NVMe Base 2.4 Chapter 4: SQE, CQE, Status, PRP, and SGL

01 The common SQE format and command fields

01.01.The common SQE fixes the locations of CDW0, NSID, metadata/data pointers, and CDW10-15. OPC selects the command, CID creates the completion association, and PSDT selects DPTR interpretation. Only after these common fields are established should command-specific CDW10-15 definitions be applied. Figures 92-94 are the coordinate system for all later command construction.

metadata

Additional information stored with a logical block; it can contain protection information or serve other purposes.

DPTR

Data Pointer, the SQE field identifying a command data buffer.

NSID

Namespace Identifier, a controller-visible numeric handle for a namespace; the identifier is not the namespace object itself.

PSDT

PRP or SGL for Data Transfer, the CDW0 field selecting PRP or SGL interpretation for DPTR.

CDW

CDW (Command Dword): a 32-bit command field. In CDW10, 10 is the field index, not a byte offset.

CID

Command Identifier, used with the SQ identifier to identify an outstanding command.

01.02.The common Admin and I/O SQE is 64 bytes. CDW0, NSID, data pointers, and CDW10-15 establish the common layout before each command defines command-specific content.

Admin

Administrative, the control path used to create, configure, query, or manage controllers and queues.

I/O

Input/Output, the class of data operations performed on a namespace.

MPTR

Metadata Pointer, the SQE field identifying a separate metadata buffer.

02 CQEs: new entries, command identity, and results

02.01.The host first uses the Phase Tag to determine whether a CQ slot contains a new completion. After ownership is established, SQID/CID recovers the command; SCT then selects the status category before SC, DNR, and CRD are interpreted. Figures 97-109 must be read in this order so a stale CQE or wrong category is not mistaken for a command failure.

Host

The system running the operating system and issuing NVMe commands.

CRD

Command Retry Delay, the status field selecting a controller-recommended retry delay.

DNR

Do Not Retry, a CQE-status bit indicating that retrying the same command is not expected to succeed.

SCT

Status Code Type: selects the completion-status category and is interpreted with SC.

CQ

Completion Queue, the queue into which a controller posts command completions.

SC

Status Code: identifies the completion result within the selected SCT category.

02.02.The common CQE is at least 16 bytes. If multiple writes construct it, the Phase Tag shall be updated in the last write so the host does not consume a partial entry.

03 How PRPs describe data across pages

03.01.PRP1 may address any byte within the first memory page, so first-segment capacity is page_size minus offset. If data crosses that page, PRP2 represents either the second page or a PRP List depending on remaining length; later page addresses must be page aligned. Figures 110-113 define address calculation, not merely pointer names.

offset

offset: a displacement measured from a stated start. It answers “how far from the start,” unlike an index.

03.02.A fixed-size PRP entry points to a physical memory page. The first entry may contain a page offset; subsequent PRPs shall obey page alignment, and transfer length determines the required entry count.

page offset

page offset: the starting displacement inside the first memory page; after a page boundary, the next page pointer determines the location.

04 SGL data and segment descriptors

04.01.An SGL descriptor combines type/subtype, address, and length. A Data Block addresses data, Segment and Last Segment address more descriptors, and Bit Bucket represents data that need not be stored in memory. Figures 114-125 require type-first decoding; blindly following an address before decoding type is incorrect.

04.02.An SGL describes a data buffer through one or more descriptors and segments. SGL length shall equal or exceed the requested transfer length; this report covers only generic descriptors applicable to PCIe.

PCIe

PCI Express, the transport and device interconnect used by an NVMe memory-based controller.

05 Feature values, identifiers, lists, and strings

05.01.Establish what a value represents before interpreting it: Feature values distinguish active from saved state, identifiers have an identity scope, and lists and strings each have layout and length rules.

05.02.A Feature may have default, saved, and current values. Saved-value support and persistence across resets or power cycles are determined from SSFS and each Feature capability.

NUMCIDS

Number of Controller Identifiers: the count of valid controller IDs in a Controller List.

EUI64

64-bit Extended Unique Identifier, a 64-bit identifier constructed from IEEE-assigned space.

NGUID

Namespace Globally Unique Identifier, a 128-bit global identifier for a namespace.

SSVID

Subsystem Vendor ID, the PCI identifier for a subsystem vendor.

UUID

Universally Unique Identifier, a 128-bit identifier whose association scope is defined by the containing structure.

VID

Vendor ID, a PCI-SIG-assigned identifier for a vendor.

MN

Model Number, a string identifying a product model.

SN

Serial Number, a string identifying a product instance.