Merge tag 'trace-v7.3-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-tracegrafted

Pull tracing fixes from Steven Rostedt:

 - Don't destroy user event fields when removal fails

   User event fields are destroyed before the event is removed from
   visibility. But that can fail leaving the still visible event with no
   fields. Move the destroying of the fields to after the event is
   successfully removed from visibility.

 - Initialize function graph state is fork before calling
   copy_exec_state()

   For non-CLONE_VM forks, copy_exec_state() allocates a new
   task_exec_state. If that allocation fails, ftrace_graph_exit_task()
   will free the tasks ret_stack pointer. Since that pointer is still
   using the parent's ret_stack, it mistakenly frees the parent's
   pointer too.

   Call ftrace_graph_init() on the task first which will NULL out the
   new tasks's ret_stack and if the copy fails, it will not free
   anything.

 - Remove FGRAPH_MAX_INDEX

   The macro FGRAPH_MAX_INDEX was added but never used. Remove it.

 - Save ent_size in function graph printing of nested functions

   The function graph tracer needs to look at the next event to see if
   the next event is the return of the current function entry. If it is,
   it prints a single line:

	ktime_get();

   Otherwise it prints it like a nested function:

	tick_nohz_irq_exit() {
	    ktime_get();
	    kcpustat_irq_exit();
	}

   In order to look at the next event, it must save the current event so
   that it has the information to print from it. It saves the event in
   the iterator descriptor called "ent". What it doesn't save is the
   ent_size of the event which is now used to know if the function graph
   arguments are to be printed. The peek doesn't save the size so the
   size used happens to be that of the size of the last event that was
   seen.

   Save the entry event size in the iterator descriptor so that the
   correct size is used.

 - Fix several errors with freeing data in the histogram code

   The histogram code had a lot of leaked or or incorrect accounting
   when failures happen. Correct them.

 - Fix histogram regression of .percent and .graph modifiers

   Up until 6.3 histogram values could have "percent" or "graph"
   modifiers that changed how they were printed. But a change that added
   restricting histograms values from being strings, stack traces and
   other modifiers inadvertently prevented them from using the percent
   and graph modifiers, which were legal use cases for values.

   Put back the percent and graph modifiers.

 - Fix various typos in the comments

 - Set the trace_clock before initializing a histogram with clock
   argument

   The histogram API allows the user to specific which trace clock to
   use via a "clock=" string. The histogram is set up first before the
   clock is checked. If the passed in clock is not valid, it exits
   without fully fixing up the histogram leaving it on the list and a
   use-after-free can trigger.

   Update the clock argument first and if it fails then exit gracefully
   before the histogram trigger is placed on any lists.

 - Restore :mod: trailer after parsing in ftrace_set_clr_event

   The function ftrace_set_clr_event() modifies the parse string and
   needs to put it back to what was passed in. It searches for ":mod:"
   via a strsep() but fails to put back the first ':' in the string.

   Add back the ':' in the passed in string.

 - Take trace_array reference when opening a tracer options file

   The options files are dynamically created and some tracers add their
   own options. When a tracer adds their own list of options, the
   trace_array holding them has an array to hold the list of options for
   each tracer. This array increases in size via a krealloc(), and the
   new entry gets a newly allocated array to hold the options of the new
   tracer being added.

   The element in each entry of the tracer's option array holds a
   pointer back to the trace_array, a pointer to the tracer it is
   associated to, a pointer to the flags of the option.

   The issue is that these arrays are freed when the trace_array is
   freed when its instance it represents is removed from the instances
   directory. There's a race that an open of one of these options files
   can happen when the instance is being removed.

   Add a new helper function to be called by the open function of the
   options file to iterate all existing trace_arrays under a lock and
   find the one that has the given option element in one of it's tracer
   arrays. If found, then update the associated trace_array's reference
   counter to keep it from being freed. If not found, have the open call
   return -ENODEV.

 - Disable interrupts when acquiring the lock in rb_wake_up_waiters()

   The function rb_wake_up_waiters() assumes it will be called in
   interrupt context and does not disable irqs when taking
   cpu_buffer->reader_lock, which can be called in hard interrupt
   context. The issue is in PREEMPT_RT, this function is called in
   thread context leaving this lock open to a deadlock.

   Take the lock with interrupts disabled.

 - Use rcu_assign_pointer() for tmp_ops filter hash

   The tmp_ops used in update_ftrace_direct_mod() assigns its
   filter_hash field directly, but that field is annotated as __rcu and
   sparse complains. Assign it with rcu_assign_pointer()

 - Fix use-after-free in enable_trigger_private_data_free()

   The trace_event_call is accessed through the event_trigger_data's
   trace_event_file pointer to put the trace_event_call on freeing. The
   issue is that the trace_event_file data may have been freed already
   causing a use-after-free. Add a field to the event_trigger_data that
   points directly to the trace_event_call so that it can decrement its
   reference directly without needing to go through the
   trace_event_file.

 - Fix accounting of buffer data remote headers

   trace_buffer_desc_size() and trace_remote_alloc_buffer() undercount
   the number of pages is needed for the asked for size as it doesn't
   take into account the meta data on each page. Add a helper function
   to do the calculation properly and use that in these functions.

 - Catch nr_page_va overflow in ring_buffer_desc sizing

   The number of pages per remote ring buffer is capped by
   ring_buffer_desc::nr_page_va (32 bits). A buffer_size large enough to
   overflow that field would silently allocate a descriptor smaller than
   what was asked for.

 - Do not resize the subbuf order if any per_cpu buffer is disabled

   The mmapping of ring buffers disables resizing the subbuffers, but it
   is done per-cpu whereas the subbuf size change is done for all the
   per_cpu buffers under the buffer->mutex. It could change the size of
   some while the mapping is happening on others. Have the resize of the
   subbuf order check all the per_cpu buffers under the lock to see if
   any of them is disabled before starting and causing an inconsistency
   between buffers that are being mapped.

* tag 'trace-v7.3-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace: (25 commits)
  ring-buffer: Check resize_disabled before publishing the new subbuf order
  tracing/remotes: Catch nr_page_va overflow in ring_buffer_desc sizing
  tracing/remotes: Account for ring buffer page header in size calculation
  tracing: Don't dereference trace_event_file in deferred trigger free
  ftrace: Use rcu_assign_pointer() for tmp_ops filter hash
  ring-buffer: Acquire the lock with irqsave in rb_wake_up_waiters()
  tracing: Take trace_array reference when opening a tracer options file
  tracing: Fix ring_buffer_read_page_size() kernel-doc
  tracing: Restore :mod: trailer after parsing in ftrace_set_clr_event()
  tracing: Fix memory corruption from a "STACKTRACE" histogram key
  tracing: Fix memory corruption from the histogram stacktrace modifier
  tracing: Undo the registration when enabling the histogram trigger fails
  tracing: Take the reference before publishing the named histogram trigger
  tracing: Set the trace clock before registering the histogram trigger
  tracing: Fix typo "preceeded" in comment
  tracing: Fix typo "availabe" in comment
  tracing: Let histogram values keep the percent and graph modifiers
  tracing: Keep the entry count when the histogram stats allocation fails
  tracing: Free histogram the field rejected for a bad modifier
  tracing: Free histogram the var ref when its initialization fails
  ...

1 files changed, 1287 insertions, 0 deletions

diff --git a/drivers/video/fbdev/gbefb.c b/drivers/video/fbdev/gbefb.c
new file mode 100644
index 000000000..cb6ff15a2
--- /dev/null
+++ b/drivers/video/fbdev/gbefb.c
@@ -0,0 +1,1287 @@
+/*
+ *  SGI GBE frame buffer driver
+ *
+ *  Copyright (C) 1999 Silicon Graphics, Inc. - Jeffrey Newquist
+ *  Copyright (C) 2002 Vivien Chappelier <vivien.chappelier@linux-mips.org>
+ *
+ *  This file is subject to the terms and conditions of the GNU General Public
+ *  License. See the file COPYING in the main directory of this archive for
+ *  more details.
+ */
+
+#include <linux/delay.h>
+#include <linux/platform_device.h>
+#include <linux/dma-mapping.h>
+#include <linux/dma-direct.h>
+#include <linux/errno.h>
+#include <linux/gfp.h>
+#include <linux/fb.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/kernel.h>
+#include <linux/mm.h>
+#include <linux/module.h>
+#include <linux/io.h>
+
+#ifdef CONFIG_MIPS
+#include <asm/addrspace.h>
+#endif
+#include <asm/byteorder.h>
+#include <asm/tlbflush.h>
+
+#include <video/gbe.h>
+
+static struct sgi_gbe *gbe;
+
+struct gbefb_par {
+	struct fb_var_screeninfo var;
+	struct gbe_timing_info timing;
+	int wc_cookie;
+	int valid;
+};
+
+#define GBE_BASE	0x16000000 /* SGI O2 */
+
+/* macro for fastest write-though access to the framebuffer */
+#ifdef CONFIG_MIPS
+#ifdef CONFIG_CPU_R10000
+#define pgprot_fb(_prot) (((_prot) & (~_CACHE_MASK)) | _CACHE_UNCACHED_ACCELERATED)
+#else
+#define pgprot_fb(_prot) (((_prot) & (~_CACHE_MASK)) | _CACHE_CACHABLE_NO_WA)
+#endif
+#endif
+
+/*
+ *  RAM we reserve for the frame buffer. This defines the maximum screen
+ *  size
+ */
+#if CONFIG_FB_GBE_MEM > 8
+#error GBE Framebuffer cannot use more than 8MB of memory
+#endif
+
+#define TILE_SHIFT 16
+#define TILE_SIZE (1 << TILE_SHIFT)
+#define TILE_MASK (TILE_SIZE - 1)
+
+static unsigned int gbe_mem_size = CONFIG_FB_GBE_MEM * 1024*1024;
+static void *gbe_mem;
+static dma_addr_t gbe_dma_addr;
+static phys_addr_t gbe_mem_phys;
+
+static struct {
+	uint16_t *cpu;
+	dma_addr_t dma;
+} gbe_tiles;
+
+static int gbe_revision;
+
+static int ypan, ywrap;
+
+static uint32_t pseudo_palette[16];
+static uint32_t gbe_cmap[256];
+static int gbe_turned_on; /* 0 turned off, 1 turned on */
+
+static char *mode_option = NULL;
+
+/* default CRT mode */
+static struct fb_var_screeninfo default_var_CRT = {
+	/* 640x480, 60 Hz, Non-Interlaced (25.175 MHz dotclock) */
+	.xres		= 640,
+	.yres		= 480,
+	.xres_virtual	= 640,
+	.yres_virtual	= 480,
+	.xoffset	= 0,
+	.yoffset	= 0,
+	.bits_per_pixel	= 8,
+	.grayscale	= 0,
+	.red		= { 0, 8, 0 },
+	.green		= { 0, 8, 0 },
+	.blue		= { 0, 8, 0 },
+	.transp		= { 0, 0, 0 },
+	.nonstd		= 0,
+	.activate	= 0,
+	.height		= -1,
+	.width		= -1,
+	.accel_flags	= 0,
+	.pixclock	= 39722,	/* picoseconds */
+	.left_margin	= 48,
+	.right_margin	= 16,
+	.upper_margin	= 33,
+	.lower_margin	= 10,
+	.hsync_len	= 96,
+	.vsync_len	= 2,
+	.sync		= 0,
+	.vmode		= FB_VMODE_NONINTERLACED,
+};
+
+/* default LCD mode */
+static struct fb_var_screeninfo default_var_LCD = {
+	/* 1600x1024, 8 bpp */
+	.xres		= 1600,
+	.yres		= 1024,
+	.xres_virtual	= 1600,
+	.yres_virtual	= 1024,
+	.xoffset	= 0,
+	.yoffset	= 0,
+	.bits_per_pixel	= 8,
+	.grayscale	= 0,
+	.red		= { 0, 8, 0 },
+	.green		= { 0, 8, 0 },
+	.blue		= { 0, 8, 0 },
+	.transp		= { 0, 0, 0 },
+	.nonstd		= 0,
+	.activate	= 0,
+	.height		= -1,
+	.width		= -1,
+	.accel_flags	= 0,
+	.pixclock	= 9353,
+	.left_margin	= 20,
+	.right_margin	= 30,
+	.upper_margin	= 37,
+	.lower_margin	= 3,
+	.hsync_len	= 20,
+	.vsync_len	= 3,
+	.sync		= 0,
+	.vmode		= FB_VMODE_NONINTERLACED
+};
+
+/* default modedb mode */
+/* 640x480, 60 Hz, Non-Interlaced (25.172 MHz dotclock) */
+static struct fb_videomode default_mode_CRT = {
+	.refresh	= 60,
+	.xres		= 640,
+	.yres		= 480,
+	.pixclock	= 39722,
+	.left_margin	= 48,
+	.right_margin	= 16,
+	.upper_margin	= 33,
+	.lower_margin	= 10,
+	.hsync_len	= 96,
+	.vsync_len	= 2,
+	.sync		= 0,
+	.vmode		= FB_VMODE_NONINTERLACED,
+};
+/* 1600x1024 SGI flatpanel 1600sw */
+static struct fb_videomode default_mode_LCD = {
+	/* 1600x1024, 8 bpp */
+	.xres		= 1600,
+	.yres		= 1024,
+	.pixclock	= 9353,
+	.left_margin	= 20,
+	.right_margin	= 30,
+	.upper_margin	= 37,
+	.lower_margin	= 3,
+	.hsync_len	= 20,
+	.vsync_len	= 3,
+	.vmode		= FB_VMODE_NONINTERLACED,
+};
+
+static struct fb_videomode *default_mode = &default_mode_CRT;
+static struct fb_var_screeninfo *default_var = &default_var_CRT;
+
+static int flat_panel_enabled = 0;
+
+static void gbe_reset(void)
+{
+	/* Turn on dotclock PLL */
+	gbe->ctrlstat = 0x300aa000;
+}
+
+
+/*
+ * Function:	gbe_turn_off
+ * Parameters:	(None)
+ * Description:	This should turn off the monitor and gbe.  This is used
+ *              when switching between the serial console and the graphics
+ *              console.
+ */
+
+static void gbe_turn_off(void)
+{
+	int i;
+	unsigned int val, y, vpixen_off;
+
+	gbe_turned_on = 0;
+
+	/* check if pixel counter is on */
+	val = gbe->vt_xy;
+	if (GET_GBE_FIELD(VT_XY, FREEZE, val) == 1)
+		return;
+
+	/* turn off DMA */
+	val = gbe->ovr_control;
+	SET_GBE_FIELD(OVR_CONTROL, OVR_DMA_ENABLE, val, 0);
+	gbe->ovr_control = val;
+	udelay(1000);
+	val = gbe->frm_control;
+	SET_GBE_FIELD(FRM_CONTROL, FRM_DMA_ENABLE, val, 0);
+	gbe->frm_control = val;
+	udelay(1000);
+	val = gbe->did_control;
+	SET_GBE_FIELD(DID_CONTROL, DID_DMA_ENABLE, val, 0);
+	gbe->did_control = val;
+	udelay(1000);
+
+	/* We have to wait through two vertical retrace periods before
+	 * the pixel DMA is turned off for sure. */
+	for (i = 0; i < 10000; i++) {
+		val = gbe->frm_inhwctrl;
+		if (GET_GBE_FIELD(FRM_INHWCTRL, FRM_DMA_ENABLE, val)) {
+			udelay(10);
+		} else {
+			val = gbe->ovr_inhwctrl;
+			if (GET_GBE_FIELD(OVR_INHWCTRL, OVR_DMA_ENABLE, val)) {
+				udelay(10);
+			} else {
+				val = gbe->did_inhwctrl;
+				if (GET_GBE_FIELD(DID_INHWCTRL, DID_DMA_ENABLE, val)) {
+					udelay(10);
+				} else
+					break;
+			}
+		}
+	}
+	if (i == 10000)
+		printk(KERN_ERR "gbefb: turn off DMA timed out\n");
+
+	/* wait for vpixen_off */
+	val = gbe->vt_vpixen;
+	vpixen_off = GET_GBE_FIELD(VT_VPIXEN, VPIXEN_OFF, val);
+
+	for (i = 0; i < 100000; i++) {
+		val = gbe->vt_xy;
+		y = GET_GBE_FIELD(VT_XY, Y, val);
+		if (y < vpixen_off)
+			break;
+		udelay(1);
+	}
+	if (i == 100000)
+		printk(KERN_ERR
+		       "gbefb: wait for vpixen_off timed out\n");
+	for (i = 0; i < 10000; i++) {
+		val = gbe->vt_xy;
+		y = GET_GBE_FIELD(VT_XY, Y, val);
+		if (y > vpixen_off)
+			break;
+		udelay(1);
+	}
+	if (i == 10000)
+		printk(KERN_ERR "gbefb: wait for vpixen_off timed out\n");
+
+	/* turn off pixel counter */
+	val = 0;
+	SET_GBE_FIELD(VT_XY, FREEZE, val, 1);
+	gbe->vt_xy = val;
+	mdelay(10);
+	for (i = 0; i < 10000; i++) {
+		val = gbe->vt_xy;
+		if (GET_GBE_FIELD(VT_XY, FREEZE, val) != 1)
+			udelay(10);
+		else
+			break;
+	}
+	if (i == 10000)
+		printk(KERN_ERR "gbefb: turn off pixel clock timed out\n");
+
+	/* turn off dot clock */
+	val = gbe->dotclock;
+	SET_GBE_FIELD(DOTCLK, RUN, val, 0);
+	gbe->dotclock = val;
+	mdelay(10);
+	for (i = 0; i < 10000; i++) {
+		val = gbe->dotclock;
+		if (GET_GBE_FIELD(DOTCLK, RUN, val))
+			udelay(10);
+		else
+			break;
+	}
+	if (i == 10000)
+		printk(KERN_ERR "gbefb: turn off dotclock timed out\n");
+
+	/* reset the frame DMA FIFO */
+	val = gbe->frm_size_tile;
+	SET_GBE_FIELD(FRM_SIZE_TILE, FRM_FIFO_RESET, val, 1);
+	gbe->frm_size_tile = val;
+	SET_GBE_FIELD(FRM_SIZE_TILE, FRM_FIFO_RESET, val, 0);
+	gbe->frm_size_tile = val;
+}
+
+static void gbe_turn_on(void)
+{
+	unsigned int val, i;
+
+	/*
+	 * Check if pixel counter is off, for unknown reason this
+	 * code hangs Visual Workstations
+	 */
+	if (gbe_revision < 2) {
+		val = gbe->vt_xy;
+		if (GET_GBE_FIELD(VT_XY, FREEZE, val) == 0)
+			return;
+	}
+
+	/* turn on dot clock */
+	val = gbe->dotclock;
+	SET_GBE_FIELD(DOTCLK, RUN, val, 1);
+	gbe->dotclock = val;
+	mdelay(10);
+	for (i = 0; i < 10000; i++) {
+		val = gbe->dotclock;
+		if (GET_GBE_FIELD(DOTCLK, RUN, val) != 1)
+			udelay(10);
+		else
+			break;
+	}
+	if (i == 10000)
+		printk(KERN_ERR "gbefb: turn on dotclock timed out\n");
+
+	/* turn on pixel counter */
+	val = 0;
+	SET_GBE_FIELD(VT_XY, FREEZE, val, 0);
+	gbe->vt_xy = val;
+	mdelay(10);
+	for (i = 0; i < 10000; i++) {
+		val = gbe->vt_xy;
+		if (GET_GBE_FIELD(VT_XY, FREEZE, val))
+			udelay(10);
+		else
+			break;
+	}
+	if (i == 10000)
+		printk(KERN_ERR "gbefb: turn on pixel clock timed out\n");
+
+	/* turn on DMA */
+	val = gbe->frm_control;
+	SET_GBE_FIELD(FRM_CONTROL, FRM_DMA_ENABLE, val, 1);
+	gbe->frm_control = val;
+	udelay(1000);
+	for (i = 0; i < 10000; i++) {
+		val = gbe->frm_inhwctrl;
+		if (GET_GBE_FIELD(FRM_INHWCTRL, FRM_DMA_ENABLE, val) != 1)
+			udelay(10);
+		else
+			break;
+	}
+	if (i == 10000)
+		printk(KERN_ERR "gbefb: turn on DMA timed out\n");
+
+	gbe_turned_on = 1;
+}
+
+static void gbe_loadcmap(void)
+{
+	int i, j;
+
+	for (i = 0; i < 256; i++) {
+		for (j = 0; j < 1000 && gbe->cm_fifo >= 63; j++)
+			udelay(10);
+		if (j == 1000)
+			printk(KERN_ERR "gbefb: cmap FIFO timeout\n");
+
+		gbe->cmap[i] = gbe_cmap[i];
+	}
+}
+
+/*
+ *  Blank the display.
+ */
+static int gbefb_blank(int blank, struct fb_info *info)
+{
+	/* 0 unblank, 1 blank, 2 no vsync, 3 no hsync, 4 off */
+	switch (blank) {
+	case FB_BLANK_UNBLANK:		/* unblank */
+		gbe_turn_on();
+		gbe_loadcmap();
+		break;
+
+	case FB_BLANK_NORMAL:		/* blank */
+		gbe_turn_off();
+		break;
+
+	default:
+		/* Nothing */
+		break;
+	}
+	return 0;
+}
+
+/*
+ *  Setup flatpanel related registers.
+ */
+static void gbefb_setup_flatpanel(struct gbe_timing_info *timing)
+{
+	int fp_wid, fp_hgt, fp_vbs, fp_vbe;
+	u32 outputVal = 0;
+
+	SET_GBE_FIELD(VT_FLAGS, HDRV_INVERT, outputVal,
+		(timing->flags & FB_SYNC_HOR_HIGH_ACT) ? 0 : 1);
+	SET_GBE_FIELD(VT_FLAGS, VDRV_INVERT, outputVal,
+		(timing->flags & FB_SYNC_VERT_HIGH_ACT) ? 0 : 1);
+	gbe->vt_flags = outputVal;
+
+	/* Turn on the flat panel */
+	fp_wid = 1600;
+	fp_hgt = 1024;
+	fp_vbs = 0;
+	fp_vbe = 1600;
+	timing->pll_m = 4;
+	timing->pll_n = 1;
+	timing->pll_p = 0;
+
+	outputVal = 0;
+	SET_GBE_FIELD(FP_DE, ON, outputVal, fp_vbs);
+	SET_GBE_FIELD(FP_DE, OFF, outputVal, fp_vbe);
+	gbe->fp_de = outputVal;
+	outputVal = 0;
+	SET_GBE_FIELD(FP_HDRV, OFF, outputVal, fp_wid);
+	gbe->fp_hdrv = outputVal;
+	outputVal = 0;
+	SET_GBE_FIELD(FP_VDRV, ON, outputVal, 1);
+	SET_GBE_FIELD(FP_VDRV, OFF, outputVal, fp_hgt + 1);
+	gbe->fp_vdrv = outputVal;
+}
+
+struct gbe_pll_info {
+	int clock_rate;
+	int fvco_min;
+	int fvco_max;
+};
+
+static struct gbe_pll_info gbe_pll_table[2] = {
+	{ 20, 80, 220 },
+	{ 27, 80, 220 },
+};
+
+static int compute_gbe_timing(struct fb_var_screeninfo *var,
+			      struct gbe_timing_info *timing)
+{
+	int pll_m, pll_n, pll_p, error, best_m, best_n, best_p, best_error;
+	int pixclock;
+	struct gbe_pll_info *gbe_pll;
+
+	if (gbe_revision < 2)
+		gbe_pll = &gbe_pll_table[0];
+	else
+		gbe_pll = &gbe_pll_table[1];
+
+	/* Determine valid resolution and timing
+	 * GBE crystal runs at 20Mhz or 27Mhz
+	 * pll_m, pll_n, pll_p define the following frequencies
+	 * fvco = pll_m * 20Mhz / pll_n
+	 * fout = fvco / (2**pll_p) */
+	best_error = 1000000000;
+	best_n = best_m = best_p = 0;
+	for (pll_p = 0; pll_p < 4; pll_p++)
+		for (pll_m = 1; pll_m < 256; pll_m++)
+			for (pll_n = 1; pll_n < 64; pll_n++) {
+				pixclock = (1000000 / gbe_pll->clock_rate) *
+						(pll_n << pll_p) / pll_m;
+
+				error = var->pixclock - pixclock;
+
+				if (error < 0)
+					error = -error;
+
+				if (error < best_error &&
+				    pll_m / pll_n >
+				    gbe_pll->fvco_min / gbe_pll->clock_rate &&
+ 				    pll_m / pll_n <
+				    gbe_pll->fvco_max / gbe_pll->clock_rate) {
+					best_error = error;
+					best_m = pll_m;
+					best_n = pll_n;
+					best_p = pll_p;
+				}
+			}
+
+	if (!best_n || !best_m)
+		return -EINVAL;	/* Resolution to high */
+
+	pixclock = (1000000 / gbe_pll->clock_rate) *
+		(best_n << best_p) / best_m;
+
+	/* set video timing information */
+	if (timing) {
+		timing->width = var->xres;
+		timing->height = var->yres;
+		timing->pll_m = best_m;
+		timing->pll_n = best_n;
+		timing->pll_p = best_p;
+		timing->cfreq = gbe_pll->clock_rate * 1000 * timing->pll_m /
+			(timing->pll_n << timing->pll_p);
+		timing->htotal = var->left_margin + var->xres +
+				var->right_margin + var->hsync_len;
+		timing->vtotal = var->upper_margin + var->yres +
+				var->lower_margin + var->vsync_len;
+		timing->fields_sec = 1000 * timing->cfreq / timing->htotal *
+				1000 / timing->vtotal;
+		timing->hblank_start = var->xres;
+		timing->vblank_start = var->yres;
+		timing->hblank_end = timing->htotal;
+		timing->hsync_start = var->xres + var->right_margin + 1;
+		timing->hsync_end = timing->hsync_start + var->hsync_len;
+		timing->vblank_end = timing->vtotal;
+		timing->vsync_start = var->yres + var->lower_margin + 1;
+		timing->vsync_end = timing->vsync_start + var->vsync_len;
+	}
+
+	return pixclock;
+}
+
+static void gbe_set_timing_info(struct gbe_timing_info *timing)
+{
+	int temp;
+	unsigned int val;
+
+	/* setup dot clock PLL */
+	val = 0;
+	SET_GBE_FIELD(DOTCLK, M, val, timing->pll_m - 1);
+	SET_GBE_FIELD(DOTCLK, N, val, timing->pll_n - 1);
+	SET_GBE_FIELD(DOTCLK, P, val, timing->pll_p);
+	SET_GBE_FIELD(DOTCLK, RUN, val, 0);	/* do not start yet */
+	gbe->dotclock = val;
+	mdelay(10);
+
+	/* setup pixel counter */
+	val = 0;
+	SET_GBE_FIELD(VT_XYMAX, MAXX, val, timing->htotal);
+	SET_GBE_FIELD(VT_XYMAX, MAXY, val, timing->vtotal);
+	gbe->vt_xymax = val;
+
+	/* setup video timing signals */
+	val = 0;
+	SET_GBE_FIELD(VT_VSYNC, VSYNC_ON, val, timing->vsync_start);
+	SET_GBE_FIELD(VT_VSYNC, VSYNC_OFF, val, timing->vsync_end);
+	gbe->vt_vsync = val;
+	val = 0;
+	SET_GBE_FIELD(VT_HSYNC, HSYNC_ON, val, timing->hsync_start);
+	SET_GBE_FIELD(VT_HSYNC, HSYNC_OFF, val, timing->hsync_end);
+	gbe->vt_hsync = val;
+	val = 0;
+	SET_GBE_FIELD(VT_VBLANK, VBLANK_ON, val, timing->vblank_start);
+	SET_GBE_FIELD(VT_VBLANK, VBLANK_OFF, val, timing->vblank_end);
+	gbe->vt_vblank = val;
+	val = 0;
+	SET_GBE_FIELD(VT_HBLANK, HBLANK_ON, val,
+		      timing->hblank_start - 5);
+	SET_GBE_FIELD(VT_HBLANK, HBLANK_OFF, val,
+		      timing->hblank_end - 3);
+	gbe->vt_hblank = val;
+
+	/* setup internal timing signals */
+	val = 0;
+	SET_GBE_FIELD(VT_VCMAP, VCMAP_ON, val, timing->vblank_start);
+	SET_GBE_FIELD(VT_VCMAP, VCMAP_OFF, val, timing->vblank_end);
+	gbe->vt_vcmap = val;
+	val = 0;
+	SET_GBE_FIELD(VT_HCMAP, HCMAP_ON, val, timing->hblank_start);
+	SET_GBE_FIELD(VT_HCMAP, HCMAP_OFF, val, timing->hblank_end);
+	gbe->vt_hcmap = val;
+
+	val = 0;
+	temp = timing->vblank_start - timing->vblank_end - 1;
+	if (temp > 0)
+		temp = -temp;
+
+	if (flat_panel_enabled)
+		gbefb_setup_flatpanel(timing);
+
+	SET_GBE_FIELD(DID_START_XY, DID_STARTY, val, (u32) temp);
+	if (timing->hblank_end >= 20)
+		SET_GBE_FIELD(DID_START_XY, DID_STARTX, val,
+			      timing->hblank_end - 20);
+	else
+		SET_GBE_FIELD(DID_START_XY, DID_STARTX, val,
+			      timing->htotal - (20 - timing->hblank_end));
+	gbe->did_start_xy = val;
+
+	val = 0;
+	SET_GBE_FIELD(CRS_START_XY, CRS_STARTY, val, (u32) (temp + 1));
+	if (timing->hblank_end >= GBE_CRS_MAGIC)
+		SET_GBE_FIELD(CRS_START_XY, CRS_STARTX, val,
+			      timing->hblank_end - GBE_CRS_MAGIC);
+	else
+		SET_GBE_FIELD(CRS_START_XY, CRS_STARTX, val,
+			      timing->htotal - (GBE_CRS_MAGIC -
+						timing->hblank_end));
+	gbe->crs_start_xy = val;
+
+	val = 0;
+	SET_GBE_FIELD(VC_START_XY, VC_STARTY, val, (u32) temp);
+	SET_GBE_FIELD(VC_START_XY, VC_STARTX, val, timing->hblank_end - 4);
+	gbe->vc_start_xy = val;
+
+	val = 0;
+	temp = timing->hblank_end - GBE_PIXEN_MAGIC_ON;
+	if (temp < 0)
+		temp += timing->htotal;	/* allow blank to wrap around */
+
+	SET_GBE_FIELD(VT_HPIXEN, HPIXEN_ON, val, temp);
+	SET_GBE_FIELD(VT_HPIXEN, HPIXEN_OFF, val,
+		      ((temp + timing->width -
+			GBE_PIXEN_MAGIC_OFF) % timing->htotal));
+	gbe->vt_hpixen = val;
+
+	val = 0;
+	SET_GBE_FIELD(VT_VPIXEN, VPIXEN_ON, val, timing->vblank_end);
+	SET_GBE_FIELD(VT_VPIXEN, VPIXEN_OFF, val, timing->vblank_start);
+	gbe->vt_vpixen = val;
+
+	/* turn off sync on green */
+	val = 0;
+	SET_GBE_FIELD(VT_FLAGS, SYNC_LOW, val, 1);
+	gbe->vt_flags = val;
+}
+
+/*
+ *  Set the hardware according to 'par'.
+ */
+
+static int gbefb_set_par(struct fb_info *info)
+{
+	int i;
+	unsigned int val;
+	int wholeTilesX, partTilesX, maxPixelsPerTileX;
+	int height_pix;
+	int xpmax, ypmax;	/* Monitor resolution */
+	int bytesPerPixel;	/* Bytes per pixel */
+	struct gbefb_par *par = (struct gbefb_par *) info->par;
+
+	compute_gbe_timing(&info->var, &par->timing);
+
+	bytesPerPixel = info->var.bits_per_pixel / 8;
+	info->fix.line_length = info->var.xres_virtual * bytesPerPixel;
+	xpmax = par->timing.width;
+	ypmax = par->timing.height;
+
+	/* turn off GBE */
+	gbe_turn_off();
+
+	/* set timing info */
+	gbe_set_timing_info(&par->timing);
+
+	/* initialize DIDs */
+	val = 0;
+	switch (bytesPerPixel) {
+	case 1:
+		SET_GBE_FIELD(WID, TYP, val, GBE_CMODE_I8);
+		info->fix.visual = FB_VISUAL_PSEUDOCOLOR;
+		break;
+	case 2:
+		SET_GBE_FIELD(WID, TYP, val, GBE_CMODE_ARGB5);
+		info->fix.visual = FB_VISUAL_TRUECOLOR;
+		break;
+	case 4:
+		SET_GBE_FIELD(WID, TYP, val, GBE_CMODE_RGB8);
+		info->fix.visual = FB_VISUAL_TRUECOLOR;
+		break;
+	}
+	SET_GBE_FIELD(WID, BUF, val, GBE_BMODE_BOTH);
+
+	for (i = 0; i < 32; i++)
+		gbe->mode_regs[i] = val;
+
+	/* Initialize interrupts */
+	gbe->vt_intr01 = 0xffffffff;
+	gbe->vt_intr23 = 0xffffffff;
+
+	/* HACK:
+	   The GBE hardware uses a tiled memory to screen mapping. Tiles are
+	   blocks of 512x128, 256x128 or 128x128 pixels, respectively for 8bit,
+	   16bit and 32 bit modes (64 kB). They cover the screen with partial
+	   tiles on the right and/or bottom of the screen if needed.
+	   For example in 640x480 8 bit mode the mapping is:
+
+	   <-------- 640 ----->
+	   <---- 512 ----><128|384 offscreen>
+	   ^  ^
+	   | 128    [tile 0]        [tile 1]
+	   |  v
+	   ^
+	   4 128    [tile 2]        [tile 3]
+	   8  v
+	   0  ^
+	   128    [tile 4]        [tile 5]
+	   |  v
+	   |  ^
+	   v  96    [tile 6]        [tile 7]
+	   32 offscreen
+
+	   Tiles have the advantage that they can be allocated individually in
+	   memory. However, this mapping is not linear at all, which is not
+	   really convenient. In order to support linear addressing, the GBE
+	   DMA hardware is fooled into thinking the screen is only one tile
+	   large and but has a greater height, so that the DMA transfer covers
+	   the same region.
+	   Tiles are still allocated as independent chunks of 64KB of
+	   continuous physical memory and remapped so that the kernel sees the
+	   framebuffer as a continuous virtual memory. The GBE tile table is
+	   set up so that each tile references one of these 64k blocks:
+
+	   GBE -> tile list    framebuffer           TLB   <------------ CPU
+	          [ tile 0 ] -> [ 64KB ]  <- [ 16x 4KB page entries ]     ^
+	             ...           ...              ...       linear virtual FB
+	          [ tile n ] -> [ 64KB ]  <- [ 16x 4KB page entries ]     v
+
+
+	   The GBE hardware is then told that the buffer is 512*tweaked_height,
+	   with tweaked_height = real_width*real_height/pixels_per_tile.
+	   Thus the GBE hardware will scan the first tile, filing the first 64k
+	   covered region of the screen, and then will proceed to the next
+	   tile, until the whole screen is covered.
+
+	   Here is what would happen at 640x480 8bit:
+
+	   normal tiling               linear
+	   ^   11111111111111112222    11111111111111111111  ^
+	   128 11111111111111112222    11111111111111111111 102 lines
+	       11111111111111112222    11111111111111111111  v
+	   V   11111111111111112222    11111111222222222222
+	       33333333333333334444    22222222222222222222
+	       33333333333333334444    22222222222222222222
+	       <      512     >        <  256 >               102*640+256 = 64k
+
+	   NOTE: The only mode for which this is not working is 800x600 8bit,
+	   as 800*600/512 = 937.5 which is not integer and thus causes
+	   flickering.
+	   I guess this is not so important as one can use 640x480 8bit or
+	   800x600 16bit anyway.
+	 */
+
+	/* Tell gbe about the tiles table location */
+	/* tile_ptr -> [ tile 1 ] -> FB mem */
+	/*             [ tile 2 ] -> FB mem */
+	/*               ...                */
+	val = 0;
+	SET_GBE_FIELD(FRM_CONTROL, FRM_TILE_PTR, val, gbe_tiles.dma >> 9);
+	SET_GBE_FIELD(FRM_CONTROL, FRM_DMA_ENABLE, val, 0); /* do not start */
+	SET_GBE_FIELD(FRM_CONTROL, FRM_LINEAR, val, 0);
+	gbe->frm_control = val;
+
+	maxPixelsPerTileX = 512 / bytesPerPixel;
+	wholeTilesX = 1;
+	partTilesX = 0;
+
+	/* Initialize the framebuffer */
+	val = 0;
+	SET_GBE_FIELD(FRM_SIZE_TILE, FRM_WIDTH_TILE, val, wholeTilesX);
+	SET_GBE_FIELD(FRM_SIZE_TILE, FRM_RHS, val, partTilesX);
+
+	switch (bytesPerPixel) {
+	case 1:
+		SET_GBE_FIELD(FRM_SIZE_TILE, FRM_DEPTH, val,
+			      GBE_FRM_DEPTH_8);
+		break;
+	case 2:
+		SET_GBE_FIELD(FRM_SIZE_TILE, FRM_DEPTH, val,
+			      GBE_FRM_DEPTH_16);
+		break;
+	case 4:
+		SET_GBE_FIELD(FRM_SIZE_TILE, FRM_DEPTH, val,
+			      GBE_FRM_DEPTH_32);
+		break;
+	}
+	gbe->frm_size_tile = val;
+
+	/* compute tweaked height */
+	height_pix = xpmax * ypmax / maxPixelsPerTileX;
+
+	val = 0;
+	SET_GBE_FIELD(FRM_SIZE_PIXEL, FB_HEIGHT_PIX, val, height_pix);
+	gbe->frm_size_pixel = val;
+
+	/* turn off DID and overlay DMA */
+	gbe->did_control = 0;
+	gbe->ovr_width_tile = 0;
+
+	/* Turn off mouse cursor */
+	gbe->crs_ctl = 0;
+
+	/* Turn on GBE */
+	gbe_turn_on();
+
+	/* Initialize the gamma map */
+	udelay(10);
+	for (i = 0; i < 256; i++)
+		gbe->gmap[i] = (i << 24) | (i << 16) | (i << 8);
+
+	/* Initialize the color map */
+	for (i = 0; i < 256; i++)
+		gbe_cmap[i] = (i << 8) | (i << 16) | (i << 24);
+
+	gbe_loadcmap();
+
+	return 0;
+}
+
+static void gbefb_encode_fix(struct fb_fix_screeninfo *fix,
+			     struct fb_var_screeninfo *var)
+{
+	memset(fix, 0, sizeof(struct fb_fix_screeninfo));
+	strcpy(fix->id, "SGI GBE");
+	fix->smem_start = (unsigned long) gbe_mem;
+	fix->smem_len = gbe_mem_size;
+	fix->type = FB_TYPE_PACKED_PIXELS;
+	fix->type_aux = 0;
+	fix->accel = FB_ACCEL_NONE;
+	switch (var->bits_per_pixel) {
+	case 8:
+		fix->visual = FB_VISUAL_PSEUDOCOLOR;
+		break;
+	default:
+		fix->visual = FB_VISUAL_TRUECOLOR;
+		break;
+	}
+	fix->ywrapstep = 0;
+	fix->xpanstep = 0;
+	fix->ypanstep = 0;
+	fix->line_length = var->xres_virtual * var->bits_per_pixel / 8;
+	fix->mmio_start = GBE_BASE;
+	fix->mmio_len = sizeof(struct sgi_gbe);
+}
+
+/*
+ *  Set a single color register. The values supplied are already
+ *  rounded down to the hardware's capabilities (according to the
+ *  entries in the var structure). Return != 0 for invalid regno.
+ */
+
+static int gbefb_setcolreg(unsigned regno, unsigned red, unsigned green,
+			     unsigned blue, unsigned transp,
+			     struct fb_info *info)
+{
+	int i;
+
+	if (regno > 255)
+		return 1;
+	red >>= 8;
+	green >>= 8;
+	blue >>= 8;
+
+	if (info->var.bits_per_pixel <= 8) {
+		gbe_cmap[regno] = (red << 24) | (green << 16) | (blue << 8);
+		if (gbe_turned_on) {
+			/* wait for the color map FIFO to have a free entry */
+			for (i = 0; i < 1000 && gbe->cm_fifo >= 63; i++)
+				udelay(10);
+			if (i == 1000) {
+				printk(KERN_ERR "gbefb: cmap FIFO timeout\n");
+				return 1;
+			}
+			gbe->cmap[regno] = gbe_cmap[regno];
+		}
+	} else if (regno < 16) {
+		switch (info->var.bits_per_pixel) {
+		case 15:
+		case 16:
+			red >>= 3;
+			green >>= 3;
+			blue >>= 3;
+			pseudo_palette[regno] =
+				(red << info->var.red.offset) |
+				(green << info->var.green.offset) |
+				(blue << info->var.blue.offset);
+			break;
+		case 32:
+			pseudo_palette[regno] =
+				(red << info->var.red.offset) |
+				(green << info->var.green.offset) |
+				(blue << info->var.blue.offset);
+			break;
+		}
+	}
+
+	return 0;
+}
+
+/*
+ *  Check video mode validity, eventually modify var to best match.
+ */
+static int gbefb_check_var(struct fb_var_screeninfo *var, struct fb_info *info)
+{
+	unsigned int line_length;
+	struct gbe_timing_info timing;
+	int ret;
+
+	/* Limit bpp to 8, 16, and 32 */
+	if (var->bits_per_pixel <= 8)
+		var->bits_per_pixel = 8;
+	else if (var->bits_per_pixel <= 16)
+		var->bits_per_pixel = 16;
+	else if (var->bits_per_pixel <= 32)
+		var->bits_per_pixel = 32;
+	else
+		return -EINVAL;
+
+	/* Check the mode can be mapped linearly with the tile table trick. */
+	/* This requires width x height x bytes/pixel be a multiple of 512 */
+	if ((var->xres * var->yres * var->bits_per_pixel) & 4095)
+		return -EINVAL;
+
+	var->grayscale = 0;	/* No grayscale for now */
+
+	ret = compute_gbe_timing(var, &timing);
+	var->pixclock = ret;
+	if (ret < 0)
+		return -EINVAL;
+
+	/* Adjust virtual resolution, if necessary */
+	if (var->xres > var->xres_virtual || (!ywrap && !ypan))
+		var->xres_virtual = var->xres;
+	if (var->yres > var->yres_virtual || (!ywrap && !ypan))
+		var->yres_virtual = var->yres;
+
+	if (var->vmode & FB_VMODE_CONUPDATE) {
+		var->vmode |= FB_VMODE_YWRAP;
+		var->xoffset = info->var.xoffset;
+		var->yoffset = info->var.yoffset;
+	}
+
+	/* No grayscale for now */
+	var->grayscale = 0;
+
+	/* Memory limit */
+	line_length = var->xres_virtual * var->bits_per_pixel / 8;
+	if (line_length * var->yres_virtual > gbe_mem_size)
+		return -ENOMEM;	/* Virtual resolution too high */
+
+	switch (var->bits_per_pixel) {
+	case 8:
+		var->red.offset = 0;
+		var->red.length = 8;
+		var->green.offset = 0;
+		var->green.length = 8;
+		var->blue.offset = 0;
+		var->blue.length = 8;
+		var->transp.offset = 0;
+		var->transp.length = 0;
+		break;
+	case 16:		/* RGB 1555 */
+		var->red.offset = 10;
+		var->red.length = 5;
+		var->green.offset = 5;
+		var->green.length = 5;
+		var->blue.offset = 0;
+		var->blue.length = 5;
+		var->transp.offset = 0;
+		var->transp.length = 0;
+		break;
+	case 32:		/* RGB 8888 */
+		var->red.offset = 24;
+		var->red.length = 8;
+		var->green.offset = 16;
+		var->green.length = 8;
+		var->blue.offset = 8;
+		var->blue.length = 8;
+		var->transp.offset = 0;
+		var->transp.length = 8;
+		break;
+	}
+	var->red.msb_right = 0;
+	var->green.msb_right = 0;
+	var->blue.msb_right = 0;
+	var->transp.msb_right = 0;
+
+	var->left_margin = timing.htotal - timing.hsync_end;
+	var->right_margin = timing.hsync_start - timing.width;
+	var->upper_margin = timing.vtotal - timing.vsync_end;
+	var->lower_margin = timing.vsync_start - timing.height;
+	var->hsync_len = timing.hsync_end - timing.hsync_start;
+	var->vsync_len = timing.vsync_end - timing.vsync_start;
+
+	return 0;
+}
+
+static int gbefb_mmap(struct fb_info *info,
+			struct vm_area_struct *vma)
+{
+	unsigned long size = vma->vm_end - vma->vm_start;
+	unsigned long offset = vma->vm_pgoff << PAGE_SHIFT;
+	unsigned long addr;
+	unsigned long phys_addr, phys_size;
+	u16 *tile;
+
+	vma->vm_page_prot = pgprot_decrypted(vma->vm_page_prot);
+
+	/* check range */
+	if (vma->vm_pgoff > (~0UL >> PAGE_SHIFT))
+		return -EINVAL;
+	if (size > gbe_mem_size)
+		return -EINVAL;
+	if (offset > gbe_mem_size - size)
+		return -EINVAL;
+
+	/* remap using the fastest write-through mode on architecture */
+	/* try not polluting the cache when possible */
+#ifdef CONFIG_MIPS
+	pgprot_val(vma->vm_page_prot) =
+		pgprot_fb(pgprot_val(vma->vm_page_prot));
+#endif
+	/* VM_IO | VM_DONTEXPAND | VM_DONTDUMP are set by remap_pfn_range() */
+
+	/* look for the starting tile */
+	tile = &gbe_tiles.cpu[offset >> TILE_SHIFT];
+	addr = vma->vm_start;
+	offset &= TILE_MASK;
+
+	/* remap each tile separately */
+	do {
+		phys_addr = (((unsigned long) (*tile)) << TILE_SHIFT) + offset;
+		if ((offset + size) < TILE_SIZE)
+			phys_size = size;
+		else
+			phys_size = TILE_SIZE - offset;
+
+		if (remap_pfn_range(vma, addr, phys_addr >> PAGE_SHIFT,
+						phys_size, vma->vm_page_prot))
+			return -EAGAIN;
+
+		offset = 0;
+		size -= phys_size;
+		addr += phys_size;
+		tile++;
+	} while (size);
+
+	return 0;
+}
+
+static const struct fb_ops gbefb_ops = {
+	.owner		= THIS_MODULE,
+	__FB_DEFAULT_IOMEM_OPS_RDWR,
+	.fb_check_var	= gbefb_check_var,
+	.fb_set_par	= gbefb_set_par,
+	.fb_setcolreg	= gbefb_setcolreg,
+	.fb_blank	= gbefb_blank,
+	__FB_DEFAULT_IOMEM_OPS_DRAW,
+	.fb_mmap	= gbefb_mmap,
+};
+
+/*
+ * sysfs
+ */
+
+static ssize_t gbefb_show_memsize(struct device *dev, struct device_attribute *attr, char *buf)
+{
+	return sysfs_emit(buf, "%u\n", gbe_mem_size);
+}
+
+static DEVICE_ATTR(size, S_IRUGO, gbefb_show_memsize, NULL);
+
+static ssize_t gbefb_show_rev(struct device *device, struct device_attribute *attr, char *buf)
+{
+	return sysfs_emit(buf, "%d\n", gbe_revision);
+}
+
+static DEVICE_ATTR(revision, S_IRUGO, gbefb_show_rev, NULL);
+
+static struct attribute *gbefb_attrs[] = {
+	&dev_attr_size.attr,
+	&dev_attr_revision.attr,
+	NULL,
+};
+ATTRIBUTE_GROUPS(gbefb);
+
+/*
+ * Initialization
+ */
+
+static int gbefb_setup(char *options)
+{
+	char *this_opt;
+
+	if (!options || !*options)
+		return 0;
+
+	while ((this_opt = strsep(&options, ",")) != NULL) {
+		if (!strncmp(this_opt, "monitor:", 8)) {
+			if (!strncmp(this_opt + 8, "crt", 3)) {
+				flat_panel_enabled = 0;
+				default_var = &default_var_CRT;
+				default_mode = &default_mode_CRT;
+			} else if (!strncmp(this_opt + 8, "1600sw", 6) ||
+				   !strncmp(this_opt + 8, "lcd", 3)) {
+				flat_panel_enabled = 1;
+				default_var = &default_var_LCD;
+				default_mode = &default_mode_LCD;
+			}
+		} else if (!strncmp(this_opt, "mem:", 4)) {
+			gbe_mem_size = memparse(this_opt + 4, &this_opt);
+			if (gbe_mem_size > CONFIG_FB_GBE_MEM * 1024 * 1024)
+				gbe_mem_size = CONFIG_FB_GBE_MEM * 1024 * 1024;
+			if (gbe_mem_size < TILE_SIZE)
+				gbe_mem_size = TILE_SIZE;
+		} else
+			mode_option = this_opt;
+	}
+	return 0;
+}
+
+static int gbefb_probe(struct platform_device *p_dev)
+{
+	int i, ret = 0;
+	struct fb_info *info;
+	struct gbefb_par *par;
+#ifndef MODULE
+	char *options = NULL;
+#endif
+
+	info = framebuffer_alloc(sizeof(struct gbefb_par), &p_dev->dev);
+	if (!info)
+		return -ENOMEM;
+
+#ifndef MODULE
+	if (fb_get_options("gbefb", &options)) {
+		ret = -ENODEV;
+		goto out_release_framebuffer;
+	}
+	gbefb_setup(options);
+#endif
+
+	if (!request_mem_region(GBE_BASE, sizeof(struct sgi_gbe), "GBE")) {
+		printk(KERN_ERR "gbefb: couldn't reserve mmio region\n");
+		ret = -EBUSY;
+		goto out_release_framebuffer;
+	}
+
+	gbe = (struct sgi_gbe *) devm_ioremap(&p_dev->dev, GBE_BASE,
+					      sizeof(struct sgi_gbe));
+	if (!gbe) {
+		printk(KERN_ERR "gbefb: couldn't map mmio region\n");
+		ret = -ENXIO;
+		goto out_release_mem_region;
+	}
+	gbe_revision = gbe->ctrlstat & 15;
+
+	gbe_tiles.cpu = dmam_alloc_coherent(&p_dev->dev,
+				GBE_TLB_SIZE * sizeof(uint16_t),
+				&gbe_tiles.dma, GFP_KERNEL);
+	if (!gbe_tiles.cpu) {
+		printk(KERN_ERR "gbefb: couldn't allocate tiles table\n");
+		ret = -ENOMEM;
+		goto out_release_mem_region;
+	}
+
+	if (gbe_mem_phys) {
+		/* memory was allocated at boot time */
+		gbe_mem = devm_ioremap_wc(&p_dev->dev, gbe_mem_phys,
+					  gbe_mem_size);
+		if (!gbe_mem) {
+			printk(KERN_ERR "gbefb: couldn't map framebuffer\n");
+			ret = -ENOMEM;
+			goto out_release_mem_region;
+		}
+
+		gbe_dma_addr = 0;
+	} else {
+		/* try to allocate memory with the classical allocator
+		 * this has high chance to fail on low memory machines */
+		gbe_mem = dmam_alloc_attrs(&p_dev->dev, gbe_mem_size,
+				&gbe_dma_addr, GFP_KERNEL,
+				DMA_ATTR_WRITE_COMBINE);
+		if (!gbe_mem) {
+			printk(KERN_ERR "gbefb: couldn't allocate framebuffer memory\n");
+			ret = -ENOMEM;
+			goto out_release_mem_region;
+		}
+
+		gbe_mem_phys = dma_to_phys(&p_dev->dev, gbe_dma_addr);
+	}
+
+	par = info->par;
+	par->wc_cookie = arch_phys_wc_add(gbe_mem_phys, gbe_mem_size);
+
+	/* map framebuffer memory into tiles table */
+	for (i = 0; i < (gbe_mem_size >> TILE_SHIFT); i++)
+		gbe_tiles.cpu[i] = (gbe_mem_phys >> TILE_SHIFT) + i;
+
+	info->fbops = &gbefb_ops;
+	info->pseudo_palette = pseudo_palette;
+	info->screen_base = gbe_mem;
+	fb_alloc_cmap(&info->cmap, 256, 0);
+
+	/* reset GBE */
+	gbe_reset();
+
+	/* turn on default video mode */
+	if (fb_find_mode(&par->var, info, mode_option, NULL, 0,
+			 default_mode, 8) == 0)
+		par->var = *default_var;
+	info->var = par->var;
+	gbefb_check_var(&par->var, info);
+	gbefb_encode_fix(&info->fix, &info->var);
+
+	if (register_framebuffer(info) < 0) {
+		printk(KERN_ERR "gbefb: couldn't register framebuffer\n");
+		ret = -ENXIO;
+		goto out_gbe_unmap;
+	}
+
+	platform_set_drvdata(p_dev, info);
+
+	fb_info(info, "%s rev %d @ 0x%08x using %dkB memory\n",
+		info->fix.id, gbe_revision, (unsigned)GBE_BASE,
+		gbe_mem_size >> 10);
+
+	return 0;
+
+out_gbe_unmap:
+	arch_phys_wc_del(par->wc_cookie);
+out_release_mem_region:
+	release_mem_region(GBE_BASE, sizeof(struct sgi_gbe));
+out_release_framebuffer:
+	framebuffer_release(info);
+
+	return ret;
+}
+
+static void gbefb_remove(struct platform_device* p_dev)
+{
+	struct fb_info *info = platform_get_drvdata(p_dev);
+	struct gbefb_par *par = info->par;
+
+	unregister_framebuffer(info);
+	gbe_turn_off();
+	arch_phys_wc_del(par->wc_cookie);
+	release_mem_region(GBE_BASE, sizeof(struct sgi_gbe));
+	framebuffer_release(info);
+}
+
+static struct platform_driver gbefb_driver = {
+	.probe = gbefb_probe,
+	.remove = gbefb_remove,
+	.driver = {
+		.name = "gbefb",
+		.dev_groups = gbefb_groups,
+	},
+};
+
+static struct platform_device *gbefb_device;
+
+static int __init gbefb_init(void)
+{
+	int ret = platform_driver_register(&gbefb_driver);
+	if (IS_ENABLED(CONFIG_SGI_IP32) && !ret) {
+		gbefb_device = platform_device_alloc("gbefb", 0);
+		if (gbefb_device) {
+			ret = platform_device_add(gbefb_device);
+		} else {
+			ret = -ENOMEM;
+		}
+		if (ret) {
+			platform_device_put(gbefb_device);
+			platform_driver_unregister(&gbefb_driver);
+		}
+	}
+	return ret;
+}
+
+static void __exit gbefb_exit(void)
+{
+	platform_device_unregister(gbefb_device);
+	platform_driver_unregister(&gbefb_driver);
+}
+
+module_init(gbefb_init);
+module_exit(gbefb_exit);
+
+MODULE_LICENSE("GPL");