2 * Copyright (C) 2015 Michael Brown <mbrown@fensystems.co.uk>.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public License as
6 * published by the Free Software Foundation; either version 2 of the
7 * License, or (at your option) any later version.
9 * This program is distributed in the hope that it will be useful, but
10 * WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * General Public License for more details.
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write to the Free Software
16 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
19 * You can also choose to distribute this program under the terms of
20 * the Unmodified Binary Distribution Licence (as given in the file
21 * COPYING.UBDL), provided that you have satisfied its requirements.
24 FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
29 #include <ipxe/http.h>
30 #include <ipxe/iobuf.h>
31 #include <ipxe/xfer.h>
33 #include <ipxe/timer.h>
34 #include <ipxe/profile.h>
35 #include <ipxe/fault.h>
36 #include <ipxe/pccrr.h>
37 #include <ipxe/peerblk.h>
41 * Peer Content Caching and Retrieval (PeerDist) protocol block downloads
45 /** PeerDist decryption chunksize
47 * This is a policy decision.
49 #define PEERBLK_DECRYPT_CHUNKSIZE 2048
51 /** PeerDist raw block download attempt initial progress timeout
53 * This is a policy decision.
55 #define PEERBLK_RAW_OPEN_TIMEOUT ( 10 * TICKS_PER_SEC )
57 /** PeerDist raw block download attempt ongoing progress timeout
59 * This is a policy decision.
61 #define PEERBLK_RAW_RX_TIMEOUT ( 15 * TICKS_PER_SEC )
63 /** PeerDist retrieval protocol block download attempt initial progress timeout
65 * This is a policy decision.
67 #define PEERBLK_RETRIEVAL_OPEN_TIMEOUT ( 3 * TICKS_PER_SEC )
69 /** PeerDist retrieval protocol block download attempt ongoing progress timeout
71 * This is a policy decision.
73 #define PEERBLK_RETRIEVAL_RX_TIMEOUT ( 5 * TICKS_PER_SEC )
75 /** PeerDist maximum number of full download attempt cycles
77 * This is the maximum number of times that we will try a full cycle
78 * of download attempts (i.e. a retrieval protocol download attempt
79 * from each discovered peer plus a raw download attempt from the
82 * This is a policy decision.
84 #define PEERBLK_MAX_ATTEMPT_CYCLES 4
86 /** PeerDist block download profiler */
87 static struct profiler peerblk_download_profiler __profiler =
88 { .name = "peerblk.download" };
90 /** PeerDist block download attempt success profiler */
91 static struct profiler peerblk_attempt_success_profiler __profiler =
92 { .name = "peerblk.attempt.success" };
94 /** PeerDist block download attempt failure profiler */
95 static struct profiler peerblk_attempt_failure_profiler __profiler =
96 { .name = "peerblk.attempt.failure" };
98 /** PeerDist block download attempt timeout profiler */
99 static struct profiler peerblk_attempt_timeout_profiler __profiler =
100 { .name = "peerblk.attempt.timeout" };
102 /** PeerDist block download discovery success profiler */
103 static struct profiler peerblk_discovery_success_profiler __profiler =
104 { .name = "peerblk.discovery.success" };
106 /** PeerDist block download discovery timeout profiler */
107 static struct profiler peerblk_discovery_timeout_profiler __profiler =
108 { .name = "peerblk.discovery.timeout" };
111 * Get profiling timestamp
113 * @ret timestamp Timestamp
115 static inline __attribute__ (( always_inline )) unsigned long
116 peerblk_timestamp ( void ) {
126 * Free PeerDist block download
128 * @v refcnt Reference count
130 static void peerblk_free ( struct refcnt *refcnt ) {
131 struct peerdist_block *peerblk =
132 container_of ( refcnt, struct peerdist_block, refcnt );
134 uri_put ( peerblk->uri );
135 free ( peerblk->cipherctx );
140 * Reset PeerDist block download attempt
142 * @v peerblk PeerDist block download
143 * @v rc Reason for reset
145 static void peerblk_reset ( struct peerdist_block *peerblk, int rc ) {
147 /* Stop decryption process */
148 process_del ( &peerblk->process );
151 stop_timer ( &peerblk->timer );
153 /* Abort any current download attempt */
154 intf_restart ( &peerblk->raw, rc );
155 intf_restart ( &peerblk->retrieval, rc );
157 /* Empty received data buffer */
158 xferbuf_free ( &peerblk->buffer );
161 /* Reset digest and free cipher context */
162 digest_init ( peerblk->digest, peerblk->digestctx );
163 free ( peerblk->cipherctx );
164 peerblk->cipherctx = NULL;
165 peerblk->cipher = NULL;
167 /* Reset trim thresholds */
168 peerblk->start = ( peerblk->trim.start - peerblk->range.start );
169 peerblk->end = ( peerblk->trim.end - peerblk->range.start );
170 assert ( peerblk->start <= peerblk->end );
174 * Close PeerDist block download
176 * @v peerblk PeerDist block download
177 * @v rc Reason for close
179 static void peerblk_close ( struct peerdist_block *peerblk, int rc ) {
180 unsigned long now = peerblk_timestamp();
182 /* Profile overall block download */
183 profile_custom ( &peerblk_download_profiler,
184 ( now - peerblk->started ) );
186 /* Reset download attempt */
187 peerblk_reset ( peerblk, rc );
189 /* Close discovery */
190 peerdisc_close ( &peerblk->discovery );
192 /* Shut down all interfaces */
193 intf_shutdown ( &peerblk->retrieval, rc );
194 intf_shutdown ( &peerblk->raw, rc );
195 intf_shutdown ( &peerblk->xfer, rc );
199 * Calculate offset within overall download
201 * @v peerblk PeerDist block download
202 * @v pos Position within incoming data stream
203 * @ret offset Offset within overall download
205 static inline __attribute__ (( always_inline )) size_t
206 peerblk_offset ( struct peerdist_block *peerblk, size_t pos ) {
208 return ( ( pos - peerblk->start ) + peerblk->offset );
212 * Deliver download attempt data block
214 * @v peerblk PeerDist block download
215 * @v iobuf I/O buffer
216 * @v meta Original data transfer metadata
217 * @v pos Position within incoming data stream
218 * @ret rc Return status code
220 static int peerblk_deliver ( struct peerdist_block *peerblk,
221 struct io_buffer *iobuf,
222 struct xfer_metadata *meta, size_t pos ) {
223 struct xfer_metadata xfer_meta;
224 size_t len = iob_len ( iobuf );
226 size_t end = ( pos + len );
229 /* Discard zero-length packets and packets which lie entirely
230 * outside the trimmed range.
232 if ( ( start >= peerblk->end ) || ( end <= peerblk->start ) ||
238 /* Truncate data to within trimmed range */
239 if ( start < peerblk->start ) {
240 iob_pull ( iobuf, ( peerblk->start - start ) );
241 start = peerblk->start;
243 if ( end > peerblk->end ) {
244 iob_unput ( iobuf, ( end - peerblk->end ) );
248 /* Construct metadata */
249 memcpy ( &xfer_meta, meta, sizeof ( xfer_meta ) );
250 xfer_meta.flags |= XFER_FL_ABS_OFFSET;
251 xfer_meta.offset = peerblk_offset ( peerblk, start );
254 if ( ( rc = xfer_deliver ( &peerblk->xfer, iob_disown ( iobuf ),
255 &xfer_meta ) ) != 0 ) {
256 DBGC ( peerblk, "PEERBLK %p %d.%d could not deliver data: %s\n",
257 peerblk, peerblk->segment, peerblk->block,
266 * Finish PeerDist block download attempt
268 * @v peerblk PeerDist block download
269 * @v rc Reason for close
271 static void peerblk_done ( struct peerdist_block *peerblk, int rc ) {
272 struct digest_algorithm *digest = peerblk->digest;
273 uint8_t hash[digest->digestsize];
274 unsigned long now = peerblk_timestamp();
276 /* Check for errors on completion */
278 DBGC ( peerblk, "PEERBLK %p %d.%d attempt failed: %s\n",
279 peerblk, peerblk->segment, peerblk->block,
285 digest_final ( digest, peerblk->digestctx, hash );
286 if ( memcmp ( hash, peerblk->hash, peerblk->digestsize ) != 0 ) {
287 DBGC ( peerblk, "PEERBLK %p %d.%d digest mismatch:\n",
288 peerblk, peerblk->segment, peerblk->block );
289 DBGC_HDA ( peerblk, 0, hash, peerblk->digestsize );
290 DBGC_HDA ( peerblk, 0, peerblk->hash, peerblk->digestsize );
295 /* Profile successful attempt */
296 profile_custom ( &peerblk_attempt_success_profiler,
297 ( now - peerblk->attempted ) );
300 peerblk_close ( peerblk, 0 );
304 /* Record failure reason and schedule a retry attempt */
305 profile_custom ( &peerblk_attempt_failure_profiler,
306 ( now - peerblk->attempted ) );
307 peerblk_reset ( peerblk, rc );
309 start_timer_nodelay ( &peerblk->timer );
312 /******************************************************************************
314 * Raw block download attempts (using an HTTP range request)
316 ******************************************************************************
320 * Open PeerDist raw block download attempt
322 * @v peerblk PeerDist block download
323 * @ret rc Return status code
325 static int peerblk_raw_open ( struct peerdist_block *peerblk ) {
326 struct http_request_range range;
329 DBGC2 ( peerblk, "PEERBLK %p %d.%d attempting raw range request\n",
330 peerblk, peerblk->segment, peerblk->block );
332 /* Construct HTTP range */
333 memset ( &range, 0, sizeof ( range ) );
334 range.start = peerblk->range.start;
335 range.len = ( peerblk->range.end - peerblk->range.start );
337 /* Initiate range request to retrieve block */
338 if ( ( rc = http_open ( &peerblk->raw, &http_get, peerblk->uri,
339 &range, NULL ) ) != 0 ) {
340 DBGC ( peerblk, "PEERBLK %p %d.%d could not create range "
341 "request: %s\n", peerblk, peerblk->segment,
342 peerblk->block, strerror ( rc ) );
346 /* Annul HTTP connection (for testing) if applicable. Do not
347 * report as an immediate error, in order to test our ability
348 * to recover from a totally unresponsive HTTP server.
350 if ( inject_fault ( PEERBLK_ANNUL_RATE ) )
351 intf_restart ( &peerblk->raw, 0 );
357 * Receive PeerDist raw data
359 * @v peerblk PeerDist block download
360 * @v iobuf I/O buffer
361 * @v meta Data transfer metadata
362 * @ret rc Return status code
364 static int peerblk_raw_rx ( struct peerdist_block *peerblk,
365 struct io_buffer *iobuf,
366 struct xfer_metadata *meta ) {
367 size_t len = iob_len ( iobuf );
368 size_t pos = peerblk->pos;
369 size_t mid = ( ( peerblk->range.end - peerblk->range.start ) / 2 );
372 /* Corrupt received data (for testing) if applicable */
373 inject_corruption ( PEERBLK_CORRUPT_RATE, iobuf->data, len );
375 /* Fail if data is delivered out of order, since the streaming
376 * digest requires strict ordering.
378 if ( ( rc = xfer_check_order ( meta, &peerblk->pos, len ) ) != 0 )
381 /* Add data to digest */
382 digest_update ( peerblk->digest, peerblk->digestctx, iobuf->data, len );
385 if ( ( rc = peerblk_deliver ( peerblk, iob_disown ( iobuf ), meta,
389 /* Extend download attempt timer */
390 start_timer_fixed ( &peerblk->timer, PEERBLK_RAW_RX_TIMEOUT );
392 /* Stall download attempt (for testing) if applicable */
393 if ( ( pos < mid ) && ( ( pos + len ) >= mid ) &&
394 ( ( rc = inject_fault ( PEERBLK_STALL_RATE ) ) != 0 ) ) {
395 intf_restart ( &peerblk->raw, rc );
402 peerblk_done ( peerblk, rc );
407 * Close PeerDist raw block download attempt
409 * @v peerblk PeerDist block download
410 * @v rc Reason for close
412 static void peerblk_raw_close ( struct peerdist_block *peerblk, int rc ) {
414 /* Restart interface */
415 intf_restart ( &peerblk->raw, rc );
417 /* Fail immediately if we have an error */
421 /* Abort download attempt (for testing) if applicable */
422 if ( ( rc = inject_fault ( PEERBLK_ABORT_RATE ) ) != 0 )
426 /* Complete download attempt */
427 peerblk_done ( peerblk, rc );
430 /******************************************************************************
432 * Retrieval protocol block download attempts (using HTTP POST)
434 ******************************************************************************
438 * Construct PeerDist retrieval protocol URI
440 * @v location Peer location
441 * @ret uri Retrieval URI, or NULL on error
443 static struct uri * peerblk_retrieval_uri ( const char *location ) {
444 char uri_string[ 7 /* "http://" */ + strlen ( location ) +
445 sizeof ( PEERDIST_MAGIC_PATH /* includes NUL */ ) ];
447 /* Construct URI string */
448 snprintf ( uri_string, sizeof ( uri_string ),
449 ( "http://%s" PEERDIST_MAGIC_PATH ), location );
451 /* Parse URI string */
452 return parse_uri ( uri_string );
456 * Open PeerDist retrieval protocol block download attempt
458 * @v peerblk PeerDist block download
459 * @v location Peer location
460 * @ret rc Return status code
462 static int peerblk_retrieval_open ( struct peerdist_block *peerblk,
463 const char *location ) {
464 size_t digestsize = peerblk->digestsize;
465 peerdist_msg_getblks_t ( digestsize, 1, 0 ) req;
466 peerblk_msg_blk_t ( digestsize, 0, 0, 0 ) *rsp;
467 struct http_request_content content;
471 DBGC2 ( peerblk, "PEERBLK %p %d.%d attempting retrieval from %s\n",
472 peerblk, peerblk->segment, peerblk->block, location );
474 /* Construct block fetch request */
475 memset ( &req, 0, sizeof ( req ) );
476 req.getblks.hdr.version.raw = htonl ( PEERDIST_MSG_GETBLKS_VERSION );
477 req.getblks.hdr.type = htonl ( PEERDIST_MSG_GETBLKS_TYPE );
478 req.getblks.hdr.len = htonl ( sizeof ( req ) );
479 req.getblks.hdr.algorithm = htonl ( PEERDIST_MSG_AES_128_CBC );
480 req.segment.segment.digestsize = htonl ( digestsize );
481 memcpy ( req.segment.id, peerblk->id, digestsize );
482 req.ranges.ranges.count = htonl ( 1 );
483 req.ranges.range[0].first = htonl ( peerblk->block );
484 req.ranges.range[0].count = htonl ( 1 );
486 /* Construct POST request content */
487 memset ( &content, 0, sizeof ( content ) );
489 content.len = sizeof ( req );
492 if ( ( uri = peerblk_retrieval_uri ( location ) ) == NULL ) {
497 /* Update trim thresholds */
498 peerblk->start += offsetof ( typeof ( *rsp ), msg.vrf );
499 peerblk->end += offsetof ( typeof ( *rsp ), msg.vrf );
501 /* Initiate HTTP POST to retrieve block */
502 if ( ( rc = http_open ( &peerblk->retrieval, &http_post, uri,
503 NULL, &content ) ) != 0 ) {
504 DBGC ( peerblk, "PEERBLK %p %d.%d could not create retrieval "
505 "request: %s\n", peerblk, peerblk->segment,
506 peerblk->block, strerror ( rc ) );
510 /* Annul HTTP connection (for testing) if applicable. Do not
511 * report as an immediate error, in order to test our ability
512 * to recover from a totally unresponsive HTTP server.
514 if ( inject_fault ( PEERBLK_ANNUL_RATE ) )
515 intf_restart ( &peerblk->retrieval, 0 );
524 * Receive PeerDist retrieval protocol data
526 * @v peerblk PeerDist block download
527 * @v iobuf I/O buffer
528 * @v meta Data transfer metadata
529 * @ret rc Return status code
531 static int peerblk_retrieval_rx ( struct peerdist_block *peerblk,
532 struct io_buffer *iobuf,
533 struct xfer_metadata *meta ) {
534 size_t len = iob_len ( iobuf );
542 /* Some genius at Microsoft thought it would be a great idea
543 * to place the AES-CBC initialisation vector *after* the
544 * encrypted data, thereby making it logically impossible to
545 * decrypt each packet as it arrives.
547 * To work around this mindless stupidity, we deliver the
548 * ciphertext as-is and later use xfer_buffer() to obtain
549 * access to the underlying data transfer buffer in order to
550 * perform the decryption.
552 * There will be some data both before and after the bytes
553 * corresponding to the trimmed plaintext: a MSG_BLK
554 * header/footer, some block padding for the AES-CBC cipher,
555 * and a possibly large quantity of unwanted ciphertext which
556 * is excluded from the trimmed content range. We store this
557 * data in a local data transfer buffer. If the amount of
558 * data to be stored is too large, we will fail allocation and
559 * so eventually fall back to using a range request (which
560 * does not require this kind of temporary storage
564 /* Corrupt received data (for testing) if applicable */
565 inject_corruption ( PEERBLK_CORRUPT_RATE, iobuf->data, len );
567 /* Calculate start and end positions of this buffer */
568 start = peerblk->pos;
569 if ( meta->flags & XFER_FL_ABS_OFFSET )
571 start += meta->offset;
572 end = ( start + len );
574 /* Buffer any data before the trimmed content */
575 if ( ( start < peerblk->start ) && ( len > 0 ) ) {
577 /* Calculate length of data before the trimmed content */
578 before = ( peerblk->start - start );
582 /* Buffer data before the trimmed content */
583 if ( ( rc = xferbuf_write ( &peerblk->buffer, start,
584 iobuf->data, before ) ) != 0 ) {
585 DBGC ( peerblk, "PEERBLK %p %d.%d could not buffer "
586 "data: %s\n", peerblk, peerblk->segment,
587 peerblk->block, strerror ( rc ) );
592 /* Buffer any data after the trimmed content */
593 if ( ( end > peerblk->end ) && ( len > 0 ) ) {
595 /* Calculate length of data after the trimmed content */
596 after = ( end - peerblk->end );
600 /* Buffer data after the trimmed content */
601 cut = ( peerblk->end - peerblk->start );
602 if ( ( rc = xferbuf_write ( &peerblk->buffer,
603 ( end - after - cut ),
604 ( iobuf->data + len - after ),
606 DBGC ( peerblk, "PEERBLK %p %d.%d could not buffer "
607 "data: %s\n", peerblk, peerblk->segment,
608 peerblk->block, strerror ( rc ) );
613 /* Deliver any remaining data */
614 if ( ( rc = peerblk_deliver ( peerblk, iob_disown ( iobuf ), meta,
618 /* Update position */
621 /* Extend download attempt timer */
622 start_timer_fixed ( &peerblk->timer, PEERBLK_RETRIEVAL_RX_TIMEOUT );
624 /* Stall download attempt (for testing) if applicable */
625 if ( ( start < peerblk->end ) && ( end >= peerblk->end ) &&
626 ( ( rc = inject_fault ( PEERBLK_STALL_RATE ) ) != 0 ) ) {
627 intf_restart ( &peerblk->retrieval, rc );
634 peerblk_done ( peerblk, rc );
639 * Parse retrieval protocol message header
641 * @v peerblk PeerDist block download
642 * @ret rc Return status code
644 static int peerblk_parse_header ( struct peerdist_block *peerblk ) {
646 struct peerdist_msg_transport_header hdr;
647 struct peerdist_msg_header msg;
648 } __attribute__ (( packed )) *msg = peerblk->buffer.data;
649 struct cipher_algorithm *cipher;
650 size_t len = peerblk->buffer.len;
654 /* Check message length */
655 if ( len < sizeof ( *msg ) ) {
656 DBGC ( peerblk, "PEERBLK %p %d.%d message too short for header "
657 "(%zd bytes)\n", peerblk, peerblk->segment,
658 peerblk->block, len );
662 /* Check message type */
663 if ( msg->msg.type != htonl ( PEERDIST_MSG_BLK_TYPE ) ) {
664 DBGC ( peerblk, "PEERBLK %p %d.%d unexpected message type "
665 "%#08x\n", peerblk, peerblk->segment, peerblk->block,
666 ntohl ( msg->msg.type ) );
670 /* Determine cipher algorithm and key length */
671 cipher = &aes_cbc_algorithm;
672 switch ( msg->msg.algorithm ) {
673 case htonl ( PEERDIST_MSG_PLAINTEXT ) :
676 case htonl ( PEERDIST_MSG_AES_128_CBC ) :
677 keylen = ( 128 / 8 );
679 case htonl ( PEERDIST_MSG_AES_192_CBC ) :
680 keylen = ( 192 / 8 );
682 case htonl ( PEERDIST_MSG_AES_256_CBC ) :
683 keylen = ( 256 / 8 );
686 DBGC ( peerblk, "PEERBLK %p %d.%d unrecognised algorithm "
687 "%#08x\n", peerblk, peerblk->segment, peerblk->block,
688 ntohl ( msg->msg.algorithm ) );
691 DBGC2 ( peerblk, "PEERBLK %p %d.%d using %s with %zd-bit key\n",
692 peerblk, peerblk->segment, peerblk->block,
693 ( cipher ? cipher->name : "plaintext" ), ( 8 * keylen ) );
695 /* Sanity check key length against maximum secret length */
696 if ( keylen > peerblk->digestsize ) {
697 DBGC ( peerblk, "PEERBLK %p %d.%d %zd-byte secret too short "
698 "for %zd-bit key\n", peerblk, peerblk->segment,
699 peerblk->block, peerblk->digestsize, ( 8 * keylen ) );
703 /* Allocate cipher context. Freeing the cipher context (on
704 * error or otherwise) is handled by peerblk_reset().
706 peerblk->cipher = cipher;
707 assert ( peerblk->cipherctx == NULL );
708 peerblk->cipherctx = malloc ( cipher->ctxsize );
709 if ( ! peerblk->cipherctx )
712 /* Initialise cipher */
713 if ( ( rc = cipher_setkey ( cipher, peerblk->cipherctx, peerblk->secret,
715 DBGC ( peerblk, "PEERBLK %p %d.%d could not set key: %s\n",
716 peerblk, peerblk->segment, peerblk->block,
725 * Parse retrieval protocol message segment and block details
727 * @v peerblk PeerDist block download
728 * @v buf_len Length of buffered data to fill in
729 * @ret rc Return status code
731 static int peerblk_parse_block ( struct peerdist_block *peerblk,
733 size_t digestsize = peerblk->digestsize;
734 peerblk_msg_blk_t ( digestsize, 0, 0, 0 ) *msg = peerblk->buffer.data;
735 size_t len = peerblk->buffer.len;
739 /* Check message length */
740 if ( len < offsetof ( typeof ( *msg ), msg.block.data ) ) {
741 DBGC ( peerblk, "PEERBLK %p %d.%d message too short for "
742 "zero-length data (%zd bytes)\n", peerblk,
743 peerblk->segment, peerblk->block, len );
747 /* Check digest size */
748 if ( ntohl ( msg->msg.segment.segment.digestsize ) != digestsize ) {
749 DBGC ( peerblk, "PEERBLK %p %d.%d incorrect digest size %d\n",
750 peerblk, peerblk->segment, peerblk->block,
751 ntohl ( msg->msg.segment.segment.digestsize ) );
755 /* Check segment ID */
756 if ( memcmp ( msg->msg.segment.id, peerblk->id, digestsize ) != 0 ) {
757 DBGC ( peerblk, "PEERBLK %p %d.%d segment ID mismatch\n",
758 peerblk, peerblk->segment, peerblk->block );
763 if ( ntohl ( msg->msg.index ) != peerblk->block ) {
764 DBGC ( peerblk, "PEERBLK %p %d.%d block ID mismatch (got %d)\n",
765 peerblk, peerblk->segment, peerblk->block,
766 ntohl ( msg->msg.index ) );
770 /* Check for missing blocks */
771 data_len = be32_to_cpu ( msg->msg.block.block.len );
773 DBGC ( peerblk, "PEERBLK %p %d.%d block not found\n",
774 peerblk, peerblk->segment, peerblk->block );
778 /* Check for underlength blocks */
779 if ( data_len < ( peerblk->range.end - peerblk->range.start ) ) {
780 DBGC ( peerblk, "PEERBLK %p %d.%d underlength block (%zd "
781 "bytes)\n", peerblk, peerblk->segment, peerblk->block,
786 /* Calculate buffered data length (i.e. excluding data which
787 * was delivered to the final data transfer buffer).
789 *buf_len = ( data_len - ( peerblk->end - peerblk->start ) );
791 /* Describe data before the trimmed content */
792 peerblk->decrypt[PEERBLK_BEFORE].xferbuf = &peerblk->buffer;
793 peerblk->decrypt[PEERBLK_BEFORE].offset =
794 offsetof ( typeof ( *msg ), msg.block.data );
795 peerblk->decrypt[PEERBLK_BEFORE].len =
797 offsetof ( typeof ( *msg ), msg.block.data ) );
798 total = peerblk->decrypt[PEERBLK_BEFORE].len;
800 /* Describe data within the trimmed content */
801 peerblk->decrypt[PEERBLK_DURING].offset =
802 peerblk_offset ( peerblk, peerblk->start );
803 peerblk->decrypt[PEERBLK_DURING].len =
804 ( peerblk->end - peerblk->start );
805 total += peerblk->decrypt[PEERBLK_DURING].len;
807 /* Describe data after the trimmed content */
808 peerblk->decrypt[PEERBLK_AFTER].xferbuf = &peerblk->buffer;
809 peerblk->decrypt[PEERBLK_AFTER].offset = peerblk->start;
810 peerblk->decrypt[PEERBLK_AFTER].len =
811 ( offsetof ( typeof ( *msg ), msg.block.data )
812 + *buf_len - peerblk->start );
813 total += peerblk->decrypt[PEERBLK_AFTER].len;
816 assert ( total == be32_to_cpu ( msg->msg.block.block.len ) );
818 /* Initialise cipher and digest lengths */
819 peerblk->cipher_remaining = total;
820 peerblk->digest_remaining =
821 ( peerblk->range.end - peerblk->range.start );
822 assert ( peerblk->cipher_remaining >= peerblk->digest_remaining );
828 * Parse retrieval protocol message useless details
830 * @v peerblk PeerDist block download
831 * @v buf_len Length of buffered data
832 * @v vrf_len Length of uselessness to fill in
833 * @ret rc Return status code
835 static int peerblk_parse_useless ( struct peerdist_block *peerblk,
836 size_t buf_len, size_t *vrf_len ) {
837 size_t digestsize = peerblk->digestsize;
838 peerblk_msg_blk_t ( digestsize, buf_len, 0, 0 ) *msg =
839 peerblk->buffer.data;
840 size_t len = peerblk->buffer.len;
842 /* Check message length */
843 if ( len < offsetof ( typeof ( *msg ), msg.vrf.data ) ) {
844 DBGC ( peerblk, "PEERBLK %p %d.%d message too short for "
845 "zero-length uselessness (%zd bytes)\n", peerblk,
846 peerblk->segment, peerblk->block, len );
850 /* Extract length of uselessness */
851 *vrf_len = be32_to_cpu ( msg->msg.vrf.vrf.len );
857 * Parse retrieval protocol message initialisation vector details
859 * @v peerblk PeerDist block download
860 * @v buf_len Length of buffered data
861 * @v vrf_len Length of uselessness
862 * @ret rc Return status code
864 static int peerblk_parse_iv ( struct peerdist_block *peerblk, size_t buf_len,
866 size_t digestsize = peerblk->digestsize;
867 size_t blksize = peerblk->cipher->blocksize;
868 peerblk_msg_blk_t ( digestsize, buf_len, vrf_len, blksize ) *msg =
869 peerblk->buffer.data;
870 size_t len = peerblk->buffer.len;
872 /* Check message length */
873 if ( len < sizeof ( *msg ) ) {
874 DBGC ( peerblk, "PEERBLK %p %d.%d message too short for "
875 "initialisation vector (%zd bytes)\n", peerblk,
876 peerblk->segment, peerblk->block, len );
880 /* Check initialisation vector size */
881 if ( ntohl ( msg->msg.iv.iv.blksize ) != blksize ) {
882 DBGC ( peerblk, "PEERBLK %p %d.%d incorrect IV size %d\n",
883 peerblk, peerblk->segment, peerblk->block,
884 ntohl ( msg->msg.iv.iv.blksize ) );
888 /* Set initialisation vector */
889 cipher_setiv ( peerblk->cipher, peerblk->cipherctx, msg->msg.iv.data );
895 * Read from decryption buffers
897 * @v peerblk PeerDist block download
898 * @v data Data buffer
899 * @v len Length to read
900 * @ret rc Return status code
902 static int peerblk_decrypt_read ( struct peerdist_block *peerblk,
903 void *data, size_t len ) {
904 struct peerdist_block_decrypt *decrypt = peerblk->decrypt;
908 /* Read from each decryption buffer in turn */
909 for ( ; len ; decrypt++, data += frag_len, len -= frag_len ) {
911 /* Calculate length to use from this buffer */
912 frag_len = decrypt->len;
913 if ( frag_len > len )
918 /* Read from this buffer */
919 if ( ( rc = xferbuf_read ( decrypt->xferbuf, decrypt->offset,
920 data, frag_len ) ) != 0 )
928 * Write to decryption buffers and update offsets and lengths
930 * @v peerblk PeerDist block download
931 * @v data Data buffer
932 * @v len Length to read
933 * @ret rc Return status code
935 static int peerblk_decrypt_write ( struct peerdist_block *peerblk,
936 const void *data, size_t len ) {
937 struct peerdist_block_decrypt *decrypt = peerblk->decrypt;
941 /* Write to each decryption buffer in turn */
942 for ( ; len ; decrypt++, data += frag_len, len -= frag_len ) {
944 /* Calculate length to use from this buffer */
945 frag_len = decrypt->len;
946 if ( frag_len > len )
951 /* Write to this buffer */
952 if ( ( rc = xferbuf_write ( decrypt->xferbuf, decrypt->offset,
953 data, frag_len ) ) != 0 )
956 /* Update offset and length */
957 decrypt->offset += frag_len;
958 decrypt->len -= frag_len;
965 * Decrypt one chunk of PeerDist retrieval protocol data
967 * @v peerblk PeerDist block download
969 static void peerblk_decrypt ( struct peerdist_block *peerblk ) {
970 struct cipher_algorithm *cipher = peerblk->cipher;
971 struct digest_algorithm *digest = peerblk->digest;
972 struct xfer_buffer *xferbuf;
979 assert ( ( PEERBLK_DECRYPT_CHUNKSIZE % cipher->blocksize ) == 0 );
981 /* Get the underlying data transfer buffer */
982 xferbuf = xfer_buffer ( &peerblk->xfer );
984 DBGC ( peerblk, "PEERBLK %p %d.%d has no underlying data "
985 "transfer buffer\n", peerblk, peerblk->segment,
988 goto err_xfer_buffer;
990 peerblk->decrypt[PEERBLK_DURING].xferbuf = xferbuf;
992 /* Calculate cipher and digest lengths */
993 cipher_len = PEERBLK_DECRYPT_CHUNKSIZE;
994 if ( cipher_len > peerblk->cipher_remaining )
995 cipher_len = peerblk->cipher_remaining;
996 digest_len = cipher_len;
997 if ( digest_len > peerblk->digest_remaining )
998 digest_len = peerblk->digest_remaining;
999 assert ( ( cipher_len & ( cipher->blocksize - 1 ) ) == 0 );
1001 /* Allocate temporary data buffer */
1002 data = malloc ( cipher_len );
1005 goto err_alloc_data;
1008 /* Read ciphertext */
1009 if ( ( rc = peerblk_decrypt_read ( peerblk, data, cipher_len ) ) != 0 ){
1010 DBGC ( peerblk, "PEERBLK %p %d.%d could not read ciphertext: "
1011 "%s\n", peerblk, peerblk->segment, peerblk->block,
1017 cipher_decrypt ( cipher, peerblk->cipherctx, data, data, cipher_len );
1019 /* Add data to digest */
1020 digest_update ( digest, peerblk->digestctx, data, digest_len );
1022 /* Write plaintext */
1023 if ( ( rc = peerblk_decrypt_write ( peerblk, data, cipher_len ) ) != 0){
1024 DBGC ( peerblk, "PEERBLK %p %d.%d could not write plaintext: "
1025 "%s\n", peerblk, peerblk->segment, peerblk->block,
1031 peerblk->cipher_remaining -= cipher_len;
1032 peerblk->digest_remaining -= digest_len;
1034 /* Free temporary data buffer */
1037 /* Continue processing until all input is consumed */
1038 if ( peerblk->cipher_remaining )
1041 /* Complete download attempt */
1042 peerblk_done ( peerblk, 0 );
1050 peerblk_done ( peerblk, rc );
1054 * Close PeerDist retrieval protocol block download attempt
1056 * @v peerblk PeerDist block download
1057 * @v rc Reason for close
1059 static void peerblk_retrieval_close ( struct peerdist_block *peerblk, int rc ) {
1063 /* Restart interface */
1064 intf_restart ( &peerblk->retrieval, rc );
1066 /* Fail immediately if we have an error */
1070 /* Abort download attempt (for testing) if applicable */
1071 if ( ( rc = inject_fault ( PEERBLK_ABORT_RATE ) ) != 0 )
1074 /* Parse message header */
1075 if ( ( rc = peerblk_parse_header ( peerblk ) ) != 0 )
1078 /* Parse message segment and block details */
1079 if ( ( rc = peerblk_parse_block ( peerblk, &buf_len ) ) != 0 )
1082 /* If the block was plaintext, then there is nothing more to do */
1083 if ( ! peerblk->cipher )
1086 /* Parse message useless details */
1087 if ( ( rc = peerblk_parse_useless ( peerblk, buf_len, &vrf_len ) ) != 0)
1090 /* Parse message initialisation vector details */
1091 if ( ( rc = peerblk_parse_iv ( peerblk, buf_len, vrf_len ) ) != 0 )
1094 /* Fail if decryption length is not aligned to the cipher block size */
1095 if ( peerblk->cipher_remaining & ( peerblk->cipher->blocksize - 1 ) ) {
1096 DBGC ( peerblk, "PEERBLK %p %d.%d unaligned data length %zd\n",
1097 peerblk, peerblk->segment, peerblk->block,
1098 peerblk->cipher_remaining );
1103 /* Stop the download attempt timer: there is no point in
1104 * timing out while decrypting.
1106 stop_timer ( &peerblk->timer );
1108 /* Start decryption process */
1109 process_add ( &peerblk->process );
1113 /* Complete download attempt */
1114 peerblk_done ( peerblk, rc );
1117 /******************************************************************************
1121 ******************************************************************************
1125 * Handle PeerDist retry timer expiry
1127 * @v timer Retry timer
1128 * @v over Failure indicator
1130 static void peerblk_expired ( struct retry_timer *timer, int over __unused ) {
1131 struct peerdist_block *peerblk =
1132 container_of ( timer, struct peerdist_block, timer );
1133 struct peerdisc_segment *segment = peerblk->discovery.segment;
1134 struct peerdisc_peer *head;
1135 unsigned long now = peerblk_timestamp();
1136 const char *location;
1139 /* Profile discovery timeout, if applicable */
1140 if ( ( peerblk->peer == NULL ) && ( timer->timeout != 0 ) ) {
1141 profile_custom ( &peerblk_discovery_timeout_profiler,
1142 ( now - peerblk->started ) );
1143 DBGC ( peerblk, "PEERBLK %p %d.%d discovery timed out after "
1144 "%ld ticks\n", peerblk, peerblk->segment,
1145 peerblk->block, timer->timeout );
1148 /* Profile download timeout, if applicable */
1149 if ( ( peerblk->peer != NULL ) && ( timer->timeout != 0 ) ) {
1150 profile_custom ( &peerblk_attempt_timeout_profiler,
1151 ( now - peerblk->attempted ) );
1152 DBGC ( peerblk, "PEERBLK %p %d.%d timed out after %ld ticks\n",
1153 peerblk, peerblk->segment, peerblk->block,
1157 /* Abort any current download attempt */
1158 peerblk_reset ( peerblk, -ETIMEDOUT );
1160 /* Record attempt start time */
1161 peerblk->attempted = now;
1163 /* If we have exceeded our maximum number of attempt cycles
1164 * (each cycle comprising a retrieval protocol download from
1165 * each peer in the list followed by a raw download from the
1166 * origin server), then abort the overall download.
1168 head = list_entry ( &segment->peers, struct peerdisc_peer, list );
1169 if ( ( peerblk->peer == head ) &&
1170 ( ++peerblk->cycles >= PEERBLK_MAX_ATTEMPT_CYCLES ) ) {
1176 /* If we have not yet made any download attempts, then move to
1177 * the start of the peer list.
1179 if ( peerblk->peer == NULL )
1180 peerblk->peer = head;
1182 /* Attempt retrieval protocol download from next usable peer */
1183 list_for_each_entry_continue ( peerblk->peer, &segment->peers, list ) {
1185 /* Attempt retrieval protocol download from this peer */
1186 location = peerblk->peer->location;
1187 if ( ( rc = peerblk_retrieval_open ( peerblk,
1188 location ) ) != 0 ) {
1189 /* Non-fatal: continue to try next peer */
1193 /* Start download attempt timer */
1194 peerblk->rc = -ETIMEDOUT;
1195 start_timer_fixed ( &peerblk->timer,
1196 PEERBLK_RETRIEVAL_OPEN_TIMEOUT );
1200 /* Attempt raw download */
1201 if ( ( rc = peerblk_raw_open ( peerblk ) ) != 0 )
1204 /* Start download attempt timer */
1205 peerblk->rc = -ETIMEDOUT;
1206 start_timer_fixed ( &peerblk->timer, PEERBLK_RAW_OPEN_TIMEOUT );
1210 peerblk_close ( peerblk, rc );
1214 * Handle PeerDist peer discovery
1216 * @v discovery PeerDist discovery client
1218 static void peerblk_discovered ( struct peerdisc_client *discovery ) {
1219 struct peerdist_block *peerblk =
1220 container_of ( discovery, struct peerdist_block, discovery );
1221 unsigned long now = peerblk_timestamp();
1223 /* Do nothing unless we are still waiting for the initial
1224 * discovery timeout.
1226 if ( ( peerblk->peer != NULL ) || ( peerblk->timer.timeout == 0 ) )
1229 /* Schedule an immediate retry */
1230 start_timer_nodelay ( &peerblk->timer );
1232 /* Profile discovery success */
1233 profile_custom ( &peerblk_discovery_success_profiler,
1234 ( now - peerblk->started ) );
1237 /******************************************************************************
1241 ******************************************************************************
1244 /** PeerDist block download data transfer interface operations */
1245 static struct interface_operation peerblk_xfer_operations[] = {
1246 INTF_OP ( intf_close, struct peerdist_block *, peerblk_close ),
1249 /** PeerDist block download data transfer interface descriptor */
1250 static struct interface_descriptor peerblk_xfer_desc =
1251 INTF_DESC ( struct peerdist_block, xfer, peerblk_xfer_operations );
1253 /** PeerDist block download raw data interface operations */
1254 static struct interface_operation peerblk_raw_operations[] = {
1255 INTF_OP ( xfer_deliver, struct peerdist_block *, peerblk_raw_rx ),
1256 INTF_OP ( intf_close, struct peerdist_block *, peerblk_raw_close ),
1259 /** PeerDist block download raw data interface descriptor */
1260 static struct interface_descriptor peerblk_raw_desc =
1261 INTF_DESC ( struct peerdist_block, raw, peerblk_raw_operations );
1263 /** PeerDist block download retrieval protocol interface operations */
1264 static struct interface_operation peerblk_retrieval_operations[] = {
1265 INTF_OP ( xfer_deliver, struct peerdist_block *, peerblk_retrieval_rx ),
1266 INTF_OP ( intf_close, struct peerdist_block *, peerblk_retrieval_close),
1269 /** PeerDist block download retrieval protocol interface descriptor */
1270 static struct interface_descriptor peerblk_retrieval_desc =
1271 INTF_DESC ( struct peerdist_block, retrieval,
1272 peerblk_retrieval_operations );
1274 /** PeerDist block download decryption process descriptor */
1275 static struct process_descriptor peerblk_process_desc =
1276 PROC_DESC ( struct peerdist_block, process, peerblk_decrypt );
1278 /** PeerDist block download discovery operations */
1279 static struct peerdisc_client_operations peerblk_discovery_operations = {
1280 .discovered = peerblk_discovered,
1284 * Open PeerDist block download
1286 * @v xfer Data transfer interface
1287 * @v uri Original URI
1288 * @v info Content information block
1289 * @ret rc Return status code
1291 int peerblk_open ( struct interface *xfer, struct uri *uri,
1292 struct peerdist_info_block *block ) {
1293 const struct peerdist_info_segment *segment = block->segment;
1294 const struct peerdist_info *info = segment->info;
1295 struct digest_algorithm *digest = info->digest;
1296 struct peerdist_block *peerblk;
1297 unsigned long timeout;
1301 /* Allocate and initialise structure */
1302 peerblk = zalloc ( sizeof ( *peerblk ) + digest->ctxsize );
1307 ref_init ( &peerblk->refcnt, peerblk_free );
1308 intf_init ( &peerblk->xfer, &peerblk_xfer_desc, &peerblk->refcnt );
1309 intf_init ( &peerblk->raw, &peerblk_raw_desc, &peerblk->refcnt );
1310 intf_init ( &peerblk->retrieval, &peerblk_retrieval_desc,
1312 peerblk->uri = uri_get ( uri );
1313 memcpy ( &peerblk->range, &block->range, sizeof ( peerblk->range ) );
1314 memcpy ( &peerblk->trim, &block->trim, sizeof ( peerblk->trim ) );
1315 peerblk->offset = ( block->trim.start - info->trim.start );
1316 peerblk->digest = info->digest;
1317 peerblk->digestsize = digestsize = info->digestsize;
1318 peerblk->digestctx = ( ( ( void * ) peerblk ) + sizeof ( *peerblk ) );
1319 peerblk->segment = segment->index;
1320 memcpy ( peerblk->id, segment->id, sizeof ( peerblk->id ) );
1321 memcpy ( peerblk->secret, segment->secret, sizeof ( peerblk->secret ) );
1322 peerblk->block = block->index;
1323 memcpy ( peerblk->hash, block->hash, sizeof ( peerblk->hash ) );
1324 xferbuf_malloc_init ( &peerblk->buffer );
1325 process_init_stopped ( &peerblk->process, &peerblk_process_desc,
1327 peerdisc_init ( &peerblk->discovery, &peerblk_discovery_operations );
1328 timer_init ( &peerblk->timer, peerblk_expired, &peerblk->refcnt );
1329 DBGC2 ( peerblk, "PEERBLK %p %d.%d id %02x%02x%02x%02x%02x..."
1330 "%02x%02x%02x [%08zx,%08zx)", peerblk, peerblk->segment,
1331 peerblk->block, peerblk->id[0], peerblk->id[1], peerblk->id[2],
1332 peerblk->id[3], peerblk->id[4], peerblk->id[ digestsize - 3 ],
1333 peerblk->id[ digestsize - 2 ], peerblk->id[ digestsize - 1 ],
1334 peerblk->range.start, peerblk->range.end );
1335 if ( ( peerblk->trim.start != peerblk->range.start ) ||
1336 ( peerblk->trim.end != peerblk->range.end ) ) {
1337 DBGC2 ( peerblk, " covers [%08zx,%08zx)",
1338 peerblk->trim.start, peerblk->trim.end );
1340 DBGC2 ( peerblk, "\n" );
1342 /* Open discovery */
1343 if ( ( rc = peerdisc_open ( &peerblk->discovery, peerblk->id,
1344 peerblk->digestsize ) ) != 0 )
1345 goto err_open_discovery;
1347 /* Schedule a retry attempt either immediately (if we already
1348 * have some peers) or after the discovery timeout.
1350 timeout = ( list_empty ( &peerblk->discovery.segment->peers ) ?
1351 ( peerdisc_timeout_secs * TICKS_PER_SEC ) : 0 );
1352 start_timer_fixed ( &peerblk->timer, timeout );
1354 /* Record start time */
1355 peerblk->started = peerblk_timestamp();
1357 /* Attach to parent interface, mortalise self, and return */
1358 intf_plug_plug ( xfer, &peerblk->xfer );
1359 ref_put ( &peerblk->refcnt );
1363 peerblk_close ( peerblk, rc );