index.txt
5E(1) General Commands Manual 5E(1) NAME 5e - user-mode ARM emulation SYNOPSIS 5e [ -npb ] text [ arguments ] DESCRIPTION 5e simulates the execution of an ARM binary in a Plan 9 environment. Unlike its predecessor vi(1) it supports, among others, the syscalls rfork (see fork(2)) and exec(2), which allows for the execution of threaded programs (e.g., rio(1) or catclock(1)). 5e executes the specified binary text, which is prepended by /bin if it does not begin with a slash, dot or hash sign. Unless -n is specified, /bin is replaced by the union of /arm/bin and /rc/bin. Unlike vi(1), 5e(1) does not provide built-in debugging facilities. It does provide emulation of the /proc directory, if the -p flag is speci‐ fied, to attach a proper debugger like acid(1). There is no equivalent of the profiling facilities, no caches or TLBs are simulated, either. 5e(1) currently has three options. -n By default, 5e(1) replaces /bin as mentioned above and also sets the variables cputype and objtype to arm. Supplying the -n op‐ tion suppresses this behaviour. -p The -p option activates emulation of a /proc file system, which is mounted at /proc and also posted as /srv/armproc, cf. srv(3). -b Supplying -b causes failing processes to call abort(2) instead of sysfatal. See below. SOURCE /sys/src/cmd/5e SEE ALSO vi(1) BUGS The host is required to be little endian and is assumed to have a floating point implementation conforming to IEEE 754. Broken processes are simulated in a rather unsatisfactory manner. The -b option leaks memory. The emulator does not post sys: notes. Obscure opcodes, in particular uncommon operations on R15, are not im‐ plemented. Careless use of the LDREX and STREX instructions can lead to deadlock, while the ARM Architecture Reference Manual indicates a real processor will exhibit undefined behavior in these cases. Accesses spanning segment boundaries will be treated as page faults. Many syscalls such as pread(2) will shuffle data around (in most cases unnecessarily) if invoked on potentially shared segments of variable length, in particular the bss segment. FPA emulation leaves much to be desired, rounding modes are ignored, all calculations are performed at extended precision. Floating point exceptions crash the emulator. Several syscalls, most notably the segattach(2) family, are not imple‐ mented (this should not be hard to fix). The emulator notes the value of errstr(2) only under obvious circumstances; with most syscalls only if the return value is negative. /proc emulation is more than unsatisfactory. The text argument should behave more like it would if it had been en‐ tered as an argument to rc(1). HISTORY 5e first appeared in 9front (June, 2011). 5E(1)