New Attributes for Network File System Version 4, Minor Version 2

OPTIONAL New Attributes - List and Definition References The list of New OPTIONAL attributes appears in . The meaning of the columns of the table are:

Name:: The name of the attribute.
Id:: The number assigned to the attribute.
Data Type:: The XDR data type of the attribute.
Acc:: Access allowed to the attribute. R means read-only.
Defined in:: The section of this specification that describes the attribute.

Name	Id	Data Type	Acc
supported_ops	83	bitmap4	R
dir_cookie_rising	84	bool	R
seek_granularity	85	uint64_t	R
mandatory_br_locks	86	bool	R
max_xattr_len	87	uint64_t	R

Definitions of new recommended attributes

Attribute 83: supported_ops This bit vector indicates which operations are supported for objects (of the appropriate type) with an fsid matching that of the specified object. The bit vector is a counted array of 32-bit integers used to contain bit values. The position of the integer in the array that contains the bit corresponding to operation n can be computed from the expression (n / 32), and its bit within that integer is (n mod 32), where n is the operation number.

Rationale Without this attribute, an NFSv4.2 client must attempt an optional operation to determine if the server supports it. This attribute allows the NFSv4.2 client to avoid attempting an optional operation when it is not supported for the file object's file system on the server. This attribute is likely to be particularly helpful in dealing with OPTIONAL attributes whose support is likely to be different for different file systems.

Attribute 84: dir_cookie_rising TRUE, if performing the READDIR operation on directories with a matching fsid always returns monotonically increasing directory offset cookies. This includes, if named attributes are supported, directories returned by OPENATTR.

Rationale If the NFSv4.2 client knows that directory offset cookies in READDIR replies are monotonically increasing, it might make it feasible to implement client side support for directory delegation. As an example, implementing client side directory delegation support is much easier when the directory offset cookies are monotonically increasing, so they can be used to order directory entries. Further, the client needs to know that the directory offset cookies are monotonically increasing before reading a directory, so that acquisition and use of a directory delegation may be done. Client side directory caching may also benefit from monotonically increasing directory offset cookies. For example, the Linux NFSv4.2 client uses a different caching algorithm when directory offset cookies are monotonically increasing.

Attribute 85: seek_granularity This attribute indicates the granularity of unallocated regions for data objects with an fsid matching that of the specified object. Data objects include regular files and, when named attributes are supported, named attributes. 0, if the SEEK operation will not return unallocated regions (holes). 1, if the SEEK operation will return unallocated regions (holes), but of no fixed granularity > 1, if the SEEK operation will return unallocated regions (holes), which are an exact multiple of this attribute in length. If this attribute is supported for a file system that does not support the SEEK operation, a value of 0 MUST be returned.

Rationale A NFSv4.2 client can avoid performing RPCs doing the SEEK operation when this attribute is equal 0 or whenever the client can determine that holes greater or equal to the value of this attribute do not exist in the file.

Attribute 86: mandatory_br_locks TRUE, if byte range locks obtained on data objects with an fsid matching that of the specified object have mandatory semantics potentially affecting IO operations done on overlapping areas. Data objects include regular files and, when named attributes are supported, named attributes.

Rationale Applications that work with advisory byte range locks will fail with mandatory byte range locks and vice versa. Given that both forms are allowed yet incompatible, it is necessary to provide a way, other than trial-and-error, to determine which form is supported. Further, if a client is doing read/write caching of data blocks using a write back caching policy, the caching requires locking if mandatory byte range locking is being enforced by the server. For this case, the client needs to acquire an exclusive byte range lock for each byte range being cached. If it does not do so, a write-back may fail, due to a conflicting lock having been acquired by a different client. Therefore, a client doing this kind of data block caching needs to know if the server is implementing mandatory byte range locking.

Attribute 87: max_xattr_len The maximum length, in bytes, of the extended attribute that can be set by the SETXATTR operation for file system objects with an fsid matching that of the specified object. The SETXATTR operation is described in the extension to NFSv4.2. If this attribute is supported for a file system that does not support the SETXATTR operation, a value of 0 MUST be returned.

Rationale This attribute allows a NFSv4.2 client to avoid attempting a SETXATTR operation when the length of the extended attribute is greater than the maximum specified by this attribute.

Implementation Status This section records the status of known implementations of the protocol defined by this specification at the time of posting of this Internet-Draft. The description of implementations in this section is intended to assist the IETF in its decision processes in progressing drafts to RFCs. Please note that the listing of any individual implementation here does not imply endorsement by the IETF. Furthermore, no effort has been spent to verify the information presented here that was supplied by IETF contributors. This is not intended as, and must not be construed to be, a catalog of available implementations or their features. Readers are advised to note that other implementations may exist.

FreeBSD NFS server and client

Organization:: FreeBSD Project
URL:
Maturity:: Prototype software based on the current document.
Coverage:: The bulk of this specification is implemented.
Licensing:: BSD
Implementation experience:: The implementation of these attributes have allowed the NFSv4.2 client to avoid unnecessary RPCs against the server. The current client implementation does not include support for directory delegations nor makes use of the dir_cookie_rising attribute. The current client implementation does not support mandatory byte range locking.