#! /usr/bin/env stap
# Copyright (C) 2016-2018 Red Hat, Inc.
# Written by William Cohen <wcohen@redhat.com>
#
#   container_check.stp watches for use of
#   prohibited capabilities, use of prohibited syscalls, and
#   syscall failures) that would indicate that this application
#   would not operate properly in a restricted contiainer.
#
#   By default this script monitors all systemcalls system-wide.
#   To limit to limit container_check.stp to monitoring a particular
#   process and it children use the systemtap -x <pid> option
#   or -c <command> option.
#
#   By default this script lists all capabilities requested.
#   To limit it to a subset of capabilities use the following
#   option on the command line with a '-' separated list of
#   forbidden capabilites:
#
#   -G forbidden_capabilities="badcap1-badcap2"
#
#   By default this script allows all syscalls.
#   To mark syscalls as forbidden use a '-' separate list: 
#   
#   -G forbidden_syscalls="syscall1-syscall2"
#
# control-c to exit data collection

global forbidden_capabilities="" # '-' separated list of forbidden capabilities
global forbidden_syscalls=""     # '-' separated list of forbidden syscalls

global capability, cap_use
global badcaps = -1, cap_name
global cap_syscall
global badsyscall
global problem_syscall
global syscall_errno

# Determine whether t is a ancestor of target()
# returns 1  if ancestor of target()
# returns 0  if not an ancestor of target()
function child_of_target:long (t:long)
{
  if (!target()) return 1
  while(t && t != task_parent(t)) {
    if (task_pid(t) == target()) return 1
    t = task_parent(t)
  }
  return 0
}

function init_cap_name2num()
{
  /* set up the names */
  cap_name[0]="cap_chown"
  cap_name[1]="cap_dac_override"
  cap_name[2]="cap_dac_read_search"
  cap_name[3]="cap_fowner"
  cap_name[4]="cap_fsetid"
  cap_name[5]="cap_kill"
  cap_name[6]="cap_setgid"
  cap_name[7]="cap_setuid"
  cap_name[8]="cap_setpcap"
  cap_name[9]="cap_linux_immutable"
  cap_name[10]="cap_net_bind_service"
  cap_name[11]="cap_net_broadcast"
  cap_name[12]="cap_net_admin"
  cap_name[13]="cap_net_raw"
  cap_name[14]="cap_ipc_lock"
  cap_name[15]="cap_ipc_owner"
  cap_name[16]="cap_sys_module"
  cap_name[17]="cap_sys_rawio"
  cap_name[18]="cap_sys_chroot"
  cap_name[19]="cap_sys_ptrace"
  cap_name[20]="cap_sys_pacct"
  cap_name[21]="cap_sys_admin"
  cap_name[22]="cap_sys_boot"
  cap_name[23]="cap_sys_nice"
  cap_name[24]="cap_sys_resource"
  cap_name[25]="cap_sys_time"
  cap_name[26]="cap_sys_tty_config"
  cap_name[27]="cap_mknod"
  cap_name[28]="cap_lease"
  cap_name[29]="cap_audit_write"
  cap_name[30]="cap_audit_control"
  cap_name[31]="cap_setfcap"
  cap_name[32]="cap_mac_override"
  cap_name[33]="cap_mac_admin"
  cap_name[34]="cap_syslog"
  cap_name[35]="cap_wake_alarm"
  cap_name[36]="cap_block_suspend"
}

function parse_capabilities() {
  /* convert optional list of forbidden capabilities into a bitmask */
  caps = 0
  cname = tokenize(forbidden_capabilities, "-")
  while (cname != "") {
    i =36
    while(i>0) {
      if(cname == cap_name[i]) {
        caps |= 1<<i
	i = -1
      }
      i -= 1
    }
    cname = tokenize("", "-")
  }
  if (caps) badcaps = caps
}

function parse_syscalls() {
  /* The following assignment is to ensure that badsyscall has typeinfo. */
  badsyscall["no_a_syscall"]=1
  /* Put in optional list of bad syscalls. */
  sysname = tokenize(forbidden_syscalls, "-")
  while (sysname != "") {
    badsyscall[sysname] = 1
    sysname = tokenize("", "-")
  }
}

probe begin {
  init_cap_name2num()
  parse_capabilities()
  parse_syscalls()

  printf ("starting container_check.stp. monitoring %d\n", target())
}

# bool ns_capable(struct user_namespace *ns, int cap)
probe ns_capable = kprobe.function("ns_capable")
{
  cap = 1 << int_arg(2)
}

#bool ns_capable_noaudit(struct user_namespace *ns, int cap)
probe ns_capable_noaudit = kprobe.function("ns_capable_noaudit")
{
  cap = 1 << int_arg(2)
}

#bool ns_capable_setid(struct user_namespace *ns, int cap)
probe ns_capable_setid = kprobe.function("ns_capable_setid")
{
  cap = 1 << int_arg(2)
}

# bool capable(int cap)
probe capable = kprobe.function("capable")
{
  cap = 1 << int_arg(1)
}

#bool file_ns_capable(const struct file *file, struct user_namespace *ns,
#		     int cap)
probe file_ns_capable = kprobe.function("file_ns_capable")
{
  cap = 1 << int_arg(3)
}

probe ns_capable?, ns_capable_noaudit?, ns_capable_setid?,
      capable?, file_ns_capable?
{
  if ((cap & badcaps) && child_of_target(task_current()))
    cap_use[tid()] |= cap
}

probe syscall_any.return {
  # note any problem capabilities use during syscall
  cap = cap_use[tid()]
  if (cap && child_of_target(task_current())) {
     capability[execname()] |= cap
     cap_syscall[execname(), name, cap] <<< 1
     delete cap_use[tid()]
  }

  # note any prohibited systemcalls
  if (name in badsyscall &&  child_of_target(task_current())) {
    problem_syscall[execname(), name] <<< 1
  }

  # note any syscalls returning errors
  if (retval < 0 && child_of_target(task_current())) {
    syscall_errno[execname(), name, retval] <<< 1
  }
}

probe end {
  printf("\n\ncapabilities used by executables\n");
  printf("%16s: %20s\n\n", "executable", "prob capability")
  foreach(e+ in capability) {
      cap = capability[e]
      i=0
      while (cap) {
        if (cap & 1)
	  printf("%16s: %20s\n", e, cap_name[i] );
	cap = cap >> 1
	i += 1
      }
      printf("\n")
  }

  printf("\n\ncapabilities used by syscalls\n");
  printf("%16s, %20s ( %16s ) : %16s\n", "executable", "syscall", "capability", "count")
  foreach([e+,s,c] in cap_syscall){
	printf("%16s, %20s ( ", e, s);
	cap = c
	i=0
	while (cap) {
	  if (cap & 1)
	    printf("%16s ", cap_name[i] );
	  cap = cap >> 1
	  i += 1
	}
	printf(") : %16d\n", @count(cap_syscall[e,s,c]) );
  }
  
  printf("\n\nforbidden syscalls\n");
  printf("%16s, %20s: %16s\n", "executable", "syscall", "count")
  foreach([e+,s] in problem_syscall){
      printf("%16s, %20s: %16d\n", e, s, @count(problem_syscall[e,s]) );
  }

  printf("\n\nfailed syscalls\n");
  printf("%16s, %20s = %16s: %16s\n", "executable", "syscall", "errno", "count")
  foreach([e+,s,v] in syscall_errno){
      printf("%16s, %20s = %16s: %16d\n", e, s, errno_str(v),
             @count(syscall_errno[e,s,v]) );
  }
}