CHANGELOG,v 1.468 2004/09/10 04:13:41 ballen4705 Exp $ The

CHANGELOG for smartmontools
$Id: CHANGELOG,v 1.468 2004/09/10 04:13:41 ballen4705 Exp $
The most recent version of this file is:
Maintainers / Developers Key:
[BA] Bruce Allen
[EB] Erik Inge Bolsø
[SB] Stanislav Brabec
[PC] Peter Cassidy
[CD] Caper Dik
[CF] Christian Franke
[GF] Guilhem Frézou
[DG] Douglas Gilbert
[GG] Guido Guenther
[GK] Geoff Keating
[DK] Dr. David Kirkby
[KM] Kai Mäkisara
[EM] Eduard Martinescu
[FM] Frédéric L. W. Meunier
[KS] Keiji Sawada
[DS] David Snyder
[SS] Sergey Svishchev
[PW] Phil Williams
[LW] Leon Woestenberg
[RZ] Richard Zybert
smartmontools 5.33 Experimental Release
[BA] smartctl: ATA disks, if SMART ATTRIBUTE THRESHOLDS page has ID
errors with some Attributes having NULL IDs, print Attribute
info anyway (but issuing a warning to the user).
[DG] [SCSI] Decode Last n error events log page; decode track following
and positioning errors [Hitachi]
[EM] FreeBSD: another tweak, __packed__ introduced in Version 5.0040
[EM] Cleaner tweak of fixes for FreeBSD 4.x.
[EM] Fix compilation errors under FreeBSD 4.x, as it is still using
and old GCC
[EM] Remove 3ware/FreeBSD specific files and just include pieces we need
[DG] Add logic in smartd to detect 3ware, Marvell controllers and SATA
disks behind an ATA-SCSI simulator (in Linux). If specific device
types are not given and they are picked in a general SCSI device
scan then warn and skip.
[GG] insert correct path to smartd into smartd’s init script
[BA] Changed all default paths in documentation to reflect /usr/local as
default path prefix. This affects on-line man pages, primarily.
[DS] Added support for OpenBSD
[BA] Added another environment variable SMART_FULLMESSAGE set by
the smartd mailing feature, and modified examplescripts/Example1
to illustrate it.
[BA] Fixed potentially misleading messages of the form:
XXX failed: success
[DG] emit warning if SATA disk detected using libata in Linux; then exit
[PW] Added Seagate U10 family, Hitachi Travelstar 7K60, Fujitsu MHR2020AT,
and QUANTUM FIREBALLP AS20.5 to knowndrives table.
[DG] Detect 3ware and Marvell controllers from SCSI INQUIRY vendor string
and suggest usage of appropriate ’-d’ argument in smartctl.
[LW] Tested the RELEASE_5_33_WITH_MARVELL_SUPPORT branch on
actual Marvell 88SX5041 hardware, with success.
Merged into HEAD.
[BA] Fixed nasty DEVICESCAN bug
[BA] Checked in RELEASE_5_33_WITH_MARVELL_SUPPORT branch with
some Marvell support.
[BA] Additional modifications of Ed’s controller scheme. Fixed
broken 3ware support under linux, problems with scanning
devices in smartd, and other small problems.
[EM] Minor change to FreeBSD inclusion of ’twe’ include files. Add
code to check if they exising in /usr/include/sys to use those
in preference to ones added here
[EM] Very preliminary support attempt for 3Ware controllers under
FreeBSD. Also, switched ’escalade_type/escalade_port’ to
’controler_type/controller_port’ and moved away from
’tryata/tryscsi’ to using new ’controller*’ variables to
determine which controller type (ATA/SCSI/3Ware) to use.
[GK] Added initscript support for Darwin.
[CF] Windows smartd: Added ability to run smartd as a windows service,
including new commands "smartd install ..." and "smartd remove"
to install and remove the service registry entry.
[BA] smartd: warn user if -s regexp regular expression contains
characters other than 0123456789.*()|+?[-]{}:=SLCO since such
characters are ’suspicous’ and may indicate a poorly formed
regexp. Extended regular expression gurus: can this list be
reduced somewhat?
[CF] Fixed bug in Windows smartd: Missing close of config file when
configuration is reloaded by smartd daemon.
[CF] Windows smartd: Added mail warning feature using the "Blat"
( mailer as a default.
[PW] Added Maxtor DiamondMax Plus 5120 Ultra ATA 33 series and TOSHIBA
MK3017GAP to knowndrives table.
[CF] Added fixes to build smartmontools on old Linux systems
(libc < 6, Kernel 2.0.x).
[BA] Added ATA minor version identity strings for latest ATA specification
updates: ATA/ATAPI-7 T13 1532D revision 4a and ATA/ATAPI-6 published,
ANSI INCITS 361-2002
[PW] Added Hitachi Travelstar 5K80 family and Fujitsu MHTxxxxAH family to
knowndrives table.
[EM] Fix up compilation under FreeBSD < 5.x
[PW] Added QUANTUM FIREBALL EX3.2A and missing Western Digital Caviar SE
drives to knowndrives table.
[BA] Modified Hitachi Travelstar 80GN family regexp in drive database.
Thanks to [GK/CF] for problem & solution.
[GK] Added os_darwin.[ch]
[PW] Added the following drives to the knowndrives table: IBM Travelstar
48GH, 30GN, and 15GN family; IBM Deskstar 37GP and 34GXP family;
Western Digital WDC WD272AA; Maxtor DiamondMax D540X-4D family;
TOSHIBA MK2016GAP, MK2018GAP, MK2018GAS, MK2023GAS; and
[BA] smartd/smarctl now print build HOST/OS information as part
of startup slogan. This should make it slightly easier to
read bug reports from users.
[RZ] Fixed the DEVICESCAN to do what it was supposed to do - give
error message unless scanning is in progress.
[BA] Update documentation to describe 3ware character devices. Better
error detection for missing/malfunctioning devices behind 3ware
controllers. Now pack 3ware ioctl structures explicitly.
[BA] For ATA devices that support LBA mode, print capacity as part
of smartctl --info
[RZ] Made DEVICESCAN quiet about non-existing devices unless debug
is on.
[DG] treat "unit attention" SCSI warning as try again in some contexts
(test unit ready and mode sense)
[BA] on drives that store max/min rather than min/max, get order
correct in printing temp.
[BA] fixed typo in ’smartctl -h’ output.
Thanks to Gabor Z. Papp.
[BA] linux: clean-up to 3ware/AMCC support; dynamically create
or fix /dev/tw[ae][0-15] device node entries if they don’t
exist or are incorrect. One can now use the character devices
/dev/twe[0-15] OR /dev/sd? for 3ware 6000/7000/8000 series
cards. One must use /dev/twa[0-15] for 3ware 9000 series cards.
Note that selective self-tests now work via /dev/tw[ae] devices.
Next step: documentation.
[BA] linux: experimental "support" for 3ware/AMCC 9000 series
controllers that use the 3w-9xxx driver. This will be in a
state of flux for a few days. Note that this requires the
character interface /dev/twa[0-15].
[DG] linux: extend general SCSI OS interface to use the SG_IO ioctl. If
not available, use the older SCSI_IOCTL_SEND_COMMAND ioctl.
[KS] Solaris/x86: fixed system identification problem in configure
script. Thanks to Stuart Swales.
smartmontools 5.32
[BA] Update link to revised/updated IBM Deskstar Firmware
[CF] Cygwin & Windows: Added missing ASPI manager initialization
with GetASPI32SupportInfo(). Thanks to Nikolai SAOUKH for pointing
this out and providing a patch.
[BA] modified smartd init script to work on whitebox (thanks to
Michael Falzon)
[BA] removed (reverted) additional Attribute definitions from All (or most?) of these
appear to be return code values for the WD Digital Life Guard Utility.
[PW] Added Seagate Medalist 17242, 13032, 10232, 8422, and 4312 to
knowndrives table. Added missing Seagate U Series 5 drives.
[PW] Added the following QUANTUM models to knowndrives table:
[PW] Added missing Western Digital Protege drives to knowndrives table.
[PW] Added Maxtor DiamondMax 40 ATA 66 series and DiamondMax 40 VL Ultra
ATA 100 series to knowndrives table.
[PW] Added the following Hitachi/IBM drives to knowndrives table:
HITACHI_DK14FA-20B, Travelstar 40GNX series, Travelstar 4LP series,
and Travelstar DK23XXB series. Added the missing Travelstar 80GN
[PW] Added Fujitsu MPB series and MPG series to knowndrives table.
the missing Fujitsu MHSxxxxAT drives.
[KS] Solaris: added workaround for dynamic change of time-zone.
[KS] Solaris: fixed problem that cannot detect absence of
auto* tools.
[BA] smartd: added time-zone bug information to man page.
Reverted CF code for _WIN32 case.
[CF] Cygwin & Windows: Added better error messages on IDE/ATA device
open error.
[BA] added additional Attribute definitions from
[BA] smartd: reworked TimeZone bug workaround so it is only invoked
for glibc. Note: this might not be right -- a similar bug may
exist in other platform’s libcs.
[DG] SCSI smartmontools documentation updated [2004/5/6]. See:
[CF] Windows: Fixed reset of TZ=GMT in glibc timezone bug workaround.
smartmontools 5.31
[DG] move SCSI device temperature and start-stop log page output
(smartctl) into --attributes section (was in --info section).
[GG] change default installation location to /usr/local
[CF] Cygwin smartd: Fixed crash on access of SCSI devices after fork().
[PW] Added TOSHIBA MK4018GAS and the following Maxtor drive families
to knowndrives table: DiamondMax D540X-4G, Fireball 541DX,
DiamondMax 3400 Ultra ATA, DiamondMax Plus 6800 Ultra ATA 66.
[PW] Added missing Maxtor DiamondMax 16, DiamondMax D540X-4K, and
DiamondMax Plus 45 Ulta ATA 100 drives to knowndrives table.
[PW] Added ExcelStor J240, Hitachi Travelstar 80GN family, Fujitsu
MHTxxxxAT family, and IBM Deskstar 25GP and 22GXP families to
knowndrives table.
[CF] Cygwin smartd: Added workaround for missing SIGQUIT via keyboard:
To exit smartd in debug mode, type CONTROL-C twice.
[BA] smartctl: printing of the selective self-test log is now
controlled by a new option: -l selective
[BA] Added entries for Samsung firmware versions -25 to -39 based
on latest info about firmware bug fixes.
[PW] Added Seagate U Series X family, Seagate U8 family, and Seagate
Medalist 8641 family to knowndrives table.
[CF] smartd: Added exit values 5/6 for missing/unreadable config file.
[BA] smartd: now monitor the Current Pending Sector count (Attribute 197)
and the Offline Pending Sector Count (Attribute 198). Log a
warning (and send an email, if so configured) if the raw count
is nonzero. These are controlled by new Directives: -C and -U.
Currently they are enabled by default.
[CF] Added option -c FILE, --configfile=FILE to smartd to specify
an alternate configuration FILE or ’-’ for standard input.
[KS] now searches for -lnsl and -lsocket for Solaris.
[CF] Win32/native smartd: Added thread to combine several syslog output
lines into one single event log entry.
[CF] Win32 smartd: Added DEVICESCAN for SCSI/ASPI devices.
[GG] Use gethostbyname() the get the DNS domain since getdomainname()
returns the NIS domain when sending mails from smartd.
[GG] pass smartd_opts to smartd on startup, read distribution
specific configuration files if found
[SS] smartctl: added NetBSD support for Selective Self-tests.
[BA] smartd.conf example configuration file now has all examples
commented out except for ’DEVICESCAN’.
[CF] Win32/native smartd: Added ability to display warning "emails"
as message box by "-m msgbox" directive. With "-m sysmsgbox",
a system modal (always on top) message box is shown.
[BA] smartctl: printing of self-test log for disks that support
Selective self-testing now shows the status of the (optional)
read-scan after the selective self test. Also, changed format
in printing self-test log to print failing LBA in base 10 not
base 16 (more compatible with kernel error messages). Also,
in printing SMART error log, print timestamps in format
[CF] Win32 smartd: Added ability to log to stdout/stderr
(-l local1/2). Toggling debug console still works
if stdout is redirected.
[BA] smartctl: selective self-test log, print current status
in a more detailed way. Allow writing of selective self-test
log provided that no other self-test is underway.
[BA] Linux: eliminated dependency on kernel tree hdreg.h.
[BA] smartctl: -l selftest option now prints Selective self-test
log in addition to the normal self-test log.
Added additional options (-t pending, -t afterselect) to
control remaining Selective Self-test capabilities. Tested
with several Maxtor disks. Modified error message printing
so that munged option messages print at the end not the
start of output.
[CF] Added daemon support to Win32 native version of smartd.
The daemon can be controlled by commands similar to initd
scripts: "smartd status|stop|reload|restart|sigusr1|sigusr2".
[CF] Added minor support for option "-l local[0-7]" to Win32 native
(not Cygwin) version of smartd. If specified, the log output
is written to file "./smartd[1-7]?.log" instead of event log.
[BA] Added Selective Self-test to smartctl (-t selective,M-N).
Currently only supported under Linux; Solaris, NetBSD, FreeBSD
and Windows developers must add WRITE LOG functionality to
[BA] Added workaround for an annoying glibc bug: if you change
timezones, (eg, flying with a laptop from USA to Europe)
localtime() does not notice this in a running
executable, so time that appears in the system log (syslog!)
will be incorrect. See
for additional examples of this bug.
[DG] Set explicit timeouts for SCSI commands (most default to 6 seconds).
Previously a 0 second timeout was meant to be interpreted as a
default timeout but the FreeBSD port had a problem in this area.
[CF] Fixed un-thread-safe exit signal handler for Win32
[BA] Fixed un-thread-safe exit signal handler pointed out
by CF.
[BA] Changed configure script to eliminate warnings under
Solaris from sys/int_type.h conflicts with int64.h
Added header files for umask to smartd.c.
[BA] Man page format change from Werner LEMBERG.
" " changed to \&
[CF] Added os_win32/syslogevt.* event message file tool for Win32
smartd (native+cygwin). May also be useful for other cygwin
programs writing to syslog().
[CF] Added Win32 version of smartd
[BA] Made some changes to man page markup suggested by
Richard Verhoeven to work around bugs in man2html.
Tested not to break anything under Linux and Solaris.
[CF] Moved PrintOut() from utility.c to smart{ctl,d}.c to avoid
syslog() output of smartctl.
[BA] Grew worried that some time-zone names could be very long (eg,
Mitteleuropaische Zeit) and put date string lengths into a
single macro in utility.c
[EM] Updated os_freebsd.c to handle older versions of FreeBSD in a
more appropriate/obvious fashion.
[EM] Modified as FreeBSD installs automake 1.7 as
’automake17’ and NOT ’automake-1.7’
smartmontools 5.30
[PW] Added QUANTUM FIREBALLlct15 30, QUANTUM FIREBALLlct20 40, and
Maxtor 6Y060P0 (DiamondMax Plus 9 60GB) to knowndrives table.
[PW] Added Maxtor MaXLine II family to knowndrives table (thanks to
Brett Russ for submitting the patch).
[BA] Added remaining read/write commands to detailed list of
error log commands that have text descriptions of problem
printed. For commands that support it, print number of failed
sectors at problem LBA.
[BA] Made SuSE section of smartd init script more SuSE 9 compatible.
Thanks to Hans-Peter Jansen.
[CF] Windows smartd: Added IDE/ATA device scan
Added windows device names to,
[BA] smartctl/smartd: user-provided ’-F samsung’ and ’-F samsung2’
command line options/Directives did NOT over-ride preset values
unless user specified ’-P ignore’. Now they will always over-ride
preset values from the database.
[BA] Added error decoding for a few more READ and WRITE commands.
[PW] Added Maxtor MaXLine Plus II, Western Digital Caviar SE (Serial ATA)
series, Hitachi Deskstar 7K250 series, and Ultra ATA 66 models of
the Maxtor DiamondMax Plus 40 series to knowndrives table.
[BA] Added Maxtor Diamondmax 250 GB drives to database. Note that
these model numbers are not listed in Maxtor documentation, but
they exist.
[BA] Removed the ’contact developers’ phrase from the Samsung disk
warning messages.
[PW] Added TOSHIBA MK2017GAP, IBM Deskstar 14GXP and 16GP series,
Fujitsu MPC series, Seagate Barracuda ATA III family, and missing
Seagate Barracuda U Series drives to knowndrives table
[BA] smartd: wrong loglevel for message: Configuration file
/etc/smartd.conf parsed. Changed to LOG_INFO from LOG_CRIT.
Thanks to Emmanuel CHANTREAU for the report.
[CF] Checked in development version of windows code base.
smartmontools 5.29 (note: there was NO 5.28 release)
[BA] smartd: configure script did not set correct directory to search for
smartd.conf based on --prefix argument to ./configure. Thanks to
GG for identifying the problem and fix.
[BA] make clean now removes man pages (generated from *.in) files as well
as object files.
[EM] Correct copying of sense data in FreeBSD SCSI implementation. Thanks
to Sergey Svishchev for noticing the bug.
[BA] On solaris, wrong warning message if no ATA support.
concerns 3ware controller, not ATA.
Warning message
[SS] Added SCSI support for NetBSD.
[BA] on big-endian linux machines, fixed interpretation of HDIO_GET_IDENTITY
to correctly identify ATAPI bit (was byte swapped). This should
eliminate some SYSLOG noise if user queries a packet device (eg, CD
ROM or DVD reader).
[PW] Removed warning for IBM Deskstar 40GV & 75GXP series drives with
A5AA/A6AA firmware. Thanks to Gerald Schnabel.
[PW] Added Toshiba TOS MK3019GAXB SUN30G to knowndrives table
[PW] Added Western Digital Caviar AC12500, AC24300, AC25100, AC36400,
and AC38400 to knowndrives table
[BA] When printing ATA error log, print the LBA at which READ
or WRITE commands failed.
[BA] Changed syntax of error message in smartctl
[BA] Added versioning info (-V options to smartd/smartctl) for
Solaris ATA module.
smartmontools 5.27
[KS] Added ATA/IDE support for Solaris/SPARC (ATA/IDE not yet for
[BA] 3ware controllers: documented that one can monitor any of the
physical disks from any of the 3ware /dev/sd? logical devices.
Better warnings if querying a disk that does not exist.
[PW] Added Hitachi Travelstar DK23DA series, Maxtor DiamondMax Plus 40
series, Western Digital Caviar WDxxxAA, WDxxxBA, and WDxxxAB series
to knowndrives table
[BA] missing ’pragma pack’ on ATA IDENTIFY DEVICE structure may have
caused odd or incorrect results on 64-bit machines.
[BA] smartctl/smartd allow inspection of self-test and error logs even
if disk firmware claims that these don’t exist. This is needed
for some Maxtor disks whose firmware does not indicate log support
even though the disk DOES support it.
[BA] Improved porting instructions and documentation in os_generic.c
[PW] Add Western Digital Caviar WD136AA and SAMSUNG SP40A2H (RR100-07
firmware) to knowndrives table.
[EM] FreeBSD: remove extra definition of FreeNonZero
[BA] smartctl: the -q silent option was printing output for some
error conditions. Fixed. Will rename relevant variables to help
avoid these errors in the future.
[SS] NetBSD port added.
[BA] more sensible error messages for devfs and devfs-like systems.
Instead of saying that the DIRECTORY does not exist, say that
the DEVICE does not exist.
[BA] smartd: added -n Directive, to prevent disk spin-up depending
upon the power mode (SLEEP, STANDBY, or IDLE).
[PW] Added Maxtor DiamondMax 20 VL series, Fujitsu MPF series,
Maxtor DiamondMax 36 series, Maxtor DiamondMax 4320 series, and
Maxtor DiamondMax 536DX series to knowndrives table.
[BA] many warning messages now give the file name AND VERSION
[BA] smartd: when the user provides multiple address recipients
to the ’-m’ Directive in a comma-delineated list, the commas
are stripped out before passing the list of addresses to the
mailer program. (Thanks to Calin A. Culianu for pointing this out
and providing a patch.)
[BA] smartd: when the ’-M exec path’ Directive is used, any stdout OR
stderr output from the executable "path" is assumed to indicate a
problem, and is echoed to SYSLOG.
[BA] Added all missing IBM/Hitachi Deskstar 180GXP models to knowndrives
[PW] Added some missing IBM/Hitachi Deskstar 120GXP models to knowndrives
[PW] Added IBM Travelstar 14GS to knowndrives table.
[PW] Modified knowndrives table to match entire Hitachi Travelstar
DK23BA and DK23EA series of drives (thanks to Norikatsu Shigemura
for submitting the patch).
[PW] Added some missing Fujitsu MPE series drives to knowndrives table.
FIREBALLlct15 20 to knowndrives table.
[EM] Fixup example command output for FreeBSD
[PW] Added Maxtor DiamondMax 80 family to knowndrives table.
[EM] Catch up FreeBSD code to switch PROJECTHOME to PACKAGE_HOMEPAGE
[BA] smartd: now watches stdout/stderr when trying to run mail, mailx
or mail warning script, and reports any output to SYSLOG. This
gives a clearer error message if something is wrong.
[BA] smartd: Solaris init script modified to accomodate grep that
lacks ’-q’ quiet option. Also check for running process to kill
on stop.
[PW] Added some missing Seagate Barracuda 7200.7 and 7200.7 Plus drives
to knowndrives table.
[PW] Added Maxtor DiamondMax Plus 60 family and Seagate U Series 5 20413
to knowndrives table.
[BA] smartd: under Solaris, made default mailer be ’mailx’ not
’mail’, since Solaris ’mail’ does not accept a ’-s’ argument.
A workaround for Solaris users of earlier versions is to
have ’-M exec /bin/mailx’ in their smartd.conf config file.
[DG] some SCSI controllers don’t like odd length transfers so make
sure LOG SENSE transfers are rounded up to an even number when
and odd length is reported (i.e. there is a double fetch, the
first to find the length, the second gets the data)
[BA] smartd man pages: under Solaris, correct section numbers in the
’See also’ section.
[KS/BA] smartd man page: describe how to set Solaris syslog.conf
file to catch all messages. Give correct Solaris SYSLOG default
path /var/adm/messages in man pages.
[BA] smartd: incorporated Debian startup script submitted by user.
[BA] smartctl: modified
firmware can leave
actually running.
case, not assuming
printing of self-test log entry number. Seagate
’holes’ in the self-test log while a test is
We now print entry numbers consistently in this
that entries are contiguous.
[PW] Added QUANTUM FIREBALL CX10.2A and Western Digital Caviar AC23200L
to knowndrives table.
[PW] Added QUANTUM FIREBALLlct20 20 to knowndrives table.
[PW] Added Maxtor DiamondMax Plus D740X family to knowndrives table.
[PW] Added IBM Travelstar 32GH, 30GT, and 20GN family to knowndrives
[BA] Slackware init script modified to search for /etc/slackware-version
rather than /etc/slackware-release.
[PW] Added Seagate Barracuda ATA II family and TOSHIBA MK4019GAXB to
knowndrives table.
[GG] explain howto use autoreconf in
[KS] changed manual page sections for
[BA] smartd: reduced number of scheduled self-test messages if
test already run in current hour.
[PW] Added Maxtor DiamondMax Plus 8 family to knowndrives table.
[BA] linux: check for linux/hdreg.h. If it’s there, use it. If
not, provide the necessary definitions ourselves.
[PW] Removed warning for IBM Deskstar 40GV & 75GXP series drives
with TXAOA5AA firmware
[PW] Added IBM Travelstar 25GS, 18GT, and 12GN family to knowndrives
[PW] Added IBM/Hitachi Travelstar 60GH & 40GN family to knowndrives
[BA] smartd: made ’-s’ Directive more efficient. Now store
compiled regex, and re-use. If device lacks certain self-test
capabilities, track it and don’t try again.
[BA] smartd: made memory allocation for device lists completely
dynamic (eliminating compile-time maximum length constants).
[PW] Removed warning for SAMSUNG SP0802N with TK100-23 firmware
[PW] Added Seagate Barracuda ATA IV family to knowndrives table.
[BA] smartd: reduce per-device memory footprint by making
mail-warning info dynamically allocated. Also remove
potential memory leak if use has -m Directive twice and
keeps reloading the config file (highly unlikely this would
ever be noticed!)
[DG] smartd: added SCSI scheduled self-tests (Background
short or extended).
[BA] smartd: can now run scheduled offline immediate and
self-tests. See man page and -s Directive for details.
[GG] don’t include manpages in make-dist-tarball.
[BA] smartctl: on-line examples given with -h are now correct
for solaris and linux, but wrong for freebsd. Ed?
[BA] smartd: man page now explains device scanning for solaris as
well as linux and freebsd.
[BA] smartd/smartctl: man pages now report correct CVS tag release
date, and executables ’-V’ options reports more build info.
smartmontools 5.26
[BA] Improved user messages that appear from ’make install’
[PW] Removed warning for SAMSUNG SP1213N with firmware TL100-23
[BA] incorporated SuSE init script from user.
[DG] if SCSI device is read only, then open it read only.
[BA] when compiled on non-supported system (NOT linux, freebsd or solaris) then
the run-time error messages now clearly say ’your system is not supported’
and give clear directions.
[BA] ./configure script now works correctly on SuSE linux boxes
[BA] minor improvements to man pages
[BA] simplified detection of packet (ATAPI, CD) devices.
[BA] init script (redhat, mandrake, yellowdog) now uses correct
strings for translation and is slightly more standard.
[DG] smartctl: output scsi Seagate vendor pages for disks (not tapes)
smartmontools 5.25
Note: there was no ’5.24’ release. From this point on, even numbered
releases will be ’stable’ ones and odd numbered releases will be
unstable/testing/development ones.
[DG] smartd/smartctl: changed scsiClearControlGLTSD() to
scsiSetControlGLTSD() with an ’enabled’ argument so ’-S on’
and ’-S off’ work for SCSI devices (if changing GLTSD supported).
[BA] smartd/smartctl: wired in scsiClearControlGLTSD(). Could still
use a corresponding Set function. Left stubs for this purpose.
[DG] scsicmds: added scsiClearControlGLTSD() [still to be wired in]
[BA] smartctl: make SCSI -T options behave the same way as the
ATA ones.
[DG] smartctl: output scsi transport protocol if available
[DG] scsi: stop device scan in smartd and smartctl if badly formed
mode response [heuristic to filter out USB devices before we
(potentially) lock them up].
[BA] smartd: deviceclose()->CloseDevice(). Got rid of SCSIDEVELOPMENT
macro-enabled code. Added -W to list of gcc specific options to
always enable. Made code clean for -W warnings.
[PW] Added Maxtor DiamondMax VL 30 family to knowndrives table.
[DG] scsi: add warning (when ’-l error’ active) if Control mode page
GLTSD bit is set (global disable of saving log counters)
[DG] scsi: remember mode sense cmd length. Output trip temperature
from IE lpage (IBM extension) when unavailable from temp lpage.
[BA] smartd: for both SCSI and ATA now warns user if either
the number of self-test errors OR timestamp of most
recent self-test error have increased.
[DG] smartctl: output Seagate scsi Cache and Factory log pages (if
available) when vendor attributes chosen
[DG] smartd: add scsiCountFailedSelfTests() function.
[DG] Do more sanity checking of scsi log page responses.
[BA] smartd: now warns user if number of self-test errors has
increased for SCSI devices.
[BA] smartd: warn user if number of ATA self-test errors increases
(as before) OR if hour time stamp of most recent self-test
error changes.
[DG] More checks for well formed mode page responses. This has the side
effect of stopping scans on bad SCSI implementations (e.g. some
USB disks) prior to sending commands (typically log sense) that
locks them up.
[PW] Added Western Digital Caviar family and Caviar SE family to
knowndrives table.
[BA] smartd: added -l daemon (which is the default value if -l
is not used).
[PW] Added Seagate Barracuda ATA V family to knowndrives table.
[BA] smartd: added additional command line argument -l FACILITY
or --logfacility FACILITY. This can be used to redirect
messages from smartd to a different file than the one used
by other system daemons.
[PW] Added Seagate Barracuda 7200.7, Western Digital Protege WD400EB,
and Western Digital Caviar AC38400 to knowndrives table.
[BA] smartd: scanning should now also work correctly for
devfs WITHOUT traditional links /dev/hd[a-t] or /dev/sd[a-z].
[PW] Added Maxtor 4W040H3, Seagate Barracuda 7200.7 Plus,
IBM Deskstar 120GXP (40GB), Seagate U Series 20410,
Fujitsu MHM2100AT, MHL2300AT, MHM2150AT, and IBM-DARA-212000
to knowndrives table.
[PW] Added remaining Maxtor DiamondMax Plus 9 models to knowndrives
[EM] smartd: If no matches found, then return 0, rather than an error
indication, as it just means no devices of the given type exist.
Adjust FreeBSD scan code to mirror Linux version.
[BA] smartd: made device scan code simpler and more robust. If
too many devices detected, warn user but scan as many
as possible. If error in scanning, warn user but don’t
die right away.
[EM] smartd: To keep as consistent as possible, migrate FreeBSD
devicescan code to also use glob(3). Also verified clean
compile on a 4.7 FreeBSD system.
[BA] smartd: Modified device scan code to use glob(3). Previously
it appeared to have trouble when scanning devices on an XFS
file system, and used non-public interface to directory
entries. Problems were also reported when /dev/ was on an
ext2/3 file system, but there was a JFS partition on the same
[BA] Clearer error messages when device scanning finds no suitable
[EM] FreeBSD: Fixup code to allow for proper compilation under
-STABLE branch.
smartmontools 5.23
[BA] smartd: didn’t close file descriptors of ATA packet devices
that are scanned. Fixed.
[BA] Added reload/report targets to the smartmontools init script.
reload: reloads config file
report: send SIGUSR1 to check devices now
smartmontools 5.22
[EM] Fix compile issues for FreeBSD < 5-CURRENT.
[PW] Added Fujitsu MHM2200AT to knowndrives table.
[BA] To help catch bugs, clear ATA error structures before all
ioctl calls. Disable code that attempted to time-out on SCSI
devices when they hung (doesn’t work).
[BA] Documented STATUS/ERROR flags added by [PW] below.
[BA] Improved algorithm to recognize ATA packet devices. Should
no longer generate SYSLOG kernel noise when user tries either
smartd or smartctl on packet device (CD-ROM or DVD). Clearer
warning messages from smartd when scanning ATA packet device.
[PW] Added TOSHIBA MK4025GAS to knowndrives table.
[PW] Added a textual interpretation of the status and error registers
in the SMART error log (ATA). The interpretation is
command-dependent and currently only eight commands are supported
(those which produced errors in the error logs that I happen to
have seen).
[BA] added memory allocation tracking to solaris code.
Fixed solaris signal handling (reset handler to default
after first call to handler) by using sigset. Added
[CD] solaris port: added SCSI functionality to solaris
[BA] smartd: attempt to address bug report about smartd
hanging on USB devices when scanning:
Set a timeout of SCSITIMEOUT (nominally 7 seconds) before
giving up.
[EM] smartd: DEVICESCAN will follow links in a devfs filesystem and
make sure the end point is a disc. Update documentation, added
note about FreeBSD scanning
[BA] smartd: DEVICESCAN also looks for block devices in
/dev. Updated documentation. Now scans for up to
20 ATA devices /dev/hda-t rather than previous 12
[EM] smartd: mirror the FreeBSD DEVICESCAN logic for Linux,
so that smartd now scans only devices found in /dev/. Also,
make utility memory functions take a line number and file so
that we report errors with the correct location.
[GG] add a note about Debian bug #208964 to WARNINGS.
[BA] smartctl: -T verypermissive option broken. Use
-T verpermissive until the next release, please.
[BA] Syntax mods so that code also compiles on Solaris using
Sun Workshop compiler. Need -xmemalign 1i -xCC flags
for cc.
smartmontools 5.21
[DK] Changed so -Wall is only included if gcc
is used (this is a gcc specific flag) and -fsignedchar
is not used at all (this is a gcc specific compiler
[BA] Modifications so that code now compiles under solaris. Now
all that’s needed (:-) is to fill in os_solaris.[hc]. Added
os_generic.[hc] as guide to future ports. Fixed -D option
of smartd (no file name). Modified -h opt of smartd/smartctl
to work properly with solaris getopt().
[EM] Update MAN pages with notes that 3ware drives are NOT supported
under FreeBSD. Cleanup FreeBSD warning message handling.
[EM] FreeBSD only: Fix first user found bug....I guess I was making
the wrong assumption on how to convert ATA devnames to
channel/unit numbers.
[EM] Allow for option --enable-sample to append ’.sample’ to installed
smartd.conf and rc script files. Also, let rc script shell setting
be determined by configure
[EM] Minor autoconf update to include -lcam for FreeBSD
[EM] Add conditional logic to allow FreeBSD to compile pre-ATAng.
-- note, not tested
Add some documentation to INSTALL for FreeBSD.
[EM] Implement SCSI CAM support for FreeBSD. NOTE: I am not an expert
in the use of CAM. It seems to work for me, but I may be doing
something horribly wrong, so please exercise caution.
[EM] Switch over to using ’atexit’ rather than ’on_exit’ routine. This also
meant we needed to save the exit status elsewhere so our ’Goodbye’
routine could examine it.
[EM] Move the DEVICESCAN code to os specific files. Also moved some of the
smartd Memory functions to utility.c to make available to smartctl.
[EM] Code janitor work on os_freebsd.c.
[EM] Added os_freebsd.[hc] code.
Additional code janitor
[BA] Code janitor working, moving OS dependent code into
[GG] conditionally compile os_{freebsd,linux}.o depending on
host architecture
[PW] Print estimated completion time for tests
[BA] Added -F samsung2 flag to correct firmware byte swap.
All samsung drives with *-23 firmware revision string.
smartmontools 5.20
[GG] Fixed broken (zero length smartd.conf.5
was being created), fix broken uninstall/distcheck targets
[FM] Improved Slackware init script added to /etc/smartd.initd
[BA] smartctl: added ’-T verypermissive’ option which is
equivalent to giving ’-T permissive’ many times.
[BA] Try harder to identify from IDENTIFY DEVICE structure
if SMART supported/enabled. smartd now does a more
thorough job of trying to assess this before sending
a SMART status command to find out for sure.
[BA] smartctl: it’s now possible to override the program’s
guess of the device type (ATA or SCSI) with -d option.
[BA] try hard to avoid sending IDENTIFY DEVICE to packet
devices (CDROMS). They can’t do SMART, and this generates
annoying syslog messages. At the same time, identify type
of Packet device.
[BA] smartctl: Can now use permissive option more
than once, to control how far to go before giving up.
[BA] smartd: if user asked to monitor either error or self-test
logs (-l error or -l selftest) WITHOUT monitoring any of the
Attribute values, code will SEGV. For 5.1-18 and earlier,
a good workaround is to enable Auto offline (-o on).
[BA] smartctl: If enable auto offline command given, update auto
offline status before printing capabilities.
[GG] Make autotools build the default, remove autotools.diff
[GG] Add auto{conf,make} support, not enabled by default.
[BA] Eliminated #include <linux/hdreg.h> from code. This
should simplify porting to solaris, FreeBSD, etc. The
only linux-specific code is now isolated to three routines,
one for SCSI, one for Escalade, one for ATA.
smartmontools 5.1-18
[BA] smartd: fixed serious bug - Attributes not monitored unless
user told smartd to ignore at least one of them!
smartmontools 5.1-17
[BA] Default runlevels for smartd changed from 3 and 5 to
2, 3, 4, and 5.
[BA] Removed as much dynamic memory allocation as possible from
configuration file parsing. Reloading config file, even in
presence of syntax errors etc. should not cause memory leaks.
[PW] It is no longer permissible for the integer part (if any) of
arguments to --report and --device to be followed by non-digits.
For example, the "foo" in --report=ioctl,2foo was previously
ignored, but now causes an error.
[BA] smartd: added -q/--quit command line option to specify
under what circumstances smartd should exit. The old
-c/--checkonce option is now obsoleted by this more
general-purpose option.
[BA] smartd now responds to a HUP signal by re-reading its
configuration file /etc/smartd.conf. If there are
errors in this file, then the configuration file is
ignored and smartd continues to monitor the devices that
it was monitoring prior to receiving the HUP signal.
[BA] Now correctly get SMART status from disks behind 3ware
controllers, thanks to Adam Radford. Need 3w-xxxx driver
version or later. Previously the smartmontools
SMART status always returned "OK" for 3ware controllers.
[BA] Additional work on dynamic memory allocation/deallocation.
This should have no effect on smartctl, but clears that way
for smartd to dynamically add and remove entries. It should
also now be easier to modify smartd to re-read its config
file on HUP (which is easy) without leaking memory (which is
harder). The philosophy is that memory for data structures in
smartd is now allocated only on demand, the first time it
is needed.
[BA] smartd: finished cleanup. Now use create/rm functions for
cfgentries and dynamic memory allocation almost everywhere.
Philosophy: aggresively try and provoke SEGV to help find
bad code.
[BA] Added SAMSUNG SV0412H to knowndrives table.
[BA] smartd: if DEVICESCAN used then knowndrives table might not set
the -v attributes correctly -- may have been the same for all
the drives. Cleaned up some data structures and memory
allocation to try and ensure segvs if such problems are
introduced again.
[BA] Now allow -S on and -o on for the 3ware device type. For these
commands to be passed through, the stock 3ware 3w-xxxx driver
must be patched (8 lines). I’ll post a patch on the smartmontools
home page after it’s been tested by a few other people and 3ware
have had a chance to look it over.
[BA] smartd - can now monitor ATA drives behind 3ware controllers.
[BA] smartd - changed some FATAL out of memory error messages from
syslog level LOG_INFO to LOG_CRIT.
[BA] smartctl - added code to look at ATA drives behind 3ware RAID
controllers using the 3w-xxxx driver. Note that for technical
reasons related to the 3w-xxxx driver, the "Enable Autosave",
"Enable Automatic Offline" commands are not implemented.
I will add this to smartd shortly.
[BA] smartd - modified sleep loop, so that smartd no longer comes
on the run queue every second. Instead, unless interrupted,
it sleeps until the next polling time, when it wakes up. Now
smartd also tries to wake up at exactly the right
intervals (nominally 30 min) even if the user has been sending
signals to it.
[GG] add Fujitsu MHN2300AT to vendoropts_9_seconds.
[EB] Fujitsu change in knowndrives ... match the whole MPD and
MPE series for vendoropts_9_seconds.
[BA] smartd bug, might cause segv if a device can not be opened. Was
due to missing comma in char* list. Consequence is that email
failure messages might have had the wrong Subject: heading for
errorcount, FAILEDhealthcheck, FAILEDreadsmartdata, FAILEDreadsmarterrorlog,
FAILEDreadsmartsefltestlog, FAILEDopendevice were all displaced by
one. And FAILEDopendevice might have caused a segv if -m was being
used as a smartd Directive.
[BA] Cleaned up smartmontools.spec so that upgrading, removing
and other such operations correctly preserve running behavior
and booting behavior of smartd.
[BA] Improved formatting of ATA Error Log printout, and added
listing of names of commands that caused the error. Added
obsolete ATA-4 SMART feature commands to table, along with
obsolete SFF-8035i SMART feature command.
[PW] Added atacmdnames.[hc], which turn command register &
feature register pairs into ATA command names.
[BA] Added conveyance self-test. Some code added for selective
self-tests, but #ifdefed out.
[BA] Modified smartd exit status and log levels. If smartd is
"cleanly" terminated, for example with SIGTERM, then its
exit messages are now logged at LOG_INFO not LOG_CRIT
[BA] Added Attribute IDs (Fujitsu) 0xCA - 0xCE. This is decimal
202-206. Added -v switches for interpretation of Attributes
192, 198 and 201.
[BA] Made smartmontools work with any endian order machine for:
- SMART selftest log
- SMART ATA error log
- SMART Attributes values
- SMART Attributes thesholds
- IDENTIFY DEVICE information
Smartmontools is now free of endian bias and works correctly
on both little- and big-endian hardware. This has been tested by
three independent PPC users on a variety of ATA and SCSI hardware.
[DG] Check that certain SCSI command responses are well formed. If
IEC mode page response is not well formed exit smartctl. This
is to protect aacraid. smartd should ignore a aacraid device.
[BA] smartctl: added column to -A output to show if Attributes are
updated only during off-line testing or also during normal
[BA] smartd: attempt to enable/disable automatic offline testing even
if the disk appears not to support it. Now the same logic
as smartctl.
[BA] Added definition of Attribute 201, soft read error rate.
[BA] Added IBM/Hitachi IC35L120AVV207-1 (GXP-180) and corresponding
8MB Cache GXP-120 to drive database.
[BA] smartd: if DEVICESCAN Directive used in smartd.conf, and
-I, -R or -r Directives used in conjunction with this, got
segv errors. Fixed by correcting memory allocation calls.
[BA] smartd: enable automatic offline testing was broken due
to cut-and-paste error that disabled it instead of
enabling it. Thanks to Maciej W. Rozycki for pointing
out the problem and solution.
[BA] Fixed "spelling" of some Attribute names to replace spaces
in names by underscores. (Fixed field width easier for awk
style parsing.)
[BA,GF] Added mods submitted by [GF] to support Attribute 193 being
load/unload cycles. Add -v 193,loadunload option, useful for
Hitachi drive DK23EA-30, and add this drive to knowndrive.c
Add meaning of attribute 250 : Read error retry rate
[BA] Added another entry for Samsung drives to knowndrive table.
[DG] Refine SCSI log sense command to do a double fetch in most cases
(but not for the TapeAlert log page). Fix TapeAlert and Self Test
log page response truncation.
[PW] Added ’removable’ argument to -d Directive for smartd. This indicates
that smartd should continue (rather than exit) if the device does not
appear to be present.
[BA] Modified smartmontools.spec [Man pages location] and
smartd.initd [Extra space kills chkconfig!] for Redhat 6.x
compatibility (thanks to Gerald Schnabel).
[EB] Add another Fujitsu disk to knowndrives.c
[GG] match for scsi/ and ide/ in case of devfs to exclude false postives
[BA] If SCSI device listed in /etc/smartd.conf fails to open or do
SMART stuff correctly, or not enough space
to list all SCSI devices, fail with error unless
-DSCSIDEVELOPMENT set during compile-time.
[BA] Added automatic recognition of /dev/i* (example: /dev/ide/...)
as an ATA device.
[DG] Add "Device type: [disk | tape | medium changer | ...] line to
smartctl -i output for SCSI devices.
[PW] Fixed bug in smartd where test email would be sent regularly (for
example, daily if the user had specified -M daily) instead of just
once on startup.
[KM] More TapeAlert work. Added translations for media changer
alerts. TapeAlert support reported according to the log page
presence. ModeSense not attempted for non-ready tapes (all
drives do not support this after all). Get peripheral type from
Inquiry even if drive info is not printed. Add QUIETON()
QUIETOFF() to TapeAlert log check.
[BA] Stupid bug in atacmds.c minor_str[] affected ataVersionInfo().
Two missing commas meant that minor_str[] had two few elements,
leading to output like this:
Device Model:
Maxtor 6Y120L0
Serial Number:
Firmware Version: YAR41VW0
Device is:
Not in smartctl database [for details use: -P showall]
ATA Version is:
ATA Standard is: 9,minutes
Missing commas inserted.
[BA] Fixed smartd bug. On device registration, if ATA device did
not support SMART error or self-test logs but user had asked to
monitor them, an attempt would be made to read them anyway,
possibly generating "Drive Seek" errors. We now check that
the self-test and error logs are supported before trying to
access them the first time.
[GG/BA] Fixed bug where if SMART ATA error log not supported,
command was tried anyway. Changed some error printing to use
print handlers.
[GG] Makefile modifications to ease packaging
[DG] Did work for TapeAlerts (SCSI). Now can detect /dev/nst0 as a
SCSI device. Also open SCSI devices O_NONBLOCK so they don’t
hang on open awaiting media. The ATA side should worry about
this also: during a DEVICESCAN a cd/dvd device without media
will hang. Added some TapeAlert code suggested by Kai Makisara.
[PW] Extended the -F option/Directive to potentially fix other firmware
bugs in addition to the Samsung byte-order bug. Long option name is
now --firmwarebug and the option/Directive accepts an argument
indicating the type of firmware bug to fix.
[BA] Fixed a bug that prevented the enable automatic off-line
test feature from enabling. It also prevented the enable Attribute
autosave from working. See CVS entry for additional details.
[PW] Modified the -r/--report option (smartctl and smartd) to allow the
user to specify the debug level as a positive integer.
[BA] Added --log directory option to smartctl. If the disk
supports the general-purpose logging feature set (ATA-6/7)
then this option enables the Log Directory to be printed.
This Log Directory shows which device logs are available, and
their lengths in sectors.
[PW] Added -P/--presets option to smartctl and -P Directive to smartd.
[GG] Introduce different exit codes indicating the type of problem
encountered for smartd.
[DG] Add non-medium error count to ’-l error’ and extended self test
duration to ’-l selftest’. Get scsi IEs and temperature changes
working in smartd. Step over various scsi disk problems rather
than abort smartd startup.
[DG] Support -l error for SCSI disks (and tapes). Output error counter
log pages.
[BA] Added -F/--fixbyteorder option to smartctl. This allows us to read
SMART data from some disks that have byte-reversed two- and fourbyte quantities in their SMART data structures.
[BA] Fixed serious bug: the -v options in smartd.conf were all put
together and used together, not drive-by-drive.
[PW] Added knowndrives.h and knowndrives.c.
supersedes the drivewarnings array.
The knowndrives array
[GG] add {-p,--pidfile} option to smartd to write a PID file on
startup. Update the manpage accordingly.
[DG] Fix scsi smartd problem detecting SMART support. More cleaning
and fix (and rename) scsiTestUnitReady(). More scsi renaming.
[BA] Fixed smartd so that if a disk that is explictily listed is not
found, then smartd will exit with nonzero status BEFORE forking.
If a disk can’t be registered, this will also be detected before
forking, so that init scripts can react correctly.
[BA] Replaced all linux-specific ioctl() calls in atacmds.c with
a generic handler smartcommandhandler(). Now the only routine
that needs to be implemented for a given OS is os_specific_handler().
Also implemented the --report ataioctl. This provides
two levels of reporting. Using the option once gives a summary
report of device IOCTL transactions. Using the option twice give
additional info (a printout of ALL device raw 512 byte SMART
data structures). This is useful for debugging.
[DG] more scsi cleanup. Output scsi device serial number (VPD page
0x80) if available as part of ’-i’. Implement ’-t offline’ as
default self test (only self test older disks support).
[BA] Changed crit to info in loglevel of smartd complaint to syslog
if DEVICESCAN enabled and device not found.
[BA] Added -v 194,10xCelsius option/Directive. Raw Attribute number
194 is ten times the disk temperature in Celsius.
[DG] scsicmds.[hc] + scsiprint.c: clean up indentation, remove tabs.
Introduce new intermediate interface based on "struct scsi_cmnd_io"
to isolate SCSI generic commands + responses from Linux details;
should help port to FreeBSD of SCSI part of smartmontools.
Make SCSI command builders more parametric.
[BA] smartctl: if HDIO_DRIVE_TASK ioctl() is not implemented (no
kernel support, then try to assess drive health by examining
Attribute values/thresholds directly.
[BA] smartd/smartctl: added -v 200,writeerrorcount option/Directive
for Fujitsu disks.
[BA] smartd: Now send email if any of the SMART commands fails,
or if open()ing the device fails. This is often noted
as a common disk failure mode.
[BA] smartd/smartctl: Added -v N,raw8 -v N,raw16 and -v N,raw48
Directives/Options for printing Raw Attributes in different
[BA] smartd: Added -r ID and -R ID for reporting/tracking Raw
values of Attributes.
[BA] smartd/smartctl: Changed printing of spin-up-time attribute
raw value to reflect current/average as per IBM standard.
[BA] smartd/smartctl: Added -v 9,seconds option for disks which
use Attribute 9 for power-on lifetime in seconds.
[BA] smartctl: Added a warning message so that users of some IBM
disks are warned to update their firmware. Note: we may want
to add a command-line flag to disable the warning messages.
I have done this in a general way, using regexp, so that we
can add warnings about any type of disk that we wish...
[BA] smartd: Created a subdirectory examplescripts/ of source
directory that contains executable scripts for the -M exec PATH
Directive of smartd.
[BA] smartd: DEVICESCAN in /etc/smartd.conf
can now be followed by all the same Directives as a regular
device name like /dev/hda takes. This allows one to use
(for example):
in the /etc/smartd.conf file.
[BA] smartd: Added -c (--checkonce) command-line option. This checks
all devices once, then exits. The exit status can be
used to learn if devices were detected, and if smartd is
functioning correctly. This is primarily for Distribution
[BA] smartd: Implemented -M exec Directive for
smartd.conf. This makes it possible to run an
arbitrary script or mailing program with the
-m option.
[PW] smartd: Modified -M Directive so that it can be given
multiple times. Added -M exec Directive.
[BA] Fixed bug in smartctl pointed out by Pierre Gentile.
-d scsi didn’t work because tryata and tryscsi were
reversed -- now works on /devfs SCSI devices.
[BA] Fixed bug in smartctl pointed out by Gregory Goddard
<>. Manual says that bit 6 of return
value turned on if errors found in smart error log. But
this wasn’t implemented.
[BA] Modified printing format for 9,minutes to read
Xh+Ym not X h + Y m, so that fields are fixed width.
[BA] Added Attribute 240 "head flying hours"
[BA] As requested, local time/date now printed by smartctl -i
[PW] Added "help" argument to -v for smartctl
[PW] Added -D, --showdirectives option to smartd
[DG] add ’-l selftest’ capability for SCSI devices (update smartctl.8)
[BA] smartd,smartctl: added additional Attribute modification option
-v 220,temp and -v 9,temp.
[PW] Renamed smartd option -X to -d
[PW] Changed smartd.conf Directives -- see man page
[BA/DG] Fixed uncommented comment in smartd.conf
[DG] Correct ’Recommended start stop count’ for SCSI devices
[PW] Replaced smartd.conf directive -C with smartd option -i
[PW] Changed options for smartctl -- see man page.
[BA] Use strerror() to generate system call error messages.
[BA] smartd: fflush() all open streams before fork().
[BA] smartctl, smartd simplified internal handling of checksums
for simpler porting and less code.
[PW] smartd --debugmode changed to --debug
[BA] smartd/smartctl added attribute 230 Head Amplitude from
IBM DPTA-353750.
[PW] Added list of proposed new options for smartctl to README.
[PW] smartd: ParseOpts() now uses getopt_long() if HAVE_GETOPT_LONG is
defined and uses getopt() otherwise. This is controlled by CPPFLAGS in
the Makefile.
[BA] smartd: Fixed a couple of error messages done with perror()
to redirect them as needed.
[BA] smartctl: The -O option to enable an Immediate off-line test
did not print out the correct time that the test would take to
complete. This is because the test timer is volatile and not
fixed. This has been fixed, and the smartctl.8 man page has been
updated to explain how to track the Immediate offline test as it
progresses, and to further emphasize the differences between the
off-line immediate test and the self-tests.
[BA] smartd/smartctl: Added new attribute (200) Multi_Zone_Error_Rate
[BA] smartctl: modified so that arguments could have either a single as in -ea or multiple ones as in -e -a. Improved warning message for
device not opened, and fixed error in redirection of error output of
HD identity command.
[PW] smartd: added support for long options. All short options are still
supported; see manpage for available long options.
[BA] smartctl. When raw Attribute value was 2^31 or larger, did
not print correctly.
[BA] smartd: added smartd.conf Directives -T and -s. The -T Directive
enables/disables Automatic Offline Testing. The -s Directive
enables/disables Attribute Autosave. Documentation and
example configuration file updated to agree.
[BA] smartd: user can make smartd check the disks at any time
(ie, interrupt sleep) by sending signal SIGUSR1 to smartd. This
can be done for example with:
kill -USR1 <pid>
where <pid> is the process ID number of smartd.
[EB] scsi: don’t trust the data we receive from the drive too
much. It very well might have errors (like zero response length).
Seen on Megaraid logical drive, and verified in the driver source.
[BA] smartd: added Directive -m for sending test email and
for modifying email reminder behavior. Updated manual, and sample
configuration file to illustrate & explain this.
[BA] smartd: increased size of a continued smartd.conf line to
1023 characters.
[BA] Simplified Directive parsers and improved warning/error
[EB] Fixed bug in smartd where testunitready logic inverted
prevented functioning on scsi devices.
The bug in question only affects smartd users with scsi devices.
To see if your version of smartd has the testunitready() bug, do
smartd -V
If the version of the module smartd.c in a line like:
Module: smartd.c
revision: 1.66
date: 2002/11/17
has a revision greater than or equal to 1.30, and less than or equal to
1.64, then your version of the code has this problem.
This problem affected releases starting with RELEASE_5_0_16 up to and
including RELEASE_5_0_43.
[BA] Added testunitnotready to smartctl for symmetry with smartd.
[SB] added Czech descriptions to .spec file
[SB] corrected comment in smartd.conf example
[BA] Changed way that entries in the ATA error log are printed,
to make it clearer which is the most recent error and
which is the oldest one.
NOTE: All changes made prior to this point were done by Bruce Allen
[BA] although several of them had been suggested by earlier postings
by Stanislav Brabec [SB].
Changed Temperature_Centigrade to Temperature_Celsius.
The term "Centigrade" ceased to exist in 1948. (c.f
Modified SCSI device check to also send warning emails if
requested in directives file.
Added a new smartd configuration file Directive: -M ADDRESS.
This sends a single warning email to ADDRESS for failures or
errors detected with the -c, -L, -l, or -f Directives.
Modified perror() statements in atacmds.c so that printout for SMART
commands errors is properly suppressed or queued depending upon users
choices for error reporting modes.
Added Italian descriptions to smartmontools.spec file.
Started impementing send-mail-on-error for smartd; not yet enabled.
Added -P (Permissive) Directive to smartd.conf file to allow SMART
monitoring of pre-ATA-3 Rev 4 disks that have SMART but do not have
a SMART capability bit.
Removed charset encodings from smartmontools.spec file for non-English
Added manual page smartd.conf.5 for configuration file.
smartctl: Missing ANSI prototype in failuretest(); fixed.
smartctl: Checksum warnings now printed on stdout, or are silent,
depending upon -q and -Q settings.
Changed Makefile so that the -V option does not reflect file state
before commit!
smartctl: added new options -W, -U, and -P to control if and how the
smartctl exits if an error is detected in either a SMART data
structure checksum, or a SMART command returns an error.
modified manual page to break options into slightly more logical
reformatted ’usage’ message order to agree with man page ordering
modified .spec file so that locale information now contains
character set definition.
Changed pt_BR to pt since we do not use any
aspect other than language. See man setlocale.
smartctl: added new options -n and -N to force device to be ATA or SCSI
smartctl: no longer dies silently if device path does not start/dev/X
smartctl: now handles arbitrary device paths
Modified .spec file and Makefile to make them more compliant with
the "right" way of doing things.
Fixed typesetting error in man page smartd.8
Removed redundant variable (harmless) from smartd.c
Added a new directive for the configuration file. If the word
DEVICESCAN appears before any non-commented material in the
configuration file, then the confi file will be ignored and the
devices wil be scanned.
Note: it has now been confirmed that the code modifications between
5.0.23 and 5.0.24 have eliminated the GCC 3.2 problems. Note that
there is a GCC bug howerver, see #8404 at
Added new Directive for Configuration file:
-C <N> This sets the time in between disk checks to be <N>
seconds apart. Note that although you can give
this Directive multiple times on different lines of
the configuration file, only the final value that
is given has an effect, and applies to all the
disks. The default value of <N> is 1800 sec, and
the minimum allowed value is ten seconds.
Problem wasn’t the print format. F.L.W. Meunier <>
sent me a gcc 3.2 build and I ran it under a debugger. The
problem seems to be with passing the very large (2x512+4) byte
data structures as arguments. I never liked this anyway; it was
inherited from smartsuite. So I’ve changed all the heavyweight
functions (ATA ones, anyone) to just passing pointers, not hideous
kB size structures on the stack. Hopefully this will now build OK
under gcc 3.2 with any sensible compilation options.
Because of reported problems with GCC 3.2 compile, I have gone
thorough the code and explicitly changed all print format
parameters to correspond EXACTLY to int unless they have to be
promoted to long longs. To quote from the glibc bible: [From
GLIBC Manual: Since the prototype doesn’t specify types for
optional arguments, in a call to a variadic function the default
argument promotions are performed on the optional argument
values. This means the objects of type char or short int (whether
signed or not) are promoted to either int or unsigned int, as
smartd, smartctl now warn if they find an attribute whose ID
number does not match between Data and Threshold structures.
Fixed nasty bug which led to wrong number of arguments for a
varargs statement, with attendent stack corruption. Sheesh!
Have added script to CVS attic to help find such nasties in the
Eliminated some global variables out of header files and other
minor cleanup of smartd.
Did some revision of the man page for smartd and made the usage
messages for Directives 100% consistent.
smartd: prints warning message when it gets SIGHUP, saying that it
is NOT re-reading the config file.
smartctl: updated man page to say self-test commands -O,x,X,s,S,A
appear to be supported in the code. [I can’t test these, can anyone
smartctl: smartctl would previously print the LBA of a self-test
if it completed, and the LBA was not 0 or 0xff...f However
according to the specs this is not correct. According to the
specs, if the self-test completed without error then LBA is
undefined. This version fixes that. LBA value only printed if
self-test encountered an error.
smartd has changed significantly. This is the first CVS checkin of
code that extends the options available for smartd. The following
options can be placed into the /etc/smartd.conf file, and control the
behavior of smartd.
Configuration file Directives (following device name):
Device is an ATA device
Device is a SCSI device
Monitor SMART Health Status
Monitor SMART Error Log for changes
Monitor SMART Self-Test Log for new errors
Monitor for failure of any ’Usage’ Attributes
Report changes in ’Prefailure’ Attributes
Report changes in ’Usage’ Attributes
Equivalent to -p and -u Directives
Equivalent to -c -l -L -f -t Directives
-i ID Ignore Attribute ID for -f Directive
-I ID Ignore Attribute ID for -p, -u or -t Directive
Comment: text after a hash sign is ignored
Line continuation character
cleaned up functions used for printing CVS IDs.
library, as it should be.
Now use string
modified length of device name string in smartd internal structure
to accomodate max length device name strings
removed un-implemented (-e = Email notification) option from
command line arg list. We’ll put it back on when implemeneted.
smartd now logs serious (fatal) conditions in its operation at
loglevel LOG_CRIT rather than LOG_INFO before exiting with error.
smartd used to open a file descriptor for each SMART enabled
device, and then keep it open the entire time smartd was running.
This meant that some commands, like IOREADBLKPART did not work,
since the fd to the device was open. smartd now opens the device
when it needs to read values, then closes it. Also, if one time
around it can’t open the device, it simply prints a warning
message but does not give up. Have eliminated the .fd field from
data structures -- no longer gets used.
smartd now opens SCSI devices as well using O_RDONLY rather than
O_RDWR. If someone can no longer monitor a SCSI device that used
to be readable, this may well be the reason why.
smartd never checked if the number of ata or scsi devices detected
was greater than the max number it could monitor. Now it does.
smartd on startup now looks in the configuration file /etc/smartd.conf for
a list of devices which to include in its monitoring list. See man page
(man smartd) for syntax.
smartd: close file descriptors of SCSI device if not SMART capable
Closes ALL file descriptors after forking to daemon.
added new temperature attribute (231, temperature)
smartd: now open ATA disks using O_RDONLY
smartd now prints the name of a failed or changed attribute
into logfile, not just ID number
Changed name of -p (print version) option to -V
Minor change in philosophy: if a SMART command fails or the device
appears incapable of a SMART command that the user has asked for,
complain by printing an error message, but go ahead and try
anyway. Since unimplemented SMART commands should just return an
error but not cause disk problems, this should’t cause any
Added two new flags: q and Q. q is quiet mode - only print: For
the -l option, errors recorded in the SMART error log; For the -L
option, errors recorded in the device self-test log; For the -c
SMART "disk failing" status or device attributes (pre-failure or
usage) which failed either now or in the past; For the -v option
device attributes (pre-failure or usage) which failed either now
or in the past. Q is Very Quiet mode: Print no ouput. The only
way to learn about what was found is to use the exit status of
smartctl now returns sensible values (bitmask). See smartctl.h
for the values, and the man page for documentation.
The SMART status check now uses the correct ATA call. If failure
is detected we search through attributes to list the failed ones.
If the SMART status check shows GOOD, we then look to see if their
are any usage attributes or prefail attributes have failed at any
time. If so we print them.
Modified function that prints vendor attributes to say if the
attribute has currently failed or has ever failed.
-p option now prints out license info and CVS strings for all
modules in the code, nicely formatted.
Previous versions of this code (and Smartsuite) only generate
SMART failure errors if the value of an attribute is below the
threshold and the prefailure bit is set. However the ATA Spec
(ATA4 <=Rev 4) says that it is a SMART failure if the value of an
attribute is LESS THAN OR EQUAL to the threshold and the
prefailure bit is set. This is now fixed in both smartctl and
smartd. Note that this is a troubled subject -- the original
SFF 8035i specification defining SMART was inconsistent about
this. One section says that Attribute==Threshold is pass,
and another section says it is fail. However the ATA specs are
consistent and say Attribute==Threshold is a fail.
smartd did not print the correct value of any failing SMART attribute.
printed the index in the attribute table, not the attribute
ID. This is fixed.
when starting self-tests in captive mode ioctl returns EIO because
the drive has been busied out. Detect this and don’t return an eror
in this case. Check this this is correct (or how to fix it?)
fixed possible error in how to determine ATA standard support
for devices with no ATA minor revision number.
device opened only in read-only not read-write mode.
access to get smart data. Check this with Andre.
Don’t need R/W
smartctl now handles all possible choices of "multiple options"
gracefully. It goes through the following phases of operation,
Documentation has bee updated to explain the different phases of
operation. Control flow through ataPrintMain()
If reading device identity information fails, try seeing if the info
can be accessed using a "DEVICE PACKET" command. This way we can
at least get device info.
Modified Makefile to automatically tag CVS archive on issuance of
a release
Modified drive detection so minor device ID code showing ATA-3 rev
0 (no SMART) is known to not be SMART capable.
Now verify the checksum of the device ID data structure, and of the
attributes threshold structure. Before neither of these
structures had their checksums verified.
New behavior vis-a-vis checksums. If they are wrong, we log
warning messages to stdout, stderr, and syslog, but carry on
anyway. All functions now call a checksumwarning routine if the
checksum doesn’t vanish as it should.
Changed Read Hard Disk Identity function to get fresh info from
the disk on each call rather than to use the values that were read
upon boot-up into the BIOS. This is the biggest change in this
release. The ioctl(device, HDIO_GET_IDENTITY, buf ) call should
be avoided in such code. Note that if people get garbled strings
for the model, serial no and firmware versions of their drives,
then blame goes here (the BIOS does the byte swapping for you,
Function ataSmartSupport now looks at correct bits in drive
identity structure to verify first that these bits are valid,
before using them.
Function ataIsSmartEnabled() written which uses the Drive ID state
information to tell if SMART is enabled or not. We’ll carry this
along for the moment without using it.
Function ataDoesSmartWork() guaranteed to work if the device
supports SMART.
Replace some numbers by #define MACROS
Wrote Function TestTime to return test time associated with each
different type of test.
Thinking of the future, have added a new function called
ataSmartStatus2(). Eventually when I understand how to use the
TASKFILE API and am sure that this works correctly, it will
replace ataSmartStatus(). This queries the drive directly to
see if the SMART status is OK, rather than comparing thresholds to
attribute values ourselves. But I need to get some drives that fail
their SMART status to check it.
Removed extraneous space before printing in some error messages
Fixed additional typos in documentation
Fixed some character buffers that were too short for their contents.
Put project home path into all source files near the top
Corrected typos in the documentation
Modified Makefile so that Mandrake Cooker won’t increment version number
(unless they happen to be working on my machine, which I doubt!)
For IBM disks whose raw temp data includes three temps. print all
print timestamps for error log to msec precision
added -m option for Hitachi disks that store power on life in
added -L option for printing self-test error logs
in -l option, now print power on lifetime, so that one can see
when the error took place
updated SMART structure definitions to ATA-5 spec
added -p option
added -f and -F options to enable/disable autosave threshold
changed argv parsing to use getops -- elminate buffer overflow
expanded and corrected documentation
fixed problem with smartd. It did not actually call
ataSmartEnable()! Since the argument was left out, the test
always suceeded because it evaluated to a pointer to the function.
smartd: closed open file descriptors if device does not support
smart. Note: this still needs to be fixed for SCSI devices
STARTED with smartsuite-2.1-2
This is a demo version of txt2pdf v.10.1
Developed by SANFACE Software
Available at