blktrace is a block layer IO tracing mechanism which provides detailed information about request queue operations up to user space. There are three major components that are provided:
The blktrace and blkparse utilities and associated kernel patch are provided as part of the following git repository:
git://git.kernel.org/pub/scm/linux/kernel/git/axboe/blktrace.git bt
The following sections outline some quick steps towards utilizing blktrace. Some of the specific instructions below may need to be tailored to your environment.
As noted above, the kernel patch along with the blktrace and blkparse utilities are stored in a git repository. One simple way to get going would be:
% git clone git://git.kernel.org/pub/scm/linux/kernel/git/axboe/blktrace.git bt % cd bt % git checkout
A patch for a specific Linux kernel is provided in bt/kernel (where bt is the name of the directory from the above git sequence). The detailed actual patching instructions for a Linux kernel is outside the scope of this document, but the following may be used as a sample template. Note that you may skip this step, if you kernel is at least 2.6.17-rc1.
As an example, bt/kernel contains blk-trace-2.6.14-rc1-git-G2, download linux-2.6.13.tar.bz2 and patch-2.6.14-rc1.bz2
% tar xjf linux-2.6.13.tar.bz2 % mv linux-2.6.13 linux-2.6.14-rc1 % cd linux-2.6.14-rc1 % bunzip2 -c ../patch-2.6.14-rc1.bz2 | patch -p1
At this point you may (optionally) remove linux-2.6.13.tar.bz2 and patch-2.6.14-rc1.bz2.
At this point you should configure the Linux kernel for your specific system - again, outside the scope of this document - and then enable Support for tracing block io actions. To do this, run
% make menuconfig or make xconfig, or edit .config, or ...
and navigate through Device Drivers and Block devices and then down to Support for tracing block io actions and hit Y.
Install the new kernel (and modules...) and reboot.
blktrace utilizes files under the debug file system, and thus must have the mount point set up - mounted on the directory /debug. To do this one may do either of the following:
% mount -t debugfs debugfs /debug
debug /debug debugfs default 0 0
To build and install the tools, execute the following sequence (as root):
% cd bt % make && make install
Now to simply watch what is going on for a specific disk (to stop the trace, hit control-C):
% blktrace -d /dev/sda -o - | blkparse -i - 8,0 3 1 0.000000000 697 G W 223490 + 8 [kjournald] 8,0 3 2 0.000001829 697 P R [kjournald] 8,0 3 3 0.000002197 697 Q W 223490 + 8 [kjournald] 8,0 3 4 0.000005533 697 M W 223498 + 8 [kjournald] 8,0 3 5 0.000008607 697 M W 223506 + 8 [kjournald] 8,0 3 6 0.000011569 697 M W 223514 + 8 [kjournald] 8,0 3 7 0.000014407 697 M W 223522 + 8 [kjournald] 8,0 3 8 0.000017367 697 M W 223530 + 8 [kjournald] 8,0 3 9 0.000020161 697 M W 223538 + 8 [kjournald] 8,0 3 10 0.000024062 697 D W 223490 + 56 [kjournald] 8,0 1 11 0.009507758 0 C W 223490 + 56 [0] 8,0 1 12 0.009538995 697 G W 223546 + 8 [kjournald] 8,0 1 13 0.009540033 697 P R [kjournald] 8,0 1 14 0.009540313 697 Q W 223546 + 8 [kjournald] 8,0 1 15 0.009542980 697 D W 223546 + 8 [kjournald] 8,0 1 16 0.013542170 0 C W 223546 + 8 [0] ... ^C ... CPU1 (8,0): Reads Queued: 0, 0KiB Writes Queued: 7, 128KiB Read Dispatches: 0, 0KiB Write Dispatches: 7, 128KiB Reads Completed: 0, 0KiB Writes Completed: 11, 168KiB Read Merges: 0 Write Merges: 25 IO unplugs: 0 Timer unplugs: 0 ... CPU3 (8,0): Reads Queued: 0, 0KiB Writes Queued: 1, 28KiB Read Dispatches: 0, 0KiB Write Dispatches: 1, 28KiB Reads Completed: 0, 0KiB Writes Completed: 0, 0KiB Read Merges: 0 Write Merges: 6 IO unplugs: 0 Timer unplugs: 0 Total (8,0): Reads Queued: 0, 0KiB Writes Queued: 11, 168KiB Read Dispatches: 0, 0KiB Write Dispatches: 11, 168KiB Reads Completed: 0, 0KiB Writes Completed: 11, 168KiB Read Merges: 0 Write Merges: 31 IO unplugs: 0 Timer unplugs: 3 Events (8,0): 89 entries, 0 skips
A btrace script is included in the distribution to ease live tracing of devices. The above could also be accomplished by issuing:
% btrace /dev/sda
By default, btrace runs the trace in quiet mode so it will not include statistics when you break the run. Add the -S option to get that dumped as well.
The previous section showed typical file system io actions, but blktrace can also show SCSI commands going in and out of the queue as submitted by applications using the SCSI Generic (sg) interface.
% btrace /dev/cdrom [...] 3,0 0 25 0.004884107 13528 G R 0 + 0 [inquiry] 3,0 0 26 0.004890361 13528 I R 56 (12 00 00 00 38 ..) [inquiry] 3,0 0 27 0.004891223 13528 P R [inquiry] 3,0 0 28 0.004893250 13528 D R 56 (12 00 00 00 38 ..) [inquiry] 3,0 0 29 0.005344910 0 C R (12 00 00 00 38 ..) [0]
Here we see a program issuing an INQUIRY command to the CDROM device. The program requested a read of 56 bytes of data, the CDB is included in parenthesis after the data length. The completion event shows shows that the command completed successfully. Tracing SCSI commands can be very useful for debugging problems with programs talking directly to the device. An example of that would be cdrecord burning.
Another way to run blktrace is to have blktrace save data away for later formatting by blkparse. This would be useful if you want to get measurements while running specific loads.
To do this, one would specify the device (or devices) to be watched. Then go run you test cases. Stop the trace, and at your leisure utilize blkparse to see the results.
In this example, devices /dev/sdaa, /dev/sdc and /dev/sdo are used in an LVM volume called adb3/vol.
% blktrace /dev/sdaa /dev/sdc /dev/sdo & [1] 9713 % % mkfs -t ext3 /dev/adb3/vol mke2fs 1.35 (28-Feb-2004) Filesystem label= OS type: Linux Block size=4096 (log=2) Fragment size=4096 (log=2) 16793600 inodes, 33555456 blocks 1677772 blocks (5.00%) reserved for the super user First data block=0 Maximum filesystem blocks=4294967296 1025 block groups 32768 blocks per group, 32768 fragments per group 16384 inodes per group Superblock backups stored on blocks: 32768, 98304, 163840, 229376, 294912, 819200, 884736, 1605632, 2654208, 4096000, 7962624, 11239424, 20480000, 23887872 Writing inode tables: done Creating journal (8192 blocks): done Writing superblocks and filesystem accounting information: done This filesystem will be automatically checked every 27 mounts or 180 days, whichever comes first. Use tune2fs -c or -i to override. % % kill -15 9713
Then you could process the events later:
% % blkparse sdaa sdc sdo > events % less events 8,32 1 1 0.000000000 9728 G R 384 + 32 [mkfs.ext3] 8,32 1 2 0.000001959 9728 P R [mkfs.ext3] 8,32 1 3 0.000002446 9728 Q R 384 + 32 [mkfs.ext3] 8,32 1 4 0.000005110 9728 D R 384 + 32 [mkfs.ext3] 8,32 3 5 0.000200570 0 C R 384 + 32 [0] 8,224 3 1 0.021658989 9728 G R 384 + 32 [mkfs.ext3] ... 65,160 3 163392 41.117070504 0 C W 87469088 + 1376 [0] 8,32 3 163374 41.122683668 0 C W 88168160 + 1376 [0] 65,160 3 163393 41.129952433 0 C W 87905984 + 1376 [0] 65,160 3 163394 41.130049431 0 D W 89129344 + 1376 [swapper] 65,160 3 163395 41.130067135 0 D W 89216704 + 1376 [swapper] 65,160 3 163396 41.130083785 0 D W 89304096 + 1376 [swapper] 65,160 3 163397 41.130099455 0 D W 89391488 + 1376 [swapper] 65,160 3 163398 41.130114732 0 D W 89478848 + 1376 [swapper] 65,160 3 163399 41.130128885 0 D W 89481536 + 64 [swapper] 8,32 3 163375 41.134758196 0 C W 86333152 + 1376 [0] 65,160 3 163400 41.142229726 0 C W 89129344 + 1376 [0] 65,160 3 163401 41.144952314 0 C W 89481536 + 64 [0] 8,32 3 163376 41.147441930 0 C W 88342912 + 1376 [0] 65,160 3 163402 41.155869604 0 C W 89478848 + 1376 [0] 8,32 3 163377 41.159466082 0 C W 86245760 + 1376 [0] 65,160 3 163403 41.166944976 0 C W 89216704 + 1376 [0] 65,160 3 163404 41.178968252 0 C W 89304096 + 1376 [0] 65,160 3 163405 41.191860173 0 C W 89391488 + 1376 [0] ... Events (sdo): 0 entries, 0 skips CPU0 (65,160): Reads Queued: 0, 0KiB Writes Queued: 9, 5,520KiB Read Dispatches: 0, 0KiB Write Dispatches: 0, 0KiB Reads Completed: 0, 0KiB Writes Completed: 0, 0KiB Read Merges: 0 Write Merges: 336 IO unplugs: 0 Timer unplugs: 0 CPU1 (65,160): Reads Queued: 2,411, 38,576KiB Writes Queued: 769, 425,408KiB Read Dispatches: 2,407, 38,512KiB Write Dispatches: 118, 61,680KiB Reads Completed: 0, 0KiB Writes Completed: 0, 0KiB Read Merges: 0 Write Merges: 25,819 IO unplugs: 0 Timer unplugs: 4 CPU2 (65,160): Reads Queued: 2, 32KiB Writes Queued: 18, 10,528KiB Read Dispatches: 2, 32KiB Write Dispatches: 3, 1,344KiB Reads Completed: 0, 0KiB Writes Completed: 0, 0KiB Read Merges: 0 Write Merges: 640 IO unplugs: 0 Timer unplugs: 0 CPU3 (65,160): Reads Queued: 20,572, 329,152KiB Writes Queued: 594, 279,712KiB Read Dispatches: 20,576, 329,216KiB Write Dispatches: 1,474, 740,720KiB Reads Completed: 22,985, 367,760KiB Writes Completed: 1,390, 721,168KiB Read Merges: 0 Write Merges: 16,888 IO unplugs: 0 Timer unplugs: 0 Total (65,160): Reads Queued: 22,985, 367,760KiB Writes Queued: 1,390, 721,168KiB Read Dispatches: 22,985, 367,760KiB Write Dispatches: 1,595, 803,744KiB Reads Completed: 22,985, 367,760KiB Writes Completed: 1,390, 721,168KiB Read Merges: 0 Write Merges: 43,683 IO unplugs: 0 Timer unplugs: 4 ...
The blktrace utility extracts event traces from the kernel (via the relaying through the debug file system). Some background details concerning the run-time behaviour of blktrace will help to understand some of the more arcane command line options:
Alternatively, one may specify the entire mask utilizing a hexadecimal value that is version-specific. (Requires understanding of the internal representation of the filter mask.)
The device portion of the event file name can be changed via the -o option.
Short | Long | Description |
-A hex-mask | -set-mask=hex-mask | Set filter mask to hex-mask |
-a mask | -act-mask=mask | Add mask to current filter (see below for masks) |
-b size | -buffer-size=size | Specifies buffer size for event extraction (scaled by ![]() |
-d dev | -dev=dev | Adds dev as a device to trace |
-k | -kill | Kill on-going trace |
-n num-sub | -num-sub=num-sub | Specifies number of buffers to use |
-o file | -output=file | Prepend file to output file name(s) |
-r rel-path | -relay=rel-path | Specifies debugfs mount point |
-V | -version | Outputs version |
-w seconds | -stopwatch=seconds | Sets run time to the number of seconds specified |
barrier | barrier attribute |
complete | completed by driver |
fs | FS requests |
issue | issued to driver |
pc | packet command events |
queue | queue operations |
read | read traces |
requeue | requeue operations |
sync | synchronous attribute |
write | write traces |
The blkparse utility will attempt to combine streams of events for various devices on various CPUs, and produce a formatted output of the event information. As with blktrace, some details concerning blkparse will help in understanding the command line options presented below.
blkparse may be run in a live manner concurrently with blktrace by specifying -i - to blkparse, and combining it with the live option for blktrace. An example would be:
% blktrace -d /dev/sda -o - | blkparse -i -
By default, blkparse sends formatted data to standard output. This may be changed via the -o option, or text output can be disabled via the-O option. A merged binary stream can be produced using the -d option.
The output from blkparse can be tailored for specific use - in particular, to ease parsing of output, and/or limit output fields to those the user wants to see. The data for fields which can be output include:
Field | Description |
Specifier | |
a | Action, a (small) string (1 or 2 characters) - see table below for more details |
c | CPU id |
C | Command |
d | RWBS field, a (small) string (1-3 characters) - see section below for more details |
D | 7-character string containing the major and minor numbers of the event's device |
(separated by a comma). | |
e | Error value |
m | Minor number of event's device. |
M | Major number of event's device. |
n | Number of blocks |
N | Number of bytes |
p | Process ID |
P | Display packet data - series of hexadecimal values |
s | Sequence numbers |
S | Sector number |
t | Time stamp (nanoseconds) |
T | Time stamp (seconds) |
u | Elapsed value in microseconds (-t command line option) |
U | Payload unsigned integer |
Note that the user can optionally specify field display width, and optionally a left-aligned specifier. These precede field specifiers, with a '%' character, followed by the optional left-alignment specifer (-) followed by the width (a decimal number) and then the field.
Thus, to specify the command in a 12-character field that is left aligned:
-f "%-12C"
Act | Description |
A | IO was remapped to a different device |
B | IO bounced |
C | IO completion |
D | IO issued to driver |
F | IO front merged with request on queue |
G | Get request |
I | IO inserted onto request queue |
M | IO back merged with request on queue |
P | Plug request |
Q | IO handled by request queue code |
S | Sleep request |
T | Unplug due to timeout |
U | Unplug request |
X | Split |
The standard header (or initial fields displayed) include:
"%D %2c %8s %5T.%9t %5p %2a %3d "
Breaking this down:
Seeing this in action:
8,0 3 1 0.000000000 697 G W 223490 + 8 [kjournald]
The header is the data in this line up to the 223490 (starting block).
The default output for all event types includes this header.
If no payload is present, the sector and number of blocks are presented (with an intervening plus (+) character). If the -t option was specified, then the elapsed time is presented. In either case, it is followed by the error value for the completion.
If no payload is present, the sector and number of blocks are presented (with an intervening plus (+) character). If the -t option was specified, then the elapsed time is presented (in parenthesis). In either case, it is followed by the command associated with the event (surrounded by square brackets).
The blktrace facility provides an efficient event transfer mechanism which supplies block IO layer state transition data via the relay filesystem. This section provides some details as to the interfaces blktrace utilizes in the kernel to effect this. It is good background data to help understand some of the outputs and command-line options above.
BLK_TA_QUEUE | (RQ) Command queued to request_queue. |
(BIO) Command queued by elevator. | |
BLK_TA_BACKMERGE | Back merging elevator operation |
BLK_TA_FRONTMERGE | Front merging elevator operation |
BLK_TA_GETRQ | Free request retrieved. |
BLK_TA_SLEEPRQ | No requests available, device unplugged. |
BLK_TA_REQUEUE | Request requeued. |
BLK_TA_ISSUE | Command set to driver for request_queue. |
BLK_TA_COMPLETE | Command completed by driver. |
BLK_TA_PLUG | Device is plugged |
BLK_TA_UNPLUG_IO | Unplug device as IO is made available. |
BLK_TA_UNPLUG_TIMER | Unplug device after timer expired. |
BLK_TA_INSERT | Insert request into queue. |
BLK_TA_SPLIT | BIO split into 2 or more requests. |
BLK_TA_BOUNCE | BIO was bounced |
BLK_TA_REMAP | BIO was remapped |
This document was generated using the LaTeX2HTML translator Version 2002-2-1 (1.71)
Copyright © 1993, 1994, 1995, 1996,
Nikos Drakos,
Computer Based Learning Unit, University of Leeds.
Copyright © 1997, 1998, 1999,
Ross Moore,
Mathematics Department, Macquarie University, Sydney.
The command line arguments were:
latex2html blktrace.tex -split 0
The translation was initiated by krzysiek on 2007-01-30