Line data Source code
1 : /* SPDX-License-Identifier: BSD-3-Clause
2 : * Copyright (C) 2018 Intel Corporation.
3 : * All rights reserved.
4 : * Copyright (c) 2021, 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
5 : */
6 :
7 : #include "vbdev_compress.h"
8 :
9 : #include "spdk/reduce.h"
10 : #include "spdk/stdinc.h"
11 : #include "spdk/rpc.h"
12 : #include "spdk/env.h"
13 : #include "spdk/endian.h"
14 : #include "spdk/string.h"
15 : #include "spdk/thread.h"
16 : #include "spdk/util.h"
17 : #include "spdk/bdev_module.h"
18 : #include "spdk/likely.h"
19 : #include "spdk/log.h"
20 : #include "spdk/accel.h"
21 :
22 : #include "spdk/accel_module.h"
23 :
24 : #define CHUNK_SIZE (1024 * 16)
25 : #define COMP_BDEV_NAME "compress"
26 : #define BACKING_IO_SZ (4 * 1024)
27 :
28 : /* This namespace UUID was generated using uuid_generate() method. */
29 : #define BDEV_COMPRESS_NAMESPACE_UUID "c3fad6da-832f-4cc0-9cdc-5c552b225e7b"
30 :
31 : struct vbdev_comp_delete_ctx {
32 : spdk_delete_compress_complete cb_fn;
33 : void *cb_arg;
34 : int cb_rc;
35 : struct spdk_thread *orig_thread;
36 : };
37 :
38 : /* List of virtual bdevs and associated info for each. */
39 : struct vbdev_compress {
40 : struct spdk_bdev *base_bdev; /* the thing we're attaching to */
41 : struct spdk_bdev_desc *base_desc; /* its descriptor we get from open */
42 : struct spdk_io_channel *base_ch; /* IO channel of base device */
43 : struct spdk_bdev comp_bdev; /* the compression virtual bdev */
44 : struct comp_io_channel *comp_ch; /* channel associated with this bdev */
45 : struct spdk_io_channel *accel_channel; /* to communicate with the accel framework */
46 : struct spdk_thread *reduce_thread;
47 : pthread_mutex_t reduce_lock;
48 : uint32_t ch_count;
49 : TAILQ_HEAD(, spdk_bdev_io) pending_comp_ios; /* outstanding operations to a comp library */
50 : struct spdk_poller *poller; /* completion poller */
51 : struct spdk_reduce_vol_params params; /* params for the reduce volume */
52 : struct spdk_reduce_backing_dev backing_dev; /* backing device info for the reduce volume */
53 : struct spdk_reduce_vol *vol; /* the reduce volume */
54 : struct vbdev_comp_delete_ctx *delete_ctx;
55 : bool orphaned; /* base bdev claimed but comp_bdev not registered */
56 : int reduce_errno;
57 : TAILQ_HEAD(, vbdev_comp_op) queued_comp_ops;
58 : TAILQ_ENTRY(vbdev_compress) link;
59 : struct spdk_thread *thread; /* thread where base device is opened */
60 : enum spdk_accel_comp_algo comp_algo; /* compression algorithm for compress bdev */
61 : uint32_t comp_level; /* compression algorithm level */
62 : bool init_failed; /* compress bdev initialization failed */
63 : };
64 : static TAILQ_HEAD(, vbdev_compress) g_vbdev_comp = TAILQ_HEAD_INITIALIZER(g_vbdev_comp);
65 :
66 : /* The comp vbdev channel struct. It is allocated and freed on my behalf by the io channel code.
67 : */
68 : struct comp_io_channel {
69 : struct spdk_io_channel_iter *iter; /* used with for_each_channel in reset */
70 : };
71 :
72 : /* Per I/O context for the compression vbdev. */
73 : struct comp_bdev_io {
74 : struct comp_io_channel *comp_ch; /* used in completion handling */
75 : struct vbdev_compress *comp_bdev; /* vbdev associated with this IO */
76 : struct spdk_bdev_io_wait_entry bdev_io_wait; /* for bdev_io_wait */
77 : struct spdk_bdev_io *orig_io; /* the original IO */
78 : int status; /* save for completion on orig thread */
79 : };
80 :
81 : static void vbdev_compress_examine(struct spdk_bdev *bdev);
82 : static int vbdev_compress_claim(struct vbdev_compress *comp_bdev);
83 : struct vbdev_compress *_prepare_for_load_init(struct spdk_bdev_desc *bdev_desc, uint32_t lb_size,
84 : uint8_t comp_algo, uint32_t comp_level);
85 : static void vbdev_compress_submit_request(struct spdk_io_channel *ch, struct spdk_bdev_io *bdev_io);
86 : static void comp_bdev_ch_destroy_cb(void *io_device, void *ctx_buf);
87 : static void vbdev_compress_delete_done(void *cb_arg, int bdeverrno);
88 : static void _comp_reduce_resubmit_backing_io(void *_backing_io);
89 :
90 : /* for completing rw requests on the orig IO thread. */
91 : static void
92 14 : _reduce_rw_blocks_cb(void *arg)
93 : {
94 14 : struct comp_bdev_io *io_ctx = arg;
95 :
96 14 : if (spdk_likely(io_ctx->status == 0)) {
97 7 : spdk_bdev_io_complete(io_ctx->orig_io, SPDK_BDEV_IO_STATUS_SUCCESS);
98 7 : } else if (io_ctx->status == -ENOMEM) {
99 0 : spdk_bdev_io_complete(io_ctx->orig_io, SPDK_BDEV_IO_STATUS_NOMEM);
100 : } else {
101 7 : SPDK_ERRLOG("Failed to execute reduce api. %s\n", spdk_strerror(-io_ctx->status));
102 7 : spdk_bdev_io_complete(io_ctx->orig_io, SPDK_BDEV_IO_STATUS_FAILED);
103 : }
104 14 : }
105 :
106 : /* Completion callback for r/w that were issued via reducelib. */
107 : static void
108 14 : reduce_rw_blocks_cb(void *arg, int reduce_errno)
109 : {
110 14 : struct spdk_bdev_io *bdev_io = arg;
111 14 : struct comp_bdev_io *io_ctx = (struct comp_bdev_io *)bdev_io->driver_ctx;
112 14 : struct spdk_io_channel *ch = spdk_io_channel_from_ctx(io_ctx->comp_ch);
113 : struct spdk_thread *orig_thread;
114 :
115 : /* TODO: need to decide which error codes are bdev_io success vs failure;
116 : * example examine calls reading metadata */
117 :
118 14 : io_ctx->status = reduce_errno;
119 :
120 : /* Send this request to the orig IO thread. */
121 14 : orig_thread = spdk_io_channel_get_thread(ch);
122 :
123 14 : spdk_thread_exec_msg(orig_thread, _reduce_rw_blocks_cb, io_ctx);
124 14 : }
125 :
126 : static int
127 0 : _compress_operation(struct spdk_reduce_backing_dev *backing_dev, struct iovec *src_iovs,
128 : int src_iovcnt, struct iovec *dst_iovs,
129 : int dst_iovcnt, bool compress, void *cb_arg)
130 : {
131 0 : struct spdk_reduce_vol_cb_args *reduce_cb_arg = cb_arg;
132 0 : struct vbdev_compress *comp_bdev = SPDK_CONTAINEROF(backing_dev, struct vbdev_compress,
133 : backing_dev);
134 : int rc;
135 :
136 0 : if (compress) {
137 0 : assert(dst_iovcnt == 1);
138 0 : rc = spdk_accel_submit_compress_ext(comp_bdev->accel_channel, dst_iovs[0].iov_base,
139 : dst_iovs[0].iov_len, src_iovs, src_iovcnt,
140 : comp_bdev->comp_algo, comp_bdev->comp_level,
141 : &reduce_cb_arg->output_size, reduce_cb_arg->cb_fn,
142 : reduce_cb_arg->cb_arg);
143 : } else {
144 0 : rc = spdk_accel_submit_decompress_ext(comp_bdev->accel_channel, dst_iovs, dst_iovcnt,
145 : src_iovs, src_iovcnt, comp_bdev->comp_algo,
146 : &reduce_cb_arg->output_size, reduce_cb_arg->cb_fn,
147 : reduce_cb_arg->cb_arg);
148 : }
149 :
150 0 : return rc;
151 : }
152 :
153 : /* Entry point for reduce lib to issue a compress operation. */
154 : static void
155 0 : _comp_reduce_compress(struct spdk_reduce_backing_dev *dev,
156 : struct iovec *src_iovs, int src_iovcnt,
157 : struct iovec *dst_iovs, int dst_iovcnt,
158 : struct spdk_reduce_vol_cb_args *cb_arg)
159 : {
160 : int rc;
161 :
162 0 : rc = _compress_operation(dev, src_iovs, src_iovcnt, dst_iovs, dst_iovcnt, true, cb_arg);
163 0 : if (rc) {
164 0 : SPDK_ERRLOG("with compress operation code %d (%s)\n", rc, spdk_strerror(-rc));
165 0 : cb_arg->cb_fn(cb_arg->cb_arg, rc);
166 : }
167 0 : }
168 :
169 : /* Entry point for reduce lib to issue a decompress operation. */
170 : static void
171 0 : _comp_reduce_decompress(struct spdk_reduce_backing_dev *dev,
172 : struct iovec *src_iovs, int src_iovcnt,
173 : struct iovec *dst_iovs, int dst_iovcnt,
174 : struct spdk_reduce_vol_cb_args *cb_arg)
175 : {
176 : int rc;
177 :
178 0 : rc = _compress_operation(dev, src_iovs, src_iovcnt, dst_iovs, dst_iovcnt, false, cb_arg);
179 0 : if (rc) {
180 0 : SPDK_ERRLOG("with decompress operation code %d (%s)\n", rc, spdk_strerror(-rc));
181 0 : cb_arg->cb_fn(cb_arg->cb_arg, rc);
182 : }
183 0 : }
184 :
185 : static void
186 2 : _comp_submit_write(void *ctx)
187 : {
188 2 : struct spdk_bdev_io *bdev_io = ctx;
189 2 : struct vbdev_compress *comp_bdev = SPDK_CONTAINEROF(bdev_io->bdev, struct vbdev_compress,
190 : comp_bdev);
191 :
192 2 : spdk_reduce_vol_writev(comp_bdev->vol, bdev_io->u.bdev.iovs, bdev_io->u.bdev.iovcnt,
193 : bdev_io->u.bdev.offset_blocks, bdev_io->u.bdev.num_blocks,
194 : reduce_rw_blocks_cb, bdev_io);
195 2 : }
196 :
197 : static void
198 2 : _comp_submit_read(void *ctx)
199 : {
200 2 : struct spdk_bdev_io *bdev_io = ctx;
201 2 : struct vbdev_compress *comp_bdev = SPDK_CONTAINEROF(bdev_io->bdev, struct vbdev_compress,
202 : comp_bdev);
203 :
204 2 : spdk_reduce_vol_readv(comp_bdev->vol, bdev_io->u.bdev.iovs, bdev_io->u.bdev.iovcnt,
205 : bdev_io->u.bdev.offset_blocks, bdev_io->u.bdev.num_blocks,
206 : reduce_rw_blocks_cb, bdev_io);
207 2 : }
208 :
209 :
210 : /* Callback for getting a buf from the bdev pool in the event that the caller passed
211 : * in NULL, we need to own the buffer so it doesn't get freed by another vbdev module
212 : * beneath us before we're done with it.
213 : */
214 : static void
215 2 : comp_read_get_buf_cb(struct spdk_io_channel *ch, struct spdk_bdev_io *bdev_io, bool success)
216 : {
217 2 : struct vbdev_compress *comp_bdev = SPDK_CONTAINEROF(bdev_io->bdev, struct vbdev_compress,
218 : comp_bdev);
219 :
220 2 : if (spdk_unlikely(!success)) {
221 0 : SPDK_ERRLOG("Failed to get data buffer\n");
222 0 : reduce_rw_blocks_cb(bdev_io, -ENOMEM);
223 0 : return;
224 : }
225 :
226 2 : spdk_thread_exec_msg(comp_bdev->reduce_thread, _comp_submit_read, bdev_io);
227 : }
228 :
229 : struct partial_chunk_info {
230 : uint64_t chunk_idx;
231 : uint64_t block_offset;
232 : uint64_t block_length;
233 : };
234 :
235 : /*
236 : * It's a structure used to hold information needed during the execution of an unmap operation.
237 : */
238 : struct compress_unmap_split_ctx {
239 : struct spdk_bdev_io *bdev_io;
240 : int32_t status;
241 : uint32_t logical_blocks_per_chunk;
242 : /* The first chunk that can be fully covered by the unmap bdevio interval */
243 : uint64_t full_chunk_idx_b;
244 : /* The last chunk that can be fully covered by the unmap bdevio interval */
245 : uint64_t full_chunk_idx_e;
246 : uint64_t num_full_chunks;
247 : uint64_t num_full_chunks_consumed;
248 : uint32_t num_partial_chunks;
249 : uint32_t num_partial_chunks_consumed;
250 : /* Used to hold the partial chunk information. There will only be less than or equal to two,
251 : because chunks that cannot be fully covered will only appear at the beginning or end or both two. */
252 : struct partial_chunk_info partial_chunk_info[2];
253 : };
254 :
255 : static void _comp_unmap_subcmd_done_cb(void *ctx, int error);
256 :
257 : /*
258 : * This function processes the unmap operation for both full and partial chunks in a
259 : * compressed block device. It iteratively submits unmap requests until all the chunks
260 : * have been unmapped or an error occurs.
261 : */
262 : static void
263 20 : _comp_submit_unmap_split(void *ctx)
264 : {
265 20 : struct compress_unmap_split_ctx *split_ctx = ctx;
266 20 : struct spdk_bdev_io *bdev_io = split_ctx->bdev_io;
267 20 : struct vbdev_compress *comp_bdev = SPDK_CONTAINEROF(bdev_io->bdev, struct vbdev_compress,
268 : comp_bdev);
269 20 : struct partial_chunk_info *partial_chunk = NULL;
270 20 : uint64_t chunk_idx = 0;
271 20 : uint64_t block_offset = 0;
272 20 : uint64_t block_length = 0;
273 :
274 20 : if (split_ctx->status != 0 ||
275 15 : (split_ctx->num_full_chunks_consumed == split_ctx->num_full_chunks &&
276 13 : split_ctx->num_partial_chunks_consumed == split_ctx->num_partial_chunks)) {
277 10 : reduce_rw_blocks_cb(bdev_io, split_ctx->status);
278 10 : free(split_ctx);
279 10 : return;
280 : }
281 :
282 10 : if (split_ctx->num_full_chunks_consumed < split_ctx->num_full_chunks) {
283 2 : chunk_idx = split_ctx->full_chunk_idx_b + split_ctx->num_full_chunks_consumed;
284 2 : block_offset = chunk_idx * split_ctx->logical_blocks_per_chunk;
285 2 : block_length = split_ctx->logical_blocks_per_chunk;
286 :
287 2 : split_ctx->num_full_chunks_consumed++;
288 2 : spdk_reduce_vol_unmap(comp_bdev->vol,
289 : block_offset, block_length,
290 : _comp_unmap_subcmd_done_cb, split_ctx);
291 8 : } else if (split_ctx->num_partial_chunks_consumed < split_ctx->num_partial_chunks) {
292 8 : partial_chunk = &split_ctx->partial_chunk_info[split_ctx->num_partial_chunks_consumed];
293 8 : block_offset = partial_chunk->chunk_idx * split_ctx->logical_blocks_per_chunk +
294 8 : partial_chunk->block_offset;
295 8 : block_length = partial_chunk->block_length;
296 :
297 8 : split_ctx->num_partial_chunks_consumed++;
298 8 : spdk_reduce_vol_unmap(comp_bdev->vol,
299 : block_offset, block_length,
300 : _comp_unmap_subcmd_done_cb, split_ctx);
301 : } else {
302 0 : assert(false);
303 : }
304 : }
305 :
306 : /*
307 : * When mkfs or fstrim, large unmap requests may be generated.
308 : * Large request will be split into multiple subcmds and processed recursively.
309 : * Run too many subcmds recursively may cause stack overflow or monopolize the thread,
310 : * delaying other tasks. To avoid this, next subcmd need to be processed asynchronously
311 : * by 'spdk_thread_send_msg'.
312 : */
313 : static void
314 10 : _comp_unmap_subcmd_done_cb(void *ctx, int error)
315 : {
316 10 : struct compress_unmap_split_ctx *split_ctx = ctx;
317 :
318 10 : split_ctx->status = error;
319 10 : spdk_thread_send_msg(spdk_get_thread(), _comp_submit_unmap_split, split_ctx);
320 10 : }
321 :
322 : /*
323 : * This function splits the unmap operation into full and partial chunks based on the
324 : * block range specified in the 'spdk_bdev_io' structure. It calculates the start and end
325 : * chunks, as well as any partial chunks at the beginning or end of the range, and prepares
326 : * a context (compress_unmap_split_ctx) to handle these chunks. The unmap operation is
327 : * then submitted for processing through '_comp_submit_unmap_split'.
328 : * some cases to handle:
329 : * 1. start and end chunks are different
330 : * 1.1 start and end chunks are full
331 : * 1.2 start and end chunks are partial
332 : * 1.3 start or end chunk is full and the other is partial
333 : * 2. start and end chunks are the same
334 : * 2.1 full
335 : * 2.2 partial
336 : */
337 : static void
338 10 : _comp_submit_unmap(void *ctx)
339 : {
340 10 : struct spdk_bdev_io *bdev_io = ctx;
341 10 : struct vbdev_compress *comp_bdev = SPDK_CONTAINEROF(bdev_io->bdev, struct vbdev_compress,
342 : comp_bdev);
343 10 : const struct spdk_reduce_vol_params *vol_params = spdk_reduce_vol_get_params(comp_bdev->vol);
344 : struct compress_unmap_split_ctx *split_ctx;
345 : struct partial_chunk_info *partial_chunk;
346 : uint32_t logical_blocks_per_chunk;
347 : uint64_t start_chunk, end_chunk, start_offset, end_tail;
348 :
349 10 : logical_blocks_per_chunk = vol_params->chunk_size / vol_params->logical_block_size;
350 10 : start_chunk = bdev_io->u.bdev.offset_blocks / logical_blocks_per_chunk;
351 10 : end_chunk = (bdev_io->u.bdev.offset_blocks + bdev_io->u.bdev.num_blocks - 1) /
352 : logical_blocks_per_chunk;
353 10 : start_offset = bdev_io->u.bdev.offset_blocks % logical_blocks_per_chunk;
354 10 : end_tail = (bdev_io->u.bdev.offset_blocks + bdev_io->u.bdev.num_blocks) %
355 : logical_blocks_per_chunk;
356 :
357 10 : split_ctx = calloc(1, sizeof(struct compress_unmap_split_ctx));
358 10 : if (split_ctx == NULL) {
359 0 : reduce_rw_blocks_cb(bdev_io, -ENOMEM);
360 0 : return;
361 : }
362 10 : partial_chunk = split_ctx->partial_chunk_info;
363 10 : split_ctx->bdev_io = bdev_io;
364 10 : split_ctx->logical_blocks_per_chunk = logical_blocks_per_chunk;
365 :
366 10 : if (start_chunk < end_chunk) {
367 0 : if (start_offset != 0) {
368 0 : partial_chunk[split_ctx->num_partial_chunks].chunk_idx = start_chunk;
369 0 : partial_chunk[split_ctx->num_partial_chunks].block_offset = start_offset;
370 0 : partial_chunk[split_ctx->num_partial_chunks].block_length = logical_blocks_per_chunk
371 0 : - start_offset;
372 0 : split_ctx->num_partial_chunks++;
373 0 : split_ctx->full_chunk_idx_b = start_chunk + 1;
374 : } else {
375 0 : split_ctx->full_chunk_idx_b = start_chunk;
376 : }
377 :
378 0 : if (end_tail != 0) {
379 0 : partial_chunk[split_ctx->num_partial_chunks].chunk_idx = end_chunk;
380 0 : partial_chunk[split_ctx->num_partial_chunks].block_offset = 0;
381 0 : partial_chunk[split_ctx->num_partial_chunks].block_length = end_tail;
382 0 : split_ctx->num_partial_chunks++;
383 0 : split_ctx->full_chunk_idx_e = end_chunk - 1;
384 : } else {
385 0 : split_ctx->full_chunk_idx_e = end_chunk;
386 : }
387 :
388 0 : split_ctx->num_full_chunks = end_chunk - start_chunk + 1 - split_ctx->num_partial_chunks;
389 :
390 0 : if (split_ctx->num_full_chunks) {
391 0 : assert(split_ctx->full_chunk_idx_b != UINT64_MAX && split_ctx->full_chunk_idx_e != UINT64_MAX);
392 0 : assert(split_ctx->full_chunk_idx_e - split_ctx->full_chunk_idx_b + 1 == split_ctx->num_full_chunks);
393 : } else {
394 0 : assert(split_ctx->full_chunk_idx_b - split_ctx->full_chunk_idx_e == 1);
395 : }
396 10 : } else if (start_offset != 0 || end_tail != 0) {
397 8 : partial_chunk[0].chunk_idx = start_chunk;
398 8 : partial_chunk[0].block_offset = start_offset;
399 8 : partial_chunk[0].block_length =
400 8 : bdev_io->u.bdev.num_blocks;
401 8 : split_ctx->num_partial_chunks = 1;
402 : } else {
403 2 : split_ctx->full_chunk_idx_b = start_chunk;
404 2 : split_ctx->full_chunk_idx_e = end_chunk;
405 2 : split_ctx->num_full_chunks = 1;
406 : }
407 10 : assert(split_ctx->num_partial_chunks <= SPDK_COUNTOF(split_ctx->partial_chunk_info));
408 :
409 10 : _comp_submit_unmap_split(split_ctx);
410 : }
411 :
412 : /* Called when someone above submits IO to this vbdev. */
413 : static void
414 14 : vbdev_compress_submit_request(struct spdk_io_channel *ch, struct spdk_bdev_io *bdev_io)
415 : {
416 14 : struct comp_bdev_io *io_ctx = (struct comp_bdev_io *)bdev_io->driver_ctx;
417 14 : struct vbdev_compress *comp_bdev = SPDK_CONTAINEROF(bdev_io->bdev, struct vbdev_compress,
418 : comp_bdev);
419 14 : struct comp_io_channel *comp_ch = spdk_io_channel_get_ctx(ch);
420 :
421 14 : memset(io_ctx, 0, sizeof(struct comp_bdev_io));
422 14 : io_ctx->comp_bdev = comp_bdev;
423 14 : io_ctx->comp_ch = comp_ch;
424 14 : io_ctx->orig_io = bdev_io;
425 :
426 14 : switch (bdev_io->type) {
427 2 : case SPDK_BDEV_IO_TYPE_READ:
428 2 : spdk_bdev_io_get_buf(bdev_io, comp_read_get_buf_cb,
429 2 : bdev_io->u.bdev.num_blocks * bdev_io->bdev->blocklen);
430 2 : return;
431 2 : case SPDK_BDEV_IO_TYPE_WRITE:
432 2 : spdk_thread_exec_msg(comp_bdev->reduce_thread, _comp_submit_write, bdev_io);
433 2 : return;
434 10 : case SPDK_BDEV_IO_TYPE_UNMAP:
435 10 : spdk_thread_exec_msg(comp_bdev->reduce_thread, _comp_submit_unmap, bdev_io);
436 10 : return;
437 : /* TODO support RESET in future patch in the series */
438 0 : case SPDK_BDEV_IO_TYPE_RESET:
439 : case SPDK_BDEV_IO_TYPE_WRITE_ZEROES:
440 : case SPDK_BDEV_IO_TYPE_FLUSH:
441 : default:
442 0 : SPDK_ERRLOG("Unknown I/O type %d\n", bdev_io->type);
443 0 : spdk_bdev_io_complete(io_ctx->orig_io, SPDK_BDEV_IO_STATUS_FAILED);
444 0 : break;
445 : }
446 : }
447 :
448 : static bool
449 12 : vbdev_compress_io_type_supported(void *ctx, enum spdk_bdev_io_type io_type)
450 : {
451 12 : struct vbdev_compress *comp_bdev = (struct vbdev_compress *)ctx;
452 :
453 12 : switch (io_type) {
454 4 : case SPDK_BDEV_IO_TYPE_READ:
455 : case SPDK_BDEV_IO_TYPE_WRITE:
456 4 : return spdk_bdev_io_type_supported(comp_bdev->base_bdev, io_type);
457 2 : case SPDK_BDEV_IO_TYPE_UNMAP:
458 2 : return true;
459 6 : case SPDK_BDEV_IO_TYPE_RESET:
460 : case SPDK_BDEV_IO_TYPE_FLUSH:
461 : case SPDK_BDEV_IO_TYPE_WRITE_ZEROES:
462 : default:
463 6 : return false;
464 : }
465 : }
466 :
467 : /* Callback for unregistering the IO device. */
468 : static void
469 0 : _device_unregister_cb(void *io_device)
470 : {
471 0 : struct vbdev_compress *comp_bdev = io_device;
472 :
473 : /* Done with this comp_bdev. */
474 0 : pthread_mutex_destroy(&comp_bdev->reduce_lock);
475 0 : free(comp_bdev->comp_bdev.name);
476 0 : free(comp_bdev);
477 0 : }
478 :
479 : static void
480 0 : _vbdev_compress_destruct_cb(void *ctx)
481 : {
482 0 : struct vbdev_compress *comp_bdev = ctx;
483 :
484 : /* Close the underlying bdev on its same opened thread. */
485 0 : spdk_bdev_close(comp_bdev->base_desc);
486 0 : comp_bdev->vol = NULL;
487 0 : if (comp_bdev->init_failed) {
488 0 : free(comp_bdev);
489 0 : return;
490 : }
491 :
492 0 : TAILQ_REMOVE(&g_vbdev_comp, comp_bdev, link);
493 0 : spdk_bdev_module_release_bdev(comp_bdev->base_bdev);
494 :
495 0 : if (comp_bdev->orphaned == false) {
496 0 : spdk_io_device_unregister(comp_bdev, _device_unregister_cb);
497 : } else {
498 0 : vbdev_compress_delete_done(comp_bdev->delete_ctx, 0);
499 0 : _device_unregister_cb(comp_bdev);
500 : }
501 : }
502 :
503 : static void
504 0 : vbdev_compress_destruct_cb(void *cb_arg, int reduce_errno)
505 : {
506 0 : struct vbdev_compress *comp_bdev = (struct vbdev_compress *)cb_arg;
507 :
508 0 : if (reduce_errno) {
509 0 : SPDK_ERRLOG("number %d\n", reduce_errno);
510 : } else {
511 0 : if (comp_bdev->thread && comp_bdev->thread != spdk_get_thread()) {
512 0 : spdk_thread_send_msg(comp_bdev->thread,
513 : _vbdev_compress_destruct_cb, comp_bdev);
514 : } else {
515 0 : _vbdev_compress_destruct_cb(comp_bdev);
516 : }
517 : }
518 0 : }
519 :
520 : static void
521 0 : _reduce_destroy_cb(void *ctx, int reduce_errno)
522 : {
523 0 : struct vbdev_compress *comp_bdev = (struct vbdev_compress *)ctx;
524 :
525 0 : if (reduce_errno) {
526 0 : SPDK_ERRLOG("number %d\n", reduce_errno);
527 : }
528 :
529 0 : comp_bdev->vol = NULL;
530 0 : spdk_put_io_channel(comp_bdev->base_ch);
531 0 : if (comp_bdev->init_failed || comp_bdev->orphaned) {
532 0 : vbdev_compress_destruct_cb((void *)comp_bdev, 0);
533 : } else {
534 0 : spdk_bdev_unregister(&comp_bdev->comp_bdev, vbdev_compress_delete_done,
535 0 : comp_bdev->delete_ctx);
536 : }
537 :
538 0 : }
539 :
540 : static void
541 0 : _delete_vol_unload_cb(void *ctx)
542 : {
543 0 : struct vbdev_compress *comp_bdev = ctx;
544 :
545 : /* FIXME: Assert if these conditions are not satisfied for now. */
546 0 : assert(!comp_bdev->reduce_thread ||
547 : comp_bdev->reduce_thread == spdk_get_thread());
548 :
549 : /* reducelib needs a channel to comm with the backing device */
550 0 : comp_bdev->base_ch = spdk_bdev_get_io_channel(comp_bdev->base_desc);
551 :
552 : /* Clean the device before we free our resources. */
553 0 : spdk_reduce_vol_destroy(&comp_bdev->backing_dev, _reduce_destroy_cb, comp_bdev);
554 0 : }
555 :
556 : /* Called by reduceLib after performing unload vol actions */
557 : static void
558 0 : delete_vol_unload_cb(void *cb_arg, int reduce_errno)
559 : {
560 0 : struct vbdev_compress *comp_bdev = (struct vbdev_compress *)cb_arg;
561 :
562 0 : if (reduce_errno) {
563 0 : SPDK_ERRLOG("Failed to unload vol, error %s\n", spdk_strerror(-reduce_errno));
564 0 : vbdev_compress_delete_done(comp_bdev->delete_ctx, reduce_errno);
565 0 : return;
566 : }
567 :
568 0 : pthread_mutex_lock(&comp_bdev->reduce_lock);
569 0 : if (comp_bdev->reduce_thread && comp_bdev->reduce_thread != spdk_get_thread()) {
570 0 : spdk_thread_send_msg(comp_bdev->reduce_thread,
571 : _delete_vol_unload_cb, comp_bdev);
572 0 : pthread_mutex_unlock(&comp_bdev->reduce_lock);
573 : } else {
574 0 : pthread_mutex_unlock(&comp_bdev->reduce_lock);
575 :
576 0 : _delete_vol_unload_cb(comp_bdev);
577 : }
578 : }
579 :
580 : const char *
581 0 : compress_get_name(const struct vbdev_compress *comp_bdev)
582 : {
583 0 : return comp_bdev->comp_bdev.name;
584 : }
585 :
586 : struct vbdev_compress *
587 0 : compress_bdev_first(void)
588 : {
589 : struct vbdev_compress *comp_bdev;
590 :
591 0 : comp_bdev = TAILQ_FIRST(&g_vbdev_comp);
592 :
593 0 : return comp_bdev;
594 : }
595 :
596 : struct vbdev_compress *
597 0 : compress_bdev_next(struct vbdev_compress *prev)
598 : {
599 : struct vbdev_compress *comp_bdev;
600 :
601 0 : comp_bdev = TAILQ_NEXT(prev, link);
602 :
603 0 : return comp_bdev;
604 : }
605 :
606 : bool
607 0 : compress_has_orphan(const char *name)
608 : {
609 : struct vbdev_compress *comp_bdev;
610 :
611 0 : TAILQ_FOREACH(comp_bdev, &g_vbdev_comp, link) {
612 0 : if (comp_bdev->orphaned && strcmp(name, comp_bdev->comp_bdev.name) == 0) {
613 0 : return true;
614 : }
615 : }
616 0 : return false;
617 : }
618 :
619 : /* Called after we've unregistered following a hot remove callback.
620 : * Our finish entry point will be called next.
621 : */
622 : static int
623 0 : vbdev_compress_destruct(void *ctx)
624 : {
625 0 : struct vbdev_compress *comp_bdev = (struct vbdev_compress *)ctx;
626 :
627 0 : if (comp_bdev->vol != NULL) {
628 : /* Tell reducelib that we're done with this volume. */
629 0 : spdk_reduce_vol_unload(comp_bdev->vol, vbdev_compress_destruct_cb, comp_bdev);
630 : } else {
631 0 : vbdev_compress_destruct_cb(comp_bdev, 0);
632 : }
633 :
634 0 : return 0;
635 : }
636 :
637 : /* We supplied this as an entry point for upper layers who want to communicate to this
638 : * bdev. This is how they get a channel.
639 : */
640 : static struct spdk_io_channel *
641 0 : vbdev_compress_get_io_channel(void *ctx)
642 : {
643 0 : struct vbdev_compress *comp_bdev = (struct vbdev_compress *)ctx;
644 :
645 : /* The IO channel code will allocate a channel for us which consists of
646 : * the SPDK channel structure plus the size of our comp_io_channel struct
647 : * that we passed in when we registered our IO device. It will then call
648 : * our channel create callback to populate any elements that we need to
649 : * update.
650 : */
651 0 : return spdk_get_io_channel(comp_bdev);
652 : }
653 :
654 : /* This is the output for bdev_get_bdevs() for this vbdev */
655 : static int
656 0 : vbdev_compress_dump_info_json(void *ctx, struct spdk_json_write_ctx *w)
657 : {
658 0 : struct vbdev_compress *comp_bdev = (struct vbdev_compress *)ctx;
659 : const struct spdk_reduce_vol_info *vol_info;
660 0 : char *comp_algo = NULL;
661 :
662 0 : if (comp_bdev->params.comp_algo == SPDK_ACCEL_COMP_ALGO_LZ4) {
663 0 : comp_algo = "lz4";
664 0 : } else if (comp_bdev->params.comp_algo == SPDK_ACCEL_COMP_ALGO_DEFLATE) {
665 0 : comp_algo = "deflate";
666 : } else {
667 0 : assert(false);
668 : }
669 :
670 0 : spdk_json_write_name(w, "compress");
671 0 : spdk_json_write_object_begin(w);
672 0 : spdk_json_write_named_string(w, "name", spdk_bdev_get_name(&comp_bdev->comp_bdev));
673 0 : spdk_json_write_named_string(w, "base_bdev_name", spdk_bdev_get_name(comp_bdev->base_bdev));
674 0 : spdk_json_write_named_string(w, "pm_path", spdk_reduce_vol_get_pm_path(comp_bdev->vol));
675 0 : spdk_json_write_named_string(w, "comp_algo", comp_algo);
676 0 : spdk_json_write_named_uint32(w, "comp_level", comp_bdev->params.comp_level);
677 0 : spdk_json_write_named_uint32(w, "chunk_size", comp_bdev->params.chunk_size);
678 0 : spdk_json_write_named_uint32(w, "backing_io_unit_size", comp_bdev->params.backing_io_unit_size);
679 0 : vol_info = spdk_reduce_vol_get_info(comp_bdev->vol);
680 0 : spdk_json_write_named_uint64(w, "allocated_io_units", vol_info->allocated_io_units);
681 0 : spdk_json_write_object_end(w);
682 :
683 0 : return 0;
684 : }
685 :
686 : static int
687 0 : vbdev_compress_config_json(struct spdk_json_write_ctx *w)
688 : {
689 : /* Nothing to dump as compress bdev configuration is saved on physical device. */
690 0 : return 0;
691 : }
692 :
693 : struct vbdev_init_reduce_ctx {
694 : struct vbdev_compress *comp_bdev;
695 : int status;
696 : bdev_compress_create_cb cb_fn;
697 : void *cb_ctx;
698 : };
699 :
700 : static void
701 0 : _cleanup_vol_unload_cb(void *ctx)
702 : {
703 0 : struct vbdev_compress *comp_bdev = ctx;
704 :
705 0 : assert(!comp_bdev->reduce_thread ||
706 : comp_bdev->reduce_thread == spdk_get_thread());
707 :
708 0 : comp_bdev->base_ch = spdk_bdev_get_io_channel(comp_bdev->base_desc);
709 :
710 0 : spdk_reduce_vol_destroy(&comp_bdev->backing_dev, _reduce_destroy_cb, comp_bdev);
711 0 : }
712 :
713 : static void
714 0 : init_vol_unload_cb(void *ctx, int reduce_errno)
715 : {
716 0 : struct vbdev_compress *comp_bdev = (struct vbdev_compress *)ctx;
717 :
718 0 : if (reduce_errno) {
719 0 : SPDK_ERRLOG("Failed to unload vol, error %s\n", spdk_strerror(-reduce_errno));
720 : }
721 :
722 0 : pthread_mutex_lock(&comp_bdev->reduce_lock);
723 0 : if (comp_bdev->reduce_thread && comp_bdev->reduce_thread != spdk_get_thread()) {
724 0 : spdk_thread_send_msg(comp_bdev->reduce_thread,
725 : _cleanup_vol_unload_cb, comp_bdev);
726 0 : pthread_mutex_unlock(&comp_bdev->reduce_lock);
727 : } else {
728 0 : pthread_mutex_unlock(&comp_bdev->reduce_lock);
729 :
730 0 : _cleanup_vol_unload_cb(comp_bdev);
731 : }
732 0 : }
733 :
734 : static void
735 0 : _vbdev_reduce_init_cb(void *ctx)
736 : {
737 0 : struct vbdev_init_reduce_ctx *init_ctx = ctx;
738 0 : struct vbdev_compress *comp_bdev = init_ctx->comp_bdev;
739 0 : int rc = init_ctx->status;
740 :
741 0 : assert(comp_bdev->base_desc != NULL);
742 :
743 : /* We're done with metadata operations */
744 0 : spdk_put_io_channel(comp_bdev->base_ch);
745 :
746 0 : if (rc != 0) {
747 0 : goto err;
748 : }
749 :
750 0 : assert(comp_bdev->vol != NULL);
751 :
752 0 : rc = vbdev_compress_claim(comp_bdev);
753 0 : if (rc != 0) {
754 0 : comp_bdev->init_failed = true;
755 0 : spdk_reduce_vol_unload(comp_bdev->vol, init_vol_unload_cb, comp_bdev);
756 : }
757 :
758 0 : init_ctx->cb_fn(init_ctx->cb_ctx, rc);
759 0 : free(init_ctx);
760 0 : return;
761 :
762 0 : err:
763 0 : init_ctx->cb_fn(init_ctx->cb_ctx, rc);
764 : /* Close the underlying bdev on its same opened thread. */
765 0 : spdk_bdev_close(comp_bdev->base_desc);
766 0 : free(comp_bdev);
767 0 : free(init_ctx);
768 : }
769 :
770 : /* Callback from reduce for when init is complete. We'll pass the vbdev_comp struct
771 : * used for initial metadata operations to claim where it will be further filled out
772 : * and added to the global list.
773 : */
774 : static void
775 0 : vbdev_reduce_init_cb(void *cb_arg, struct spdk_reduce_vol *vol, int reduce_errno)
776 : {
777 0 : struct vbdev_init_reduce_ctx *init_ctx = cb_arg;
778 0 : struct vbdev_compress *comp_bdev = init_ctx->comp_bdev;
779 :
780 0 : if (reduce_errno == 0) {
781 0 : comp_bdev->vol = vol;
782 : } else {
783 0 : SPDK_ERRLOG("for vol %s, error %s\n",
784 : spdk_bdev_get_name(comp_bdev->base_bdev), spdk_strerror(-reduce_errno));
785 : }
786 :
787 0 : init_ctx->status = reduce_errno;
788 :
789 0 : if (comp_bdev->thread && comp_bdev->thread != spdk_get_thread()) {
790 0 : spdk_thread_send_msg(comp_bdev->thread, _vbdev_reduce_init_cb, init_ctx);
791 : } else {
792 0 : _vbdev_reduce_init_cb(init_ctx);
793 : }
794 0 : }
795 :
796 : /* Callback for the function used by reduceLib to perform IO to/from the backing device. We just
797 : * call the callback provided by reduceLib when it called the read/write/unmap function and
798 : * free the bdev_io.
799 : */
800 : static void
801 0 : comp_reduce_io_cb(struct spdk_bdev_io *bdev_io, bool success, void *arg)
802 : {
803 0 : struct spdk_reduce_vol_cb_args *cb_args = arg;
804 : int reduce_errno;
805 :
806 0 : if (success) {
807 0 : reduce_errno = 0;
808 : } else {
809 0 : reduce_errno = -EIO;
810 : }
811 0 : spdk_bdev_free_io(bdev_io);
812 0 : cb_args->cb_fn(cb_args->cb_arg, reduce_errno);
813 0 : }
814 :
815 : static void
816 0 : _comp_backing_bdev_queue_io_wait(struct vbdev_compress *comp_bdev,
817 : struct spdk_reduce_backing_io *backing_io)
818 : {
819 : struct spdk_bdev_io_wait_entry *waitq_entry;
820 : int rc;
821 :
822 0 : waitq_entry = (struct spdk_bdev_io_wait_entry *) &backing_io->user_ctx;
823 0 : waitq_entry->bdev = spdk_bdev_desc_get_bdev(comp_bdev->base_desc);
824 0 : waitq_entry->cb_fn = _comp_reduce_resubmit_backing_io;
825 0 : waitq_entry->cb_arg = backing_io;
826 :
827 0 : rc = spdk_bdev_queue_io_wait(waitq_entry->bdev, comp_bdev->base_ch, waitq_entry);
828 0 : if (rc) {
829 0 : SPDK_ERRLOG("Queue io failed in _comp_backing_bdev_queue_io_wait, rc=%d.\n", rc);
830 0 : assert(false);
831 : backing_io->backing_cb_args->cb_fn(backing_io->backing_cb_args->cb_arg, rc);
832 : }
833 0 : }
834 :
835 : static void
836 0 : _comp_backing_bdev_read(struct spdk_reduce_backing_io *backing_io)
837 : {
838 0 : struct spdk_reduce_vol_cb_args *backing_cb_args = backing_io->backing_cb_args;
839 0 : struct vbdev_compress *comp_bdev = SPDK_CONTAINEROF(backing_io->dev, struct vbdev_compress,
840 : backing_dev);
841 : int rc;
842 :
843 0 : rc = spdk_bdev_readv_blocks(comp_bdev->base_desc, comp_bdev->base_ch,
844 0 : backing_io->iov, backing_io->iovcnt,
845 0 : backing_io->lba, backing_io->lba_count,
846 : comp_reduce_io_cb,
847 : backing_cb_args);
848 :
849 0 : if (rc) {
850 0 : if (rc == -ENOMEM) {
851 0 : _comp_backing_bdev_queue_io_wait(comp_bdev, backing_io);
852 0 : return;
853 : } else {
854 0 : SPDK_ERRLOG("submitting readv request, rc=%d\n", rc);
855 : }
856 0 : backing_cb_args->cb_fn(backing_cb_args->cb_arg, rc);
857 : }
858 : }
859 :
860 : static void
861 0 : _comp_backing_bdev_write(struct spdk_reduce_backing_io *backing_io)
862 : {
863 0 : struct spdk_reduce_vol_cb_args *backing_cb_args = backing_io->backing_cb_args;
864 0 : struct vbdev_compress *comp_bdev = SPDK_CONTAINEROF(backing_io->dev, struct vbdev_compress,
865 : backing_dev);
866 : int rc;
867 :
868 0 : rc = spdk_bdev_writev_blocks(comp_bdev->base_desc, comp_bdev->base_ch,
869 0 : backing_io->iov, backing_io->iovcnt,
870 0 : backing_io->lba, backing_io->lba_count,
871 : comp_reduce_io_cb,
872 : backing_cb_args);
873 :
874 0 : if (rc) {
875 0 : if (rc == -ENOMEM) {
876 0 : _comp_backing_bdev_queue_io_wait(comp_bdev, backing_io);
877 0 : return;
878 : } else {
879 0 : SPDK_ERRLOG("error submitting writev request, rc=%d\n", rc);
880 : }
881 0 : backing_cb_args->cb_fn(backing_cb_args->cb_arg, rc);
882 : }
883 : }
884 :
885 : static void
886 0 : _comp_backing_bdev_unmap(struct spdk_reduce_backing_io *backing_io)
887 : {
888 0 : struct spdk_reduce_vol_cb_args *backing_cb_args = backing_io->backing_cb_args;
889 0 : struct vbdev_compress *comp_bdev = SPDK_CONTAINEROF(backing_io->dev, struct vbdev_compress,
890 : backing_dev);
891 : int rc;
892 :
893 0 : rc = spdk_bdev_unmap_blocks(comp_bdev->base_desc, comp_bdev->base_ch,
894 0 : backing_io->lba, backing_io->lba_count,
895 : comp_reduce_io_cb,
896 : backing_cb_args);
897 :
898 0 : if (rc) {
899 0 : if (rc == -ENOMEM) {
900 0 : _comp_backing_bdev_queue_io_wait(comp_bdev, backing_io);
901 0 : return;
902 : } else {
903 0 : SPDK_ERRLOG("submitting unmap request, rc=%d\n", rc);
904 : }
905 0 : backing_cb_args->cb_fn(backing_cb_args->cb_arg, rc);
906 : }
907 : }
908 :
909 : /* This is the function provided to the reduceLib for sending reads/writes/unmaps
910 : * directly to the backing device.
911 : */
912 : static void
913 0 : _comp_reduce_submit_backing_io(struct spdk_reduce_backing_io *backing_io)
914 : {
915 0 : switch (backing_io->backing_io_type) {
916 0 : case SPDK_REDUCE_BACKING_IO_WRITE:
917 0 : _comp_backing_bdev_write(backing_io);
918 0 : break;
919 0 : case SPDK_REDUCE_BACKING_IO_READ:
920 0 : _comp_backing_bdev_read(backing_io);
921 0 : break;
922 0 : case SPDK_REDUCE_BACKING_IO_UNMAP:
923 0 : _comp_backing_bdev_unmap(backing_io);
924 0 : break;
925 0 : default:
926 0 : SPDK_ERRLOG("Unknown I/O type %d\n", backing_io->backing_io_type);
927 0 : backing_io->backing_cb_args->cb_fn(backing_io->backing_cb_args->cb_arg, -EINVAL);
928 0 : break;
929 : }
930 0 : }
931 :
932 : static void
933 0 : _comp_reduce_resubmit_backing_io(void *_backing_io)
934 : {
935 0 : struct spdk_reduce_backing_io *backing_io = _backing_io;
936 :
937 0 : _comp_reduce_submit_backing_io(backing_io);
938 0 : }
939 :
940 : /* Called by reduceLib after performing unload vol actions following base bdev hotremove */
941 : static void
942 0 : bdev_hotremove_vol_unload_cb(void *cb_arg, int reduce_errno)
943 : {
944 0 : struct vbdev_compress *comp_bdev = (struct vbdev_compress *)cb_arg;
945 :
946 0 : if (reduce_errno) {
947 0 : SPDK_ERRLOG("number %d\n", reduce_errno);
948 : }
949 :
950 0 : comp_bdev->vol = NULL;
951 0 : spdk_bdev_unregister(&comp_bdev->comp_bdev, NULL, NULL);
952 0 : }
953 :
954 : static void
955 0 : vbdev_compress_base_bdev_hotremove_cb(struct spdk_bdev *bdev_find)
956 : {
957 : struct vbdev_compress *comp_bdev, *tmp;
958 :
959 0 : TAILQ_FOREACH_SAFE(comp_bdev, &g_vbdev_comp, link, tmp) {
960 0 : if (bdev_find == comp_bdev->base_bdev) {
961 : /* Tell reduceLib that we're done with this volume. */
962 0 : spdk_reduce_vol_unload(comp_bdev->vol, bdev_hotremove_vol_unload_cb, comp_bdev);
963 : }
964 : }
965 0 : }
966 :
967 : /* Called when the underlying base bdev triggers asynchronous event such as bdev removal. */
968 : static void
969 0 : vbdev_compress_base_bdev_event_cb(enum spdk_bdev_event_type type, struct spdk_bdev *bdev,
970 : void *event_ctx)
971 : {
972 0 : switch (type) {
973 0 : case SPDK_BDEV_EVENT_REMOVE:
974 0 : vbdev_compress_base_bdev_hotremove_cb(bdev);
975 0 : break;
976 0 : default:
977 0 : SPDK_NOTICELOG("Unsupported bdev event: type %d\n", type);
978 0 : break;
979 : }
980 0 : }
981 :
982 : /* TODO: determine which parms we want user configurable, HC for now
983 : * params.vol_size
984 : * params.chunk_size
985 : * compression PMD, algorithm, window size, comp level, etc.
986 : * DEV_MD_PATH
987 : */
988 :
989 : /* Common function for init and load to allocate and populate the minimal
990 : * information for reducelib to init or load.
991 : */
992 : struct vbdev_compress *
993 0 : _prepare_for_load_init(struct spdk_bdev_desc *bdev_desc, uint32_t lb_size, uint8_t comp_algo,
994 : uint32_t comp_level)
995 : {
996 : struct vbdev_compress *comp_bdev;
997 : struct spdk_bdev *bdev;
998 :
999 0 : comp_bdev = calloc(1, sizeof(struct vbdev_compress));
1000 0 : if (comp_bdev == NULL) {
1001 0 : SPDK_ERRLOG("failed to alloc comp_bdev\n");
1002 0 : return NULL;
1003 : }
1004 :
1005 0 : comp_bdev->backing_dev.submit_backing_io = _comp_reduce_submit_backing_io;
1006 0 : comp_bdev->backing_dev.compress = _comp_reduce_compress;
1007 0 : comp_bdev->backing_dev.decompress = _comp_reduce_decompress;
1008 :
1009 0 : comp_bdev->base_desc = bdev_desc;
1010 0 : bdev = spdk_bdev_desc_get_bdev(bdev_desc);
1011 0 : comp_bdev->base_bdev = bdev;
1012 :
1013 0 : comp_bdev->backing_dev.blocklen = bdev->blocklen;
1014 0 : comp_bdev->backing_dev.blockcnt = bdev->blockcnt;
1015 :
1016 0 : comp_bdev->backing_dev.user_ctx_size = sizeof(struct spdk_bdev_io_wait_entry);
1017 :
1018 0 : comp_bdev->comp_algo = comp_algo;
1019 0 : comp_bdev->comp_level = comp_level;
1020 0 : comp_bdev->params.comp_algo = comp_algo;
1021 0 : comp_bdev->params.comp_level = comp_level;
1022 0 : comp_bdev->params.chunk_size = CHUNK_SIZE;
1023 0 : if (lb_size == 0) {
1024 0 : comp_bdev->params.logical_block_size = bdev->blocklen;
1025 : } else {
1026 0 : comp_bdev->params.logical_block_size = lb_size;
1027 : }
1028 :
1029 0 : comp_bdev->params.backing_io_unit_size = BACKING_IO_SZ;
1030 0 : return comp_bdev;
1031 : }
1032 :
1033 : /* Call reducelib to initialize a new volume */
1034 : static int
1035 0 : vbdev_init_reduce(const char *bdev_name, const char *pm_path, uint32_t lb_size, uint8_t comp_algo,
1036 : uint32_t comp_level, bdev_compress_create_cb cb_fn, void *cb_arg)
1037 : {
1038 0 : struct spdk_bdev_desc *bdev_desc = NULL;
1039 : struct vbdev_init_reduce_ctx *init_ctx;
1040 : struct vbdev_compress *comp_bdev;
1041 : int rc;
1042 :
1043 0 : init_ctx = calloc(1, sizeof(*init_ctx));
1044 0 : if (init_ctx == NULL) {
1045 0 : SPDK_ERRLOG("failed to alloc init contexts\n");
1046 0 : return - ENOMEM;
1047 : }
1048 :
1049 0 : init_ctx->cb_fn = cb_fn;
1050 0 : init_ctx->cb_ctx = cb_arg;
1051 :
1052 0 : rc = spdk_bdev_open_ext(bdev_name, true, vbdev_compress_base_bdev_event_cb,
1053 : NULL, &bdev_desc);
1054 0 : if (rc) {
1055 0 : SPDK_ERRLOG("could not open bdev %s, error %s\n", bdev_name, spdk_strerror(-rc));
1056 0 : free(init_ctx);
1057 0 : return rc;
1058 : }
1059 :
1060 0 : comp_bdev = _prepare_for_load_init(bdev_desc, lb_size, comp_algo, comp_level);
1061 0 : if (comp_bdev == NULL) {
1062 0 : free(init_ctx);
1063 0 : spdk_bdev_close(bdev_desc);
1064 0 : return -EINVAL;
1065 : }
1066 :
1067 0 : init_ctx->comp_bdev = comp_bdev;
1068 :
1069 : /* Save the thread where the base device is opened */
1070 0 : comp_bdev->thread = spdk_get_thread();
1071 :
1072 0 : comp_bdev->base_ch = spdk_bdev_get_io_channel(comp_bdev->base_desc);
1073 :
1074 0 : spdk_reduce_vol_init(&comp_bdev->params, &comp_bdev->backing_dev,
1075 : pm_path,
1076 : vbdev_reduce_init_cb,
1077 : init_ctx);
1078 0 : return 0;
1079 : }
1080 :
1081 : /* We provide this callback for the SPDK channel code to create a channel using
1082 : * the channel struct we provided in our module get_io_channel() entry point. Here
1083 : * we get and save off an underlying base channel of the device below us so that
1084 : * we can communicate with the base bdev on a per channel basis. If we needed
1085 : * our own poller for this vbdev, we'd register it here.
1086 : */
1087 : static int
1088 0 : comp_bdev_ch_create_cb(void *io_device, void *ctx_buf)
1089 : {
1090 0 : struct vbdev_compress *comp_bdev = io_device;
1091 :
1092 : /* Now set the reduce channel if it's not already set. */
1093 0 : pthread_mutex_lock(&comp_bdev->reduce_lock);
1094 0 : if (comp_bdev->ch_count == 0) {
1095 : /* We use this queue to track outstanding IO in our layer. */
1096 0 : TAILQ_INIT(&comp_bdev->pending_comp_ios);
1097 :
1098 : /* We use this to queue up compression operations as needed. */
1099 0 : TAILQ_INIT(&comp_bdev->queued_comp_ops);
1100 :
1101 0 : comp_bdev->base_ch = spdk_bdev_get_io_channel(comp_bdev->base_desc);
1102 0 : comp_bdev->reduce_thread = spdk_get_thread();
1103 0 : comp_bdev->accel_channel = spdk_accel_get_io_channel();
1104 : }
1105 0 : comp_bdev->ch_count++;
1106 0 : pthread_mutex_unlock(&comp_bdev->reduce_lock);
1107 :
1108 0 : return 0;
1109 : }
1110 :
1111 : static void
1112 0 : _channel_cleanup(struct vbdev_compress *comp_bdev)
1113 : {
1114 0 : spdk_put_io_channel(comp_bdev->base_ch);
1115 0 : spdk_put_io_channel(comp_bdev->accel_channel);
1116 0 : comp_bdev->reduce_thread = NULL;
1117 0 : }
1118 :
1119 : /* Used to reroute destroy_ch to the correct thread */
1120 : static void
1121 0 : _comp_bdev_ch_destroy_cb(void *arg)
1122 : {
1123 0 : struct vbdev_compress *comp_bdev = arg;
1124 :
1125 0 : pthread_mutex_lock(&comp_bdev->reduce_lock);
1126 0 : _channel_cleanup(comp_bdev);
1127 0 : pthread_mutex_unlock(&comp_bdev->reduce_lock);
1128 0 : }
1129 :
1130 : /* We provide this callback for the SPDK channel code to destroy a channel
1131 : * created with our create callback. We just need to undo anything we did
1132 : * when we created. If this bdev used its own poller, we'd unregister it here.
1133 : */
1134 : static void
1135 0 : comp_bdev_ch_destroy_cb(void *io_device, void *ctx_buf)
1136 : {
1137 0 : struct vbdev_compress *comp_bdev = io_device;
1138 :
1139 0 : pthread_mutex_lock(&comp_bdev->reduce_lock);
1140 0 : comp_bdev->ch_count--;
1141 0 : if (comp_bdev->ch_count == 0) {
1142 : /* Send this request to the thread where the channel was created. */
1143 0 : if (comp_bdev->reduce_thread != spdk_get_thread()) {
1144 0 : spdk_thread_send_msg(comp_bdev->reduce_thread,
1145 : _comp_bdev_ch_destroy_cb, comp_bdev);
1146 : } else {
1147 0 : _channel_cleanup(comp_bdev);
1148 : }
1149 : }
1150 0 : pthread_mutex_unlock(&comp_bdev->reduce_lock);
1151 0 : }
1152 :
1153 : static int
1154 0 : _check_compress_bdev_comp_algo(enum spdk_accel_comp_algo algo, uint32_t comp_level)
1155 : {
1156 : uint32_t min_level, max_level;
1157 : int rc;
1158 :
1159 0 : rc = spdk_accel_get_compress_level_range(algo, &min_level, &max_level);
1160 0 : if (rc != 0) {
1161 0 : return rc;
1162 : }
1163 :
1164 : /* If both min_level and max_level are 0, the compression level can be ignored.
1165 : * The back-end implementation hardcodes the compression level.
1166 : */
1167 0 : if (min_level == 0 && max_level == 0) {
1168 0 : return 0;
1169 : }
1170 :
1171 0 : if (comp_level > max_level || comp_level < min_level) {
1172 0 : return -EINVAL;
1173 : }
1174 :
1175 0 : return 0;
1176 : }
1177 :
1178 : /* RPC entry point for compression vbdev creation. */
1179 : int
1180 0 : create_compress_bdev(const char *bdev_name, const char *pm_path, uint32_t lb_size,
1181 : uint8_t comp_algo, uint32_t comp_level,
1182 : bdev_compress_create_cb cb_fn, void *cb_arg)
1183 : {
1184 0 : struct vbdev_compress *comp_bdev = NULL;
1185 : struct stat info;
1186 : int rc;
1187 :
1188 0 : if (stat(pm_path, &info) != 0) {
1189 0 : SPDK_ERRLOG("PM path %s does not exist.\n", pm_path);
1190 0 : return -EINVAL;
1191 0 : } else if (!S_ISDIR(info.st_mode)) {
1192 0 : SPDK_ERRLOG("PM path %s is not a directory.\n", pm_path);
1193 0 : return -EINVAL;
1194 : }
1195 :
1196 0 : if ((lb_size != 0) && (lb_size != LB_SIZE_4K) && (lb_size != LB_SIZE_512B)) {
1197 0 : SPDK_ERRLOG("Logical block size must be 512 or 4096\n");
1198 0 : return -EINVAL;
1199 : }
1200 :
1201 0 : rc = _check_compress_bdev_comp_algo(comp_algo, comp_level);
1202 0 : if (rc != 0) {
1203 0 : SPDK_ERRLOG("Compress bdev doesn't support compression algo(%u) or level(%u)\n",
1204 : comp_algo, comp_level);
1205 0 : return rc;
1206 : }
1207 :
1208 0 : TAILQ_FOREACH(comp_bdev, &g_vbdev_comp, link) {
1209 0 : if (strcmp(bdev_name, comp_bdev->base_bdev->name) == 0) {
1210 0 : SPDK_ERRLOG("Bass bdev %s already being used for a compress bdev\n", bdev_name);
1211 0 : return -EBUSY;
1212 : }
1213 : }
1214 0 : return vbdev_init_reduce(bdev_name, pm_path, lb_size, comp_algo, comp_level, cb_fn, cb_arg);
1215 : }
1216 :
1217 : static int
1218 0 : vbdev_compress_init(void)
1219 : {
1220 0 : return 0;
1221 : }
1222 :
1223 : /* Called when the entire module is being torn down. */
1224 : static void
1225 0 : vbdev_compress_finish(void)
1226 : {
1227 : /* TODO: unload vol in a future patch */
1228 0 : }
1229 :
1230 : /* During init we'll be asked how much memory we'd like passed to us
1231 : * in bev_io structures as context. Here's where we specify how
1232 : * much context we want per IO.
1233 : */
1234 : static int
1235 0 : vbdev_compress_get_ctx_size(void)
1236 : {
1237 0 : return sizeof(struct comp_bdev_io);
1238 : }
1239 :
1240 : /* When we register our bdev this is how we specify our entry points. */
1241 : static const struct spdk_bdev_fn_table vbdev_compress_fn_table = {
1242 : .destruct = vbdev_compress_destruct,
1243 : .submit_request = vbdev_compress_submit_request,
1244 : .io_type_supported = vbdev_compress_io_type_supported,
1245 : .get_io_channel = vbdev_compress_get_io_channel,
1246 : .dump_info_json = vbdev_compress_dump_info_json,
1247 : .write_config_json = NULL,
1248 : };
1249 :
1250 : static struct spdk_bdev_module compress_if = {
1251 : .name = "compress",
1252 : .module_init = vbdev_compress_init,
1253 : .get_ctx_size = vbdev_compress_get_ctx_size,
1254 : .examine_disk = vbdev_compress_examine,
1255 : .module_fini = vbdev_compress_finish,
1256 : .config_json = vbdev_compress_config_json
1257 : };
1258 :
1259 1 : SPDK_BDEV_MODULE_REGISTER(compress, &compress_if)
1260 :
1261 0 : static int _set_compbdev_name(struct vbdev_compress *comp_bdev)
1262 : {
1263 : struct spdk_bdev_alias *aliases;
1264 :
1265 0 : if (!TAILQ_EMPTY(spdk_bdev_get_aliases(comp_bdev->base_bdev))) {
1266 0 : aliases = TAILQ_FIRST(spdk_bdev_get_aliases(comp_bdev->base_bdev));
1267 0 : comp_bdev->comp_bdev.name = spdk_sprintf_alloc("COMP_%s", aliases->alias.name);
1268 0 : if (!comp_bdev->comp_bdev.name) {
1269 0 : SPDK_ERRLOG("could not allocate comp_bdev name for alias\n");
1270 0 : return -ENOMEM;
1271 : }
1272 : } else {
1273 0 : comp_bdev->comp_bdev.name = spdk_sprintf_alloc("COMP_%s", comp_bdev->base_bdev->name);
1274 0 : if (!comp_bdev->comp_bdev.name) {
1275 0 : SPDK_ERRLOG("could not allocate comp_bdev name for unique name\n");
1276 0 : return -ENOMEM;
1277 : }
1278 : }
1279 0 : return 0;
1280 : }
1281 :
1282 : static int
1283 0 : vbdev_compress_claim(struct vbdev_compress *comp_bdev)
1284 : {
1285 : struct spdk_uuid ns_uuid;
1286 : int rc;
1287 :
1288 0 : if (_set_compbdev_name(comp_bdev)) {
1289 0 : return -EINVAL;
1290 : }
1291 :
1292 : /* Note: some of the fields below will change in the future - for example,
1293 : * blockcnt specifically will not match (the compressed volume size will
1294 : * be slightly less than the base bdev size)
1295 : */
1296 0 : comp_bdev->comp_bdev.product_name = COMP_BDEV_NAME;
1297 0 : comp_bdev->comp_bdev.write_cache = comp_bdev->base_bdev->write_cache;
1298 :
1299 0 : comp_bdev->comp_bdev.optimal_io_boundary =
1300 0 : comp_bdev->params.chunk_size / comp_bdev->params.logical_block_size;
1301 :
1302 0 : comp_bdev->comp_bdev.split_on_optimal_io_boundary = true;
1303 :
1304 0 : comp_bdev->comp_bdev.blocklen = comp_bdev->params.logical_block_size;
1305 0 : comp_bdev->comp_bdev.blockcnt = comp_bdev->params.vol_size / comp_bdev->comp_bdev.blocklen;
1306 0 : assert(comp_bdev->comp_bdev.blockcnt > 0);
1307 :
1308 : /* This is the context that is passed to us when the bdev
1309 : * layer calls in so we'll save our comp_bdev node here.
1310 : */
1311 0 : comp_bdev->comp_bdev.ctxt = comp_bdev;
1312 0 : comp_bdev->comp_bdev.fn_table = &vbdev_compress_fn_table;
1313 0 : comp_bdev->comp_bdev.module = &compress_if;
1314 :
1315 : /* Generate UUID based on namespace UUID + base bdev UUID. */
1316 0 : spdk_uuid_parse(&ns_uuid, BDEV_COMPRESS_NAMESPACE_UUID);
1317 0 : rc = spdk_uuid_generate_sha1(&comp_bdev->comp_bdev.uuid, &ns_uuid,
1318 0 : (const char *)&comp_bdev->base_bdev->uuid, sizeof(struct spdk_uuid));
1319 0 : if (rc) {
1320 0 : SPDK_ERRLOG("Unable to generate new UUID for compress bdev, error %s\n", spdk_strerror(-rc));
1321 0 : return -EINVAL;
1322 : }
1323 :
1324 0 : pthread_mutex_init(&comp_bdev->reduce_lock, NULL);
1325 :
1326 : /* Save the thread where the base device is opened */
1327 0 : comp_bdev->thread = spdk_get_thread();
1328 :
1329 0 : spdk_io_device_register(comp_bdev, comp_bdev_ch_create_cb, comp_bdev_ch_destroy_cb,
1330 : sizeof(struct comp_io_channel),
1331 0 : comp_bdev->comp_bdev.name);
1332 :
1333 0 : rc = spdk_bdev_module_claim_bdev(comp_bdev->base_bdev, comp_bdev->base_desc,
1334 : comp_bdev->comp_bdev.module);
1335 0 : if (rc) {
1336 0 : SPDK_ERRLOG("could not claim bdev %s, error %s\n", spdk_bdev_get_name(comp_bdev->base_bdev),
1337 : spdk_strerror(-rc));
1338 0 : goto error_claim;
1339 : }
1340 :
1341 0 : rc = spdk_bdev_register(&comp_bdev->comp_bdev);
1342 0 : if (rc < 0) {
1343 0 : SPDK_ERRLOG("trying to register bdev, error %s\n", spdk_strerror(-rc));
1344 0 : goto error_bdev_register;
1345 : }
1346 :
1347 0 : TAILQ_INSERT_TAIL(&g_vbdev_comp, comp_bdev, link);
1348 :
1349 0 : SPDK_NOTICELOG("registered io_device and virtual bdev for: %s\n", comp_bdev->comp_bdev.name);
1350 :
1351 0 : return 0;
1352 :
1353 : /* Error cleanup paths. */
1354 0 : error_bdev_register:
1355 0 : spdk_bdev_module_release_bdev(comp_bdev->base_bdev);
1356 0 : error_claim:
1357 0 : spdk_io_device_unregister(comp_bdev, NULL);
1358 0 : free(comp_bdev->comp_bdev.name);
1359 0 : return rc;
1360 : }
1361 :
1362 : static void
1363 0 : _vbdev_compress_delete_done(void *_ctx)
1364 : {
1365 0 : struct vbdev_comp_delete_ctx *ctx = _ctx;
1366 :
1367 0 : ctx->cb_fn(ctx->cb_arg, ctx->cb_rc);
1368 :
1369 0 : free(ctx);
1370 0 : }
1371 :
1372 : static void
1373 0 : vbdev_compress_delete_done(void *cb_arg, int bdeverrno)
1374 : {
1375 0 : struct vbdev_comp_delete_ctx *ctx = cb_arg;
1376 :
1377 0 : ctx->cb_rc = bdeverrno;
1378 :
1379 0 : if (ctx->orig_thread != spdk_get_thread()) {
1380 0 : spdk_thread_send_msg(ctx->orig_thread, _vbdev_compress_delete_done, ctx);
1381 : } else {
1382 0 : _vbdev_compress_delete_done(ctx);
1383 : }
1384 0 : }
1385 :
1386 : void
1387 0 : bdev_compress_delete(const char *name, spdk_delete_compress_complete cb_fn, void *cb_arg)
1388 : {
1389 0 : struct vbdev_compress *comp_bdev = NULL;
1390 : struct vbdev_comp_delete_ctx *ctx;
1391 :
1392 0 : TAILQ_FOREACH(comp_bdev, &g_vbdev_comp, link) {
1393 0 : if (strcmp(name, comp_bdev->comp_bdev.name) == 0) {
1394 0 : break;
1395 : }
1396 : }
1397 :
1398 0 : if (comp_bdev == NULL) {
1399 0 : cb_fn(cb_arg, -ENODEV);
1400 0 : return;
1401 : }
1402 :
1403 0 : ctx = calloc(1, sizeof(*ctx));
1404 0 : if (ctx == NULL) {
1405 0 : SPDK_ERRLOG("Failed to allocate delete context\n");
1406 0 : cb_fn(cb_arg, -ENOMEM);
1407 0 : return;
1408 : }
1409 :
1410 : /* Save these for after the vol is destroyed. */
1411 0 : ctx->cb_fn = cb_fn;
1412 0 : ctx->cb_arg = cb_arg;
1413 0 : ctx->orig_thread = spdk_get_thread();
1414 :
1415 0 : comp_bdev->delete_ctx = ctx;
1416 :
1417 : /* Tell reducelib that we're done with this volume. */
1418 0 : if (comp_bdev->orphaned == false) {
1419 0 : spdk_reduce_vol_unload(comp_bdev->vol, delete_vol_unload_cb, comp_bdev);
1420 : } else {
1421 0 : delete_vol_unload_cb(comp_bdev, 0);
1422 : }
1423 : }
1424 :
1425 : static void
1426 0 : _vbdev_reduce_load_unload_cb(void *ctx, int reduce_errno)
1427 : {
1428 0 : }
1429 :
1430 : static void
1431 0 : _vbdev_reduce_load_cb(void *ctx)
1432 : {
1433 0 : struct vbdev_compress *comp_bdev = ctx;
1434 : int rc;
1435 :
1436 0 : assert(comp_bdev->base_desc != NULL);
1437 :
1438 : /* Done with metadata operations */
1439 0 : spdk_put_io_channel(comp_bdev->base_ch);
1440 :
1441 0 : if (comp_bdev->reduce_errno == 0) {
1442 0 : rc = vbdev_compress_claim(comp_bdev);
1443 0 : if (rc != 0) {
1444 0 : spdk_reduce_vol_unload(comp_bdev->vol, _vbdev_reduce_load_unload_cb, NULL);
1445 0 : goto err;
1446 : }
1447 0 : } else if (comp_bdev->reduce_errno == -ENOENT) {
1448 0 : if (_set_compbdev_name(comp_bdev)) {
1449 0 : goto err;
1450 : }
1451 :
1452 : /* Save the thread where the base device is opened */
1453 0 : comp_bdev->thread = spdk_get_thread();
1454 :
1455 0 : comp_bdev->comp_bdev.module = &compress_if;
1456 0 : pthread_mutex_init(&comp_bdev->reduce_lock, NULL);
1457 0 : rc = spdk_bdev_module_claim_bdev(comp_bdev->base_bdev, comp_bdev->base_desc,
1458 : comp_bdev->comp_bdev.module);
1459 0 : if (rc) {
1460 0 : SPDK_ERRLOG("could not claim bdev %s, error %s\n", spdk_bdev_get_name(comp_bdev->base_bdev),
1461 : spdk_strerror(-rc));
1462 0 : free(comp_bdev->comp_bdev.name);
1463 0 : goto err;
1464 : }
1465 :
1466 0 : comp_bdev->orphaned = true;
1467 0 : TAILQ_INSERT_TAIL(&g_vbdev_comp, comp_bdev, link);
1468 : } else {
1469 0 : if (comp_bdev->reduce_errno != -EILSEQ) {
1470 0 : SPDK_ERRLOG("for vol %s, error %s\n", spdk_bdev_get_name(comp_bdev->base_bdev),
1471 : spdk_strerror(-comp_bdev->reduce_errno));
1472 : }
1473 0 : goto err;
1474 : }
1475 :
1476 0 : spdk_bdev_module_examine_done(&compress_if);
1477 0 : return;
1478 :
1479 0 : err:
1480 : /* Close the underlying bdev on its same opened thread. */
1481 0 : spdk_bdev_close(comp_bdev->base_desc);
1482 0 : free(comp_bdev);
1483 0 : spdk_bdev_module_examine_done(&compress_if);
1484 : }
1485 :
1486 : /* Callback from reduce for then load is complete. We'll pass the vbdev_comp struct
1487 : * used for initial metadata operations to claim where it will be further filled out
1488 : * and added to the global list.
1489 : */
1490 : static void
1491 0 : vbdev_reduce_load_cb(void *cb_arg, struct spdk_reduce_vol *vol, int reduce_errno)
1492 : {
1493 0 : struct vbdev_compress *comp_bdev = cb_arg;
1494 :
1495 0 : if (reduce_errno == 0) {
1496 : /* Update information following volume load. */
1497 0 : comp_bdev->vol = vol;
1498 0 : memcpy(&comp_bdev->params, spdk_reduce_vol_get_params(vol),
1499 : sizeof(struct spdk_reduce_vol_params));
1500 0 : comp_bdev->comp_algo = comp_bdev->params.comp_algo;
1501 0 : comp_bdev->comp_level = comp_bdev->params.comp_level;
1502 : }
1503 :
1504 0 : comp_bdev->reduce_errno = reduce_errno;
1505 :
1506 0 : if (comp_bdev->thread && comp_bdev->thread != spdk_get_thread()) {
1507 0 : spdk_thread_send_msg(comp_bdev->thread, _vbdev_reduce_load_cb, comp_bdev);
1508 : } else {
1509 0 : _vbdev_reduce_load_cb(comp_bdev);
1510 : }
1511 :
1512 0 : }
1513 :
1514 : /* Examine_disk entry point: will do a metadata load to see if this is ours,
1515 : * and if so will go ahead and claim it.
1516 : */
1517 : static void
1518 0 : vbdev_compress_examine(struct spdk_bdev *bdev)
1519 : {
1520 0 : struct spdk_bdev_desc *bdev_desc = NULL;
1521 : struct vbdev_compress *comp_bdev;
1522 : int rc;
1523 :
1524 0 : if (strcmp(bdev->product_name, COMP_BDEV_NAME) == 0) {
1525 0 : spdk_bdev_module_examine_done(&compress_if);
1526 0 : return;
1527 : }
1528 :
1529 0 : rc = spdk_bdev_open_ext(spdk_bdev_get_name(bdev), false,
1530 : vbdev_compress_base_bdev_event_cb, NULL, &bdev_desc);
1531 0 : if (rc) {
1532 0 : SPDK_ERRLOG("could not open bdev %s, error %s\n", spdk_bdev_get_name(bdev),
1533 : spdk_strerror(-rc));
1534 0 : spdk_bdev_module_examine_done(&compress_if);
1535 0 : return;
1536 : }
1537 :
1538 0 : comp_bdev = _prepare_for_load_init(bdev_desc, 0, SPDK_ACCEL_COMP_ALGO_DEFLATE, 1);
1539 0 : if (comp_bdev == NULL) {
1540 0 : spdk_bdev_close(bdev_desc);
1541 0 : spdk_bdev_module_examine_done(&compress_if);
1542 0 : return;
1543 : }
1544 :
1545 : /* Save the thread where the base device is opened */
1546 0 : comp_bdev->thread = spdk_get_thread();
1547 :
1548 0 : comp_bdev->base_ch = spdk_bdev_get_io_channel(comp_bdev->base_desc);
1549 0 : spdk_reduce_vol_load(&comp_bdev->backing_dev, vbdev_reduce_load_cb, comp_bdev);
1550 : }
1551 :
1552 1 : SPDK_LOG_REGISTER_COMPONENT(vbdev_compress)
|
