lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-ID: <20091011071911.54347f2b@pedra.chehab.org>
Date:	Sun, 11 Oct 2009 07:19:11 -0300
From:	Mauro Carvalho Chehab <mchehab@...radead.org>
To:	Andy Whitcroft <apw@...onical.com>
Cc:	LKML <linux-kernel@...r.kernel.org>, Andy Walls <awalls@...ix.net>
Subject: checkpatch is not detecting bad whitespacing when using macros on
 formulas

Hi Andy Whitcroft,

While reviewing a patch from Andy Walls, I noticed that checkpatch v 0.29
didn't get any errors when bad whitespacing is used on some formulas:

$ checkpatch.pl cx23888-ir-part2_2.patch
total: 0 errors, 0 warnings, 1237 lines checked

cx23888-ir-part2_2.patch has no obvious style problems and is ready for submission.

However, there are several places at the patch like:

+static inline u16 ns_to_clock_divider(unsigned int ns)
+{
+       return count_to_clock_divider(
+                 DIV_ROUND_CLOSEST(CX23888_IR_REFCLK_FREQ/1000000 * ns, 1000));
+}

I tested even to run checkpatch with --strict:

$ checkpatch.pl --strict cx23888-ir-part2_2.patch
CHECK: struct mutex definition without comment
#126: FILE: linux/drivers/media/video/cx23885/cx23888-ir.c:123:
+       struct mutex rx_params_lock;

CHECK: spinlock_t definition without comment
#131: FILE: linux/drivers/media/video/cx23885/cx23888-ir.c:128:
+       spinlock_t rx_kfifo_lock;

CHECK: struct mutex definition without comment
#134: FILE: linux/drivers/media/video/cx23885/cx23888-ir.c:131:
+       struct mutex tx_params_lock;

CHECK: spinlock_t definition without comment
#138: FILE: linux/drivers/media/video/cx23885/cx23888-ir.c:135:
+       spinlock_t tx_kfifo_lock;

CHECK: if this code is redundant consider removing it
#316: FILE: linux/drivers/media/video/cx23885/cx23888-ir.c:285:
+#if 0

CHECK: multiple assignments should be avoided
#794: FILE: linux/drivers/media/video/cx23885/cx23888-ir.c:763:
+       o->mode = p->mode = V4L2_SUBDEV_IR_MODE_PULSE_WIDTH;

CHECK: multiple assignments should be avoided
#796: FILE: linux/drivers/media/video/cx23885/cx23888-ir.c:765:
+       o->bytes_per_data_element = p->bytes_per_data_element = sizeof(u32);

CHECK: multiple assignments should be avoided
#811: FILE: linux/drivers/media/video/cx23885/cx23888-ir.c:780:
+               o->duty_cycle = p->duty_cycle = 50;

CHECK: if this code is redundant consider removing it
#860: FILE: linux/drivers/media/video/cx23885/cx23888-ir.c:829:
+#if 0

CHECK: multiple assignments should be avoided
#953: FILE: linux/drivers/media/video/cx23885/cx23888-ir.c:922:
+       o->mode = p->mode = V4L2_SUBDEV_IR_MODE_PULSE_WIDTH;

CHECK: multiple assignments should be avoided
#955: FILE: linux/drivers/media/video/cx23885/cx23888-ir.c:924:
+       o->bytes_per_data_element = p->bytes_per_data_element = sizeof(u32);

total: 0 errors, 0 warnings, 11 checks, 1237 lines checked

cx23888-ir-part2_2.patch has style problems, please review.  If any of these errors
are false positives report them to the maintainer, see
CHECKPATCH in MAINTAINERS.

For your reference, I'm enclosing the patch with the bad whitespaces.

Could you please take a look on it



Cheers,
Mauro


---


cx23885: Complete CX23888 IR subdev implementation for Rx & almost for Tx

From: Andy Walls <awalls@...ix.net>

This change completes the v4l2_subdev implementation for IR receive for the
IR controller built into the CX23888.

This changes almost completes the IR transmit side also, but doesn't.  Instead
notes in the comments describe what needs to be done for IR Tx to work in the
subdevice implementation.

The current Tx behavior is skeletal and benign.  If left alone, it does nothing.
It will only ever generate a Tx interrupt on Tx init by a caller or when the
tx_write() method is called.  The ISR, when called, will then disable the Tx
FIFO service interrupt.

Priority: normal

Signed-off-by: Andy Walls <awalls@...ix.net>

--- a/linux/drivers/media/video/cx23885/cx23888-ir.c	Sun Sep 27 16:50:04 2009 -0400
+++ b/linux/drivers/media/video/cx23885/cx23888-ir.c	Sun Sep 27 17:31:37 2009 -0400
@@ -21,25 +21,79 @@
  *  02110-1301, USA.
  */
 
+#include <linux/kfifo.h>
+
 #include <media/v4l2-device.h>
 #include <media/v4l2-chip-ident.h>
 
 #include "cx23885.h"
 
+static unsigned int ir_888_debug;
+module_param(ir_888_debug, int, 0644);
+MODULE_PARM_DESC(ir_888_debug, "enable debug messages [CX23888 IR controller]");
+
 #define CX23888_IR_REG_BASE 	0x170000
 /*
  * These CX23888 register offsets have a straightforward one to one mapping
  * to the CX23885 register offsets of 0x200 through 0x218
  */
 #define CX23888_IR_CNTRL_REG	0x170000
+#define CNTRL_WIN_3_3	0x00000000
+#define CNTRL_WIN_4_3	0x00000001
+#define CNTRL_WIN_3_4	0x00000002
+#define CNTRL_WIN_4_4	0x00000003
+#define CNTRL_WIN	0x00000003
+#define CNTRL_EDG_NONE	0x00000000
+#define CNTRL_EDG_FALL	0x00000004
+#define CNTRL_EDG_RISE	0x00000008
+#define CNTRL_EDG_BOTH	0x0000000C
+#define CNTRL_EDG	0x0000000C
+#define CNTRL_DMD	0x00000010
+#define CNTRL_MOD	0x00000020
+#define CNTRL_RFE	0x00000040
+#define CNTRL_TFE	0x00000080
+#define CNTRL_RXE	0x00000100
+#define CNTRL_TXE	0x00000200
+#define CNTRL_RIC	0x00000400
+#define CNTRL_TIC	0x00000800
+#define CNTRL_CPL	0x00001000
+#define CNTRL_LBM	0x00002000
+#define CNTRL_R		0x00004000
+
 #define CX23888_IR_TXCLK_REG	0x170004
+#define TXCLK_TCD	0x0000FFFF
+
 #define CX23888_IR_RXCLK_REG	0x170008
+#define RXCLK_RCD	0x0000FFFF
+
 #define CX23888_IR_CDUTY_REG	0x17000C
+#define CDUTY_CDC	0x0000000F
+
 #define CX23888_IR_STATS_REG	0x170010
+#define STATS_RTO	0x00000001
+#define STATS_ROR	0x00000002
+#define STATS_RBY	0x00000004
+#define STATS_TBY	0x00000008
+#define STATS_RSR	0x00000010
+#define STATS_TSR	0x00000020
+
 #define CX23888_IR_IRQEN_REG	0x170014
+#define IRQEN_RTE	0x00000001
+#define IRQEN_ROE	0x00000002
+#define IRQEN_RSE	0x00000010
+#define IRQEN_TSE	0x00000020
+
 #define CX23888_IR_FILTR_REG	0x170018
+#define FILTR_LPF	0x0000FFFF
+
 /* This register doesn't follow the pattern; it's 0x23C on a CX23885 */
 #define CX23888_IR_FIFO_REG	0x170040
+#define FIFO_RXTX	0x0000FFFF
+#define FIFO_RXTX_LVL	0x00010000
+#define FIFO_RXTX_RTO	0x0001FFFF
+#define FIFO_RX_NDV	0x00020000
+#define FIFO_RX_DEPTH	8
+#define FIFO_TX_DEPTH	8
 
 /* CX23888 unique registers */
 #define CX23888_IR_SEEDP_REG	0x17001C
@@ -53,12 +107,32 @@
 #define CX23888_IR_DPIPG_REG	0x17003C
 #define CX23888_IR_LEARN_REG	0x170044
 
+#define CX23888_VIDCLK_FREQ	108000000 /* 108 MHz, BT.656 */
+#define CX23888_IR_REFCLK_FREQ	(CX23888_VIDCLK_FREQ/2)
+
+#define CX23888_IR_RX_KFIFO_SIZE	(512 * sizeof(u32))
+#define CX23888_IR_TX_KFIFO_SIZE	(512 * sizeof(u32))
 
 struct cx23888_ir_state {
 	struct v4l2_subdev sd;
 	struct cx23885_dev *dev;
 	u32 id;
 	u32 rev;
+
+	struct v4l2_subdev_ir_parameters rx_params;
+	struct mutex rx_params_lock;
+	atomic_t rxclk_divider;
+	atomic_t rx_invert;
+
+	struct kfifo *rx_kfifo;
+	spinlock_t rx_kfifo_lock;
+
+	struct v4l2_subdev_ir_parameters tx_params;
+	struct mutex tx_params_lock;
+	atomic_t txclk_divider;
+
+	struct kfifo *tx_kfifo;
+	spinlock_t tx_kfifo_lock;
 };
 
 static inline struct cx23888_ir_state *to_state(struct v4l2_subdev *sd)
@@ -66,17 +140,9 @@
 	return v4l2_get_subdevdata(sd);
 }
 
-static int cx23888_ir_write(struct cx23885_dev *dev, u32 addr, u8 value)
-{
-	u32 reg = (addr & ~3);
-	int shift = (addr & 3) * 8;
-	u32 x = cx_read(reg);
-
-	x = (x & ~(0xff << shift)) | ((u32)value << shift);
-	cx_write(reg, x);
-	return 0;
-}
-
+/*
+ * IR register block read and write functions
+ */
 static
 inline int cx23888_ir_write4(struct cx23885_dev *dev, u32 addr, u32 value)
 {
@@ -84,41 +150,959 @@
 	return 0;
 }
 
-static u8 cx23888_ir_read(struct cx23885_dev *dev, u32 addr)
-{
-	u32 x = cx_read((addr & ~3));
-	int shift = (addr & 3) * 8;
-
-	return (x >> shift) & 0xff;
-}
-
 static inline u32 cx23888_ir_read4(struct cx23885_dev *dev, u32 addr)
 {
 	return cx_read(addr);
 }
 
-static int cx23888_ir_and_or(struct cx23885_dev *dev, u32 addr,
-			     unsigned and_mask, u8 or_value)
-{
-	return cx23888_ir_write(dev, addr,
-				(cx23888_ir_read(dev, addr) & and_mask) |
-				or_value);
-}
-
 static inline int cx23888_ir_and_or4(struct cx23885_dev *dev, u32 addr,
 				     u32 and_mask, u32 or_value)
 {
-	cx_andor(addr, and_mask, or_value);
+	cx_andor(addr, ~and_mask, or_value);
 	return 0;
 }
 
+/*
+ * Rx and Tx Clock Divider register computations
+ *
+ * Note the largest clock divider value of 0xffff corresponds to:
+ * 	(0xffff + 1) * 1000 / 108/2 MHz = 1,213,629.629... ns
+ * which fits in 21 bits, so we'll use unsigned int for time arguments.
+ */
+static inline u16 count_to_clock_divider(unsigned int d)
+{
+	if (d > RXCLK_RCD+1)
+		d = RXCLK_RCD;
+	else if (d < 2)
+		d = 1;
+	else
+		d--;
+	return (u16) d;
+}
+
+static inline u16 ns_to_clock_divider(unsigned int ns)
+{
+	return count_to_clock_divider(
+		  DIV_ROUND_CLOSEST(CX23888_IR_REFCLK_FREQ/1000000 * ns, 1000));
+}
+
+static inline unsigned int clock_divider_to_ns(unsigned int divider)
+{
+	/* Period of the Rx or Tx clock in ns */
+	return DIV_ROUND_CLOSEST((divider + 1) * 1000,
+				 CX23888_IR_REFCLK_FREQ/1000000);
+}
+
+static inline u16 carrier_freq_to_clock_divider(unsigned int freq)
+{
+	return count_to_clock_divider(
+			  DIV_ROUND_CLOSEST(CX23888_IR_REFCLK_FREQ, freq * 16));
+}
+
+static inline unsigned int clock_divider_to_carrier_freq(unsigned int divider)
+{
+	return DIV_ROUND_CLOSEST(CX23888_IR_REFCLK_FREQ, (divider + 1) * 16);
+}
+
+static inline u16 freq_to_clock_divider(unsigned int freq,
+					unsigned int rollovers)
+{
+	return count_to_clock_divider(
+		   DIV_ROUND_CLOSEST(CX23888_IR_REFCLK_FREQ, freq * rollovers));
+}
+
+static inline unsigned int clock_divider_to_freq(unsigned int divider,
+						 unsigned int rollovers)
+{
+	return DIV_ROUND_CLOSEST(CX23888_IR_REFCLK_FREQ,
+				 (divider + 1) * rollovers);
+}
+
+/*
+ * Low Pass Filter register calculations
+ *
+ * Note the largest count value of 0xffff corresponds to:
+ * 	0xffff * 1000 / 108/2 MHz = 1,213,611.11... ns
+ * which fits in 21 bits, so we'll use unsigned int for time arguments.
+ */
+static inline u16 count_to_lpf_count(unsigned int d)
+{
+	if (d > FILTR_LPF)
+		d = FILTR_LPF;
+	else if (d < 4)
+		d = 0;
+	return (u16) d;
+}
+
+static inline u16 ns_to_lpf_count(unsigned int ns)
+{
+	return count_to_lpf_count(
+		  DIV_ROUND_CLOSEST(CX23888_IR_REFCLK_FREQ/1000000 * ns, 1000));
+}
+
+static inline unsigned int lpf_count_to_ns(unsigned int count)
+{
+	/* Duration of the Low Pass Filter rejection window in ns */
+	return DIV_ROUND_CLOSEST(count * 1000, CX23888_IR_REFCLK_FREQ/1000000);
+}
+
+static inline unsigned int lpf_count_to_us(unsigned int count)
+{
+	/* Duration of the Low Pass Filter rejection window in us */
+	return DIV_ROUND_CLOSEST(count, CX23888_IR_REFCLK_FREQ/1000000);
+}
+
+/*
+ * FIFO register pulse width count compuations
+ */
+static u32 clock_divider_to_resolution(u16 divider)
+{
+	/*
+	 * Resolution is the duration of 1 tick of the readable portion of
+	 * of the pulse width counter as read from the FIFO.  The two lsb's are
+	 * not readable, hence the << 2.  This function returns ns.
+	 */
+	return DIV_ROUND_CLOSEST((1 << 2)  * ((u32) divider + 1) * 1000,
+				 CX23888_IR_REFCLK_FREQ/1000000);
+}
+
+static u64 pulse_width_count_to_ns(u16 count, u16 divider)
+{
+	u64 n;
+	u32 rem;
+
+	/*
+	 * The 2 lsb's of the pulse width timer count are not readable, hence
+	 * the (count << 2) | 0x3
+	 */
+	n = (((u64) count << 2) | 0x3) * (divider + 1) * 1000; /* millicycles */
+	rem = do_div(n, CX23888_IR_REFCLK_FREQ/1000000);       /* / MHz => ns */
+	if (rem >= CX23888_IR_REFCLK_FREQ/1000000/2)
+		n++;
+	return n;
+}
+
+#if 0
+/* Keep as we will need this for Transmit functionality */
+static u16 ns_to_pulse_width_count(u32 ns, u16 divider)
+{
+	u64 n;
+	u32 d;
+	u32 rem;
+
+	/*
+	 * The 2 lsb's of the pulse width timer count are not accessable, hence
+	 * the (1 << 2)
+	 */
+	n = ((u64) ns) * CX23888_IR_REFCLK_FREQ/1000000; /* millicycles */
+	d = (1 << 2) * ((u32) divider + 1) * 1000; /* millicycles/count */
+	rem = do_div(n, d);
+	if (rem >= d/2)
+		n++;
+
+	if (n > FIFO_RXTX)
+		n = FIFO_RXTX;
+	else if (n == 0)
+		n = 1;
+	return (u16) n;
+}
+
+#endif
+static unsigned int pulse_width_count_to_us(u16 count, u16 divider)
+{
+	u64 n;
+	u32 rem;
+
+	/*
+	 * The 2 lsb's of the pulse width timer count are not readable, hence
+	 * the (count << 2) | 0x3
+	 */
+	n = (((u64) count << 2) | 0x3) * (divider + 1);  /* cycles      */
+	rem = do_div(n, CX23888_IR_REFCLK_FREQ/1000000); /* / MHz => us */
+	if (rem >= CX23888_IR_REFCLK_FREQ/1000000/2)
+		n++;
+	return (unsigned int) n;
+}
+
+/*
+ * Pulse Clocks computations: Combined Pulse Width Count & Rx Clock Counts
+ *
+ * The total pulse clock count is an 18 bit pulse width timer count as the most
+ * significant part and (up to) 16 bit clock divider count as a modulus.
+ * When the Rx clock divider ticks down to 0, it increments the 18 bit pulse
+ * width timer count's least significant bit.
+ */
+static u64 ns_to_pulse_clocks(u32 ns)
+{
+	u64 clocks;
+	u32 rem;
+	clocks = CX23888_IR_REFCLK_FREQ/1000000 * (u64) ns; /* millicycles    */
+	rem = do_div(clocks, 1000);                         /* /1000 = cycles */
+	if (rem >= 1000/2)
+		clocks++;
+	return clocks;
+}
+
+static u16 pulse_clocks_to_clock_divider(u64 count)
+{
+	u32 rem;
+
+	rem = do_div(count, (FIFO_RXTX << 2) | 0x3);
+
+	/* net result needs to be rounded down and decremented by 1 */
+	if (count > RXCLK_RCD+1)
+		count = RXCLK_RCD;
+	else if (count < 2)
+		count = 1;
+	else
+		count--;
+	return (u16) count;
+}
+
+/*
+ * IR Control Register helpers
+ */
+enum tx_fifo_watermark {
+	TX_FIFO_HALF_EMPTY = 0,
+	TX_FIFO_EMPTY      = CNTRL_TIC,
+};
+
+enum rx_fifo_watermark {
+	RX_FIFO_HALF_FULL = 0,
+	RX_FIFO_NOT_EMPTY = CNTRL_RIC,
+};
+
+static inline void control_tx_irq_watermark(struct cx23885_dev *dev,
+					    enum tx_fifo_watermark level)
+{
+	cx23888_ir_and_or4(dev, CX23888_IR_CNTRL_REG, ~CNTRL_TIC, level);
+}
+
+static inline void control_rx_irq_watermark(struct cx23885_dev *dev,
+					    enum rx_fifo_watermark level)
+{
+	cx23888_ir_and_or4(dev, CX23888_IR_CNTRL_REG, ~CNTRL_RIC, level);
+}
+
+static inline void control_tx_enable(struct cx23885_dev *dev, bool enable)
+{
+	cx23888_ir_and_or4(dev, CX23888_IR_CNTRL_REG, ~(CNTRL_TXE | CNTRL_TFE),
+			   enable ? (CNTRL_TXE | CNTRL_TFE) : 0);
+}
+
+static inline void control_rx_enable(struct cx23885_dev *dev, bool enable)
+{
+	cx23888_ir_and_or4(dev, CX23888_IR_CNTRL_REG, ~(CNTRL_RXE | CNTRL_RFE),
+			   enable ? (CNTRL_RXE | CNTRL_RFE) : 0);
+}
+
+static inline void control_tx_modulation_enable(struct cx23885_dev *dev,
+						bool enable)
+{
+	cx23888_ir_and_or4(dev, CX23888_IR_CNTRL_REG, ~CNTRL_MOD,
+			   enable ? CNTRL_MOD : 0);
+}
+
+static inline void control_rx_demodulation_enable(struct cx23885_dev *dev,
+						  bool enable)
+{
+	cx23888_ir_and_or4(dev, CX23888_IR_CNTRL_REG, ~CNTRL_DMD,
+			   enable ? CNTRL_DMD : 0);
+}
+
+static inline void control_rx_s_edge_detection(struct cx23885_dev *dev,
+					       u32 edge_types)
+{
+	cx23888_ir_and_or4(dev, CX23888_IR_CNTRL_REG, ~CNTRL_EDG_BOTH,
+			   edge_types & CNTRL_EDG_BOTH);
+}
+
+static void control_rx_s_carrier_window(struct cx23885_dev *dev,
+					unsigned int carrier,
+					unsigned int *carrier_range_low,
+					unsigned int *carrier_range_high)
+{
+	u32 v;
+	unsigned int c16 = carrier * 16;
+
+	if (*carrier_range_low < DIV_ROUND_CLOSEST(c16, 16 + 3)) {
+		v = CNTRL_WIN_3_4;
+		*carrier_range_low = DIV_ROUND_CLOSEST(c16, 16 + 4);
+	} else {
+		v = CNTRL_WIN_3_3;
+		*carrier_range_low = DIV_ROUND_CLOSEST(c16, 16 + 3);
+	}
+
+	if (*carrier_range_high > DIV_ROUND_CLOSEST(c16, 16 - 3)) {
+		v |= CNTRL_WIN_4_3;
+		*carrier_range_high = DIV_ROUND_CLOSEST(c16, 16 - 4);
+	} else {
+		v |= CNTRL_WIN_3_3;
+		*carrier_range_high = DIV_ROUND_CLOSEST(c16, 16 - 3);
+	}
+	cx23888_ir_and_or4(dev, CX23888_IR_CNTRL_REG, ~CNTRL_WIN, v);
+}
+
+static inline void control_tx_polarity_invert(struct cx23885_dev *dev,
+					      bool invert)
+{
+	cx23888_ir_and_or4(dev, CX23888_IR_CNTRL_REG, ~CNTRL_CPL,
+			   invert ? CNTRL_CPL : 0);
+}
+
+/*
+ * IR Rx & Tx Clock Register helpers
+ */
+static unsigned int txclk_tx_s_carrier(struct cx23885_dev *dev,
+				       unsigned int freq,
+				       u16 *divider)
+{
+	*divider = carrier_freq_to_clock_divider(freq);
+	cx23888_ir_write4(dev, CX23888_IR_TXCLK_REG, *divider);
+	return clock_divider_to_carrier_freq(*divider);
+}
+
+static unsigned int rxclk_rx_s_carrier(struct cx23885_dev *dev,
+				       unsigned int freq,
+				       u16 *divider)
+{
+	*divider = carrier_freq_to_clock_divider(freq);
+	cx23888_ir_write4(dev, CX23888_IR_RXCLK_REG, *divider);
+	return clock_divider_to_carrier_freq(*divider);
+}
+
+static u32 txclk_tx_s_max_pulse_width(struct cx23885_dev *dev, u32 ns,
+				      u16 *divider)
+{
+	u64 pulse_clocks;
+
+	if (ns > V4L2_SUBDEV_IR_PULSE_MAX_WIDTH_NS)
+		ns = V4L2_SUBDEV_IR_PULSE_MAX_WIDTH_NS;
+	pulse_clocks = ns_to_pulse_clocks(ns);
+	*divider = pulse_clocks_to_clock_divider(pulse_clocks);
+	cx23888_ir_write4(dev, CX23888_IR_TXCLK_REG, *divider);
+	return (u32) pulse_width_count_to_ns(FIFO_RXTX, *divider);
+}
+
+static u32 rxclk_rx_s_max_pulse_width(struct cx23885_dev *dev, u32 ns,
+				      u16 *divider)
+{
+	u64 pulse_clocks;
+
+	if (ns > V4L2_SUBDEV_IR_PULSE_MAX_WIDTH_NS)
+		ns = V4L2_SUBDEV_IR_PULSE_MAX_WIDTH_NS;
+	pulse_clocks = ns_to_pulse_clocks(ns);
+	*divider = pulse_clocks_to_clock_divider(pulse_clocks);
+	cx23888_ir_write4(dev, CX23888_IR_RXCLK_REG, *divider);
+	return (u32) pulse_width_count_to_ns(FIFO_RXTX, *divider);
+}
+
+/*
+ * IR Tx Carrier Duty Cycle register helpers
+ */
+static unsigned int cduty_tx_s_duty_cycle(struct cx23885_dev *dev,
+					  unsigned int duty_cycle)
+{
+	u32 n;
+	n = DIV_ROUND_CLOSEST(duty_cycle * 100, 625); /* 16ths of 100% */
+	if (n != 0)
+		n--;
+	if (n > 15)
+		n = 15;
+	cx23888_ir_write4(dev, CX23888_IR_CDUTY_REG, n);
+	return DIV_ROUND_CLOSEST((n+1) * 100, 16);
+}
+
+/*
+ * IR Filter Register helpers
+ */
+static u32 filter_rx_s_min_width(struct cx23885_dev *dev, u32 min_width_ns)
+{
+	u32 count = ns_to_lpf_count(min_width_ns);
+	cx23888_ir_write4(dev, CX23888_IR_FILTR_REG, count);
+	return lpf_count_to_ns(count);
+}
+
+/*
+ * IR IRQ Enable Register helpers
+ */
+static inline void irqenable_rx(struct cx23885_dev *dev, u32 mask)
+{
+	mask &= (IRQEN_RTE | IRQEN_ROE | IRQEN_RSE);
+	cx23888_ir_and_or4(dev, CX23888_IR_IRQEN_REG,
+			   ~(IRQEN_RTE | IRQEN_ROE | IRQEN_RSE), mask);
+}
+
+static inline void irqenable_tx(struct cx23885_dev *dev, u32 mask)
+{
+	mask &= IRQEN_TSE;
+	cx23888_ir_and_or4(dev, CX23888_IR_IRQEN_REG, ~IRQEN_TSE, mask);
+}
+
+/*
+ * V4L2 Subdevice IR Ops
+ */
+static int cx23888_ir_irq_handler(struct v4l2_subdev *sd, u32 status,
+				  bool *handled)
+{
+	struct cx23888_ir_state *state = to_state(sd);
+	struct cx23885_dev *dev = state->dev;
+
+	u32 cntrl = cx23888_ir_read4(dev, CX23888_IR_CNTRL_REG);
+	u32 irqen = cx23888_ir_read4(dev, CX23888_IR_IRQEN_REG);
+	u32 stats = cx23888_ir_read4(dev, CX23888_IR_STATS_REG);
+
+	u32 rx_data[FIFO_RX_DEPTH];
+	int i, j, k;
+	u32 events, v;
+	int tsr, rsr, rto, ror, tse, rse, rte, roe, kror;
+
+	tsr = stats & STATS_TSR; /* Tx FIFO Service Request */
+	rsr = stats & STATS_RSR; /* Rx FIFO Service Request */
+	rto = stats & STATS_RTO; /* Rx Pulse Width Timer Time Out */
+	ror = stats & STATS_ROR; /* Rx FIFO Over Run */
+
+	tse = irqen & IRQEN_TSE; /* Tx FIFO Service Request IRQ Enable */
+	rse = irqen & IRQEN_RSE; /* Rx FIFO Service Reuqest IRQ Enable */
+	rte = irqen & IRQEN_RTE; /* Rx Pulse Width Timer Time Out IRQ Enable */
+	roe = irqen & IRQEN_ROE; /* Rx FIFO Over Run IRQ Enable */
+
+	*handled = false;
+	v4l2_dbg(2, ir_888_debug, sd, "IRQ Status:  %s %s %s %s %s %s\n",
+		 tsr ? "tsr" : "   ", rsr ? "rsr" : "   ",
+		 rto ? "rto" : "   ", ror ? "ror" : "   ",
+		 stats & STATS_TBY ? "tby" : "   ",
+		 stats & STATS_RBY ? "rby" : "   ");
+
+	v4l2_dbg(2, ir_888_debug, sd, "IRQ Enables: %s %s %s %s\n",
+		 tse ? "tse" : "   ", rse ? "rse" : "   ",
+		 rte ? "rte" : "   ", roe ? "roe" : "   ");
+
+	/*
+	 * Transmitter interrupt service
+	 */
+	if (tse && tsr) {
+		/*
+		 * TODO:
+		 * Check the watermark threshold setting
+		 * Pull FIFO_TX_DEPTH or FIFO_TX_DEPTH/2 entries from tx_kfifo
+		 * Push the data to the hardware FIFO.
+		 * If there was nothing more to send in the tx_kfifo, disable
+		 *	the TSR IRQ and notify the v4l2_device.
+		 * If there was something in the tx_kfifo, check the tx_kfifo
+		 *      level and notify the v4l2_device, if it is low.
+		 */
+		/* For now, inhibit TSR interrupt until Tx is implemented */
+		irqenable_tx(dev, 0);
+		events = V4L2_SUBDEV_IR_TX_FIFO_SERVICE_REQ;
+		v4l2_subdev_notify(sd, V4L2_SUBDEV_IR_TX_NOTIFY, &events);
+		*handled = true;
+	}
+
+	/*
+	 * Receiver interrupt service
+	 */
+	kror = 0;
+	if ((rse && rsr) || (rte && rto)) {
+		/*
+		 * Receive data on RSR to clear the STATS_RSR.
+		 * Receive data on RTO, since we may not have yet hit the RSR
+		 * watermark when we receive the RTO.
+		 */
+		for (i = 0, v = FIFO_RX_NDV;
+		     (v & FIFO_RX_NDV) && !kror; i = 0) {
+			for (j = 0;
+			     (v & FIFO_RX_NDV) && j < FIFO_RX_DEPTH; j++) {
+				v = cx23888_ir_read4(dev, CX23888_IR_FIFO_REG);
+				rx_data[i++] = v & ~FIFO_RX_NDV;
+			}
+			if (i == 0)
+				break;
+			j = i * sizeof(u32);
+			k = kfifo_put(state->rx_kfifo,
+				      (unsigned char *) rx_data, j);
+			if (k != j)
+				kror++; /* rx_kfifo over run */
+		}
+		*handled = true;
+	}
+
+	events = 0;
+	v = 0;
+	if (kror) {
+		events |= V4L2_SUBDEV_IR_RX_SW_FIFO_OVERRUN;
+		v4l2_err(sd, "IR receiver software FIFO overrun\n");
+	}
+	if (roe && ror) {
+		/*
+		 * The RX FIFO Enable (CNTRL_RFE) must be toggled to clear
+		 * the Rx FIFO Over Run status (STATS_ROR)
+		 */
+		v |= CNTRL_RFE;
+		events |= V4L2_SUBDEV_IR_RX_HW_FIFO_OVERRUN;
+		v4l2_err(sd, "IR receiver hardware FIFO overrun\n");
+	}
+	if (rte && rto) {
+		/*
+		 * The IR Receiver Enable (CNTRL_RXE) must be toggled to clear
+		 * the Rx Pulse Width Timer Time Out (STATS_RTO)
+		 */
+		v |= CNTRL_RXE;
+		events |= V4L2_SUBDEV_IR_RX_END_OF_RX_DETECTED;
+	}
+	if (v) {
+		/* Clear STATS_ROR & STATS_RTO as needed by reseting hardware */
+		cx23888_ir_write4(dev, CX23888_IR_CNTRL_REG, cntrl & ~v);
+		cx23888_ir_write4(dev, CX23888_IR_CNTRL_REG, cntrl);
+		*handled = true;
+	}
+	if (kfifo_len(state->rx_kfifo) >= CX23888_IR_RX_KFIFO_SIZE/2)
+		events |= V4L2_SUBDEV_IR_RX_FIFO_SERVICE_REQ;
+
+	if (events)
+		v4l2_subdev_notify(sd, V4L2_SUBDEV_IR_RX_NOTIFY, &events);
+	return 0;
+}
+
+/* Receiver */
+static int cx23888_ir_rx_read(struct v4l2_subdev *sd, u8 *buf, size_t count,
+			      ssize_t *num)
+{
+	struct cx23888_ir_state *state = to_state(sd);
+	bool invert = (bool) atomic_read(&state->rx_invert);
+	u16 divider = (u16) atomic_read(&state->rxclk_divider);
+
+	unsigned int i, n;
+	u32 *p;
+	u32 u, v;
+
+	n = count / sizeof(u32) * sizeof(u32);
+	if (n == 0) {
+		*num = 0;
+		return 0;
+	}
+
+	n = kfifo_get(state->rx_kfifo, buf, n);
+
+	n /= sizeof(u32);
+	*num = n * sizeof(u32);
+
+	for (p = (u32 *) buf, i = 0; i < n; p++, i++) {
+		if ((*p & FIFO_RXTX_RTO) == FIFO_RXTX_RTO) {
+			*p = V4L2_SUBDEV_IR_PULSE_RX_SEQ_END;
+			v4l2_dbg(2, ir_888_debug, sd, "rx read: end of rx\n");
+			continue;
+		}
+
+		u = (*p & FIFO_RXTX_LVL) ? V4L2_SUBDEV_IR_PULSE_LEVEL_MASK : 0;
+		if (invert)
+			u = u ? 0 : V4L2_SUBDEV_IR_PULSE_LEVEL_MASK;
+
+		v = (u32) pulse_width_count_to_ns((u16) (*p & FIFO_RXTX),
+						  divider);
+		if (v >= V4L2_SUBDEV_IR_PULSE_MAX_WIDTH_NS)
+			v = V4L2_SUBDEV_IR_PULSE_MAX_WIDTH_NS - 1;
+
+		*p = u | v;
+
+		v4l2_dbg(2, ir_888_debug, sd, "rx read: %10u ns  %s\n",
+			 v, u ? "mark" : "space");
+	}
+	return 0;
+}
+
+static int cx23888_ir_rx_g_parameters(struct v4l2_subdev *sd,
+				      struct v4l2_subdev_ir_parameters *p)
+{
+	struct cx23888_ir_state *state = to_state(sd);
+	mutex_lock(&state->rx_params_lock);
+	memcpy(p, &state->rx_params, sizeof(struct v4l2_subdev_ir_parameters));
+	mutex_unlock(&state->rx_params_lock);
+	return 0;
+}
+
+static int cx23888_ir_rx_shutdown(struct v4l2_subdev *sd)
+{
+	struct cx23888_ir_state *state = to_state(sd);
+	struct cx23885_dev *dev = state->dev;
+
+	mutex_lock(&state->rx_params_lock);
+
+	/* Disable or slow down all IR Rx circuits and counters */
+	irqenable_rx(dev, 0);
+	control_rx_enable(dev, false);
+	control_rx_demodulation_enable(dev, false);
+	control_rx_s_edge_detection(dev, CNTRL_EDG_NONE);
+	filter_rx_s_min_width(dev, 0);
+	cx23888_ir_write4(dev, CX23888_IR_RXCLK_REG, RXCLK_RCD);
+
+	state->rx_params.shutdown = true;
+
+	mutex_unlock(&state->rx_params_lock);
+	return 0;
+}
+
+static int cx23888_ir_rx_s_parameters(struct v4l2_subdev *sd,
+				      struct v4l2_subdev_ir_parameters *p)
+{
+	struct cx23888_ir_state *state = to_state(sd);
+	struct cx23885_dev *dev = state->dev;
+	struct v4l2_subdev_ir_parameters *o = &state->rx_params;
+	u16 rxclk_divider;
+
+	if (p->shutdown)
+		return cx23888_ir_rx_shutdown(sd);
+
+	if (p->mode != V4L2_SUBDEV_IR_MODE_PULSE_WIDTH)
+		return -ENOSYS;
+
+	mutex_lock(&state->rx_params_lock);
+
+	o->shutdown = p->shutdown;
+
+	o->mode = p->mode = V4L2_SUBDEV_IR_MODE_PULSE_WIDTH;
+
+	o->bytes_per_data_element = p->bytes_per_data_element = sizeof(u32);
+
+	/* Before we tweak the hardware, we have to disable the receiver */
+	irqenable_rx(dev, 0);
+	control_rx_enable(dev, false);
+
+	control_rx_demodulation_enable(dev, p->modulation);
+	o->modulation = p->modulation;
+
+	if (p->modulation) {
+		p->carrier_freq = rxclk_rx_s_carrier(dev, p->carrier_freq,
+						     &rxclk_divider);
+
+		o->carrier_freq = p->carrier_freq;
+
+		o->duty_cycle = p->duty_cycle = 50;
+
+		control_rx_s_carrier_window(dev, p->carrier_freq,
+					    &p->carrier_range_lower,
+					    &p->carrier_range_upper);
+		o->carrier_range_lower = p->carrier_range_lower;
+		o->carrier_range_upper = p->carrier_range_upper;
+	} else {
+		p->max_pulse_width =
+			    rxclk_rx_s_max_pulse_width(dev, p->max_pulse_width,
+						       &rxclk_divider);
+		o->max_pulse_width = p->max_pulse_width;
+	}
+	atomic_set(&state->rxclk_divider, rxclk_divider);
+
+	p->noise_filter_min_width =
+			  filter_rx_s_min_width(dev, p->noise_filter_min_width);
+	o->noise_filter_min_width = p->noise_filter_min_width;
+
+	p->resolution = clock_divider_to_resolution(rxclk_divider);
+	o->resolution = p->resolution;
+
+	/* FIXME - make this dependent on resolution for better performance */
+	control_rx_irq_watermark(dev, RX_FIFO_HALF_FULL);
+
+	control_rx_s_edge_detection(dev, CNTRL_EDG_BOTH);
+
+	o->invert = p->invert;
+	atomic_set(&state->rx_invert, p->invert);
+
+	o->interrupt_enable = p->interrupt_enable;
+	o->enable = p->enable;
+	if (p->enable) {
+		kfifo_reset(state->rx_kfifo);
+		if (p->interrupt_enable)
+			irqenable_rx(dev, IRQEN_RSE | IRQEN_RTE | IRQEN_ROE);
+		control_rx_enable(dev, p->enable);
+	}
+
+	mutex_unlock(&state->rx_params_lock);
+	return 0;
+}
+
+/* Transmitter */
+static int cx23888_ir_tx_write(struct v4l2_subdev *sd, u8 *buf, size_t count,
+			       ssize_t *num)
+{
+	struct cx23888_ir_state *state = to_state(sd);
+	struct cx23885_dev *dev = state->dev;
+#if 0
+	/*
+	 * FIXME - the code below is an incomplete and untested sketch of what
+	 * may need to be done.  The critical part is to get 4 (or 8) pulses
+	 * from the tx_kfifo, or converted from ns to the proper units from the
+	 * input, and push them off to the hardware Tx FIFO right away, if the
+	 * HW TX fifo needs service.  The rest can be pushed to the tx_kfifo in
+	 * a less critical timeframe.  Also watch out for overruning the
+	 * tx_kfifo - don't let it happen and let the caller know not all his
+	 * pulses were written.
+	 */
+	u32 *ns_pulse = (u32 *) buf;
+	unsigned int n;
+	u32 fifo_pulse[FIFO_TX_DEPTH];
+	u32 mark;
+
+	/* Compute how much we can fit in the tx kfifo */
+	n = CX23888_IR_TX_KFIFO_SIZE - kfifo_len(state->tx_kfifo);
+	n = min(n, (unsigned int) count);
+	n /= sizeof(u32);
+
+	/* FIXME - turn on Tx Fifo service interrupt
+	 * check hardware fifo level, and other stuff
+	 */
+	for (i = 0; i < n; ) {
+		for (j = 0; j < FIFO_TX_DEPTH/2 && i < n; j++) {
+			mark = ns_pulse[i] & V4L2_SUBDEV_IR_PULSE_LEVEL_MASK;
+			fifo_pulse[j] = ns_to_pulse_width_count(
+					 ns_pulse[i] &
+					       ~V4L2_SUBDEV_IR_PULSE_LEVEL_MASK,
+					 state->txclk_divider);
+			if (mark)
+				fifo_pulse[j] &= FIFO_RXTX_LVL;
+			i++;
+		}
+		kfifo_put(state->tx_kfifo, (u8 *) fifo_pulse, j * sizeof(u32));
+	}
+	*num = n * sizeof(u32);
+#else
+	/* For now enable the Tx FIFO Service interrupt & pretend we did work */
+	irqenable_tx(dev, IRQEN_TSE);
+	*num = count;
+#endif
+	return 0;
+}
+
+static int cx23888_ir_tx_g_parameters(struct v4l2_subdev *sd,
+				      struct v4l2_subdev_ir_parameters *p)
+{
+	struct cx23888_ir_state *state = to_state(sd);
+	mutex_lock(&state->tx_params_lock);
+	memcpy(p, &state->tx_params, sizeof(struct v4l2_subdev_ir_parameters));
+	mutex_unlock(&state->tx_params_lock);
+	return 0;
+}
+
+static int cx23888_ir_tx_shutdown(struct v4l2_subdev *sd)
+{
+	struct cx23888_ir_state *state = to_state(sd);
+	struct cx23885_dev *dev = state->dev;
+
+	mutex_lock(&state->tx_params_lock);
+
+	/* Disable or slow down all IR Tx circuits and counters */
+	irqenable_tx(dev, 0);
+	control_tx_enable(dev, false);
+	control_tx_modulation_enable(dev, false);
+	cx23888_ir_write4(dev, CX23888_IR_TXCLK_REG, TXCLK_TCD);
+
+	state->tx_params.shutdown = true;
+
+	mutex_unlock(&state->tx_params_lock);
+	return 0;
+}
+
+static int cx23888_ir_tx_s_parameters(struct v4l2_subdev *sd,
+				      struct v4l2_subdev_ir_parameters *p)
+{
+	struct cx23888_ir_state *state = to_state(sd);
+	struct cx23885_dev *dev = state->dev;
+	struct v4l2_subdev_ir_parameters *o = &state->tx_params;
+	u16 txclk_divider;
+
+	if (p->shutdown)
+		return cx23888_ir_tx_shutdown(sd);
+
+	if (p->mode != V4L2_SUBDEV_IR_MODE_PULSE_WIDTH)
+		return -ENOSYS;
+
+	mutex_lock(&state->tx_params_lock);
+
+	o->shutdown = p->shutdown;
+
+	o->mode = p->mode = V4L2_SUBDEV_IR_MODE_PULSE_WIDTH;
+
+	o->bytes_per_data_element = p->bytes_per_data_element = sizeof(u32);
+
+	/* Before we tweak the hardware, we have to disable the transmitter */
+	irqenable_tx(dev, 0);
+	control_tx_enable(dev, false);
+
+	control_tx_modulation_enable(dev, p->modulation);
+	o->modulation = p->modulation;
+
+	if (p->modulation) {
+		p->carrier_freq = txclk_tx_s_carrier(dev, p->carrier_freq,
+						     &txclk_divider);
+		o->carrier_freq = p->carrier_freq;
+
+		p->duty_cycle = cduty_tx_s_duty_cycle(dev, p->duty_cycle);
+		o->duty_cycle = p->duty_cycle;
+	} else {
+		p->max_pulse_width =
+			    txclk_tx_s_max_pulse_width(dev, p->max_pulse_width,
+						       &txclk_divider);
+		o->max_pulse_width = p->max_pulse_width;
+	}
+	atomic_set(&state->txclk_divider, txclk_divider);
+
+	p->resolution = clock_divider_to_resolution(txclk_divider);
+	o->resolution = p->resolution;
+
+	/* FIXME - make this dependent on resolution for better performance */
+	control_tx_irq_watermark(dev, TX_FIFO_HALF_EMPTY);
+
+	control_tx_polarity_invert(dev, p->invert);
+	o->invert = p->invert;
+
+	o->interrupt_enable = p->interrupt_enable;
+	o->enable = p->enable;
+	if (p->enable) {
+		kfifo_reset(state->tx_kfifo);
+		if (p->interrupt_enable)
+			irqenable_tx(dev, IRQEN_TSE);
+		control_tx_enable(dev, p->enable);
+	}
+
+	mutex_unlock(&state->tx_params_lock);
+	return 0;
+}
+
+
+/*
+ * V4L2 Subdevice Core Ops
+ */
 static int cx23888_ir_log_status(struct v4l2_subdev *sd)
 {
 	struct cx23888_ir_state *state = to_state(sd);
 	struct cx23885_dev *dev = state->dev;
-	u8 cntrl = cx23888_ir_read(dev, CX23888_IR_CNTRL_REG+1);
-	v4l2_info(sd, "receiver    %sabled\n", cntrl & 0x1 ? "en" : "dis");
-	v4l2_info(sd, "transmitter %sabled\n", cntrl & 0x2 ? "en" : "dis");
+	char *s;
+	int i, j;
+
+	u32 cntrl = cx23888_ir_read4(dev, CX23888_IR_CNTRL_REG);
+	u32 txclk = cx23888_ir_read4(dev, CX23888_IR_TXCLK_REG) & TXCLK_TCD;
+	u32 rxclk = cx23888_ir_read4(dev, CX23888_IR_RXCLK_REG) & RXCLK_RCD;
+	u32 cduty = cx23888_ir_read4(dev, CX23888_IR_CDUTY_REG) & CDUTY_CDC;
+	u32 stats = cx23888_ir_read4(dev, CX23888_IR_STATS_REG);
+	u32 irqen = cx23888_ir_read4(dev, CX23888_IR_IRQEN_REG);
+	u32 filtr = cx23888_ir_read4(dev, CX23888_IR_FILTR_REG) & FILTR_LPF;
+
+	v4l2_info(sd, "IR Receiver:\n");
+	v4l2_info(sd, "\tEnabled:                           %s\n",
+		  cntrl & CNTRL_RXE ? "yes" : "no");
+	v4l2_info(sd, "\tDemodulation from a carrier:       %s\n",
+		  cntrl & CNTRL_DMD ? "enabled" : "disabled");
+	v4l2_info(sd, "\tFIFO:                              %s\n",
+		  cntrl & CNTRL_RFE ? "enabled" : "disabled");
+	switch (cntrl & CNTRL_EDG) {
+	case CNTRL_EDG_NONE:
+		s = "disabled";
+		break;
+	case CNTRL_EDG_FALL:
+		s = "falling edge";
+		break;
+	case CNTRL_EDG_RISE:
+		s = "rising edge";
+		break;
+	case CNTRL_EDG_BOTH:
+		s = "rising & falling edges";
+		break;
+	default:
+		s = "??? edge";
+		break;
+	}
+	v4l2_info(sd, "\tPulse timers' start/stop trigger:  %s\n", s);
+	v4l2_info(sd, "\tFIFO data on pulse timer overflow: %s\n",
+		  cntrl & CNTRL_R ? "not loaded" : "overflow marker");
+	v4l2_info(sd, "\tFIFO interrupt watermark:          %s\n",
+		  cntrl & CNTRL_RIC ? "not empty" : "half full or greater");
+	v4l2_info(sd, "\tLoopback mode:                     %s\n",
+		  cntrl & CNTRL_LBM ? "loopback active" : "normal receive");
+	if (cntrl & CNTRL_DMD) {
+		v4l2_info(sd, "\tExpected carrier (16 clocks):      %u Hz\n",
+			  clock_divider_to_carrier_freq(rxclk));
+		switch (cntrl & CNTRL_WIN) {
+		case CNTRL_WIN_3_3:
+			i = 3;
+			j = 3;
+			break;
+		case CNTRL_WIN_4_3:
+			i = 4;
+			j = 3;
+			break;
+		case CNTRL_WIN_3_4:
+			i = 3;
+			j = 4;
+			break;
+		case CNTRL_WIN_4_4:
+			i = 4;
+			j = 4;
+			break;
+		default:
+			i = 0;
+			j = 0;
+			break;
+		}
+		v4l2_info(sd, "\tNext carrier edge window:          16 clocks "
+			  "-%1d/+%1d, %u to %u Hz\n", i, j,
+			  clock_divider_to_freq(rxclk, 16 + j),
+			  clock_divider_to_freq(rxclk, 16 - i));
+	} else {
+		v4l2_info(sd, "\tMax measurable pulse width:        %u us, "
+			  "%llu ns\n",
+			  pulse_width_count_to_us(FIFO_RXTX, rxclk),
+			  pulse_width_count_to_ns(FIFO_RXTX, rxclk));
+	}
+	v4l2_info(sd, "\tLow pass filter:                   %s\n",
+		  filtr ? "enabled" : "disabled");
+	if (filtr)
+		v4l2_info(sd, "\tMin acceptable pulse width (LPF):  %u us, "
+			  "%u ns\n",
+			  lpf_count_to_us(filtr),
+			  lpf_count_to_ns(filtr));
+	v4l2_info(sd, "\tPulse width timer timed-out:       %s\n",
+		  stats & STATS_RTO ? "yes" : "no");
+	v4l2_info(sd, "\tPulse width timer time-out intr:   %s\n",
+		  irqen & IRQEN_RTE ? "enabled" : "disabled");
+	v4l2_info(sd, "\tFIFO overrun:                      %s\n",
+		  stats & STATS_ROR ? "yes" : "no");
+	v4l2_info(sd, "\tFIFO overrun interrupt:            %s\n",
+		  irqen & IRQEN_ROE ? "enabled" : "disabled");
+	v4l2_info(sd, "\tBusy:                              %s\n",
+		  stats & STATS_RBY ? "yes" : "no");
+	v4l2_info(sd, "\tFIFO service requested:            %s\n",
+		  stats & STATS_RSR ? "yes" : "no");
+	v4l2_info(sd, "\tFIFO service request interrupt:    %s\n",
+		  irqen & IRQEN_RSE ? "enabled" : "disabled");
+
+	v4l2_info(sd, "IR Transmitter:\n");
+	v4l2_info(sd, "\tEnabled:                           %s\n",
+		  cntrl & CNTRL_TXE ? "yes" : "no");
+	v4l2_info(sd, "\tModulation onto a carrier:         %s\n",
+		  cntrl & CNTRL_MOD ? "enabled" : "disabled");
+	v4l2_info(sd, "\tFIFO:                              %s\n",
+		  cntrl & CNTRL_TFE ? "enabled" : "disabled");
+	v4l2_info(sd, "\tFIFO interrupt watermark:          %s\n",
+		  cntrl & CNTRL_TIC ? "not empty" : "half full or less");
+	v4l2_info(sd, "\tSignal polarity:                   %s\n",
+		  cntrl & CNTRL_CPL ? "0:mark 1:space" : "0:space 1:mark");
+	if (cntrl & CNTRL_MOD) {
+		v4l2_info(sd, "\tCarrier (16 clocks):               %u Hz\n",
+			  clock_divider_to_carrier_freq(txclk));
+		v4l2_info(sd, "\tCarrier duty cycle:                %2u/16\n",
+			  cduty + 1);
+	} else {
+		v4l2_info(sd, "\tMax pulse width:                   %u us, "
+			  "%llu ns\n",
+			  pulse_width_count_to_us(FIFO_RXTX, txclk),
+			  pulse_width_count_to_ns(FIFO_RXTX, txclk));
+	}
+	v4l2_info(sd, "\tBusy:                              %s\n",
+		  stats & STATS_TBY ? "yes" : "no");
+	v4l2_info(sd, "\tFIFO service requested:            %s\n",
+		  stats & STATS_TSR ? "yes" : "no");
+	v4l2_info(sd, "\tFIFO service request interrupt:    %s\n",
+		  irqen & IRQEN_TSE ? "enabled" : "disabled");
+
 	return 0;
 }
 
@@ -187,19 +1171,81 @@
 #endif
 };
 
+static const struct v4l2_subdev_ir_ops cx23888_ir_ir_ops = {
+	.interrupt_service_routine = cx23888_ir_irq_handler,
+
+	.rx_read = cx23888_ir_rx_read,
+	.rx_g_parameters = cx23888_ir_rx_g_parameters,
+	.rx_s_parameters = cx23888_ir_rx_s_parameters,
+
+	.tx_write = cx23888_ir_tx_write,
+	.tx_g_parameters = cx23888_ir_tx_g_parameters,
+	.tx_s_parameters = cx23888_ir_tx_s_parameters,
+};
+
 static const struct v4l2_subdev_ops cx23888_ir_controller_ops = {
 	.core = &cx23888_ir_core_ops,
+	.ir = &cx23888_ir_ir_ops,
+};
+
+static const struct v4l2_subdev_ir_parameters default_rx_params = {
+	.bytes_per_data_element = sizeof(u32),
+	.mode = V4L2_SUBDEV_IR_MODE_PULSE_WIDTH,
+
+	.enable = false,
+	.interrupt_enable = false,
+	.shutdown = true,
+
+	.modulation = true,
+	.carrier_freq = 36000, /* 36 kHz - RC-5, RC-6, and RC-6A carrier */
+
+	/* RC-5:    666,667 ns = 1/36 kHz * 32 cycles * 1 mark * 0.75 */
+	/* RC-6A:   333,333 ns = 1/36 kHz * 16 cycles * 1 mark * 0.75 */
+	.noise_filter_min_width = 333333, /* ns */
+	.carrier_range_lower = 35000,
+	.carrier_range_upper = 37000,
+	.invert = false,
+};
+
+static const struct v4l2_subdev_ir_parameters default_tx_params = {
+	.bytes_per_data_element = sizeof(u32),
+	.mode = V4L2_SUBDEV_IR_MODE_PULSE_WIDTH,
+
+	.enable = false,
+	.interrupt_enable = false,
+	.shutdown = true,
+
+	.modulation = true,
+	.carrier_freq = 36000, /* 36 kHz - RC-5 carrier */
+	.duty_cycle = 25,      /* 25 %   - RC-5 carrier */
+	.invert = false,
 };
 
 int cx23888_ir_probe(struct cx23885_dev *dev)
 {
 	struct cx23888_ir_state *state;
 	struct v4l2_subdev *sd;
+	struct v4l2_subdev_ir_parameters default_params;
+	int ret;
 
 	state = kzalloc(sizeof(struct cx23888_ir_state), GFP_KERNEL);
 	if (state == NULL)
 		return -ENOMEM;
 
+	spin_lock_init(&state->rx_kfifo_lock);
+	state->rx_kfifo = kfifo_alloc(CX23888_IR_RX_KFIFO_SIZE, GFP_KERNEL,
+				      &state->rx_kfifo_lock);
+	if (state->rx_kfifo == NULL)
+		return -ENOMEM;
+
+	spin_lock_init(&state->tx_kfifo_lock);
+	state->tx_kfifo = kfifo_alloc(CX23888_IR_TX_KFIFO_SIZE, GFP_KERNEL,
+				      &state->tx_kfifo_lock);
+	if (state->tx_kfifo == NULL) {
+		kfifo_free(state->rx_kfifo);
+		return -ENOMEM;
+	}
+
 	state->dev = dev;
 	state->id = V4L2_IDENT_CX23888_IR;
 	state->rev = 0;
@@ -210,7 +1256,30 @@
 	/* FIXME - fix the formatting of dev->v4l2_dev.name and use it */
 	snprintf(sd->name, sizeof(sd->name), "%s/888-ir", dev->name);
 	sd->grp_id = CX23885_HW_888_IR;
-	return v4l2_device_register_subdev(&dev->v4l2_dev, sd);
+
+	ret = v4l2_device_register_subdev(&dev->v4l2_dev, sd);
+	if (ret == 0) {
+		/*
+		 * Ensure no interrupts arrive from '888 specific conditions,
+		 * since we ignore them in this driver to have commonality with
+		 * similar IR controller cores.
+		 */
+		cx23888_ir_write4(dev, CX23888_IR_IRQEN_REG, 0);
+
+		mutex_init(&state->rx_params_lock);
+		memcpy(&default_params, &default_rx_params,
+		       sizeof(struct v4l2_subdev_ir_parameters));
+		v4l2_subdev_call(sd, ir, rx_s_parameters, &default_params);
+
+		mutex_init(&state->tx_params_lock);
+		memcpy(&default_params, &default_tx_params,
+		       sizeof(struct v4l2_subdev_ir_parameters));
+		v4l2_subdev_call(sd, ir, tx_s_parameters, &default_params);
+	} else {
+		kfifo_free(state->rx_kfifo);
+		kfifo_free(state->tx_kfifo);
+	}
+	return ret;
 }
 
 int cx23888_ir_remove(struct cx23885_dev *dev)
@@ -222,11 +1291,13 @@
 	if (sd == NULL)
 		return -ENODEV;
 
-	/* Disable receiver and transmitter */
-	cx23888_ir_and_or(dev, CX23888_IR_CNTRL_REG+1, 0xfc, 0);
+	cx23888_ir_rx_shutdown(sd);
+	cx23888_ir_tx_shutdown(sd);
 
 	state = to_state(sd);
 	v4l2_device_unregister_subdev(sd);
+	kfifo_free(state->rx_kfifo);
+	kfifo_free(state->tx_kfifo);
 	kfree(state);
 	/* Nothing more to free() as state held the actual v4l2_subdev object */
 	return 0;

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ