NFSv4 Working Group David L. Black Internet Draft Stephen Fridella Expires: June 2006 EMC Corporation December 30, 2005 pNFS Block/Volume Layout draft-ietf-nfsv4-pnfs-block-00.txt Status of this Memo By submitting this Internet-Draft, each author represents that any applicable patent or other IPR claims of which he or she is aware have been or will be disclosed, and any of which he or she becomes aware will be disclosed, in accordance with Section 6 of BCP 79. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet- Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress." The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/1id-abstracts.txt The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html This Internet-Draft will expire in April 2006. Abstract Parallel NFS (pNFS) extends NFSv4 to allow clients to directly access file data on the storage used by the NFSv4 server. This ability to bypass the server for data access can increase both performance and parallelism, but requires additional client functionality for data access, some of which is dependent on the class of storage used. The main pNFS operations draft specifies storage-class-independent extensions to NFS; this draft specifies the additional extensions (primarily data structures) for use of pNFS with block and volume based storage. Black Expires June 2006 [Page 1] Internet-Draft pNFS Block/Volume Layout December 2005 Conventions used in this document The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC-2119 [RFC2119]. Table of Contents 1. Introduction...................................................3 2. Background and Architecture....................................3 2.1. Data Structures: Extents, Extent Lists, and Volumes.......4 2.1.1. Layout Requests and Extent Lists.....................8 2.1.2. Layout Commits.......................................9 2.1.3. Layout Returns.......................................9 2.1.4. Client Copy-on-Write Processing.....................10 2.1.5. Extents are Permissions.............................11 2.1.6. End-of-file Processing..............................12 2.2. Volume Identification....................................14 3. New Operations Issues.........................................15 3.1. Server Controlling Client Access to Block Devices........15 3.1.1. Guarantees Provided by Layouts......................15 3.1.2. I/O In-flight Issues................................15 3.1.3. Crash Recovery Issues...............................15 3.2. End-of-File (EOF) Handling issues........................16 3.2.1. Truncation..........................................16 3.2.2. Extension...........................................17 3.2.3. Extension vs. Readable Layouts......................17 4. Security Considerations.......................................18 5. Conclusions...................................................19 6. IANA Considerations...........................................19 7. Revision History..............................................19 8. Acknowledgments...............................................20 9. References....................................................20 9.1. Normative References.....................................20 9.2. Informative References...................................20 Author's Addresses...............................................21 Intellectual Property Statement..................................21 Disclaimer of Validity...........................................22 Copyright Statement..............................................22 Acknowledgment...................................................22 Black Expires June 2006 [Page 2] Internet-Draft pNFS Block/Volume Layout December 2005 1. Introduction Figure 1 shows the overall architecture of a pNFS system: +-----------+ |+-----------+ +-----------+ ||+-----------+ | | ||| | NFSv4 + pNFS | | +|| Clients |<------------------------------>| Server | +| | | | +-----------+ | | ||| +-----------+ ||| | ||| | ||| +-----------+ | ||| |+-----------+ | ||+----------------||+-----------+ | |+-----------------||| | | +------------------+|| Storage |------------+ +| Systems | +-----------+ Figure 1 pNFS Architecture The overall approach is that pNFS-enhanced clients obtain sufficient information from the server to enable them to access the underlying storage (on the Storage Systems) directly. See the pNFS portion of [NFSV4.1] for more details. This draft is concerned with access from pNFS clients to Storage Systems over storage protocols based on blocks and volumes, such as the SCSI protocol family (e.g., parallel SCSI, FCP for Fibre Channel, iSCSI, SAS). This class of storage is referred to as block/volume storage. While the Server to Storage System protocol is not of concern for interoperability here, it will typically also be a block/volume protocol when clients use block/volume protocols. 2. Background and Architecture The fundamental storage abstraction supported by block/volume storage is a storage volume consisting of a sequential series of fixed size blocks. This can be thought of as a logical disk; it may be realized by the Storage System as a physical disk, a portion of a physical disk or something more complex (e.g., concatenation, striping, RAID, and combinations thereof) involving multiple physical disks or portions thereof. Black Expires June 2006 [Page 3] Internet-Draft pNFS Block/Volume Layout December 2005 A pNFS layout for this block/volume class of storage is responsible for mapping from an NFS file (or portion of a file) to the blocks of storage volumes that contain the file. The blocks are expressed as extents with 64 bit offsets and lengths using the existing NFSv4 offset4 and length4 types. Clients must be able to perform I/O to the block extents without affecting additional areas of storage (especially important for writes), therefore extents MUST be aligned to 512-byte boundaries, and SHOULD be aligned to the block size used by the NFSv4 server in managing the actual filesystem (4 kilobytes and 8 kilobytes are common block sizes). This block size is available as an NFSv4 attribute - see Section 11.4 of [NFSV4.1]. The pNFS operation for requesting a layout (LAYOUTGET) includes the "pnfs_layoutiomode4 iomode" argument which indicates whether the requested layout is for read-only use or read-write use. A read-only layout may contain holes that are read as zero, whereas a read-write layout will contain allocated, but uninitialized storage in those holes (read as zero, can be written by client). This draft also supports client participation in copy on write by providing both read-only and uninitialized storage for the same range in a layout. Reads are initially performed on the read-only storage, with writes going to the uninitialized storage. After the first write that initializes the uninitialized storage, all reads are performed to that now-initialized writeable storage, and the corresponding read- only storage is no longer used. 2.1. Data Structures: Extents, Extent Lists, and Volumes A pNFS block layout is a list of extents within a flat array of 512- byte data blocks known as a volume. A volume may correspond to a single logical unit in a SAN, or a more complex aggregation of multiple logical units. The block layout describes both the topology of the volume as well as the individual block extents on the volume that make up the file. Each individual extent MUST be at least 512- byte aligned. Black Expires June 2006 [Page 4] Internet-Draft pNFS Block/Volume Layout December 2005 enum extentState4 { READ_WRITE_DATA = 0, /* the data located by this extent is valid for reading and writing. */ READ_DATA = 1, /* the data located by this extent is valid for reading only; it may not be written. */ INVALID_DATA = 2, /* the location is valid; the data is invalid. It is a newly (pre-) allocated extent. There is physical space on the volume. */ NONE_DATA = 3, /* the location is invalid. It is a hole in the file. There is no physical space on the volume. */ }; struct pnfs_block_extent { offset4 offset; /* the starting offset in the file */ length4 length; /* the size of the extent */ offset4 storage_offset; /* the starting offset in the volume */ extentState4 es; /* the state of this extent */ }; enum pnfs_block_volume_type { VOLUME_SIMPLE = 0, /* volume maps to a single LU */ VOLUME_SLICE = 1, /* volume is a slice of another volume */ VOLUME_CONCAT = 2, /* volume is a concatenation of multiple volumes */ VOLUME_STRIPE = 3, /* volume is striped across multiple volumes */ }; Black Expires June 2006 [Page 5] Internet-Draft pNFS Block/Volume Layout December 2005 struct pnfs_block_slice_volume_info { offset4 start; /* block-offset of the start of the slice */ length4 length; /* length of slice in blocks */ pnfs_block_volume volume; /* volume which is sliced */ }; struct pnfs_block_concat_volume_info { pnfs_block_volume volumes<>; /* volumes which are concatenated */ }; struct pnfs_block_stripe_volume_info { length4 stripe_unit; /* size of stripe */ pnfs_block_volume volumes<>; /* volumes which are striped across*/ }; union pnfs_block_volume switch (pnfs_block_volume_type type) { case VOLUME_SIMPLE: pnfs_deviceid4 volume_ID; case VOLUME_SLICE: pnfs_block_slice_volume_info slice_info; case VOLUME_CONCAT: pnfs_block_concat_volume_info concat_info; case VOLUME_STRIPE: pnfs_block_stripe_volume_info stripe_info; }; Black Expires June 2006 [Page 6] Internet-Draft pNFS Block/Volume Layout December 2005 struct pnfs_block_layout { pnfs_block_volume volume; /* topology of the volume on which file is stored. */ pnfs_block_extent extents<>; /* extents which make up this layout. */ }; The block layout consists of information describing the topology of the logical volume on which the file is stored, followed by a list of extents which map the logical regions of the file to physical locations on the volume. The "pnfs_block_volume" union is a recursive structure that allows arbitrarily complex nested volume structures to be encoded. The types of aggregations that are allowed are stripes, concatenations, and slices. The base case is a volume which maps simply to one logical unit in the SAN, identified by the "pnfs_deviceid4" (see discussion of volume identification, section 2.2 below). The "storage_offset" field within each extent identifies a location on the logical volume described by the "volume" field in the layout. The client is responsible for translating this logical offset into an offset on the appropriate underlying SAN logical unit. Each extent maps a logical region of the file onto a portion of the specified logical volume. The file_offset, extent_length, and es fields for an extent returned from the server are always valid. The interpretation of the storage_offset field depends on the value of es as follows: o READ_WRITE_DATA means that storage_offset is valid, and points to valid/initialized data that can be read and written. o READ_DATA means that storage_offset is valid and points to valid/ initialized data which can only be read. Write operations are prohibited; the client may need to request a read-write layout. o INVALID_DATA means that storage_offset is valid, but points to invalid uninitialized data. This data must not be physically read from the disk until it has been initialized. A read request for an INVALID_DATA extent must fill the user buffer with zeros. Write requests must write whole server-sized blocks to the disk with bytes not initialized by the user must be set to zero. Any write to storage in an INVALID_DATA extent changes the written portion of the extent to READ_WRITE_DATA; the pNFS client is responsible for reporting this change via LAYOUTCOMMIT. Black Expires June 2006 [Page 7] Internet-Draft pNFS Block/Volume Layout December 2005 o NONE_DATA means that storage_offset is not valid, and this extent may not be used to satisfy write requests. Read requests may be satisfied by zero-filling as for INVALID_DATA. NONE_DATA extents are returned by requests for readable extents; they are never returned if the request was for a writeable extent. The extent list lists all relevant extents in increasing order of the file_offset of each extent; any ties are broken by increasing order of the extent state (es). 2.1.1. Layout Requests and Extent Lists Each request for a layout specifies at least three parameters: offset, desired size, and minimum size (the desired size is missing from the operations draft - see Section 3). If the status of a request indicates success, the extent list returned must meet the following criteria: o A request for a readable (but not writeable) layout returns only READ_DATA or NONE_DATA extents (but not INVALID_DATA or READ_WRITE_DATA extents). o A request for a writeable layout returns READ_WRITE_DATA or INVALID_DATA extents (but not NONE_DATA extents). It may also return READ_DATA extents only when the offset ranges in those extents are also covered by INVALID_DATA extents to permit writes. o The first extent in the list MUST contain the starting offset. o The total size of extents in the extent list MUST cover at least the minimum size and no more than the desired size. One exception is allowed: the total size MAY be smaller if only readable extents were requested and EOF is encountered. o Extents in the extent list MUST be logically contiguous for a read-only layout. For a read-write layout, the set of writable extents (i.e., excluding READ_DATA extents) MUST be logically contiguous. Every READ_DATA extent in a read-write layout MUST be covered by an INVALID_DATA extent. This overlap of READ_DATA and INVALID_DATA extents is the only permitted extent overlap. o Extents MUST be ordered in the list by starting offset, with READ_DATA extents preceding INVALID_DATA extents in the case of equal file_offsets. Black Expires June 2006 [Page 8] Internet-Draft pNFS Block/Volume Layout December 2005 2.1.2. Layout Commits struct pnfs_block_layoutupdate { pnfs_block_extent commit_list; /* list of extents to which now contain valid data. */ bool make_version; /* client requests server to create copy-on-write image of this file. */ } The "pnfs_block_layoutupdate" structure is used by the client as the block-protocol specific argument in a LAYOUTCOMMIT operation. The "commit_list" field is an extent list covering regions of the file layout that were previously in the INVALID_DATA state, but have been written by the client and should now be considered in the READ_WRITE_DATA state. It should be noted that the server may be unable to commit regions at a granularity smaller than a file-system block (typically 4KB or 8KB). As noted above, the block-size that the server uses is available as an NFSv4 attribute, and any extents included in the "commit_list" must be aligned on this granularity. If the client believes that its actions have moved the end-of-file into the middle of a block being committed, the client MUST write zeroes from the end-of-file to the end of that block before committing the block. Failure to do so may result in junk (uninitialized data) appearing in that area if the file is subsequently extended by moving the end-of-file. The "make_version" field of the structure is a flag that the client may set to request that the server create a copy-on-write image of the file (see section 2.1.4, below). In anticipation of this operation the client which sets the "make_version" flag in the LAYOUTCOMMIT operation should immediately mark all extents in the layout that is possesses as state READ_DATA. Future writes to the file require a new LAYOUTGET operation to the server with an "iomode" set to LAYOUTIOMODE_RW. 2.1.3. Layout Returns struct pnfs_block_layoutreturn { pnfs_block_extent rel_list; /* list of extents the client will no longer use. */ } Black Expires June 2006 [Page 9] Internet-Draft pNFS Block/Volume Layout December 2005 The "rel_list" field is an extent list covering regions of the file layout that are no longer needed by the client. Including extents in the "rel_list" for a LAYOUTRETURN operation represents an explicit release of resources by the client, usually done for the purpose of avoiding unnecessary CB_LAYOUTRECALL operations in the future. 2.1.4. Client Copy-on-Write Processing Distinguishing the READ_WRITE_DATA and READ_DATA extent types combined with the allowed overlap of READ_DATA extents with INVALID_DATA extents allows copy-on-write processing to be done by pNFS clients. In classic NFS, this operation would be done by the server. Since pNFS enables clients to do direct block access, it is useful for clients to participate in copy-on-write operations. All block/volume pNFS clients MUST support this copy-on-write processing. When a client wishes to write data covered by a READ_DATA extent, it MUST have requested a writable layout from the server; that layout will contain INVALID_DATA extents to cover all the data ranges of that layout's READ_DATA extents. More precisely, for any file_offset range covered by one or more READ_DATA extents in a writable layout, the server MUST include one or more INVALID_DATA extents in the layout that cover the same file_offset range. The client MUST logically copy the data from the READ_DATA extent for any partial blocks of file_offset and range, merge in the changes to be written, and write the result to the INVALID_DATA extent for the blocks for that file_offset and range. That is, if entire blocks of data are to be overwritten by an operation, the corresponding READ_DATA blocks need not be fetched, but any partial-block writes must be merged with data fetched via READ_DATA extents before storing the result via INVALID_DATA extents. For the purposes of this discussion, "entire blocks" and "partial blocks" refer to the server's file-system block size. Storing of data in an INVALID_DATA extent converts the written portion of the INVALID_DATA extent to a READ_WRITE_DATA extent; all subsequent reads MUST be performed from this extent; the corresponding portion of the READ_DATA extent MUST NOT be used after storing data in an INVALID_DATA extent. In the LAYOUTCOMMIT operation that normally sends updated layout information back to the server, for writable data, some INVALID_DATA extents may be committed as READ_WRITE_DATA extents, signifying that the storage at the corresponding storage_offset values has been stored into and is now to be considered as valid data to be read. READ_DATA extents need not be sent to the server. For extents that the client receives via LAYOUTGET as INVALID_DATA and returns via LAYOUTCOMMIT as READ_WRITE_DATA, the server will understand that the Black Expires June 2006 [Page 10] Internet-Draft pNFS Block/Volume Layout December 2005 READ_DATA mapping for that extent is no longer valid or necessary for that file. 2.1.5. Extents are Permissions Layout extents returned to pNFS clients grant permission to read or write; READ_DATA and NONE_DATA are read-only (NONE_DATA reads as zeroes), READ_WRITE_DATA and INVALID_DATA are read/write, (INVALID_DATA reads as zeros, any write converts it to READ_WRITE_DATA). This is the only client means of obtaining permission to perform direct I/O to storage devices; a pNFS client MUST NOT perform direct I/O operations that are not permitted by an extent held by the client. Client adherence to this rule places the pNFS server in control of potentially conflicting storage device operations, enabling the server to determine what does conflict and how to avoid conflicts by granting and recalling extents to/from clients. Block/volume class storage devices are not required to perform read and write operations atomically. Overlapping concurrent read and write operations to the same data may cause the read to return a mixture of before-write and after-write data. Overlapping write operations can be worse, as the result could be a mixture of data from the two write operations; this can be particularly nasty if the underlying storage is striped and the operations complete in different orders on different stripes. A pNFS server can avoid these conflicts by implementing a single writer XOR multiple readers concurrency control policy when there are multiple clients who wish to access the same data. This policy SHOULD be implemented when storage devices do not provide atomicity for concurrent read/write and write/write operations to the same data. A client that makes a layout request that conflicts with an existing layout delegation will be rejected with the error NFS4ERR_LAYOUTTRYLATER. This client is then expected to retry the request after a short interval. During this interval the server needs to recall the conflicting portion of the layout delegation from the client that currently holds it. This reject-and-retry approach does not prevent client starvation when there is contention for the layout of a particular file. For this reason a pNFS server SHOULD implement a mechanism to prevent starvation. One possibility is that the server can maintain a queue of rejected layout requests. Each new layout request can be checked to see if it conflicts with a previous rejected request, and if so, the newer request can be rejected. Once the original requesting client retries its request, its entry in the rejected request queue can be cleared, or the entry Black Expires June 2006 [Page 11] Internet-Draft pNFS Block/Volume Layout December 2005 in the rejected request queue can be removed when it reaches a certain age. NFSv4 supports mandatory locks and share reservations. These are mechanisms that clients can use to restrict the set of I/O operations that are permissible to other clients. Since all I/O operations ultimately arrive at the NFSv4 server for processing, the server is in a position to enforce these restrictions. However, with pNFS layout delegations, I/Os will be issued from the clients that hold the delegations directly to the storage devices that host the data. These devices have no knowledge of files, mandatory locks, or share reservations, and are not in a position to enforce such restrictions. For this reason the NFSv4 server MUST NOT grant layout delegations that conflict with mandatory locks or share reservations. Further, if a conflicting mandatory lock request or a conflicting open request arrives at the server, the server MUST recall the part of the layout delegation in conflict with the request before processing the request. 2.1.6. End-of-file Processing To avoid file-system corruption, close coordination between pNFS clients and the server is required when an NFSv4 client changes the end-of-file marker via the SETATTR call. Whenever the end-of-file is set into the middle of a file-system block, the portion of the block which comes after the end-of-file must be zeroed on disk. The pNFS clients and server share the responsibility for this zeroing as follows: 2.1.6.1. Server End-of-file Processing When an NFSv4 client changes the end-of-file marker via a SETATTR operation, the server MUST send a CB_SIZECHANGED notification to each pNFS client (with the exception of the client which sent the SETATTR) that holds a layout for the file. Clients process this callback as described in Section 2.1.6.2. The CB_SIZECHANGED notification has the effect of invalidating all data beyond the new end-of-file. Once the server receives a successful response to the CB_SIZECHANGED notification from a client, it will consider any portion of any layout held by the client beyond the new end-of-file to be invalid. This requires that: Black Expires June 2006 [Page 12] Internet-Draft pNFS Block/Volume Layout December 2005 o READ_WRITE_DATA extents are changed INVALID_DATA extents. o READ_DATA extents are changed to NONE_DATA extents. If this results in any portion of a NONE_DATA extent overlapping an INVALID_DATA extent, that portion of the NONE_DATA extent (which may be the entire extent) is immediately discarded as INVALID_DATA extents are not permitted to overlap NONE_DATA extents. If the new end-of-file is not set on a file-system block boundary, the server must ensure that zeroes are written to the partial block from the new end-of-file to the end of the block containing it. If a pNFS client holds a writeable layout covering that block, that pNFS client is expected to perform this function, but the CB_SIZECHANGED callback response needs to have a way for the client to communicate that it has done so (see Section 3.2.2). If no client performs this function, the server must do the zeroing, including recalling all layouts (readable and writable) for that block. 2.1.6.2. Client End-of-file Processing When a pNFS client receives the CB_SIZECHANGED notification from the server, or when a pNFS client issues a SETATTR operation to set the end-of-file marker for a file on which it holds a layout, it should proceed the same way. First, it should consider any portions of any readable layout beyond the new end-of-file to be in the NONE_DATA state and it should immediately stop servicing read requests from such extents. If the client holds a readable layout for the block containing a new end-of-file that is not at a block boundary, it SHOULD return at least that block before replying to the callback; this avoids a callback from the server in order to zero the partial block beyond the new end-of-file (how to do this is an open issue - see Section 3.2.3). Next, the client should consider any portions of a writeable layout beyond the new end-of-file to be in the INVALID_DATA state, and should henceforth return zeroes in response to any read requests for these extents. Data can, of course, be rewritten to these extents, turning them back into READ_WRITE_DATA extents, as desired. Finally, if the new end-of-file marker is not on a file-system block boundary, and if the client has a writeable layout which covers the block containing the new end-of-file, then the client MUST zero-fill the portion of the block after the end-of- file marker and write this block to the storage before responding to the CB_SIZECHANGED notification. In this case the new end-of-file block MUST be considered to be in the READ_WRITE_DATA state. Black Expires June 2006 [Page 13] Internet-Draft pNFS Block/Volume Layout December 2005 2.2. Volume Identification Storage Systems such as storage arrays can have multiple physical network ports that need not be connected to a common network, resulting in a pNFS client having simultaneous multipath access to the same storage volumes via different ports on different networks. The networks may not even be the same technology - for example, access to the same volume via both iSCSI and Fibre Channel is possible, hence network address are difficult to use for volume identification. For this reason, this pNFS block layout identifies storage volumes by content, for example providing the means to match (unique portions of) labels used by volume managers. Any block pNFS system using this layout MUST support a means of content-based unique volume identification that can be employed via the data structure given here. struct sigComponent { /* disk signature component */ offset4 sig_offset; /* byte offset of component */ length4 sig_length; /* byte length of component */ opaque contents<>; /* contents of this component of the signature (this is opaque) */ }; struct pnfs_block_deviceaddr4 { sigComponent ds; /* disk signature */ }; A volume is content-identified by a disk signature made up of extents within blocks and contents that must match. The "pnfs_block_deviceaddr4" structure is returned by the server as the storage-protocol-specific opaque field in the "pnfs_deviceaddr4" structure, in response to the GETDEVICEINFO or GETDEVICELIST operations. Note that the opaque "contents" field in the "sigComponent" structure MUST NOT be interpreted as a zero-terminated string, as it may contain embedded zero-valued octets. It contains exactly sig_length octets. There are no restrictions on alignment (e.g., neither sig_offset nor sig_length are required to be multiples of 4). Black Expires June 2006 [Page 14] Internet-Draft pNFS Block/Volume Layout December 2005 3. New Operations Issues This section collects issues in the [NFVV4.1] draft encountered in writing this block/volume layout draft. 3.1. Server Controlling Client Access to Block Devices Typically, SAN disk arrays and SAN protocols provide access control mechanisms (access-logics, lun masking, etc.) which operate at the granularity of individual hosts. The functionality provided by such mechanisms makes it possible for the server to "fence" individual client machines from certain physical disks---that is to say, to prevent individual client machines from reading or writing to certain physical disks. Finer-grained access control methods are not generally available. This affects the ability of the block/volume storage protocol to meet the requirements set out in the [NFSV4.1] draft in the following ways: 3.1.1. Guarantees Provided by Layouts See first paragraph of Security Considerations (Section 4). 3.1.2. I/O In-flight Issues See third paragraph of Security Considerations (Section 4). 3.1.3. Crash Recovery Issues The main [NFSV4.1] draft does not currently include a "reclaim bit" in the LAYOUTGET operation. This means that after a server crash, during the server's recovery grace period, a client cannot be guaranteed that it will be able to reclaim a writable layout that it held before the crash (due to possibly conflicting requests from other clients), nor can the client be guaranteed that a writeable layout that it reclaims will have the same on-disk layout as the one it held prior to the crash. There is one possible scenario where this lack of guarantees could lead to data corruption. The problematic case is when the server crashes while the client holds a writable layout, but has yet to commit some allocated block extents back to the server with a LAYOUTCOMMIT operation. If the data that has been written to these extents is still cached by the client, the client can simply re-write the data via NFSv4, once the server has come back online. However, the uncommitted extents may represent more data than can fit in the client's cache at one time. In this case, the data cannot be rewritten to the server---indeed, it cannot even be safely re-read from storage in the absence of a pNFS Black Expires June 2006 [Page 15] Internet-Draft pNFS Block/Volume Layout December 2005 layout which covers those disk blocks. The ideal solution, from the client's perspective, would be a method that allows the client to reliably reclaim the previously held layout from the server. For this, the client would need a "reclaim bit" in the LAYOUTGET operation to allow the server to recognize, and give priority to, a reclaim request during the server's recovery grace period, as well as a way of supplying an extent list indicating the blocks (including storage location) that were previously held in the writable layout. Without this support, the server would need to make sure that provisionally allocated extents that are supplied in response to writable LAYOUTGET commands are persistently stored and recoverable in the face of server crashes. Otherwise, the client would have no guarantee that the new writable layout it receives in response to a reclaim request covers the same disk blocks as the old layout it held before the server crash. The "reclaim bit" could be implemented as another iomode (e.g., LAYOUTIOMODE_RW_RECOVER). If this is not done, the result may be that the amount of written-but-uncomitted data held by a client may need to be limited to the client's cache size, resulting in less effective cache usage and more commit traffic. 3.2. End-of-File (EOF) Handling issues The main pNFS draft [NFSV4.1] includes a callback (CB_SIZECHANGED) for a pNFS server to communicate a file size change (new EOF) to pNFS clients. The callback documentation indicates that "the client should update its internal size for the file". This is insufficient in a number of ways described below. Some of these issues are unique to the block/ volume layout as this layout requires that either the client or server zero beyond the new EOF to the end of the block that contains the new EOF (both the file and object layouts can make this the responsibility of the storage server(s)). 3.2.1. Truncation There is a race between a pNFS client extending a file via direct pNFS writes and an ordinary NFS client truncating the file via a SETATTR - the order of these operations may be visible to the application(s) using NFS, and hence it is necessary that they be done in the right order. List discussion suggests that the absence of operation sequencing in pNFS (in contrast to HighRoad) results in the best approach being to require a recall of the portions of the layout that are invalidated by a truncation so that the server can perform the truncation (this MUST include the entire layout beyond the new EOF). For the block/volume layout, the server is responsible for zeroing the partial block beyond the new EOF in this case. Black Expires June 2006 [Page 16] Internet-Draft pNFS Block/Volume Layout December 2005 3.2.2. Extension When extending a file via SETATTR, a similar race can arise between that SETATTR and direct pNFS writes. When CB_SIZECHANGED is used to pass the new end-of-file to a client that holds a writable layout for the block containing the new end-of-file, and that new end-of-file is not at a filesystem block boundary, it is desirable for the client holding the writable layout to take responsibility for zeroing the partial block beyond end-of file. Note that the client may not be able to do so in all cases, as the client may have been in the process of returning that portion of the layout when the callback arrives, or the server and client may not agree on whether the client holds a writable layout containing that block. Hence the CB_SIZECHANGED reply needs to carry an indication from the client as to whether the client has zeroed that partial block. Open Issue: How to convey that indication. The least intrusive way to do this may be to define an NFS Error that indicates that the callback was successful and the client zeroed the resulting partial block. Open Issue: Can callback reply be delayed until this write is done? Immediate callback reply with promise to zero the partial block followed by client crash results in potential data corruption due to failure to zero the partial block (junk appears if the file is extended by moving end-of-file forward). 3.2.3. Extension vs. Readable Layouts There is an analogous issue for a client holding a readable layout (READ_DATA) including the block that contains a new non-block-aligned end-of-file. In this case, the client cannot zero the partial block itself, as it does not hold a writable layout, but if it retains the readable layout, the server is likely to immediately issue a recall callback to revoke the readable layout before the server writes the zeroes. This second callback should be avoided. Issue: How? Allowing an arbitrary return in the response to CB_SIZECHANGED would get the job done, but is only needed by the block/volume layout, and complicates the end-of-file change code on the server by potentially returning unrelated areas of the layout. Delaying the callback reply to allow the client to do a RETURN helps some (2 round-trips instead of 3 if the recall were used) at the cost of delaying the recall reply. Best bet may be to enlarge the CB_SIZECHANGED callback so the server can tell the client what needs to be returned (in the readable layout case, the server asks for one block) and for the client to accept/reject that (client accepts if it Black Expires June 2006 [Page 17] Internet-Draft pNFS Block/Volume Layout December 2005 can do so immediately - it would reject if it needed to do a commit); that should yield one round-trip with the server getting exactly what it needs back (and server MUST NOT use this to recall a writable layout). 4. Security Considerations Certain security responsibilities are delegated to pNFS clients. Block/volume storage systems generally control access at a volume granularity, and hence pNFS clients have to be trusted to only perform accesses allowed by the layout extents they currently hold (e.g., and not access storage for files on which a layout extent is not held). In general, the server will not be able to prevent a client which holds a layout for a file from accessing parts of the physical disk not covered by the layout. Similarly, the server will not be able to prevent a client from accessing blocks covered by a layout that it has already returned. This block-based level of protection must be provided by the client software. In environments where the security requirements are such that client-side protection from access to storage outside of the layout is not sufficient, pNFS block/volume storage layouts for pNFS SHOULD NOT be used. This also has implications for some NFSv4 functionality outside pNFS. For instance, if a file is covered by a mandatory read-only lock, the server can ensure that only readable layouts for the file are granted to pNFS clients. However, it is up to each pNFS client to ensure that the readable layout is used only to service read requests, and not to allow writes to the existing parts of the file. Since block/volume storage systems are generally not capable of enforcing such file-based security, in environments where pNFS clients cannot be trusted to enforce such policies, pNFS block/volume storage layouts SHOULD NOT be used. When a layout lease timer expires, it is important for the sake of correctness that any in-flight I/Os that the client issued before the expiration of the timer are rejected at the storage. For the block/volume protocol, this is possible by fencing a client with an expired layout timer from the physical storage. Note, however, that the granularity of this operation can only be at the host/logical- unit level. Thus, if a client's lease timer expires for a single layout, it will effectively render useless *all* of the clients layouts for files in the containing filesystem. Access to block/volume storage is logically at a lower layer of the I/O stack than NFSv4, and hence NFSv4 security is not directly applicable to protocols that access such storage directly. Depending on the protocol, some of the security mechanisms provided by NFSv4 Black Expires June 2006 [Page 18] Internet-Draft pNFS Block/Volume Layout December 2005 (e.g., encryption, cryptographic integrity) may not be available, or may be provided via different means. At one extreme, pNFS with block/volume storage can be used with storage access protocols (e.g., parallel SCSI) that provide essentially no security functionality. At the other extreme, pNFS may be used with storage protocols such as iSCSI that provide significant functionality. It is the responsibility of those administering and deploying pNFS with a block/volume storage access protocol to ensure that appropriate protection is provided to that protocol (physical security is a common means for protocols not based on IP). In environments where the security requirements for the storage protocol cannot be met, pNFS block/volume storage layouts SHOULD NOT be used. When security is available for a storage protocol, it is generally at a different granularity and with a different notion of identity than NFSv4 (e.g., NFSv4 controls user access to files, iSCSI controls initiator access to volumes). The responsibility for enforcing appropriate correspondences between these security layers is placed upon the pNFS client. As with the issues in the first paragraph of this section, in environments where the security requirements are such that client-side protection from access to storage outside of the layout is not sufficient, pNFS block/volume storage layouts SHOULD NOT be used. 5. Conclusions This draft specifies the block/volume layout type for pNFS and associated functionality. 6. IANA Considerations There are no IANA considerations in this document. All pNFS IANA Considerations are covered in [NFSV4.1]. 7. Revision History -00: Initial Version as draft-black-pnfs-block-00 -01: Rework discussion of extents as locks to talk about extents granting access permissions. Rewrite operation ordering section to discuss deadlocks and races that can cause problems. Add new section on recall completion. Add client copy-on-write based on text from Craig Everhart. -02: Fix glitches in extent state descriptions. Describe most issues as RESOLVED. Most of Section 3 has been incorporated into the the main PNFD draft, add NOTE to that effect and say that it will be Black Expires June 2006 [Page 19] Internet-Draft pNFS Block/Volume Layout December 2005 deleted in the next version of this draft (which should be a draft- ietf-nfsv4 draft). Cleaning up a number of things have been left to that draft revision, including the interlocks with the types in the main pNFS draft, layout striping support, and finishing the Security Considerations section. -00: New version as draft-ietf-nfsv4-pnfs-block. Removed resolved operations issues (Section 3). Align types with main pNFS draft (which is now part of the NFSv4.1 minor version draft), add volume striping and slicing support. New operations issues are in Section 3 - the need for a "reclaim bit" and EOF concerns are the two major issues. Extended and improved the Security Considerations section, but it still needs work. Added 1-sentence conclusion that also still needs work. 8. Acknowledgments This draft draws extensively on the authors' familiarity with the mapping functionality and protocol in EMC's HighRoad system [HighRoad]. The protocol used by HighRoad is called FMP (File Mapping Protocol); it is an add-on protocol that runs in parallel with filesystem protocols such as NFSv3 to provide pNFS-like functionality for block/volume storage. While drawing on HighRoad FMP, the data structures and functional considerations in this draft differ in significant ways, based on lessons learned and the opportunity to take advantage of NFSv4 features such as COMPOUND operations. The design to support pNFS client participation in copy- on-write is based on text and ideas contributed by Craig Everhart (formerly with IBM). 9. References 9.1. Normative References [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997. [NFSV4.1] Shepler, S., ed., "NFSv4 Minor Version 1", draft-ietf- nfsv4-minorversion1-01.txt, Work in Progress, December 2005. 9.2. Informative References [HighRoad] EMC Corporation, "EMC Celerra HighRoad", EMC C819.1 white paper, available at: http://www.emc.com/pdf/products/celerra_file_server/HighRoad_wp.pdf link checked 30 December 2005. Black Expires June 2006 [Page 20] Internet-Draft pNFS Block/Volume Layout December 2005 Author's Addresses David L. Black EMC Corporation 176 South Street Hopkinton, MA 01748 Phone: +1 (508) 293-7953 Email: black_david@emc.com Stephen Fridella EMC Corporation 32 Coslin Drive Southboro, MA 01772 Phone: +1 (508) 305-8512 Email: fridella_stephen@emc.com Intellectual Property Statement The IETF takes no position regarding the validity or scope of any Intellectual Property Rights or other rights that might be claimed to pertain to the implementation or use of the technology described in this document or the extent to which any license under such rights might or might not be available; nor does it represent that it has made any independent effort to identify any such rights. Information on the procedures with respect to rights in RFC documents can be found in BCP 78 and BCP 79. Copies of IPR disclosures made to the IETF Secretariat and any assurances of licenses to be made available, or the result of an attempt made to obtain a general license or permission for the use of such proprietary rights by implementers or users of this specification can be obtained from the IETF on-line IPR repository at http://www.ietf.org/ipr. The IETF invites any interested party to bring to its attention any copyrights, patents or patent applications, or other proprietary rights that may cover technology that may be required to implement this standard. Please address the information to the IETF at ietf- ipr@ietf.org. Black Expires June 2006 [Page 21] Internet-Draft pNFS Block/Volume Layout December 2005 Disclaimer of Validity This document and the information contained herein are provided on an "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Copyright Statement Copyright (C) The Internet Society (2005). This document is subject to the rights, licenses and restrictions contained in BCP 78, and except as set forth therein, the authors retain all their rights. Acknowledgment Funding for the RFC Editor function is currently provided by the Internet Society. Black Expires June 2006 [Page 22]