diff --git a/drivers/gpu/drm/udl/udl_drv.h b/drivers/gpu/drm/udl/udl_drv.h
index 072582570a4fa4508a18585fa95c6514f08d64a1..e9e9b1ff678ee0a81d0d4b100b816b19122c7f0c 100644
--- a/drivers/gpu/drm/udl/udl_drv.h
+++ b/drivers/gpu/drm/udl/udl_drv.h
@@ -113,7 +113,7 @@ udl_fb_user_fb_create(struct drm_device *dev,
 		      struct drm_file *file,
 		      const struct drm_mode_fb_cmd2 *mode_cmd);
 
-int udl_render_hline(struct drm_device *dev, int bpp, struct urb **urb_ptr,
+int udl_render_hline(struct drm_device *dev, int log_bpp, struct urb **urb_ptr,
 		     const char *front, char **urb_buf_ptr,
 		     u32 byte_offset, u32 device_byte_offset, u32 byte_width,
 		     int *ident_ptr, int *sent_ptr);
diff --git a/drivers/gpu/drm/udl/udl_fb.c b/drivers/gpu/drm/udl/udl_fb.c
index d5583190f3e44de77560d76c2d5a5a4cf7db62de..dbb62f6eb48a5fba107acd9981b3cfae45f8f6a4 100644
--- a/drivers/gpu/drm/udl/udl_fb.c
+++ b/drivers/gpu/drm/udl/udl_fb.c
@@ -90,7 +90,10 @@ int udl_handle_damage(struct udl_framebuffer *fb, int x, int y,
 	int bytes_identical = 0;
 	struct urb *urb;
 	int aligned_x;
-	int bpp = fb->base.format->cpp[0];
+	int log_bpp;
+
+	BUG_ON(!is_power_of_2(fb->base.format->cpp[0]));
+	log_bpp = __ffs(fb->base.format->cpp[0]);
 
 	if (!fb->active_16)
 		return 0;
@@ -125,12 +128,12 @@ int udl_handle_damage(struct udl_framebuffer *fb, int x, int y,
 
 	for (i = y; i < y + height ; i++) {
 		const int line_offset = fb->base.pitches[0] * i;
-		const int byte_offset = line_offset + (x * bpp);
-		const int dev_byte_offset = (fb->base.width * bpp * i) + (x * bpp);
-		if (udl_render_hline(dev, bpp, &urb,
+		const int byte_offset = line_offset + (x << log_bpp);
+		const int dev_byte_offset = (fb->base.width * i + x) << log_bpp;
+		if (udl_render_hline(dev, log_bpp, &urb,
 				     (char *) fb->obj->vmapping,
 				     &cmd, byte_offset, dev_byte_offset,
-				     width * bpp,
+				     width << log_bpp,
 				     &bytes_identical, &bytes_sent))
 			goto error;
 	}
@@ -149,7 +152,7 @@ int udl_handle_damage(struct udl_framebuffer *fb, int x, int y,
 error:
 	atomic_add(bytes_sent, &udl->bytes_sent);
 	atomic_add(bytes_identical, &udl->bytes_identical);
-	atomic_add(width*height*bpp, &udl->bytes_rendered);
+	atomic_add((width * height) << log_bpp, &udl->bytes_rendered);
 	end_cycles = get_cycles();
 	atomic_add(((unsigned int) ((end_cycles - start_cycles)
 		    >> 10)), /* Kcycles */
@@ -175,7 +178,7 @@ static int udl_fb_mmap(struct fb_info *info, struct vm_area_struct *vma)
 
 	pos = (unsigned long)info->fix.smem_start + offset;
 
-	pr_notice("mmap() framebuffer addr:%lu size:%lu\n",
+	pr_debug("mmap() framebuffer addr:%lu size:%lu\n",
 		  pos, size);
 
 	/* We don't want the framebuffer to be mapped encrypted */
@@ -221,7 +224,7 @@ static int udl_fb_open(struct fb_info *info, int user)
 
 		struct fb_deferred_io *fbdefio;
 
-		fbdefio = kmalloc(sizeof(struct fb_deferred_io), GFP_KERNEL);
+		fbdefio = kzalloc(sizeof(struct fb_deferred_io), GFP_KERNEL);
 
 		if (fbdefio) {
 			fbdefio->delay = DL_DEFIO_WRITE_DELAY;
@@ -233,7 +236,7 @@ static int udl_fb_open(struct fb_info *info, int user)
 	}
 #endif
 
-	pr_notice("open /dev/fb%d user=%d fb_info=%p count=%d\n",
+	pr_debug("open /dev/fb%d user=%d fb_info=%p count=%d\n",
 		  info->node, user, info, ufbdev->fb_count);
 
 	return 0;
@@ -258,7 +261,7 @@ static int udl_fb_release(struct fb_info *info, int user)
 	}
 #endif
 
-	pr_warn("released /dev/fb%d user=%d count=%d\n",
+	pr_debug("released /dev/fb%d user=%d count=%d\n",
 		info->node, user, ufbdev->fb_count);
 
 	return 0;
diff --git a/drivers/gpu/drm/udl/udl_main.c b/drivers/gpu/drm/udl/udl_main.c
index d518de8f496bc75524d246dff580fa7a2a7bcf36..f455f095a14685d234569eaed56121d6269393b1 100644
--- a/drivers/gpu/drm/udl/udl_main.c
+++ b/drivers/gpu/drm/udl/udl_main.c
@@ -170,25 +170,19 @@ static void udl_free_urb_list(struct drm_device *dev)
 	struct list_head *node;
 	struct urb_node *unode;
 	struct urb *urb;
-	int ret;
-	unsigned long flags;
 
 	DRM_DEBUG("Waiting for completes and freeing all render urbs\n");
 
 	/* keep waiting and freeing, until we've got 'em all */
 	while (count--) {
+		down(&udl->urbs.limit_sem);
 
-		/* Getting interrupted means a leak, but ok at shutdown*/
-		ret = down_interruptible(&udl->urbs.limit_sem);
-		if (ret)
-			break;
-
-		spin_lock_irqsave(&udl->urbs.lock, flags);
+		spin_lock_irq(&udl->urbs.lock);
 
 		node = udl->urbs.list.next; /* have reserved one with sem */
 		list_del_init(node);
 
-		spin_unlock_irqrestore(&udl->urbs.lock, flags);
+		spin_unlock_irq(&udl->urbs.lock);
 
 		unode = list_entry(node, struct urb_node, entry);
 		urb = unode->urb;
@@ -205,17 +199,22 @@ static void udl_free_urb_list(struct drm_device *dev)
 static int udl_alloc_urb_list(struct drm_device *dev, int count, size_t size)
 {
 	struct udl_device *udl = dev->dev_private;
-	int i = 0;
 	struct urb *urb;
 	struct urb_node *unode;
 	char *buf;
+	size_t wanted_size = count * size;
 
 	spin_lock_init(&udl->urbs.lock);
 
+retry:
 	udl->urbs.size = size;
 	INIT_LIST_HEAD(&udl->urbs.list);
 
-	while (i < count) {
+	sema_init(&udl->urbs.limit_sem, 0);
+	udl->urbs.count = 0;
+	udl->urbs.available = 0;
+
+	while (udl->urbs.count * size < wanted_size) {
 		unode = kzalloc(sizeof(struct urb_node), GFP_KERNEL);
 		if (!unode)
 			break;
@@ -231,11 +230,16 @@ static int udl_alloc_urb_list(struct drm_device *dev, int count, size_t size)
 		}
 		unode->urb = urb;
 
-		buf = usb_alloc_coherent(udl->udev, MAX_TRANSFER, GFP_KERNEL,
+		buf = usb_alloc_coherent(udl->udev, size, GFP_KERNEL,
 					 &urb->transfer_dma);
 		if (!buf) {
 			kfree(unode);
 			usb_free_urb(urb);
+			if (size > PAGE_SIZE) {
+				size /= 2;
+				udl_free_urb_list(dev);
+				goto retry;
+			}
 			break;
 		}
 
@@ -246,16 +250,14 @@ static int udl_alloc_urb_list(struct drm_device *dev, int count, size_t size)
 
 		list_add_tail(&unode->entry, &udl->urbs.list);
 
-		i++;
+		up(&udl->urbs.limit_sem);
+		udl->urbs.count++;
+		udl->urbs.available++;
 	}
 
-	sema_init(&udl->urbs.limit_sem, i);
-	udl->urbs.count = i;
-	udl->urbs.available = i;
-
-	DRM_DEBUG("allocated %d %d byte urbs\n", i, (int) size);
+	DRM_DEBUG("allocated %d %d byte urbs\n", udl->urbs.count, (int) size);
 
-	return i;
+	return udl->urbs.count;
 }
 
 struct urb *udl_get_urb(struct drm_device *dev)
@@ -265,7 +267,6 @@ struct urb *udl_get_urb(struct drm_device *dev)
 	struct list_head *entry;
 	struct urb_node *unode;
 	struct urb *urb = NULL;
-	unsigned long flags;
 
 	/* Wait for an in-flight buffer to complete and get re-queued */
 	ret = down_timeout(&udl->urbs.limit_sem, GET_URB_TIMEOUT);
@@ -276,14 +277,14 @@ struct urb *udl_get_urb(struct drm_device *dev)
 		goto error;
 	}
 
-	spin_lock_irqsave(&udl->urbs.lock, flags);
+	spin_lock_irq(&udl->urbs.lock);
 
 	BUG_ON(list_empty(&udl->urbs.list)); /* reserved one with limit_sem */
 	entry = udl->urbs.list.next;
 	list_del_init(entry);
 	udl->urbs.available--;
 
-	spin_unlock_irqrestore(&udl->urbs.lock, flags);
+	spin_unlock_irq(&udl->urbs.lock);
 
 	unode = list_entry(entry, struct urb_node, entry);
 	urb = unode->urb;
diff --git a/drivers/gpu/drm/udl/udl_modeset.c b/drivers/gpu/drm/udl/udl_modeset.c
index 5bcae76497959bf3a14ba483df525eb052dd6300..7e37765cf5acd1ec0d2d74798b308c1a4be0bc5c 100644
--- a/drivers/gpu/drm/udl/udl_modeset.c
+++ b/drivers/gpu/drm/udl/udl_modeset.c
@@ -243,7 +243,7 @@ static int udl_crtc_write_mode_to_hw(struct drm_crtc *crtc)
 
 	memcpy(buf, udl->mode_buf, udl->mode_buf_len);
 	retval = udl_submit_urb(dev, urb, udl->mode_buf_len);
-	DRM_INFO("write mode info %d\n", udl->mode_buf_len);
+	DRM_DEBUG("write mode info %d\n", udl->mode_buf_len);
 	return retval;
 }
 
@@ -366,7 +366,6 @@ static int udl_crtc_page_flip(struct drm_crtc *crtc,
 {
 	struct udl_framebuffer *ufb = to_udl_fb(fb);
 	struct drm_device *dev = crtc->dev;
-	unsigned long flags;
 
 	struct drm_framebuffer *old_fb = crtc->primary->fb;
 	if (old_fb) {
@@ -377,10 +376,10 @@ static int udl_crtc_page_flip(struct drm_crtc *crtc,
 
 	udl_handle_damage(ufb, 0, 0, fb->width, fb->height);
 
-	spin_lock_irqsave(&dev->event_lock, flags);
+	spin_lock_irq(&dev->event_lock);
 	if (event)
 		drm_crtc_send_vblank_event(crtc, event);
-	spin_unlock_irqrestore(&dev->event_lock, flags);
+	spin_unlock_irq(&dev->event_lock);
 	crtc->primary->fb = fb;
 
 	return 0;
diff --git a/drivers/gpu/drm/udl/udl_transfer.c b/drivers/gpu/drm/udl/udl_transfer.c
index b992644c17e6b565b414351b1e262d4b61c9f38d..ce87661e544f7a4cc37bb6a03d89104d3ea97e72 100644
--- a/drivers/gpu/drm/udl/udl_transfer.c
+++ b/drivers/gpu/drm/udl/udl_transfer.c
@@ -13,7 +13,6 @@
 #include <linux/module.h>
 #include <linux/slab.h>
 #include <linux/fb.h>
-#include <linux/prefetch.h>
 #include <asm/unaligned.h>
 
 #include <drm/drmP.h>
@@ -51,9 +50,6 @@ static int udl_trim_hline(const u8 *bback, const u8 **bfront, int *width_bytes)
 	int start = width;
 	int end = width;
 
-	prefetch((void *) front);
-	prefetch((void *) back);
-
 	for (j = 0; j < width; j++) {
 		if (back[j] != front[j]) {
 			start = j;
@@ -83,12 +79,12 @@ static inline u16 pixel32_to_be16(const uint32_t pixel)
 		((pixel >> 8) & 0xf800));
 }
 
-static inline u16 get_pixel_val16(const uint8_t *pixel, int bpp)
+static inline u16 get_pixel_val16(const uint8_t *pixel, int log_bpp)
 {
-	u16 pixel_val16 = 0;
-	if (bpp == 2)
+	u16 pixel_val16;
+	if (log_bpp == 1)
 		pixel_val16 = *(const uint16_t *)pixel;
-	else if (bpp == 4)
+	else
 		pixel_val16 = pixel32_to_be16(*(const uint32_t *)pixel);
 	return pixel_val16;
 }
@@ -125,8 +121,9 @@ static void udl_compress_hline16(
 	const u8 *const pixel_end,
 	uint32_t *device_address_ptr,
 	uint8_t **command_buffer_ptr,
-	const uint8_t *const cmd_buffer_end, int bpp)
+	const uint8_t *const cmd_buffer_end, int log_bpp)
 {
+	const int bpp = 1 << log_bpp;
 	const u8 *pixel = *pixel_start_ptr;
 	uint32_t dev_addr  = *device_address_ptr;
 	uint8_t *cmd = *command_buffer_ptr;
@@ -139,8 +136,6 @@ static void udl_compress_hline16(
 		const u8 *cmd_pixel_start, *cmd_pixel_end = NULL;
 		uint16_t pixel_val16;
 
-		prefetchw((void *) cmd); /* pull in one cache line at least */
-
 		*cmd++ = 0xaf;
 		*cmd++ = 0x6b;
 		*cmd++ = (uint8_t) ((dev_addr >> 16) & 0xFF);
@@ -153,12 +148,11 @@ static void udl_compress_hline16(
 		raw_pixels_count_byte = cmd++; /*  we'll know this later */
 		raw_pixel_start = pixel;
 
-		cmd_pixel_end = pixel + min3(MAX_CMD_PIXELS + 1UL,
-					(unsigned long)(pixel_end - pixel) / bpp,
-					(unsigned long)(cmd_buffer_end - 1 - cmd) / 2) * bpp;
+		cmd_pixel_end = pixel + (min3(MAX_CMD_PIXELS + 1UL,
+					(unsigned long)(pixel_end - pixel) >> log_bpp,
+					(unsigned long)(cmd_buffer_end - 1 - cmd) / 2) << log_bpp);
 
-		prefetch_range((void *) pixel, cmd_pixel_end - pixel);
-		pixel_val16 = get_pixel_val16(pixel, bpp);
+		pixel_val16 = get_pixel_val16(pixel, log_bpp);
 
 		while (pixel < cmd_pixel_end) {
 			const u8 *const start = pixel;
@@ -170,7 +164,7 @@ static void udl_compress_hline16(
 			pixel += bpp;
 
 			while (pixel < cmd_pixel_end) {
-				pixel_val16 = get_pixel_val16(pixel, bpp);
+				pixel_val16 = get_pixel_val16(pixel, log_bpp);
 				if (pixel_val16 != repeating_pixel_val16)
 					break;
 				pixel += bpp;
@@ -179,10 +173,10 @@ static void udl_compress_hline16(
 			if (unlikely(pixel > start + bpp)) {
 				/* go back and fill in raw pixel count */
 				*raw_pixels_count_byte = (((start -
-						raw_pixel_start) / bpp) + 1) & 0xFF;
+						raw_pixel_start) >> log_bpp) + 1) & 0xFF;
 
 				/* immediately after raw data is repeat byte */
-				*cmd++ = (((pixel - start) / bpp) - 1) & 0xFF;
+				*cmd++ = (((pixel - start) >> log_bpp) - 1) & 0xFF;
 
 				/* Then start another raw pixel span */
 				raw_pixel_start = pixel;
@@ -192,14 +186,14 @@ static void udl_compress_hline16(
 
 		if (pixel > raw_pixel_start) {
 			/* finalize last RAW span */
-			*raw_pixels_count_byte = ((pixel-raw_pixel_start) / bpp) & 0xFF;
+			*raw_pixels_count_byte = ((pixel - raw_pixel_start) >> log_bpp) & 0xFF;
 		} else {
 			/* undo unused byte */
 			cmd--;
 		}
 
-		*cmd_pixels_count_byte = ((pixel - cmd_pixel_start) / bpp) & 0xFF;
-		dev_addr += ((pixel - cmd_pixel_start) / bpp) * 2;
+		*cmd_pixels_count_byte = ((pixel - cmd_pixel_start) >> log_bpp) & 0xFF;
+		dev_addr += ((pixel - cmd_pixel_start) >> log_bpp) * 2;
 	}
 
 	if (cmd_buffer_end <= MIN_RLX_CMD_BYTES + cmd) {
@@ -222,19 +216,19 @@ static void udl_compress_hline16(
  * (that we can only write to, slowly, and can never read), and (optionally)
  * our shadow copy that tracks what's been sent to that hardware buffer.
  */
-int udl_render_hline(struct drm_device *dev, int bpp, struct urb **urb_ptr,
+int udl_render_hline(struct drm_device *dev, int log_bpp, struct urb **urb_ptr,
 		     const char *front, char **urb_buf_ptr,
 		     u32 byte_offset, u32 device_byte_offset,
 		     u32 byte_width,
 		     int *ident_ptr, int *sent_ptr)
 {
 	const u8 *line_start, *line_end, *next_pixel;
-	u32 base16 = 0 + (device_byte_offset / bpp) * 2;
+	u32 base16 = 0 + (device_byte_offset >> log_bpp) * 2;
 	struct urb *urb = *urb_ptr;
 	u8 *cmd = *urb_buf_ptr;
 	u8 *cmd_end = (u8 *) urb->transfer_buffer + urb->transfer_buffer_length;
 
-	BUG_ON(!(bpp == 2 || bpp == 4));
+	BUG_ON(!(log_bpp == 1 || log_bpp == 2));
 
 	line_start = (u8 *) (front + byte_offset);
 	next_pixel = line_start;
@@ -244,7 +238,7 @@ int udl_render_hline(struct drm_device *dev, int bpp, struct urb **urb_ptr,
 
 		udl_compress_hline16(&next_pixel,
 			     line_end, &base16,
-			     (u8 **) &cmd, (u8 *) cmd_end, bpp);
+			     (u8 **) &cmd, (u8 *) cmd_end, log_bpp);
 
 		if (cmd >= cmd_end) {
 			int len = cmd - (u8 *) urb->transfer_buffer;