Jens Axboe [Tue, 27 Mar 2007 11:10:00 +0000 (13:10 +0200)]
Cleanup generic_file_open()
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 27 Mar 2007 10:49:56 +0000 (12:49 +0200)]
Don't completely fail for block device flushing failure
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 27 Mar 2007 10:45:34 +0000 (12:45 +0200)]
Make sure to set io size on non-files
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 27 Mar 2007 09:09:40 +0000 (11:09 +0200)]
Trim the rbtree stuff fio doesn't use
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 27 Mar 2007 09:03:28 +0000 (11:03 +0200)]
Remove kernel headers from rbtree.h
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 27 Mar 2007 09:04:59 +0000 (11:04 +0200)]
GUASI engine: should use log_err() during runtime
stderr may be redirected somewhere else.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 27 Mar 2007 08:59:49 +0000 (10:59 +0200)]
Improve random verify block sorting
Add an option to control it and also check the nosort flag in file
layout to further catch a case where sorting isn't needed.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 27 Mar 2007 08:36:12 +0000 (10:36 +0200)]
Avoid using the rbtree if we don't have to
Basically reinstate the old logic of not sorting when it's
not a win for reading the data back.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 27 Mar 2007 08:21:25 +0000 (10:21 +0200)]
Fix problems with rb code
If the offset match, we must not break. Instead follow the left branch
and things will work as expected.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 27 Mar 2007 08:16:21 +0000 (10:16 +0200)]
Update rbtree code
Grab from the later kernels, they embed the color in the pointer
to save some space.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 27 Mar 2007 08:03:09 +0000 (10:03 +0200)]
Remove redundant double casts in time functions
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 27 Mar 2007 07:56:08 +0000 (09:56 +0200)]
NULL engine: ->commit() has side effects, make sure it always works
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 27 Mar 2007 06:48:48 +0000 (08:48 +0200)]
Include syncs in the stored io log
We want to replay those as well.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 27 Mar 2007 06:46:33 +0000 (08:46 +0200)]
Remove !td->o.read_io_log test that is always true
We test it further up.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 27 Mar 2007 06:32:48 +0000 (08:32 +0200)]
Only call into log_io_piece() when verify is set
Don't know why it currently always logs the IO, it's pretty
pointless.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Davide Libenzi [Tue, 27 Mar 2007 06:14:13 +0000 (08:14 +0200)]
GUASI engine: free lingering requests from the buffer on cleanup
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Davide Libenzi [Mon, 26 Mar 2007 10:20:49 +0000 (12:20 +0200)]
GUASI update
The GUASI library has been changed to avoid using pthread defaults stacks
(that burns 8MB of VM area for each thread). Also, a new paramter has been
added to guasi_create(), to allow the caller to specify a stack size.
I noticed a different behaviour of FIO in randrw mode, from GUASI to libaio.
In GUASI, the amount of data read is about the same +-2-3% of the amount
of data written. Whereas in libaio, there can be a huge difference between
the two (up to 2-3 times the other). This is consistent.
Besides that, GUASI performance follows pretty much the libaio one, up to
an io-depth of 1000-1500, after that shows about 10-15% less performance.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 26 Mar 2007 08:32:30 +0000 (10:32 +0200)]
No need to fill random bytes for VERIFY_NULL
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 26 Mar 2007 08:23:34 +0000 (10:23 +0200)]
Add 'null' verify option
For testing purposes.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 26 Mar 2007 08:09:12 +0000 (10:09 +0200)]
->ddir_nr must be 1 by default, otherwise we'll do sequential IO
if no :foo postfix is appended.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 26 Mar 2007 07:32:22 +0000 (09:32 +0200)]
Fix io piece logging to not have O(n) runtime
Use an rbtree for that log instead.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Fri, 23 Mar 2007 17:31:49 +0000 (18:31 +0100)]
Merge branch 'master' of ssh://router/data/git/fio
Jens Axboe [Fri, 23 Mar 2007 14:57:56 +0000 (15:57 +0100)]
Free/alloc every FIO_OPT_STR_STORE string
Makes ownership clear, though the freeing is a bit anal.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Fri, 23 Mar 2007 14:29:45 +0000 (15:29 +0100)]
Make sure the ->files array is job private
Otherwise we introduce differences between threads and processes,
we don't want to do that.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Fri, 23 Mar 2007 14:20:54 +0000 (15:20 +0100)]
Fix file unlinking
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Fri, 23 Mar 2007 13:26:23 +0000 (14:26 +0100)]
Fix bug with numjobs > 1, directory and filename given
Reported by gurudas pai <gurudas.pai@oracle.com>
If numjobs > 1, add_job() will append the directory name
several times. Fix this by doing the directory append in
add_file() instead, it cleans up the code as well.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Fri, 23 Mar 2007 07:23:51 +0000 (08:23 +0100)]
Print help info on huge page shmget() returning EINVAL
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Fri, 23 Mar 2007 07:23:30 +0000 (08:23 +0100)]
Move the huge page size defines into arch code
It's 2MB on x86-64 by default, and 4MB on x86.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Fri, 23 Mar 2007 07:04:27 +0000 (08:04 +0100)]
Cleanup get_next_offset()
Split into two parts, one for generation a random offset, and one
for just setting a new sequential offset.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 22 Mar 2007 18:24:09 +0000 (19:24 +0100)]
Add option to disable fadvise() hints
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 22 Mar 2007 18:04:24 +0000 (19:04 +0100)]
Leftover debug printf()
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 22 Mar 2007 18:01:48 +0000 (19:01 +0100)]
Improve mixed random append option
If we set ->ddir_nr to o.ddir_nr initally, we can remove the double
else in get_next_offset().
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 22 Mar 2007 17:56:45 +0000 (18:56 +0100)]
Add support for mixing a random IO workload a bit
Now you can append a number to randread/randwrite (and others) to
specify only getting a random offset for every X number of ios.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Davide Libenzi [Thu, 22 Mar 2007 06:43:55 +0000 (07:43 +0100)]
Add few debug statements and limit thread usage.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Davide Libenzi [Thu, 22 Mar 2007 06:43:50 +0000 (07:43 +0100)]
Avoid "ts" going out of scope.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Davide Libenzi [Thu, 22 Mar 2007 06:43:07 +0000 (07:43 +0100)]
Adapth guasi.c to the new FIO strctures, and free the requests.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Davide Libenzi [Thu, 22 Mar 2007 06:42:57 +0000 (07:42 +0100)]
Introduce EXTFLAGS and EXTLIBS in Makefile.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 21 Mar 2007 13:20:20 +0000 (14:20 +0100)]
Improve data direction runtime logging
If we cut a data direction short, log that time so we don't
account the full runtime.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 21 Mar 2007 13:05:39 +0000 (14:05 +0100)]
Improve rwmix states
We didn't use to look at the previous bytes done for the
data directions, which skewed the mix for buffered IO.
Fix that up.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 21 Mar 2007 12:07:54 +0000 (13:07 +0100)]
Dump actual filename in failure to open
Helpful for debugging.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 21 Mar 2007 09:32:54 +0000 (10:32 +0100)]
Catch too large memory allocations that cause size_t to wrap
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Davide Libenzi [Wed, 21 Mar 2007 08:15:13 +0000 (09:15 +0100)]
Update to guasi 0.5 diff
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 21 Mar 2007 07:51:56 +0000 (08:51 +0100)]
Mention GUASI in the HOWTO for the ioengine= option
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 21 Mar 2007 07:48:26 +0000 (08:48 +0100)]
Expose guasi IO engine in ioengine=
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Davide Libenzi [Wed, 21 Mar 2007 07:46:18 +0000 (08:46 +0100)]
GUASI IO engine
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 20 Mar 2007 13:38:11 +0000 (14:38 +0100)]
Fix divide-by-zero
Another file creation crap out, this really needs a rewrite and
soon.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 20 Mar 2007 13:28:34 +0000 (14:28 +0100)]
Warn if default value is given for string option without value
That would not make sense, as the option would then always be set
and there would be no way to unset it.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 20 Mar 2007 13:23:26 +0000 (14:23 +0100)]
'new_group' option
By default, jobs in a file are part of the same reporting group
unless seperated by a stone wall or if they define a group by
themselves (numjobs=x is used). Inserting a 'new_group' parameter
in a job description, will mark the start of a new reporting group.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 20 Mar 2007 12:54:53 +0000 (13:54 +0100)]
Clear ->file_map after free()
Just in case...
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 20 Mar 2007 12:53:39 +0000 (13:53 +0100)]
Sometimes we allocated too little memory for buffers
Fix it by adding page_mask to the allocation size, then the
alignment is guarenteed to fit.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 20 Mar 2007 09:47:45 +0000 (10:47 +0100)]
Print clue to reduce queue depth if engine init fails
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 20 Mar 2007 09:42:07 +0000 (10:42 +0100)]
libaio engine: queue init error handling
It fails for repeated large io depths, so make sure we retrieve the
error and log it.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 20 Mar 2007 09:33:11 +0000 (10:33 +0100)]
Always make sure buffer sizes are aligned for allocation
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 20 Mar 2007 09:11:07 +0000 (10:11 +0100)]
File creation fix
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 20 Mar 2007 09:02:06 +0000 (10:02 +0100)]
'dev' is 'stdev' in the print out now
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 19 Mar 2007 13:00:54 +0000 (14:00 +0100)]
Make sure mmaphuge file backed file is the size we need
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 19 Mar 2007 12:15:23 +0000 (13:15 +0100)]
Sort option possible value map
We need to match longest option first, since we have to use
strncmp() to avoid fiddling with postfix options.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 19 Mar 2007 12:14:03 +0000 (13:14 +0100)]
Revert
b370e46b370e46... Put longest options first
This reverts commit
b370e46abdd2bfe764e28efb41c4caadc632b937.
We should not have to do this, and the commit was even buggy since
it mistakenly put mmap inside the mmaphuge ifdef section. Just make
parse sort the posval entries.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 19 Mar 2007 10:36:36 +0000 (11:36 +0100)]
Helpful indication to run as root
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 19 Mar 2007 10:32:46 +0000 (11:32 +0100)]
No need to use perror() after td_verror()
That'll just result in the error being displayed twice.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 19 Mar 2007 09:51:49 +0000 (10:51 +0100)]
Put longest options first
We need to have things like shmhuge before shm, otherwise the parser
will match shm. The parser needs to be fixed, this is just a quick fix.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 19 Mar 2007 09:50:47 +0000 (10:50 +0100)]
We can't stringify something the parser doesn't handle
The size needs to be written out in full.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Fri, 16 Mar 2007 19:27:27 +0000 (20:27 +0100)]
Improve rwmix buffered split
For buffered IO, the rwmix split can often be grossly unfair
towards reads, because dirtying tons of memory is done much
faster than reading data. Improve the split for such workloads
by not only looking at time.
Note that it'll still be somewhat unfair, there's only so much
we can reliably do. But it's better.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Fri, 16 Mar 2007 09:24:07 +0000 (10:24 +0100)]
If the rate setting is based on IOPS, reflect that in eta output
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 15 Mar 2007 19:49:25 +0000 (20:49 +0100)]
Untangle the file creation mess
Still needs a lot of work, will do a rewrite of this soonish. At
least it should work now.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 15 Mar 2007 14:04:43 +0000 (15:04 +0100)]
Compile error with FIO_USE_TIMEOUT defined
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 15 Mar 2007 14:01:33 +0000 (15:01 +0100)]
Move thread options into a seperate structure
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 15 Mar 2007 13:12:05 +0000 (14:12 +0100)]
Add some more options for commands
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 15 Mar 2007 13:11:38 +0000 (14:11 +0100)]
Show alias in command help
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 15 Mar 2007 13:11:10 +0000 (14:11 +0100)]
Style cleanup
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 15 Mar 2007 13:09:28 +0000 (14:09 +0100)]
Command help match also on aliases
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 15 Mar 2007 11:09:57 +0000 (12:09 +0100)]
Move setup_rate() out of log.c
Dunno why it ended up in there, doesn't make any sense.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 15 Mar 2007 11:09:39 +0000 (12:09 +0100)]
Improve rate accounting
Don't assume usec_sleep() is extremely precise.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 15 Mar 2007 10:41:11 +0000 (11:41 +0100)]
Support for setting rated based on IOPS
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 15 Mar 2007 10:35:33 +0000 (11:35 +0100)]
blockdev_invalidate_cache(): return -1 for error
This follows the normal system calls, and makes fio pickup errno
appropriately.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 15 Mar 2007 10:06:32 +0000 (11:06 +0100)]
Failure to put job in add_job() failure case
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 15 Mar 2007 09:48:13 +0000 (10:48 +0100)]
Split option handling out of init.c
It grew way too large.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 15 Mar 2007 08:35:56 +0000 (09:35 +0100)]
More file creation improvements
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 15 Mar 2007 08:33:15 +0000 (09:33 +0100)]
Remember to account for existing files in size setup
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 15 Mar 2007 08:22:23 +0000 (09:22 +0100)]
Use log_info() throughout
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 15 Mar 2007 08:15:51 +0000 (09:15 +0100)]
siint range parsing HOWTO update
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 15 Mar 2007 08:15:12 +0000 (09:15 +0100)]
Accept '-' as siint delimter as well.
Hope this doesn't break any parsing...
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 15 Mar 2007 08:14:47 +0000 (09:14 +0100)]
Add 'filesize' option
Allows the user to define the range of file sizes generated.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 14 Mar 2007 19:41:42 +0000 (20:41 +0100)]
Fix build on x86-64 distros without __NR_pread64/pwrite64
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 14 Mar 2007 13:29:16 +0000 (14:29 +0100)]
Fix build of syslet engine on x86-64
Broke on some distros, apparently. Reported by
gurudas pai <gurudas.pai@oracle.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 14 Mar 2007 13:24:11 +0000 (14:24 +0100)]
Fio 1.14a
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 14 Mar 2007 13:23:54 +0000 (14:23 +0100)]
Fixup new and wrong file name behaviour
We used to default to using the jobname as the filename,
if it wasn't a regular existing file. So do that again.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 14 Mar 2007 13:15:45 +0000 (14:15 +0100)]
Merge branch 'master' of ssh://brick.kernel.dk/data/git/fio
Jens Axboe [Wed, 14 Mar 2007 13:14:48 +0000 (14:14 +0100)]
posixaio engine: better handling of partial completions
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 14 Mar 2007 12:58:54 +0000 (13:58 +0100)]
Don't free ->files
It may be shared.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 14 Mar 2007 12:28:31 +0000 (13:28 +0100)]
Help update
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 14 Mar 2007 12:24:07 +0000 (13:24 +0100)]
Add suboption help texts
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 14 Mar 2007 12:02:53 +0000 (13:02 +0100)]
Left justify help output
The right justification may look prettier, but it's harder
to read and find the option.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 14 Mar 2007 10:37:39 +0000 (11:37 +0100)]
Fio version 1.14
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 14 Mar 2007 10:21:27 +0000 (11:21 +0100)]
New/old file mix fix
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 14 Mar 2007 09:59:33 +0000 (10:59 +0100)]
Update io engine comments
Some were plain wrong, mainly because they were born through
copying an existing engine.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 14 Mar 2007 09:34:47 +0000 (10:34 +0100)]
mmap engine: make sure that page unaligned syncs work
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 14 Mar 2007 08:39:06 +0000 (09:39 +0100)]
recurse_dir(): Move . and .. check earlier
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 14 Mar 2007 08:16:09 +0000 (09:16 +0100)]
Better handling of file creation vs existing files
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 14 Mar 2007 08:15:42 +0000 (09:15 +0100)]
Fix for bs shrinkage
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>