On 2018-04-02 07:10 AM, Mahesh Rajashekhara wrote:
Hello,
I am RAID HBA driver engineer here at Microsemi. We are working on linux driver
development for Microsemi SAS/SATA RAID HBA controllers.
As per our understanding, while a drive is processing the SANITIZE command:
- The drive should still be exposed to the OS.
- The drive will fail all commands other than REQUEST SENSE
Not sure that the drive should ever knowingly fail INQUIRY and REPORT
LUNS. That is to allow a discovery process to occur. In the INQUIRY
response the peripheral qualifier should be 000b by my reading.
Go look at sbc4r15.pdf, specifically clause 4.11.2 to see all the
commands that should be responded to during a SANITIZE (there are
9 of them).
When we initiate SCSI SANITIZE operation on disk drive, from the OS dmesg, we
could see that continuous SCSI READ (10) commands have been sent by the OS
upper layer drivers
and drive reports ASC: 04 ASCQ: 1B "LOGICAL UNIT NOT READY, SANITIZE IN
PROGRESS". Since the sanitize operation is in progress, the drive will fail the
commands.
Although, OS could see drive is sanitizing, we are not sure why does OS upper
layer driver is sending SCSI READ (10).
During the OS boot, if the drive is sanitizing, OS boot is taking very long
time and call trace issue has been occurred.
It would be of great help if you could shed some light on this issue.
Snippet of dmesg:
[ 202.659196] sd 0:0:11:0: [sdm] CDB: Read(10) 28 00 00 00 00 00 00 00 01 00
[ 202.659198] blk_update_request: I/O error, dev sdm, sector 0
[ 202.659201] Buffer I/O error on dev sdm, logical block 0, async page read
[ 202.659294] sd 0:0:11:0: [sdm] FAILED Result: hostbyte=DID_OK
driverbyte=DRIVER_SENSE
[ 202.659299] sd 0:0:11:0: [sdm] Sense Key : Not Ready [current]
[ 202.659303] sd 0:0:11:0: [sdm] Add. Sense: Logical unit not ready, sanitize
in progress
[ 202.659308] sd 0:0:11:0: [sdm] CDB: Read(10) 28 00 00 00 00 00 00 00 01 00
[ 202.659311] blk_update_request: I/O error, dev sdm, sector 0
[ 202.659314] Buffer I/O error on dev sdm, logical block 0, async page read
[ 202.659322] sdm: unable to read partition table
[ 202.659633] sd 0:0:11:0: [sdm] FAILED Result: hostbyte=DID_OK
driverbyte=DRIVER_SENSE
[ 202.659638] sd 0:0:11:0: [sdm] Sense Key : Not Ready [current]
[ 202.659641] sd 0:0:11:0: [sdm] Add. Sense: Logical unit not ready, sanitize
in progress
[ 202.659645] sd 0:0:11:0: [sdm] CDB: Read(10) 28 00 00 00 00 00 00 00 01 00
[ 202.659648] blk_update_request: I/O error, dev sdm, sector 0
[ 202.660621] sd 0:0:11:0: [sdm] Spinning up disk...
[ 204.067155] Buffer I/O error on dev sdm, logical block 488378624, async
page read
[ 204.122152] SGI XFS with ACLs, security attributes, no debug enabled
[ 204.123494] XFS (dm-0): Mounting V5 Filesystem
[ 204.289817] XFS (dm-0): Ending clean mount
[ 240.069283] INFO: task systemd-udevd:251 blocked for more than 120 seconds.
[ 240.069289] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables
this message
[ 203.661324] .....................................
[ 240.069293] systemd-udevd D
[ 240.069296] ffffffffc00c3580 0 251 1 0x00000004
[ 240.069301] ffff88007e0e7d18 0000000000000082 ffff88007e0ceeb0
ffff88007e0e7fd8
[ 240.069306] ffff88007e0e7fd8 ffff88007e0e7fd8 ffff88007e0ceeb0
ffffffffffffffff
[ 240.069311] 0000000000000000 ffffffffc00c35d0 0000000000000001
ffffffffc00c3580
[ 240.069316] Call Trace:
[ 240.069328] [<ffffffff816a94c9>] schedule+0x29/0x70
[ 240.069334] [<ffffffff810b7e85>] async_synchronize_cookie_domain+0x85/0x150
[ 240.069340] [<ffffffff810b1910>] ? wake_up_atomic_t+0x30/0x30
[ 240.069346] [<ffffffff810b7f87>] async_synchronize_full+0x17/0x20
[ 240.069351] [<ffffffff81100790>] load_module+0x1fc0/0x29e0
[ 240.069358] [<ffffffff8134bbf0>] ? ddebug_proc_write+0xf0/0xf0
[ 240.069365] [<ffffffff81101275>] SyS_init_module+0xc5/0x110
[ 240.069371] [<ffffffff816b4fc9>] system_call_fastpath+0x16/0x1b
The scsi mid-level and/or the sd driver sends TEST UNIT READY (TUR)
commands and should take note of the NOT READY sense key and should
not send READ commands until that condition clears. If we are talking
about modern SCSI devices then REQUEST SENSE should respond with progress
information.
Doug Gilbert
