[manual index][section index]


tiny: sh, broke, kill, rm - reduced command line interface to the Inferno system


tiny/sh [ -n ] [ -ccommand ] [ file ]


tiny/kill [ -g ] [ pid ... ] [ module ... ]

tiny/rm [ file ... ]


The tiny commands are smaller, simpler versions of more capable but larger Inferno commands. They are provided for use on devices where a certain level of functionality might be useful for configuration or maintenance (or development), but device constraints are such as to make the use of the normal, fleshier versions of the commands unattractive. For example, the Dis object files can be as much as 5 times smaller (or better) than the mainstream alternatives. They are also useful when initially porting the system. They live in the directory /dis/tiny, but could be placed in the /dis of a small device (eg, via root(3)

Broke kills broken processes and prints their process IDs.

Kill terminates each process (for a numeric process ID pid) or process running a given module (for a non-numeric module name), by writing a kill message to the corresponding process's control file in prog(3). The -g option causes kill to write a killgrp message instead, killing all processes in the given process's process group (see sys-pctl(2)). Processes running a module are identified by their status file, and the process ID of each such process is printed on standard output.

Rm removes files and empty directories, subject to the permission rules given in rm(1). There are no options.

Sh provides a simple user level interface (a shell) to the Inferno system. (It was once the only Inferno shell.) It reads input lines, identifies a command and arguments for that command, and arranges for execution of the corresponding Inferno module. There are features that allow input/output redirection, creating pipelines, and performing tasks in background. It is nevertheless a rudimentary shell designed for starting and debugging applications. It is not intended to serve as a general-purpose programmable shell.

If a file is named as a command line argument, that file is the source of input; otherwise, standard input is read.

Options are:

Don't fork the namespace. By default, sh forks the namespace, making subsequent namespace changes invisible to the previous namespace group.
Execute the single command rather than prompting to read commands from the standard input.

Command line syntax
Each line consists of one or more command pipelines each separated by either an ampersand (&) which indicates that the pipeline should be run in background or a semi-colon (;). The semi-colon need not be provided for the last command pipeline on a line.

Command pipelines are not allowed to span lines.

Each command pipeline consists of one or more commands separated by a vertical bar (|) character. The standard output of one command is made the standard input of the next command to the right.

Redirection of input/output to pipes takes precedence over redirection from/to files.

In the limit case, a command pipeline consists of a single command with no pipes.

A command consists of one or more fields. The first (leftmost) field is the command field. It is used to determined the executable file to be loaded and run; see below. The remaining fields are parsed and become command line arguments that are passed to the module's init function as a list of strings.

Any input following a # on a line is discarded as comment.

Finding the module
The command field is converted to the pathname of the Dis file of some module. That field can be either an absolute pathname, starting from /, or a relative pathname from the current directory.

As a convenience, the user need not specify the .dis suffix to the filename. If missing, it will be added by the shell.

If the load fails there is, in general, a second attempt to load the module by resolving the pathname relative to the /dis directory (or any directory bound to the /dis directory in the current namespace).

There are two exceptions to this second attempt. The second load attempt is not performed if the command field provides an absolute pathname or a relative pathname starting with dot-slash (./). Such explicit naming is taken to mean that the user will accept no substitutions.

The shell requires that the Dis file implement a module with an interface equivalent to the Command module as specified in /module/sh.m (see command(2)). Otherwise, the named file will not load.

In lieu of a path mechanism, a process can create a union directory at /dis.

File name expansion
Command line arguments (including the command field itself) are expanded by the shell according to the regular expression rules described in filepat(2).

This expansion is not applied to the filenames used for input/output redirection.

The shell special characters can be stripped of their meaning and treated as literals by enclosing them in single quotes. Inside a quoted string, the special meaning of the single quote can be removed by immediately following it with another single quote. Command lines with un-terminated quoted strings are rejected and cause an error message.

For example:

$ echo ''''
$ echo 'don''t'
$ echo 'hello' 'world
sh: unmatched quote
$ echo 'a'b
$ echo a'b'

Shell special characters
The following characters are treated specially by sh and must be quoted to be taken literally:

white space, except in a quoted string
white space, except in a quoted string
command line terminator
Start of comment
Start of/end of quoted string (single quote)
Interface between commands in a command pipeline.
Terminator for command pipelines to be run in background.
Terminator for command pipelines to be run synchronously by the shell.
Output re-direction: create file if it does not exist; truncate file if it exists
Output re-direction: create file if it does not exist; append to file if it exists
Input re-direction.

The shell uses a prompt consisting of the system name as provided by /dev/sysname suffixed by $.

Input/output re-directions
By default, standard input is the console keyboard and standard output the console display. Each command can specify that standard input be taken from a file and standard output be written to a file.

Attempts to redirect standard input to a non-existing file will fail. Redirecting standard output to a non-existing file will cause that file to be created. If the destination file already exists, it will be overwritten. Any previous contents are lost.

In cases of competing re-direction mechanisms (re-direct to a file and to a pipe), the pipe has precedence.

Background tasks
In general, the shell waits for the termination of a command pipeline before continuing execution, for example, prompting the user for the next command. However, if the command pipeline is terminated by an ampersand (&) character, the wait stage is skipped and the shell continues execution immediately, in this case the command pipeline executes as a background task.

Name space concerns
When started, the shell creates an independent file name space that is a copy of the file name space of the shell's creator.

Command pipelines started by the shell are executed by threads that share the shell's name space. If those commands modify the file name space (and they have not mimicked the shell in creating their own independent name space), those modifications will be perceived by the shell when it continues execution. See bind(1) and sys-pctl(2).






bind(1), sh(1), filepat(2), command(2), sys-pctl(2), cons(3), pipe(3), prog(3)

TINY(1 ) Rev:  Tue Mar 31 02:42:38 GMT 2015