Parallel NFS (pNFS) SCSI Layout

The fundamental storage abstraction supported by SCSI storage devices is a Logical Unit (LU) 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. A pNFS layout for this SCSI 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 writable extents MUST be aligned to the block size used by the NFSv4 server in managing the actual file system (4 kilobytes and 8 kilobytes are common block sizes). This block size is available as the NFSv4.1 layout_blksize attribute. . Readable extents SHOULD be aligned to the block size used by the NFSv4 server, but in order to support legacy file systems with fragments, alignment to 512-byte boundaries is acceptable. The pNFS operation for requesting a layout (LAYOUTGET) includes the "layoutiomode4 loga_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 un-initialized storage in those holes (read as zero, can be written by client). This document also supports client participation in copy-on-write (e.g., for file systems with snapshots) by providing both read-only and un- initialized storage for the same range in a layout. Reads are initially performed on the read-only storage, with writes going to the un-initialized storage. After the first write that initializes the un-initialized storage, all reads are performed to that now- initialized writable storage, and the corresponding read-only storage is no longer used. The SCSI layout solution expands the security responsibilities of the pNFS clients, and there are a number of environments where the mandatory to implement security properties for NFS cannot be satisfied. The additional security responsibilities of the client follow, and a full discussion is present im , "Security Considerations". Typically, SCSI storage devices provide access control mechanisms (e.g., Logical Unit Number (LUN) mapping and/or masking), which operate at the granularity of individual hosts, not individual blocks. For this reason, block-based protection must be provided by the client software. Similarly, SCSI storage devices typically are not able to validate NFS locks that apply to file regions. 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 SCSI storage devices are generally not capable of enforcing such file-based security, in environments where pNFS clients cannot be trusted to enforce such policies, pNFS SCSI layouts SHOULD NOT be used.

The layout4 type defined in is extended with a new value as follows:

enum layouttype4 { LAYOUT4_NFSV4_1_FILES = 1, LAYOUT4_OSD2_OBJECTS = 2, LAYOUT4_BLOCK_VOLUME = 3, LAYOUT4_SCSI = 0x80000005 [[RFC Editor: please modify the LAYOUT4_SCSI to be the layouttype assigned by IANA]] }; This document defines structure associated with the layouttype4 value LAYOUT4_SCSI. specifies the loc_body structure as an XDR type "opaque". The opaque layout is uninterpreted by the generic pNFS client layers, but obviously must be interpreted by the Layout Type implementation.

SCSI targets implementing export unique LU names for each LU through the Device Identification VPD page, which can be obtained using the INQUIRY command. This document uses a subset of this information to identify LUs backing pNFS SCSI layouts. It is similar to the "Identification Descriptor Target Descriptor" specified in , but limits the allowed values to those that uniquely identify a LU. Device Identification VPD page descriptors used to identify LUs for use with pNFS SCSI layouts must adhere to the following restrictions: The "ASSOCIATION" must be set to 0 (The DESIGNATOR field is associated with the addressed logical unit). The "DESIGNATOR TYPE" must be set to one of three values explicitly listed in the "pnfs_scsi_designator_type" enumerations. The "CODE SET" VPD page field is stored in the "sbv_code_set" field of the "pnfs_scsi_base_volume_info4" structure, the "DESIGNATOR TYPE" is stored in "sbv_designator_type", and the DESIGNATOR is stored in "sbv_designator". Due to the use of a XDR array the "DESIGNATOR LENGTH" field does not need to be set separately. Only certain combinations of "sbv_code_set" and "sbv_designator_type" are valid, please refer to for details, and note that ASCII may be used as the code set for UTF-8 text that contains only ASCII characters. Note that a Device Identification VPD page MAY contain multiple descriptors with the same association, code set and designator type. NFS clients thus MUST iterate the descriptors until a match for "sbv_code_set", "sbv_designator_type" and "sbv_designator" is found, or until the end of VPD page. Storage devices 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. Selection of one or multiple ports to access the storage device is left up to the client. Additionally the server returns a Persistent Reservation key in the "sbv_pr_key" field. See for more details on the use of Persistent Reservations.

The pNFS SCSI layout volume topology is expressed as an arbitrary combination of base volume types enumerated in the following data structures. The individual components of the topology are contained in an array and components may refer to other components by using array indices.

/// enum pnfs_scsi_volume_type4 { /// PNFS_SCSI_VOLUME_SLICE = 1, /* volume is a slice of /// another volume */ /// PNFS_SCSI_VOLUME_CONCAT = 2, /* volume is a /// concatenation of /// multiple volumes */ /// PNFS_SCSI_VOLUME_STRIPE = 3 /* volume is striped across /// multiple volumes */ /// PNFS_SCSI_VOLUME_BASE = 4, /* volume maps to a single /// LU */ /// }; ///

/// /* /// * Code sets from SPC-3. /// */ /// enum pnfs_scsi_code_set { /// PS_CODE_SET_BINARY = 1, /// PS_CODE_SET_ASCII = 2, /// PS_CODE_SET_UTF8 = 3 /// }; /// /// /* /// * Designator types from taken from SPC-3. /// * /// * Other values are allocated in SPC-3, but not mandatory to /// * implement or aren't Logical Unit names. /// */ /// enum pnfs_scsi_designator_type { /// PS_DESIGNATOR_EUI64 = 2, /// PS_DESIGNATOR_NAA = 3, /// PS_DESIGNATOR_NAME = 8 /// }; /// /// /* /// * Logical Unit name + reservation key. /// */ /// struct pnfs_scsi_base_volume_info4 { /// pnfs_scsi_code_set sbv_code_set; /// pnfs_scsi_designator_type sbv_designator_type; /// opaque sbv_designator<>; /// uint64_t sbv_pr_key; /// }; ///

/// /// struct pnfs_scsi_slice_volume_info4 { /// offset4 ssv_start; /* offset of the start of the /// slice in bytes */ /// length4 ssv_length; /* length of slice in bytes */ /// uint32_t ssv_volume; /* array index of sliced /// volume */ /// };

/// /// struct pnfs_scsi_concat_volume_info4 { /// uint32_t scv_volumes<>; /* array indices of volumes /// which are concatenated */ /// };

/// /// struct pnfs_scsi_stripe_volume_info4 { /// length4 ssv_stripe_unit; /* size of stripe in bytes */ /// uint32_t ssv_volumes<>; /* array indices of volumes /// which are striped across -- /// MUST be same size */ /// };

/// /// union pnfs_scsi_volume4 switch (pnfs_scsi_volume_type4 type) { /// case PNFS_SCSI_VOLUME_BASE: /// pnfs_scsi_base_volume_info4 sv_simple_info; /// case PNFS_SCSI_VOLUME_SLICE: /// pnfs_scsi_slice_volume_info4 sv_slice_info; /// case PNFS_SCSI_VOLUME_CONCAT: /// pnfs_scsi_concat_volume_info4 sv_concat_info; /// case PNFS_SCSI_VOLUME_STRIPE: /// pnfs_scsi_stripe_volume_info4 sv_stripe_info; /// }; ///

/// /* SCSI layout specific type for da_addr_body */ /// struct pnfs_scsi_deviceaddr4 { /// pnfs_scsi_volume4 sda_volumes<>; /* array of volumes */ /// }; /// The "pnfs_scsi_deviceaddr4" data structure is a structure that allows arbitrarily complex nested volume structures to be encoded. The types of aggregations that are allowed are stripes, concatenations, and slices. Note that the volume topology expressed in the pnfs_scsi_deviceaddr4 data structure will always resolve to a set of pnfs_scsi_volume_type4 PNFS_SCSI_VOLUME_BASE. The array of volumes is ordered such that the root of the volume hierarchy is the last element of the array. Concat, slice, and stripe volumes MUST refer to volumes defined by lower indexed elements of the array. The "pnfs_scsi_device_addr4" data structure is returned by the server as the storage-protocol-specific opaque field da_addr_body in the "device_addr4" structure by a successful GETDEVICEINFO operation . As noted above, all device_addr4 structures eventually resolve to a set of volumes of type PNFS_SCSI_VOLUME_BASE. Complicated volume hierarchies may be composed of dozens of volumes each with several signature components; thus, the device address may require several kilobytes. The client SHOULD be prepared to allocate a large buffer to contain the result. In the case of the server returning NFS4ERR_TOOSMALL, the client SHOULD allocate a buffer of at least gdir_mincount_bytes to contain the expected result and retry the GETDEVICEINFO request.

A pNFS SCSI layout is a list of extents within a flat array of data blocks in a logical volume. The details of the volume topology can be determined by using the GETDEVICEINFO operation. The SCSI layout describes the individual block extents on the volume that make up the file. The offsets and length contained in an extent are specified in units of bytes.

/// enum pnfs_scsi_extent_state4 { /// PNFS_SCSI_READ_WRITE_DATA = 0, /* the data located by this /// extent is valid /// for reading and writing. */ /// PNFS_SCSI_READ_DATA = 1, /* the data located by this /// extent is valid for reading /// only; it may not be /// written. */ /// PNFS_SCSI_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. */ /// PNFS_SCSI_NONE_DATA = 3 /* the location is invalid. /// It is a hole in the file. /// There is no physical space /// on the volume. */ /// };

/// /// struct pnfs_scsi_extent4 { /// deviceid4 se_vol_id; /* id of logical volume on /// which extent of file is /// stored. */ /// offset4 se_file_offset; /* starting byte offset /// in the file */ /// length4 se_length; /* size in bytes of the /// extent */ /// offset4 se_storage_offset; /* starting byte offset /// in the volume */ /// pnfs_scsi_extent_state4 se_state; /// /* state of this extent */ /// };

/// /// /* SCSI layout specific type for loc_body */ /// struct pnfs_scsi_layout4 { /// pnfs_scsi_extent4 sl_extents<>; /// /* extents which make up this /// layout. */ /// }; /// The SCSI layout consists of a list of extents that map the logical regions of the file to physical locations on a volume. The "se_storage_offset" field within each extent identifies a location on the logical volume specified by the "se_vol_id" field in the extent. The se_vol_id itself is shorthand for the whole topology of the logical volume on which the file is stored. The client is responsible for translating this logical offset into an offset on the appropriate underlying SCSI LU. In most cases, all extents in a layout will reside on the same volume and thus have the same se_vol_id. In the case of copy-on-write file systems, the PNFS_SCSI_READ_DATA extents may have a different se_vol_id from the writable extents. Each extent maps a logical region of the file onto a portion of the specified LU. The se_file_offset, se_length, and se_state fields for an extent returned from the server are valid for all extents. In contrast, the interpretation of the se_storage_offset field depends on the value of se_state as follows (in increasing order): means that se_storage_offset is valid, and points to valid/initialized data that can be read and written. means that se_storage_offset is valid andpoints to valid/initialized data that can only be read. Write operations are prohibited; the client may need to request a read-write layout. means that se_storage_offset is valid, but points to invalid un-initialized data. This data must not be physically read from the disk until it has been initialized. A read request for a PNFS_SCSI_INVALID_DATA extent must fill the user buffer with zeros, unless the extent is covered by a PNFS_SCSI_READ_DATA extent of a copy-on-write file system. Write requests must write whole server-sized blocks to the disk; bytes not initialized by the user must be set to zero. Any write to storage in a PNFS_SCSI_INVALID_DATA extent changes the written portion of the extent to PNFS_SCSI_READ_WRITE_DATA; the pNFS client is responsible for reporting this change via LAYOUTCOMMIT. means that se_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 PNFS_SCSI_INVALID_DATA. PNFS_SCSI_NONE_DATA extents may be returned by requests for readable extents; they are never returned if the request was for a writable extent. An extent list contains all relevant extents in increasing order of the se_file_offset of each extent; any ties are broken by increasing order of the extent state (se_state).

Each request for a layout specifies at least three parameters: file offset, desired size, and minimum size. If the status of a request indicates success, the extent list returned must meet the following criteria: A request for a readable (but not writable) layout returns only PNFS_SCSI_READ_DATA or PNFS_SCSI_NONE_DATA extents (but not PNFS_SCSI_INVALID_DATA or PNFS_SCSI_READ_WRITE_DATA extents). A request for a writable layout returns PNFS_SCSI_READ_WRITE_DATA or PNFS_SCSI_INVALID_DATA extents (but not PNFS_SCSI_NONE_DATA extents). It may also return PNFS_SCSI_READ_DATA extents only when the offset ranges in those extents are also covered by PNFS_SCSI_INVALID_DATA extents to permit writes. The first extent in the list MUST contain the requested starting offset. The total size of extents within the requested range MUST cover at least the minimum size. One exception is allowed: the total size MAY be smaller if only readable extents were requested and EOF is encountered. 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 PNFS_SCSI_READ_DATA extents) MUST be logically contiguous. Every PNFS_SCSI_READ_DATA extent in a read-write layout MUST be covered by one or more PNFS_SCSI_INVALID_DATA extents. This overlap of PNFS_SCSI_READ_DATA and PNFS_SCSI_INVALID_DATA extents is the only permitted extent overlap. Extents MUST be ordered in the list by starting offset, with PNFS_SCSI_READ_DATA extents preceding PNFS_SCSI_INVALID_DATA extents in the case of equal se_file_offsets. If the minimum requested size, loga_minlength, is zero, this is an indication to the metadata server that the client desires any layout at offset loga_offset or less that the metadata server has "readily available". Readily is subjective, and depends on the layout type and the pNFS server implementation. For SCSI layout servers, readily available SHOULD be interpreted such that readable layouts are always available, even if some extents are in the PNFS_SCSI_NONE_DATA state. When processing requests for writable layouts, a layout is readily available if extents can be returned in the PNFS_SCSI_READ_WRITE_DATA state.

/// /// /* SCSI layout specific type for lou_body */ /// /// struct pnfs_scsi_range4 { /// offset4 sr_file_offset; /* starting byte offset /// in the file */ /// length4 sr_length; /* size in bytes */ /// }; /// /// struct pnfs_scsi_layoutupdate4 { /// pnfs_scsi_range4 slu_commit_list<>; /// /* list of extents which /// * now contain valid data. /// */ /// }; The "pnfs_scsi_layoutupdate4" structure is used by the client as the SCSI layout specific argument in a LAYOUTCOMMIT operation. The "slu_commit_list" field is a list covering regions of the file layout that were previously in the PNFS_SCSI_INVALID_DATA state, but have been written by the client and should now be considered in the PNFS_SCSI_READ_WRITE_DATA state. The extents in the commit list MUST be disjoint and MUST be sorted by sr_file_offset. Implementors should be aware that a server may be unable to commit regions at a granularity smaller than a file-system block (typically 4 KB or 8 KB). As noted above, the block-size that the server uses is available as an NFSv4 attribute, and any extents included in the "slu_commit_list" MUST be aligned to this granularity and have a size that is a multiple of 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 (un-initialized data) appearing in that area if the file is subsequently extended by moving the end-of-file.

The LAYOUTRETURN operation is done without any SCSI layout specific data. When the LAYOUTRETURN operation specifies a LAYOUTRETURN4_FILE_return type, then the layoutreturn_file4 data structure specifies the region of the file layout that is no longer needed by the client. The opaque "lrf_body" field of the "layoutreturn_file4" data structure MUST have length zero. 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. The client may return disjoint regions of the file by using multiple LAYOUTRETURN operations within a single COMPOUND operation. Note that the SCSI layout supports unilateral layout revocation. When a layout is unilaterally revoked by the server, usually due to the client's lease time expiring, or a delegation being recalled, or the client failing to return a layout in a timely manner, it is important for the sake of correctness that any in-flight I/Os that the client issued before the layout was revoked are rejected at the storage. For the SCSI 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/LU level. Thus, if one of a client's layouts is unilaterally revoked by the server, it will effectively render useless *all* of the client's layouts for files located on the storage units comprising the logical volume. This may render useless the client's layouts for files in other file systems.

Copy-on-write is a mechanism used to support file and/or file system snapshots. When writing to unaligned regions, or to regions smaller than a file system block, the writer must copy the portions of the original file data to a new location on disk. This behavior can either be implemented on the client or the server. The paragraphs below describe how a pNFS SCSI layout client implements access to a file that requires copy-on-write semantics. Distinguishing the PNFS_SCSI_READ_WRITE_DATA and PNFS_SCSI_READ_DATA extent types in combination with the allowed overlap of PNFS_SCSI_READ_DATA extents with PNFS_SCSI_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 SCSI pNFS clients MUST support this copy-on-write processing. When a client wishes to write data covered by a PNFS_SCSI_READ_DATA extent, it MUST have requested a writable layout from the server; that layout will contain PNFS_SCSI_INVALID_DATA extents to cover all the data ranges of that layout's PNFS_SCSI_READ_DATA extents. More precisely, for any se_file_offset range covered by one or more PNFS_SCSI_READ_DATA extents in a writable layout, the server MUST include one or more PNFS_SCSI_INVALID_DATA extents in the layout that cover the same se_file_offset range. When performing a write to such an area of a layout, the client MUST effectively copy the data from the PNFS_SCSI_READ_DATA extent for any partial blocks of se_file_offset and range, merge in the changes to be written, and write the result to the PNFS_SCSI_INVALID_DATA extent for the blocks for that se_file_offset and range. That is, if entire blocks of data are to be overwritten by an operation, the corresponding PNFS_SCSI_READ_DATA blocks need not be fetched, but any partial- block writes must be merged with data fetched via PNFS_SCSI_READ_DATA extents before storing the result via PNFS_SCSI_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 a PNFS_SCSI_INVALID_DATA extent converts the written portion of the PNFS_SCSI_INVALID_DATA extent to a PNFS_SCSI_READ_WRITE_DATA extent; all subsequent reads MUST be performed from this extent; the corresponding portion of the PNFS_SCSI_READ_DATA extent MUST NOT be used after storing data in a PNFS_SCSI_INVALID_DATA extent. If a client writes only a portion of an extent, the extent may be split at block aligned boundaries. When a client wishes to write data to a PNFS_SCSI_INVALID_DATA extent that is not covered by a PNFS_SCSI_READ_DATA extent, it MUST treat this write identically to a write to a file not involved with copy-on-write semantics. Thus, data must be written in at least block-sized increments, aligned to multiples of block-sized offsets, and unwritten portions of blocks must be zero filled.

Layout extents returned to pNFS clients grant permission to read or write; PNFS_SCSI_READ_DATA and PNFS_SCSI_NONE_DATA are read-only (PNFS_SCSI_NONE_DATA reads as zeroes), PNFS_SCSI_READ_WRITE_DATA and PNFS_SCSI_INVALID_DATA are read/write, (PNFS_SCSI_INVALID_DATA reads as zeros, any write converts it to PNFS_SCSI_READ_WRITE_DATA). This is the only means a client has 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. SCSI storage devices do not provide byte granularity access and can only perform read and write operations atomically on a block granularity, and thus require read-modify-write cycles to write data smaller than the block size. Overlapping concurrent read and write operations to the same data thus will cause the read to return a mixture of before-write and after-write data. Additionally, data corruption can occur if the underlying storage is striped and the operations complete in different orders on different stripes. When there are multiple clients who wish to access the same data, a pNFS server MUST avoid these conflicts by implementing a concurrency control policy of single writer XOR multiple readers for a given data region. If a client makes a layout request that conflicts with an existing layout delegation, the request 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 SHOULD 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 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 layouts, I/Os will be issued from the clients that hold the layouts 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 layouts 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 in conflict with the request before granting the request.

The end-of-file location can be changed in two ways: implicitly as the result of a WRITE or LAYOUTCOMMIT beyond the current end-of-file, or explicitly as the result of a SETATTR request. Typically, when a file is truncated by an NFSv4 client via the SETATTR call, the server frees any disk blocks belonging to the file that are beyond the new end-of-file byte, and MUST write zeros to the portion of the new end-of-file block beyond the new end-of-file byte. These actions render any pNFS layouts that refer to the blocks that are freed or written semantically invalid. Therefore, the server MUST recall from clients the portions of any pNFS layouts that refer to blocks that will be freed or written by the server before processing the truncate request. These recalls may take time to complete; as explained in , if the server cannot respond to the client SETATTR request in a reasonable amount of time, it SHOULD reply to the client with the error NFS4ERR_DELAY. Blocks in the PNFS_SCSI_INVALID_DATA state that lie beyond the new end-of-file block present a special case. The server has reserved these blocks for use by a pNFS client with a writable layout for the file, but the client has yet to commit the blocks, and they are not yet a part of the file mapping on disk. The server MAY free these blocks while processing the SETATTR request. If so, the server MUST recall any layouts from pNFS clients that refer to the blocks before processing the truncate. If the server does not free the PNFS_SCSI_INVALID_DATA blocks while processing the SETATTR request, it need not recall layouts that refer only to the PNFS_SCSI_INVALID_DATA blocks. When a file is extended implicitly by a WRITE or LAYOUTCOMMIT beyond the current end-of-file, or extended explicitly by a SETATTR request, the server need not recall any portions of any pNFS layouts.

The SETATTR operation supports a layout hint attribute . Clients MUST NOT set a layout hint with a layout type (the loh_type field) of LAYOUT4_SCSI_VOLUME.

The pNFS SCSI protocol must handle situations in which a system failure, typically a network connectivity issue, requires the server to unilaterally revoke extents from one client in order to transfer the extents to another client. The pNFS server implementation MUST ensure that when resources are transferred to another client, they are not used by the client originally owning them, and this must be ensured against any possible combination of partitions and delays among all of the participants to the protocol (server, storage and client). The pNFS SCSI protocol implements fencing using Persistent Reservations (PRs), similar to the fencing method used by existing shared disk file systems. By placing a PR of type "Exclusive Access – All Registrants" on each SCSI LU exported to pNFS clients the MDS prevents access from any client that does not have an outstanding device device ID that gives the client a reservation key to access the LU, and allows the MDS to revoke access to the logic unit at any time.

To allow fencing individual systems, each system must use a unique Persistent Reservation key. does not specify a way to generate keys. This document assigns the burden to generate unique keys to the MDS, which must generate a key for itself before exporting a volume, and one for each client that accesses a volume. The MDS MAY either generate a key for each client that accesses logic units exported by the MDS, or generate a key for each [LU, client] combination. If using a single key per client, the MDS needs to be aware of the per-client fencing granularity.

Before returning a PNFS_SCSI_VOLUME_BASE volume to the client, the MDS needs to prepare the volume for fencing using PRs. This is done by registering the reservation generated for the MDS with the device using the "PERSISTENT RESERVE OUT" command with a service action of "REGISTER", followed by a "PERSISTENT RESERVE OUT" command, with a service action of "RESERVE" and the type field set to 8h (Exclusive Access – All Registrants). To make sure all I_T nexuses are registered, the MDS SHOULD set the "All Target Ports" (ALL_TG_PT) bit when registering the key, or otherwise ensure the registration is performed for each initiator port.

Before performing the first IO to a device returned from a GETDEVICEINFO operation the client will register the registration key returned in sbv_pr_key with the storage device by issuing a "PERSISTENT RESERVE OUT" command with a service action of REGISTER with the "SERVICE ACTION RESERVATION KEY" set to the reservation key returned in sbv_pr_key. To make sure all I_T nexus are registered, the client SHOULD set the "All Target Ports" (ALL_TG_PT) bit when registering the key, or otherwise ensure the registration is performed for each initiator port. When a client stops using a device earlier returned by GETDEVICEINFO it MUST unregister the earlier registered key by issuing a "PERSISTENT RESERVE OUT" command with a service action of "REGISTER" with the "RESERVATION KEY" set to the earlier registered reservation key.

In case of a non-responding client the MDS MUST fence the client by issuing a "PERSISTENT RESERVE OUT" command with the service action set to "PREEMPT" or "PREEMPT AND ABORT", the reservation key field set to the server's reservation key, the service action reservation key field set to the reservation key associated with the non-responding client, and the type field set to 8h (Exclusive Access – All Registrants). After the MDS preempts a client, all client I/O to the LU fails. The client should at this point return any layout that refers to the device ID that points to the LU. Note that the client can distinguish I/O errors due to fencing from other errors based on the "RESERVATION CONFLICT" status. Refer to for details.

A client that detects I/O errors on the storage devices MUST commit through the MDS, return all outstanding layouts for the device, forget the device ID and unregister the reservation key. Future GETDEVICEINFO calls may refer to the storage device again, in which case a new registration will be performed.

A critical requirement in crash recovery is that both the client and the server know when the other has failed. Additionally, it is required that a client sees a consistent view of data across server restarts. These requirements and a full discussion of crash recovery issues are covered in the "Crash Recovery" section of the NFSv41 specification . This document contains additional crash recovery material specific only to the SCSI layout. When the server crashes while the client holds a writable layout, and the client has written data to blocks covered by the layout, and the blocks are still in the PNFS_SCSI_INVALID_DATA state, the client has two options for recovery. If the data that has been written to these blocks is still cached by the client, the client can simply re-write the data via NFSv4, once the server has come back online. However, if the data is no longer in the client's cache, the client MUST NOT attempt to source the data from the data servers. Instead, it should attempt to commit the blocks in question to the server during the server's recovery grace period, by sending a LAYOUTCOMMIT with the "loca_reclaim" flag set to true. This process is described in detail in Section 18.42.4 of .

The server may decide that it cannot hold all of the state for layouts without running out of resources. In such a case, it is free to recall individual layouts using CB_LAYOUTRECALL to reduce the load, or it may choose to request that the client return any layout. The NFSv4.1 spec defines the following types:

const RCA4_TYPE_MASK_BLK_LAYOUT = 4; struct CB_RECALL_ANY4args { uint32_t craa_objects_to_keep; bitmap4 craa_type_mask; }; When the server sends a CB_RECALL_ANY request to a client specifying the RCA4_TYPE_MASK_BLK_LAYOUT bit in craa_type_mask, the client should immediately respond with NFS4_OK, and then asynchronously return complete file layouts until the number of files with layouts cached on the client is less than craa_object_to_keep.

The server may respond to LAYOUTGET with a variety of error statuses. These errors can convey transient conditions or more permanent conditions that are unlikely to be resolved soon. The transient errors, NFS4ERR_RECALLCONFLICT and NFS4ERR_TRYLATER, are used to indicate that the server cannot immediately grant the layout to the client. In the former case, this is because the server has recently issued a CB_LAYOUTRECALL to the requesting client, whereas in the case of NFS4ERR_TRYLATER, the server cannot grant the request possibly due to sharing conflicts with other clients. In either case, a reasonable approach for the client is to wait several milliseconds and retry the request. The client SHOULD track the number of retries, and if forward progress is not made, the client SHOULD send the READ or WRITE operation directly to the server. The error NFS4ERR_LAYOUTUNAVAILABLE may be returned by the server if layouts are not supported for the requested file or its containing file system. The server may also return this error code if the server is the progress of migrating the file from secondary storage, or for any other reason that causes the server to be unable to supply the layout. As a result of receiving NFS4ERR_LAYOUTUNAVAILABLE, the client SHOULD send future READ and WRITE requests directly to the server. It is expected that a client will not cache the file's layoutunavailable state forever, particular if the file is closed, and thus eventually, the client MAY reissue a LAYOUTGET operation.

Many storage devices implement volatile write caches that require an explicit flush to persist the data from write operations to stable storage. When a volatile write cache is used, the pNFS server must ensure the volatile write cache has been committed to stable storage before the LAYOUTCOMMIT operation returns.