linux/linux-5.4.31/drivers/media/pci/cx88/cx88-core.c

// SPDX-License-Identifier: GPL-2.0-or-later
/*
 * device driver for Conexant 2388x based TV cards
 * driver core
 *
 * (c) 2003 Gerd Knorr <kraxel@bytesex.org> [SuSE Labs]
 *
 * (c) 2005-2006 Mauro Carvalho Chehab <mchehab@kernel.org>
 *     - Multituner support
 *     - video_ioctl2 conversion
 *     - PAL/M fixes
 */

#include "cx88.h"

#include <linux/init.h>
#include <linux/list.h>
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/slab.h>
#include <linux/kmod.h>
#include <linux/sound.h>
#include <linux/interrupt.h>
#include <linux/pci.h>
#include <linux/delay.h>
#include <linux/videodev2.h>
#include <linux/mutex.h>

#include <media/v4l2-common.h>
#include <media/v4l2-ioctl.h>

MODULE_DESCRIPTION("v4l2 driver module for cx2388x based TV cards");
MODULE_AUTHOR("Gerd Knorr <kraxel@bytesex.org> [SuSE Labs]");
MODULE_LICENSE("GPL v2");

/* ------------------------------------------------------------------ */

unsigned int cx88_core_debug;
module_param_named(core_debug, cx88_core_debug, int, 0644);
MODULE_PARM_DESC(core_debug, "enable debug messages [core]");

static unsigned int nicam;
module_param(nicam, int, 0644);
MODULE_PARM_DESC(nicam, "tv audio is nicam");

static unsigned int nocomb;
module_param(nocomb, int, 0644);
MODULE_PARM_DESC(nocomb, "disable comb filter");

#define dprintk0(fmt, arg...)				\
	printk(KERN_DEBUG pr_fmt("%s: core:" fmt),	\
		__func__, ##arg)			\

#define dprintk(level, fmt, arg...)	do {			\
	if (cx88_core_debug >= level)				\
		printk(KERN_DEBUG pr_fmt("%s: core:" fmt),	\
		       __func__, ##arg);			\
} while (0)

static unsigned int cx88_devcount;
static LIST_HEAD(cx88_devlist);
static DEFINE_MUTEX(devlist);

#define NO_SYNC_LINE (-1U)

/*
 * @lpi: lines per IRQ, or 0 to not generate irqs. Note: IRQ to be
 * generated _after_ lpi lines are transferred.
 */
static __le32 *cx88_risc_field(__le32 *rp, struct scatterlist *sglist,
			       unsigned int offset, u32 sync_line,
			       unsigned int bpl, unsigned int padding,
			       unsigned int lines, unsigned int lpi, bool jump)
{
	struct scatterlist *sg;
	unsigned int line, todo, sol;

	if (jump) {
		(*rp++) = cpu_to_le32(RISC_JUMP);
		(*rp++) = 0;
	}

	/* sync instruction */
	if (sync_line != NO_SYNC_LINE)
		*(rp++) = cpu_to_le32(RISC_RESYNC | sync_line);

	/* scan lines */
	sg = sglist;
	for (line = 0; line < lines; line++) {
		while (offset && offset >= sg_dma_len(sg)) {
			offset -= sg_dma_len(sg);
			sg = sg_next(sg);
		}
		if (lpi && line > 0 && !(line % lpi))
			sol = RISC_SOL | RISC_IRQ1 | RISC_CNT_INC;
		else
			sol = RISC_SOL;
		if (bpl <= sg_dma_len(sg) - offset) {
			/* fits into current chunk */
			*(rp++) = cpu_to_le32(RISC_WRITE | sol |
					      RISC_EOL | bpl);
			*(rp++) = cpu_to_le32(sg_dma_address(sg) + offset);
			offset += bpl;
		} else {
			/* scanline needs to be split */
			todo = bpl;
			*(rp++) = cpu_to_le32(RISC_WRITE | sol |
					      (sg_dma_len(sg) - offset));
			*(rp++) = cpu_to_le32(sg_dma_address(sg) + offset);
			todo -= (sg_dma_len(sg) - offset);
			offset = 0;
			sg = sg_next(sg);
			while (todo > sg_dma_len(sg)) {
				*(rp++) = cpu_to_le32(RISC_WRITE |
						      sg_dma_len(sg));
				*(rp++) = cpu_to_le32(sg_dma_address(sg));
				todo -= sg_dma_len(sg);
				sg = sg_next(sg);
			}
			*(rp++) = cpu_to_le32(RISC_WRITE | RISC_EOL | todo);
			*(rp++) = cpu_to_le32(sg_dma_address(sg));
			offset += todo;
		}
		offset += padding;
	}

	return rp;
}

int cx88_risc_buffer(struct pci_dev *pci, struct cx88_riscmem *risc,
		     struct scatterlist *sglist,
		     unsigned int top_offset, unsigned int bottom_offset,
		     unsigned int bpl, unsigned int padding, unsigned int lines)
{
	u32 instructions, fields;
	__le32 *rp;

	fields = 0;
	if (top_offset != UNSET)
		fields++;
	if (bottom_offset != UNSET)
		fields++;

	/*
	 * estimate risc mem: worst case is one write per page border +
	 * one write per scan line + syncs + jump (all 2 dwords).  Padding
	 * can cause next bpl to start close to a page border.  First DMA
	 * region may be smaller than PAGE_SIZE
	 */
	instructions  = fields * (1 + ((bpl + padding) * lines) /
				  PAGE_SIZE + lines);
	instructions += 4;
	risc->size = instructions * 8;
	risc->dma = 0;
	risc->cpu = pci_zalloc_consistent(pci, risc->size, &risc->dma);
	if (!risc->cpu)
		return -ENOMEM;

	/* write risc instructions */
	rp = risc->cpu;
	if (top_offset != UNSET)
		rp = cx88_risc_field(rp, sglist, top_offset, 0,
				     bpl, padding, lines, 0, true);
	if (bottom_offset != UNSET)
		rp = cx88_risc_field(rp, sglist, bottom_offset, 0x200,
				     bpl, padding, lines, 0,
				     top_offset == UNSET);

	/* save pointer to jmp instruction address */
	risc->jmp = rp;
	WARN_ON((risc->jmp - risc->cpu + 2) * sizeof(*risc->cpu) > risc->size);
	return 0;
}
EXPORT_SYMBOL(cx88_risc_buffer);

int cx88_risc_databuffer(struct pci_dev *pci, struct cx88_riscmem *risc,
			 struct scatterlist *sglist, unsigned int bpl,
			 unsigned int lines, unsigned int lpi)
{
	u32 instructions;
	__le32 *rp;

	/*
	 * estimate risc mem: worst case is one write per page border +
	 * one write per scan line + syncs + jump (all 2 dwords).  Here
	 * there is no padding and no sync.  First DMA region may be smaller
	 * than PAGE_SIZE
	 */
	instructions  = 1 + (bpl * lines) / PAGE_SIZE + lines;
	instructions += 3;
	risc->size = instructions * 8;
	risc->dma = 0;
	risc->cpu = pci_zalloc_consistent(pci, risc->size, &risc->dma);
	if (!risc->cpu)
		return -ENOMEM;

	/* write risc instructions */
	rp = risc->cpu;
	rp = cx88_risc_field(rp, sglist, 0, NO_SYNC_LINE, bpl, 0,
			     lines, lpi, !lpi);

	/* save pointer to jmp instruction address */
	risc->jmp = rp;
	WARN_ON((risc->jmp - risc->cpu + 2) * sizeof(*risc->cpu) > risc->size);
	return 0;
}
EXPORT_SYMBOL(cx88_risc_databuffer);

/*
 * our SRAM memory layout
 */

/*
 * we are going to put all thr risc programs into host memory, so we
 * can use the whole SDRAM for the DMA fifos.  To simplify things, we
 * use a static memory layout.  That surely will waste memory in case
 * we don't use all DMA channels at the same time (which will be the
 * case most of the time).  But that still gives us enough FIFO space
 * to be able to deal with insane long pci latencies ...
 *
 * FIFO space allocations:
 *    channel  21    (y video)  - 10.0k
 *    channel  22    (u video)  -  2.0k
 *    channel  23    (v video)  -  2.0k
 *    channel  24    (vbi)      -  4.0k
 *    channels 25+26 (audio)    -  4.0k
 *    channel  28    (mpeg)     -  4.0k
 *    channel  27    (audio rds)-  3.0k
 *    TOTAL                     = 29.0k
 *
 * Every channel has 160 bytes control data (64 bytes instruction
 * queue and 6 CDT entries), which is close to 2k total.
 *
 * Address layout:
 *    0x0000 - 0x03ff    CMDs / reserved
 *    0x0400 - 0x0bff    instruction queues + CDs
 *    0x0c00 -           FIFOs
 */

const struct sram_channel cx88_sram_channels[] = {
	[SRAM_CH21] = {
		.name       = "video y / packed",
		.cmds_start = 0x180040,
		.ctrl_start = 0x180400,
		.cdt        = 0x180400 + 64,
		.fifo_start = 0x180c00,
		.fifo_size  = 0x002800,
		.ptr1_reg   = MO_DMA21_PTR1,
		.ptr2_reg   = MO_DMA21_PTR2,
		.cnt1_reg   = MO_DMA21_CNT1,
		.cnt2_reg   = MO_DMA21_CNT2,
	},
	[SRAM_CH22] = {
		.name       = "video u",
		.cmds_start = 0x180080,
		.ctrl_start = 0x1804a0,
		.cdt        = 0x1804a0 + 64,
		.fifo_start = 0x183400,
		.fifo_size  = 0x000800,
		.ptr1_reg   = MO_DMA22_PTR1,
		.ptr2_reg   = MO_DMA22_PTR2,
		.cnt1_reg   = MO_DMA22_CNT1,
		.cnt2_reg   = MO_DMA22_CNT2,
	},
	[SRAM_CH23] = {
		.name       = "video v",
		.cmds_start = 0x1800c0,
		.ctrl_start = 0x180540,
		.cdt        = 0x180540 + 64,
		.fifo_start = 0x183c00,
		.fifo_size  = 0x000800,
		.ptr1_reg   = MO_DMA23_PTR1,
		.ptr2_reg   = MO_DMA23_PTR2,
		.cnt1_reg   = MO_DMA23_CNT1,
		.cnt2_reg   = MO_DMA23_CNT2,
	},
	[SRAM_CH24] = {
		.name       = "vbi",
		.cmds_start = 0x180100,
		.ctrl_start = 0x1805e0,
		.cdt        = 0x1805e0 + 64,
		.fifo_start = 0x184400,
		.fifo_size  = 0x001000,
		.ptr1_reg   = MO_DMA24_PTR1,
		.ptr2_reg   = MO_DMA24_PTR2,
		.cnt1_reg   = MO_DMA24_CNT1,
		.cnt2_reg   = MO_DMA24_CNT2,
	},
	[SRAM_CH25] = {
		.name       = "audio from",
		.cmds_start = 0x180140,
		.ctrl_start = 0x180680,
		.cdt        = 0x180680 + 64,
		.fifo_start = 0x185400,
		.fifo_size  = 0x001000,
		.ptr1_reg   = MO_DMA25_PTR1,
		.ptr2_reg   = MO_DMA25_PTR2,
		.cnt1_reg   = MO_DMA25_CNT1,
		.cnt2_reg   = MO_DMA25_CNT2,
	},
	[SRAM_CH26] = {
		.name       = "audio to",
		.cmds_start = 0x180180,
		.ctrl_start = 0x180720,
		.cdt        = 0x180680 + 64,  /* same as audio IN */
		.fifo_start = 0x185400,       /* same as audio IN */
		.fifo_size  = 0x001000,       /* same as audio IN */
		.ptr1_reg   = MO_DMA26_PTR1,
		.ptr2_reg   = MO_DMA26_PTR2,
		.cnt1_reg   = MO_DMA26_CNT1,
		.cnt2_reg   = MO_DMA26_CNT2,
	},
	[SRAM_CH28] = {
		.name       = "mpeg",
		.cmds_start = 0x180200,
		.ctrl_start = 0x1807C0,
		.cdt        = 0x1807C0 + 64,
		.fifo_start = 0x186400,
		.fifo_size  = 0x001000,
		.ptr1_reg   = MO_DMA28_PTR1,
		.ptr2_reg   = MO_DMA28_PTR2,
		.cnt1_reg   = MO_DMA28_CNT1,
		.cnt2_reg   = MO_DMA28_CNT2,
	},
	[SRAM_CH27] = {
		.name       = "audio rds",
		.cmds_start = 0x1801C0,
		.ctrl_start = 0x180860,
		.cdt        = 0x180860 + 64,
		.fifo_start = 0x187400,
		.fifo_size  = 0x000C00,
		.ptr1_reg   = MO_DMA27_PTR1,
		.ptr2_reg   = MO_DMA27_PTR2,
		.cnt1_reg   = MO_DMA27_CNT1,
		.cnt2_reg   = MO_DMA27_CNT2,
	},
};
EXPORT_SYMBOL(cx88_sram_channels);

int cx88_sram_channel_setup(struct cx88_core *core,
			    const struct sram_channel *ch,
			    unsigned int bpl, u32 risc)
{
	unsigned int i, lines;
	u32 cdt;

	bpl   = (bpl + 7) & ~7; /* alignment */
	cdt   = ch->cdt;
	lines = ch->fifo_size / bpl;
	if (lines > 6)
		lines = 6;
	WARN_ON(lines < 2);

	/* write CDT */
	for (i = 0; i < lines; i++)
		cx_write(cdt + 16 * i, ch->fifo_start + bpl * i);

	/* write CMDS */
	cx_write(ch->cmds_start +  0, risc);
	cx_write(ch->cmds_start +  4, cdt);
	cx_write(ch->cmds_start +  8, (lines * 16) >> 3);
	cx_write(ch->cmds_start + 12, ch->ctrl_start);
	cx_write(ch->cmds_start + 16, 64 >> 2);
	for (i = 20; i < 64; i += 4)
		cx_write(ch->cmds_start + i, 0);

	/* fill registers */
	cx_write(ch->ptr1_reg, ch->fifo_start);
	cx_write(ch->ptr2_reg, cdt);
	cx_write(ch->cnt1_reg, (bpl >> 3) - 1);
	cx_write(ch->cnt2_reg, (lines * 16) >> 3);

	dprintk(2, "sram setup %s: bpl=%d lines=%d\n", ch->name, bpl, lines);
	return 0;
}
EXPORT_SYMBOL(cx88_sram_channel_setup);

/* ------------------------------------------------------------------ */
/* debug helper code                                                  */

static int cx88_risc_decode(u32 risc)
{
	static const char * const instr[16] = {
		[RISC_SYNC    >> 28] = "sync",
		[RISC_WRITE   >> 28] = "write",
		[RISC_WRITEC  >> 28] = "writec",
		[RISC_READ    >> 28] = "read",
		[RISC_READC   >> 28] = "readc",
		[RISC_JUMP    >> 28] = "jump",
		[RISC_SKIP    >> 28] = "skip",
		[RISC_WRITERM >> 28] = "writerm",
		[RISC_WRITECM >> 28] = "writecm",
		[RISC_WRITECR >> 28] = "writecr",
	};
	static int const incr[16] = {
		[RISC_WRITE   >> 28] = 2,
		[RISC_JUMP    >> 28] = 2,
		[RISC_WRITERM >> 28] = 3,
		[RISC_WRITECM >> 28] = 3,
		[RISC_WRITECR >> 28] = 4,
	};
	static const char * const bits[] = {
		"12",   "13",   "14",   "resync",
		"cnt0", "cnt1", "18",   "19",
		"20",   "21",   "22",   "23",
		"irq1", "irq2", "eol",  "sol",
	};
	int i;

	dprintk0("0x%08x [ %s", risc,
		 instr[risc >> 28] ? instr[risc >> 28] : "INVALID");
	for (i = ARRAY_SIZE(bits) - 1; i >= 0; i--)
		if (risc & (1 << (i + 12)))
			pr_cont(" %s", bits[i]);
	pr_cont(" count=%d ]\n", risc & 0xfff);
	return incr[risc >> 28] ? incr[risc >> 28] : 1;
}

void cx88_sram_channel_dump(struct cx88_core *core,
			    const struct sram_channel *ch)
{
	static const char * const name[] = {
		"initial risc",
		"cdt base",
		"cdt size",
		"iq base",
		"iq size",
		"risc pc",
		"iq wr ptr",
		"iq rd ptr",
		"cdt current",
		"pci target",
		"line / byte",
	};
	u32 risc;
	unsigned int i, j, n;

	dprintk0("%s - dma channel status dump\n", ch->name);
	for (i = 0; i < ARRAY_SIZE(name); i++)
		dprintk0("   cmds: %-12s: 0x%08x\n",
			 name[i], cx_read(ch->cmds_start + 4 * i));
	for (n = 1, i = 0; i < 4; i++) {
		risc = cx_read(ch->cmds_start + 4 * (i + 11));
		pr_cont("  risc%d: ", i);
		if (--n)
			pr_cont("0x%08x [ arg #%d ]\n", risc, n);
		else
			n = cx88_risc_decode(risc);
	}
	for (i = 0; i < 16; i += n) {
		risc = cx_read(ch->ctrl_start + 4 * i);
		dprintk0("  iq %x: ", i);
		n = cx88_risc_decode(risc);
		for (j = 1; j < n; j++) {
			risc = cx_read(ch->ctrl_start + 4 * (i + j));
			pr_cont("  iq %x: 0x%08x [ arg #%d ]\n",
				i + j, risc, j);
		}
	}

	dprintk0("fifo: 0x%08x -> 0x%x\n",
		 ch->fifo_start, ch->fifo_start + ch->fifo_size);
	dprintk0("ctrl: 0x%08x -> 0x%x\n",
		 ch->ctrl_start, ch->ctrl_start + 6 * 16);
	dprintk0("  ptr1_reg: 0x%08x\n", cx_read(ch->ptr1_reg));
	dprintk0("  ptr2_reg: 0x%08x\n", cx_read(ch->ptr2_reg));
	dprintk0("  cnt1_reg: 0x%08x\n", cx_read(ch->cnt1_reg));
	dprintk0("  cnt2_reg: 0x%08x\n", cx_read(ch->cnt2_reg));
}
EXPORT_SYMBOL(cx88_sram_channel_dump);

static const char *cx88_pci_irqs[32] = {
	"vid", "aud", "ts", "vip", "hst", "5", "6", "tm1",
	"src_dma", "dst_dma", "risc_rd_err", "risc_wr_err",
	"brdg_err", "src_dma_err", "dst_dma_err", "ipb_dma_err",
	"i2c", "i2c_rack", "ir_smp", "gpio0", "gpio1"
};

void cx88_print_irqbits(const char *tag, const char *strings[],
			int len, u32 bits, u32 mask)
{
	unsigned int i;

	dprintk0("%s [0x%x]", tag, bits);
	for (i = 0; i < len; i++) {
		if (!(bits & (1 << i)))
			continue;
		if (strings[i])
			pr_cont(" %s", strings[i]);
		else
			pr_cont(" %d", i);
		if (!(mask & (1 << i)))
			continue;
		pr_cont("*");
	}
	pr_cont("\n");
}
EXPORT_SYMBOL(cx88_print_irqbits);

/* ------------------------------------------------------------------ */

int cx88_core_irq(struct cx88_core *core, u32 status)
{
	int handled = 0;

	if (status & PCI_INT_IR_SMPINT) {
		cx88_ir_irq(core);
		handled++;
	}
	if (!handled)
		cx88_print_irqbits("irq pci",
				   cx88_pci_irqs, ARRAY_SIZE(cx88_pci_irqs),
				   status, core->pci_irqmask);
	return handled;
}
EXPORT_SYMBOL(cx88_core_irq);

void cx88_wakeup(struct cx88_core *core,
		 struct cx88_dmaqueue *q, u32 count)
{
	struct cx88_buffer *buf;

	buf = list_entry(q->active.next,
			 struct cx88_buffer, list);
	buf->vb.vb2_buf.timestamp = ktime_get_ns();
	buf->vb.field = core->field;
	buf->vb.sequence = q->count++;
	list_del(&buf->list);
	vb2_buffer_done(&buf->vb.vb2_buf, VB2_BUF_STATE_DONE);
}
EXPORT_SYMBOL(cx88_wakeup);

void cx88_shutdown(struct cx88_core *core)
{
	/* disable RISC controller + IRQs */
	cx_write(MO_DEV_CNTRL2, 0);

	/* stop dma transfers */
	cx_write(MO_VID_DMACNTRL, 0x0);
	cx_write(MO_AUD_DMACNTRL, 0x0);
	cx_write(MO_TS_DMACNTRL, 0x0);
	cx_write(MO_VIP_DMACNTRL, 0x0);
	cx_write(MO_GPHST_DMACNTRL, 0x0);

	/* stop interrupts */
	cx_write(MO_PCI_INTMSK, 0x0);
	cx_write(MO_VID_INTMSK, 0x0);
	cx_write(MO_AUD_INTMSK, 0x0);
	cx_write(MO_TS_INTMSK, 0x0);
	cx_write(MO_VIP_INTMSK, 0x0);
	cx_write(MO_GPHST_INTMSK, 0x0);

	/* stop capturing */
	cx_write(VID_CAPTURE_CONTROL, 0);
}
EXPORT_SYMBOL(cx88_shutdown);

int cx88_reset(struct cx88_core *core)
{
	dprintk(1, "");
	cx88_shutdown(core);

	/* clear irq status */
	cx_write(MO_VID_INTSTAT, 0xFFFFFFFF); // Clear PIV int
	cx_write(MO_PCI_INTSTAT, 0xFFFFFFFF); // Clear PCI int
	cx_write(MO_INT1_STAT,   0xFFFFFFFF); // Clear RISC int

	/* wait a bit */
	msleep(100);

	/* init sram */
	cx88_sram_channel_setup(core, &cx88_sram_channels[SRAM_CH21],
				720 * 4, 0);
	cx88_sram_channel_setup(core, &cx88_sram_channels[SRAM_CH22], 128, 0);
	cx88_sram_channel_setup(core, &cx88_sram_channels[SRAM_CH23], 128, 0);
	cx88_sram_channel_setup(core, &cx88_sram_channels[SRAM_CH24], 128, 0);
	cx88_sram_channel_setup(core, &cx88_sram_channels[SRAM_CH25], 128, 0);
	cx88_sram_channel_setup(core, &cx88_sram_channels[SRAM_CH26], 128, 0);
	cx88_sram_channel_setup(core, &cx88_sram_channels[SRAM_CH28],
				188 * 4, 0);
	cx88_sram_channel_setup(core, &cx88_sram_channels[SRAM_CH27], 128, 0);

	/* misc init ... */
	cx_write(MO_INPUT_FORMAT, ((1 << 13) |   // agc enable
				   (1 << 12) |   // agc gain
				   (1 << 11) |   // adaptibe agc
				   (0 << 10) |   // chroma agc
				   (0 <<  9) |   // ckillen
				   (7)));

	/* setup image format */
	cx_andor(MO_COLOR_CTRL, 0x4000, 0x4000);

	/* setup FIFO Thresholds */
	cx_write(MO_PDMA_STHRSH,   0x0807);
	cx_write(MO_PDMA_DTHRSH,   0x0807);

	/* fixes flashing of image */
	cx_write(MO_AGC_SYNC_TIP1, 0x0380000F);
	cx_write(MO_AGC_BACK_VBI,  0x00E00555);

	cx_write(MO_VID_INTSTAT,   0xFFFFFFFF); // Clear PIV int
	cx_write(MO_PCI_INTSTAT,   0xFFFFFFFF); // Clear PCI int
	cx_write(MO_INT1_STAT,     0xFFFFFFFF); // Clear RISC int

	/* Reset on-board parts */
	cx_write(MO_SRST_IO, 0);
	usleep_range(10000, 20000);
	cx_write(MO_SRST_IO, 1);

	return 0;
}
EXPORT_SYMBOL(cx88_reset);

/* ------------------------------------------------------------------ */

static inline unsigned int norm_swidth(v4l2_std_id norm)
{
	return (norm & (V4L2_STD_MN & ~V4L2_STD_PAL_Nc)) ? 754 : 922;
}

static inline unsigned int norm_hdelay(v4l2_std_id norm)
{
	return (norm & (V4L2_STD_MN & ~V4L2_STD_PAL_Nc)) ? 135 : 186;
}

static inline unsigned int norm_vdelay(v4l2_std_id norm)
{
	return (norm & V4L2_STD_625_50) ? 0x24 : 0x18;
}

static inline unsigned int norm_fsc8(v4l2_std_id norm)
{
	if (norm & V4L2_STD_PAL_M)
		return 28604892;      // 3.575611 MHz

	if (norm & (V4L2_STD_PAL_Nc))
		return 28656448;      // 3.582056 MHz

	if (norm & V4L2_STD_NTSC) // All NTSC/M and variants
		return 28636360;      // 3.57954545 MHz +/- 10 Hz

	/*
	 * SECAM have also different sub carrier for chroma,
	 * but step_db and step_dr, at cx88_set_tvnorm already handles that.
	 *
	 * The same FSC applies to PAL/BGDKIH, PAL/60, NTSC/4.43 and PAL/N
	 */

	return 35468950;      // 4.43361875 MHz +/- 5 Hz
}

static inline unsigned int norm_htotal(v4l2_std_id norm)
{
	unsigned int fsc4 = norm_fsc8(norm) / 2;

	/* returns 4*FSC / vtotal / frames per seconds */
	return (norm & V4L2_STD_625_50) ?
				((fsc4 + 312) / 625 + 12) / 25 :
				((fsc4 + 262) / 525 * 1001 + 15000) / 30000;
}

static inline unsigned int norm_vbipack(v4l2_std_id norm)
{
	return (norm & V4L2_STD_625_50) ? 511 : 400;
}

int cx88_set_scale(struct cx88_core *core, unsigned int width,
		   unsigned int height, enum v4l2_field field)
{
	unsigned int swidth  = norm_swidth(core->tvnorm);
	unsigned int sheight = norm_maxh(core->tvnorm);
	u32 value;

	dprintk(1, "set_scale: %dx%d [%s%s,%s]\n", width, height,
		V4L2_FIELD_HAS_TOP(field)    ? "T" : "",
		V4L2_FIELD_HAS_BOTTOM(field) ? "B" : "",
		v4l2_norm_to_name(core->tvnorm));
	if (!V4L2_FIELD_HAS_BOTH(field))
		height *= 2;

	// recalc H delay and scale registers
	value = (width * norm_hdelay(core->tvnorm)) / swidth;
	value &= 0x3fe;
	cx_write(MO_HDELAY_EVEN,  value);
	cx_write(MO_HDELAY_ODD,   value);
	dprintk(1, "set_scale: hdelay  0x%04x (width %d)\n", value, swidth);

	value = (swidth * 4096 / width) - 4096;
	cx_write(MO_HSCALE_EVEN,  value);
	cx_write(MO_HSCALE_ODD,   value);
	dprintk(1, "set_scale: hscale  0x%04x\n", value);

	cx_write(MO_HACTIVE_EVEN, width);
	cx_write(MO_HACTIVE_ODD,  width);
	dprintk(1, "set_scale: hactive 0x%04x\n", width);

	// recalc V scale Register (delay is constant)
	cx_write(MO_VDELAY_EVEN, norm_vdelay(core->tvnorm));
	cx_write(MO_VDELAY_ODD,  norm_vdelay(core->tvnorm));
	dprintk(1, "set_scale: vdelay  0x%04x\n", norm_vdelay(core->tvnorm));

	value = (0x10000 - (sheight * 512 / height - 512)) & 0x1fff;
	cx_write(MO_VSCALE_EVEN,  value);
	cx_write(MO_VSCALE_ODD,   value);
	dprintk(1, "set_scale: vscale  0x%04x\n", value);

	cx_write(MO_VACTIVE_EVEN, sheight);
	cx_write(MO_VACTIVE_ODD,  sheight);
	dprintk(1, "set_scale: vactive 0x%04x\n", sheight);

	// setup filters
	value = 0;
	value |= (1 << 19);        // CFILT (default)
	if (core->tvnorm & V4L2_STD_SECAM) {
		value |= (1 << 15);
		value |= (1 << 16);
	}
	if (INPUT(core->input).type == CX88_VMUX_SVIDEO)
		value |= (1 << 13) | (1 << 5);
	if (field == V4L2_FIELD_INTERLACED)
		value |= (1 << 3); // VINT (interlaced vertical scaling)
	if (width < 385)
		value |= (1 << 0); // 3-tap interpolation
	if (width < 193)
		value |= (1 << 1); // 5-tap interpolation
	if (nocomb)
		value |= (3 << 5); // disable comb filter

	cx_andor(MO_FILTER_EVEN,  0x7ffc7f, value); /* preserve PEAKEN, PSEL */
	cx_andor(MO_FILTER_ODD,   0x7ffc7f, value);
	dprintk(1, "set_scale: filter  0x%04x\n", value);

	return 0;
}
EXPORT_SYMBOL(cx88_set_scale);

static const u32 xtal = 28636363;

static int set_pll(struct cx88_core *core, int prescale, u32 ofreq)
{
	static const u32 pre[] = { 0, 0, 0, 3, 2, 1 };
	u64 pll;
	u32 reg;
	int i;

	if (prescale < 2)
		prescale = 2;
	if (prescale > 5)
		prescale = 5;

	pll = ofreq * 8 * prescale * (u64)(1 << 20);
	do_div(pll, xtal);
	reg = (pll & 0x3ffffff) | (pre[prescale] << 26);
	if (((reg >> 20) & 0x3f) < 14) {
		pr_err("pll out of range\n");
		return -1;
	}

	dprintk(1, "set_pll:    MO_PLL_REG       0x%08x [old=0x%08x,freq=%d]\n",
		reg, cx_read(MO_PLL_REG), ofreq);
	cx_write(MO_PLL_REG, reg);
	for (i = 0; i < 100; i++) {
		reg = cx_read(MO_DEVICE_STATUS);
		if (reg & (1 << 2)) {
			dprintk(1, "pll locked [pre=%d,ofreq=%d]\n",
				prescale, ofreq);
			return 0;
		}
		dprintk(1, "pll not locked yet, waiting ...\n");
		usleep_range(10000, 20000);
	}
	dprintk(1, "pll NOT locked [pre=%d,ofreq=%d]\n", prescale, ofreq);
	return -1;
}

int cx88_start_audio_dma(struct cx88_core *core)
{
	/* constant 128 made buzz in analog Nicam-stereo for bigger fifo_size */
	int bpl = cx88_sram_channels[SRAM_CH25].fifo_size / 4;

	int rds_bpl = cx88_sram_channels[SRAM_CH27].fifo_size / AUD_RDS_LINES;

	/* If downstream RISC is enabled, bail out; ALSA is managing DMA */
	if (cx_read(MO_AUD_DMACNTRL) & 0x10)
		return 0;

	/* setup fifo + format */
	cx88_sram_channel_setup(core, &cx88_sram_channels[SRAM_CH25], bpl, 0);
	cx88_sram_channel_setup(core, &cx88_sram_channels[SRAM_CH26], bpl, 0);
	cx88_sram_channel_setup(core, &cx88_sram_channels[SRAM_CH27],
				rds_bpl, 0);

	cx_write(MO_AUDD_LNGTH, bpl); /* fifo bpl size */
	cx_write(MO_AUDR_LNGTH, rds_bpl); /* fifo bpl size */

	/* enable Up, Down and Audio RDS fifo */
	cx_write(MO_AUD_DMACNTRL, 0x0007);

	return 0;
}

int cx88_stop_audio_dma(struct cx88_core *core)
{
	/* If downstream RISC is enabled, bail out; ALSA is managing DMA */
	if (cx_read(MO_AUD_DMACNTRL) & 0x10)
		return 0;

	/* stop dma */
	cx_write(MO_AUD_DMACNTRL, 0x0000);

	return 0;
}

static int set_tvaudio(struct cx88_core *core)
{
	v4l2_std_id norm = core->tvnorm;

	if (INPUT(core->input).type != CX88_VMUX_TELEVISION &&
	    INPUT(core->input).type != CX88_VMUX_CABLE)
		return 0;

	if (V4L2_STD_PAL_BG & norm) {
		core->tvaudio = WW_BG;

	} else if (V4L2_STD_PAL_DK & norm) {
		core->tvaudio = WW_DK;

	} else if (V4L2_STD_PAL_I & norm) {
		core->tvaudio = WW_I;

	} else if (V4L2_STD_SECAM_L & norm) {
		core->tvaudio = WW_L;

	} else if ((V4L2_STD_SECAM_B | V4L2_STD_SECAM_G | V4L2_STD_SECAM_H) &
		   norm) {
		core->tvaudio = WW_BG;

	} else if (V4L2_STD_SECAM_DK & norm) {
		core->tvaudio = WW_DK;

	} else if ((V4L2_STD_NTSC_M & norm) ||
		   (V4L2_STD_PAL_M  & norm)) {
		core->tvaudio = WW_BTSC;

	} else if (V4L2_STD_NTSC_M_JP & norm) {
		core->tvaudio = WW_EIAJ;

	} else {
		pr_info("tvaudio support needs work for this tv norm [%s], sorry\n",
			v4l2_norm_to_name(core->tvnorm));
		core->tvaudio = WW_NONE;
		return 0;
	}

	cx_andor(MO_AFECFG_IO, 0x1f, 0x0);
	cx88_set_tvaudio(core);
	/* cx88_set_stereo(dev,V4L2_TUNER_MODE_STEREO); */

/*
 * This should be needed only on cx88-alsa. It seems that some cx88 chips have
 * bugs and does require DMA enabled for it to work.
 */
	cx88_start_audio_dma(core);
	return 0;
}

int cx88_set_tvnorm(struct cx88_core *core, v4l2_std_id norm)
{
	u32 fsc8;
	u32 adc_clock;
	u32 vdec_clock;
	u32 step_db, step_dr;
	u64 tmp64;
	u32 bdelay, agcdelay, htotal;
	u32 cxiformat, cxoformat;

	if (norm == core->tvnorm)
		return 0;
	if (core->v4ldev && (vb2_is_busy(&core->v4ldev->vb2_vidq) ||
			     vb2_is_busy(&core->v4ldev->vb2_vbiq)))
		return -EBUSY;
	if (core->dvbdev && vb2_is_busy(&core->dvbdev->vb2_mpegq))
		return -EBUSY;
	core->tvnorm = norm;
	fsc8       = norm_fsc8(norm);
	adc_clock  = xtal;
	vdec_clock = fsc8;
	step_db    = fsc8;
	step_dr    = fsc8;

	if (norm & V4L2_STD_NTSC_M_JP) {
		cxiformat = VideoFormatNTSCJapan;
		cxoformat = 0x181f0008;
	} else if (norm & V4L2_STD_NTSC_443) {
		cxiformat = VideoFormatNTSC443;
		cxoformat = 0x181f0008;
	} else if (norm & V4L2_STD_PAL_M) {
		cxiformat = VideoFormatPALM;
		cxoformat = 0x1c1f0008;
	} else if (norm & V4L2_STD_PAL_N) {
		cxiformat = VideoFormatPALN;
		cxoformat = 0x1c1f0008;
	} else if (norm & V4L2_STD_PAL_Nc) {
		cxiformat = VideoFormatPALNC;
		cxoformat = 0x1c1f0008;
	} else if (norm & V4L2_STD_PAL_60) {
		cxiformat = VideoFormatPAL60;
		cxoformat = 0x181f0008;
	} else if (norm & V4L2_STD_NTSC) {
		cxiformat = VideoFormatNTSC;
		cxoformat = 0x181f0008;
	} else if (norm & V4L2_STD_SECAM) {
		step_db = 4250000 * 8;
		step_dr = 4406250 * 8;

		cxiformat = VideoFormatSECAM;
		cxoformat = 0x181f0008;
	} else { /* PAL */
		cxiformat = VideoFormatPAL;
		cxoformat = 0x181f0008;
	}

	dprintk(1, "set_tvnorm: \"%s\" fsc8=%d adc=%d vdec=%d db/dr=%d/%d\n",
		v4l2_norm_to_name(core->tvnorm), fsc8, adc_clock, vdec_clock,
		step_db, step_dr);
	set_pll(core, 2, vdec_clock);

	dprintk(1, "set_tvnorm: MO_INPUT_FORMAT  0x%08x [old=0x%08x]\n",
		cxiformat, cx_read(MO_INPUT_FORMAT) & 0x0f);
	/*
	 * Chroma AGC must be disabled if SECAM is used, we enable it
	 * by default on PAL and NTSC
	 */
	cx_andor(MO_INPUT_FORMAT, 0x40f,
		 norm & V4L2_STD_SECAM ? cxiformat : cxiformat | 0x400);

	// FIXME: as-is from DScaler
	dprintk(1, "set_tvnorm: MO_OUTPUT_FORMAT 0x%08x [old=0x%08x]\n",
		cxoformat, cx_read(MO_OUTPUT_FORMAT));
	cx_write(MO_OUTPUT_FORMAT, cxoformat);

	// MO_SCONV_REG = adc clock / video dec clock * 2^17
	tmp64  = adc_clock * (u64)(1 << 17);
	do_div(tmp64, vdec_clock);
	dprintk(1, "set_tvnorm: MO_SCONV_REG     0x%08x [old=0x%08x]\n",
		(u32)tmp64, cx_read(MO_SCONV_REG));
	cx_write(MO_SCONV_REG, (u32)tmp64);

	// MO_SUB_STEP = 8 * fsc / video dec clock * 2^22
	tmp64  = step_db * (u64)(1 << 22);
	do_div(tmp64, vdec_clock);
	dprintk(1, "set_tvnorm: MO_SUB_STEP      0x%08x [old=0x%08x]\n",
		(u32)tmp64, cx_read(MO_SUB_STEP));
	cx_write(MO_SUB_STEP, (u32)tmp64);

	// MO_SUB_STEP_DR = 8 * 4406250 / video dec clock * 2^22
	tmp64  = step_dr * (u64)(1 << 22);
	do_div(tmp64, vdec_clock);
	dprintk(1, "set_tvnorm: MO_SUB_STEP_DR   0x%08x [old=0x%08x]\n",
		(u32)tmp64, cx_read(MO_SUB_STEP_DR));
	cx_write(MO_SUB_STEP_DR, (u32)tmp64);

	// bdelay + agcdelay
	bdelay   = vdec_clock * 65 / 20000000 + 21;
	agcdelay = vdec_clock * 68 / 20000000 + 15;
	dprintk(1,
		"set_tvnorm: MO_AGC_BURST     0x%08x [old=0x%08x,bdelay=%d,agcdelay=%d]\n",
		(bdelay << 8) | agcdelay, cx_read(MO_AGC_BURST),
		bdelay, agcdelay);
	cx_write(MO_AGC_BURST, (bdelay << 8) | agcdelay);

	// htotal
	tmp64 = norm_htotal(norm) * (u64)vdec_clock;
	do_div(tmp64, fsc8);
	htotal = (u32)tmp64;
	dprintk(1,
		"set_tvnorm: MO_HTOTAL        0x%08x [old=0x%08x,htotal=%d]\n",
		htotal, cx_read(MO_HTOTAL), (u32)tmp64);
	cx_andor(MO_HTOTAL, 0x07ff, htotal);

	// vbi stuff, set vbi offset to 10 (for 20 Clk*2 pixels), this makes
	// the effective vbi offset ~244 samples, the same as the Bt8x8
	cx_write(MO_VBI_PACKET, (10 << 11) | norm_vbipack(norm));

	// this is needed as well to set all tvnorm parameter
	cx88_set_scale(core, 320, 240, V4L2_FIELD_INTERLACED);

	// audio
	set_tvaudio(core);

	// tell i2c chips
	call_all(core, video, s_std, norm);

	/*
	 * The chroma_agc control should be inaccessible
	 * if the video format is SECAM
	 */
	v4l2_ctrl_grab(core->chroma_agc, cxiformat == VideoFormatSECAM);

	// done
	return 0;
}
EXPORT_SYMBOL(cx88_set_tvnorm);

/* ------------------------------------------------------------------ */

void cx88_vdev_init(struct cx88_core *core,
		    struct pci_dev *pci,
		    struct video_device *vfd,
		    const struct video_device *template_,
		    const char *type)
{
	*vfd = *template_;

	/*
	 * The dev pointer of v4l2_device is NULL, instead we set the
	 * video_device dev_parent pointer to the correct PCI bus device.
	 * This driver is a rare example where there is one v4l2_device,
	 * but the video nodes have different parent (PCI) devices.
	 */
	vfd->v4l2_dev = &core->v4l2_dev;
	vfd->dev_parent = &pci->dev;
	vfd->release = video_device_release_empty;
	vfd->lock = &core->lock;
	snprintf(vfd->name, sizeof(vfd->name), "%s %s (%s)",
		 core->name, type, core->board.name);
}
EXPORT_SYMBOL(cx88_vdev_init);

struct cx88_core *cx88_core_get(struct pci_dev *pci)
{
	struct cx88_core *core;

	mutex_lock(&devlist);
	list_for_each_entry(core, &cx88_devlist, devlist) {
		if (pci->bus->number != core->pci_bus)
			continue;
		if (PCI_SLOT(pci->devfn) != core->pci_slot)
			continue;

		if (cx88_get_resources(core, pci) != 0) {
			mutex_unlock(&devlist);
			return NULL;
		}
		refcount_inc(&core->refcount);
		mutex_unlock(&devlist);
		return core;
	}

	core = cx88_core_create(pci, cx88_devcount);
	if (core) {
		cx88_devcount++;
		list_add_tail(&core->devlist, &cx88_devlist);
	}

	mutex_unlock(&devlist);
	return core;
}
EXPORT_SYMBOL(cx88_core_get);

void cx88_core_put(struct cx88_core *core, struct pci_dev *pci)
{
	release_mem_region(pci_resource_start(pci, 0),
			   pci_resource_len(pci, 0));

	if (!refcount_dec_and_test(&core->refcount))
		return;

	mutex_lock(&devlist);
	cx88_ir_fini(core);
	if (core->i2c_rc == 0) {
		if (core->i2c_rtc)
			i2c_unregister_device(core->i2c_rtc);
		i2c_del_adapter(&core->i2c_adap);
	}
	list_del(&core->devlist);
	iounmap(core->lmmio);
	cx88_devcount--;
	mutex_unlock(&devlist);
	v4l2_ctrl_handler_free(&core->video_hdl);
	v4l2_ctrl_handler_free(&core->audio_hdl);
	v4l2_device_unregister(&core->v4l2_dev);
	kfree(core);
}
EXPORT_SYMBOL(cx88_core_put);