Welcome to Jaehyek's Blog!

Here is my front-end learning path

eMMC & UFS About eMMC (1)

2014-05-22

Jaehyek

eMMC-UFS

eMMC UFS About

Here is the document about eMMC ppt doc

Multi Partition

eMMC 4.3  2 boot partitions and 1 user data partition
eMMC 4.4  RPMB partition is added
4 general purpose partitions and enhanced user data area can be set in normal user data area

001

Partitions	NAND type	Default Size	Remarks
Boot Area Partition 1	SLC Mode	128KB	Size as multiple of 128KB (max. 32MB)
Boot Area Partition 2	SLC Mode	128KB	Size as multiple of 128KB (max. 32MB)
RPMB Area Partition	SLC Mode	128KB	Size as multiple of 128KB (max. 32MB)
General Purpose Partitions	MLC “or”Enhanced Area	0KB	Available size can be seen by following: (EXT_CSD[145]x 82 + EXT_CSD[144]x 81 +EXT_CSD[143]) x HC_WP_GPR_SIZE x HC_ERASE_GPR_SIZE x 512KB byte
User Data Area ( Enhanced Area)	SLC Mode	0KB	Start address  multiple of Write Protect Group size
User Data Area ( Default Area)	MLC Mode	93.1%

RPMB(Replay Protect Memory Block)

Only handled by security key (SHA-256)
When host tries to access to RPMB

a) Host reads key the counter value to generate MAC (token) using SHA-256 algorithm
b) The host sends the generated MAC to eMMC controller
c) eMMC controller reads the key which was pre-loaded by host and the counter value to generate MAC using SHA-256
d) Compares two MAC. If two keys are identical, RPMB access is allowed

002

TRIM

Mark data that is invalid
Similar to STL_Delete()
Trim data should be shown as “0” or “1” (by JEDEC)

Secure TRIM/Erase

Secure Trim/Erase
- Step 1 : marks data which is logically deleted
- Step 2 : Actual secure purge operation is executed

-> secure purge: overwrite single character not only to trimed blocks but also to data which was generated by internal operation like wear-leveling, and then erase physically

003

Secure Trim and Secure Erase are identical except operation unit

Secure Erase: erase unit is Erase Group size (512KB)
Secure Trim : step1: sector based size(512byte) step2: purge unit is Write Group size(512KB)

HPI(High Priority Interrupt)

stops ongoing write operation and then read data within a specific time
aborts ongoing writing operation by CMD 12
If read/ write partition is different, additional partition switching time is needed for changing partition

004

Background Operation

Host gives free time to the device. During this time device can do housekeeping like merge operation
Host should check that necessity (4 levels) of background operation
Helps to improve performance

[Background operation]

Level	Descriptions
0	No operation required
1	Background operation which is not critical
2	Background operation which impacts on performance
3	Background operation which is critical

Data Atomicity

Normal Reliable Write

The old data that logical address point must retain unchanged until a new data is successfully written to the same logical address
Reliable write sector (EXT_CSD[222]) : 1 or 255

005

Enhanced Reliable Write – Sector based Atomicity

006

eMMC 4.5 Features

Category	Features	Meaning
Performance	HS200	MMC interface speed increases (104200Mbps)
	Packed CMD	Vectored read/write command (ex. readv/writev)
	Cache	Write-back & flush like HDD
Hint	Discard	Mark unused LBA without zero-filling (ex. Deleted files)
	Context ID	Specify context ID for write command (ex. File, task)
	Data tag	Specify data tag for write command (ex. FS meta)
	Partition type	Specify GPAP type (system code, non-persistent)
	Power-off notify	Host shall not intentionally power-off device without notification
	RTC	Send RTC info to device
	Dynamic capacity	Device request host for capacity reduction
Security	Sanitize	Purge all unused LBA

Trim & Discard

Benefits

Reduce write timeout by reducing garbage collection cost
Level 3 over-provision

Trim vs. Discard – misalignment handling

007

Trim is useful both for higher throughput and for reducing performance fluctuation

Packed Command

Reduces # of discrete I/O CMD processing (usually random pattern)  higher read and write IOPS
Internal data processing can be pipelined if there are multiple flash chips that can operate in parallel

008

Cache

Buffers write operations that usually come in burst  higher write IOPS, shorter write latency
May interleave read I/O if there is no ongoing flush operation

009

Power-off notification

eMMC FW could use power-off signal for
- marking where the last write operation happened
- flushing dirty metadata of FW into nonvolatile memory
As a result, eMMC initialization process could be shortened

Sanitize

Secure erase/trim (4.4) – mark data to be securely deleted and purge at once
Sanitize (4.5) – purge unused data at once  may take huge time for completion

010

Read All

11/11