6 fs64 - cached-worm file server, dump
26 is a cached-worm file server that runs
27 as a user-mode program and can
28 maintain file systems created by
30 the original Plan 9 file server
31 that had its own kernel and operated
32 a standalone system with disks and
33 optical-disc jukebox attached.
36 which could only accept 9P connections over IL/IPv4 on Ethernets
37 (or over Datakit and Cyclones, long ago),
39 accepts 9P connections over any network medium and protocol
40 that it can announce on,
41 by default TCP (over IPv4 or IPv6).
42 Given suitable 9P clients,
43 one could even run 9P over
50 implements a 16K file system block size
51 and 32-bit disk addresses,
52 in order to be compatible with some existing file systems, notably
54 These parameters can be changed by recompilation.
57 places its server 9P connection in
59 and its command pipe in
60 .BI /srv/ name\f1 .cmd,
70 other than the default
80 use a newer, faster, and incompatible cache-device layout.
81 To convert an old file system's cache to the new layout,
82 dump the file system, note the last superblock number,
90 the file system, and start
97 enter the file server's configuration mode
98 before starting normal operation.
103 name of the file server's configuration.
106 Post file descriptor zero in
108 and read and write protocol messages on file descriptor one.
113 contains a simple device name
116 and a replacement per line.
117 The device name is in the usual
121 The replacement can be the name of an existing file
125 or another such device name.
126 For example, the file
136 would map accesses to device
140 and accesses to device
150 The file server normally requires all users except
152 to provide authentication tickets on each
154 This can be disabled using the
156 configuration command (see
159 The group numbered 9999, normally called
162 on the file server. Any user belonging to that group has
163 attenuated access privileges. Specifically, when checking such
164 a user's access to files, the file's permission bits are first ANDed
165 with 0770 for normal files or 0771 for directories. The effect is
166 to deny world access permissions to
169 when walking directories.
173 is always allowed to attach to
175 without authentication but has minimal permissions.
178 maintains three file systems
179 on a combination of disks and
180 write-once-read-many (WORM) magneto-optical disks.
183 is a simple disk-based file system similar to
187 is a worm-based file system with a disk-based
191 to overcome the write-once property of the worm.
192 The cache also holds recently accessed
193 non-modified blocks to
194 speed up the effective access time of the worm.
196 (usually daily at 5AM) the modified blocks in the
200 traffic to the file system is halted and the
201 modified blocks are relabeled to the unwritten
204 the file system traffic is continued and
205 the relabeled blocks are copied to the worm by
206 a background process.
209 Each time the main file system is dumped,
210 its root is appended to a subdirectory of the dump file system.
211 Since the dump file system is not mirrored with a disk
214 The name of the newly added root is created from the date
216 .BI / yyyy / mmdds\f1.
223 is the day number and
225 is a sequence number if more than
226 one dump is done in a day.
230 For the subsequent dumps
234 The root of the main file system
235 that is frozen on the first dump
239 in the dump file system.
240 .SS "Changes from fs"
242 IP configuration is ignored and the underlying system's is used.
246 commands have been omitted since they (or equivalents) can now be
247 executed directly on the underlying CPU server,
255 Files can be used directly as wren devices by giving
256 a file name rooted at
258 or using double or single quotes. Such a file name
259 can be appended to the
261 device instead of specifying
272 for Marvell SATA disks are not supported; use
274 to map wren devices to appropriate names under
277 The file server kernel seems to have scanned PCI buses
278 in reverse order from the other Plan 9 kernels,
279 so systems with multiple SCSI cards may find controller
282 can be used to compensate for this if you don't want to change
286 The file server kernel's
288 field in NVRAM was overloaded in recent times to hold a
290 key for the CPU hostowner.
293 runs on a CPU kernel,
294 the location of its configuration block must be supplied on the command line.
296 Disk labels are now implemented for
299 At the first access of a side,
301 will attempt to read the label and verify that it has the correct side
302 number and byte order; if either is wrong, it will issue a warning.
303 If the label cannot be read,
305 will attempt to write a new label.
307 The original file server reserved the rest of the machines RAM for
310 running under the Plan 9 kernel reserves a settable percentage
311 of the remaining user pages. The percentage is read from the
314 which when not set is assumed to be 25% (default).
316 Place the root of the
320 and show the modified times of the MIPS C compiler
321 over all dumps in February, 1992:
326 ls -l /n/dump/1992/02??/mips/bin/vc
329 To get only one line of output for each version of the compiler:
332 ls -lp /n/dump/1992/02??/mips/bin/vc | uniq
346 ``A Cached WORM File System'',
348 Software \- Practice and Experience,
353 ``The 64-bit Standalone Plan 9 File Server''
356 the file server serves both the old (9P1) and new (9P2000) versions of 9P,
357 deciding which to serve by sniffing the first packet on each connection.
359 File system block size and disk address size (32- or 64-bit) are fixed
360 at compilation time, and this is not easily changed.