Version 13 (modified by 5 years ago) ( diff ) | ,
---|
SAT with UAS under Linux
Table of Contents
- Overview
- Technical Details
- Workarounds
- Permanent Flag Settings
- Temporary Settings
- Additional Information
Overview
As mentioned on the USB devices and smartmontools page, in certain situations the Linux "uas" driver disables SAT transfers, which prevents smartmontools
(and other tools, e.g. hdparm
) from communicating properly with the attached SATA device. See related tickets for examples.
This is because the Linux kernel rejects SAT ATA pass-through commands for certain devices with broken pass-through support. This list of devices affected depends on the kernel version, and may include all Seagate enclosures, Initio INIC-3069 and VIA VL711.
Technical Details
The problem is that certain USB-SATA-bridge chipsets do not properly pass SAT commands on to the SATA device when running in UAS mode, and to work around this problem Linux automatically enables the NO_ATA_1X flag for those chipsets. This turns off SAT passthrough... which in turn means that smartmontools is prevented from communicating with the device.
For many of these chipsets, SAT passthrough works correctly in the old BOT (usb-storage
driver) mode, and so smartmontools may suddenly stop working when the kernel is upgraded to a version which supports UAS, or from one which did did not apply the flag for that device to a newer kernel which does (see NO_ATA_1X kernel versions).
For more information on the usb-storage flags, search for the 'usb-storage.quirks' entry in the Kernel Parameters Documentation.
You can easily check to see whether the kernel using uas or usbstorage for a particular device using the lsusb -t
command:
# lsusb -t /: Bus 04.Port 1: Dev 1, Class=root_hub, Driver=xhci_hcd/2p, 10000M /: Bus 03.Port 1: Dev 1, Class=root_hub, Driver=xhci_hcd/2p, 480M /: Bus 02.Port 1: Dev 1, Class=root_hub, Driver=xhci_hcd/4p, 5000M |__ Port 1: Dev 11, If 0, Class=Mass Storage, Driver=usb-storage, 5000M [...]
or
# lsusb -t /: Bus 04.Port 1: Dev 1, Class=root_hub, Driver=xhci_hcd/2p, 10000M /: Bus 03.Port 1: Dev 1, Class=root_hub, Driver=xhci_hcd/2p, 480M /: Bus 02.Port 1: Dev 1, Class=root_hub, Driver=xhci_hcd/4p, 5000M |__ Port 1: Dev 8, If 0, Class=Mass Storage, Driver=uas, 5000M [...]
Alternatively you can check the messages that appear in kern.log when the device is first connected and look for the driver name mentioned at the end of the "scsi hostN" line:
Line | |
---|---|
1 | usb 2-1: new SuperSpeed USB device number 8 using xhci_hcd |
2 | usb 2-1: New USB device found, idVendor=NNNN, idProduct=NNN |
3 | usb 2-1: New USB device strings: Mfr=1, Product=2, SerialNumber=3 |
4 | [...] |
5 | scsi host4: uas |
6 | scsi 4:0:0:0: Direct-Access VENDOR-NAME 0204 PQ: 0 ANSI: 6 |
7 | sd 4:0:0:0: Attached scsi generic sg2 type 0 |
8 | sd 4:0:0:0: [sdc] 1953525168 512-byte logical blocks: (1.00 TB/932 GiB) |
9 | sd 4:0:0:0: [sdc] 4096-byte physical blocks |
10 | sd 4:0:0:0: [sdc] Write Protect is off |
11 | sd 4:0:0:0: [sdc] Mode Sense: 53 00 00 08 |
12 | sd 4:0:0:0: [sdc] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA |
13 | sdc: sdc1 sdc2 |
14 | sd 4:0:0:0: [sdc] Attached SCSI disk |
or
Line | |
---|---|
1 | usb 2-1: new SuperSpeed USB device number 9 using xhci_hcd |
2 | usb 2-1: New USB device found, idVendor=NNNN, idProduct=NNNN |
3 | usb 2-1: New USB device strings: Mfr=1, Product=2, SerialNumber=3 |
4 | [...] |
5 | usb-storage 2-1:1.0: USB Mass Storage device detected |
6 | scsi host4: usb-storage 2-1:1.0 |
7 | scsi 4:0:0:0: Direct-Access VENDOR-NAME 0204 PQ: 0 ANSI: 6 |
8 | sd 4:0:0:0: Attached scsi generic sg2 type 0 |
9 | sd 4:0:0:0: 1953525168 512-byte logical blocks: (1.00 TB/932 GiB) |
10 | sd 4:0:0:0: [sdc] Write Protect is off |
11 | sd 4:0:0:0: [sdc] Mode Sense: 47 00 00 08 |
12 | sd 4:0:0:0: [sdc] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA |
13 | sdc: sdc1 sdc2 |
14 | sd 4:0:0:0: [sdc] Attached |
The detailed selection logic used by the kernel to decide which devices get this flag can be found in the uas_use_uas_driver()
function in the drivers/usb/storage/uas-detect.h combined with the list of entries found in the drivers/usb/storage/unusual_uas.h
file. (As far as smartmontools/SAT passthrough is concerned, only the NO_ATA_1X flag is significant.)
For the uas driver, there doesn't seem to be a definitive way to confirm that the NO_ATA_1X flag is in fact in effect for a particular device (see note on /proc/scsi/usb-storage), but if it is the symptom will be that commands which rely on SAT show invalid data, e.g.
# smartctl -d sat -a /dev/sdb smartctl 6.6 2016-05-31 r4324 [x86_64-linux-4.15.0-33-generic] (local build) Copyright (C) 2002-16, Bruce Allen, Christian Franke, www.smartmontools.org Read Device Identity failed: scsi error unsupported field in scsi command A mandatory SMART command failed: exiting. To continue, add one or more '-T permissive' options. # smartctl -d sat -T permissive -a /dev/sdb smartctl 6.6 2016-05-31 r4324 [x86_64-linux-4.15.0-33-generic] (local build) Copyright (C) 2002-16, Bruce Allen, Christian Franke, www.smartmontools.org Read Device Identity failed: scsi error unsupported field in scsi command === START OF INFORMATION SECTION === Device Model: [No Information Found] Serial Number: [No Information Found] Firmware Version: [No Information Found] Device is: Not in smartctl database [for details use: -P showall] ATA Version is: [No Information Found] Local Time is: Mon Aug 27 20:44:01 2018 CEST SMART support is: Ambiguous - ATA IDENTIFY DEVICE words 82-83 don't show if SMART supported. SMART support is: Ambiguous - ATA IDENTIFY DEVICE words 85-87 don't show if SMART is enabled. A mandatory SMART command failed: exiting. To continue, add one or more '-T permissive
or
# hdparm -I /dev/sdc /dev/sdc: SG_IO: bad/missing sense data, sb[]: 70 00 05 00 00 00 00 0a 00 00 00 00 24 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ATA device, with non-removable media Standards: Likely used: 1 Configuration: Logical max current cylinders 0 0 heads 0 0 sectors/track 0 0 -- Logical/Physical Sector size: 512 bytes device size with M = 1024*1024: 0 MBytes device size with M = 1000*1000: 0 MBytes cache/buffer size = unknown Capabilities: IORDY not likely Cannot perform double-word IO R/W multiple sector transfer: not supported DMA: not supported PIO: pio0
Workarounds
There are two approaches to getting smartmontools to work with one of these devices, both involving setting a "quirks" flag for the device in question.
- If you just want to get smartmontools working again and don't care about the performance improvements the come from using UAS mode, you can simply use the IGNORE_UAS flag to tell Linux to stick with the usb-storage mode instead, by using the "u" flag on the
usb-storage.quirks
setting.
- If you want to continue using UAS mode and are sure your particular device does properly support SAT mode even though the kernel detects it as a device that does not (e.g. as described in ticket:971#comment:12) , you can override that default configuration by passing an empty flag list in the
usb-storage.quirks
setting.
Note that in either case, you will need to find the idVendor and idProduct fields for your device, in 4-digit hex format. This information is found in the New USB device found, idVendor=NNNN, idProduct=NNNN
kern.log line, or in the output from "lsusb". (In the option strings below, the 4-digit numbers can optionally be preceded a "0x" prefix.)
Permanent Flag Settings
The usb-storage.quirks
option is part of the more general mechanism for setting kernel/module options, and the exact mechanisms available for setting it will vary depending on your distribution. In brief:
- if your system has the "usb-storage" driver compiled as a loadable kernel module, you can set it by creating a file (e.g.
disable_uas.conf
) under/etc/modprobe.d/
containing the line (with the proper id numbers substituted)options usb-storage quirks=0bc2:231a:u
to disable UAS for that device (or
options usb-storage quirks=0bc2:231a:to try overriding the default NO_ATA_1X flag).
You will then probably need to follow your distribution's instructions for rebuilding your initramdisk in order to make the change apply when usb-storage drives are detected upon initial bootup.
- If the "usb-storage" driver is built in to the kernel, or if you simply prefer to make this change at the boot level rather than under /etc/modproble.d/, you can pass the option to the kernel as part of your boot configuration. For example, if your system uses grub2, you can edit
/boot/grub/default
so that in includes a line similar toGRUB_CMDLINE_LINUX="usb_storage.quirks=0bc2:231a:u"
and then run
update-grub
, and then reboot.
(Note that while the basic information used in these two approaches in the same, the punctuation used is different.)
In either case, you can then try plugging your USB device in again, and confirm using one of the methods above that the change had taken effect. (Note that if you use the :u
flag, the kern.log messages generated when the device is detected should now include
Line | |
---|---|
1 | usb 2-1: UAS is blacklisted for this device, using usb-storage instead |
2 | usb-storage 2-1:1.0: USB Mass Storage device detected |
3 | usb-storage 2-1:1.0: Quirks match for vid 0bc2 pid ab38: 800000 |
4 | scsi host4: usb-storage 2-1:1.0 |
For more information, see the introduction to the Kernel Parameters document, as well as documentation for your specific Linux distribution.
Temporary Settings
If you would like to temporarily set the quirks flag rather than making a permanent change (e.g. in order to run smartctl manually at some point but continuing to use UAS mode normally when using the drive for data transfers), you can do so by dynamically updating the module configuration (also described in the introductory paragraphs of the Kernel Parameters document).
First, unmount/unplug the USB device, then run the following commands (as root):
# cat /sys/module/usb_storage/parameters/quirks # echo "0x0bc2:0x231a:u" > /sys/module/usb_storage/parameters/quirks # cat /sys/module/usb_storage/parameters/quirks 0x0bc2:0x231a:u
(Note that the "echo" command completely replaces the current setting, so the first "cat" command makes sure that no quirks have been previously configured; it should just show a blank line.)
At this point you can plug the device's USB cable back in, and (after confirming via lsusb/kern.log that the device is connected to the usb-storage driver) you can then proceed to access the device using SAT.
A quirk set in this manner will automatically get cleared the next time the system is rebooted, but if you want to clear it more quickly you can simply execute:
# echo "" > /sys/module/usb_storage/parameters/quirks
to clear the configuration (i.e. use the kernel's compiled-in defaults) for the next time the device is plugged in.
Additional Information
"Quirks match" kern.log lines:
- The final number on the
Quirks match for vid 0bc2 pid ab38: 800000
kern.log line represents the set of flag bits that have been enabled (by the "quirks" option processing) for this particular USB device. (Note that the USB id numbers as well as the flag value are shown in hexadecimal.) - For SAT pass-through issues, the important flags are
IGNORE_UAS 0x00800000 :u NO_ATA_1X 0x02000000 :t
- The full list of flag bits is defined in the include/linux/usb_usual.h file in the kernel source (while the quirks-option flag letters and fuller explanations of each flag's meaning are found in the
usb-storage.quirks=
section of the Kernel Parameters documentation). - The "Quirks match" line will be output if quirks are found in either the compiled-in quirks list or a list is specified using one of the kernel-module override methods, and thus can be used to check which set of flags is actually getting applied. However, the blanket application of the NO_ATA_1X flag for all Seagate devices (USB vendor ID of 0x0bc2) is coded separately and does not generate any kern.log output indicating it has been applied. Also, if you specify an empty flag list (e.g. "
0bc2:231a:
") in an override and thus have no flag bits being set, no "Quirks match" log line will be generated.
Using /proc/scsi/usb-storage/[HOSTNUMBER} to display active flag bits:
- If a device is attached to the usb-storage driver, there will be an associated file under
/proc/scsi/usb-storage
which will display information related to that particular device driver instance's internal state... and in particular the Quirks: line displays a list of the active flag bits for this device -- whether set by a manual override, an entry in the unusual_devs.h or unusual_uas.h files, or by dynamically by logic coded within the driver.
For example, after setting the quirks string for this particular USB ID to ":u" and pugging in the device, the kernel detected it with a
scsi host4: usb-storage 2-1:1.0
message, and the /proc info file showed:============ # cat /proc/scsi/usb-storage/4 Host scsi4: usb-storage Vendor: Seagate Product: Backup+ RD Serial Number: NA5CGDGD Protocol: Transparent SCSI Transport: Bulk Quirks: IGNORE_UAS ============(thus confirming the quirks string had taken effect).
- Unfortunately the uas driver does not create an equivalent
/proc/scsi/uas/*
file, so this method doesn't help when you are attempting to determine (for example) whether or not the NO_ATA_1X flag is set on a uas device....
NO_ATA_1X Kernel Versions
- Because SAT pass-through behavior will change suddenly depending on the NO_ATA_1X flag, it can important to determine whether SAT issues are an (unexpected) side-effect of a kernel upgrade:
- For most devices this setting will be turned on by a change to the unusual_uas.h file... in which case the "Quirks mode" line in kern.log should show that the 0x02000000 bit is indeed getting set as part of the device initialization. This doesn't differentiate between entries from the unusual_uas.h file and manual settings from the quirks option string, but if a kernel upgrade (with no change to the quirk options in use) suddenly causes that message to appear, it's very likely a patch was made to unusal_uas.h between the old and new kernel versions.
- However, for the patch setting this flag for all Seagate devices there is no specific kern.log message to look for, so one may have to guess the patch status based on the kernel version number. The patch was created in November 2017 and applied to various kernel.org branches in December of 2017 and early 2018: here is a list some kernel branches with the first release within that branch to include the patch:
- Linux distributions may backport patches to their own kernel branches (and those may use somewhat different version numbering schemes), but certainly kernel packages released up through Nov 2017 will not have the patch, and mostly like ones based on the upstream 4.15-or-later branches will include it.
Web links:
Linux SAT-with-UAS bug reports for other projects:
General info:
- StackExchange: How can I check whether USB3.0 UASP (USB Attached SCSI Protocol) mode is enabled in Linux?
- Suze:USB 3.0 Hard Drive troubleshooting
Examples of the problems that originally triggered the current kernel behavior: