xref: /openbmc/linux/drivers/media/pci/tw68/tw68-risc.c (revision d3964221)
1 /*
2  *  tw68_risc.c
3  *  Part of the device driver for Techwell 68xx based cards
4  *
5  *  Much of this code is derived from the cx88 and sa7134 drivers, which
6  *  were in turn derived from the bt87x driver.  The original work was by
7  *  Gerd Knorr; more recently the code was enhanced by Mauro Carvalho Chehab,
8  *  Hans Verkuil, Andy Walls and many others.  Their work is gratefully
9  *  acknowledged.  Full credit goes to them - any problems within this code
10  *  are mine.
11  *
12  *  Copyright (C) 2009  William M. Brack
13  *
14  *  Refactored and updated to the latest v4l core frameworks:
15  *
16  *  Copyright (C) 2014 Hans Verkuil <hverkuil@xs4all.nl>
17  *
18  *  This program is free software; you can redistribute it and/or modify
19  *  it under the terms of the GNU General Public License as published by
20  *  the Free Software Foundation; either version 2 of the License, or
21  *  (at your option) any later version.
22  *
23  *  This program is distributed in the hope that it will be useful,
24  *  but WITHOUT ANY WARRANTY; without even the implied warranty of
25  *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
26  *  GNU General Public License for more details.
27  */
28 
29 #include "tw68.h"
30 
31 /**
32  *  @rp		pointer to current risc program position
33  *  @sglist	pointer to "scatter-gather list" of buffer pointers
34  *  @offset	offset to target memory buffer
35  *  @sync_line	0 -> no sync, 1 -> odd sync, 2 -> even sync
36  *  @bpl	number of bytes per scan line
37  *  @padding	number of bytes of padding to add
38  *  @lines	number of lines in field
39  *  @jump	insert a jump at the start
40  */
41 static __le32 *tw68_risc_field(__le32 *rp, struct scatterlist *sglist,
42 			    unsigned int offset, u32 sync_line,
43 			    unsigned int bpl, unsigned int padding,
44 			    unsigned int lines, bool jump)
45 {
46 	struct scatterlist *sg;
47 	unsigned int line, todo, done;
48 
49 	if (jump) {
50 		*(rp++) = cpu_to_le32(RISC_JUMP);
51 		*(rp++) = 0;
52 	}
53 
54 	/* sync instruction */
55 	if (sync_line == 1)
56 		*(rp++) = cpu_to_le32(RISC_SYNCO);
57 	else
58 		*(rp++) = cpu_to_le32(RISC_SYNCE);
59 	*(rp++) = 0;
60 
61 	/* scan lines */
62 	sg = sglist;
63 	for (line = 0; line < lines; line++) {
64 		/* calculate next starting position */
65 		while (offset && offset >= sg_dma_len(sg)) {
66 			offset -= sg_dma_len(sg);
67 			sg = sg_next(sg);
68 		}
69 		if (bpl <= sg_dma_len(sg) - offset) {
70 			/* fits into current chunk */
71 			*(rp++) = cpu_to_le32(RISC_LINESTART |
72 					      /* (offset<<12) |*/  bpl);
73 			*(rp++) = cpu_to_le32(sg_dma_address(sg) + offset);
74 			offset += bpl;
75 		} else {
76 			/*
77 			 * scanline needs to be split.  Put the start in
78 			 * whatever memory remains using RISC_LINESTART,
79 			 * then the remainder into following addresses
80 			 * given by the scatter-gather list.
81 			 */
82 			todo = bpl;	/* one full line to be done */
83 			/* first fragment */
84 			done = (sg_dma_len(sg) - offset);
85 			*(rp++) = cpu_to_le32(RISC_LINESTART |
86 						(7 << 24) |
87 						done);
88 			*(rp++) = cpu_to_le32(sg_dma_address(sg) + offset);
89 			todo -= done;
90 			sg = sg_next(sg);
91 			/* succeeding fragments have no offset */
92 			while (todo > sg_dma_len(sg)) {
93 				*(rp++) = cpu_to_le32(RISC_INLINE |
94 						(done << 12) |
95 						sg_dma_len(sg));
96 				*(rp++) = cpu_to_le32(sg_dma_address(sg));
97 				todo -= sg_dma_len(sg);
98 				sg = sg_next(sg);
99 				done += sg_dma_len(sg);
100 			}
101 			if (todo) {
102 				/* final chunk - offset 0, count 'todo' */
103 				*(rp++) = cpu_to_le32(RISC_INLINE |
104 							(done << 12) |
105 							todo);
106 				*(rp++) = cpu_to_le32(sg_dma_address(sg));
107 			}
108 			offset = todo;
109 		}
110 		offset += padding;
111 	}
112 
113 	return rp;
114 }
115 
116 /**
117  * tw68_risc_buffer
118  *
119  *	This routine is called by tw68-video.  It allocates
120  *	memory for the dma controller "program" and then fills in that
121  *	memory with the appropriate "instructions".
122  *
123  *	@pci_dev	structure with info about the pci
124  *			slot which our device is in.
125  *	@risc		structure with info about the memory
126  *			used for our controller program.
127  *	@sglist		scatter-gather list entry
128  *	@top_offset	offset within the risc program area for the
129  *			first odd frame line
130  *	@bottom_offset	offset within the risc program area for the
131  *			first even frame line
132  *	@bpl		number of data bytes per scan line
133  *	@padding	number of extra bytes to add at end of line
134  *	@lines		number of scan lines
135  */
136 int tw68_risc_buffer(struct pci_dev *pci,
137 			struct tw68_buf *buf,
138 			struct scatterlist *sglist,
139 			unsigned int top_offset,
140 			unsigned int bottom_offset,
141 			unsigned int bpl,
142 			unsigned int padding,
143 			unsigned int lines)
144 {
145 	u32 instructions, fields;
146 	__le32 *rp;
147 
148 	fields = 0;
149 	if (UNSET != top_offset)
150 		fields++;
151 	if (UNSET != bottom_offset)
152 		fields++;
153 	/*
154 	 * estimate risc mem: worst case is one write per page border +
155 	 * one write per scan line + syncs + 2 jumps (all 2 dwords).
156 	 * Padding can cause next bpl to start close to a page border.
157 	 * First DMA region may be smaller than PAGE_SIZE
158 	 */
159 	instructions  = fields * (1 + (((bpl + padding) * lines) /
160 			 PAGE_SIZE) + lines) + 4;
161 	buf->size = instructions * 8;
162 	buf->cpu = pci_alloc_consistent(pci, buf->size, &buf->dma);
163 	if (buf->cpu == NULL)
164 		return -ENOMEM;
165 
166 	/* write risc instructions */
167 	rp = buf->cpu;
168 	if (UNSET != top_offset)	/* generates SYNCO */
169 		rp = tw68_risc_field(rp, sglist, top_offset, 1,
170 				     bpl, padding, lines, true);
171 	if (UNSET != bottom_offset)	/* generates SYNCE */
172 		rp = tw68_risc_field(rp, sglist, bottom_offset, 2,
173 				     bpl, padding, lines, top_offset == UNSET);
174 
175 	/* save pointer to jmp instruction address */
176 	buf->jmp = rp;
177 	buf->cpu[1] = cpu_to_le32(buf->dma + 8);
178 	/* assure risc buffer hasn't overflowed */
179 	BUG_ON((buf->jmp - buf->cpu + 2) * sizeof(buf->cpu[0]) > buf->size);
180 	return 0;
181 }
182 
183 #if 0
184 /* ------------------------------------------------------------------ */
185 /* debug helper code                                                  */
186 
187 static void tw68_risc_decode(u32 risc, u32 addr)
188 {
189 #define	RISC_OP(reg)	(((reg) >> 28) & 7)
190 	static struct instr_details {
191 		char *name;
192 		u8 has_data_type;
193 		u8 has_byte_info;
194 		u8 has_addr;
195 	} instr[8] = {
196 		[RISC_OP(RISC_SYNCO)]	  = {"syncOdd", 0, 0, 0},
197 		[RISC_OP(RISC_SYNCE)]	  = {"syncEven", 0, 0, 0},
198 		[RISC_OP(RISC_JUMP)]	  = {"jump", 0, 0, 1},
199 		[RISC_OP(RISC_LINESTART)] = {"lineStart", 1, 1, 1},
200 		[RISC_OP(RISC_INLINE)]	  = {"inline", 1, 1, 1},
201 	};
202 	u32 p;
203 
204 	p = RISC_OP(risc);
205 	if (!(risc & 0x80000000) || !instr[p].name) {
206 		pr_debug("0x%08x [ INVALID ]\n", risc);
207 		return;
208 	}
209 	pr_debug("0x%08x %-9s IRQ=%d",
210 		risc, instr[p].name, (risc >> 27) & 1);
211 	if (instr[p].has_data_type)
212 		pr_debug(" Type=%d", (risc >> 24) & 7);
213 	if (instr[p].has_byte_info)
214 		pr_debug(" Start=0x%03x Count=%03u",
215 			(risc >> 12) & 0xfff, risc & 0xfff);
216 	if (instr[p].has_addr)
217 		pr_debug(" StartAddr=0x%08x", addr);
218 	pr_debug("\n");
219 }
220 
221 void tw68_risc_program_dump(struct tw68_core *core, struct tw68_buf *buf)
222 {
223 	const __le32 *addr;
224 
225 	pr_debug("%s: risc_program_dump: risc=%p, buf->cpu=0x%p, buf->jmp=0x%p\n",
226 		  core->name, buf, buf->cpu, buf->jmp);
227 	for (addr = buf->cpu; addr <= buf->jmp; addr += 2)
228 		tw68_risc_decode(*addr, *(addr+1));
229 }
230 #endif
231