq_lock, irqflags); - BUG_ON(dev->irq_enabled && dev_priv->user_irq_refcount <= 0); - if (dev->irq_enabled && (--dev_priv->user_irq_refcount == 0)) { + BUG_ON(dev->irq_enabled && ring->user_irq_refcount <= 0); + if (dev->irq_enabled && (--ring->user_irq_refcount == 0)) { if (HAS_PCH_SPLIT(dev)) ironlake_disable_graphics_irq(dev_priv, GT_PIPE_NOTIFY); else @@ -212,20 +326,31 @@ void i915_user_irq_put(struct drm_device *dev) spin_unlock_irqrestore(&dev_priv->user_irq_lock, irqflags); } -/** Dispatch a batchbuffer to the ring - */ -int -i915_dispatch_gem_execbuffer(struct drm_device *dev, - struct drm_i915_gem_execbuffer2 *exec, - struct drm_clip_rect *cliprects, - uint64_t exec_offset) +static void render_setup_status_page(struct drm_device *dev, + struct intel_ring_buffer *ring) +{ + drm_i915_private_t *dev_priv = dev->dev_private; + if (IS_GEN6(dev)) { + I915_WRITE(HWS_PGA_GEN6, ring->status_page.gfx_addr); + I915_READ(HWS_PGA_GEN6); /* posting read */ + } else { + I915_WRITE(HWS_PGA, ring->status_page.gfx_addr); + I915_READ(HWS_PGA); /* posting read */ + } + +} + +static int +render_ring_dispatch_gem_execbuffer(struct drm_device *dev, + struct intel_ring_buffer *ring, + struct drm_i915_gem_execbuffer2 *exec, + struct drm_clip_rect *cliprects, + uint64_t exec_offset) { drm_i915_private_t *dev_priv = dev->dev_private; int nbox = exec->num_cliprects; int i = 0, count; uint32_t exec_start, exec_len; - RING_LOCALS; - exec_start = (uint32_t) exec_offset + exec->batch_start_offset; exec_len = (uint32_t) exec->batch_len; @@ -242,74 +367,61 @@ i915_dispatch_gem_execbuffer(struct drm_device *dev, } if (IS_I830(dev) || IS_845G(dev)) { - BEGIN_LP_RING(4); - OUT_RING(MI_BATCH_BUFFER); - OUT_RING(exec_start | MI_BATCH_NON_SECURE); - OUT_RING(exec_start + exec_len - 4); - OUT_RING(0); - ADVANCE_LP_RING(); + intel_ring_begin(dev, ring, 4); + intel_ring_emit(dev, ring, MI_BATCH_BUFFER); + intel_ring_emit(dev, ring, + exec_start | MI_BATCH_NON_SECURE); + intel_ring_emit(dev, ring, exec_start + exec_len - 4); + intel_ring_emit(dev, ring, 0); } else { - BEGIN_LP_RING(2); + intel_ring_begin(dev, ring, 4); if (IS_I965G(dev)) { - OUT_RING(MI_BATCH_BUFFER_START | - (2 << 6) | - MI_BATCH_NON_SECURE_I965); - OUT_RING(exec_start); + intel_ring_emit(dev, ring, + MI_BATCH_BUFFER_START | (2 << 6) + | MI_BATCH_NON_SECURE_I965); + intel_ring_emit(dev, ring, exec_start); } else { - OUT_RING(MI_BATCH_BUFFER_START | - (2 << 6)); - OUT_RING(exec_start | MI_BATCH_NON_SECURE); + intel_ring_emit(dev, ring, MI_BATCH_BUFFER_START + | (2 << 6)); + intel_ring_emit(dev, ring, exec_start | + MI_BATCH_NON_SECURE); } - ADVANCE_LP_RING(); } + intel_ring_advance(dev, ring); } /* XXX breadcrumb */ return 0; } -static void -i915_gem_cleanup_hws(struct drm_device *dev) +static void cleanup_status_page(struct drm_device *dev, + struct intel_ring_buffer *ring) { drm_i915_private_t *dev_priv = dev->dev_private; struct drm_gem_object *obj; struct drm_i915_gem_object *obj_priv; - if (dev_priv->hws_obj == NULL) + obj = ring->status_page.obj; + if (obj == NULL) return; - - obj = dev_priv->hws_obj; obj_priv = to_intel_bo(obj); kunmap(obj_priv->pages[0]); i915_gem_object_unpin(obj); drm_gem_object_unreference(obj); - dev_priv->hws_obj = NULL; + ring->status_page.obj = NULL; memset(&dev_priv->hws_map, 0, sizeof(dev_priv->hws_map)); - dev_priv->hw_status_page = NULL; - - if (HAS_PIPE_CONTROL(dev)) - i915_gem_cleanup_pipe_control(dev); - - /* Write high address into HWS_PGA when disabling. */ - I915_WRITE(HWS_PGA, 0x1ffff000); } -static int -i915_gem_init_hws(struct drm_device *dev) +static int init_status_page(struct drm_device *dev, + struct intel_ring_buffer *ring) { drm_i915_private_t *dev_priv = dev->dev_private; struct drm_gem_object *obj; struct drm_i915_gem_object *obj_priv; int ret; - /* If we need a physical address for the status page, it's already - * initialized at driver load time. - */ - if (!I915_NEED_GFX_HWS(dev)) - return 0; - obj = i915_gem_alloc_object(dev, 4096); if (obj == NULL) { DRM_ERROR("Failed to allocate status page\n"); @@ -321,36 +433,21 @@ i915_gem_init_hws(struct drm_device *dev) ret = i915_gem_object_pin(obj, 4096); if (ret != 0) { - drm_gem_object_unreference(obj); goto err_unref; } - dev_priv->status_gfx_addr = obj_priv->gtt_offset; - - dev_priv->hw_status_page = kmap(obj_priv->pages[0]); - if (dev_priv->hw_status_page == NULL) { - DRM_ERROR("Failed to map status page.\n"); + ring->status_page.gfx_addr = obj_priv->gtt_offset; + ring->status_page.page_addr = kmap(obj_priv->pages[0]); + if (ring->status_page.page_addr == NULL) { memset(&dev_priv->hws_map, 0, sizeof(dev_priv->hws_map)); - ret = -EINVAL; goto err_unpin; } + ring->status_page.obj = obj; + memset(ring->status_page.page_addr, 0, PAGE_SIZE); - if (HAS_PIPE_CONTROL(dev)) { - ret = i915_gem_init_pipe_control(dev); - if (ret) - goto err_unpin; - } - - dev_priv->hws_obj = obj; - memset(dev_priv->hw_status_page, 0, PAGE_SIZE); - if (IS_GEN6(dev)) { - I915_WRITE(HWS_PGA_GEN6, dev_priv->status_gfx_addr); - I915_READ(HWS_PGA_GEN6); /* posting read */ - } else { - I915_WRITE(HWS_PGA, dev_priv->status_gfx_addr); - I915_READ(HWS_PGA); /* posting read */ - } - DRM_DEBUG_DRIVER("hws offset: 0x%08x\n", dev_priv->status_gfx_addr); + ring->setup_status_page(dev, ring); + DRM_DEBUG_DRIVER("%s hws offset: 0x%08x\n", + ring->name, ring->status_page.gfx_addr); return 0; @@ -359,43 +456,42 @@ err_unpin: err_unref: drm_gem_object_unreference(obj); err: - return 0; + return ret; } -int -i915_gem_init_ringbuffer(struct drm_device *dev) + +int intel_init_ring_buffer(struct drm_device *dev, + struct intel_ring_buffer *ring) { - drm_i915_private_t *dev_priv = dev->dev_private; - struct drm_gem_object *obj; - struct drm_i915_gem_object *obj_priv; - drm_i915_ring_buffer_t *ring = &dev_priv->render_ring; int ret; - u32 head; + struct drm_i915_gem_object *obj_priv; + struct drm_gem_object *obj; + ring->dev = dev; - ret = i915_gem_init_hws(dev); - if (ret != 0) - return ret; + if (I915_NEED_GFX_HWS(dev)) { + ret = init_status_page(dev, ring); + if (ret) + return ret; + } - obj = i915_gem_alloc_object(dev, 128 * 1024); + obj = i915_gem_alloc_object(dev, ring->size); if (obj == NULL) { DRM_ERROR("Failed to allocate ringbuffer\n"); - i915_gem_cleanup_hws(dev); - return -ENOMEM; + ret = -ENOMEM; + goto cleanup; } - obj_priv = to_intel_bo(obj); - ret = i915_gem_object_pin(obj, 4096); + ring->gem_object = obj; + + ret = i915_gem_object_pin(obj, ring->alignment); if (ret != 0) { drm_gem_object_unreference(obj); - i915_gem_cleanup_hws(dev); - return ret; + goto cleanup; } - /* Set up the kernel mapping for the ring. */ - ring->Size = obj->size; - + obj_priv = to_intel_bo(obj); + ring->map.size = ring->size; ring->map.offset = dev->agp->base + obj_priv->gtt_offset; - ring->map.size = obj->size; ring->map.type = 0; ring->map.flags = 0; ring->map.mtrr = 0; @@ -403,143 +499,85 @@ i915_gem_init_ringbuffer(struct drm_device *dev) drm_core_ioremap_wc(&ring->map, dev); if (ring->map.handle == NULL) { DRM_ERROR("Failed to map ringbuffer.\n"); - memset(&dev_priv->render_ring, 0, sizeof(dev_priv->render_ring)); i915_gem_object_unpin(obj); drm_gem_object_unreference(obj); - i915_gem_cleanup_hws(dev); - return -EINVAL; - } - ring->ring_obj = obj; - ring->virtual_start = ring->map.handle; - - /* Stop the ring if it's running. */ - I915_WRITE(PRB0_CTL, 0); - I915_WRITE(PRB0_TAIL, 0); - I915_WRITE(PRB0_HEAD, 0); - - /* Initialize the ring. */ - I915_WRITE(PRB0_START, obj_priv->gtt_offset); - head = I915_READ(PRB0_HEAD) & HEAD_ADDR; - - /* G45 ring initialization fails to reset head to zero */ - if (head != 0) { - DRM_ERROR("Ring head not reset to zero " - "ctl %08x head %08x tail %08x start %08x\n", - I915_READ(PRB0_CTL), - I915_READ(PRB0_HEAD), - I915_READ(PRB0_TAIL), - I915_READ(PRB0_START)); - I915_WRITE(PRB0_HEAD, 0); - - DRM_ERROR("Ring head forced to zero " - "ctl %08x head %08x tail %08x start %08x\n", - I915_READ(PRB0_CTL), - I915_READ(PRB0_HEAD), - I915_READ(PRB0_TAIL), - I915_READ(PRB0_START)); + ret = -EINVAL; + goto cleanup; } - I915_WRITE(PRB0_CTL, - ((obj->size - 4096) & RING_NR_PAGES) | - RING_NO_REPORT | - RING_VALID); - - head = I915_READ(PRB0_HEAD) & HEAD_ADDR; - - /* If the head is still not zero, the ring is dead */ - if (head != 0) { - DRM_ERROR("Ring initialization failed " - "ctl %08x head %08x tail %08x start %08x\n", - I915_READ(PRB0_CTL), - I915_READ(PRB0_HEAD), - I915_READ(PRB0_TAIL), - I915_READ(PRB0_START)); - return -EIO; + ring->virtual_start = ring->map.handle; + ret = ring->init(dev, ring); + if (ret != 0) { + intel_cleanup_ring_buffer(dev, ring); + return ret; } - /* Update our cache of the ring state */ if (!drm_core_check_feature(dev, DRIVER_MODESET)) i915_kernel_lost_context(dev); else { - ring->head = I915_READ(PRB0_HEAD) & HEAD_ADDR; - ring->tail = I915_READ(PRB0_TAIL) & TAIL_ADDR; + ring->head = ring->get_head(dev, ring); + ring->tail = ring->get_tail(dev, ring); ring->space = ring->head - (ring->tail + 8); if (ring->space < 0) - ring->space += ring->Size; + ring->space += ring->size; } - - if (IS_I9XX(dev) && !IS_GEN3(dev)) { - I915_WRITE(MI_MODE, - (VS_TIMER_DISPATCH) << 16 | VS_TIMER_DISPATCH); - } - - return 0; + INIT_LIST_HEAD(&ring->active_list); + INIT_LIST_HEAD(&ring->request_list); + return ret; +cleanup: + cleanup_status_page(dev, ring); + return ret; } -void -i915_gem_cleanup_ringbuffer(struct drm_device *dev) +void intel_cleanup_ring_buffer(struct drm_device *dev, + struct intel_ring_buffer *ring) { - drm_i915_private_t *dev_priv = dev->dev_private; - - if (dev_priv->render_ring.ring_obj == NULL) + if (ring->gem_object == NULL) return; - drm_core_ioremapfree(&dev_priv->render_ring.map, dev); - - i915_gem_object_unpin(dev_priv->render_ring.ring_obj); - drm_gem_object_unreference(dev_priv->render_ring.ring_obj); - dev_priv->render_ring.ring_obj = NULL; - memset(&dev_priv->render_ring, 0, sizeof(dev_priv->render_ring)); + drm_core_ioremapfree(&ring->map, dev); - i915_gem_cleanup_hws(dev); + i915_gem_object_unpin(ring->gem_object); + drm_gem_object_unreference(ring->gem_object); + ring->gem_object = NULL; + cleanup_status_page(dev, ring); } -/* As a ringbuffer is only allowed to wrap between instructions, fill - * the tail with NOOPs. - */ -int i915_wrap_ring(struct drm_device *dev) +int intel_wrap_ring_buffer(struct drm_device *dev, + struct intel_ring_buffer *ring) { - drm_i915_private_t *dev_priv = dev->dev_private; - volatile unsigned int *virt; + unsigned int *virt; int rem; + rem = ring->size - ring->tail; - rem = dev_priv->render_ring.Size - dev_priv->render_ring.tail; - if (dev_priv->render_ring.space < rem) { - int ret = i915_wait_ring(dev, rem, __func__); + if (ring->space < rem) { + int ret = intel_wait_ring_buffer(dev, ring, rem); if (ret) return ret; } - dev_priv->render_ring.space -= rem; - virt = (unsigned int *) - (dev_priv->render_ring.virtual_start + dev_priv->render_ring.tail); + virt = (unsigned int *)(ring->virtual_start + ring->tail); rem /= 4; while (rem--) *virt++ = MI_NOOP; - dev_priv->render_ring.tail = 0; + ring->tail = 0; return 0; } -int i915_wait_ring(struct drm_device * dev, int n, const char *caller) +int intel_wait_ring_buffer(struct drm_device *dev, + struct intel_ring_buffer *ring, int n) { - drm_i915_private_t *dev_priv = dev->dev_private; - drm_i915_ring_buffer_t *ring = &(dev_priv->render_ring); - u32 acthd_reg = IS_I965G(dev) ? ACTHD_I965 : ACTHD; - u32 last_acthd = I915_READ(acthd_reg); - u32 acthd; - u32 last_head = I915_READ(PRB0_HEAD) & HEAD_ADDR; - int i; + unsigned long end; trace_i915_ring_wait_begin (dev); - - for (i = 0; i < 100000; i++) { - ring->head = I915_READ(PRB0_HEAD) & HEAD_ADDR; - acthd = I915_READ(acthd_reg); + end = jiffies + 3 * HZ; + do { + ring->head = ring->get_head(dev, ring); ring->space = ring->head - (ring->tail + 8); if (ring->space < 0) - ring->space += ring->Size; + ring->space += ring->size; if (ring->space >= n) { trace_i915_ring_wait_end (dev); return 0; @@ -550,19 +588,97 @@ int i915_wait_ring(struct drm_device * dev, int n, const char *caller) if (master_priv->sarea_priv) master_priv->sarea_priv->perf_boxes |= I915_BOX_WAIT; } + yield(); + } while (!time_after(jiffies, end)); + trace_i915_ring_wait_end (dev); + return -EBUSY; +} +void intel_ring_begin(struct drm_device *dev, + struct intel_ring_buffer *ring, int n) +{ + if (unlikely(ring->tail + n > ring->size)) + intel_wrap_ring_buffer(dev, ring); + if (unlikely(ring->space < n)) + intel_wait_ring_buffer(dev, ring, n); +} - if (ring->head != last_head) - i = 0; - if (acthd != last_acthd) - i = 0; +void intel_ring_emit(struct drm_device *dev, + struct intel_ring_buffer *ring, unsigned int data) +{ + unsigned int *virt = ring->virtual_start + ring->tail; + *virt = data; + ring->tail += 4; + ring->tail &= ring->size - 1; + ring->space -= 4; +} - last_head = ring->head; - last_acthd = acthd; - msleep_interruptible(10); +void intel_ring_advance(struct drm_device *dev, + struct intel_ring_buffer *ring) +{ + ring->advance_ring(dev, ring); +} - } +void intel_fill_struct(struct drm_device *dev, + struct intel_ring_buffer *ring, + void *data, + unsigned int len) +{ + unsigned int *virt = ring->virtual_start + ring->tail; + BUG_ON((len&~(4-1)) != 0); + intel_ring_begin(dev, ring, len); + memcpy(virt, data, len); + ring->tail += len; + ring->tail &= ring->size - 1; + ring->space -= len; + intel_ring_advance(dev, ring); +} - trace_i915_ring_wait_end (dev); - return -EBUSY; +u32 intel_ring_get_seqno(struct drm_device *dev, + struct intel_ring_buffer *ring) +{ + u32 seqno; + seqno = ring->next_seqno; + + /* reserve 0 for non-seqno */ + if (++ring->next_seqno == 0) + ring->next_seqno = 1; + return seqno; } + +struct intel_ring_buffer render_ring = { + .name = "render ring", + .regs = { + .ctl = PRB0_CTL, + .head = PRB0_HEAD, + .tail = PRB0_TAIL, + .start = PRB0_START + }, + .ring_flag = I915_EXEC_RENDER, + .size = 32 * PAGE_SIZE, + .alignment = PAGE_SIZE, + .virtual_start = NULL, + .dev = NULL, + .gem_object = NULL, + .head = 0, + .tail = 0, + .space = 0, + .next_seqno = 1, + .user_irq_refcount = 0, + .irq_gem_seqno = 0, + .waiting_gem_seqno = 0, + .setup_status_page = render_setup_status_page, + .init = init_render_ring, + .get_head = render_ring_get_head, + .get_tail = render_ring_get_tail, + .get_active_head = render_ring_get_active_head, + .advance_ring = render_ring_advance_ring, + .flush = render_ring_flush, + .add_request = render_ring_add_request, + .get_gem_seqno = render_ring_get_gem_seqno, + .user_irq_get = render_ring_get_user_irq, + .user_irq_put = render_ring_put_user_irq, + .dispatch_gem_execbuffer = render_ring_dispatch_gem_execbuffer, + .status_page = {NULL, 0, NULL}, + .map = {0,} +}; diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h new file mode 100644 index 0000000..d5568d3 --- /dev/null +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -0,0 +1,124 @@ +#ifndef _INTEL_RINGBUFFER_H_ +#define _INTEL_RINGBUFFER_H_ + +struct intel_hw_status_page { + void *page_addr; + unsigned int gfx_addr; + struct drm_gem_object *obj; +}; + +struct drm_i915_gem_execbuffer2; +struct intel_ring_buffer { + const char *name; + struct ring_regs { + u32 ctl; + u32 head; + u32 tail; + u32 start; + } regs; + unsigned int ring_flag; + unsigned long size; + unsigned int alignment; + void *virtual_start; + struct drm_device *dev; + struct drm_gem_object *gem_object; + + unsigned int head; + unsigned int tail; + unsigned int space; + u32 next_seqno; + struct intel_hw_status_page status_page; + + u32 irq_gem_seqno; /* last seq seem at irq time */ + u32 waiting_gem_seqno; + int user_irq_refcount; + void (*user_irq_get)(struct drm_device *dev, + struct intel_ring_buffer *ring); + void (*user_irq_put)(struct drm_device *dev, + struct intel_ring_buffer *ring); + void (*setup_status_page)(struct drm_device *dev, + struct intel_ring_buffer *ring); + + int (*init)(struct drm_device *dev, + struct intel_ring_buffer *ring); + + unsigned int (*get_head)(struct drm_device *dev, + struct intel_ring_buffer *ring); + unsigned int (*get_tail)(struct drm_device *dev, + struct intel_ring_buffer *ring); + unsigned int (*get_active_head)(struct drm_device *dev, + struct intel_ring_buffer *ring); + void (*advance_ring)(struct drm_device *dev, + struct intel_ring_buffer *ring); + void (*flush)(struct drm_device *dev, + struct intel_ring_buffer *ring, + u32 invalidate_domains, + u32 flush_domains); + u32 (*add_request)(struct drm_device *dev, + struct intel_ring_buffer *ring, + struct drm_file *file_priv, + u32 flush_domains); + u32 (*get_gem_seqno)(struct drm_device *dev, + struct intel_ring_buffer *ring); + int (*dispatch_gem_execbuffer)(struct drm_device *dev, + struct intel_ring_buffer *ring, + struct drm_i915_gem_execbuffer2 *exec, + struct drm_clip_rect *cliprects, + uint64_t exec_offset); + + /** + * List of objects currently involved in rendering from the + * ringbuffer. + * + * Includes buffers having the contents of their GPU caches + * flushed, not necessarily primitives. last_rendering_seqno + * represents when the rendering involved will be completed. + * + * A reference is held on the buffer while on this list. + */ + struct list_head active_list; + + /** + * List of breadcrumbs associated with GPU requests currently + * outstanding. + */ + struct list_head request_list; + + wait_queue_head_t irq_queue; + drm_local_map_t map; +}; + +static inline u32 +intel_read_status_page(struct intel_ring_buffer *ring, + int reg) +{ + u32 *regs = ring->status_page.page_addr; + return regs[reg]; +} + +int intel_init_ring_buffer(struct drm_device *dev, + struct intel_ring_buffer *ring); +void intel_cleanup_ring_buffer(struct drm_device *dev, + struct intel_ring_buffer *ring); +int intel_wait_ring_buffer(struct drm_device *dev, + struct intel_ring_buffer *ring, int n); +int intel_wrap_ring_buffer(struct drm_device *dev, + struct intel_ring_buffer *ring); +void intel_ring_begin(struct drm_device *dev, + struct intel_ring_buffer *ring, int n); +void intel_ring_emit(struct drm_device *dev, + struct intel_ring_buffer *ring, u32 data); +void intel_fill_struct(struct drm_device *dev, + struct intel_ring_buffer *ring, + void *data, + unsigned int len); +void intel_ring_advance(struct drm_device *dev, + struct intel_ring_buffer *ring); + +u32 intel_ring_get_seqno(struct drm_device *dev, + struct intel_ring_buffer *ring); + +extern struct intel_ring_buffer render_ring; +extern struct intel_ring_buffer bsd_ring; + +#endif /* _INTEL_RINGBUFFER_H_ */ diff --git a/include/drm/i915_drm.h b/include/drm/i915_drm.h index b64a8d7..e916870 100644 --- a/include/drm/i915_drm.h +++ b/include/drm/i915_drm.h @@ -616,7 +616,9 @@ struct drm_i915_gem_execbuffer2 { __u32 num_cliprects; /** This is a struct drm_clip_rect *cliprects */ __u64 cliprects_ptr; - __u64 flags; /* currently unused */ +#define I915_EXEC_RENDER (1<<0) +#define I915_EXEC_BSD (1<<1) + __u64 flags; __u64 rsvd1; __u64 rsvd2; }; |