1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
|
/*D:400
* The Guest block driver
*
* This is a simple block driver, which appears as /dev/lgba, lgbb, lgbc etc.
* The mechanism is simple: we place the information about the request in the
* device page, then use SEND_DMA (containing the data for a write, or an empty
* "ping" DMA for a read).
:*/
/* Copyright 2006 Rusty Russell <rusty@rustcorp.com.au> IBM Corporation
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*/
//#define DEBUG
#include <linux/init.h>
#include <linux/types.h>
#include <linux/blkdev.h>
#include <linux/interrupt.h>
#include <linux/lguest_bus.h>
static char next_block_index = 'a';
/*D:420 Here is the structure which holds all the information we need about
* each Guest block device.
*
* I'm sure at this stage, you're wondering "hey, where was the adventure I was
* promised?" and thinking "Rusty sucks, I shall say nasty things about him on
* my blog". I think Real adventures have boring bits, too, and you're in the
* middle of one. But it gets better. Just not quite yet. */
struct blockdev
{
/* The block queue infrastructure wants a spinlock: it is held while it
* calls our block request function. We grab it in our interrupt
* handler so the responses don't mess with new requests. */
spinlock_t lock;
/* The disk structure registered with kernel. */
struct gendisk *disk;
/* The major device number for this disk, and the interrupt. We only
* really keep them here for completeness; we'd need them if we
* supported device unplugging. */
int major;
int irq;
/* The physical address of this device's memory page */
unsigned long phys_addr;
/* The mapped memory page for convenient acces. */
struct lguest_block_page *lb_page;
/* We only have a single request outstanding at a time: this is it. */
struct lguest_dma dma;
struct request *req;
};
/*D:495 We originally used end_request() throughout the driver, but it turns
* out that end_request() is deprecated, and doesn't actually end the request
* (which seems like a good reason to deprecate it!). It simply ends the first
* bio. So if we had 3 bios in a "struct request" we would do all 3,
* end_request(), do 2, end_request(), do 1 and end_request(): twice as much
* work as we needed to do.
*
* This reinforced to me that I do not understand the block layer.
*
* Nonetheless, Jens Axboe gave me this nice helper to end all chunks of a
* request. This improved disk speed by 130%. */
static void end_entire_request(struct request *req, int uptodate)
{
if (end_that_request_first(req, uptodate, req->hard_nr_sectors))
BUG();
add_disk_randomness(req->rq_disk);
blkdev_dequeue_request(req);
end_that_request_last(req, uptodate);
}
/* I'm told there are only two stories in the world worth telling: love and
* hate. So there used to be a love scene here like this:
*
* Launcher: We could make beautiful I/O together, you and I.
* Guest: My, that's a big disk!
*
* Unfortunately, it was just too raunchy for our otherwise-gentle tale. */
/*D:490 This is the interrupt handler, called when a block read or write has
* been completed for us. */
static irqreturn_t lgb_irq(int irq, void *_bd)
{
/* We handed our "struct blockdev" as the argument to request_irq(), so
* it is passed through to us here. This tells us which device we're
* dealing with in case we have more than one. */
struct blockdev *bd = _bd;
unsigned long flags;
/* We weren't doing anything? Strange, but could happen if we shared
* interrupts (we don't!). */
if (!bd->req) {
pr_debug("No work!\n");
return IRQ_NONE;
}
/* Not done yet? That's equally strange. */
if (!bd->lb_page->result) {
pr_debug("No result!\n");
return IRQ_NONE;
}
/* We have to grab the lock before ending the request. */
spin_lock_irqsave(&bd->lock, flags);
/* "result" is 1 for success, 2 for failure: end_entire_request() wants
* to know whether this succeeded or not. */
end_entire_request(bd->req, bd->lb_page->result == 1);
/* Clear out request, it's done. */
bd->req = NULL;
/* Reset incoming DMA for next time. */
bd->dma.used_len = 0;
/* Ready for more reads or writes */
blk_start_queue(bd->disk->queue);
spin_unlock_irqrestore(&bd->lock, flags);
/* The interrupt was for us, we dealt with it. */
return IRQ_HANDLED;
}
/*D:480 The block layer's "struct request" contains a number of "struct bio"s,
* each of which contains "struct bio_vec"s, each of which contains a page, an
* offset and a length.
*
* Fortunately there are iterators to help us walk through the "struct
* request". Even more fortunately, there were plenty of places to steal the
* code from. We pack the "struct request" into our "struct lguest_dma" and
* return the total length. */
static unsigned int req_to_dma(struct request *req, struct lguest_dma *dma)
{
unsigned int i = 0, idx, len = 0;
struct bio *bio;
rq_for_each_bio(bio, req) {
struct bio_vec *bvec;
bio_for_each_segment(bvec, bio, idx) {
/* We told the block layer not to give us too many. */
BUG_ON(i == LGUEST_MAX_DMA_SECTIONS);
/* If we had a zero-length segment, it would look like
* the end of the data referred to by the "struct
* lguest_dma", so make sure that doesn't happen. */
BUG_ON(!bvec->bv_len);
/* Convert page & offset to a physical address */
dma->addr[i] = page_to_phys(bvec->bv_page)
+ bvec->bv_offset;
dma->len[i] = bvec->bv_len;
len += bvec->bv_len;
i++;
}
}
/* If the array isn't full, we mark the end with a 0 length */
if (i < LGUEST_MAX_DMA_SECTIONS)
dma->len[i] = 0;
return len;
}
/* This creates an empty DMA, useful for prodding the Host without sending data
* (ie. when we want to do a read) */
static void empty_dma(struct lguest_dma *dma)
{
dma->len[0] = 0;
}
/*D:470 Setting up a request is fairly easy: */
static void setup_req(struct blockdev *bd,
int type, struct request *req, struct lguest_dma *dma)
{
/* The type is 1 (write) or 0 (read). */
bd->lb_page->type = type;
/* The sector on disk where the read or write starts. */
bd->lb_page->sector = req->sector;
/* The result is initialized to 0 (unfinished). */
bd->lb_page->result = 0;
/* The current request (so we can end it in the interrupt handler). */
bd->req = req;
/* The number of bytes: returned as a side-effect of req_to_dma(),
* which packs the block layer's "struct request" into our "struct
* lguest_dma" */
bd->lb_page->bytes = req_to_dma(req, dma);
}
/*D:450 Write is pretty straightforward: we pack the request into a "struct
* lguest_dma", then use SEND_DMA to send the request. */
static void do_write(struct blockdev *bd, struct request *req)
{
struct lguest_dma send;
pr_debug("lgb: WRITE sector %li\n", (long)req->sector);
setup_req(bd, 1, req, &send);
lguest_send_dma(bd->phys_addr, &send);
}
/* Read is similar to write, except we pack the request into our receive
* "struct lguest_dma" and send through an empty DMA just to tell the Host that
* there's a request pending. */
static void do_read(struct blockdev *bd, struct request *req)
{
struct lguest_dma ping;
pr_debug("lgb: READ sector %li\n", (long)req->sector);
setup_req(bd, 0, req, &bd->dma);
empty_dma(&ping);
lguest_send_dma(bd->phys_addr, &ping);
}
/*D:440 This where requests come in: we get handed the request queue and are
* expected to pull a "struct request" off it until we've finished them or
* we're waiting for a reply: */
static void do_lgb_request(struct request_queue *q)
{
struct blockdev *bd;
struct request *req;
again:
/* This sometimes returns NULL even on the very first time around. I
* wonder if it's something to do with letting elves handle the request
* queue... */
req = elv_next_request(q);
if (!req)
return;
/* We attached the struct blockdev to the disk: get it back */
bd = req->rq_disk->private_data;
/* Sometimes we get repeated requests after blk_stop_queue(), but we
* can only handle one at a time. */
if (bd->req)
return;
/* We only do reads and writes: no tricky business! */
if (!blk_fs_request(req)) {
pr_debug("Got non-command 0x%08x\n", req->cmd_type);
req->errors++;
end_entire_request(req, 0);
goto again;
}
if (rq_data_dir(req) == WRITE)
do_write(bd, req);
else
do_read(bd, req);
/* We've put out the request, so stop any more coming in until we get
* an interrupt, which takes us to lgb_irq() to re-enable the queue. */
blk_stop_queue(q);
}
/*D:430 This is the "struct block_device_operations" we attach to the disk at
* the end of lguestblk_probe(). It doesn't seem to want much. */
static struct block_device_operations lguestblk_fops = {
.owner = THIS_MODULE,
};
/*D:425 Setting up a disk device seems to involve a lot of code. I'm not sure
* quite why. I do know that the IDE code sent two or three of the maintainers
* insane, perhaps this is the fringe of the same disease?
*
* As in the console code, the probe function gets handed the generic
* lguest_device from lguest_bus.c: */
static int lguestblk_probe(struct lguest_device *lgdev)
{
struct blockdev *bd;
int err;
int irqflags = IRQF_SHARED;
/* First we allocate our own "struct blockdev" and initialize the easy
* fields. */
bd = kmalloc(sizeof(*bd), GFP_KERNEL);
if (!bd)
return -ENOMEM;
spin_lock_init(&bd->lock);
bd->irq = lgdev_irq(lgdev);
bd->req = NULL;
bd->dma.used_len = 0;
bd->dma.len[0] = 0;
/* The descriptor in the lguest_devices array provided by the Host
* gives the Guest the physical page number of the device's page. */
bd->phys_addr = (lguest_devices[lgdev->index].pfn << PAGE_SHIFT);
/* We use lguest_map() to get a pointer to the device page */
bd->lb_page = lguest_map(bd->phys_addr, 1);
if (!bd->lb_page) {
err = -ENOMEM;
goto out_free_bd;
}
/* We need a major device number: 0 means "assign one dynamically". */
bd->major = register_blkdev(0, "lguestblk");
if (bd->major < 0) {
err = bd->major;
goto out_unmap;
}
/* This allocates a "struct gendisk" where we pack all the information
* about the disk which the rest of Linux sees. We ask for one minor
* number; I do wonder if we should be asking for more. */
bd->disk = alloc_disk(1);
if (!bd->disk) {
err
|