1POHMELFS network protocol.
2
3Basic structure used in network communication is following command:
4
5struct netfs_cmd
6{
7	__u16			cmd;	/* Command number */
8	__u16			csize;	/* Attached crypto information size */
9	__u16			cpad;	/* Attached padding size */
10	__u16			ext;	/* External flags */
11	__u32			size;	/* Size of the attached data */
12	__u32			trans;	/* Transaction id */
13	__u64			id;	/* Object ID to operate on. Used for feedback.*/
14	__u64			start;	/* Start of the object. */
15	__u64			iv;	/* IV sequence */
16	__u8			data[0];
17};
18
19Commands can be embedded into transaction command (which in turn has own command),
20so one can extend protocol as needed without breaking backward compatibility as long
21as old commands are supported. All string lengths include tail 0 byte.
22
23All commands are transferred over the network in big-endian. CPU endianness is used at the end peers.
24
25@cmd - command number, which specifies command to be processed. Following
26	commands are used currently:
27
28	NETFS_READDIR	= 1,	/* Read directory for given inode number */
29	NETFS_READ_PAGE,	/* Read data page from the server */
30	NETFS_WRITE_PAGE,	/* Write data page to the server */
31	NETFS_CREATE,		/* Create directory entry */
32	NETFS_REMOVE,		/* Remove directory entry */
33	NETFS_LOOKUP,		/* Lookup single object */
34	NETFS_LINK,		/* Create a link */
35	NETFS_TRANS,		/* Transaction */
36	NETFS_OPEN,		/* Open intent */
37	NETFS_INODE_INFO,	/* Metadata cache coherency synchronization message */
38	NETFS_PAGE_CACHE,	/* Page cache invalidation message */
39	NETFS_READ_PAGES,	/* Read multiple contiguous pages in one go */
40	NETFS_RENAME,		/* Rename object */
41	NETFS_CAPABILITIES,	/* Capabilities of the client, for example supported crypto */
42	NETFS_LOCK,		/* Distributed lock message */
43	NETFS_XATTR_SET,	/* Set extended attribute */
44	NETFS_XATTR_GET,	/* Get extended attribute */
45
46@ext - external flags. Used by different commands to specify some extra arguments
47	like partial size of the embedded objects or creation flags.
48
49@size - size of the attached data. For NETFS_READ_PAGE and NETFS_READ_PAGES no data is attached,
50	but size of the requested data is incorporated here. It does not include size of the command
51	header (struct netfs_cmd) itself.
52
53@id - id of the object this command operates on. Each command can use it for own purpose.
54
55@start - start of the object this command operates on. Each command can use it for own purpose.
56
57@csize, @cpad - size and padding size of the (attached if needed) crypto information.
58
59Command specifications.
60
61@NETFS_READDIR
62This command is used to sync content of the remote dir to the client.
63
64@ext - length of the path to object.
65@size - the same.
66@id - local inode number of the directory to read.
67@start - zero.
68
69
70@NETFS_READ_PAGE
71This command is used to read data from remote server.
72Data size does not exceed local page cache size.
73
74@id - inode number.
75@start - first byte offset.
76@size - number of bytes to read plus length of the path to object.
77@ext - object path length.
78
79
80@NETFS_CREATE
81Used to create object.
82It does not require that all directories on top of the object were
83already created, it will create them automatically. Each object has
84associated @netfs_path_entry data structure, which contains creation
85mode (permissions and type) and length of the name as long as name itself.
86
87@start - 0
88@size - size of the all data structures needed to create a path
89@id - local inode number
90@ext - 0
91
92
93@NETFS_REMOVE
94Used to remove object.
95
96@ext - length of the path to object.
97@size - the same.
98@id - local inode number.
99@start - zero.
100
101
102@NETFS_LOOKUP
103Lookup information about object on server.
104
105@ext - length of the path to object.
106@size - the same.
107@id - local inode number of the directory to look object in.
108@start - local inode number of the object to look at.
109
110
111@NETFS_LINK
112Create hard of symlink.
113Command is sent as "object_path|target_path".
114
115@size - size of the above string.
116@id - parent local inode number.
117@start - 1 for symlink, 0 for hardlink.
118@ext - size of the "object_path" above.
119
120
121@NETFS_TRANS
122Transaction header.
123
124@size - incorporates all embedded command sizes including theirs header sizes.
125@start - transaction generation number - unique id used to find transaction.
126@ext - transaction flags. Unused at the moment.
127@id - 0.
128
129
130@NETFS_OPEN
131Open intent for given transaction.
132
133@id - local inode number.
134@start - 0.
135@size - path length to the object.
136@ext - open flags (O_RDWR and so on).
137
138
139@NETFS_INODE_INFO
140Metadata update command.
141It is sent to servers when attributes of the object are changed and received
142when data or metadata were updated. It operates with the following structure:
143
144struct netfs_inode_info
145{
146	unsigned int		mode;
147	unsigned int		nlink;
148	unsigned int		uid;
149	unsigned int		gid;
150	unsigned int		blocksize;
151	unsigned int		padding;
152	__u64			ino;
153	__u64			blocks;
154	__u64			rdev;
155	__u64			size;
156	__u64			version;
157};
158
159It effectively mirrors stat(2) returned data.
160
161
162@ext - path length to the object.
163@size - the same plus size of the netfs_inode_info structure.
164@id - local inode number.
165@start - 0.
166
167
168@NETFS_PAGE_CACHE
169Command is only received by clients. It contains information about
170page to be marked as not up-to-date.
171
172@id - client's inode number.
173@start - last byte of the page to be invalidated. If it is not equal to
174	current inode size, it will be vmtruncated().
175@size - 0
176@ext - 0
177
178
179@NETFS_READ_PAGES
180Used to read multiple contiguous pages in one go.
181
182@start - first byte of the contiguous region to read.
183@size - contains of two fields: lower 8 bits are used to represent page cache shift
184	used by client, another 3 bytes are used to get number of pages.
185@id - local inode number.
186@ext - path length to the object.
187
188
189@NETFS_RENAME
190Used to rename object.
191Attached data is formed into following string: "old_path|new_path".
192
193@id - local inode number.
194@start - parent inode number.
195@size - length of the above string.
196@ext - length of the old path part.
197
198
199@NETFS_CAPABILITIES
200Used to exchange crypto capabilities with server.
201If crypto capabilities are not supported by server, then client will disable it
202or fail (if 'crypto_fail_unsupported' mount options was specified).
203
204@id - superblock index. Used to specify crypto information for group of servers.
205@size - size of the attached capabilities structure.
206@start - 0.
207@size - 0.
208@scsize - 0.
209
210@NETFS_LOCK
211Used to send lock request/release messages. Although it sends byte range request
212and is capable of flushing pages based on that, it is not used, since all Linux
213filesystems lock the whole inode.
214
215@id - lock generation number.
216@start - start of the locked range.
217@size - size of the locked range.
218@ext - lock type: read/write. Not used actually. 15'th bit is used to determine,
219	if it is lock request (1) or release (0).
220
221@NETFS_XATTR_SET
222@NETFS_XATTR_GET
223Used to set/get extended attributes for given inode.
224@id - attribute generation number or xattr setting type
225@start - size of the attribute (request or attached)
226@size - name length, path len and data size for given attribute
227@ext - path length for given object
228