NVMe on RHEL7

Posted 2021-01-22 tcicy

原文地址https://www.dell.com/support/article/cn/zh/cnbsd1/sln312382/nvme-on-rhel7?lang=en

 

Posted on behalf of Lakshmi Narayanan Durairajan ([email protected])

What is NVMe?

NVM Express [NVMe] or Non-Volatile Memory Host Controller Interface Specification (NVMHCI), is a specification for accessing solid-state drives (SSDs) attached through the PCI Express (PCIe) bus. NVM is an acronym for non-volatile memory, as used in SSDs.

NVMe defines optimized register interface, command set and feature set for PCIe SSD’s. NVMe focuses to standardize the PCIe SSD’s and improve the performance

PCIe SSD devices designed based on the NVMe specification are NVMe based PCIeSSD’s

For more details on the NVMe please refer the link http://www.nvmexpress.org/ .The NVMe devices used currently are NVMe 1.0c compliant

In this blog we will be looking into RHEL 7 support for the NVMe devices.

Currently DELL support the NVMe devices with RHEL 7 out of box [vendor based] driver

Following are the list of the things that we will look into:

• NVMe - Features Supported
• NVMe Device : Listing the device and its Capabilities
• Checking MaxPayLoad
• NVMe Driver : List the driver information
• NVMe Device Node and Naming conventions
• Formatting with xfs and mounting the device
• Using ledmon utility to manage backplane LEDs for NVMe device

NVMe- Features Supported

NVMe driver exposes the following features

• Basic IO operations
• Hot Plug
• Boot Support [UEFI and Legacy]

The following table lists the RHEL 7 [Out of box] driver supported features for NVMe on 12G and 13 G machines

Generation	Basic IO	Hot Plug	UEFI Boot	Legacy Boot
13 G	Yes	Yes	Yes	No
12 G	Yes	Yes	No	No

Table 1: RHEL 7 Driver Support

NVMe Device: Listing the device and its Capabilities

1) List the RHEL 7 OS information

[[email protected] ~]# uname -a

Linux localhost.localdomain 3.10.0-123.el7.x86_64 #1 SMP Mon May 5 11:16:57 EDT 2014 x86_64 x86_64 x86_64 GNU/Linux 

2) Get the device details by using the lspci utility

a) We support Samsung based NVMe drives. First get the pci slot id by using the following command

[[email protected] ~]# lspci | grep -i Samsung

45:00.0 Non-Volatile memory controller: Samsung Electronics Co Ltd NVMe SSD Controller 171X (rev 03)

47:00.0 Non-Volatile memory controller: Samsung Electronics Co Ltd NVMe SSD Controller 171X (rev 03)

b) The slot id will be listed as shown in the below [ Fig 1]. Here ‘’45:00.0"and "47:00.0"are the slots on which the drives are connected.


Figure 1: lspci listing the slot id

a) Use the slot id and use the following lspci options to get the device details ,capabilities and the corresponding driver

[[email protected] ~]# lspci -s 45:00.0 -v

45:00.0 Non-Volatile memory controller: Samsung Electronics Co Ltd NVMe SSD Controller 171X (rev 03) (prog-if 02)

Subsystem: Dell Express Flash NVMe XS1715 SSD 800GB

Physical Slot: 25

Flags: bus master, fast devsel, latency 0, IRQ 76

Memory at d47fc000 (64-bit, non-prefetchable) [size=16K]

Capabilities: [c0] Power Management version 3

Capabilities: [c8] MSI: Enable- Count=1/32 Maskable+ 64bit+

Capabilities: [e0] MSI-X: Enable+ Count=129 Masked-

Capabilities: [70] Express Endpoint, MSI 00

Capabilities: [40] Vendor Specific Information: Len=24 <?>

Capabilities: [100] Advanced Error Reporting

Capabilities: [180] #19

Capabilities: [150] Vendor Specific Information: ID=0001 Rev=1 Len=02c <?>

Kernel driver in use: nvme


The below [Fig 2] shows the Samsung NVMe device and the device details listed. It also shows name of the driver ‘nvme’ in this case for this device


Figure 2: lspci listing NVMe device details

Checking MaxPayLoad

Check the MaxPayload value by executing the following commands. It should set it to 256 bytes [Fig.3]

[[email protected] home]# lspci | grep -i Samsung

45:00.0 Non-Volatile memory controller: Samsung Electronics Co Ltd NVMe SSD Controller 171X (rev 03) 

[[email protected] home]# lspci -vvv -s 45:00.0


Figure 3: MaxPayload set to 256 bytes

NVMe Driver: List the driver information

1) Use modinfo command to list the diver details

[[email protected] ~]# modinfo nvme

filename: /lib/modules/3.10.0-123.el7.x86_64/extra/nvme/nvme.ko

version: 0.8-dell1.17

license: GPL

author: Samsung Electronics Corporation

srcversion: AB81DD9D63DD5DADDED9253

alias: pci:v0000144Dd0000A820sv*sd*bc*sc*i*

depends: 

vermagic: 3.10.0-123.el7.x86_64 SMP mod_unload modversions

parm: nvme_major:int

parm: use_threaded_interrupts:int 

The below [Fig 4] shows details of the NVMe driver nvme.ko 


Figure 4: Modinfo listing driver information

NVMe Device Node and Naming conventions

1) cat /proc/partitions displays the device node of nvme.

a) Following command run lists the nvme device as nvme0n1 and nvme1n1

[[email protected] ~]# cat /proc/partitions

major minor #blocks name 

259 0 781412184 nvme0n1

8 0 1952448512 sda

8 1 512000 sda1

8 2 1951935488 sda2

11 0 1048575 sr0

253 0 52428800 dm-0

253 1 16523264 dm-1

253 2 1882980352 dm-2

259 3 390711384 nvme1n1 

Partition the device using the any partitioning tools (fdisk,parted)

b) Executing the following command again, lists nvme device along with partitions

[[email protected] ~]# cat /proc/partitions

major minor #blocks name 

259 0 781412184 nvme0n1

259 1 390705068 nvme0n1p1

259 2 390706008 nvme0n1p2

8 0 1952448512 sda

8 1 512000 sda1

8 2 1951935488 sda2

11 0 1048575 sr0

253 0 52428800 dm-0

253 1 16523264 dm-1

253 2 1882980352 dm-2

259 3 390711384 nvme1n1

259 4 195354668 nvme1n1p1

259 5 195354712 nvme1n1p2 

Naming conventions:

The below [Fig 5] explains the naming convention of the device nodes

The number immediately after the string "nvme" is the device number

Example:

nvme0n1 – Here the device number is 0

Partitions are appended after the device name with the prefix ‘p’

Example:

nvme0n1p1 – partition 1

nvme1n1p2 – partition 2

Example:

nvme0n1p1 – partition 1 of device 0

nvme0n1p2 – partition 2 of device 0

nvme1n1p1 – partition 1 of device 1

nvme1n1p2 – partition 2 of device 1 


Figure 5: Device node naming conventions

Formatting with xfs and mounting the device

1) The following command formats the nvme partition 1 on device 1 to xfs 

[[email protected] ~]# mkfs.xfs /dev/nvme1n1p1

meta-data=/dev/nvme1n1p1 isize=256 agcount=4, agsize=12209667 blks

= sectsz=512 attr=2, projid32bit=1

= crc=0

data = bsize=4096 blocks=48838667, imaxpct=25

= sunit=0 swidth=0 blks

naming =version 2 bsize=4096 ascii-ci=0 ftype=0

log =internal log bsize=4096 blocks=23847, version=2

= sectsz=512 sunit=0 blks, lazy-count=1

realtime =none extsz=4096 blocks=0, rtextents=0 

2) Mount the device to a mount point and list the same 

[[email protected] ~]# mount /dev/nvme1n1p1 /mnt/

[[email protected] ~]# mount | grep -i nvme

/dev/nvme1n1p1 on /mnt type xfs (rw,relatime,seclabel,attr2,inode64,noquota) 

Using ledmon utility to manage backplane LEDs for NVMe device

Ledmon and ledctl are two utilities for Linux that can be used to control LED status on drive backplanes. Normally drive backplane LEDs are controlled by a hardware RAID controller (PERC), but when using Software RAID on Linux (mdadm) for NVMe PCIE SSD, the ledmon daemon will monitor the status of the drive array and update the status of drive LEDs.

For extra reading check the link 

https://www.dell.com/support/article/SLN310523/

Following are the steps to install and use the ledmon/ledctl utility

1) Installing OpenIPMI and ledmon/ledctl utilities:

Execute the following commands to install OpenIPMI and ledmon

[[email protected] ~]# yum install OpenIPMI

[[email protected] ~]# yum install ledmon-0.79-3.el7.x86_64.rpm 

2) Use ledmod/ledctl utilities 

Running ledctl and ledmon concurrently, ledmon will eventually override the ledctl settings

a) Start and check the status of ipmi as shown in the [Fig.6] using the following command

[[email protected] ~]# systemctl start ipmi


Figure 6: IPMI start and status

a) Start the ledmod

[[email protected] ~]# ledmon

b) [Fig 7] shows LED status after executing ledmon for the working state of the device

Figure 7: LED status after ledmon run for working state of the device (green) 

a) The below command will blink drive LED [on the device node /dev/nvme0n1 ]

[[email protected] ~]# ledctl locate=/dev/nvme0n1

Below command will blink both the drive LEDs [on the device node /dev/nvme0n1 and /dev/nvme1n1]

[[email protected] ~]# ledctl locate={ /dev/nvme0n1 /dev/nvme1n1 }

And the following command will turn off the locate LED

[[email protected] ~]# ledctl locate_off=/dev/nvme0n1
???????