P
ULSAR
.2 SAS P
RODUCT
M
ANUAL
, R
EV
. C
38
9.2
D
RIVE ERROR RECOVERY PROCEDURES
When an error occurs during drive operation, the drive performs error recovery procedures to attempt to recover the data. The error
recovery procedures used are not user changeable.
9.3
SAS
SYSTEM ERRORS
Information on the reporting of operational errors across the interface is given in the
SAS Interface Manual
. The SSP Response returns
information to the host about numerous kinds of errors. The Receive Diagnostic Results reports the results of diagnostic operations
performed by the drive.
Status returned by the drive to the initiator is described in the
SAS Interface Manual.
Status reporting plays a role in systems error
management and its use in that respect is described in sections where the various commands are discussed.
9.4
B
ACKGROUND
M
EDIA
S
CAN
Background Media Scan (BMS) is a self-initiated media scan. BMS is defined in the T10 document SPC-4 available from the T10
committee. BMS performs reads across the entire addressable space of the media while the drive is idle. In RAID arrays, BMS allows hot
spare drives to be scanned for defects prior to being put into service by the host system. On regular duty drives, if the host system makes
use of the BMS Log Page, it can avoid placing data in suspect locations on the media. Unreadable and recovered error sites will be logged
and reallocated.
With BMS, the host system can consume less power and system overhead by only checking BMS status and results rather than tying up
the bus and consuming power in the process of host-initiated media scanning activity.
Since the background scan functions are only done during idle periods, BMS causes a negligible impact to system performance. The BMS
scan is performed after 500ms of idle time. Other features that normally use idle time to function will function normally because BMS
functions for bursts of 500ms and then suspends activity for 100ms to allow other background functions to operate.
BMS interrupts immediately to service host commands from the interface bus while performing reads. BMS will complete any BMS-initiated
error recovery prior to returning to service host-initiated commands. Overhead associated with a return to host-servicing activity from BMS
only impacts the first command that interrupted BMS, this results in a typical delay of about 1ms.
9.5
A
UTO
-R
EALLOCATION
Auto-Reallocation allows the drive to reallocate unreadable locations on a subsequent write command if the recovery process deems the
location to be defective. The drive performs auto-reallocation on every WRITE command. With each write to a Logical LBA, the drive
writes the data to a different physical media location. Physical locations that return unrecoverable errors are retired during future WRITE
attempts and associated recovery process.
This is in contrast to the system having to use the REASSIGN BLOCKS command to reassign a location that was unreadable and then
generate a WRITE command to rewrite the data. This operation requires that AWRE and ARRE are enabled—this is the default setting
from the Seagate factory.
