As mentioned in #131 it would be nice if we had an automated test that all syscalls are available via FTRACE_SYSCALLS. Failing that we should just do it by hand.
Part of the problem is that checking against NR_syscalls doesnt' work because there are gaps in the syscall number space.
Looking at 64-bit, we currently have NR_syscalls == 387, but that's only 326 actual syscalls.
On a big endian system I see 309 in /sys/kernel/debug/tracing/events/syscalls and 311 on a little endian system.
So even if the 326 number is off by one or two, we seem to be missing quite a few.
As mentioned in #131 it would be nice if we had an automated test that all syscalls are available via FTRACE_SYSCALLS. Failing that we should just do it by hand.
Part of the problem is that checking against
NR_syscalls
doesnt' work because there are gaps in the syscall number space.Looking at 64-bit, we currently have
NR_syscalls
== 387, but that's only 326 actual syscalls.On a big endian system I see 309 in
/sys/kernel/debug/tracing/events/syscalls
and 311 on a little endian system.So even if the 326 number is off by one or two, we seem to be missing quite a few.