Zhang, Yanmin [Wed, 20 May 2009 09:30:55 +0000 (11:30 +0200)]
Add pre_read option
With this option set, files will be pre-read into memory before
starting the given IO operation(s).
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 20 May 2009 08:49:44 +0000 (10:49 +0200)]
Add appropriate warning in mmap engine for large maps failing
On 32-bit archs, this can trigger for large files. This should be fixed by
mapping lazily, for now just catch the error.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 20 May 2009 08:45:12 +0000 (10:45 +0200)]
Clear f on error get_next_file_rr()
Otherwise we could return an invalid non-open file and trigger and
assertion failure later in the process.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 19 May 2009 08:35:38 +0000 (10:35 +0200)]
Return EILSEQ on verify errors instead of EIO
Suggested by Randy Dunlap <randy.dunlap@oracle.com>
This makes it easier to differentiate between an error to read/write
from disk versus getting the wrong data back.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 18 May 2009 11:08:12 +0000 (13:08 +0200)]
Include transfer size in the clat/slat measurements
It's useful to know the block size that matches the given latency, so
add that to the end.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Fabio Checconi [Mon, 18 May 2009 10:54:50 +0000 (12:54 +0200)]
Fix garbling of clat/slat logs if a shared log file is used
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 13 May 2009 05:51:19 +0000 (07:51 +0200)]
Merge branch 'master' of ssh://router/data/git/fio
Randy Dunlap [Wed, 13 May 2009 05:51:05 +0000 (07:51 +0200)]
fio howto typo fixes
Fix a few typos in the fio HOWTO file.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Randy Dunlap [Wed, 13 May 2009 05:50:38 +0000 (07:50 +0200)]
fio docs corrections
Correct fio mailing list address and unit of time for cpuchunks.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 11 May 2009 08:37:33 +0000 (10:37 +0200)]
net engine: fix compilation warnings on Solaris
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 7 May 2009 12:21:59 +0000 (14:21 +0200)]
Fio 1.26.2
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 30 Apr 2009 06:07:43 +0000 (08:07 +0200)]
Merge branch 'master' of ssh://brick.kernel.dk/data/git/fio
Jens Axboe [Wed, 29 Apr 2009 07:52:10 +0000 (09:52 +0200)]
'V' is a running state, not an idle state. Correct HOWTO.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 29 Apr 2009 07:48:04 +0000 (09:48 +0200)]
Align io units to processor cache line size
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Radha Ramachandran [Wed, 29 Apr 2009 05:23:26 +0000 (07:23 +0200)]
Fix name of the slat log file
Currently the slat and clat logs go to the same clat log file when the
latency log file is specified.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Radha Ramachandran [Tue, 28 Apr 2009 06:12:56 +0000 (08:12 +0200)]
Don't populate random data for verify reads
I was seeing some performance drop during the read verification phase
of a test, and from the code in io_u.c in get_io_u function, we
prepare/populate the buffer in the io_u structure based on the verify
patterns/options.
This makes sense when we are doing writes, but I dont understand why
we do this for the read phase when this data is going to be
overwritten anyways(and in case of truncated reads, we do modify the
buf_len).
So based on that I changed the code to populate the buffer(io_u->buf)
only if its a write with verify enabled.
This works for my tests, but I do not know if there was a reason why
this was populated for reads as well to begin with.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 27 Apr 2009 06:05:45 +0000 (08:05 +0200)]
Fio 1.26.1
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 27 Apr 2009 06:00:48 +0000 (08:00 +0200)]
Filename must be parsed after directory option
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Fri, 24 Apr 2009 12:23:59 +0000 (14:23 +0200)]
Update bssplit documentation in fio.1
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Kedar Sovani [Thu, 23 Apr 2009 11:47:41 +0000 (13:47 +0200)]
fio: Include support for ARM
Assembly instructions picked from include/asm-arm/system.h
Signed-off-by: Kedar Sovani <kedars@marvell.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 22 Apr 2009 07:48:34 +0000 (09:48 +0200)]
Fio 1.26
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 22 Apr 2009 07:47:43 +0000 (09:47 +0200)]
Add sample job for doing a quick SSD performance analysis
Nothing SSD special about it really, but this sort of test is
useful on SSD drives to see if they are crap or not.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Radha Ramachandran [Wed, 22 Apr 2009 06:17:42 +0000 (08:17 +0200)]
Fix issue with random offset being off if io_size != real_file_size
get_next_free_block() would generate offset that were larger than the
requested size, either causing failures or just causing the test
to exit prematurely.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 21 Apr 2009 07:17:31 +0000 (09:17 +0200)]
Remove debug printf()
Left in by mistake.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 21 Apr 2009 06:29:55 +0000 (08:29 +0200)]
Add support for different weights for reads and writes with bssplit
You can now do:
bssplit=4k/20:8k/80,1k/50:4k/50
To have 20% 4k reads and 80% 8k reads, while having 50% 1k writes
and 50% 4k writes. This is identical to how the bs= option treats
reads and writes.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 21 Apr 2009 06:28:17 +0000 (08:28 +0200)]
Allow IO logging even for diskless engines
Helps to verify IO patterns.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 20 Apr 2009 10:39:17 +0000 (12:39 +0200)]
splice engine: fix vmsplice() problem with block sizes larger than a pipe
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 20 Apr 2009 09:39:04 +0000 (11:39 +0200)]
Fio 1.25.2
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 20 Apr 2009 08:06:02 +0000 (10:06 +0200)]
Fix throughput reporting
It was off by 1024/1024 :/
Initial patch from Carl Henrik Lunde <chlunde@ping.uio.no>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Carl Henrik Lunde [Mon, 20 Apr 2009 06:41:37 +0000 (08:41 +0200)]
Fix crash with multiple files
There's an off-by-one in add_file() when it calculates the size
for the file array.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 8 Apr 2009 05:54:20 +0000 (07:54 +0200)]
Fio 1.25.1
Silly segfault with cpuio/null engines.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 7 Apr 2009 20:21:28 +0000 (22:21 +0200)]
Merge branch 'master' of ssh://brick.kernel.dk/data/git/fio
Jens Axboe [Tue, 7 Apr 2009 20:20:56 +0000 (22:20 +0200)]
Use td_io_get_file_size() instead of calling ->get_file_size()
It may be NULL on some engines, causing a crash.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 7 Apr 2009 11:40:04 +0000 (13:40 +0200)]
Fio 1.25
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 7 Apr 2009 11:28:12 +0000 (13:28 +0200)]
Reseed random generator for loop/time based runs
Then we get repeatable 2nd/3rd/etc runs. This is important
if the file layout was random, otherwise subsequent runs will
get very different results.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 7 Apr 2009 11:06:43 +0000 (13:06 +0200)]
File reset should clear random map as well
Fixes infinite loop on time based runs, where it repeatedly
fails getting a new offset.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 7 Apr 2009 10:52:08 +0000 (12:52 +0200)]
Merge branch 'master' of ssh://router/data/git/fio
Jens Axboe [Tue, 7 Apr 2009 10:51:45 +0000 (12:51 +0200)]
aio: fixed a signedness issue that could cause abort on strace
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Sat, 14 Mar 2009 19:17:36 +0000 (20:17 +0100)]
Add some missing options to the man page
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Sat, 14 Mar 2009 19:08:47 +0000 (20:08 +0100)]
Man page typo on zero_buffers
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 11 Mar 2009 10:00:13 +0000 (11:00 +0100)]
Add blockalign/ba option
Allows the job to specify alignment of the IO specifically, instead
of relying on using blocksize as the offset alignment.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 10 Mar 2009 07:57:34 +0000 (08:57 +0100)]
Fix integer store bug in parser
With commit
f7fa2653589f9ecf0d3d106286f17620bcf32de0 I forgot to check
the type, so we would mis-store on platforms where sizeof(int) is different
from sizeof(long).
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 9 Mar 2009 13:22:56 +0000 (14:22 +0100)]
Add some missing options to the man page
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 9 Mar 2009 13:20:20 +0000 (14:20 +0100)]
Get rid of int vs siint difference
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 9 Mar 2009 13:16:47 +0000 (14:16 +0100)]
Update HOWTO with 0x base for siint
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 9 Mar 2009 13:15:21 +0000 (14:15 +0100)]
Allow 0x (or 0X) prefix for any int value to indicate hex base
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Fri, 6 Mar 2009 19:29:29 +0000 (20:29 +0100)]
Fix failure to strip end of line with comment
We left a space at the end, which confused the postfix checking.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Fri, 6 Mar 2009 07:11:00 +0000 (08:11 +0100)]
Merge branch 'master' of ssh://brick.kernel.dk/data/git/fio
Jens Axboe [Thu, 5 Mar 2009 20:23:18 +0000 (21:23 +0100)]
Update smalloc section in README
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 5 Mar 2009 20:14:21 +0000 (21:14 +0100)]
Use strict matching for --section command line option
Don't limit to string length of passed in section, otherwise we could
be matching more than one section.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 5 Mar 2009 09:13:58 +0000 (10:13 +0100)]
Avoid opening files until they are used
Fio still opens and creates all files, just to check the size of them. Add
a specialized IO engine op for getting the size of a file and use that
instead.
This also cleans a lot of things up. Note that the IO engine version is now
bumped to 10, meaning that external engines will have to separate the
file open from the size checking.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 4 Mar 2009 19:17:22 +0000 (20:17 +0100)]
Need to clear ->files_size when ->files is cleared
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 4 Mar 2009 15:03:49 +0000 (16:03 +0100)]
More file setting bug fixes
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 4 Mar 2009 14:37:52 +0000 (15:37 +0100)]
Botch edit in files setup
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 4 Mar 2009 14:28:31 +0000 (15:28 +0100)]
Missing fio.h update
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 4 Mar 2009 14:27:42 +0000 (15:27 +0100)]
Realloc td->files to full size immediately
For adding many files, we do as many reallocs() as we have files. This is
slow, so do full allocs instead.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 4 Mar 2009 11:53:13 +0000 (12:53 +0100)]
Add create_on_open option
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Please enter the commit message for your changes. Lines starting
Jens Axboe [Wed, 4 Mar 2009 11:52:01 +0000 (12:52 +0100)]
Fix file open leak
Introduced by commit
87b10676354d10be02864ad615e1e9fe5a2376ab
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 4 Mar 2009 08:39:47 +0000 (09:39 +0100)]
Fix bug with openfiles
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 4 Mar 2009 07:41:31 +0000 (08:41 +0100)]
Fix bug in file selection
We need to check both the OPEN and CLOSING flag, since CLOSE will not
be set until the file has really been closed.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 4 Mar 2009 07:27:37 +0000 (08:27 +0100)]
Add 'sequential' file_service_type
This service type will keep a file open until IO to it is completely
done, before moving on to the next available file.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 19 Feb 2009 20:32:12 +0000 (21:32 +0100)]
Allow 'norandommap' with verify
We used to turn off verify completely if norandommap was given, since
fio does not track block rewrites. But we can easily track rewrites,
as long as we don't have to track various extent sizes. So relax
the restriction to only cover norandommap with differing block sizes
and adapt log_io_piece() to always use rbtree inserts for cases where
we do need to check for potential overwrites.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Fri, 6 Feb 2009 13:12:03 +0000 (14:12 +0100)]
Fio 1.24
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 2 Feb 2009 14:07:37 +0000 (15:07 +0100)]
Check for a valid block device size
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 19 Jan 2009 12:26:36 +0000 (13:26 +0100)]
net engine: notify receiver of UDP link close
Also optimize the send/receive path, only use poll() if the non-blocking
send/recv returned EAGAIN.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 19 Jan 2009 11:49:48 +0000 (12:49 +0100)]
Fix broken rate check
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 19 Jan 2009 11:38:27 +0000 (12:38 +0100)]
net engine: termination fixes
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 19 Jan 2009 11:33:51 +0000 (12:33 +0100)]
net engine: don't pass in flags
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 19 Jan 2009 09:17:06 +0000 (10:17 +0100)]
net engine: use poll() always for sending/receiving
Avoids getting stuck waiting for data and not accepting signals.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Fri, 16 Jan 2009 18:17:35 +0000 (19:17 +0100)]
Add comment on UDP on examples/netio
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Fri, 16 Jan 2009 18:15:33 +0000 (19:15 +0100)]
Missing text in iodepth_batch* help entries
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Fri, 16 Jan 2009 18:15:07 +0000 (19:15 +0100)]
libaio engine: get rid of non-async behaviour with buffered IO
It may not be true, just get rid of it.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Fri, 16 Jan 2009 12:22:26 +0000 (13:22 +0100)]
net engine: accept both upper and lowercase tcp/udp.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Fri, 16 Jan 2009 12:21:15 +0000 (13:21 +0100)]
net engine: add UDP support
The filename= option now accepts a third parameter designating the
protocol type, udp or tcp.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 7 Jan 2009 09:05:14 +0000 (10:05 +0100)]
Add platform section to README
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 6 Jan 2009 13:22:30 +0000 (14:22 +0100)]
OS includes need errno.h
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 6 Jan 2009 10:30:08 +0000 (11:30 +0100)]
smalloc: get rid of MP_SAFE define and lock checks
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 5 Jan 2009 18:07:13 +0000 (19:07 +0100)]
Some more pthread_* call mishandled errors
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 5 Jan 2009 18:06:39 +0000 (19:06 +0100)]
Add --debug=mutex
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 5 Jan 2009 18:04:15 +0000 (19:04 +0100)]
Make sure os mutex define is visible in mutex.c
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 5 Jan 2009 08:56:29 +0000 (09:56 +0100)]
Only use process shared mutexes on support platforms
On FreeBSD, suggest the use of threads instead.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 5 Jan 2009 08:45:13 +0000 (09:45 +0100)]
Fixup handling of pthread_mutex*() error returns
They return the errno, not 0/-1. Thanks to
Fabio Checconi <fabio@gandalf.sssup.it>
for noticing this.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 5 Jan 2009 08:40:13 +0000 (09:40 +0100)]
Add fio_cpuset_exit() empty handler for platforms that don't support affinity
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 22 Dec 2008 10:29:50 +0000 (11:29 +0100)]
Include crc32c and crc32c-intel checksums in FreeBSD/Solaris
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 22 Dec 2008 10:28:26 +0000 (11:28 +0100)]
Solaris pset fixes
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Fri, 12 Dec 2008 19:54:15 +0000 (20:54 +0100)]
Solaris compilation fix
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Fri, 12 Dec 2008 19:51:40 +0000 (20:51 +0100)]
CPU set creation and destruction can fail on some platforms
Check for that and error out if necessary.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Fri, 12 Dec 2008 19:42:26 +0000 (20:42 +0100)]
Support for CPU binding in Solaris
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Fri, 12 Dec 2008 19:40:27 +0000 (20:40 +0100)]
Add a cpuset exit function
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Fri, 12 Dec 2008 19:02:58 +0000 (20:02 +0100)]
Prepare for support for CPU binding on other platforms
The interface was Linux/glibc centric, since that is what got there first.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Wed, 10 Dec 2008 12:24:12 +0000 (13:24 +0100)]
__usec_sleep() really just spins, rename it
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Tue, 9 Dec 2008 07:04:29 +0000 (08:04 +0100)]
Long line style fix + comment
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 8 Dec 2008 19:31:43 +0000 (20:31 +0100)]
Don't allow higher CPU counts than what the system has
Basically just catch the condition before setting up the job, job files,
etc.
Also fixes a big with cpumask, it set the wrong part.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 8 Dec 2008 14:03:12 +0000 (15:03 +0100)]
Make sure that the given CPU range is within the bounds of the OS
glibc provides CPU_SETSIZE as the maximum CPU count.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 8 Dec 2008 13:10:52 +0000 (14:10 +0100)]
Add gtod_cpu option for pinning gettimeofday() to a single CPU
Similar to what real life products sometimes do, offload gettimeofday()
calls to a single CPU and have that update the current time into a shared
memory location. This option pins a specific CPU for that job, and excludes
it from participating in any of the IO jobs.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 8 Dec 2008 13:06:42 +0000 (14:06 +0100)]
Properly strip end-of-option blanks
We don't want to fail with
option_foo = 0
when
option_foo=0
works.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Mon, 8 Dec 2008 10:37:01 +0000 (11:37 +0100)]
Allow ranges of CPUs in cpus_allowed option
For a larger number of CPUs, it's a bit of a bother having to set
each one individually. Plus, you typically want to bind to a range of
CPUs, to keep the job core local (or whatever your preference). So extend
cpus_allowed to take ranges of CPU numbers as well. Now you can do:
cpus_allowed=0-7
and have your job confined to CPUs 0 through 7.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Fri, 5 Dec 2008 12:17:16 +0000 (13:17 +0100)]
Fix bogus 'may be used uninitialized' warning from older gcc versions
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Fri, 5 Dec 2008 11:57:11 +0000 (12:57 +0100)]
Add comment on cpumask not working with a CPU count > 32
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Jens Axboe [Thu, 4 Dec 2008 13:33:41 +0000 (14:33 +0100)]
smalloc: pool->file is only used in add_pool(), so make it local
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>