Validate spl_hostid, control loading spl.ko

- Parse spl_hostid and spl.spl_hostid on the kernel command line.
  Validate that either the base 10 or base 16 value is valid,
  normalize to 8 hex digits

- Manually load spl.ko via insmod, setting the spl_hostid module
  parameter to 0. This way an invalid spl.spl_hostid on the kernel
  command line won't result in a (previously) unrecoverable module load
  error.

- Add in insmod/modinfo/lsmod/depmod as symlink aliases to the
  multi-call kmod binary.

Closes #184

90zfsbootmenu/module-setup.sh

@@ -84,6 +84,10 @@ install() {
     "chmod"
     "od"
     "stty"
+    "insmod"
+    "modinfo"
+    "lsmod"
+    "depmod"
   )
 
   for _exec in "${essential_execs[@]}"; do

90zfsbootmenu/zfsbootmenu-init.sh

@@ -19,9 +19,6 @@ if ! is_lib_sourced > /dev/null 2>&1 ; then
   exec /bin/bash
 fi
 
-# Make sure /dev/zfs exists, otherwise drop to a recovery shell
-[ -e /dev/zfs ] || emergency_shell "/dev/zfs missing, check that kernel modules are loaded"
-
 if [ -z "${BASE}" ]; then
   export BASE="/zfsbootmenu"
 fi
@@ -30,14 +27,38 @@ mkdir -p "${BASE}"
 
 # Write out a default or overridden hostid
 if [ -n "${spl_hostid}" ] ; then
-  zinfo "writing /etc/hostid from command line: ${spl_hostid}"
-  write_hostid "${spl_hostid}"
+  if write_hostid "${spl_hostid}" ; then
+    zinfo "writing /etc/hostid from command line: ${spl_hostid}"
+  else
+    # write_hostid logs an error for us, just note the new value
+    # shellcheck disable=SC2154
+    write_hostid "${default_hostid}"
+    zinfo "defaulting hostid to ${default_hostid}"
+  fi
 elif [ ! -e /etc/hostid ]; then
   zinfo "no hostid found on kernel command line or /etc/hostid"
-  zinfo "defaulting hostid to 00bab10c"
-  write_hostid 00bab10c
+  # shellcheck disable=SC2154
+  zinfo "defaulting hostid to ${default_hostid}"
+  write_hostid "${default_hostid}"
+fi
+
+# Capture the filename for spl.ko
+_modfilename="$( modinfo -F filename spl )"
+zinfo "loading ${_modfilename}"
+
+# Load with a hostid of 0, so that /etc/hostid takes precedence
+if ! _modload="$( insmod "${_modfilename}" "spl_hostid=0" 2>&1 )" ; then
+  zdebug "${_modload}"
+  emergency_shell "unable to load SPL kernel module"
 fi
 
+if ! _modload="$( modprobe zfs 2>&1 )" ; then
+  zdebug "${_modload}"
+  emergency_shell "unable to load ZFS kernel modules"
+fi
+
+udevadm settle
+
 # Prefer a specific pool when checking for a bootfs value
 # shellcheck disable=SC2154
 if [ "${root}" = "zfsbootmenu" ]; then

90zfsbootmenu/zfsbootmenu-parse-commandline.sh

@@ -21,10 +21,33 @@ fi
 
 # Let the command line override our host id.
 # shellcheck disable=SC2034
-spl_hostid=$(getarg spl_hostid=)
+cli_spl_hostid=$( getarg spl.spl_hostid ) || cli_spl_hostid=$( getarg spl_hostid )
+if [ -n "${cli_spl_hostid}" ] ; then
+  # Start empty so only valid hostids are set for future use
+  spl_hostid=
+
+  # Test for decimal
+  if (( 10#${cli_spl_hostid} )) >/dev/null 2>&1 ; then
+    spl_hostid="$( printf "%08x" "${cli_spl_hostid}" )"
+  # Test for hex. Requires 0x, if present, to be stripped
+  # The change to cli_spl_hostid isn't saved outside of the test
+  elif (( 16#${cli_spl_hostid#0x} )) >/dev/null 2>&1 ; then
+    spl_hostid="${cli_spl_hostid#0x}"
+    # printf will strip leading 0s if there are more than 8 hex digits
+    # normalize to a maximum of 8, then run through printf to fill in
+    # if there are fewer than 8 digits.
+    spl_hostid="$( printf "%08x" "0x${spl_hostid:0:8}" )"
+  # base 10 / base 16 tests fail on 0
+  elif [ "${cli_spl_hostid#0x}" -eq "0" ] >/dev/null 2>&1 ; then
+    spl_hostid=0
+  # Not valid hex or dec, log
+  else
+    warn "ZFSBootMenu: invalid hostid value ${cli_spl_hostid}, ignoring"
+  fi
+fi
 
 # Use the last defined console= to control menu output
-control_term=$( getarg console=)
+control_term=$( getarg console )
 if [ -n "${control_term}" ]; then
   control_term="/dev/${control_term%,*}"
   info "ZFSBootMenu: setting controlling terminal to: ${control_term}"
@@ -35,7 +58,7 @@ fi
 
 # Use loglevel to determine logging to /dev/kmsg
 min_logging=4
-loglevel=$( getarg loglevel=)
+loglevel=$( getarg loglevel )
 if [ -n "${loglevel}" ]; then
   # minimum log level of 4, so we never lose error or warning messages
   [ "${loglevel}" -ge ${min_logging} ] || loglevel=${min_logging}
@@ -94,7 +117,7 @@ else
 fi
 
 zbm_import_delay=$( getarg zbm.import_delay )
-if [ "${zbm_import_delay:-0}" -gt 0 ] 2>/dev/null ; then 
+if [ "${zbm_import_delay:-0}" -gt 0 ] 2>/dev/null ; then
   # Again, this validates that zbm_import_delay is numeric in addition to logging
   info "ZFSBootMenu: import retry delay is ${zbm_import_delay} seconds"
 else
@@ -103,13 +126,13 @@ fi
 
 # Allow setting of console size; there are no defaults here
 # shellcheck disable=SC2034
-zbm_lines=$( getarg zbm.lines=)
+zbm_lines=$( getarg zbm.lines )
 # shellcheck disable=SC2034
-zbm_columns=$( getarg zbm.columns=)
+zbm_columns=$( getarg zbm.columns )
 
 # Allow sorting based on a key
 zbm_sort=
-sort_key=$( getarg zbm.sort_key=)
+sort_key=$( getarg zbm.sort_key )
 if [ -n "${sort_key}" ] ; then
   valid_keys=( "name" "creation" "used" )
   for key in "${valid_keys[@]}"; do

90zfsbootmenu/zfsbootmenu-preinit.sh

@@ -31,6 +31,7 @@ export menu_timeout="${menu_timeout}"
 export loglevel="${loglevel}"
 export root="${root}"
 export zbm_require_bpool="${zbm_require_bpool}"
+export default_hostid=00bab10c
 export zbm_sort="${zbm_sort}"
 export zbm_set_hostid="${zbm_set_hostid}"
 export zbm_import_delay="${zbm_import_delay}"
@@ -41,9 +42,6 @@ getcmdline > "${BASE}/zbm.cmdline"
 # Set a non-empty hostname so we show up in zpool history correctly
 echo "ZFSBootMenu" > /proc/sys/kernel/hostname
 
-modprobe zfs 2>/dev/null
-udevadm settle
-
 # try to set console options for display and interaction
 # this is sometimes run as an initqueue hook, but cannot be guaranteed
 #shellcheck disable=SC2154