Castor Fu [Fri, 20 Dec 2013 07:00:46 +0000 (23:00 -0800)]
Make configure tests pass pickier compiler options.
Signed-off-by: Castor Fu <castor@alumni.caltech.edu>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Mon, 9 Dec 2013 19:38:53 +0000 (12:38 -0700)]
Add options to have fio latency profile a device
This adds three new options:
- latency_target. This defines a specific latency target, in usec.
- latency_window. This defines the period over which fio samples.
- latency_percentile. This defines the percentage of IOs that must
meet the criteria specified by latency_target/latency_window.
With these options set, fio will run the described workload and
vary the queue depth between 1 and iodepth= to find the best
performing spot that meets the criteria specified by the three
options.
A sample job file is also added to demonstrate how to use this.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Fri, 6 Dec 2013 19:02:10 +0000 (12:02 -0700)]
Calloc() cleanup
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Wed, 27 Nov 2013 03:40:00 +0000 (20:40 -0700)]
Merge branch 'master' of ssh://git.kernel.dk/data/git/fio
Jens Axboe [Wed, 27 Nov 2013 03:39:35 +0000 (20:39 -0700)]
init: escape description option string
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Wed, 27 Nov 2013 03:19:59 +0000 (20:19 -0700)]
init: allow FIO_OPT_STR_SET options to have an optional argument
Makes them work like a bool.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Bruce Cran [Thu, 7 Nov 2013 14:04:00 +0000 (14:04 +0000)]
Windows: add CONFIG_TLS_THREAD=y and swap pthreadGC2 to libwinpthread-1.
Since GCC 4.4, MinGW has supported TLS via the __thread keyword, so
generate config-host.mak with CONFIG_TLS_THREAD=y.
MinGW now uses a new threading library libwinpthread-1.dll: install it
instead of pthreadGC2.dll.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Thu, 21 Nov 2013 18:13:12 +0000 (11:13 -0700)]
Error out gracefully if we don't find the replay device for log replay
It would segfault before because no files got added to replay
against. Handle this more gracefully and ensure that fio exits
with a non-zero exit status.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Thu, 21 Nov 2013 16:55:49 +0000 (09:55 -0700)]
blktrace: add support for non-native endian format
The blktrace format is stored in the native endianness of
the machine it is run on. So to reply traces on a machine
with a different endianness, we need to swap the trace
fields. Detect and do this automatically.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Sat, 16 Nov 2013 18:15:12 +0000 (11:15 -0700)]
Fio 2.1.4
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Chris Mason [Fri, 15 Nov 2013 22:52:58 +0000 (15:52 -0700)]
Add support for O_ATOMIC
O_ATOMIC makes writes atomic, meaning that they are fully stable
on media (in the event of a power cut) when acknowledged by the
device and OS.
This only truly works on Linux with the pending patches to
add O_ATOMIC.
Updated by Jens to:
- Add man page and HOWTO description of the option
- Make O_ATOMIC imply O_DIRECT, so that it actually works if you
don't set O_DIRECT manually.
- Add the option to the conversion list so it works for
client/server.
- Error handling so that if atomic=1 is set and the OS does not
support it, error out instead of just pretending it works.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Fri, 8 Nov 2013 17:57:08 +0000 (10:57 -0700)]
Fix access-outside-array of o->rwmix[]
gcc 4.9 rightfully complains about it, though in a very obtuse way:
CC cconv.o
In file included from thread_options.h:5:0,
from cconv.c:3:
cconv.c: In function 'convert_thread_options_to_cpu':
os/os.h:197:16: warning: iteration 2u invokes undefined behavior
[-Waggressive-loop-optimizations]
__le32_to_cpu(*__val); \
^
os/os.h:176:28: note: in definition of macro '__le32_to_cpu'
#define __le32_to_cpu(x) (x)
^
cconv.c:78:17: note: in expansion of macro 'le32_to_cpu'
o->rwmix[i] = le32_to_cpu(top->rwmix[i]);
^
cconv.c:63:2: note: containing loop
for (i = 0; i < DDIR_RWDIR_CNT; i++) {
^
In file included from thread_options.h:5:0,
from cconv.c:3:
cconv.c: In function 'convert_thread_options_to_net':
os/os.h:209:16: warning: iteration 2u invokes undefined behavior
[-Waggressive-loop-optimizations]
__cpu_to_le32(*__val); \
^
os/os.h:179:28: note: in definition of macro '__cpu_to_le32'
#define __cpu_to_le32(x) (x)
^
cconv.c:372:19: note: in expansion of macro 'cpu_to_le32'
top->rwmix[i] = cpu_to_le32(o->rwmix[i]);
^
cconv.c:352:2: note: containing loop
for (i = 0; i < DDIR_RWDIR_CNT; i++) {
^
In file included from thread_options.h:5:0,
from cconv.c:3:
cconv.c: In function 'fio_test_cconv':
os/os.h:209:16: warning: iteration 2u invokes undefined behavior
[-Waggressive-loop-optimizations]
__cpu_to_le32(*__val); \
^
os/os.h:179:28: note: in definition of macro '__cpu_to_le32'
#define __cpu_to_le32(x) (x)
^
cconv.c:372:19: note: in expansion of macro 'cpu_to_le32'
top->rwmix[i] = cpu_to_le32(o->rwmix[i]);
^
cconv.c:352:2: note: containing loop
for (i = 0; i < DDIR_RWDIR_CNT; i++) {
^
In file included from thread_options.h:5:0,
from cconv.c:3:
os/os.h:197:16: warning: iteration 2u invokes undefined behavior
[-Waggressive-loop-optimizations]
__le32_to_cpu(*__val); \
^
os/os.h:176:28: note: in definition of macro '__le32_to_cpu'
#define __le32_to_cpu(x) (x)
^
cconv.c:78:17: note: in expansion of macro 'le32_to_cpu'
o->rwmix[i] = le32_to_cpu(top->rwmix[i]);
^
cconv.c:63:2: note: containing loop
for (i = 0; i < DDIR_RWDIR_CNT; i++) {
^
Fix it by extending the rwmix indexes to the full rwdir count,
even though it doesn't support trim as part of the mix (yet).
Reported-by: Bruce Cran <bruce@cran.org.uk>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Thu, 7 Nov 2013 20:37:09 +0000 (13:37 -0700)]
server: make job/connection lists private
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Wed, 6 Nov 2013 22:37:35 +0000 (15:37 -0700)]
solaris: ensure that -D_REENTRANT gets set
Apparently some Solaris' require this for threadsafe
errno.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Wed, 6 Nov 2013 21:47:22 +0000 (14:47 -0700)]
posixaio: fix thread problem with using errno
If we fail queueing a read or a write, use aio_error() to
retrieve the right error value. This fixes an issue on
(at least) Solaris where we get EAGAIN due to system
shortage of resources, but treat that as a random type
of error due to using errno.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Wed, 6 Nov 2013 18:19:42 +0000 (11:19 -0700)]
server: get rid of compile warning on platforms where pid_t isn't int
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Bruce Cran [Mon, 4 Nov 2013 11:59:43 +0000 (11:59 +0000)]
Fix a couple of typos in fio2gnuplot.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Bruce Cran [Mon, 4 Nov 2013 11:52:08 +0000 (11:52 +0000)]
Use #!/usr/bin/env to allow for interpreters in different directories.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Bruce Cran [Fri, 1 Nov 2013 18:32:40 +0000 (12:32 -0600)]
server: ensure that fio_time_init() is called before option parsing
When run in server mode, jobs are processed before fio_time_init()
is called, which causes pthread_setspecific() to be called before
pthread_key_create(). Work around this by calling it in thread_main()
too.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Bruce Cran [Fri, 1 Nov 2013 18:31:03 +0000 (12:31 -0600)]
Implement writev on Windows
The network code depends on writev(), so implement it for Windows.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Bruce Cran [Tue, 8 Oct 2013 14:05:27 +0000 (15:05 +0100)]
Update Windows code to fix build breakage
MinGW has been updated: this causes problems due to changes in header
files, including COM stuff being pulled in resulting in 'interface' being
a reserved keyword, and changes to a few POSIX definitions.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Castor Fu [Thu, 31 Oct 2013 17:00:54 +0000 (11:00 -0600)]
configure: add option to disable libnuma usage
Signed-off-by: Castor Fu <castor@alumni.caltech.edu>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Castor Fu [Thu, 31 Oct 2013 17:00:34 +0000 (11:00 -0600)]
Add json output for client/server mode
In client/server mode, this adds support for json mode. Each job's
details are gradually accumulated and upon completion the client
dumps the full json tree. The tree is annotated with the server
host and port.
Signed-off-by: Castor Fu <castor@alumni.caltech.edu>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Thu, 31 Oct 2013 16:56:53 +0000 (10:56 -0600)]
Fix bad asm constrants for x86-64 cpuid()
Bruce reports:
It seems there's a bug in the x86_64 version of do_cpuid() that causes
fio to segfault when built with clang at lower optimization levels:
static inline void do_cpuid(unsigned int *eax, unsigned int *ebx,
unsigned int *ecx, unsigned int *edx)
{
asm volatile("cpuid"
: "=a" (*eax), "=b" (*ebx), "=r" (*ecx), "=d" (*edx)
: "0" (*eax), "2" (*ecx)
: "memory");
}
via Tijl Coosemans:
Should be: "=c" (*ecx)
But you can also use the '+' modifier and remove the input operands:
: "+a" (*eax), "=b" (*ebx), "+c" (*ecx), "=d" (*edx)
:
: "memory"
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Hong Zhiguo [Wed, 16 Oct 2013 14:35:12 +0000 (08:35 -0600)]
client: fix free of wrong pointer
Signed-off-by: Hong Zhiguo <zhiguohong@tencent.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Fri, 11 Oct 2013 16:27:28 +0000 (10:27 -0600)]
ppc: disable CPU clock until we can detect whether we have it or not
The child segfault test should catch it, however it does not on
AIX at least.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Akash Verma [Fri, 11 Oct 2013 16:24:26 +0000 (10:24 -0600)]
arch-ppc.h: Fix typo in header guard
Signed-off-by: Akash Verma <akashv@google.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Akinobu Mita [Wed, 9 Oct 2013 15:32:34 +0000 (09:32 -0600)]
make ioengine specific command line options work
If the ioengine specific parameter is given on the command line,
unrelated value in thread_data.o will be corrupted unexpectedly.
Specifically, if libaio specific parameter userspace_reap is given
on the command line, thread_data.o.description will be set to 1.
This fixes it by passing the correct pointer of ioengine specific
options to the command line parser.
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Tue, 8 Oct 2013 14:49:25 +0000 (08:49 -0600)]
Merge branch 'master' of ssh://git.kernel.dk/data/git/fio
Hong Zhiguo [Tue, 8 Oct 2013 14:48:47 +0000 (08:48 -0600)]
fix crash: free of wrong pointer in bssplit_ddir
I got SIGABRT when I passed "bssplit=4k/10:64k/50:/40" to fio.
It should be "bssplit=4k/10:64k/50:32k/40" but I missed "32k" in
my typing. The bug turns out to be another typo just like mine:)
*** glibc detected *** /viki/lkml/fio/fio: free(): invalid pointer:
0x000000000071fd30 ***
======= Backtrace: =========
/lib/libc.so.6(+0x78bb6)[0x7ffff700cbb6]
/lib/libc.so.6(cfree+0x73)[0x7ffff7013483]
/viki/lkml/fio/fio[0x42c108]
/viki/lkml/fio/fio[0x42c475]
/viki/lkml/fio/fio[0x429194]
/viki/lkml/fio/fio[0x42a888]
/viki/lkml/fio/fio(parse_option+0x101)[0x42ac47]
/viki/lkml/fio/fio(fio_options_parse+0x95)[0x42e5b3]
/viki/lkml/fio/fio(parse_jobs_ini+0x3ae)[0x4103ae]
/viki/lkml/fio/fio(parse_options+0x112)[0x41198a]
/viki/lkml/fio/fio(main+0x3b)[0x460f5b]
/lib/libc.so.6(__libc_start_main+0xfd)[0x7ffff6fb2c4d]
/viki/lkml/fio/fio[0x40a7c9]
Signed-off-by: Hong Zhiguo <zhiguohong@tencent.com>
Tested-by: Hong Zhiguo <zhiguohong@tencent.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Sitsofe Wheeler [Fri, 4 Oct 2013 22:09:42 +0000 (23:09 +0100)]
howto: Fix redundant entries
Reported-by: Konstantin Ruvinsky <Konstantin.Ruvinsky@emc.com>
Signed-off-by: Sitsofe Wheeler <sitsofe@yahoo.com>
Sitsofe Wheeler [Fri, 4 Oct 2013 21:07:23 +0000 (22:07 +0100)]
man page: Bump up release date
Signed-off-by: Sitsofe Wheeler <sitsofe@yahoo.com>
Sitsofe Wheeler [Fri, 4 Oct 2013 21:07:23 +0000 (22:07 +0100)]
man page: fix broken formatting before OUTPUT section
Signed-off-by: Sitsofe Wheeler <sitsofe@yahoo.com>
Sitsofe Wheeler [Fri, 4 Oct 2013 21:07:23 +0000 (22:07 +0100)]
man page: manpagify things that look like syscalls.
Use the bold style with section number to mark syscalls.
Signed-off-by: Sitsofe Wheeler <sitsofe@yahoo.com>
Sitsofe Wheeler [Fri, 4 Oct 2013 21:07:23 +0000 (22:07 +0100)]
man page: Fix typos
Signed-off-by: Sitsofe Wheeler <sitsofe@yahoo.com>
Jens Axboe [Fri, 4 Oct 2013 18:33:11 +0000 (12:33 -0600)]
man page: correct typoe
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Fri, 4 Oct 2013 15:48:15 +0000 (09:48 -0600)]
Merge branch 'manpage' of https://github.com/sitsofe/fio
Jens Axboe [Mon, 30 Sep 2013 18:17:34 +0000 (12:17 -0600)]
log: disable log, if realloc fails
Right now we just segfault, if realloc() returns NULL. Handle
this a bit more gracefully, by just disabling logging if that
happens. Fio will also print an error for that case.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Sitsofe Wheeler [Fri, 27 Sep 2013 12:17:59 +0000 (13:17 +0100)]
Update man page
Many of fio's options are not described in the man page. Copy the short
text to the man page for those commands listed by:
./fio --cmdhelp=all | sed 's/ *:.*$//' | sed 's/ \+//' \
| while read line; do grep "\b$line\b" -q fio.1; if [[ $? -eq 1 ]]; then \
echo $line; fi; done
Signed-off-by: Sitsofe Wheeler <sitsofe@yahoo.com>
Sitsofe Wheeler [Fri, 27 Sep 2013 12:17:59 +0000 (13:17 +0100)]
Update man page
Fix verifysort typo, broken formatting after roundrobin and tidy up some
grammar.
Signed-off-by: Sitsofe Wheeler <sitsofe@yahoo.com>
Anatol Pomozov [Thu, 26 Sep 2013 23:31:34 +0000 (16:31 -0700)]
Fix misspellings using 'codespell' tool
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Tue, 24 Sep 2013 14:42:24 +0000 (08:42 -0600)]
Fio 2.1.3
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Juan Casse [Fri, 20 Sep 2013 15:20:52 +0000 (09:20 -0600)]
Fixes bug: stale LAST_POS(f) is not being reset
Problem:
When running fio with ioengine=sync, LAST_POS(f) state is not
reset after the file is closed. This causes workloads with
readwrite=randread and loops > 1 to fail verification if the
state of LAST_POS(f) at the beginning of the next loop is the
same as the io_u->offset. If that is the case, lseek in not
invoked, but the file position is at 0 and not at io_u->offset.
Proposed Solution:
Other ioengines, such as binject and fusion-aw, set f->engine_data
to 0 when closing the file. The sync ioengin uses f->engine_data to
store the LAST_POS(f) state. The proposed solution is to set
f->engine_data = 0 when closing a file in generic_close_file().
Signed-off-by: Juan Casse <jcasse@chromium.org>
Reviewed-by: Grant Grundler <grundler@chromium.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Juan Casse [Mon, 16 Sep 2013 16:22:10 +0000 (09:22 -0700)]
Add condition to stop issuing io in do_io().
Problem:
When specifying iodepth > 1 while doing random async io
without a random map (and using the default random generator), the
main while loop in do_io() will issue extra io in the amount of io
allowed to accumulate in the queue. For example, a job with size=24k
and bs=8k will perform 4 io instead of the expected 3.
Reason:
The above behavior occurs because the while loop in do_io()
will continue until the amount of "completed io" >= size, but io
in the queue are not accounted for because they have not been
completed by io_u_queued_complete().
Exceptions:
The above behavior does not manifest itself when:
using random_generator=lfsr bacause lfsr knows when to stop
generating, and get_io_u() inside the while loop returns NULL,
breaking out of the loop.
using random map because random map ensures that all blocks
are hit, and after they are all hit, no more are generated.
Proposed Solution:
Stop while loop based on bytes issued instead of bytes completed.
Signed-off-by: Juan Casse <jcasse@chromium.org>
Reviewed-by: Grant Grundler <grundler@chromium.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Wed, 11 Sep 2013 14:23:45 +0000 (08:23 -0600)]
debug: ensure that __dprint() is also logged over the network
We force local logging now, but that's not terribly useful
if you can't see the output from the backend. Make it follow
the usual rules of logging over the network if a backend, to
syslog, or locally.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Mon, 9 Sep 2013 21:16:28 +0000 (15:16 -0600)]
Fix broken requeue of io_us
Suresh reports:
-----
Does a requeued io_u work get submitted again through __get_io_u() ? If
so, the io_u item is taken out of requeue list & then the io_u->file is
set to NULL. We don't make a distinction about item being already setup
Vs a new io_u in __get_io_u(). In get_io_u(), when an io_u may come out
of requeue list, we depend on io_u->file to be non NULL to detect the
work as old pending one.
Am I missing something here or is this a bug?
1274 io_u = __get_io_u(td);
1275 if (!io_u) {
1276 dprint(FD_IO, "__get_io_u failed\n");
1277 return NULL;
1278 }
1279
1280 if (check_get_verify(td, io_u))
1281 goto out;
1282 if (check_get_trim(td, io_u))
1283 goto out;
1284
1285 /*
1286 * from a requeue, io_u already setup
1287 */
1288 if (io_u->file)
1289 goto out;
1290
"io_u.c" line 1286 of 1667 --77%-- col 3-10
1130 struct io_u *__get_io_u(struct thread_data *td)
1131 {
1132 struct io_u *io_u;
1133
1134 td_io_u_lock(td);
1135
1136 again:
1137 if (!io_u_rempty(&td->io_u_requeues))
1138 io_u = io_u_rpop(&td->io_u_requeues);
1139 else if (!io_u_qempty(&td->io_u_freelist))
1140 io_u = io_u_qpop(&td->io_u_freelist);
1141
1142 if (io_u) {
1143 io_u->buflen = 0;
1144 io_u->resid = 0;
1145 io_u->file = NULL;
1146 io_u->end_io = NULL;
1147 }
"io_u.c" line 1130 of 1667 --67%-- col 1
-----
We want a retry of a requeue to be identical, not some
new chunk (and drop the latter one on the floor). This got broken with
the introduction of the rings instead of lists, where a single if
section was split into two.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Fri, 6 Sep 2013 16:07:37 +0000 (10:07 -0600)]
Catch more places where the callback should just return on parse/check
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Fri, 6 Sep 2013 15:59:56 +0000 (09:59 -0600)]
Fix problem with --showcmd and callbacks that verify
David reports:
I'm using "fio --showcmd file.job" to convert a script to plain
commandline arguments and this fails if the directory specified in job
file does not exist. While this has to be an error if the job is being
executed, it should not be in context of --showcmd.
To reproduce:
$ cat job.fio
[global]
directory=/x
$ fio --showcmd job.fio
fio: /x is not a directory
fio: failed parsing directory=/x
fio: job global dropped
Expected output:
fio --directory=/x
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Grant Grundler [Wed, 4 Sep 2013 20:12:09 +0000 (14:12 -0600)]
Remove duplicate LDFLAGS on -o lines
Only need to list LDFLAGS once on each line.
Signed-off-by: Grant Grundler <grundler@chromium.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Grant Grundler [Wed, 4 Sep 2013 14:26:52 +0000 (08:26 -0600)]
Rework descriptions for mailing list, options, job files
In general, make the existing document less ambiguous:
o Use consistent formatting/layout when describing options
o always refer to options with "--" prefix.
o reduce use of pronouns: "its" and "this" are often ambigous.
o remove use of "you" to describe user actions/choices
o add reference to examples/ directory
Signed-off-by: Grant Grundler <grundler@chromium.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Fri, 30 Aug 2013 16:13:36 +0000 (10:13 -0600)]
client: fix compile warning on platforms where pfds->fd is not an int
It's apparently a long on some platforms.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Alex Pyrgiotis [Tue, 27 Aug 2013 16:58:34 +0000 (10:58 -0600)]
LFSR: Do not ignore returning the seed
Fix a situation where we would spin prematurely and would hop over the
seed value.
Also, add some more comments to the code.
Signed-off-by: Alex Pyrgiotis <apyrgio@grnet.gr>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Martin Steigerwald [Tue, 27 Aug 2013 15:33:35 +0000 (09:33 -0600)]
Fix spelling error in fio man page
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Fri, 9 Aug 2013 20:31:06 +0000 (14:31 -0600)]
verify: fix problem with hole punching on newer Linux kernels
Jeff Moyer writes:
Our QE team noticed fio failures on recent kernels. I simplified the
job file to the following:
[global]
name=fio-mmap
rw=write
bs=4K
direct=1
end_fsync=1
verify=crc32
[file3]
size=100M
ioengine=mmap
mem=malloc
direct=1
After fio completes (and returns verify errors), the file is completely
full of zeroes.
Here is what the verify logic is doing:
static void do_verify(struct thread_data *td, uint64_t verify_bytes)
{
...
/*
* sync io first and invalidate cache, to make sure we really
* read from disk.
*/
for_each_file(td, f, i) {
if (!fio_file_open(f))
continue;
if (fio_io_sync(td, f))
break;
if (file_invalidate_cache(td, f)) <--------
break;
}
That invalidate cache call looks like so:
static int __file_invalidate_cache(struct thread_data *td, struct fio_file *f,
unsigned long long off,
unsigned long long len)
{
...
/*
* FIXME: add blockdev flushing too
*/
if (f->mmap_ptr) {
ret = posix_madvise(f->mmap_ptr, f->mmap_sz, POSIX_MADV_DONTNEED);
(void) posix_madvise(f->mmap_ptr, f->mmap_sz, FIO_MADV_FREE); <-------
FIO_MADV_FREE can be defined as MADV_REMOVE, which will actually punch a
hole in the file (a hole the size of the entire file, btw). Of course,
unallocated blocks are returned as zeroes by the file system, so that
explains that!
---
For now, only do the MADV_FREE on non-files to avoid punching holes.
Further investigation is needed on the blockdev side, but at least
this should cure the immediate problem.
Reported-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Fri, 9 Aug 2013 18:53:44 +0000 (12:53 -0600)]
Add number_ios= option
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Fri, 9 Aug 2013 15:17:09 +0000 (09:17 -0600)]
powerpc: disable use of ATBU clock
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Erwan Velu [Thu, 8 Aug 2013 08:59:46 +0000 (10:59 +0200)]
genfio: Don't consider only /dev/ disk device
Prior this patch, genfio was considering the disks __NOT__ having the "/dev"
name in it.
That's a serious limitation as
- we can't test regular files
- that doesn't works for complex /dev/ naming like /dev/disk/by-id/...
This patch removes this requirement and let user giving the complete
path.
Erwan Velu [Wed, 7 Aug 2013 14:02:25 +0000 (16:02 +0200)]
fio2gnuplot: Adding manpage
As requested by users, let's add a manpage to make things easier to
understand.
I'm providing here the raw file used to generate the manpage. It can be
converted to a real manpage by using txt2man.
I'm also providing the manpage file to avoid dependencies at build time,
anyway the makefile target can be easily used to rebuild it if
necessary.
Erwan Velu [Wed, 7 Aug 2013 13:32:02 +0000 (15:32 +0200)]
fio2gnuplot: Removing .py extension
It's pretty boring having the .py extension to the tool.
Let's remove it !
Erwan Velu [Wed, 7 Aug 2013 12:32:50 +0000 (14:32 +0200)]
fio2gnuplot: Fixing help message
Adding missing features & fix indent
Erwan Velu [Wed, 7 Aug 2013 12:23:43 +0000 (14:23 +0200)]
fio2gnuplot: Managing temporary files in a better way
If gnuplot success at rendering, let's delete all temporary files.
If gnuplot fails at rendering, let's keep all temporary files.
If user don't want to run gnuplot, let's keep temporary files.
If user ask to keep the temporary files (option -k), let's always keep
them.
Erwan Velu [Wed, 7 Aug 2013 12:11:38 +0000 (14:11 +0200)]
fio2gnuplot: Deleting temporary files by default
Erwan Velu [Wed, 7 Aug 2013 11:28:58 +0000 (13:28 +0200)]
fio2gnuplot: Adding verbose option
Reducing default's verbosity and add a -v option to increase it.
This avoids flooding end-users and let an option for thoses who need to
get a complete trace.
Erwan Velu [Wed, 7 Aug 2013 11:20:56 +0000 (13:20 +0200)]
fio2gnuplot: Fixing typo in error message
Erwan Velu [Wed, 7 Aug 2013 08:52:05 +0000 (10:52 +0200)]
fio2gnuplot: Adding getoplong support for cmdline
Help message was misleading since long options were not supported....
This patch adds all the long options to be supported.
Erwan Velu [Wed, 7 Aug 2013 08:46:55 +0000 (10:46 +0200)]
fio2gnplot: Print specific message if cmdline parsing fails
If the cmdline isn't consistent regarding the getopt, let's print a
specific message to inform user about this situation instead of printing
the help.
Showing the help without any specific message was very confusing.
Jens Axboe [Tue, 6 Aug 2013 15:39:54 +0000 (09:39 -0600)]
Fio 2.1.2
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Erwan Velu [Fri, 2 Aug 2013 14:39:40 +0000 (16:39 +0200)]
iolog: Logging blocksize in IOPS traces
When saving bandwidth's with write_bw_log option, the last column
provides the blocksize used during the run.
When doing the same with write_io_log option, the last column was always
set to 0 which is confusing when reading the log file later.
This patch put both write_bw_log & write_io_log using the same semantic
by reporting the blocksize for every sample.
Erwan Velu [Fri, 2 Aug 2013 12:29:00 +0000 (14:29 +0200)]
fio2gnuplot: Display average value in label
When plotting 2D or mathematical graphs, let's show the numerical value
of the average value to ease the reading.
Erwan Velu [Fri, 2 Aug 2013 12:16:14 +0000 (14:16 +0200)]
fio2gnuplot: Plotting average value for compare graphs
When comparing in 2D several plots, it's pretty useful to have the
average value printed in the same graph.
Erwan Velu [Fri, 2 Aug 2013 08:52:50 +0000 (10:52 +0200)]
fio2gnuplot: Fixing plotting issues on shorten files
fio2gnuplot was mixing files when some got shorten.
The position inside the array was used to remember what file was read
leading to mistakes once some values got ignored.
As a fix, we shall count the index for every single line we read and add
to the time & perf we keep. While writing data to the temporary files,
the extraction of the index insure we write to the proper file.
Before this patch, we observed at the end of the traces some stupid artefacts like loops... was
looking like a 4years old kid drawing...
That's now fixed.
Erwan Velu [Fri, 2 Aug 2013 08:51:25 +0000 (10:51 +0200)]
fio2gnplot: Renaming file variable to myfile
'file' could be reserved on some modules or in the python.
It's better to rename it to avoid troubles.
Erwan Velu [Thu, 1 Aug 2013 13:11:23 +0000 (15:11 +0200)]
fio2gnuplot: Adding time selection options
It's could be very handy being able to extract a portion of a trace to
plot it.
This patch adds two new options to select the starting time and the end
time were the logs shall be considered for plotting.
Erwan Velu [Thu, 1 Aug 2013 11:41:02 +0000 (13:41 +0200)]
fio2gnuplot: Print line if int() conversion fails
When data aren't properly formated the int() conversion could fails but
doesn't print the impacted line which is pretty difficult to estimate
where the bug is.
This patch simply reports the line that cause the trouble.
Erwan Velu [Thu, 1 Aug 2013 11:39:32 +0000 (13:39 +0200)]
fio2gnuplot: Fixing typo when inserting fake data
Commit
f9846c392d176582916c39b1925a436e79f23de1 introduced fake data
when matched files doesn't have the exact same amount of traces. This
patch introduced a typo that prevented the int comparison to work
properly.
Martin Steigerwald [Mon, 5 Aug 2013 14:57:33 +0000 (08:57 -0600)]
Make test for gnuplot work with empty strings.
When I launch fio_generate_plots on a system where gnuplot is not installed I
get this error :
$ fio_generate_plots test
Making bw logs
/usr/bin/fio_generate_plots: 42: /usr/bin/fio_generate_plots: -: not found
That's because the test checking whether gnuplot is installed is failing
because of an empty variable :
GNUPLOT=$(which gnuplot)
if [ ! -x $GNUPLOT ]; then
echo You need gnuplot installed to generate graphs
exit 1
fi
Indeed the command "test -x" is exiting with code 0.
To correct this we should enclose the variable with quotes :
if [ ! -x "$GNUPLOT" ]; then
Then the script is going into the test properly :
$ fio_generate_plots test
You need gnuplot installed to generate graphs
I copied problem description by Hervé from the Debian bug report below.
Fixes:
Bug#700580: /usr/bin/fio_generate_plots: -: not found
http://bugs.debian.org/700580
Reported-By: Hervé Werner <hwerner@score-md.com>
Tested-By: Hervé Werner <hwerner@score-md.com>
Tested-By: Martin Steigerwald <ms@teamix.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Erwan Velu [Wed, 31 Jul 2013 10:32:40 +0000 (12:32 +0200)]
fio2gnuplot: Adding filename as title for 2D graphs
While plotting 2D graphs, it's pretty handy to keep the filename as
title for the plotting trace. This way, the png file still keep the
original filename used for plotting.
Erwan Velu [Wed, 31 Jul 2013 10:21:10 +0000 (12:21 +0200)]
fio2gnuplot: Adding information steps
While executing fio2gnuplot, we were lacking of some information of
what's going on.
This patch makes things easier to understand/follow for end users.
Erwan Velu [Wed, 31 Jul 2013 10:15:24 +0000 (12:15 +0200)]
fio2gnuplot: Fixing help message
Some options were missing on the main description.
Erwan Velu [Wed, 31 Jul 2013 10:07:32 +0000 (12:07 +0200)]
fio2gnuplot: Adding comparing 2D graphs
It's a pretty common need to plot several files on the same graph to
compare behavior during various benchs.
If 2 or more files are selected by the pattern, new plots will be
generated to compare the traces into a single graph. 3 kinds of graphs
will be generated (raw, smooth and trends) like for the individual
graphs.
The output filename will be prefixed with "compare-" like :
compare-vm1-1-4m-vdb-read-para.results_bw-2Draw.png
compare-vm1-1-4m-vdb-read-para.results_bw-2Dsmooth.png
compare-vm1-1-4m-vdb-read-para.results_bw-2Dtrend.png
Erwan Velu [Tue, 30 Jul 2013 12:13:51 +0000 (14:13 +0200)]
fio2gnuplot: Printing number of selected files
when using pattern matching on a big set of files, it's pretty hard to
estimate if we missed some or not.
Let's print the number of matched items with the selected pattern to
help user estimate if it's ok or not.
Erwan Velu [Tue, 30 Jul 2013 11:50:07 +0000 (13:50 +0200)]
fio2gnuplot: Using default install dir for gpm files
That was pretty boring trying to read the gpm files from the same dir as
the logs. Since gpm files are now installed in a specific dir at make
install time, let's use them from here.
Two locations will be considered at runtime : /usr/share/fio and
/usr/local/share/fio
Erwan Velu [Tue, 30 Jul 2013 09:54:45 +0000 (11:54 +0200)]
Tree: Moving fio_generate_plots in tools/
As fio2gnplot.py & genfio are located in tools, it sounds better to put
all the additional tools in the same directory.
Erwan Velu [Tue, 30 Jul 2013 09:47:16 +0000 (11:47 +0200)]
Makefile: Installing genfio at install time
genfio wasn't installed at all during 'make install'.
Erwan Velu [Tue, 30 Jul 2013 09:45:18 +0000 (11:45 +0200)]
Makefile: Installing fio2gnuplot & sons
Let's install fio2gnuplot & other gnuplot files (.gpm) while installing
fio.
Gnuplot files are installed into $(prefix}/share/fio/ subdirectory
Erwan Velu [Tue, 30 Jul 2013 09:34:53 +0000 (11:34 +0200)]
fio2gnuplot: Option to select gnuplot's output dir
This patch add the '-d' option to select where gnuplot shall render its
files.
When plotting big series of files, it's pretty anoying having all the
files in the current directory as there is many many of them.
So let's use this option to allow users specifying where all the
graphics shall be rendered
Jens Axboe [Fri, 26 Jul 2013 19:11:52 +0000 (13:11 -0600)]
Update sample atomic write job file
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Thu, 25 Jul 2013 19:01:11 +0000 (13:01 -0600)]
Make it clear in job output whether we are using bs_is_seq_rand or not
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Thu, 25 Jul 2013 18:45:26 +0000 (12:45 -0600)]
Add support for bs_is_seq_rand
This option will switch the meaning of the read,write part of any
blocksize setting to mean sequential,random instead.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Thu, 25 Jul 2013 18:30:06 +0000 (12:30 -0600)]
init: remove leftover unused variable
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Jens Axboe [Thu, 25 Jul 2013 16:20:45 +0000 (10:20 -0600)]
Add support for randomness of any IO direction
sequential_random used to be applied to all of the IO
directions, with no possibility of having different settings
for reads, writes, and trims. Now it supports setting each of
them individually.
By default, if you do:
sequential_random=50
it will still apply to all three. If you do:
sequential_random=10,90,80
you would get reads 10% random, writes 90% random, and trims
80% random.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Yufei Ren [Tue, 23 Jul 2013 02:58:05 +0000 (20:58 -0600)]
Update for RDMA io engine's compatibility
Since the RDMA engine is not able to be successfully compiled, this
patch accommodates it to the upstream version of fio.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Erwan Velu [Mon, 22 Jul 2013 21:48:48 +0000 (23:48 +0200)]
core: Insure "il" is initialized in all cases
Prior this commit, "il" was not initialized and check_int() was supposed
to setup it's value before usage.
In fact, check_int() can let "il" uninitialized leading to a dprint()
with a unintialized value.
Erwan Velu [Mon, 22 Jul 2013 21:46:10 +0000 (23:46 +0200)]
core: Don't leak memory if error is unknown
If error is unknown, the "return 1" call was not freeing error which was
malloced at the beginning of the function.
Erwan Velu [Mon, 22 Jul 2013 21:39:18 +0000 (23:39 +0200)]
core: Removing duplicated code
Some variable are assigned during initialization and another time with
the same value a very few later.
That's clearly useless and could lead to an error if only one is changed.
Erwan Velu [Mon, 22 Jul 2013 10:59:04 +0000 (12:59 +0200)]
fio2gnuplot: Don't plot fake data
As we did add some fake data to keep as long a possible the coherency
between all the fio files, let's ignore them when rendering.
As a consequence, we are keeping the time associated to each
performance we store instead of a single time for all performance
traces. This will increase precision while plotting.
Erwan Velu [Mon, 22 Jul 2013 10:56:26 +0000 (12:56 +0200)]
fio2gnuplot: Don't truncate fio log files
The code was stopping at reading once the shorter stopped. But in the
case of a very big difference of performances between hosts, some could
be much longer/shorter than others.
This patch does insure that until _all_ input files got read complety we
do had fake data with a negative time.
This way, all the traces will be complete as per fio's output.
Erwan Velu [Mon, 22 Jul 2013 10:54:39 +0000 (12:54 +0200)]
fio2gnuplot: Keep original filename in temp. files
It was pretty hard to understand what trace file was linked to which fio
output file. This simple commit add a simpler header to inform the
debugger which fio outputfile got used.
Shawn Bohrer [Mon, 22 Jul 2013 14:57:44 +0000 (09:57 -0500)]
net: Add UDP multicast example job file
This adds a pingpong UDP multicast example with two readers.
Signed-off-by: Shawn Bohrer <sbohrer@rgmadvisors.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Shawn Bohrer [Fri, 19 Jul 2013 18:24:08 +0000 (13:24 -0500)]
net: Add option to set outgoing multicast TTL
Signed-off-by: Shawn Bohrer <sbohrer@rgmadvisors.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>