spufs

SPU file system과 mount option

spufs는 Cell Broadband Engine Architecture를 구현한 PowerPC에서 SPU에 접근하는 파일 시스템이다. POSIX shared memory나 message queue와 비슷한 namespace를 제공하며, 파일 시스템에 쓰기 권한이 있는 사용자는 `spu_create(2)`로 spufs root에 SPU context를 만들 수 있다.

각 context는 논리 SPU 상태를 조작하는 미리 정의된 파일 집합의 directory로 표현된다. 사용자는 파일 권한을 바꿀 수 있지만 파일 자체를 추가하거나 제거할 수는 없다.

mount option `uid=<uid>`와 `gid=<gid>`는 mountpoint 소유 사용자와 그룹을 정하며 둘 다 기본값은 0(root)이다.

.. SPDX-License-Identifier: GPL-2.0

=====
spufs
=====

Name
====

       spufs - the SPU file system


Description
===========

       The SPU file system is used on PowerPC machines that implement the Cell
       Broadband Engine Architecture in order to access Synergistic  Processor
       Units (SPUs).

       The file system provides a name space similar to posix shared memory or
       message queues. Users that have write permissions on  the  file  system
       can use spu_create(2) to establish SPU contexts in the spufs root.

       Every SPU context is represented by a directory containing a predefined
       set of files. These files can be used for manipulating the state of the
       logical SPU. Users can change permissions on those files, but not actu-
       ally add or remove files.


Mount Options
=============

       uid=<uid>
              set the user owning the mount point, the default is 0 (root).

       gid=<gid>
              set the group owning the mount point, the default is 0 (root).

공통 파일 규칙과 `/mem`

spufs 파일은 대체로 `read(2)`·`write(2)`의 일반 동작을 따르지만 각 파일은 일부 연산만 지원한다. read를 지원하면 `readv(2)`, write를 지원하면 `writev(2)`도 지원한다. 모든 파일은 `access(2)`와 stat family를 지원하지만 `struct stat`에서 믿을 수 있는 필드는 `st_mode`, `st_nlink`, `st_uid`, `st_gid`뿐이다.

모든 파일에 chmod·fchmod·chown·fchown을 적용할 수 있지만, `/wbox`의 read 권한처럼 실제 지원 연산과 모순되는 권한은 부여할 수 없다.

`/mem`은 code와 data를 포함한 SPU local storage memory다. `read`, `pread`, `write`, `pwrite`, `lseek`를 지원하지만 EOF 너머 seek·write·pwrite는 허용되지 않는다. 보통 크기는 256KB다. `mmap`으로 process address space에 local storage를 보일 수 있으며 `MAP_SHARED`만 허용한다.

Files
=====

       The files in spufs mostly follow the standard behavior for regular sys-
       tem  calls like read(2) or write(2), but often support only a subset of
       the operations supported on regular file systems. This list details the
       supported  operations  and  the  deviations  from  the behaviour in the
       respective man pages.

       All files that support the read(2) operation also support readv(2)  and
       all  files  that support the write(2) operation also support writev(2).
       All files support the access(2) and stat(2) family of  operations,  but
       only  the  st_mode,  st_nlink,  st_uid and st_gid fields of struct stat
       contain reliable information.

       All files support the chmod(2)/fchmod(2) and chown(2)/fchown(2)  opera-
       tions,  but  will  not be able to grant permissions that contradict the
       possible operations, e.g. read access on the wbox file.

       The current set of files is:


   /mem
       the contents of the local storage memory  of  the  SPU.   This  can  be
       accessed  like  a regular shared memory file and contains both code and
       data in the address space of the SPU.  The possible  operations  on  an
       open mem file are:

       read(2), pread(2), write(2), pwrite(2), lseek(2)
              These  operate  as  documented, with the exception that seek(2),
              write(2) and pwrite(2) are not supported beyond the end  of  the
              file. The file size is the size of the local storage of the SPU,
              which normally is 256 kilobytes.

       mmap(2)
              Mapping mem into the process address space gives access  to  the
              SPU  local  storage  within  the  process  address  space.  Only
              MAP_SHARED mappings are allowed.

`/mbox`와 `/ibox`

`/mbox`는 첫 번째 SPU→CPU mailbox다. read-only이고 32-bit 단위로 읽으며 nonblocking mode만 사용할 수 있다. `poll()`도 block하지 않는다. 4 byte보다 작은 count는 `EINVAL`, data가 없으면 `EAGAIN`, 성공하면 buffer에 4 byte를 놓고 4를 반환한다.

`/ibox`는 두 번째 SPU→CPU interrupt mailbox다. `/mbox`와 달리 blocking I/O와 poll 대기를 지원한다. 4 byte 미만은 `EINVAL`; `O_NONBLOCK`에서 data가 없으면 `EAGAIN`; blocking mode면 SPU가 interrupt mailbox channel에 쓸 때까지 기다린다. 성공 시 4 byte와 반환값 4다.

`poll(2)`은 `/ibox`에 읽을 data가 생기면 `POLLIN | POLLRDNORM`을 반환한다.

   /mbox
       The first SPU to CPU communication mailbox. This file is read-only  and
       can  be  read  in  units of 32 bits.  The file can only be used in non-
       blocking mode and it even poll() will not block on  it.   The  possible
       operations on an open mbox file are:

       read(2)
              If  a  count smaller than four is requested, read returns -1 and
              sets errno to EINVAL.  If there is no data available in the mail
              box,  the  return  value  is set to -1 and errno becomes EAGAIN.
              When data has been read successfully, four bytes are  placed  in
              the data buffer and the value four is returned.


   /ibox
       The  second  SPU  to CPU communication mailbox. This file is similar to
       the first mailbox file, but can be read in blocking I/O mode,  and  the
       poll  family of system calls can be used to wait for it.  The  possible
       operations on an open ibox file are:

       read(2)
              If a count smaller than four is requested, read returns  -1  and
              sets errno to EINVAL.  If there is no data available in the mail
              box and the file descriptor has been opened with O_NONBLOCK, the
              return value is set to -1 and errno becomes EAGAIN.

              If  there  is  no  data  available  in the mail box and the file
              descriptor has been opened without  O_NONBLOCK,  the  call  will
              block  until  the  SPU  writes to its interrupt mailbox channel.
              When data has been read successfully, four bytes are  placed  in
              the data buffer and the value four is returned.

       poll(2)
              Poll  on  the  ibox  file returns (POLLIN | POLLRDNORM) whenever
              data is available for reading.

`/wbox`와 mailbox status

`/wbox`는 CPU→SPU mailbox다. write-only이며 32-bit 단위로 쓴다. 가득 차면 blocking descriptor는 SPU가 PPE mailbox channel에서 읽을 때까지 기다리고, `O_NONBLOCK`이면 `EAGAIN`이다. 4 byte 미만 요청은 `EINVAL`, 성공은 4 byte와 반환값 4다.

`poll(2)`은 쓸 공간이 생기면 `POLLOUT | POLLWRNORM`을 반환한다. 원문에는 ibox file이라고 쓰였지만 문맥상 `/wbox`의 poll 의미를 설명한다.

`/mbox_stat`, `/ibox_stat`, `/wbox_stat`는 현재 queue 길이를 나타내는 read-only 파일이다. 앞의 둘은 block 없이 읽을 수 있는 word 수, 뒤는 쓸 수 있는 word 수를 4-byte big-endian binary integer로 반환한다. 4 byte 미만 count는 `EINVAL`이다.

   /wbox
       The CPU to SPU communation mailbox. It is write-only and can be written
       in  units  of  32  bits. If the mailbox is full, write() will block and
       poll can be used to wait for it becoming  empty  again.   The  possible
       operations  on  an open wbox file are: write(2) If a count smaller than
       four is requested, write returns -1 and sets errno to EINVAL.  If there
       is  no space available in the mail box and the file descriptor has been
       opened with O_NONBLOCK, the return value is set to -1 and errno becomes
       EAGAIN.

       If  there is no space available in the mail box and the file descriptor
       has been opened without O_NONBLOCK, the call will block until  the  SPU
       reads  from  its PPE mailbox channel.  When data has been read success-
       fully, four bytes are placed in the data buffer and the value  four  is
       returned.

       poll(2)
              Poll  on  the  ibox file returns (POLLOUT | POLLWRNORM) whenever
              space is available for writing.


   /mbox_stat, /ibox_stat, /wbox_stat
       Read-only files that contain the length of the current queue, i.e.  how
       many  words  can  be  read  from  mbox or ibox or how many words can be
       written to wbox without blocking.  The files can be read only in 4-byte
       units  and  return  a  big-endian  binary integer number.  The possible
       operations on an open ``*box_stat`` file are:

       read(2)
              If a count smaller than four is requested, read returns  -1  and
              sets errno to EINVAL.  Otherwise, a four byte value is placed in
              the data buffer, containing the number of elements that  can  be
              read  from  (for  mbox_stat  and  ibox_stat)  or written to (for
              wbox_stat) the respective mail box without blocking or resulting
              in EAGAIN.

ASCII SPU register 파일

`/npc`, `/decr`, `/decr_status`, `/spu_tag_mask`, `/event_mask`, `/srr0`는 SPU 내부 register를 numeric ASCII string으로 표현한다. debug를 위해 read/write할 수 있지만 `npc` 외 register 접근은 SPU context save를 요구해 매우 비효율적이므로 정상 program 동작이 의존해서는 안 된다.

read buffer가 값과 newline 전체보다 짧으면 같은 descriptor의 후속 read가 처음 snapshot의 나머지 문자열을 완성한다. 완전한 문자열을 읽은 뒤에는 EOF인 0 byte를 반환하며 새 값을 읽으려면 파일을 다시 열어야 한다.

write는 buffer 처음부터 첫 non-numeric 문자 또는 끝까지 숫자를 parse해 register를 설정한다. 같은 descriptor의 후속 write는 이전 설정을 덮어쓴다.

   /npc, /decr, /decr_status, /spu_tag_mask, /event_mask, /srr0
       Internal  registers  of  the SPU. The representation is an ASCII string
       with the numeric value of the next instruction to  be  executed.  These
       can  be  used in read/write mode for debugging, but normal operation of
       programs should not rely on them because access to any of  them  except
       npc requires an SPU context save and is therefore very inefficient.

       The contents of these files are:

       =================== ===================================
       npc                 Next Program Counter
       decr                SPU Decrementer
       decr_status         Decrementer Status
       spu_tag_mask        MFC tag mask for SPU DMA
       event_mask          Event mask for SPU interrupts
       srr0                Interrupt Return address register
       =================== ===================================


       The   possible   operations   on   an   open  npc,  decr,  decr_status,
       spu_tag_mask, event_mask or srr0 file are:

       read(2)
              When the count supplied to the read call  is  shorter  than  the
              required  length for the pointer value plus a newline character,
              subsequent reads from the same file descriptor  will  result  in
              completing  the string, regardless of changes to the register by
              a running SPU task.  When a complete string has been  read,  all
              subsequent read operations will return zero bytes and a new file
              descriptor needs to be opened to read the value again.

       write(2)
              A write operation on the file results in setting the register to
              the  value  given  in  the string. The string is parsed from the
              beginning to the first non-numeric character or the end  of  the
              buffer.  Subsequent writes to the same file descriptor overwrite
              the previous setting.

`/fpcr` 4-byte register

`/fpcr`은 Floating Point Status and Control Register를 정확히 4-byte 파일로 노출한다. read count가 4보다 작으면 `EINVAL`, 아니면 현재 register 값 4 byte를 buffer에 놓는다. write도 4보다 작으면 `EINVAL`, 아니면 buffer의 4 byte로 fpcr을 갱신한다.

   /fpcr
       This file gives access to the Floating Point Status and Control  Regis-
       ter as a four byte long file. The operations on the fpcr file are:

       read(2)
              If  a  count smaller than four is requested, read returns -1 and
              sets errno to EINVAL.  Otherwise, a four byte value is placed in
              the data buffer, containing the current value of the fpcr regis-
              ter.

       write(2)
              If a count smaller than four is requested, write returns -1  and
              sets  errno  to  EINVAL.  Otherwise, a four byte value is copied
              from the data buffer, updating the value of the fpcr register.

Signal notification channel

`/signal1`과 `/signal2`는 32-bit word로 동작하는 SPU signal notification channel이다. 쓰면 SPU interrupt가 발생하며, SPU channel read 또는 host user space 파일 read로 값을 읽을 수 있다. SPU가 읽고 나면 값은 0으로 reset된다.

read·write 모두 count가 4보다 작으면 `EINVAL`이고, 그렇지 않으면 지정 signal register와 4 byte를 주고받는다. write가 기존 값을 교체할지 bitwise OR로 누적할지는 각각 `signal1_type`, `signal2_type`의 내용이 정한다.

type 파일은 numeric ASCII `0` 또는 `1`이다. mode 0은 overwrite, mode 1은 logical OR다. 짧은 read는 같은 descriptor에서 문자열의 나머지를 이어 읽고, 완전한 문자열 뒤에는 0 byte를 반환하므로 새 값을 보려면 다시 열어야 한다. write는 첫 non-numeric 문자까지 parse하며 후속 write가 이전 설정을 덮어쓴다.

   /signal1, /signal2
       The two signal notification channels of an SPU.  These  are  read-write
       files  that  operate  on  a 32 bit word.  Writing to one of these files
       triggers an interrupt on the SPU.  The  value  written  to  the  signal
       files can be read from the SPU through a channel read or from host user
       space through the file.  After the value has been read by the  SPU,  it
       is  reset  to zero.  The possible operations on an open signal1 or sig-
       nal2 file are:

       read(2)
              If a count smaller than four is requested, read returns  -1  and
              sets errno to EINVAL.  Otherwise, a four byte value is placed in
              the data buffer, containing the current value of  the  specified
              signal notification register.

       write(2)
              If  a count smaller than four is requested, write returns -1 and
              sets errno to EINVAL.  Otherwise, a four byte  value  is  copied
              from the data buffer, updating the value of the specified signal
              notification register.  The signal  notification  register  will
              either be replaced with the input data or will be updated to the
              bitwise OR of the old value and the input data, depending on the
              contents  of  the  signal1_type,  or  signal2_type respectively,
              file.


   /signal1_type, /signal2_type
       These two files change the behavior of the signal1 and signal2  notifi-
       cation  files.  The  contain  a numerical ASCII string which is read as
       either "1" or "0".  In mode 0 (overwrite), the  hardware  replaces  the
       contents of the signal channel with the data that is written to it.  in
       mode 1 (logical OR), the hardware accumulates the bits that are  subse-
       quently written to it.  The possible operations on an open signal1_type
       or signal2_type file are:

       read(2)
              When the count supplied to the read call  is  shorter  than  the
              required  length  for the digit plus a newline character, subse-
              quent reads from the same file descriptor will  result  in  com-
              pleting  the  string.  When a complete string has been read, all
              subsequent read operations will return zero bytes and a new file
              descriptor needs to be opened to read the value again.

       write(2)
              A write operation on the file results in setting the register to
              the value given in the string. The string  is  parsed  from  the
              beginning  to  the first non-numeric character or the end of the
              buffer.  Subsequent writes to the same file descriptor overwrite
              the previous setting.

fstab 예제와 관련 문서

`/etc/fstab` 예제는 `none /spu spufs gid=spu 0 0`으로 spufs를 `/spu`에 mount하고 소유 group을 `spu`로 지정한다.

저자는 Arnd Bergmann, Mark Nutter, Ulrich Weigand다. 관련 manual page는 `capabilities(7)`, `close(2)`, `spu_create(2)`, `spu_run(2)`, `spufs(7)`이다.

Examples
========
       /etc/fstab entry
              none      /spu      spufs     gid=spu   0    0


Authors
=======
       Arnd  Bergmann  <arndb@de.ibm.com>,  Mark  Nutter <mnutter@us.ibm.com>,
       Ulrich Weigand <Ulrich.Weigand@de.ibm.com>

See Also
========
       capabilities(7), close(2), spu_create(2), spu_run(2), spufs(7)