1SG_RESET(8) SG3_UTILS SG_RESET(8)
2
3
4
6 sg_reset - sends SCSI device, target, bus or host reset; or checks
7 reset state
8
10 sg_reset [--bus] [--device] [--help] [--host] [--no-esc] [--target]
11 [--verbose] [--version] DEVICE
12
14 The sg_reset utility with no options (just a DEVICE) reports on the
15 reset state (e.g. if a reset is underway) of the DEVICE. When given a
16 --device, --target, --bus or --host option it requests a device, tar‐
17 get, bus or host reset respectively.
18
19 A device reset is applied to the Logical Unit (LU) corresponding to
20 DEVICE. It is most likely implemented by a Low level Driver (LLD) in
21 Linux as a LOGICAL UNIT RESET task management function.
22
23 The ability to reset a SCSI target was added in Linux kernel 2.6.27 . A
24 LLD may send Low level Drivers (LLDs) the I_T NEXUS RESET task manage‐
25 ment function. Alternatively it may use a transport mechanism to do the
26 same thing (e.g. a hard reset on the link containing a SAS target).
27
28 In the Linux kernel 2.6 and 3 series this utility can be called on sd,
29 sr (cd/dvd), st or sg device nodes; if the user has appropriate permis‐
30 sions.
31
32 Users of this utility can check whether a reset recovery is already
33 underway before trying to send a new reset with this utility. Calling
34 this utility with no options, just the DEVICE, will do such a check.
35
37 -b, --bus
38 attempt a SCSI bus reset. A bus reset is a SCSI Parallel Inter‐
39 face (SPI) concept not found in modern transports. A recent LLD
40 may implement it as a series of resets on targets that might be
41 considered as siblings to the target on the DEVICE path.
42
43 -d, --device
44 attempt a SCSI device reset. This would typically involve send‐
45 ing a LOGICAL UNIT RESET task management function to DEVICE.
46
47 -z, --help
48 print the usage message then exit.
49
50 -H, --host
51 attempt a host reset. The "host" in this context is often called
52 a Host Bus Adapter (HBA) and contains one or more SCSI initia‐
53 tors.
54
55 -N, --no-esc
56 without this option, if a device reset (--device) fails then it
57 will escalate to a target reset. And if a target reset (--tar‐
58 get) fails then it will escalate to a bus reset. And if a bus
59 reset (--bus) fails then it will escalate to a host reset. With
60 this option only the requested reset is attempted. An alternate
61 option name of --no-escalate is also accepted.
62
63 --no-escalate
64 The same as -N, --no-esc.
65
66 -t, --target
67 attempt a SCSI target reset. A SCSI target contains one or more
68 LUs. This would typically involve sending a I_T NEXUS RESET task
69 management function to DEVICE There may be a transport action
70 that is equivalent (e.g. in SAS a hard reset on the link that
71 contains the target).
72
73 -v, --verbose
74 increase the degree of verbosity (debug messages).
75
76 -V, --version
77 prints the version string then exits.
78
80 The error recovery code within the Linux kernel (SCSI mid-level) when
81 faced with a SCSI command timing out and no response from the device
82 (LU) does the following. First it tries a device reset and if that is
83 not successful tries a target reset. If that is not successful it tries
84 a bus reset. If that is not successful it tries a host reset. The
85 "device,target,bus,host" order is the reset escalation that the --no-
86 esc option attempts to stop. In large storage configurations the esca‐
87 lation may be (very) undesirable.
88
89 This utility calls the SG_SCSI_RESET ioctl and as of lk 3.10.7 the
90 --no-esc option is not supported. Patches to implement this functional‐
91 ity have not been accepted. If you want it, post your concern to linux-
92 scsi@vger.kernel.org which is the linux-scsi list.
93
94 SAM-4 and 5 define a hard reset, a LOGICAL UNIT RESET and a I_T NEXUS
95 RESET. A hard reset is defined to be a power on condition, a microcode
96 change or a transport reset event. LOGICAL UNIT RESET and I_T NEXUS
97 RESET can be requested via task management functions (and support for
98 LOGICAL UNIT RESET is mandatory). In Linux the SCSI subsystem leaves it
99 up to the LLDs as to exactly what type (if any) of reset is performed.
100 The "bus reset" is SCSI Parallel Interface (SPI) concept that may not
101 map well to recent SCSI transports so it may be a dummy operation. A
102 "host reset" attempts to re-initialize the HBA that the request passes
103 through en route to the DEVICE. Note that a "host reset" and a "bus
104 reset" may cause collateral damage.
105
106 This utility does not allow individual SCSI commands to be aborted.
107 SAM-4 defines ABORT TASK and ABORT TASK SET task management functions
108 for that.
109
110 Prior to SAM-3 there was a TARGET RESET task management function. And
111 in SAM-4 I_T NEXUS RESET appeared which seems closely related: the
112 "I_T" stands for Initiator-Target.
113
114 Transports may have their own types of resets not supported by this
115 utility. For example SAS has a link reset in which both ends of a
116 physical link (e.g. between a SAS expander and a SAS tape drive) rene‐
117 gotiate their connection.
118
119 Prior to version 0.57 of this utility the command line had short
120 options only (e.g. -d but not --device). Also -h invoked a host reset
121 while in the current version -h is equivalent to --help and both -H and
122 --host invoke a host reset. For backward compatibility define the envi‐
123 ronment variable SG3_UTILS_OLD_OPTS or SG_RESET_OLD_OPTS . In this case
124 -h will invoke a host reset and the output will be verbose as it was
125 previously (equivalent to using the --verbose option now). For exam‐
126 ple:
127
128 SG_RESET_OLD_OPTS=1 sg_reset -h /dev/sg1
129 sg_reset: starting host reset
130 sg_reset: completed host reset
131
133 Written by Douglas Gilbert.
134
136 Copyright © 1999-2013 Douglas Gilbert
137 This software is distributed under the GPL version 2. There is NO war‐
138 ranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PUR‐
139 POSE.
140
141
142
143sg3_utils-1.37 August 2013 SG_RESET(8)