[PATCH 3/3] drm/omap: DMM based hardware scrolling console

From: Rob Clark rob@ti.com

Add support for YWRAP scrolling by shuffling pages around in DMM instead of sw blits.

Note that fbcon only utilizes this mode if the y resolution is divided evenly by the font height. So, for example, a 1920x1080 display using a 16 pixel tall font will not utilize this, but a 1280x1024 display would.

Signed-off-by: Rob Clark rob@ti.com --- drivers/staging/omapdrm/omap_dmm_priv.h | 4 +- drivers/staging/omapdrm/omap_dmm_tiler.c | 35 +++++++++----- drivers/staging/omapdrm/omap_dmm_tiler.h | 3 +- drivers/staging/omapdrm/omap_drv.h | 3 + drivers/staging/omapdrm/omap_fb.c | 22 +-------- drivers/staging/omapdrm/omap_fbdev.c | 73 +++++++++++++++++++++++++++-- drivers/staging/omapdrm/omap_gem.c | 59 +++++++++++++++++++++--- 7 files changed, 153 insertions(+), 46 deletions(-)

diff --git a/drivers/staging/omapdrm/omap_dmm_priv.h b/drivers/staging/omapdrm/omap_dmm_priv.h index 65b990c..2f529ab 100644 --- a/drivers/staging/omapdrm/omap_dmm_priv.h +++ b/drivers/staging/omapdrm/omap_dmm_priv.h @@ -79,12 +79,12 @@ #define DMM_PATSTATUS_ERR_UPD_DATA (1<<14) #define DMM_PATSTATUS_ERR_ACCESS (1<<15)

+/* note: don't treat DMM_PATSTATUS_ERR_ACCESS as an error */ #define DMM_PATSTATUS_ERR (DMM_PATSTATUS_ERR_INV_DESCR | \ DMM_PATSTATUS_ERR_INV_DATA | \ DMM_PATSTATUS_ERR_UPD_AREA | \ DMM_PATSTATUS_ERR_UPD_CTRL | \ - DMM_PATSTATUS_ERR_UPD_DATA | \ - DMM_PATSTATUS_ERR_ACCESS) + DMM_PATSTATUS_ERR_UPD_DATA)

diff --git a/drivers/staging/omapdrm/omap_dmm_tiler.c b/drivers/staging/omapdrm/omap_dmm_tiler.c index 9ed5215..b182de5 100644 --- a/drivers/staging/omapdrm/omap_dmm_tiler.c +++ b/drivers/staging/omapdrm/omap_dmm_tiler.c @@ -170,7 +170,7 @@ static struct dmm_txn *dmm_txn_init(struct dmm *dmm, struct tcm *tcm) * corresponding slot is cleared (ie. dummy_pa is programmed) */ static int dmm_txn_append(struct dmm_txn *txn, struct pat_area *area, - struct page **pages) + struct page **pages, uint32_t npages, uint32_t roll) { dma_addr_t pat_pa = 0; uint32_t *data; @@ -197,8 +197,11 @@ static int dmm_txn_append(struct dmm_txn *txn, struct pat_area *area, data = alloc_dma(txn, 4*i, &pat->data_pa);

while (i--) { - data[i] = (pages && pages[i]) ? - page_to_phys(pages[i]) : engine->dmm->dummy_pa; + int n = i + roll; + if (n >= npages) + n -= npages; + data[i] = (pages && pages[n]) ? + page_to_phys(pages[n]) : engine->dmm->dummy_pa; }

/* fill in lut with new addresses */ @@ -262,7 +265,8 @@ cleanup: /* * DMM programming */ -static int fill(struct tcm_area *area, struct page **pages, bool wait) +static int fill(struct tcm_area *area, struct page **pages, + uint32_t npages, uint32_t roll, bool wait) { int ret = 0; struct tcm_area slice, area_s; @@ -278,12 +282,11 @@ static int fill(struct tcm_area *area, struct page **pages, bool wait) .x1 = slice.p1.x, .y1 = slice.p1.y, };

- ret = dmm_txn_append(txn, &p_area, pages); + ret = dmm_txn_append(txn, &p_area, pages, npages, roll); if (ret) goto fail;

- if (pages) - pages += tcm_sizeof(slice); + roll += tcm_sizeof(slice); }

ret = dmm_txn_commit(txn, wait); @@ -298,11 +301,12 @@ fail:

/* note: slots for which pages[i] == NULL are filled w/ dummy page */ -int tiler_pin(struct tiler_block *block, struct page **pages, bool wait) +int tiler_pin(struct tiler_block *block, struct page **pages, + uint32_t npages, uint32_t roll, bool wait) { int ret;

- ret = fill(&block->area, pages, wait); + ret = fill(&block->area, pages, npages, roll, wait);

if (ret) tiler_unpin(block); @@ -312,7 +316,7 @@ int tiler_pin(struct tiler_block *block, struct page **pages, bool wait)

int tiler_unpin(struct tiler_block *block) { - return fill(&block->area, NULL, false); + return fill(&block->area, NULL, 0, 0, false); }

/* @@ -558,8 +562,13 @@ int omap_dmm_init(struct drm_device *dev) goto fail; }

- /* enable some interrupts! */ - writel(0xfefefefe, omap_dmm->base + DMM_PAT_IRQENABLE_SET); + /* Enable all interrupts for each refill engine except + * ERR_LUT_MISS<n> (which is just advisory, and we don't care + * about because we want to be able to refill live scanout + * buffers for accelerated pan/scroll) and FILL_DSC<n> which + * we just generally don't care about. + */ + writel(0x7e7e7e7e, omap_dmm->base + DMM_PAT_IRQENABLE_SET);

lut_table_size = omap_dmm->lut_width * omap_dmm->lut_height * omap_dmm->num_lut; @@ -658,7 +667,7 @@ int omap_dmm_init(struct drm_device *dev) /* initialize all LUTs to dummy page entries */ for (i = 0; i < omap_dmm->num_lut; i++) { area.tcm = omap_dmm->tcm[i]; - if (fill(&area, NULL, true)) + if (fill(&area, NULL, 0, 0, true)) dev_err(omap_dmm->dev, "refill failed"); }

diff --git a/drivers/staging/omapdrm/omap_dmm_tiler.h b/drivers/staging/omapdrm/omap_dmm_tiler.h index 7e63b6b..58aa046 100644 --- a/drivers/staging/omapdrm/omap_dmm_tiler.h +++ b/drivers/staging/omapdrm/omap_dmm_tiler.h @@ -77,7 +77,8 @@ int omap_dmm_init(struct drm_device *dev); int omap_dmm_remove(void);

/* pin/unpin */ -int tiler_pin(struct tiler_block *block, struct page **pages, bool wait); +int tiler_pin(struct tiler_block *block, struct page **pages, + uint32_t npages, uint32_t roll, bool wait); int tiler_unpin(struct tiler_block *block);

/* reserve/release */ diff --git a/drivers/staging/omapdrm/omap_drv.h b/drivers/staging/omapdrm/omap_drv.h index 9d0783d..263057a 100644 --- a/drivers/staging/omapdrm/omap_drv.h +++ b/drivers/staging/omapdrm/omap_drv.h @@ -47,6 +47,8 @@ struct omap_drm_private { struct drm_connector *connectors[8];

struct drm_fb_helper *fbdev; + + bool has_dmm; };

struct drm_fb_helper *omap_fbdev_init(struct drm_device *dev); @@ -107,6 +109,7 @@ int omap_gem_op_finish(struct drm_gem_object *obj, enum omap_gem_op op); int omap_gem_op_sync(struct drm_gem_object *obj, enum omap_gem_op op); int omap_gem_op_async(struct drm_gem_object *obj, enum omap_gem_op op, void (*fxn)(void *arg), void *arg); +int omap_gem_roll(struct drm_gem_object *obj, uint32_t roll); int omap_gem_get_paddr(struct drm_gem_object *obj, dma_addr_t *paddr, bool remap); int omap_gem_put_paddr(struct drm_gem_object *obj); diff --git a/drivers/staging/omapdrm/omap_fb.c b/drivers/staging/omapdrm/omap_fb.c index 3f62505..0b50c5b 100644 --- a/drivers/staging/omapdrm/omap_fb.c +++ b/drivers/staging/omapdrm/omap_fb.c @@ -218,25 +218,9 @@ struct drm_framebuffer *omap_framebuffer_init(struct drm_device *dev,

size = PAGE_ALIGN(mode_cmd->pitch * mode_cmd->height);

- if (bo) { - DBG("using existing %d byte buffer (needed %d)", bo->size, size); - if (size > bo->size) { - dev_err(dev->dev, "provided buffer object is too small!\n"); - goto fail; - } - } else { - /* for convenience of all the various callers who don't want - * to be bothered to allocate their own buffer.. - */ - union omap_gem_size gsize = { - .bytes = size, - }; - DBG("allocating %d bytes for fb %d", size, dev->primary->index); - bo = omap_gem_new(dev, gsize, OMAP_BO_SCANOUT | OMAP_BO_WC); - if (!bo) { - dev_err(dev->dev, "failed to allocate buffer object\n"); - goto fail; - } + if (size > bo->size) { + dev_err(dev->dev, "provided buffer object is too small!\n"); + goto fail; }

omap_fb->bo = bo; diff --git a/drivers/staging/omapdrm/omap_fbdev.c b/drivers/staging/omapdrm/omap_fbdev.c index 048077c..d8962e8 100644 --- a/drivers/staging/omapdrm/omap_fbdev.c +++ b/drivers/staging/omapdrm/omap_fbdev.c @@ -31,9 +31,11 @@ struct omap_fbdev { struct drm_fb_helper base; struct drm_framebuffer *fb; + struct drm_gem_object *bo; };

static void omap_fbdev_flush(struct fb_info *fbi, int x, int y, int w, int h); +static struct drm_fb_helper *get_fb(struct fb_info *fbi);

static ssize_t omap_fbdev_write(struct fb_info *fbi, const char __user *buf, size_t count, loff_t *ppos) @@ -68,6 +70,31 @@ static void omap_fbdev_imageblit(struct fb_info *fbi, image->width, image->height); }

+static int omap_fbdev_pan_display(struct fb_var_screeninfo *var, + struct fb_info *fbi) +{ + struct drm_fb_helper *helper = get_fb(fbi); + struct omap_fbdev *fbdev = to_omap_fbdev(helper); + struct omap_drm_private *priv; + int npages; + + if (!helper) + goto fallback; + + priv = helper->dev->dev_private; + if (!priv->has_dmm) + goto fallback; + + /* DMM roll shifts in 4K pages: */ + npages = fbi->fix.line_length >> PAGE_SHIFT; + omap_gem_roll(fbdev->bo, var->yoffset * npages); + + return 0; + +fallback: + return drm_fb_helper_pan_display(var, fbi); +} + static struct fb_ops omap_fb_ops = { .owner = THIS_MODULE,

@@ -82,7 +109,7 @@ static struct fb_ops omap_fb_ops = {

.fb_check_var = drm_fb_helper_check_var, .fb_set_par = drm_fb_helper_set_par, - .fb_pan_display = drm_fb_helper_pan_display, + .fb_pan_display = omap_fbdev_pan_display, .fb_blank = drm_fb_helper_blank, .fb_setcmap = drm_fb_helper_setcmap,

@@ -95,7 +122,9 @@ static int omap_fbdev_create(struct drm_fb_helper *helper, { struct omap_fbdev *fbdev = to_omap_fbdev(helper); struct drm_device *dev = helper->dev; + struct omap_drm_private *priv = dev->dev_private; struct drm_framebuffer *fb = NULL; + union omap_gem_size gsize; struct fb_info *fbi = NULL; struct drm_mode_fb_cmd mode_cmd = {0}; dma_addr_t paddr; @@ -109,8 +138,9 @@ static int omap_fbdev_create(struct drm_fb_helper *helper, sizes->surface_bpp = 32; sizes->surface_depth = 32;

- DBG("create fbdev: %dx%d@%d", sizes->surface_width, - sizes->surface_height, sizes->surface_bpp); + DBG("create fbdev: %dx%d@%d (%dx%d)", sizes->surface_width, + sizes->surface_height, sizes->surface_bpp, + sizes->fb_width, sizes->fb_height);

mode_cmd.width = sizes->surface_width; mode_cmd.height = sizes->surface_height; @@ -118,7 +148,27 @@ static int omap_fbdev_create(struct drm_fb_helper *helper, mode_cmd.bpp = sizes->surface_bpp; mode_cmd.depth = sizes->surface_depth;

- fb = omap_framebuffer_init(dev, &mode_cmd, NULL); + mode_cmd.pitch = align_pitch( + mode_cmd.width * ((mode_cmd.bpp + 7) / 8), + mode_cmd.width, mode_cmd.bpp); + + if (priv->has_dmm) { + /* need to align pitch to page size if using DMM scrolling */ + mode_cmd.pitch = ALIGN(mode_cmd.pitch, PAGE_SIZE); + } + + /* allocate backing bo */ + gsize = (union omap_gem_size){ + .bytes = PAGE_ALIGN(mode_cmd.pitch * mode_cmd.height), + }; + DBG("allocating %d bytes for fb %d", gsize.bytes, dev->primary->index); + fbdev->bo = omap_gem_new(dev, gsize, OMAP_BO_SCANOUT | OMAP_BO_WC); + if (!fbdev->bo) { + dev_err(dev->dev, "failed to allocate buffer object\n"); + goto fail; + } + + fb = omap_framebuffer_init(dev, &mode_cmd, fbdev->bo); if (!fb) { dev_err(dev->dev, "failed to allocate fb\n"); ret = -ENOMEM; @@ -153,7 +203,7 @@ static int omap_fbdev_create(struct drm_fb_helper *helper, }

drm_fb_helper_fill_fix(fbi, fb->pitch, fb->depth); - drm_fb_helper_fill_var(fbi, helper, fb->width, fb->height); + drm_fb_helper_fill_var(fbi, helper, sizes->fb_width, sizes->fb_height);

size = omap_framebuffer_get_buffer(fb, 0, 0, &vaddr, &paddr, &screen_width); @@ -165,6 +215,15 @@ static int omap_fbdev_create(struct drm_fb_helper *helper, fbi->fix.smem_start = paddr; fbi->fix.smem_len = size;

+ /* if we have DMM, then we can use it for scrolling by just + * shuffling pages around in DMM rather than doing sw blit. + */ + if (priv->has_dmm) { + DRM_INFO("Enabling DMM ywrap scrolling\n"); + fbi->flags |= FBINFO_HWACCEL_YWRAP | FBINFO_READS_FAST; + fbi->fix.ywrapstep = 1; + } + DBG("par=%p, %dx%d", fbi->par, fbi->var.xres, fbi->var.yres); DBG("allocated %dx%d fb", fbdev->fb->width, fbdev->fb->height);

@@ -300,5 +359,9 @@ void omap_fbdev_free(struct drm_device *dev)

kfree(fbdev);

+ /* this will free the backing object */ + if (fbdev->fb) + fbdev->fb->funcs->destroy(fbdev->fb); + priv->fbdev = NULL; } diff --git a/drivers/staging/omapdrm/omap_gem.c b/drivers/staging/omapdrm/omap_gem.c index a451c57..9684891 100644 --- a/drivers/staging/omapdrm/omap_gem.c +++ b/drivers/staging/omapdrm/omap_gem.c @@ -50,6 +50,9 @@ struct omap_gem_object { /** width/height for tiled formats (rounded up to slot boundaries) */ uint16_t width, height;

+ /** roll applied when mapping to DMM */ + uint32_t roll; + /** * If buffer is allocated physically contiguous, the OMAP_BO_DMA flag * is set and the paddr is valid. Also if the buffer is remapped in @@ -338,7 +341,7 @@ static int fault_2d(struct drm_gem_object *obj, memset(pages + slots, 0, sizeof(struct page *) * (usergart[fmt].height - slots));

- ret = tiler_pin(entry->block, pages, true); + ret = tiler_pin(entry->block, pages, ARRAY_SIZE(pages), 0, true); if (ret) { dev_err(obj->dev->dev, "failed to pin: %d\n", ret); return ret; @@ -521,6 +524,41 @@ fail: return ret; }

+/* Set scrolling position. This allows us to implement fast scrolling + * for console. + */ +int omap_gem_roll(struct drm_gem_object *obj, uint32_t roll) +{ + struct omap_gem_object *omap_obj = to_omap_bo(obj); + uint32_t npages = obj->size >> PAGE_SHIFT; + int ret = 0; + + if (roll > npages) { + dev_err(obj->dev->dev, "invalid roll: %d\n", roll); + return -EINVAL; + } + + mutex_lock(&obj->dev->struct_mutex); + + omap_obj->roll = roll; + + /* if we aren't mapped yet, we don't need to do anything */ + if (omap_obj->block) { + struct page **pages; + ret = get_pages(obj, &pages); + if (ret) + goto fail; + ret = tiler_pin(omap_obj->block, pages, npages, roll, true); + if (ret) + dev_err(obj->dev->dev, "could not repin: %d\n", ret); + } + +fail: + mutex_unlock(&obj->dev->struct_mutex); + + return ret; +} + /* Get physical address for DMA.. if 'remap' is true, and the buffer is not * already contiguous, remap it to pin in physically contiguous memory.. (ie. * map in TILER) @@ -528,23 +566,25 @@ fail: int omap_gem_get_paddr(struct drm_gem_object *obj, dma_addr_t *paddr, bool remap) { + struct omap_drm_private *priv = obj->dev->dev_private; struct omap_gem_object *omap_obj = to_omap_bo(obj); int ret = 0;

mutex_lock(&obj->dev->struct_mutex);

- if (remap && is_shmem(obj)) { + if (remap && is_shmem(obj) && priv->has_dmm) { if (omap_obj->paddr_cnt == 0) { struct page **pages; + uint32_t npages = obj->size >> PAGE_SHIFT; enum tiler_fmt fmt = gem2fmt(omap_obj->flags); struct tiler_block *block; + BUG_ON(omap_obj->block);

ret = get_pages(obj, &pages); if (ret) goto fail;

- if (omap_obj->flags & OMAP_BO_TILED) { block = tiler_reserve_2d(fmt, omap_obj->width, @@ -561,7 +601,8 @@ int omap_gem_get_paddr(struct drm_gem_object *obj, }

/* TODO: enable async refill.. */ - ret = tiler_pin(block, pages, true); + ret = tiler_pin(block, pages, npages, + omap_obj->roll, true); if (ret) { tiler_release(block); dev_err(obj->dev->dev, @@ -1002,6 +1043,7 @@ int omap_gem_new_handle(struct drm_device *dev, struct drm_file *file, struct drm_gem_object *omap_gem_new(struct drm_device *dev, union omap_gem_size gsize, uint32_t flags) { + struct omap_drm_private *priv = dev->dev_private; struct omap_gem_object *omap_obj; struct drm_gem_object *obj = NULL; size_t size; @@ -1043,8 +1085,10 @@ struct drm_gem_object *omap_gem_new(struct drm_device *dev,

obj = &omap_obj->base;

- if (flags & OMAP_BO_SCANOUT) { - /* attempt to allocate contiguous memory */ + if ((flags & OMAP_BO_SCANOUT) && !priv->has_dmm) { + /* attempt to allocate contiguous memory if we don't + * have DMM for remappign discontiguous buffers + */ omap_obj->vaddr = dma_alloc_writecombine(dev->dev, size, &omap_obj->paddr, GFP_KERNEL); if (omap_obj->vaddr) { @@ -1081,6 +1125,7 @@ fail: /* init/cleanup.. if DMM is used, we need to set some stuff up.. */ void omap_gem_init(struct drm_device *dev) { + struct omap_drm_private *priv = dev->dev_private; const enum tiler_fmt fmts[] = { TILFMT_8BIT, TILFMT_16BIT, TILFMT_32BIT }; @@ -1130,6 +1175,8 @@ void omap_gem_init(struct drm_device *dev) usergart[i].stride_pfn << PAGE_SHIFT); } } + + priv->has_dmm = true; }

void omap_gem_deinit(struct drm_device *dev)