| 1 |
#!/opt/bin/perl |
| 2 |
|
| 3 |
#system "echo $$ >fmd.pid"; |
| 4 |
#use Coro::Debug; our $server = new_unix_server Coro::Debug "debug"; |
| 5 |
|
| 6 |
=head1 fmd - the freenet mass downloader |
| 7 |
|
| 8 |
Fmd is at a very early stage of development (and very hackish, too), as I |
| 9 |
am learning the basics of freenet myself. |
| 10 |
|
| 11 |
However, I use it in production, and since it verifies everything it |
| 12 |
decodes etc., it seems to be quite safe to use, that is, if you know how |
| 13 |
to debug perl :) |
| 14 |
|
| 15 |
=head2 FEATURES |
| 16 |
|
| 17 |
- decoding is done "in place", i.e. all non-checkblock blocks are |
| 18 |
stored in-place and will not be moved, only checkblocks will |
| 19 |
be moved to their final position in the file before decoding. |
| 20 |
- high resistance against failures of fred or fmd |
| 21 |
- extremely persistent retry behaviour - there is no such thing |
| 22 |
as a permanent failure. |
| 23 |
- handles hundreds of simultaneous downloads with grace. |
| 24 |
|
| 25 |
=head2 ENVIRONMENT |
| 26 |
|
| 27 |
Set FMD_HOME to a directory (default ~/fmd) where fmd will store it's |
| 28 |
files. It will not store your porn files or other freenet data outside |
| 29 |
that directory. |
| 30 |
|
| 31 |
The subdirectory db will contain a database (soon to go away), while tmp |
| 32 |
contains queue files and partial splitfiles. All finished files will be |
| 33 |
moved to the done subdir. |
| 34 |
|
| 35 |
FREDHOST and FREDPORT do the obvious. Fix these docs if you disagree. |
| 36 |
|
| 37 |
Also edit the fmd executable for the number of threads and other |
| 38 |
non-useful constants. The default number (250) works for me, probably not |
| 39 |
for you. |
| 40 |
|
| 41 |
=head2 COMMANDS |
| 42 |
|
| 43 |
=over 4 |
| 44 |
|
| 45 |
=item CHK@... (space or "/") filename |
| 46 |
|
| 47 |
Just pasting a CHK and a filename sperated by one or more spaces or |
| 48 |
slashes will add files to the queue. The line can have leading garbage, |
| 49 |
i.e. you can paste full uris. |
| 50 |
|
| 51 |
=item <attachment>...</attachment> |
| 52 |
|
| 53 |
=item <attach>...</attach> |
| 54 |
|
| 55 |
=item <attached>...</attached> |
| 56 |
|
| 57 |
Will add frosts mentally deranged pseudo-xml format, soon to be replaced |
| 58 |
by something even more horrible. |
| 59 |
|
| 60 |
=item l |
| 61 |
|
| 62 |
List all jobs by number. |
| 63 |
|
| 64 |
=item <number> (optionally trailing command) |
| 65 |
|
| 66 |
Selects the job with the given number for further comamnds that require a current job. |
| 67 |
|
| 68 |
=item s |
| 69 |
|
| 70 |
Show the current job. Bot useful right now. |
| 71 |
|
| 72 |
=item k |
| 73 |
|
| 74 |
Kill the current job. This will keep the temp. file around. Sorry. (But it |
| 75 |
could be used to reconstruct the download.. hmm..) |
| 76 |
|
| 77 |
=item pri<number> |
| 78 |
|
| 79 |
Set the current job priority to <pri>. The default is 1. A job with higher |
| 80 |
priority will (on average) get more requests. |
| 81 |
|
| 82 |
The useful range is probably 0..100, but be careful and limit yourself to |
| 83 |
small values (<10), otherwise your other downloads might starve! |
| 84 |
|
| 85 |
=item q |
| 86 |
|
| 87 |
Kill the command prompt. Yupp. |
| 88 |
|
| 89 |
=back |
| 90 |
|
| 91 |
=cut |
| 92 |
|
| 93 |
use strict; |
| 94 |
|
| 95 |
use EV; |
| 96 |
use Coro::EV; |
| 97 |
|
| 98 |
use Net::FCP; |
| 99 |
use Storable (); |
| 100 |
use Time::HiRes; |
| 101 |
use Coro; |
| 102 |
use Coro::Channel; |
| 103 |
use Coro::Handle; |
| 104 |
use Coro::Signal; |
| 105 |
use Coro::Timer; |
| 106 |
use Coro::AIO; |
| 107 |
use Coro::Storable (); |
| 108 |
use List::Util; |
| 109 |
use Digest::SHA1; |
| 110 |
use Algorithm::FEC; |
| 111 |
use Digest::SHA1; |
| 112 |
use Net::FCP::Util; |
| 113 |
use POSIX (); |
| 114 |
|
| 115 |
Coro::AIO::max_poll_time 0.01; |
| 116 |
Coro::AIO::min_parallel 1; |
| 117 |
Coro::AIO::max_parallel 32; |
| 118 |
|
| 119 |
EV::set_io_collect_interval 0.2; |
| 120 |
EV::set_timeout_collect_interval 0.05; |
| 121 |
|
| 122 |
$|=1; |
| 123 |
|
| 124 |
our $MAX_TXN = 400; # use max. this many transactions in parallel |
| 125 |
our @HTL = (5,20); |
| 126 |
our $VERIFY_CHK = 1; # verify all blocks, again and again (only useful to debugging). |
| 127 |
our $FMD_HOME = $ENV{FMD_HOME} || "$ENV{HOME}/fmd"; |
| 128 |
|
| 129 |
our $FCP = new Net::FCP; |
| 130 |
|
| 131 |
defined $FMD_HOME |
| 132 |
or die "you currently must define FMD_HOME to a persistent directory"; |
| 133 |
|
| 134 |
mkdir $FMD_HOME, 0700; |
| 135 |
|
| 136 |
our $QUEUE_HOME = "$FMD_HOME/job"; |
| 137 |
mkdir $QUEUE_HOME, 0700; |
| 138 |
our $DATA_HOME = "$FMD_HOME/tmp"; |
| 139 |
mkdir $DATA_HOME, 0700; |
| 140 |
our $DONE_HOME = "$FMD_HOME/done"; |
| 141 |
mkdir $DONE_HOME, 0700; |
| 142 |
|
| 143 |
our %job; |
| 144 |
|
| 145 |
sub push_key { |
| 146 |
my ($key, $title) = @_; |
| 147 |
|
| 148 |
for my $job (values %job) { |
| 149 |
if ($job->{p}{key} eq $key) { |
| 150 |
warn "job $job->{id} already works on this, not adding"; |
| 151 |
return; |
| 152 |
} |
| 153 |
} |
| 154 |
|
| 155 |
my $job = job->new_from_key ($key, $title); |
| 156 |
warn "added as job $job->{id}\n"; |
| 157 |
$job; |
| 158 |
} |
| 159 |
|
| 160 |
sub cmdline { |
| 161 |
my ($i, $o) = @_; |
| 162 |
|
| 163 |
my $cmd = async { |
| 164 |
my $job; |
| 165 |
while (print $o "> " and defined ($_ = <$i>)) { |
| 166 |
chomp; |
| 167 |
if (/<attach(?:ment|ed|)>(.*) \* (CHK[^<]+)<\/attach/) { |
| 168 |
$job = push_key $2, $1; |
| 169 |
} elsif (/(CHK\@[a-zA-Z0-9,~\-]{54})[\/ ]+(.*?)\s*$/) { |
| 170 |
$job = push_key $1, $2; |
| 171 |
} elsif (s/^(\d+)//) { |
| 172 |
if ($job = $job{$1}) { |
| 173 |
# print $o delete $job->{log}; |
| 174 |
# if ($job->{input}) { |
| 175 |
# } |
| 176 |
} |
| 177 |
redo; |
| 178 |
} elsif (/^q/) { |
| 179 |
close $i; |
| 180 |
close $o; |
| 181 |
} elsif (/^l/) { |
| 182 |
for my $job (sort { $a->{id} <=> $b->{id} } values %job) { |
| 183 |
print $o "$_ $job->{id}: $job->{p}{key} $job->{p}{title} $job->{status}\n"; |
| 184 |
} |
| 185 |
# } elsif (/^pri\s*(\d+)/) { |
| 186 |
# if ($job) { |
| 187 |
# $job->{p}{pri} = $1; |
| 188 |
# $job->save; |
| 189 |
# } |
| 190 |
} elsif (/^s$/) { |
| 191 |
print $o $job->show if $job; |
| 192 |
} elsif (/^k$/) { |
| 193 |
$job->kill if $job; |
| 194 |
} elsif (/^p(.*)/) { |
| 195 |
print $o +(join " ", eval $1), "\n"; |
| 196 |
} elsif (/\S/) { |
| 197 |
# |
| 198 |
} else { |
| 199 |
print $o "?\n"; |
| 200 |
} |
| 201 |
for my $job (sort { $a->{id} <=> $b->{id} } values %job) { |
| 202 |
if ($job->{_input}) { |
| 203 |
print $o "> $job->{id} $job->{title} $job->{status}\n"; |
| 204 |
} |
| 205 |
} |
| 206 |
} |
| 207 |
}; |
| 208 |
} |
| 209 |
|
| 210 |
package job; |
| 211 |
|
| 212 |
use strict; |
| 213 |
|
| 214 |
use Coro; |
| 215 |
use Coro::AIO; |
| 216 |
use Fcntl; |
| 217 |
use IO::Handle; |
| 218 |
use Array::Heap2; |
| 219 |
|
| 220 |
our $count = 0; |
| 221 |
|
| 222 |
sub new { |
| 223 |
my $class = shift; |
| 224 |
|
| 225 |
my $self = bless { @_ }, $class; |
| 226 |
|
| 227 |
for my $job (values %job) { |
| 228 |
if ($job->{p}{key} eq $self->{p}{key}) { |
| 229 |
warn "job $job->{id} already works on this, not adding"; |
| 230 |
aio_unlink $self->{job}; |
| 231 |
return; |
| 232 |
} |
| 233 |
} |
| 234 |
|
| 235 |
$self->{p}{title} !~ /\// |
| 236 |
or die "$self->{p}{title}: malformed key\n"; |
| 237 |
|
| 238 |
$self->{p}{title} =~ s/\s+$//; |
| 239 |
$self->{p}{title} =~ s/\n/_/g; |
| 240 |
|
| 241 |
$self->{id} = ++$count; |
| 242 |
$self->{job} ||= "$QUEUE_HOME/" . Time::HiRes::time . ":$count.j"; |
| 243 |
|
| 244 |
$job{$self->{id}} = $self; |
| 245 |
|
| 246 |
$self->start; |
| 247 |
$self |
| 248 |
} |
| 249 |
|
| 250 |
sub new_from_key { |
| 251 |
my ($class, $key, $title) = @_; |
| 252 |
$class->new (p => { key => $key, title => $title, state => "examine" }); |
| 253 |
} |
| 254 |
|
| 255 |
sub new_from_file { |
| 256 |
my ($class, $path) = @_; |
| 257 |
$class->new (job => $path, p => Storable::retrieve $path); |
| 258 |
} |
| 259 |
|
| 260 |
sub save { |
| 261 |
my ($self) = @_; |
| 262 |
if (my $fh = aio_open "$self->{job}~", O_CREAT|O_TRUNC|O_WRONLY, 0600) { |
| 263 |
aio_write $fh, 0, undef, Coro::Storable::blocking_nfreeze $self->{p}, 0; |
| 264 |
aio_fsync $fh; |
| 265 |
aio_close $fh; |
| 266 |
aio_rename "$self->{job}~", $self->{job}; |
| 267 |
aio_pathsync $QUEUE_HOME; |
| 268 |
} |
| 269 |
} |
| 270 |
|
| 271 |
sub clean { |
| 272 |
my ($self) = @_; |
| 273 |
|
| 274 |
delete $job{$self->{id}}; |
| 275 |
$self->save; |
| 276 |
system "mv", $self->{job}, "$DONE_HOME/$self->{p}{title}.job"; |
| 277 |
unlink $self->{job}; |
| 278 |
} |
| 279 |
|
| 280 |
sub kill { |
| 281 |
my ($self) = @_; |
| 282 |
|
| 283 |
$self->clean; |
| 284 |
$self->{coro}->cancel; |
| 285 |
} |
| 286 |
|
| 287 |
our @queue; |
| 288 |
our $queue_change = new Coro::Signal; |
| 289 |
our $queue_alloc = 0; |
| 290 |
|
| 291 |
async { |
| 292 |
for (;;) { |
| 293 |
while (@queue |
| 294 |
and (($queue[0][0] > 10 and $queue_alloc < $MAX_TXN) |
| 295 |
or ($queue[0][0] > 1 and $queue_alloc < $MAX_TXN - 3) |
| 296 |
or $queue_alloc < $MAX_TXN - 5)) { |
| 297 |
(pop_heap @queue)->[1]->send; |
| 298 |
$queue_alloc++; |
| 299 |
Coro::Timer::sleep 0.05; |
| 300 |
} |
| 301 |
$queue_change->wait; |
| 302 |
} |
| 303 |
}; |
| 304 |
|
| 305 |
sub txn_begin { |
| 306 |
my ($pri) = @_; |
| 307 |
my $sig = new Coro::Signal; |
| 308 |
|
| 309 |
#warn "txn_begin $pri\n";#d# |
| 310 |
push_heap @queue, [$pri, $sig]; |
| 311 |
$queue_change->send; |
| 312 |
$sig->wait; |
| 313 |
} |
| 314 |
|
| 315 |
sub txn_end { |
| 316 |
$queue_alloc--; |
| 317 |
$queue_change->send; |
| 318 |
} |
| 319 |
|
| 320 |
sub txn_client_get { |
| 321 |
my %arg = @_; |
| 322 |
|
| 323 |
txn_begin $arg{pri}; |
| 324 |
$FCP->txn_client_get ($arg{uri}, $arg{htl})->cb (unblock_sub { |
| 325 |
txn_end; |
| 326 |
|
| 327 |
$arg{cb}->(@_); |
| 328 |
}); |
| 329 |
} |
| 330 |
|
| 331 |
sub fetch_uri { |
| 332 |
my ($pri, $uri) = @_; |
| 333 |
|
| 334 |
for(my $count = 1; ; $count += 0.3) { |
| 335 |
for my $htl (@HTL) { |
| 336 |
txn_begin time + $htl + $count; |
| 337 |
my ($meta, $data) = eval { @{ $FCP->client_get ($uri, $htl) } }; |
| 338 |
txn_end; |
| 339 |
if ($@) { |
| 340 |
if (UNIVERSAL::isa ($@, Net::FCP::Exception::)) { |
| 341 |
if ($@->type ("data_not_found") |
| 342 |
|| $@->type ("route_not_found")) { |
| 343 |
next; |
| 344 |
} |
| 345 |
if ($@->type ("short_data")) { |
| 346 |
warn "(short_data, redo)\n"; |
| 347 |
redo; |
| 348 |
} |
| 349 |
die; |
| 350 |
} |
| 351 |
} |
| 352 |
if (defined $data) { |
| 353 |
return ($meta, $data); |
| 354 |
} |
| 355 |
} |
| 356 |
} |
| 357 |
|
| 358 |
die; |
| 359 |
} |
| 360 |
|
| 361 |
sub log { |
| 362 |
my ($self, $text) = @_; |
| 363 |
my $time = POSIX::strftime "%H:%M:%S", localtime time; |
| 364 |
warn "$time $self->{id},$self->{p}{pri}: $text\n"; |
| 365 |
$self->{p}{log} .= "$time $text\n"; |
| 366 |
$self->save; |
| 367 |
} |
| 368 |
|
| 369 |
sub feedback { |
| 370 |
my ($self, $prompt) = @_; |
| 371 |
$self->{input} = [$Coro::current, $prompt]; |
| 372 |
Coro::schedule; |
| 373 |
} |
| 374 |
|
| 375 |
sub show { |
| 376 |
my ($self) = @_; |
| 377 |
|
| 378 |
"\n$self->{p}{log}\n" |
| 379 |
. "ID: $self->{id}\n" |
| 380 |
. "Title: $self->{p}{title}\n" |
| 381 |
. "Blocks#: " . @{$self->{p}{blk} || []} . "\n" |
| 382 |
. "Blocks: " . (join "", map { |
| 383 |
$_->{done} ? "+" : "-" |
| 384 |
} @{$self->{p}{blk} || []}) . "\n" . |
| 385 |
"" |
| 386 |
} |
| 387 |
|
| 388 |
our $id; |
| 389 |
|
| 390 |
sub MAXSEG (){ 128*1024*1024 } |
| 391 |
sub MINSEG (){ 6* 128*1024 } |
| 392 |
|
| 393 |
sub blocksize($) { |
| 394 |
return |
| 395 |
$_[0] >= 64*1024*1024 ? 1024*1024 |
| 396 |
: $_[0] >= 32*1024*1024 ? 512*1024 |
| 397 |
: $_[0] >= 1024*1024 ? 256*1024 |
| 398 |
: 128*1024; |
| 399 |
} |
| 400 |
|
| 401 |
sub start { |
| 402 |
my ($self) = @_; |
| 403 |
|
| 404 |
$self->{p}{pri} ||= 1; |
| 405 |
$self->{p}{state} ||= "examine"; |
| 406 |
|
| 407 |
$self->{job} =~ /\/([^\/]*)\.j$/ or die "$self->{job}: missing .j"; |
| 408 |
$self->{file} = "$DATA_HOME/$1.d"; |
| 409 |
sysopen $self->{fh}, $self->{file}, O_RDWR|O_CREAT, 0600 |
| 410 |
#$self->{fh} = aio_open $self->{file}, O_RDWR|O_CREAT, 0600 |
| 411 |
or die "$self->{file}: $!"; |
| 412 |
|
| 413 |
$self->{status} = "starting"; |
| 414 |
$self->{coro} = async { |
| 415 |
$self->save; |
| 416 |
|
| 417 |
for(;;) { |
| 418 |
my ($state, @args) = ref $self->{p}{state} ? @{$self->{p}{state}} : $self->{p}{state}; |
| 419 |
my $next = eval { $self->can ("state_$state")->($self, @args) }; |
| 420 |
if ($@) { |
| 421 |
$self->log ($@); |
| 422 |
$next = $self->feedback ("continue with state: "); |
| 423 |
} |
| 424 |
$self->log ($self->{status} = "STATE CHANGE: ". join ", ", ref $next ? @$next : $next); |
| 425 |
$self->{p}{state} = $next; |
| 426 |
$self->save; |
| 427 |
} |
| 428 |
}; |
| 429 |
} |
| 430 |
|
| 431 |
sub state_finish { |
| 432 |
my ($self, $save) = @_; |
| 433 |
|
| 434 |
if ($save) { |
| 435 |
aio_fsync $self->{fh}; |
| 436 |
close $self->{fh}; |
| 437 |
|
| 438 |
aio_unlink "$DONE_HOME/$self->{p}{title}"; |
| 439 |
aio_link $self->{file}, "$DONE_HOME/$self->{p}{title}" |
| 440 |
and die "link: $self->{file} => $DONE_HOME/$self->{p}{title}: $!"; |
| 441 |
aio_pathsync $DONE_HOME; |
| 442 |
} |
| 443 |
$self->clean; |
| 444 |
|
| 445 |
aio_unlink $self->{file}; |
| 446 |
|
| 447 |
$self->{status} = "finished"; |
| 448 |
$self->feedback ("finished"); |
| 449 |
terminate; |
| 450 |
} |
| 451 |
|
| 452 |
sub state_examine { |
| 453 |
my ($self) = @_; |
| 454 |
my $p = $self->{p}; |
| 455 |
|
| 456 |
$self->{status} = "initial fetch"; |
| 457 |
|
| 458 |
for (;;) { |
| 459 |
$self->log ("fetching $p->{key} (=$p->{title})"); |
| 460 |
|
| 461 |
($p->{meta}, $p->{data}) = fetch_uri 100, "freenet:$p->{key}"; |
| 462 |
$self->save; |
| 463 |
#use PApp::Util; print STDERR PApp::Util::dumpval [keys %{$meta->{document}[0]{split_file}}]; |
| 464 |
$self->log ("type $p->{meta}{document}[0]{info}{format}"); |
| 465 |
|
| 466 |
if (my $splitfile = $p->{meta}{document}[0]{split_file}) { |
| 467 |
return "splitfile"; |
| 468 |
} elsif ((defined $p->{data}) and (length $p->{data})) { |
| 469 |
syswrite $self->{fh}, delete $p->{data}; |
| 470 |
aio_fsync $self->{fh}; |
| 471 |
return ["finish", 1]; |
| 472 |
} |
| 473 |
|
| 474 |
$self->log ("EMPTY, retrying in an hour"); |
| 475 |
Coro::Timer::sleep 3600; |
| 476 |
} |
| 477 |
|
| 478 |
} |
| 479 |
|
| 480 |
sub state_splitfile { |
| 481 |
my ($self) = @_; |
| 482 |
my $p = $self->{p}; |
| 483 |
|
| 484 |
my $splitfile = $p->{meta}{document}[0]{split_file}; |
| 485 |
my $filesize = hex $splitfile->{size}; |
| 486 |
|
| 487 |
if ($splitfile->{algo_name} eq "OnionFEC_a_1_2") { |
| 488 |
my $data_packets = hex $splitfile->{block_count}; |
| 489 |
my $check_packets = hex $splitfile->{check_block_count}; |
| 490 |
|
| 491 |
my $blk = ($p->{blk} ||= []); |
| 492 |
|
| 493 |
unless (@$blk) { |
| 494 |
for (1..$data_packets) { |
| 495 |
push @$blk, { |
| 496 |
uri => $splitfile->{block}{sprintf "%x", $_}, |
| 497 |
}; |
| 498 |
} |
| 499 |
for (1..$check_packets) { |
| 500 |
push @$blk, { |
| 501 |
uri => $splitfile->{check_block}{sprintf "%x", $_}, |
| 502 |
}; |
| 503 |
} |
| 504 |
} |
| 505 |
|
| 506 |
my @segments; |
| 507 |
my $segments = 0; |
| 508 |
|
| 509 |
{ |
| 510 |
# that is a horrible algorithm :(, these freenet freaks are... java-disabled |
| 511 |
# hardcoding lots of magic parameters is soo dumb. |
| 512 |
my $size = $filesize; |
| 513 |
my $offset = 0; |
| 514 |
my $offset2 = ($filesize & ~(1024*1024-1)) + 1024*1024; # leave enough space after last data block |
| 515 |
my $idx = 0; |
| 516 |
my $idx2 = $data_packets; |
| 517 |
my @redundandy = (0,1,2); # maybe OnionFAC_a_1_2 means 1/2 redundancy(?) |
| 518 |
|
| 519 |
while ($size > 0) { |
| 520 |
my $segsize = $size >= MAXSEG ? MAXSEG : $size <= MINSEG ? MINSEG : $size; |
| 521 |
my $blksize = blocksize $segsize; |
| 522 |
my $seg = |
| 523 |
{ |
| 524 |
id => $segments++, |
| 525 |
todo => int (($segsize + $blksize - 1) / $blksize), |
| 526 |
done => 0, |
| 527 |
blk => [], |
| 528 |
blksize => $blksize, |
| 529 |
}; |
| 530 |
|
| 531 |
push @segments, $seg; |
| 532 |
$size -= $segsize; |
| 533 |
|
| 534 |
while ($segsize > 0) { |
| 535 |
push @{$seg->{blk}}, $idx; |
| 536 |
for ($blk->[$idx++]) { |
| 537 |
$_->{offset} = $offset; |
| 538 |
#$_->{size} = $blksize > $segsize ? $segsize : $blksize; # WRONG |
| 539 |
$_->{size} = $blksize; |
| 540 |
$_->{seg} = $seg; |
| 541 |
} |
| 542 |
|
| 543 |
$segsize -= $blksize; |
| 544 |
$offset += $blksize; |
| 545 |
|
| 546 |
if (($redundandy[0] += $redundandy[1]) >= $redundandy[2]) { |
| 547 |
$redundandy[0] -= $redundandy[2]; |
| 548 |
|
| 549 |
push @{$seg->{blk}}, $idx2; |
| 550 |
for ($blk->[$idx2++]) { |
| 551 |
$_->{offset} = $offset2; |
| 552 |
$_->{size} = $blksize; |
| 553 |
$_->{seg} = $seg; |
| 554 |
} |
| 555 |
|
| 556 |
$offset2 += $blksize; |
| 557 |
} |
| 558 |
} |
| 559 |
} |
| 560 |
|
| 561 |
$idx == $data_packets |
| 562 |
or die "$self->{id}/$p->{tile} $self->{job}\nidx $idx != data_packets $data_packets"; |
| 563 |
$idx2 == $data_packets + $check_packets |
| 564 |
or die "$self->{id}/$p->{tile} $self->{job}\nidx2 $idx2 != data_packets $data_packets + check_packets $check_packets"; |
| 565 |
} |
| 566 |
|
| 567 |
for (@$blk) { |
| 568 |
++$_->{seg}{done} if $_->{done}; |
| 569 |
delete $_->{htl}; |
| 570 |
} |
| 571 |
|
| 572 |
my $fail = 0; |
| 573 |
my $sig = new Coro::Signal; |
| 574 |
|
| 575 |
$self->{status} = "splitfile fetch (" . @$blk . " blocks)"; |
| 576 |
|
| 577 |
my @txn; |
| 578 |
|
| 579 |
for (;;) { |
| 580 |
for my $id (0 .. $#$blk) { |
| 581 |
my $blk = $blk->[$id]; |
| 582 |
|
| 583 |
next if $txn[$id] || $blk->{done} || $blk->{seg}{todo} <= $blk->{seg}{done}; |
| 584 |
|
| 585 |
my $htl = $HTL[$blk->{htl}++ % @HTL]; |
| 586 |
my $pri = int time + $htl / 20 * 7200 * rand; |
| 587 |
|
| 588 |
#warn $self->{id} . ", GET<$htl, $pri>\n";#d# |
| 589 |
$txn[$id] ||= txn_client_get pri => $pri, uri => $blk->{uri}, htl => $htl, cb => sub { |
| 590 |
undef $txn[$id]; |
| 591 |
|
| 592 |
my $seg = $blk->{seg}; |
| 593 |
|
| 594 |
my ($meta, $data) = eval { @{ $_[0]->result } }; |
| 595 |
|
| 596 |
if (defined $data) { |
| 597 |
$blk->{size} == length $data |
| 598 |
or die sprintf "block $id expected size %d, got %d\n", $blk->{size}, length $data; |
| 599 |
|
| 600 |
(length $data) == (aio_write $self->{fh}, $blk->{offset}, (length $data), $data, 0) |
| 601 |
or die "unable to write chunk to disk, not setting valid flag"; |
| 602 |
aio_fsync $self->{fh}; |
| 603 |
|
| 604 |
$blk->{done} = 1; |
| 605 |
$blk->{meta} = $meta->{raw} if length $meta->{raw};#d# |
| 606 |
$seg->{done}++; |
| 607 |
$self->save; |
| 608 |
|
| 609 |
$::htl_sum += $htl; |
| 610 |
$::htl_cnt++; |
| 611 |
|
| 612 |
$self->log (sprintf "got block $seg->{id}.$id %d ($seg->{done}/$seg->{todo}) at htl $htl (%f) and pri $pri", |
| 613 |
length $data, $::htl_sum / $::htl_cnt); |
| 614 |
} else { |
| 615 |
if ($@) { |
| 616 |
if ($@->type ("data_not_found")) { |
| 617 |
# nop |
| 618 |
} elsif ($@->type ("network_error")) { |
| 619 |
$self->log ("$@, retrying in 1s"); |
| 620 |
CORE::sleep 1; |
| 621 |
} else { |
| 622 |
$self->log ("$@"); |
| 623 |
} |
| 624 |
} |
| 625 |
++$fail; |
| 626 |
} |
| 627 |
$self->{status} = "splitfile fetch ($seg->{done}/$seg->{todo}, $fail failed)"; |
| 628 |
|
| 629 |
$sig->send; |
| 630 |
}; |
| 631 |
} |
| 632 |
|
| 633 |
for my $seg (@segments) { |
| 634 |
if ($seg->{done} >= $seg->{todo} && !$seg->{finished}) { |
| 635 |
|
| 636 |
$self->log ("segment done, cancelling segment $seg->{id}"); |
| 637 |
for my $id (@{$seg->{blk}}) { |
| 638 |
(delete $txn[$id])->cancel if $txn[$id]; |
| 639 |
} |
| 640 |
|
| 641 |
my $verify; |
| 642 |
for my $id (@{$seg->{blk}}) { |
| 643 |
my $blk = $blk->[$id]; |
| 644 |
if ($blk->{done}) { |
| 645 |
aio_read $self->{fh}, $blk->{offset}, $blk->{size}, my $buf, 0; |
| 646 |
|
| 647 |
my $k1 = Net::FCP::Util::extract_chk_hash $blk->{uri}; |
| 648 |
my $k2 = Net::FCP::Util::generate_chk_hash $blk->{meta}, $buf; |
| 649 |
|
| 650 |
if ($k1 ne $k2) { |
| 651 |
$verify .= "v"; |
| 652 |
#warn sprintf "$p->{title} block $id BROKEN (%s != %s)", (unpack "H*", $k1), (unpack "H*", $k2); |
| 653 |
$blk->{done} = 0; |
| 654 |
$seg->{done}--; |
| 655 |
$self->save; |
| 656 |
} else { |
| 657 |
$verify .= "V"; |
| 658 |
} |
| 659 |
} |
| 660 |
} |
| 661 |
$self->log ("verified segment $seg->{id}"); |
| 662 |
$self->log ($verify); |
| 663 |
|
| 664 |
if ($seg->{done} >= $seg->{todo} && !$seg->{finished}) { |
| 665 |
$self->log ("verified segment OK $seg->{id}"); |
| 666 |
$seg->{finished}++; |
| 667 |
$segments--; |
| 668 |
} else { |
| 669 |
$self->log ("verified segment NOT OK $seg->{id}"); |
| 670 |
} |
| 671 |
} |
| 672 |
} |
| 673 |
|
| 674 |
last unless $segments; |
| 675 |
|
| 676 |
$sig->wait; |
| 677 |
} |
| 678 |
|
| 679 |
$self->log ("decoding < $self->{job} $self->{file} $filesize >"); |
| 680 |
|
| 681 |
for my $seg (@segments) { |
| 682 |
my @part; |
| 683 |
my @idx; |
| 684 |
my @blk = map $blk->[$_], sort { $a <=> $b } @{$seg->{blk}}; |
| 685 |
|
| 686 |
for my $id (0 .. $#blk) { |
| 687 |
my $blk = $blk[$id]; |
| 688 |
next unless $blk->{done}; |
| 689 |
|
| 690 |
push @part, [$self->{fh}, $blk->{offset}]; |
| 691 |
push @idx, $id; |
| 692 |
|
| 693 |
last if @idx == $seg->{todo}; |
| 694 |
} |
| 695 |
|
| 696 |
my $fec = new Algorithm::FEC |
| 697 |
$seg->{todo}, |
| 698 |
scalar @blk, |
| 699 |
$seg->{blksize}; |
| 700 |
|
| 701 |
$fec->shuffle (\@part, \@idx); |
| 702 |
|
| 703 |
# now copy check blocks to their destination position |
| 704 |
for my $i (0 .. $#idx) { |
| 705 |
next if $idx[$i] == $i; |
| 706 |
|
| 707 |
my $src = $part[$i]; |
| 708 |
$part[$i] = [$self->{fh}, $blk[$i]{offset}]; |
| 709 |
$fec->copy ($src, $part[$i]); |
| 710 |
} |
| 711 |
|
| 712 |
$fec->set_decode_blocks (\@part, \@idx); |
| 713 |
$fec->decode; |
| 714 |
} |
| 715 |
|
| 716 |
my $sha1 = new Digest::SHA1; |
| 717 |
open my $dd, "-|", "head -c$filesize \Q$self->{file}\E" |
| 718 |
or die "DD: $!"; |
| 719 |
#$dd = Coro::Handle::unblock $dd; |
| 720 |
$sha1->addfile ($dd); |
| 721 |
$sha1 = $sha1->hexdigest; |
| 722 |
|
| 723 |
if (exists $p->{meta}{document}[0]{info}{checksum} |
| 724 |
and $p->{meta}{document}[0]{info}{checksum} ne $sha1) { |
| 725 |
$self->log ("META: $p->{meta}{document}[0]{info}{checksum} and real checksum $sha1 for $filesize DIFFER"); |
| 726 |
# $self->feedback ("CHECKSUM ERROR"); |
| 727 |
# terminate; |
| 728 |
} |
| 729 |
|
| 730 |
truncate $self->{fh}, $filesize; |
| 731 |
sysseek $self->{fh}, 0, 0; |
| 732 |
|
| 733 |
return ["finish", 1]; |
| 734 |
} else { |
| 735 |
$self->log ("splitfile algo '$splitfile->{algo_name}' unknown"); |
| 736 |
$self->feedback ("algo unknown"); |
| 737 |
terminate; |
| 738 |
} |
| 739 |
} |
| 740 |
|
| 741 |
package main; |
| 742 |
|
| 743 |
$|=1; |
| 744 |
|
| 745 |
for (<\Q$QUEUE_HOME\E/*.j>) { |
| 746 |
job->new_from_file ($_); |
| 747 |
print "J"; |
| 748 |
} |
| 749 |
print "\n"; |
| 750 |
|
| 751 |
open my $stdin , "<&0" or die; |
| 752 |
open my $stdout, ">&1" or die; |
| 753 |
cmdline unblock $stdin, unblock $stdout; |
| 754 |
|
| 755 |
EV::loop; |
| 756 |
|
