****************** Release of 1.4.8.fsl.0.8 ************************ 2010-12-30 16:43 vass * configure.ac: Updating the version to 0.8. 2010-12-30 16:43 vass * workloads/Makefile.am: Removing deleted workloads from a Makefile. 2010-12-30 16:33 vass * workloads/postmark.f: postmark = varmail. 2010-12-30 15:41 vass * workloads/deletefiles.f: Deletefiles is the same as filemicro_delete.f 2010-12-30 15:14 vass * workloads/bringover.f: Bringover workload is the same as copyfiles workload. Remove it. 2010-12-30 14:43 vass * parser_gram.y: Forgot to commit parser_gram.y in the previous commit. 2010-12-29 19:42 vass * flowop.c, ipc.c, ipc.h, procflow.c, procflow.h, threadflow.c: Refactoring the process spanning procedure. 2010-12-29 15:12 gyumin * flowop.c, ipc.c, ipc.h: Restored shm_flowop_find_lock (flag -> rwlock) 2010-12-29 13:35 vass * TODO: Updated TODO. 2010-12-29 12:45 vass * ipc.c: Updating a comment about filebench_log() initialization. 2010-12-24 14:02 gyumin * Makefile.am, filebench.h, flag.h, flowop.c, ipc.c, ipc.h, parser_gram.y, procflow.c: Replaced pthread_rwlock to flag. Look for USE_RWLOCK features. 2010-12-23 12:08 gyumin * ipc.c: Bug-fix: Segmentation Fault on FreeBSD 2010-12-19 17:10 vass * ipc.c: Initialize _msg_mutex before we use it in filebench_log(). BUG reported by Gyumin Sim. 2010-12-11 18:48 vass * ipc.c, parser_gram.y: Displaying the shared memory region size allocated by Filebench. 2010-12-11 17:43 vass * filebench.h, misc.c, misc.h, parser_gram.y: we need to set rlimite per process, not only in master process. filebench_init() renamed to clock_init() because that's what it actually does. 2010-12-11 17:17 vass * fileset.c: Increasing the precision of fileset size output. 2010-11-09 14:25 gyumin * fb_localfs.c: Fixed a compile warning on OpenSolaris 2010-11-09 13:57 gyumin * stats.c: Fixed a compile error on OpenSolaris 2010-11-01 17:34 vass * TODO, misc.c, parser_gram.y: use shared memory only after we have attached to it. 2010-11-01 15:47 vass * TODO, filebench.h, flowop.c, misc.c, parser_gram.y, parser_lex.l, parsertypes.h, utils.h: Multiple changes: 1) main() revised 2) variables: some removed from shared region, some changed to static 3) -noproc option removed. 2010-10-29 16:08 vass * README_COND_COMP, configure.ac, fb_localfs.c, ipc.c, parser_gram.y, procflow.c: Removing Thread Model support, it is outdated. 2010-10-29 14:08 vass * README_COND_COMP, configure.ac, eventgen.c, ipc.c, ipc.h, parser_gram.y, procflow.c, stats.c: Multiple fixes. Some of them: 1) only use mkstemp() function. tmpnam() is outdated. 2) fix statistics output (redundant "\n" was added when we added min-max latency support) 3) ipc_init() and ipc_attach() refactoring 4) Use LOG_FATAL in case of errors in touched code. 5) synchonize workers' forks via rwlock, not via conditional variable 6) eliminating caddr_t in touched code 7) check for MAP_FAILED return in mmap calls and check that the returned virtual address is fixed accross all the processes 8) pass address between processes using "%p". This allows to eliminate dependency on the bitness (32 vs 64) of the architecture 2010-10-04 18:12 vass * TODO: TODO updated. 2010-09-28 15:29 vass * stats.c: Fixing the conversion from nanoseconds to miliseconds. 2010-09-27 14:50 vass * flowop.c, stats.c, stats.h: Adding min and max latency per-flowop to FB. 2010-09-27 13:27 vass * stats.c, stats.h: Setting osprof bucket number to 64 as a compile-time constant. 2010-09-27 12:36 vass * flowop.c, flowop.h, procflow.h, stats.c, stats.h, threadflow.h: Replacing 'flowstat_t' by 'struct flowstats' 2010-09-25 10:54 vass * filebench.h, flowop.c, ipc.h, parser_gram.y, stats.c: Almost any global parameter in FB must be stored in filebench_shm structure, not in global variable. If one stores global parameter in a global variable, then when main FB process forkes+execs new processes, global variable's value in the new processes is not initialized. This was the case with osprof_enabled variable. Fixed now. 2010-09-12 01:38 vass * Makefile.am, TODO, flowop.c, ioprio.c, ioprio.h, parser_gram.y, parser_lex.l, threadflow.h: Now one can specify ioprio for each Filebench thread in the workload file. 2010-08-24 19:16 vass * configure.ac: Update FB version to 0.7. 2010-08-24 19:13 vass * ChangeLog: Updating ChangeLog. ****************** Release of 1.4.8.fsl.0.7 ************************ 2010-08-24 19:09 vass * NEWS: Updating the NEWS file for the fresh FB release. 2010-07-13 23:45 vass * fb_random.c: fb_urandom() must return value less or equal than (max - round). 2010-07-10 23:14 vass * ipc.c, ipc.h: Arranging ipc_malloc/ipc_free code in a consisten way. 2010-07-10 20:57 vass * fileset.c: filesizegamma was used instead of dirsizegamma, resulting in wrong directory depth. E.g., setting dirsizegamma to 0, haven't had an expected result of fixed number of files in every dir. 2010-07-10 20:33 vass * ipc.h, procflow.c: Moving all main FB object lists & locks to the opening part of filbench_shm. Renaming proclist to procflowlist, as it is supposed to be. 2010-07-10 20:10 vass * ipc.h: seting shm_marker size to 0, since it is only used to mark zeroed region. 2010-07-10 19:55 vass * auto_comp.h, eventgen.h, fb_avl.h, filebench.h, fileset.h, flowop.h, gamma_dist.h, ipc.h, misc.h, parsertypes.h, procflow.h, stats.h, threadflow.h, utils.h: removing annoying #ifdef __cplusplus ****************** Release of 1.4.8.fsl.0.6 ************************ 2010-06-18 17:52 vass * NEWS, README, configure.ac: Update version to 0.6, posset feature description, NEWS updated. 2010-06-18 17:19 vass * flowop.c, flowop_library.c, parser_gram.y, posset.c, posset.h: First version of possition set feature is completed. 2010-06-18 02:34 vass * TODO, fb_random.c, flowop.c, flowop.h, ipc.c, ipc.h, parser_gram.y, parser_lex.l, posset.c, posset.h: Propagating the posssetname to each flowop instance. 2010-06-17 21:24 vass * fb_random.c, ipc.c, parser_gram.y, posset.c, posset.h: some bug fixes related to previous commit. 2010-06-17 19:19 vass * TODO, fb_localfs.c, fb_random.c, fb_random.h, filebench.h, fileset.c, flowop_library.c, misc.c, parser_gram.y, posset.c, posset.h: 1) functions related to /dev/urandom reworked 2) more code for posset support. 2010-06-16 21:57 vass * ipc.c, ipc.h, posset.c, posset.h, vars.h: 1) ipc_malloc() function now gets the number of preallocated entries through the new preallocated_entries() function 2) posset_alloc() function allocates posset and sets up all the fields in posset. 2010-06-14 00:53 vass * procflow.c: Patch from Evgeny Budilovsky that fixes deadlock when spanning process instances. Observed especially on Ubuntu, but possible everythere. Citation from Evgeny's e-mail: "I think I discovered what is causing the deadlock in the code. The problem was in "procflow_init" function which creates thread "procflow_createnwait" which is responsible to notify condition on which "procflow_init" waits next. The problem is that the notify happens before the first thread has chance to call wait, since there is no proper locking." ****************** Release of 1.4.8.fsl.0.5 ************************ 2010-04-06 20:24 vass * configure.ac: Updated to version 0.5 2010-04-06 20:22 vass * Makefile.am, configure.ac, filebench.h: Fixes to be compilable by SUN Studio compilers. ****************** Release of 1.4.8.fsl.0.4 ************************ 2010-04-05 16:40 vass * configure.ac: Updating the version to 0.4 2010-04-05 16:36 vass * Makefile.am, README_COND_COMP, TODO, configure.ac, fb_localfs.c, filebench.h, fileset.c, flowop.c, flowop_library.c, ipc.c, misc.c, parser_gram.y, procflow.c: - After switching to PROCESS_MODEL we need to make certain hacks on FreeBSD: a) emulate waitid() by waitpid() b) no posix_fadvise() support, so just print a warning c) no PROCESS_SHARED attribute for pthread objects. Not considering it as a fatal problem now, just printing a warning. If we have multiple processes (not threads) in a workload, we might have problems because of that. - Use directio() on Solaris to perform direct I/O - compilation warnings 2010-04-01 17:39 vass * flowop_library.c: Enabling O_DIRECT and readahead options for raw devices. 2010-03-31 23:33 psehgal * fileset.c, flowop.h, flowop_library.c, parser_gram.y, parser_lex.l: Added code to advise the kernel to switch off read aheads 2010-03-31 18:29 vass * stats.c: Incraesing the ops/sec precision. 2010-03-31 17:18 vass * fileset.c, flowop_library.c: working O_DIRECT support. 2010-03-25 20:48 vass * fileset.c, flowop_library.c: Support of RAW devices for Filebench resurected. 2010-03-24 22:58 vass * TODO: dd fs imaging to Filebench TODO list 2010-03-24 20:58 vass * TODO: Adding remark on Filebench hang to TODO list. 2010-03-21 15:11 vass * stats.c: Removing some of new-lines in Filebench output. 2010-03-11 11:26 vass * stats.c, stats.h: Making it possible to snap the statistics multiple times during the run. 2010-03-11 10:37 vass * stats.c, stats.h: flowstats.fs_children is not used - remove. 2010-03-11 10:28 vass * flowop.c, stats.h, threadflow.h: .fs_stime and .fs_etime are used ONLY to note the statistics collection period in globalstats. So, no need to update these fields in per-thread or per-flowop flowstat. ****************** Release of 1.4.8.fsl.0.3 ************************ 2010-02-04 17:28 vass * configure.ac: Changing the Filebench version to 0.3. 2010-02-04 17:24 vass * README_COND_COMP, TODO, configure.ac: Switching to more accurate process model. 2010-02-04 11:51 vass * stats.c: Compilation warnings in OSPROF part. 2010-01-08 13:57 vass * TODO: process vs thread model verification to TODO 2010-01-06 16:07 vass * TODO: Updating TODO. 2010-01-06 12:17 vass * TODO: Some additional points to the TODO list. 2010-01-05 20:49 saumitra * filebench.h, flowop.c, parser_gram.y, parser_lex.l, stats.c, stats.h: Added osprof_enable/disable commands in filebench to facilitate the display of histogram per flowop if desired. 2010-01-05 12:43 vass * README: Updating README with info about ENOMEM on pthread_create(). 2010-01-05 02:29 vass * procflow.c, threadflow.c: Adding correct error propagation if we use thread model and thread creation failed. This allows to properly detect that we ran out of virtual memory on OLTP workload (if we use thread model). This can be workaround by setting the stack ulimit to unlimited (indirectly affects the amount of memory pthread_create mmaps for stack). 2010-01-05 01:56 vass * README_COND_COMP, configure.ac, parser_gram.y: Increasing /proc/sys/kernel/shmmax to 256MB in FileBench. 2010-01-05 01:10 vass * parser_gram.y: Increasing the ulimit for number of opened files. 2010-01-04 15:52 vass * fb_localfs.c: Fixing aio operations in FileBench on Linux: we just need to bzero aiocb. 2010-01-04 00:17 vass * README_COND_COMP, filebench.h, flowop.c, ipc.c, misc.c, procflow.c: Improving error reporting slightly (added while fixing OLTP problem): 1) report tid, not pid on Linux in logs 2) report error code of some functions. 2009-12-18 14:55 vass * README_COND_COMP, auto_comp.h, configure.ac, eventgen.h, fb_avl.h, fb_random.h, filebench.h, fileset.h, flowop.h, fsplug.h, gamma_dist.h, ipc.h, misc.h, parsertypes.h, procflow.h, stats.h, threadflow.h, utils.h, vars.h: To compile on OpenSolaris for 64 bit we need to specify: ./configure CFLAFS="-g -O2 -m64" During this compilation the problem arises: if compiling for 64bit, Solaris redefines aio_write64 as aio_write, so configure doesn't detect aio_write64. That is correct, but configure detects aiocb64_t (in a separate check), so we got compilation errors: aio_write(aoicb64_t). To fix it we assume that whenever aio_write64 is not defined, aiocb64_t is also not defined. The small change described above revealed a lot of mess in header files. It is fixed in this commit too. everything is in filbench.h 2009-12-17 00:42 vass * .cvsignore, ChangeLog, README_COND_COMP, configure.ac: Updating version to 0.2 and generating ChangeLog ****************** Release of 1.4.8.fsl.0.2 ************************ 2009-12-16 13:32 sujay * stats.c, vars.c: added 64bit wordsize check in fscanf/printf code to print or read hrtime_t correctly for all architectures. 2009-12-16 11:29 vass * Makefile.am, README_COND_COMP, configure.ac, flowop.c, parser_gram.y: Fixing the compilation of FreeBSD: 1) sigignore() is not available on FreeBSD, so emulate it via sigaction 2) -Werror removed, because yacc on FreeBSD does not generate the declaration of yyparse, so we error on: warning: implicit declaration of function 'yyparse' 3) other minor things 2009-12-15 22:14 vass * README_COND_COMP, configure.ac, filebench.h, flowop.c, misc.c: Fixing compilation for OpenSolaris. 2009-12-15 21:49 vass * README_COND_COMP, configure.ac, filebench.h, flowop.c: some clean-up for the FreeBSD compilation. 2009-12-15 21:12 sujay * configure.ac, filebench.h, flowop.c: Added changes to support compilation for FreeBSD - added condition check for various 64bit related functionality in Linux and added code to filebench.h to act accordingly. 2009-12-14 23:01 sujay * fb_localfs.c: fixed small typo mistake LOG_ERR to LOG_ERROR 2009-12-14 20:50 vass * Makefile.am, TODO, fb_localfs.c, fb_random.c, fileset.c, flowop.c, flowop_library.c, ipc.c, misc.c, multi_client_sync.c, parser_gram.y, parser_lex.l, procflow.c, vars.c: Adding -Wall -Werror -Wno-unknown-pragmas to compilation and fixing all corresponding warnings. Most of them are simply missing brackets, not used variables, non-initialized vars, etc. But some errors are quite interesting. Was: ncompleted < todom, aio != NULL; (in the loop condition!) fixed to: ncompleted < todo && aio != NULL; That might have triggered AIO bug. 2009-12-14 20:24 vass * fb_avl.h: Fixing my nasty mistake in ifdefs: ifdef -> ifndef 2009-12-14 18:30 vass * README_COND_COMP, TODO, configure.ac: Fixing the AIO comilation. This finalizes all current conditional compilation. Woo-hooo :) Just need to keep this list up-to-date. 2009-12-14 15:23 vass * README_COND_COMP, fb_avl.c, fb_avl.h: AVL trees implementation depends on 64/32 bit arch. Explain this case and add || __WORDSIZE in the missing places. 2009-12-14 14:20 vass * README_COND_COMP, filebench.h: Explaining __STDC__ usage: we only use it correctly handle the filbench_log() with variable number of arguments. 2009-12-14 13:54 vass * Makefile.am, README_COND_COMP, configure.ac, fb_random.c: --user-process conditional compilation added (USER_PROCESS_MODEL). Additionally, while fixing the compilation the following thing came up: we should never use __SOME_MACRO, since they are internal to libc. E.g., instead __USE_GNU, _GNU_SOURCE should be used. This things are fixed accordingly. 2009-12-14 13:21 vass * README_COND_COMP, filebench.h, threadflow.h: Fixing all the conditional compilations related to various types definitions. 2009-12-14 12:50 vass * README_COND_COMP, configure.ac: USE_SYSTEM autoconfiguration, if we want system() to be used instead of execlp(). 2009-12-13 17:11 sujay * configure.ac, flowop.c, flowop.h: changes to support correct per thread CPU usage for Linux flaovors. This gives correct op-cpu value in final stats. 2009-12-12 23:57 vass * flowop.c, threadflow.h: Removing LINUX_PORT: we don't need it any more. 2009-12-12 23:54 vass * README_COND_COMP, configure.ac, fileset.c, parser_gram.y, parsertypes.h, utils.c, utils.h: Eliminatint "sun" conditional compilation: 1) HAVE_STL[cat|cpy] introduced 2) O_SYNC used for sync operations, not O_DSYNC/O_FSYNC 3) Not using directio() function on Solaris 4) Don't use uchar_t, use just unsigned char 2009-12-12 23:08 vass * misc.c: Allowing to use filbench_log() even before filebench_shm is not initialized. 2009-12-12 23:04 vass * README_COND_COMP, configure.ac, misc.c, misc.h, stats.h: gethrtime() related problems solved: 1) Nasty linking problem solved: for functions that return uint64_t, function definition must be in the file, where it is called. 2) All conditional compilation through HAVE_GETHRTIME and USE_RDTSC. 3) RDTSC instruction fixed: overflow issue is fixed. 4) hrtime_t type is used consistently 2009-12-10 19:38 vass * README_COND_COMP, TODO, configure.ac, fb_localfs.c, threadflow.c: List of the conditional compilation constats is created and appropriate additional changes: - IPC_RMID detection for semget() - Redundant HAVE_LWP is removed - Warning when off64_t to off_t conversion in ftruncate() - TODO updated 2009-12-10 17:33 vass * TODO: TODO update 2009-12-10 13:56 sujay * configure.ac, flowop.c, threadflow.h: convert all definitions of HAVE_PROCFS to HAVE_PROC_PID_LWP 2009-12-10 00:03 vass * TODO: Updating TODO 2009-12-09 23:38 vass * fileset.c: Follow-up fix for the prev. patch: not %lld, but %llu shoud be used. 2009-12-09 23:14 vass * fileset.c: Fixing the printf() format for Fileset info, so that it displays Fileset size and other correctly. 2009-12-09 23:13 vass * Makefile.am, configure.ac, fb_localfs.c, filebench.h, fileset.h: On Linux we should use -D_LARGEFILE64_SOURCE to have off64_t, but not the tricks that were used before. 2009-12-09 20:52 vass * auto_comp.c: Fixing the compilation warnings in the auto-completion code. 2009-12-09 20:34 vass * fb_random.c: Preventing the warning about undefined round() funcion. 2009-12-09 19:43 vass * configure.ac, misc.h, stats.c: Fixing the /proc/stat related functionality on Linux: 1) Use libkstat ifdef HAVE_LIBKSTAT (on Solaris), use /proc/stat ifdef HAVE_PROC_STAT (on Linux), dont use anything otherwise (on FreeBSD). Corresponding cleanup. 2) Return hrtime_t type from kstats_read_cpu() function, as it is supposed to be. Also removed one unnecessary HAVE_HRTIME check. 3) Close /proc/stat after each read, otherwise we read from the wrong offset. rewind() doesnt work on /proc files. 4) Correct check of error if couldn't open /proc/stat and if scanf() didn't succeed. 5) Fixing conversion from Jiffies to Nanoseconds (one zero was missing) 6) Use of %llu everythere, so that input/output happens correctly. As a result, global mumber of cpu microseconds per operation is calculated and displayed correctly on Linux. ****************** Release of 1.4.8.fsl.0.1 ************************ 2009-12-03 01:02 vass * filebench.h: Reporting correct FileBench version in the prompt. 2009-12-03 00:21 vass * TODO: TODO list added. 2009-12-03 00:20 vass * Makefile.am: Adding _REENTRANT definition for now 2009-12-03 00:07 vass * configure.ac: Handling pthreads and semaphores related stuff properly. 2009-12-02 23:22 vass * Makefile.am, configure.ac: adding -lm and -lpthreads through the .ac file, not hardcoding 2009-12-02 21:46 vass * flowop_library.c: Removing sys/asynch.h as it is excessive. 2009-12-02 21:41 vass * configure.ac, fb_localfs.c, threadflow.h: Adding the support of async operations if librt supports it. 2009-12-02 20:19 vass * configure.ac: Use ftruncate64 when available. 2009-12-02 20:13 vass * Makefile.am, configure.ac: Adding autocompletion support if libtecla is available. 2009-12-02 19:13 vass * configure.ac: Use /proc//lwp/* for per-thread accounting on Solaris. 2009-12-02 16:47 vass * configure.ac: use mkstemp() instead tempnam() if possible. 2009-12-02 16:36 vass * configure.ac: Use cftime() if available instead of strftime() 2009-12-02 16:28 vass * configure.ac: Use libkstat or /proc/stat for obtaining CPU usage statistics. 2009-12-02 16:12 vass * stats.c: sys/sysinfo.h is not required ever. 2009-12-02 16:10 vass * fileset.c, flowop_library.c: Remove RAW device support (works(?) only on Solaris, we don't need it). We may resurect this later. 2009-12-02 15:44 vass * configure.ac: Use semtimedop() instead semop() if possible. 2009-12-02 15:31 vass * configure.ac: Using SYSV semaphores if available. Otherwise use POSIX semaphores. 2009-12-02 15:17 vass * flowop_library.c: We dont care about utility.h 2009-12-02 15:05 vass * configure.ac: Use sigsend() on Solaris. 2009-12-02 14:57 vass * workloads/.cvsignore: Adding Makefile to .cvsignore 2009-12-02 14:56 vass * configure.ac: using set/getrilimit to set the maximum number of files per process. 2009-12-02 14:45 vass * configure.ac: Using fork1() on Solaris if possible. 2009-12-02 14:34 vass * configure.ac: Cleaning up configure.ac 2009-12-02 04:09 vass * Makefile.am, configure.ac, workloads/.cvsignore, workloads/Makefile.am: Adding workloads directory to distribution pacakage. 2009-12-02 03:04 vass * workloads/: bringover.f, compflow_demo.f, copyfiles.f, createfiles.f, deletefiles.f, filemicro_create.f, filemicro_createfiles.f, filemicro_createrand.f, filemicro_delete.f, filemicro_rread.f, filemicro_rwrite.f, filemicro_rwritedsync.f, filemicro_rwritefsync.f, filemicro_seqread.f, filemicro_seqwrite.f, filemicro_seqwriterand.f, filemicro_seqwriterandvargam.f, filemicro_seqwriterandvartab.f, filemicro_statfile.f, filemicro_writefsync.f, fileserver.f, listdirs.f, makedirs.f, mongo.f, multistreamread.f, multistreamreaddirect.f, multistreamwrite.f, multistreamwritedirect.f, netsfs.f, networkfs.f, oltp.f, openfiles.f, postmark.f, randomfileaccess.f, randomread.f, randomrw.f, randomwrite.f, ratelimcopyfiles.f, removedirs.f, singlestreamread.f, singlestreamreaddirect.f, singlestreamwrite.f, singlestreamwritedirect.f, tpcso.f, varmail.f, videoserver.f, webproxy.f, webserver.f: Adding workload personalities. 2009-12-02 03:01 vass * .cvsignore: Adding package archive to .cvsignore. 2009-12-02 02:49 vass * Makefile.am: Dont remove parser_gram.c and similar on distclean, because they should go to the distribution. 2009-12-02 02:46 vass * .cvsignore, filebench.h, multi_client_sync.c, multi_client_sync.h, parser_gram.y: Fixing the compilation on Linux by: 1) Disabling multi client sync with master node 2) Disabling fscheck command 3) Moving off64_t up 4) removing UINT64_MAX definition Plus appropriate records are added to .cvsignore. 2009-12-02 02:20 vass * .cvsignore: Adding .cvsignore with the entries that appear during aclocal && autoconf && autoheader && automake && configure 2009-12-02 02:09 vass * AUTHORS, COPYING, ChangeLog, INSTALL, LICENSE, Makefile.am, Makefile.in, NEWS, README, configure.ac: Adding our building infrastructure based on the Autotoolsset. We only store files that we can't regenerate. E.g. configure script can be regenerated, so we don't store it. configure.ac on the othe hand can't be regenerated, so we store it. According to this principle Makefile.in was removed. - AUTHORS, COPYING, and LICESE files come from FileBench distibution. - README is also from FileBench ditribution, but slightly adjusted to the fact that we maintain only go_filebench, not the whole FileBench. - ChangeLog, NEWS are empty at the moment, but required by autotools 2009-12-02 01:27 vass * config_solaris.h: We dont need a separate config for Solaris. 2009-12-02 01:23 vass * Makefile.am, Makefile.in, auto_comp.c, auto_comp.h, config_solaris.h, eventgen.c, eventgen.h, fb_avl.c, fb_avl.h, fb_localfs.c, fb_random.c, fb_random.h, filebench.h, fileset.c, fileset.h, flowop.c, flowop.h, flowop_library.c, fsplug.h, gamma_dist.c, gamma_dist.h, ipc.c, ipc.h, misc.c, misc.h, multi_client_sync.c, multi_client_sync.h, parser_gram.y, parser_lex.l, parsertypes.h, procflow.c, procflow.h, stats.c, stats.h, threadflow.c, threadflow.h, utils.c, utils.h, vars.c, vars.h: Original code from FileBench-1.4.8 package from Sourceforge. We import only filebench/ subdirectory that contains sources for go_filebench binary. In this commit we do only minimal cosmetic changes (filebench.h.BAK and .deps/ are removed, plus permissions changed), so that in the following commits we can track the changes we add on top of legacy FileBench.