diff options
author | Geert Uytterhoeven <Geert.Uytterhoeven@sonycom.com> | 2009-03-05 21:54:09 -0500 |
---|---|---|
committer | Benjamin Herrenschmidt <benh@kernel.crashing.org> | 2009-03-13 01:07:19 -0400 |
commit | f507cd22035fdadd5dbb476dd05e9e7ee21c3b84 (patch) | |
tree | 6d152e75c9ae1ba59349c969ae4b96b1e9f61304 /drivers/block/ps3vram.c | |
parent | 9ead64974b05501bbac0d63a47c99fa786d064ba (diff) |
ps3/block: Replace mtd/ps3vram by block/ps3vram
Convert the PS3 Video RAM Storage Driver from an MTD driver to a plain block
device driver.
The ps3vram driver exposes unused video RAM on the PS3 as a block device
suitable for storage or swap. Fast data transfer is achieved using a local
cache in system RAM and DMA transfers via the GPU.
The new driver is ca. 50% faster for reading, and ca. 10% for writing.
Signed-off-by: Geert Uytterhoeven <Geert.Uytterhoeven@sonycom.com>
Acked-by: Geoff Levand <geoffrey.levand@am.sony.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Diffstat (limited to 'drivers/block/ps3vram.c')
-rw-r--r-- | drivers/block/ps3vram.c | 865 |
1 files changed, 865 insertions, 0 deletions
diff --git a/drivers/block/ps3vram.c b/drivers/block/ps3vram.c new file mode 100644 index 000000000000..393ed6760d78 --- /dev/null +++ b/drivers/block/ps3vram.c | |||
@@ -0,0 +1,865 @@ | |||
1 | /* | ||
2 | * ps3vram - Use extra PS3 video ram as MTD block device. | ||
3 | * | ||
4 | * Copyright 2009 Sony Corporation | ||
5 | * | ||
6 | * Based on the MTD ps3vram driver, which is | ||
7 | * Copyright (c) 2007-2008 Jim Paris <jim@jtan.com> | ||
8 | * Added support RSX DMA Vivien Chappelier <vivien.chappelier@free.fr> | ||
9 | */ | ||
10 | |||
11 | #include <linux/blkdev.h> | ||
12 | #include <linux/delay.h> | ||
13 | #include <linux/proc_fs.h> | ||
14 | #include <linux/seq_file.h> | ||
15 | |||
16 | #include <asm/firmware.h> | ||
17 | #include <asm/lv1call.h> | ||
18 | #include <asm/ps3.h> | ||
19 | |||
20 | |||
21 | #define DEVICE_NAME "ps3vram" | ||
22 | |||
23 | |||
24 | #define XDR_BUF_SIZE (2 * 1024 * 1024) /* XDR buffer (must be 1MiB aligned) */ | ||
25 | #define XDR_IOIF 0x0c000000 | ||
26 | |||
27 | #define FIFO_BASE XDR_IOIF | ||
28 | #define FIFO_SIZE (64 * 1024) | ||
29 | |||
30 | #define DMA_PAGE_SIZE (4 * 1024) | ||
31 | |||
32 | #define CACHE_PAGE_SIZE (256 * 1024) | ||
33 | #define CACHE_PAGE_COUNT ((XDR_BUF_SIZE - FIFO_SIZE) / CACHE_PAGE_SIZE) | ||
34 | |||
35 | #define CACHE_OFFSET CACHE_PAGE_SIZE | ||
36 | #define FIFO_OFFSET 0 | ||
37 | |||
38 | #define CTRL_PUT 0x10 | ||
39 | #define CTRL_GET 0x11 | ||
40 | #define CTRL_TOP 0x15 | ||
41 | |||
42 | #define UPLOAD_SUBCH 1 | ||
43 | #define DOWNLOAD_SUBCH 2 | ||
44 | |||
45 | #define NV_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN 0x0000030c | ||
46 | #define NV_MEMORY_TO_MEMORY_FORMAT_NOTIFY 0x00000104 | ||
47 | |||
48 | #define L1GPU_CONTEXT_ATTRIBUTE_FB_BLIT 0x601 | ||
49 | |||
50 | #define CACHE_PAGE_PRESENT 1 | ||
51 | #define CACHE_PAGE_DIRTY 2 | ||
52 | |||
53 | struct ps3vram_tag { | ||
54 | unsigned int address; | ||
55 | unsigned int flags; | ||
56 | }; | ||
57 | |||
58 | struct ps3vram_cache { | ||
59 | unsigned int page_count; | ||
60 | unsigned int page_size; | ||
61 | struct ps3vram_tag *tags; | ||
62 | unsigned int hit; | ||
63 | unsigned int miss; | ||
64 | }; | ||
65 | |||
66 | struct ps3vram_priv { | ||
67 | struct request_queue *queue; | ||
68 | struct gendisk *gendisk; | ||
69 | |||
70 | u64 size; | ||
71 | |||
72 | u64 memory_handle; | ||
73 | u64 context_handle; | ||
74 | u32 *ctrl; | ||
75 | u32 *reports; | ||
76 | u8 __iomem *ddr_base; | ||
77 | u8 *xdr_buf; | ||
78 | |||
79 | u32 *fifo_base; | ||
80 | u32 *fifo_ptr; | ||
81 | |||
82 | struct ps3vram_cache cache; | ||
83 | |||
84 | /* Used to serialize cache/DMA operations */ | ||
85 | struct mutex lock; | ||
86 | }; | ||
87 | |||
88 | |||
89 | static int ps3vram_major; | ||
90 | |||
91 | |||
92 | static struct block_device_operations ps3vram_fops = { | ||
93 | .owner = THIS_MODULE, | ||
94 | }; | ||
95 | |||
96 | |||
97 | #define DMA_NOTIFIER_HANDLE_BASE 0x66604200 /* first DMA notifier handle */ | ||
98 | #define DMA_NOTIFIER_OFFSET_BASE 0x1000 /* first DMA notifier offset */ | ||
99 | #define DMA_NOTIFIER_SIZE 0x40 | ||
100 | #define NOTIFIER 7 /* notifier used for completion report */ | ||
101 | |||
102 | static char *size = "256M"; | ||
103 | module_param(size, charp, 0); | ||
104 | MODULE_PARM_DESC(size, "memory size"); | ||
105 | |||
106 | static u32 *ps3vram_get_notifier(u32 *reports, int notifier) | ||
107 | { | ||
108 | return (void *)reports + DMA_NOTIFIER_OFFSET_BASE + | ||
109 | DMA_NOTIFIER_SIZE * notifier; | ||
110 | } | ||
111 | |||
112 | static void ps3vram_notifier_reset(struct ps3_system_bus_device *dev) | ||
113 | { | ||
114 | struct ps3vram_priv *priv = dev->core.driver_data; | ||
115 | u32 *notify = ps3vram_get_notifier(priv->reports, NOTIFIER); | ||
116 | int i; | ||
117 | |||
118 | for (i = 0; i < 4; i++) | ||
119 | notify[i] = 0xffffffff; | ||
120 | } | ||
121 | |||
122 | static int ps3vram_notifier_wait(struct ps3_system_bus_device *dev, | ||
123 | unsigned int timeout_ms) | ||
124 | { | ||
125 | struct ps3vram_priv *priv = dev->core.driver_data; | ||
126 | u32 *notify = ps3vram_get_notifier(priv->reports, NOTIFIER); | ||
127 | unsigned long timeout = jiffies + msecs_to_jiffies(timeout_ms); | ||
128 | |||
129 | do { | ||
130 | if (!notify[3]) | ||
131 | return 0; | ||
132 | msleep(1); | ||
133 | } while (time_before(jiffies, timeout)); | ||
134 | |||
135 | return -ETIMEDOUT; | ||
136 | } | ||
137 | |||
138 | static void ps3vram_init_ring(struct ps3_system_bus_device *dev) | ||
139 | { | ||
140 | struct ps3vram_priv *priv = dev->core.driver_data; | ||
141 | |||
142 | priv->ctrl[CTRL_PUT] = FIFO_BASE + FIFO_OFFSET; | ||
143 | priv->ctrl[CTRL_GET] = FIFO_BASE + FIFO_OFFSET; | ||
144 | } | ||
145 | |||
146 | static int ps3vram_wait_ring(struct ps3_system_bus_device *dev, | ||
147 | unsigned int timeout_ms) | ||
148 | { | ||
149 | struct ps3vram_priv *priv = dev->core.driver_data; | ||
150 | unsigned long timeout = jiffies + msecs_to_jiffies(timeout_ms); | ||
151 | |||
152 | do { | ||
153 | if (priv->ctrl[CTRL_PUT] == priv->ctrl[CTRL_GET]) | ||
154 | return 0; | ||
155 | msleep(1); | ||
156 | } while (time_before(jiffies, timeout)); | ||
157 | |||
158 | dev_warn(&dev->core, "FIFO timeout (%08x/%08x/%08x)\n", | ||
159 | priv->ctrl[CTRL_PUT], priv->ctrl[CTRL_GET], | ||
160 | priv->ctrl[CTRL_TOP]); | ||
161 | |||
162 | return -ETIMEDOUT; | ||
163 | } | ||
164 | |||
165 | static void ps3vram_out_ring(struct ps3vram_priv *priv, u32 data) | ||
166 | { | ||
167 | *(priv->fifo_ptr)++ = data; | ||
168 | } | ||
169 | |||
170 | static void ps3vram_begin_ring(struct ps3vram_priv *priv, u32 chan, u32 tag, | ||
171 | u32 size) | ||
172 | { | ||
173 | ps3vram_out_ring(priv, (size << 18) | (chan << 13) | tag); | ||
174 | } | ||
175 | |||
176 | static void ps3vram_rewind_ring(struct ps3_system_bus_device *dev) | ||
177 | { | ||
178 | struct ps3vram_priv *priv = dev->core.driver_data; | ||
179 | int status; | ||
180 | |||
181 | ps3vram_out_ring(priv, 0x20000000 | (FIFO_BASE + FIFO_OFFSET)); | ||
182 | |||
183 | priv->ctrl[CTRL_PUT] = FIFO_BASE + FIFO_OFFSET; | ||
184 | |||
185 | /* asking the HV for a blit will kick the FIFO */ | ||
186 | status = lv1_gpu_context_attribute(priv->context_handle, | ||
187 | L1GPU_CONTEXT_ATTRIBUTE_FB_BLIT, 0, | ||
188 | 0, 0, 0); | ||
189 | if (status) | ||
190 | dev_err(&dev->core, | ||
191 | "%s: lv1_gpu_context_attribute failed %d\n", __func__, | ||
192 | status); | ||
193 | |||
194 | priv->fifo_ptr = priv->fifo_base; | ||
195 | } | ||
196 | |||
197 | static void ps3vram_fire_ring(struct ps3_system_bus_device *dev) | ||
198 | { | ||
199 | struct ps3vram_priv *priv = dev->core.driver_data; | ||
200 | int status; | ||
201 | |||
202 | mutex_lock(&ps3_gpu_mutex); | ||
203 | |||
204 | priv->ctrl[CTRL_PUT] = FIFO_BASE + FIFO_OFFSET + | ||
205 | (priv->fifo_ptr - priv->fifo_base) * sizeof(u32); | ||
206 | |||
207 | /* asking the HV for a blit will kick the FIFO */ | ||
208 | status = lv1_gpu_context_attribute(priv->context_handle, | ||
209 | L1GPU_CONTEXT_ATTRIBUTE_FB_BLIT, 0, | ||
210 | 0, 0, 0); | ||
211 | if (status) | ||
212 | dev_err(&dev->core, | ||
213 | "%s: lv1_gpu_context_attribute failed %d\n", __func__, | ||
214 | status); | ||
215 | |||
216 | if ((priv->fifo_ptr - priv->fifo_base) * sizeof(u32) > | ||
217 | FIFO_SIZE - 1024) { | ||
218 | dev_dbg(&dev->core, "FIFO full, rewinding\n"); | ||
219 | ps3vram_wait_ring(dev, 200); | ||
220 | ps3vram_rewind_ring(dev); | ||
221 | } | ||
222 | |||
223 | mutex_unlock(&ps3_gpu_mutex); | ||
224 | } | ||
225 | |||
226 | static void ps3vram_bind(struct ps3_system_bus_device *dev) | ||
227 | { | ||
228 | struct ps3vram_priv *priv = dev->core.driver_data; | ||
229 | |||
230 | ps3vram_begin_ring(priv, UPLOAD_SUBCH, 0, 1); | ||
231 | ps3vram_out_ring(priv, 0x31337303); | ||
232 | ps3vram_begin_ring(priv, UPLOAD_SUBCH, 0x180, 3); | ||
233 | ps3vram_out_ring(priv, DMA_NOTIFIER_HANDLE_BASE + NOTIFIER); | ||
234 | ps3vram_out_ring(priv, 0xfeed0001); /* DMA system RAM instance */ | ||
235 | ps3vram_out_ring(priv, 0xfeed0000); /* DMA video RAM instance */ | ||
236 | |||
237 | ps3vram_begin_ring(priv, DOWNLOAD_SUBCH, 0, 1); | ||
238 | ps3vram_out_ring(priv, 0x3137c0de); | ||
239 | ps3vram_begin_ring(priv, DOWNLOAD_SUBCH, 0x180, 3); | ||
240 | ps3vram_out_ring(priv, DMA_NOTIFIER_HANDLE_BASE + NOTIFIER); | ||
241 | ps3vram_out_ring(priv, 0xfeed0000); /* DMA video RAM instance */ | ||
242 | ps3vram_out_ring(priv, 0xfeed0001); /* DMA system RAM instance */ | ||
243 | |||
244 | ps3vram_fire_ring(dev); | ||
245 | } | ||
246 | |||
247 | static int ps3vram_upload(struct ps3_system_bus_device *dev, | ||
248 | unsigned int src_offset, unsigned int dst_offset, | ||
249 | int len, int count) | ||
250 | { | ||
251 | struct ps3vram_priv *priv = dev->core.driver_data; | ||
252 | |||
253 | ps3vram_begin_ring(priv, UPLOAD_SUBCH, | ||
254 | NV_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN, 8); | ||
255 | ps3vram_out_ring(priv, XDR_IOIF + src_offset); | ||
256 | ps3vram_out_ring(priv, dst_offset); | ||
257 | ps3vram_out_ring(priv, len); | ||
258 | ps3vram_out_ring(priv, len); | ||
259 | ps3vram_out_ring(priv, len); | ||
260 | ps3vram_out_ring(priv, count); | ||
261 | ps3vram_out_ring(priv, (1 << 8) | 1); | ||
262 | ps3vram_out_ring(priv, 0); | ||
263 | |||
264 | ps3vram_notifier_reset(dev); | ||
265 | ps3vram_begin_ring(priv, UPLOAD_SUBCH, | ||
266 | NV_MEMORY_TO_MEMORY_FORMAT_NOTIFY, 1); | ||
267 | ps3vram_out_ring(priv, 0); | ||
268 | ps3vram_begin_ring(priv, UPLOAD_SUBCH, 0x100, 1); | ||
269 | ps3vram_out_ring(priv, 0); | ||
270 | ps3vram_fire_ring(dev); | ||
271 | if (ps3vram_notifier_wait(dev, 200) < 0) { | ||
272 | dev_warn(&dev->core, "%s: Notifier timeout\n", __func__); | ||
273 | return -1; | ||
274 | } | ||
275 | |||
276 | return 0; | ||
277 | } | ||
278 | |||
279 | static int ps3vram_download(struct ps3_system_bus_device *dev, | ||
280 | unsigned int src_offset, unsigned int dst_offset, | ||
281 | int len, int count) | ||
282 | { | ||
283 | struct ps3vram_priv *priv = dev->core.driver_data; | ||
284 | |||
285 | ps3vram_begin_ring(priv, DOWNLOAD_SUBCH, | ||
286 | NV_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN, 8); | ||
287 | ps3vram_out_ring(priv, src_offset); | ||
288 | ps3vram_out_ring(priv, XDR_IOIF + dst_offset); | ||
289 | ps3vram_out_ring(priv, len); | ||
290 | ps3vram_out_ring(priv, len); | ||
291 | ps3vram_out_ring(priv, len); | ||
292 | ps3vram_out_ring(priv, count); | ||
293 | ps3vram_out_ring(priv, (1 << 8) | 1); | ||
294 | ps3vram_out_ring(priv, 0); | ||
295 | |||
296 | ps3vram_notifier_reset(dev); | ||
297 | ps3vram_begin_ring(priv, DOWNLOAD_SUBCH, | ||
298 | NV_MEMORY_TO_MEMORY_FORMAT_NOTIFY, 1); | ||
299 | ps3vram_out_ring(priv, 0); | ||
300 | ps3vram_begin_ring(priv, DOWNLOAD_SUBCH, 0x100, 1); | ||
301 | ps3vram_out_ring(priv, 0); | ||
302 | ps3vram_fire_ring(dev); | ||
303 | if (ps3vram_notifier_wait(dev, 200) < 0) { | ||
304 | dev_warn(&dev->core, "%s: Notifier timeout\n", __func__); | ||
305 | return -1; | ||
306 | } | ||
307 | |||
308 | return 0; | ||
309 | } | ||
310 | |||
311 | static void ps3vram_cache_evict(struct ps3_system_bus_device *dev, int entry) | ||
312 | { | ||
313 | struct ps3vram_priv *priv = dev->core.driver_data; | ||
314 | struct ps3vram_cache *cache = &priv->cache; | ||
315 | |||
316 | if (!(cache->tags[entry].flags & CACHE_PAGE_DIRTY)) | ||
317 | return; | ||
318 | |||
319 | dev_dbg(&dev->core, "Flushing %d: 0x%08x\n", entry, | ||
320 | cache->tags[entry].address); | ||
321 | if (ps3vram_upload(dev, CACHE_OFFSET + entry * cache->page_size, | ||
322 | cache->tags[entry].address, DMA_PAGE_SIZE, | ||
323 | cache->page_size / DMA_PAGE_SIZE) < 0) { | ||
324 | dev_err(&dev->core, | ||
325 | "Failed to upload from 0x%x to " "0x%x size 0x%x\n", | ||
326 | entry * cache->page_size, cache->tags[entry].address, | ||
327 | cache->page_size); | ||
328 | } | ||
329 | cache->tags[entry].flags &= ~CACHE_PAGE_DIRTY; | ||
330 | } | ||
331 | |||
332 | static void ps3vram_cache_load(struct ps3_system_bus_device *dev, int entry, | ||
333 | unsigned int address) | ||
334 | { | ||
335 | struct ps3vram_priv *priv = dev->core.driver_data; | ||
336 | struct ps3vram_cache *cache = &priv->cache; | ||
337 | |||
338 | dev_dbg(&dev->core, "Fetching %d: 0x%08x\n", entry, address); | ||
339 | if (ps3vram_download(dev, address, | ||
340 | CACHE_OFFSET + entry * cache->page_size, | ||
341 | DMA_PAGE_SIZE, | ||
342 | cache->page_size / DMA_PAGE_SIZE) < 0) { | ||
343 | dev_err(&dev->core, | ||
344 | "Failed to download from 0x%x to 0x%x size 0x%x\n", | ||
345 | address, entry * cache->page_size, cache->page_size); | ||
346 | } | ||
347 | |||
348 | cache->tags[entry].address = address; | ||
349 | cache->tags[entry].flags |= CACHE_PAGE_PRESENT; | ||
350 | } | ||
351 | |||
352 | |||
353 | static void ps3vram_cache_flush(struct ps3_system_bus_device *dev) | ||
354 | { | ||
355 | struct ps3vram_priv *priv = dev->core.driver_data; | ||
356 | struct ps3vram_cache *cache = &priv->cache; | ||
357 | int i; | ||
358 | |||
359 | dev_dbg(&dev->core, "FLUSH\n"); | ||
360 | for (i = 0; i < cache->page_count; i++) { | ||
361 | ps3vram_cache_evict(dev, i); | ||
362 | cache->tags[i].flags = 0; | ||
363 | } | ||
364 | } | ||
365 | |||
366 | static unsigned int ps3vram_cache_match(struct ps3_system_bus_device *dev, | ||
367 | loff_t address) | ||
368 | { | ||
369 | struct ps3vram_priv *priv = dev->core.driver_data; | ||
370 | struct ps3vram_cache *cache = &priv->cache; | ||
371 | unsigned int base; | ||
372 | unsigned int offset; | ||
373 | int i; | ||
374 | static int counter; | ||
375 | |||
376 | offset = (unsigned int) (address & (cache->page_size - 1)); | ||
377 | base = (unsigned int) (address - offset); | ||
378 | |||
379 | /* fully associative check */ | ||
380 | for (i = 0; i < cache->page_count; i++) { | ||
381 | if ((cache->tags[i].flags & CACHE_PAGE_PRESENT) && | ||
382 | cache->tags[i].address == base) { | ||
383 | cache->hit++; | ||
384 | dev_dbg(&dev->core, "Found entry %d: 0x%08x\n", i, | ||
385 | cache->tags[i].address); | ||
386 | return i; | ||
387 | } | ||
388 | } | ||
389 | |||
390 | /* choose a random entry */ | ||
391 | i = (jiffies + (counter++)) % cache->page_count; | ||
392 | dev_dbg(&dev->core, "Using entry %d\n", i); | ||
393 | |||
394 | ps3vram_cache_evict(dev, i); | ||
395 | ps3vram_cache_load(dev, i, base); | ||
396 | |||
397 | cache->miss++; | ||
398 | return i; | ||
399 | } | ||
400 | |||
401 | static int ps3vram_cache_init(struct ps3_system_bus_device *dev) | ||
402 | { | ||
403 | struct ps3vram_priv *priv = dev->core.driver_data; | ||
404 | |||
405 | priv->cache.page_count = CACHE_PAGE_COUNT; | ||
406 | priv->cache.page_size = CACHE_PAGE_SIZE; | ||
407 | priv->cache.tags = kzalloc(sizeof(struct ps3vram_tag) * | ||
408 | CACHE_PAGE_COUNT, GFP_KERNEL); | ||
409 | if (priv->cache.tags == NULL) { | ||
410 | dev_err(&dev->core, "Could not allocate cache tags\n"); | ||
411 | return -ENOMEM; | ||
412 | } | ||
413 | |||
414 | dev_info(&dev->core, "Created ram cache: %d entries, %d KiB each\n", | ||
415 | CACHE_PAGE_COUNT, CACHE_PAGE_SIZE / 1024); | ||
416 | |||
417 | return 0; | ||
418 | } | ||
419 | |||
420 | static void ps3vram_cache_cleanup(struct ps3_system_bus_device *dev) | ||
421 | { | ||
422 | struct ps3vram_priv *priv = dev->core.driver_data; | ||
423 | |||
424 | ps3vram_cache_flush(dev); | ||
425 | kfree(priv->cache.tags); | ||
426 | } | ||
427 | |||
428 | static int ps3vram_read(struct ps3_system_bus_device *dev, loff_t from, | ||
429 | size_t len, size_t *retlen, u_char *buf) | ||
430 | { | ||
431 | struct ps3vram_priv *priv = dev->core.driver_data; | ||
432 | unsigned int cached, count; | ||
433 | |||
434 | dev_dbg(&dev->core, "%s: from=0x%08x len=0x%zx\n", __func__, | ||
435 | (unsigned int)from, len); | ||
436 | |||
437 | if (from >= priv->size) | ||
438 | return -EIO; | ||
439 | |||
440 | if (len > priv->size - from) | ||
441 | len = priv->size - from; | ||
442 | |||
443 | /* Copy from vram to buf */ | ||
444 | count = len; | ||
445 | while (count) { | ||
446 | unsigned int offset, avail; | ||
447 | unsigned int entry; | ||
448 | |||
449 | offset = (unsigned int) (from & (priv->cache.page_size - 1)); | ||
450 | avail = priv->cache.page_size - offset; | ||
451 | |||
452 | mutex_lock(&priv->lock); | ||
453 | |||
454 | entry = ps3vram_cache_match(dev, from); | ||
455 | cached = CACHE_OFFSET + entry * priv->cache.page_size + offset; | ||
456 | |||
457 | dev_dbg(&dev->core, "%s: from=%08x cached=%08x offset=%08x " | ||
458 | "avail=%08x count=%08x\n", __func__, | ||
459 | (unsigned int)from, cached, offset, avail, count); | ||
460 | |||
461 | if (avail > count) | ||
462 | avail = count; | ||
463 | memcpy(buf, priv->xdr_buf + cached, avail); | ||
464 | |||
465 | mutex_unlock(&priv->lock); | ||
466 | |||
467 | buf += avail; | ||
468 | count -= avail; | ||
469 | from += avail; | ||
470 | } | ||
471 | |||
472 | *retlen = len; | ||
473 | return 0; | ||
474 | } | ||
475 | |||
476 | static int ps3vram_write(struct ps3_system_bus_device *dev, loff_t to, | ||
477 | size_t len, size_t *retlen, const u_char *buf) | ||
478 | { | ||
479 | struct ps3vram_priv *priv = dev->core.driver_data; | ||
480 | unsigned int cached, count; | ||
481 | |||
482 | if (to >= priv->size) | ||
483 | return -EIO; | ||
484 | |||
485 | if (len > priv->size - to) | ||
486 | len = priv->size - to; | ||
487 | |||
488 | /* Copy from buf to vram */ | ||
489 | count = len; | ||
490 | while (count) { | ||
491 | unsigned int offset, avail; | ||
492 | unsigned int entry; | ||
493 | |||
494 | offset = (unsigned int) (to & (priv->cache.page_size - 1)); | ||
495 | avail = priv->cache.page_size - offset; | ||
496 | |||
497 | mutex_lock(&priv->lock); | ||
498 | |||
499 | entry = ps3vram_cache_match(dev, to); | ||
500 | cached = CACHE_OFFSET + entry * priv->cache.page_size + offset; | ||
501 | |||
502 | dev_dbg(&dev->core, "%s: to=%08x cached=%08x offset=%08x " | ||
503 | "avail=%08x count=%08x\n", __func__, (unsigned int)to, | ||
504 | cached, offset, avail, count); | ||
505 | |||
506 | if (avail > count) | ||
507 | avail = count; | ||
508 | memcpy(priv->xdr_buf + cached, buf, avail); | ||
509 | |||
510 | priv->cache.tags[entry].flags |= CACHE_PAGE_DIRTY; | ||
511 | |||
512 | mutex_unlock(&priv->lock); | ||
513 | |||
514 | buf += avail; | ||
515 | count -= avail; | ||
516 | to += avail; | ||
517 | } | ||
518 | |||
519 | *retlen = len; | ||
520 | return 0; | ||
521 | } | ||
522 | |||
523 | static int ps3vram_proc_show(struct seq_file *m, void *v) | ||
524 | { | ||
525 | struct ps3vram_priv *priv = m->private; | ||
526 | |||
527 | seq_printf(m, "hit:%u\nmiss:%u\n", priv->cache.hit, priv->cache.miss); | ||
528 | return 0; | ||
529 | } | ||
530 | |||
531 | static int ps3vram_proc_open(struct inode *inode, struct file *file) | ||
532 | { | ||
533 | return single_open(file, ps3vram_proc_show, PDE(inode)->data); | ||
534 | } | ||
535 | |||
536 | static const struct file_operations ps3vram_proc_fops = { | ||
537 | .owner = THIS_MODULE, | ||
538 | .open = ps3vram_proc_open, | ||
539 | .read = seq_read, | ||
540 | .llseek = seq_lseek, | ||
541 | .release = single_release, | ||
542 | }; | ||
543 | |||
544 | static void __devinit ps3vram_proc_init(struct ps3_system_bus_device *dev) | ||
545 | { | ||
546 | struct ps3vram_priv *priv = dev->core.driver_data; | ||
547 | struct proc_dir_entry *pde; | ||
548 | |||
549 | pde = proc_create(DEVICE_NAME, 0444, NULL, &ps3vram_proc_fops); | ||
550 | if (!pde) { | ||
551 | dev_warn(&dev->core, "failed to create /proc entry\n"); | ||
552 | return; | ||
553 | } | ||
554 | |||
555 | pde->owner = THIS_MODULE; | ||
556 | pde->data = priv; | ||
557 | } | ||
558 | |||
559 | static int ps3vram_make_request(struct request_queue *q, struct bio *bio) | ||
560 | { | ||
561 | struct ps3_system_bus_device *dev = q->queuedata; | ||
562 | int write = bio_data_dir(bio) == WRITE; | ||
563 | const char *op = write ? "write" : "read"; | ||
564 | loff_t offset = bio->bi_sector << 9; | ||
565 | int error = 0; | ||
566 | struct bio_vec *bvec; | ||
567 | unsigned int i; | ||
568 | |||
569 | dev_dbg(&dev->core, "%s\n", __func__); | ||
570 | |||
571 | bio_for_each_segment(bvec, bio, i) { | ||
572 | /* PS3 is ppc64, so we don't handle highmem */ | ||
573 | char *ptr = page_address(bvec->bv_page) + bvec->bv_offset; | ||
574 | size_t len = bvec->bv_len, retlen; | ||
575 | |||
576 | dev_dbg(&dev->core, " %s %zu bytes at offset %llu\n", op, | ||
577 | len, offset); | ||
578 | if (write) | ||
579 | error = ps3vram_write(dev, offset, len, &retlen, ptr); | ||
580 | else | ||
581 | error = ps3vram_read(dev, offset, len, &retlen, ptr); | ||
582 | |||
583 | if (error) { | ||
584 | dev_err(&dev->core, "%s failed\n", op); | ||
585 | goto out; | ||
586 | } | ||
587 | |||
588 | if (retlen != len) { | ||
589 | dev_err(&dev->core, "Short %s\n", op); | ||
590 | goto out; | ||
591 | } | ||
592 | |||
593 | offset += len; | ||
594 | } | ||
595 | |||
596 | dev_dbg(&dev->core, "%s completed\n", op); | ||
597 | |||
598 | out: | ||
599 | bio_endio(bio, error); | ||
600 | return 0; | ||
601 | } | ||
602 | |||
603 | static int __devinit ps3vram_probe(struct ps3_system_bus_device *dev) | ||
604 | { | ||
605 | struct ps3vram_priv *priv; | ||
606 | int error, status; | ||
607 | struct request_queue *queue; | ||
608 | struct gendisk *gendisk; | ||
609 | u64 ddr_lpar, ctrl_lpar, info_lpar, reports_lpar, ddr_size, | ||
610 | reports_size; | ||
611 | char *rest; | ||
612 | |||
613 | priv = kzalloc(sizeof(*priv), GFP_KERNEL); | ||
614 | if (!priv) { | ||
615 | error = -ENOMEM; | ||
616 | goto fail; | ||
617 | } | ||
618 | |||
619 | mutex_init(&priv->lock); | ||
620 | dev->core.driver_data = priv; | ||
621 | |||
622 | priv = dev->core.driver_data; | ||
623 | |||
624 | /* Allocate XDR buffer (1MiB aligned) */ | ||
625 | priv->xdr_buf = (void *)__get_free_pages(GFP_KERNEL, | ||
626 | get_order(XDR_BUF_SIZE)); | ||
627 | if (priv->xdr_buf == NULL) { | ||
628 | dev_err(&dev->core, "Could not allocate XDR buffer\n"); | ||
629 | error = -ENOMEM; | ||
630 | goto fail_free_priv; | ||
631 | } | ||
632 | |||
633 | /* Put FIFO at begginning of XDR buffer */ | ||
634 | priv->fifo_base = (u32 *) (priv->xdr_buf + FIFO_OFFSET); | ||
635 | priv->fifo_ptr = priv->fifo_base; | ||
636 | |||
637 | /* XXX: Need to open GPU, in case ps3fb or snd_ps3 aren't loaded */ | ||
638 | if (ps3_open_hv_device(dev)) { | ||
639 | dev_err(&dev->core, "ps3_open_hv_device failed\n"); | ||
640 | error = -EAGAIN; | ||
641 | goto out_close_gpu; | ||
642 | } | ||
643 | |||
644 | /* Request memory */ | ||
645 | status = -1; | ||
646 | ddr_size = ALIGN(memparse(size, &rest), 1024*1024); | ||
647 | if (!ddr_size) { | ||
648 | dev_err(&dev->core, "Specified size is too small\n"); | ||
649 | error = -EINVAL; | ||
650 | goto out_close_gpu; | ||
651 | } | ||
652 | |||
653 | while (ddr_size > 0) { | ||
654 | status = lv1_gpu_memory_allocate(ddr_size, 0, 0, 0, 0, | ||
655 | &priv->memory_handle, | ||
656 | &ddr_lpar); | ||
657 | if (!status) | ||
658 | break; | ||
659 | ddr_size -= 1024*1024; | ||
660 | } | ||
661 | if (status) { | ||
662 | dev_err(&dev->core, "lv1_gpu_memory_allocate failed %d\n", | ||
663 | status); | ||
664 | error = -ENOMEM; | ||
665 | goto out_free_xdr_buf; | ||
666 | } | ||
667 | |||
668 | /* Request context */ | ||
669 | status = lv1_gpu_context_allocate(priv->memory_handle, 0, | ||
670 | &priv->context_handle, &ctrl_lpar, | ||
671 | &info_lpar, &reports_lpar, | ||
672 | &reports_size); | ||
673 | if (status) { | ||
674 | dev_err(&dev->core, "lv1_gpu_context_allocate failed %d\n", | ||
675 | status); | ||
676 | error = -ENOMEM; | ||
677 | goto out_free_memory; | ||
678 | } | ||
679 | |||
680 | /* Map XDR buffer to RSX */ | ||
681 | status = lv1_gpu_context_iomap(priv->context_handle, XDR_IOIF, | ||
682 | ps3_mm_phys_to_lpar(__pa(priv->xdr_buf)), | ||
683 | XDR_BUF_SIZE, 0); | ||
684 | if (status) { | ||
685 | dev_err(&dev->core, "lv1_gpu_context_iomap failed %d\n", | ||
686 | status); | ||
687 | error = -ENOMEM; | ||
688 | goto out_free_context; | ||
689 | } | ||
690 | |||
691 | priv->ddr_base = ioremap_flags(ddr_lpar, ddr_size, _PAGE_NO_CACHE); | ||
692 | |||
693 | if (!priv->ddr_base) { | ||
694 | dev_err(&dev->core, "ioremap DDR failed\n"); | ||
695 | error = -ENOMEM; | ||
696 | goto out_free_context; | ||
697 | } | ||
698 | |||
699 | priv->ctrl = ioremap(ctrl_lpar, 64 * 1024); | ||
700 | if (!priv->ctrl) { | ||
701 | dev_err(&dev->core, "ioremap CTRL failed\n"); | ||
702 | error = -ENOMEM; | ||
703 | goto out_unmap_vram; | ||
704 | } | ||
705 | |||
706 | priv->reports = ioremap(reports_lpar, reports_size); | ||
707 | if (!priv->reports) { | ||
708 | dev_err(&dev->core, "ioremap REPORTS failed\n"); | ||
709 | error = -ENOMEM; | ||
710 | goto out_unmap_ctrl; | ||
711 | } | ||
712 | |||
713 | mutex_lock(&ps3_gpu_mutex); | ||
714 | ps3vram_init_ring(dev); | ||
715 | mutex_unlock(&ps3_gpu_mutex); | ||
716 | |||
717 | priv->size = ddr_size; | ||
718 | |||
719 | ps3vram_bind(dev); | ||
720 | |||
721 | mutex_lock(&ps3_gpu_mutex); | ||
722 | error = ps3vram_wait_ring(dev, 100); | ||
723 | mutex_unlock(&ps3_gpu_mutex); | ||
724 | if (error < 0) { | ||
725 | dev_err(&dev->core, "Failed to initialize channels\n"); | ||
726 | error = -ETIMEDOUT; | ||
727 | goto out_unmap_reports; | ||
728 | } | ||
729 | |||
730 | ps3vram_cache_init(dev); | ||
731 | ps3vram_proc_init(dev); | ||
732 | |||
733 | queue = blk_alloc_queue(GFP_KERNEL); | ||
734 | if (!queue) { | ||
735 | dev_err(&dev->core, "blk_alloc_queue failed\n"); | ||
736 | error = -ENOMEM; | ||
737 | goto out_cache_cleanup; | ||
738 | } | ||
739 | |||
740 | priv->queue = queue; | ||
741 | queue->queuedata = dev; | ||
742 | blk_queue_make_request(queue, ps3vram_make_request); | ||
743 | blk_queue_max_phys_segments(queue, MAX_PHYS_SEGMENTS); | ||
744 | blk_queue_max_hw_segments(queue, MAX_HW_SEGMENTS); | ||
745 | blk_queue_max_segment_size(queue, MAX_SEGMENT_SIZE); | ||
746 | blk_queue_max_sectors(queue, SAFE_MAX_SECTORS); | ||
747 | |||
748 | gendisk = alloc_disk(1); | ||
749 | if (!gendisk) { | ||
750 | dev_err(&dev->core, "alloc_disk failed\n"); | ||
751 | error = -ENOMEM; | ||
752 | goto fail_cleanup_queue; | ||
753 | } | ||
754 | |||
755 | priv->gendisk = gendisk; | ||
756 | gendisk->major = ps3vram_major; | ||
757 | gendisk->first_minor = 0; | ||
758 | gendisk->fops = &ps3vram_fops; | ||
759 | gendisk->queue = queue; | ||
760 | gendisk->private_data = dev; | ||
761 | gendisk->driverfs_dev = &dev->core; | ||
762 | strlcpy(gendisk->disk_name, DEVICE_NAME, sizeof(gendisk->disk_name)); | ||
763 | set_capacity(gendisk, priv->size >> 9); | ||
764 | |||
765 | dev_info(&dev->core, "%s: Using %lu MiB of GPU memory\n", | ||
766 | gendisk->disk_name, get_capacity(gendisk) >> 11); | ||
767 | |||
768 | add_disk(gendisk); | ||
769 | return 0; | ||
770 | |||
771 | fail_cleanup_queue: | ||
772 | blk_cleanup_queue(queue); | ||
773 | out_cache_cleanup: | ||
774 | remove_proc_entry(DEVICE_NAME, NULL); | ||
775 | ps3vram_cache_cleanup(dev); | ||
776 | out_unmap_reports: | ||
777 | iounmap(priv->reports); | ||
778 | out_unmap_ctrl: | ||
779 | iounmap(priv->ctrl); | ||
780 | out_unmap_vram: | ||
781 | iounmap(priv->ddr_base); | ||
782 | out_free_context: | ||
783 | lv1_gpu_context_free(priv->context_handle); | ||
784 | out_free_memory: | ||
785 | lv1_gpu_memory_free(priv->memory_handle); | ||
786 | out_close_gpu: | ||
787 | ps3_close_hv_device(dev); | ||
788 | out_free_xdr_buf: | ||
789 | free_pages((unsigned long) priv->xdr_buf, get_order(XDR_BUF_SIZE)); | ||
790 | fail_free_priv: | ||
791 | kfree(priv); | ||
792 | dev->core.driver_data = NULL; | ||
793 | fail: | ||
794 | return error; | ||
795 | } | ||
796 | |||
797 | static int ps3vram_remove(struct ps3_system_bus_device *dev) | ||
798 | { | ||
799 | struct ps3vram_priv *priv = dev->core.driver_data; | ||
800 | |||
801 | del_gendisk(priv->gendisk); | ||
802 | put_disk(priv->gendisk); | ||
803 | blk_cleanup_queue(priv->queue); | ||
804 | remove_proc_entry(DEVICE_NAME, NULL); | ||
805 | ps3vram_cache_cleanup(dev); | ||
806 | iounmap(priv->reports); | ||
807 | iounmap(priv->ctrl); | ||
808 | iounmap(priv->ddr_base); | ||
809 | lv1_gpu_context_free(priv->context_handle); | ||
810 | lv1_gpu_memory_free(priv->memory_handle); | ||
811 | ps3_close_hv_device(dev); | ||
812 | free_pages((unsigned long) priv->xdr_buf, get_order(XDR_BUF_SIZE)); | ||
813 | kfree(priv); | ||
814 | dev->core.driver_data = NULL; | ||
815 | return 0; | ||
816 | } | ||
817 | |||
818 | static struct ps3_system_bus_driver ps3vram = { | ||
819 | .match_id = PS3_MATCH_ID_GPU, | ||
820 | .match_sub_id = PS3_MATCH_SUB_ID_GPU_RAMDISK, | ||
821 | .core.name = DEVICE_NAME, | ||
822 | .core.owner = THIS_MODULE, | ||
823 | .probe = ps3vram_probe, | ||
824 | .remove = ps3vram_remove, | ||
825 | .shutdown = ps3vram_remove, | ||
826 | }; | ||
827 | |||
828 | |||
829 | static int __init ps3vram_init(void) | ||
830 | { | ||
831 | int error; | ||
832 | |||
833 | if (!firmware_has_feature(FW_FEATURE_PS3_LV1)) | ||
834 | return -ENODEV; | ||
835 | |||
836 | error = register_blkdev(0, DEVICE_NAME); | ||
837 | if (error <= 0) { | ||
838 | pr_err("%s: register_blkdev failed %d\n", DEVICE_NAME, error); | ||
839 | return error; | ||
840 | } | ||
841 | ps3vram_major = error; | ||
842 | |||
843 | pr_info("%s: registered block device major %d\n", DEVICE_NAME, | ||
844 | ps3vram_major); | ||
845 | |||
846 | error = ps3_system_bus_driver_register(&ps3vram); | ||
847 | if (error) | ||
848 | unregister_blkdev(ps3vram_major, DEVICE_NAME); | ||
849 | |||
850 | return error; | ||
851 | } | ||
852 | |||
853 | static void __exit ps3vram_exit(void) | ||
854 | { | ||
855 | ps3_system_bus_driver_unregister(&ps3vram); | ||
856 | unregister_blkdev(ps3vram_major, DEVICE_NAME); | ||
857 | } | ||
858 | |||
859 | module_init(ps3vram_init); | ||
860 | module_exit(ps3vram_exit); | ||
861 | |||
862 | MODULE_LICENSE("GPL"); | ||
863 | MODULE_DESCRIPTION("PS3 Video RAM Storage Driver"); | ||
864 | MODULE_AUTHOR("Sony Corporation"); | ||
865 | MODULE_ALIAS(PS3_MODULE_ALIAS_GPU_RAMDISK); | ||