16 June 2007 e07129r1 ATA8-ACS Endianness clarifications

Transcription

1 16 June 2007 e07129r1 ATA8-ACS Endianness clarifications To: T13 Technical Committee From: Rob Elliott, HP Date: 16 June 2007 Subject: e07129r1 ATA8-ACS Endianness clarifications Revision history Revision 0 (13 February 2007) First revision Revision 1 (16 June 2007) Updated to be based on ATA8-ACS revision 4a. Added editor s notes with rationale explaining the changes. Related documents d1699r4a - AT Attachment - 8 ATA/ATAPI Command Set (ATA8-ACS) revision 4a Overview Although most multi-byte fields in ATA are little-endian, some fields are unusual: a) The IDENTIFY (PACKET) DEVICE data SERIAL NUMBER field, MODEL NUMBER field, and FIRMWARE REVISION field, and possibly the MEDIA SERIAL NUMBER field each contain ASCII characters representing an ASCII string. However, they do not contain the characters in normal string order (with the first character at byte 0, the second character at byte 1, etc.). For historical reasons, each pair of bytes is swapped (byte 0 with byte 1, byte 2 with byte 3, byte 4 with byte 5, etc.). ATA/ATAPI-7 and prior versions included text describing that, but that text was lost as of ATA8-ACS revision 3f as part of removing text dependent upon parallel ATA. The special string order needs to be described again. A new name ATA string is proposed to describe strings in this order. b) The IDENTIFY (PACKET) DEVICE data WORLD WIDE NAME field contains all the bytes of an NAA format identifier, but it carries them neither byte-by-byte (as if the field were a single 8-byte (QWord) big-endian value) nor with all 8 bytes swapped (as if the field were a single 8-byte little-endian value). Instead, it carries the identifier in an almost-big-endian format, with each pair of bytes swapped (byte 0 with byte 1, byte 2 with byte 3, byte 4 with byte 5, and byte 6 with byte 7). This differs from how NAA identifiers are used in Fibre Channel and Serial Attached SCSI (SAS). Other multi-word fields in IDENTIFY DEVICE data are true little-endian and don t require unusual swapping. Since that format survived ATA/ATAPI-7 letter ballot, this proposal assumes it cannot be changed now (which could cause ATA/ATAPI-7 and ATA8-ACS implementations to have colliding worldwide names). In ATA8-ACS, the description is technically not inconsistent - just not obviously different, since the fields are shown defined as words. A byte-view of the data structure is proposed to highlight that the bytes are in different order, and that treating it as a 8-byte (QWord) little-endian value does not correct the problem. c) The CFA TRANSLATE SECTOR data contains LBA field and logical sector write cycles count field are both big-endian. Changes are proposed to clarify that ATA is little-endian (except for these exceptions). Suggested changes 2.1 Approved references Table 1 lists approved ANSI standards, approved international and regional standards (ISO, IEC, CEN/CENELEC, ITUT), may be obtained from the international and regional organizations who control them. 1

2 e07129r1 ATA8-ACS Endianness clarifications 16 June 2007 To obtain copies of these documents, contact Global Engineering or INCITS. Additional information may be available at and Table 1 Approved ANSI References Name Reference Information Systems - Coded Character Sets - 7-Bit American National Standard Code for Information Interchange (7-Bit ASCII) ANSI INCITS (R2002) Editor s Note 1: Need to define ASCII for completeness. Even this reference is incomplete, since INCITS 4 defines 7 bits not 8 bits (this typically goes into bits 6:0, but that s not said in ASCII itself). ISO/IEC solves that, but it calls the character set Latin-1 and defines characters above 7Fh that ATA does not use. 3.1 Definitions and abbreviations ASCII Character: Designates 8-bit value that is encoded using the ASCII Character set. 3.1.xx ASCII: American Standard Code for Information Interchange. 3.1.xx ASCII character: A byte containing a 7-bit ASCII pattern in bits 6:0 with bit 7 cleared to zero. 3.1.xx ATA string: A set of ASCII characters (see 3.1.xx) in the format specified in 3.2.xx. 3.1.xx byte: A sequence of eight contiguous bits considered as a unit. See xx DWord: A sequence of four contiguous bytes considered as a unit. See xx QWord: A sequence of eight contiguous bytes considered as a unit. See xx word: A sequence of two contiguous bytes considered as a unit. See xx big-endian: A format for storage or transmission of binary data in which the most significant byte appears first. In a multi-byte value, the byte containing the most significant bit is stored in the lowest memory address and transmitted first and the byte containing the least significant bit is stored in the highest memory address and transmitted last (e.g., for the value 0080h, the byte containing 00h is stored in the lowest memory address and the byte containing 80h is stored in the highest memory address). 3.1.xx little-endian: A format for storage or transmission of binary data in which the least significant byte appears first. In a multi-byte value, the byte containing the least significant bit is stored in the lowest memory address and transmitted first and the byte containing the most significant bit is stored in the highest memory address and transmitted last (e.g., for the value 0080h, the byte containing 80h is stored in the lowest memory address and the byte containing 00h is stored in the highest memory address). Editor s Note 2: Almost everything in ATA8-ACS is little endian. Defining both endian terms is helpful since a) SATA drives (little-endian) are being used in SAS systems (big-endian), where the host has to understand both types of endianness at the same time, and b) ATAPI is mainly used for running SCSI commands (which are big-endian) over ATA. Also, ATA is starting to define more fields larger than 2-bytes (e.g., 48-bit LBAs) where it needs to be clear that little-endian format should be followed, not some hybrid (as was inadvertantly done for the ATA strings and WWN field). These definitions match those in the T10 SAS and SAT standards. 3.1.xxx (least significant bit): In a binary code, the bit or bit position with the smallest numerical weighting in a group of bits that, when taken as a whole, represent a numerical value (e.g., in the number 0001b, the bit that is set to one). 2

3 16 June 2007 e07129r1 ATA8-ACS Endianness clarifications 3.1.xxx (most significant bit): In a binary code, the bit or bit position with the largest numerical weighting in a group of bits that, when taken as a whole, represent a numerical value (e.g., in the number 1000b, the bit that is set to one). Editor s Note 3: It is often confusing whether these acronyms mean byte or bit WWN (world wide name): A 64-bit worldwide unique name based upon a company s IEEE organizationally unique identifier (OUI), reported in IDENTIFY DEVICE data words (111:108)(see ) and IDENTIFY PACKET DEVICE data words (111:108)(see )(See IDENTIFY DEVICE data words (111:108), ). Editor s Note 4: SAT uses IDENTIFY (PACKET) DEVICE but ACS prefers to list them both (this was made consistent a while back). 3.2 Conventions Byte, word, DWord, and QWord Rrelationships Bytes contain 8 bits. Words contain 16 bits (i.e., 2 bytes). DWords contain 32 bits (i.e., 2 words or 4 bytes). QWords contain 64 bits (i.e., 2 DWords, 4 words, or 8 bytes). Editor s Note 5: In some CPU architectures, word means 4 bytes (hword means 2 bytes), so it is important to define these terms. IEEE thinks byte is so unclear they use octet everywhere; T13 should just define it. Keep the existing capitalization of byte word DWord and QWord, as those are now consistent in ata8-acs-r4. The preceding paragraph could be dropped in favor of just including the definitions in 3.1.xx (or the definitions could be dropped). Most tables that define data transferred by a command (e.g., the IDENTIFY DEVICE data in table 21) indicate the locations of fields by their word address. At each word address n, bits 15:8 of the word are stored at byte 2n+1 and bits 7:0 of the word are stored at byte 2n. Unless stated or defined otherwise, in a field containing a multi-byte value (e.g., a word, DWord, or QWord), the byte containing the is stored at the lowest address and the byte containing the is stored at the highest address (i.e., little-endian byte ordering). Exceptions include: a) each field containing an ATA string (e.g., the IDENTIFY DEVICE data and IDENTIFY PACKET DEVICE data Serial number, Firmware revision, and Model number fields) is considered to be an array of bytes, not a multi-byte value, and is handled as described in 3.2.xx; Editor s Note 6: The IDENTIFY DEVICE data Media serial number might also be an ATA string b) the IDENTIFY DEVICE data and IDENTIFY PACKET DEVICE data World Wide Name field is treated as four word fields rather than one QWord field and is handled as described in ; and c) the CFA TRANSLATE SECTOR data LBA and logical sector write cycles count fields are three-byte big endian fields (see table 16 in 7.4.5). Editor s Note 7: The only statement of the above rules (that SATA is little-endian) is in the IDENTIFY DEVICE data overview, but the rules apply to other commands as well like READ LOG EXT log pages. 3

4 e07129r1 ATA8-ACS Endianness clarifications 16 June 2007 For example: a) if the two-byte field (i.e., word) in SCT command (see table 68 in 8.3.1) word 0 contains 0007h, then: A) byte 0 contains 07h; and B) byte 1 contains 00h; b) if the four-byte field (i.e., DWord) at IDENTIFY DEVICE data words 61:60 (see table 22 in ) contains h (i.e., ), then: A) byte 120 contains 03h; B) byte 121 contains 02h; C) byte 122 contains 01h; and D) byte 123 contains 80h; and c) if a eight-byte field (i.e., QWord) in the WRITE SAME command words 5:2 (see table 73 in 8.3.3) contains _ h), then: A) byte 4 contains 00h; B) byte 5 contains 01h; C) byte 6 contains 02h; D) byte 7 contains 03h; E) byte 8 contains 04h; F) byte 9 contains 05h; G) byte 10 contains 00h; and H) byte 11 contains 00h; Editor s Note 8: Specific examples from real drives: For a 500 GB drive, IDENTIFY DEVICE data bytes (words 60-61) contain FF FF FF 0Fh = 0FFFFFFFh (the maximum value). Bytes (words ) contain A h = A386030h = 976,773,168d; that * 512 = 500,107,862,016 (i.e., just over 500 GB). For a 120 GB drive, bytes contain B0 4B F9 0Dh = 0DF94BB0h = 234,441,648, that * 512 = 120,034,123,776 (just over 120 GB). 4

5 16 June 2007 e07129r1 ATA8-ACS Endianness clarifications Figure 3 illustrates the relationship between bytes, words, DWords, and QWords used by this standard. Byte Word at offset n Byte (little-endian) DWord at word offset n D/ (little-endian) QWord at word offset n Q/ (little-endian) D/2+1 D/2 Figure 3 Byte, word, DWord and QWord relationships [redrawn with changes] Editor s Note 9: In the figure, made the addresses of the values be explicitly n based rather than implicitly 0-based so these concepts are more easily applied to tables such as IDENTIFY DEVICE data. Made n be the word-based value, since that is the unit used by most tables in the standard. Separated the byte, word, DWord, and QWord sections, which currently run together. Showed the bits inside the QWord. Added / labels. Added little-endian comment to clarify that ATA is indeed little-endian. 3.2.xx ATA string convention [new section] ATA strings are sequences of bytes containing ASCII graphic characters in the range of 20h-7Eh. ATA strings shall not contain values in the range of 00h-1Fh or 7Fh-FFh. 5

6 e07129r1 ATA8-ACS Endianness clarifications 16 June 2007 Three fields in IDENTIFY DEVICE data (see ) and IDENTIFY PACKET DEVICE data (see ) contain ATA strings: a) Serial number (words 10-19)(see ); b) Firmware revision (words 23-26)(see ); and c) Model number (word 27-46)(see ). Editor s Note 10: The IDENTIFY DEVICE media serial number might also be an ATA string However, these ATA strings are not stored in the normal ASCII string format where byte offset 0 contains the first character, byte offset 1 contains the second character, etc. Each pair of bytes in ATA strings is swapped as shown in table 2. Table 2 ATA string byte swapping Word Byte Character in string 0 0 Second character 1 First character 2 Fourth character 1 3 Third character n 2n Last character 2n+1 Second-to-last character For example, if the Firmware revision number field (words 23-26) contains the string abcdefg (including one padding space character at the end), its word and byte representations are shown in table 3. Table 3 ATA firmware revision example Word Value Byte Value h (i.e., ba ) h (i.e., dc ) h (i.e., fe ) h (i.e., g ) 36 62h (i.e., b ) 37 61h (i.e., a ) 38 64h (i.e., d ) 39 63h (i.e., c ) 40 66h (i.e., f ) 41 65h (i.e., e ) 42 20h (i.e.,, the space character) 43 67h (i.e., g ) Editor s Note 11: The rule about unusual ATA strings was lost in ATA8-ACS revision 3f. This section clearly defines how they are laid out, gives them a special name of ATA string, and provides a sample string value. describes cases where ATAPI drives have misunderstood the ATA string ordering. [end of all new section] 4 Feature set definitions 6

7 16 June 2007 e07129r1 ATA8-ACS Endianness clarifications 4.1 Overview Commands may be delivered in two forms. For devices that implement the General feature set, all commands and command parameters are delivered by transmitting a Command Block to the device. Devices that implement the PACKET feature set use DEVICE RESET and PACKET commands as well as a subset of the General feature set to control the device. The content of command packets delivered during execution of the PACKET command are defined in the standard indicated by IDENTIFY PACKET DEVICE data word 0 bits (12:8) (see ) and are not described in this standard. NOTE 1 - If the command packet contains a SCSI command, the command packet contains big-endian fields (see SPC-3) and parameter lists and parameter data transferred by the command contain big-endian fields (see SPC-3 and the applicable command set standard). Editor s Note 12: That change highlights where the contents of the packets are defined, and highlights a big-endian and little-endian difference. If the host issues a command that is indicated as not supported in the IDENTIFY DEVICE data, the device shall abort the command The PACKET command The PACKET command allows a host to send a command to the device via a command packet. The command packet contains the command and command parameters that the device is to execute (See clause 1). The protocol for handling the transmission of the PACKET command and associated data is transport specific. 7.4 CFA TRANSLATE SECTOR - 87h, PIO data-in Output Data Structure 7

8 e07129r1 ATA8-ACS Endianness clarifications 16 June bytes of data are transferred to the host (see table 16). Table 4 CFA TRANSLATE SECTOR data Bytes Description 00h-03h Obsolete 04h LBA bits (23:16) 05h LBA bits (15:8) 06h LBA bits (7:0) 07h-12h Reserved 13h Logical sector erased flag 14h-17h Reserved 18h Logical sector write cycles count bits (23:16) 19h Logical sector write cycles count bits (15:8) 1Ah Logical sector write cycles count bits (7:0) 1Bh - 1FFh Reserved 0 to 3 Obsolete 4 LBA (23:16) a 5 LBA (15:8) a 6 LBA (7:0) a 7 to 18 Reserved 19 Logical sector erased flag 20 to 23 Reserved 24 Logical sector write cycles count (23:16) a 25 Logical sector write cycles count (15:8) a 26 Logical sector write cycles count (7:0) a 27 to 511 Reserved a This byte is part of a big-endian 3-byte quantity. Editor s Note 13: Note in the above table that the of the LBA field is in the lowest offset 04h, and the is in the highest offset 06h - i.e., it is big-endian. It s not a DWord, though, since only 3 bytes are used - there is no name for a 3-byte value. Same for the Logical sector write cycles count field IDENTIFY DEVICE - ECh, PIO Data-in Feature Set This command is mandatory for all devices Description The IDENTIFY DEVICE command enables the host to receive a 512-byte block of data from the device. See for a description of the return data. 8

9 16 June 2007 e07129r1 ATA8-ACS Endianness clarifications Some devices may have to read the media in order for all applicable IDENTIFY DEVICE data fields to be valid. The IDENTIFY DEVICE data contains information regarding optional feature or command support. If the host issues a command that is indicated as not supported in the IDENTIFY DEVICE data, the device shall abort the command. Some parameters are defined as a 16-bit value. A word that is defined as a 16-bit value transmits the most significant byte of the value the first and the least significant byte second. Some parameters are defined as 32-bit values (e.g., words 60-61). Such values are transferred using two word transfers. The device shall first transfer the least significant word followed by the most significant value. Some parameters are defined as ASCII strings. ASCII strings are a list of values that range from 20h to 7Eh which the host should interpret ASCII Characters. Values 00h-1Fh and values 7Fh-FFh shall not appear in ASCII strings. Editor s Note 14: The first two stricken paragraphs are general, not just for IDENTIFY DEVICE. The last paragraph doesn t mention the unusual ordering of the strings Input Data Word O/M S/P F/V Description Table 21 IDENTIFY DEVICE data M B F Serial number (ASCII StringATA string) M B F Firmware revision (ASCII StringATA string) M B F Model number (ASCII StringATA string) 108 M B F 15:12 NAA (3:0) 11:0 IEEE OUI (23:12) 109 M B F 15:4 IEEE OUI (11:0) 3:0 Unique ID (35:32) 110 M B F 15:0 Unique ID (31:16) 111 M B F 15:0 Unique ID (15:0) O F Reserved for world wide name extension to 128 bits Words 10-19: Serial number This field contains the serial number of the device. The contents of this field is an ASCII stringata string of twenty bytes in the format defined in 3.2.xx. The device shall pad the character string with spaces (20h), if necessary, to ensure that the string is the proper length. The combination of Serial number (words 10-19) and Model number (words 27-46) shall be unique for a given manufacturer (See 0) Words 23-26: Firmware revision 9

10 e07129r1 ATA8-ACS Endianness clarifications 16 June 2007 This field contains the firmware revision number of the device. The contents of this field is an ASCII stringata string of eight bytes in the format defined in 3.2.xx. The device shall pad the character string with spaces (20h), if necessary, to ensure that the string is the proper length (See 0) Words 27-46: Model number This field contains the model number of the device. The contents of this field is an ASCII stringata string of forty bytes in the format defined in 3.2.xx. The device shall pad the character string with spaces (20h), if necessary, to ensure that the string is the proper length. The combination of Serial number (words 10-19) and Model number (words 27-46) shall be unique for a given manufacturer Words : World wide name Words shall contain the mandatory value of the world wide name (WWN) for the device. Words contain a mandatory World Wide Name (WWN) in the NAA IEEE Registered identifier format. Word 108 bits shall contain 5h, indicating that the naming authority is IEEE. All other values are reserved. Words 108 bits 11-0 and word 109 bits (15:4) shall contain the Organization Unique Identifier (OUI) (i.e., company ID) for the device manufacturer. The OUI shall be assigned by the IEEE/RAC as specified by ISO/IEC 13213:1994 assigned by the IEEE. The identifier may be obtained from: Institute of Electrical and Electronic Engineers, Inc. Registration Authority Committee 445 Hoes Lane Piscataway, NJ Word 109 bits (3:0), word 110, and word 111 shall contain a value assigned by the vendor that is unique for the device in the OUI domain. NOTE 2 - The NAA IEEE Registered identifier contained in words is represented by four little-endian words rather than one little-endian QWord. To convert it into a single big-endian QWord, swap each byte within each word (i.e., swap byte 0 with byte 1, byte 2 with byte 3, byte 4 with byte 5, and byte 6 with byte 7). Editor s Note 15: This format is neither true big-endian (like used in SCSI or Fibre Channel), nor true little-endian (like other 8-byte (i.e., QWord) fields in this standard such as words 103:100). In the word-based view shown in the table 22, the fields look nicely aligned; however, when viewed as bytes, it is more obvious that they are not contiguous. This note highlights that there is a difference, and the new tables proposed below visually show the differences. 10

11 16 June 2007 e07129r1 ATA8-ACS Endianness clarifications The IDENTIFY DEVICE data World Wide Name field is shown by word number in table 23. Table 23 IDENTIFY DEVICE data World Wide Name field (word-based view) Word Bit number within each word NAA (5h) 109 (bit 11) () (bit 23) IEEE OUI () (bit 0) () (bit 35) (bit 12) (bit 32) 110 (bit 31) (bit 16) 111 (bit 15) UNIQUE ID () (bit 0) The World Wide Name field is shown by byte number in table 24. Table 24 IDENTIFY DEVICE data World Wide Name field (byte-based view) Word Byte Bit number within each byte (bit 19) IEEE OUI (bit 12) 217 NAA (5h) 218 (bit 3) () (bit 0) () (bit 23) () (bit 35) (bit 20) UNIQUE ID (bit 32) 219 (bit 11) IEEE OUI (bit 4) 220 (bit 23) UNIQUE ID (bit 16) 221 (bit 31) (bit 24) 222 (bit 7) () (bit 0) 223 (bit 15) (bit 8) Words : Reserved for a 128-bit world wide name 7.17 IDENTIFY PACKET DEVICE - A1h, PIO Data-in Overview Table 24 IDENTIFY PACKET DEVICE data Word O/M S/P F/V Description 0 M B F 12:8 Field indicates command packet set used by the device 11

12 e07129r1 ATA8-ACS Endianness clarifications 16 June Word 0: General configuration Bits (15:14) of word 0 indicate the type of device. Bit 15 shall be set to one and bit 14 shall be cleared to zero to indicate the device implements the PACKET feature set. Bits (12:8) of word 0 indicate the command packet set implementedused by the device. This value follows the peripheral device type value as defined in SPC-3 (e.g., 05h indicates a CD/DVD device). Editor s Note 16: Mention the most common value used by CD/DVD devices Words 10-19: Serial number The use of these words is optional. If not implemented, the content shall be zeros. If implemented, the content shall be as described infor IDENTIFY DEVICE data words See Words 20-22: Reserved Words 23-26: Firmware revision Words shall have the content described for IDENTIFY DEVICE data words Words 27-46: Model number Words shall have the content described for IDENTIFY DEVICE data words Words : World wide name Words shall contain the mandatory value of the world wide name (WWN) for the device. Word 108 bits shall contain 5h, indicating that the naming authority is IEEE. All other values are reserved. Words 108 bits 11-0 and word 109 bits (15:4) shall contain the Organization Unique Identifier (OUI) for the device manufacturer. The OUI shall be assigned by the IEEE/RAC as specified by ISO/IEC 13213:1994. The identifier may be obtained from: Institute of Electrical and Electronic Engineers, Inc. Registration Authority Committee 445 Hoes Lane Piscataway, NJ Word 109 bits (3:0), word 110, and word 111 shall contain a value assigned by the vendor that is unique for the device in the OUI domain. Words shall have the content described for IDENTIFY DEVICE data word Editor s Note 17: All the other IDENTIFY PACKET DEVICE file descriptions just point to IDENTIFY DEVICE rather than restate the rules (and risk getting out of sync) Words : Reserved for a 128-bit world wide name SMART READ DATA - B0h/D0h, PIO data-in Feature Set This command is optional for devices that implement the SMART feature set Description This command returns the Device SMART data structure to the host. 12

13 16 June 2007 e07129r1 ATA8-ACS Endianness clarifications Table 49 defines the 512 bytes that make up the Device SMART data structure. All multi-byte fields shown in this structure follow the byte ordering described in Editor s Note 18: Although true, that statement is also true throughout the whole standard. Stating it a few times but not everywhere implies it might not be true in other places. A.4 Comprehensive SMART Error Log (Log Address 02h) A.4.1 Overview Table A.4 defines the format of each of the log pages that are part of the SMART comprehensive error log. The SMART Comprehensive error log provides logging for 28-bit addressing only. For 48-bit addressing, see A.7. The maximum size of the SMART comprehensive error log shall be 51 log pages. Devices may support fewer than 51 log pages. All multi-byte fields shown in this structure follow the byte ordering described in The comprehensive error log data structures shall include UNC errors, IDNF errors for which the address requested was valid, servo errors, write fault errors, etc. Comprehensive SMART Error Log data structures shall not include errors attributed to the receipt of faulty commands such as command codes not supported by the device or requests with invalid parameters or invalid addresses. A.7 Extended Comprehensive SMART Error log (Log Address 03h) A.7.1 Overview Table A.6 defines the format of each of the log pages that define the Extended Comprehensive SMART error log. The maximum size of the Extended Comprehensive SMART error log is 16,383 log pages. Devices may support fewer than 16,383 log pages. All multi-byte fields shown in this structure follow the byte ordering described in Error log data structures shall include Uncorrectable errors (See 6.2.9), ID Not Found errors (See 6.2.4) for which the address requested was valid, servo errors, write fault errors, etc. Error log data structures shall not include errors attributed to the receipt of faulty commands such as command codes not implemented by the device or requests with invalid parameters or invalid addresses. A.8 Extended SMART Self-Test Log (Log Address 07h) A.8.1 Overview Table A.11 defines the format of each of the log pages that define the Extended SMART Self-test log. The maximum size of the self-test log is 3,449 log pages. Devices may support fewer than 3,449 log pages. All multi-byte fields shown in this structure follow the byte ordering described in A.14 SMART Self-Test Log (Log Address 06h) A.14.1 Overview Table A.19 defines the log page that make up the SMART self-test log. All multi-byte fields shown in this structure follow the byte ordering described in The SMART Self-Test log supports 28-bit addressing only. A.15 Summary SMART Error Log (Log Address 01h) A.15.1 Overview Table A.21 defines the log page that make up the SMART summary error log. All multi-byte fields shown in this structure follow the byte ordering described in Summary SMART Error log data structures shall 13

14 e07129r1 ATA8-ACS Endianness clarifications 16 June 2007 include UNC errors, IDNF errors for which the address requested was valid, servo errors, write fault errors, etc. Summary error log data structures shall not include errors attributed to the receipt of faulty commands such as command codes not implemented by the device or requests with invalid parameters or invalid addresses. If the device supports the Comprehensive SMART Error log, then the Summary SMART Error log duplicates the last five error entries in the Comprehensive SMART Error log. The Summary SMART Error log supports 28-bit addressing only. Supplementary material (not proposed for ATA8-ACS) Thisi table shows both big-endian and little-endian. 14

15 16 June 2007 e07129r1 ATA8-ACS Endianness clarifications Word at offset n (big-endian) (little-endian) DWord at word offset n D/ (little-endian) D/ (big-endian) QWord at word offset n Q/ (little-endian) D/2+1 D/2 Q/ (big-endian) D/2 D/2+1 Figure 4 Big-endian and little-endian 15

16 e07129r1 ATA8-ACS Endianness clarifications 16 June