Add support to discover and assume a hostid, as well as fix arguments passed to a BE

90zfsbootmenu/zfsbootmenu-countdown.sh

@@ -13,25 +13,24 @@ fi
 
 mkdir -p "${BASE}"
 
-# Attempt to import all pools read-only
-read_write='' all_pools=yes import_pool
-
-# Make sure at least one pool can be imported; if not,
-# drop to an emergency shell to allow the user to attempt recovery
-import_success=0
 while true; do
+  # Attempt to import all pools read-only
+  read_write='' all_pools=yes import_pool
+
+  # Make sure at least one was imported
+  import_success=0
   while IFS=$'\t' read -r _pool _health; do
     [ -n "${_pool}" ] || continue
-
     import_success=1
+
     if [ "${_health}" != "ONLINE" ]; then
       echo "${_pool}" >> "${BASE}/degraded"
+      zerror "prohibiting read/write operations on ${_pool}"
     fi
   done <<<"$( zpool list -H -o name,health )"
 
-  if [ "${import_success}" -ne 1 ]; then
-    emergency_shell "unable to successfully import a pool"
-  else
+  if [ "${import_success}" -eq 1 ]; then
+    # With at least one imported pool, just move on
     zdebug "$(
       echo "zpool list" ; \
       zpool list
@@ -42,6 +41,28 @@ while true; do
     )"
     break
   fi
+
+  # shellcheck disable=SC2154
+  if [ "${import_policy}" == "hostid" ] && masked="$( match_hostid )"; then
+    zdebug "match_hostid returned: ${masked}"
+
+    pool="${masked%%;*}"
+    spl_hostid="${masked##*;}"
+
+    export spl_hostid
+
+    zerror "imported ${pool} with assumed hostid ${spl_hostid}"
+    zerror "set spl_hostid=${spl_hostid} on ZBM KCL or regenerate with corrected /etc/hostid"
+
+    # potentially useful later?
+    : > "${BASE}/masked_hostid"
+
+    # Retry the import cycle with the masked hostid
+    continue
+  fi
+
+  # Allow the user to attempt recovery
+  emergency_shell "unable to successfully import a pool"
 done
 
 # Prefer a specific pool when checking for a bootfs value

90zfsbootmenu/zfsbootmenu-exec.sh

@@ -2,11 +2,12 @@
 # vim: softtabstop=2 shiftwidth=2 expandtab
 
 export spl_hostid
-export force_import
+export import_policy
 export menu_timeout
 export loglevel
 export root
 export zbm_sort
+export zbm_set_hostid
 
 # Disable all kernel messages to the console
 echo 0 > /proc/sys/kernel/printk

90zfsbootmenu/zfsbootmenu-lib.sh

@@ -105,6 +105,69 @@ center_string() {
   printf "%*s" $(( (${#1} + _WIDTH ) / 2)) "${1}"
 }
 
+
+# args: no arguments
+# prints: <imported pool>;<hostid>
+# returns: 0 on successful pool import, 1 on failure
+
+match_hostid() {
+  local importable pool state hostid
+  importable=()
+
+  # If root is defined and not zfsbootmenu, prefer that pool for masked import
+  # shellcheck disable=SC2154
+  if [ -n "${root}" ] && [ "${root}" != "zfsbootmenu" ]; then
+    pool="${root%%/*}"
+    importable+=( "${pool}" )
+  fi
+
+  while read -r line; do
+    case "$line" in
+      pool*)
+        pool="${line#pool: }"
+        ;;
+      state*)
+        state="${line#state: }"
+        if [ "${state}" == "ONLINE" ] && [ -n "${pool}" ]; then
+          importable+=("${pool}")
+          pool=""
+        fi
+        ;;
+    esac
+  done <<<"$( zpool import )"
+
+  zdebug "importable pools: ${importable[*]}"
+
+  for pool in "${importable[@]}"; do
+    zdebug "trying to import: ${pool}"
+    hostid="$( zpool import -o readonly=on -N "${pool}" 2>&1 | grep -E -o "hostid=[A-Za-z0-9]{1,8}")"
+
+    if [ -n "${hostid}" ]; then
+      hostid="${hostid##*=}"
+      zdebug "discovered pool owner hostid: ${hostid}"
+    else
+      zdebug "unable to scrape hostid for ${pool}, skipping"
+      continue
+    fi
+
+    if [ "${hostid}" == "0" ]; then
+      hostid="00000000"
+    fi
+
+    zdebug "setting hostid to: ${hostid}"
+    echo -ne "\\x${hostid:6:2}\\x${hostid:4:2}\\x${hostid:2:2}\\x${hostid:0:2}" > "/etc/hostid"
+
+    if read_write='' import_pool "${pool}"; then
+      zdebug "successfully imported ${pool}"
+      echo "${pool};${hostid}"
+      return 0
+    fi
+  done
+
+  # no pools could be imported, we failed to match a hostid
+  return 1
+}
+
 # arg1: ZFS filesystem name
 # prints: mountpoint
 # returns: 0 on success
@@ -981,6 +1044,54 @@ preload_be_cmdline() {
   fi
 }
 
+# arg1: key(and associated value) to suppress from KCL
+# arg2..argN: kernel command line
+# prints: supressed kernel command line
+# returns: 0 on success
+
+suppress_kcl_arg() {
+  arg=$1
+  shift
+
+  if [ -z "${arg}" ]; then
+    echo "$*"
+    return 0
+  fi
+
+  awk <<< "$*" '
+    BEGIN {
+      quot = 0;
+      supp = 0;
+      ORS = " ";
+    }
+
+    {
+      for (i=1; i <= NF; i++) {
+        if ( quot == 0 ) {
+          # If unquoted, determine if output should be suppressed
+          if ( $(i) ~ /^'"${arg}"'=/ ) {
+            # Suppress unwanted argument
+             supp = 1;
+          } else {
+            # Nothing else is suppressed
+            supp = 0;
+          }
+        }
+
+        # If output is not suppressed, print the field
+        if ( supp == 0 && length($(i)) > 0 ) {
+          print $(i);
+        }
+
+        # If an odd number of quotes are in this field, toggle quoting
+        if ( gsub(/"/, "\"", $(i)) % 2 == 1 ) {
+          quot = (quot + 1) % 2;
+        }
+      }
+    }
+  '  
+}
+
 # arg1: ZFS filesystem
 # prints: kernel command line arguments
 # returns: nothing
@@ -1012,39 +1123,15 @@ load_be_cmdline() {
   if [ -e "${BASE}/noresume" ]; then
     zdebug "${BASE}/noresume set, processing ${zfsbe_args}"
     # Must replace resume= arguments and append a noresume
-    zfsbe_args="$( awk <<< "${zfsbe_args}" '
-      BEGIN {
-        quot = 0;
-        supp = 0;
-        ORS = " ";
-      }
-
-      {
-        for (i=1; i <= NF; i++) {
-          if ( quot == 0 ) {
-            # If unquoted, determine if output should be suppressed
-            if ( $(i) ~ /^resume=/ ) {
-              # Argument starts with "resume=", suppress
-              supp = 1;
-            } else {
-              # Nothing else is suppressed
-              supp = 0;
-            }
-          }
-
-          # If output is not suppressed, print the field
-          if ( supp == 0 && length($(i)) > 0 ) {
-            print $(i);
-          }
+    zfsbe_args="$( suppress_kcl_arg resume "${zfsbe_args}" ) noresume"
+  fi
 
-          # If an odd number of quotes are in this field, toggle quoting
-          if ( gsub(/"/, "\"", $(i)) % 2 == 1 ) {
-            quot = (quot + 1) % 2;
-          }
-        }
-        printf "noresume";
-      }
-    ' )"
+  # shellcheck disable=SC2154
+  if [ "${zbm_set_hostid}" -eq 1 ]; then
+    zdebug "removing spl_hostid in: ${zfsbe_args}"
+    zfsbe_args="$( suppress_kcl_arg spl_hostid "${zfsbe_args}" )"
+    zdebug "rewriting spl.spl_hostid in ${zfsbe_args}"
+    zfsbe_args="spl.spl_hostid=0x${spl_hostid} $( suppress_kcl_arg spl.spl_hostid "${zfsbe_args}" )"
   fi
 
   zdebug "processed commandline: ${zfsbe_args}"
@@ -1056,7 +1143,7 @@ load_be_cmdline() {
 # returns: 0 on success, 1 on failure
 
 # Accepted environment variables
-# force_import=1: enable force importing of a pool
+# import_policy=force: enable force importing of a pool
 # read_write=1: import read-write, defaults to read-only
 # rewind_to_checkpoint=1: enable --rewind-to-checkpoint
 # all_pools=1: import all pools instead of a single specific pool
@@ -1073,9 +1160,9 @@ import_pool() {
   import_args=( "-N" )
 
   # shellcheck disable=SC2154
-  if [ -n "${force_import}" ]; then
+  if [ "${import_policy}" == "force" ]; then
     import_args+=( "-f" )
-    zdebug "force_import set: ${force_import}"
+    zdebug "import_policy set: ${import_policy}"
   fi
 
   # shellcheck disable=SC2154
@@ -1774,7 +1861,7 @@ emergency_shell() {
 
   echo -n "Launching emergency shell: "
   echo -e "${message}\n"
-  /bin/bash --rcfile <( test -f /lib/zfsbootmenu-lib.sh && echo "source /lib/zfsbootmenu-lib.sh" ) 
+  /bin/bash --rcfile <( test -f /lib/zfsbootmenu-lib.sh && echo "source /lib/zfsbootmenu-lib.sh" )
 }
 
 # prints: nothing

90zfsbootmenu/zfsbootmenu-parse-commandline.sh

@@ -10,10 +10,13 @@ if [ -n "${spl_hostid}" ] ; then
   info "ZFSBootMenu: using hostid from command line: ${spl_hostid}"
   echo -ne "\\x${spl_hostid:6:2}\\x${spl_hostid:4:2}\\x${spl_hostid:2:2}\\x${spl_hostid:0:2}" >/etc/hostid
 elif [ -f "/etc/hostid" ] ; then
-  info "ZFSBootMenu: using hostid from /etc/hostid: $(hostid)"
+  spl_hostid="$(hostid)"
+  info "ZFSBootMenu: using hostid from /etc/hostid: ${spl_hostid}"
 else
   warn "ZFSBootMenu: no hostid found on kernel command line or /etc/hostid"
-  warn "ZFSBootMenu: pools may not import correctly"
+  warn "ZFSBootMenu: defaulting hostid to 00000000"
+  echo -ne "\\x00\\x00\\x00\\x00" >/etc/hostid
+  spl_hostid="00000000"
 fi
 
 # Use the last defined console= to control menu output
@@ -36,11 +39,29 @@ else
   loglevel=3
 fi
 
-# Force import pools only when explicitly told to do so
-if getargbool 0 zbm.force_import -d force_import ; then
-  # shellcheck disable=SC2034
-  force_import="yes"
-  info "ZFSBootMenu: enabling force import of ZFS pools"
+import_policy=$( getarg zbm.import_policy )
+if [ -n "${import_policy}" ]; then
+  case "${import_policy}" in
+    hostid)
+      info "ZFSBootMenu: setting import_policy to hostid matching"
+      ;;
+    force)
+      info "ZFSBootMenu: setting import_policy to force"
+      ;;
+    legacy)
+      info "ZFSBootMenu: setting import_policy to legacy"
+      ;;
+    *)
+      info "ZFSBootMenu: unknown import policy ${import_policy}, defaulting to legacy"
+      import_policy="legacy"
+      ;;
+  esac
+elif getargbool 0 zbm.force_import -d force_import ; then
+  import_policy="force"
+  info "ZFSBootMenu: setting import_policy to force"
+else
+  info "ZFSBootMenu: defaulting import_policy to hostid matching"
+  import_policy="legacy"
 fi
 
 # zbm.timeout= overrides timeout=
@@ -98,10 +119,21 @@ fi
 # Turn on tmux integrations
 # shellcheck disable=SC2034
 if getargbool 0 zbm.tmux ; then
-  zbm_tmux="yes"
+  zbm_tmux=1
   info "ZFSBootMenu: enabling tmux integrations"
 fi
 
+# Do not automatically set spl_hostid on the BE KCL
+# shellcheck disable=SC2034
+if getargbool 0 zbm.set_hostid ; then
+  zbm_set_hostid=0
+  info "ZFSBootMenu: disabling automatic replacement of spl_hostid"
+else
+  zbm_set_hostid=1
+  info "ZFSBootMenu: defaulting automatic replacement of spl_hostid to on"
+fi
+
+
 wait_for_zfs=0
 case "${root}" in
   ""|zfsbootmenu|zfsbootmenu:)

90zfsbootmenu/zlogtail.sh

@@ -6,7 +6,7 @@ trap 'rm -f ${PID_FILE}' EXIT
 
 #shellcheck disable=SC2154
 LOG_LEVEL="0,1,2,3,4,5,6,7"
-FACILITY="kern,user"
+FACILITY="kern,user,daemon"
 FOLLOW=""
 ALLOW_EXIT=1
 while getopts "cfnl:F:" opt; do