add --version, bump to v0.28

We were saying unknown instead of vulnerable when the count of lfence opcodes was low
This was not impacting batch mode or the final decision, just the human-readable output of the script.

README.md

@@ -3,7 +3,7 @@ Spectre & Meltdown Checker
 
 A simple shell script to tell if your Linux installation is vulnerable against the 3 "speculative execution" CVEs that were made public early 2018.
 
-Without options, it'll inspect you currently running kernel. 
+Without options, it'll inspect your currently running kernel. 
 You can also specify a kernel image on the command line, if you'd like to inspect a kernel you're not running.
 
 The script will do its best to detect mitigations, including backported non-vanilla patches, regardless of the advertised kernel version number.
@@ -40,6 +40,6 @@ However, some mitigations could also exist in your kernel that this script doesn
 
 Your system exposure also depends on your CPU. As of now, AMD and ARM processors are marked as immune to some or all of these vulnerabilities (except some specific ARM models). All Intel processors manufactured since circa 1995 are thought to be vulnerable. Whatever processor one uses, one might seek more information from the manufacturer of that processor and/or of the device in which it runs.
 
-The nature of the discovered vulnerabilities being quite new, the landscape of vulnerable processors can be expected to change over time, which is why this script makes the assumption that all CPUs are vulnerable, except if the manufacturer explicitely stated otherwise in a verifiable public announcement.
+The nature of the discovered vulnerabilities being quite new, the landscape of vulnerable processors can be expected to change over time, which is why this script makes the assumption that all CPUs are vulnerable, except if the manufacturer explicitly stated otherwise in a verifiable public announcement.
 
 This tool has been released in the hope that it'll be useful, but don't use it to jump to conclusions about your security.

spectre-meltdown-checker.sh

@@ -8,7 +8,7 @@
 #
 # Stephane Lesimple
 #
-VERSION=0.25
+VERSION=0.28
 
 # Script configuration
 show_usage()
@@ -22,7 +22,7 @@ show_usage()
 		Two modes are available.
 
 		First mode is the "live" mode (default), it does its best to find information about the currently running kernel.
-		To run under this mode, just start the script without any option (you can also use --live explicitely)
+		To run under this mode, just start the script without any option (you can also use --live explicitly)
 
 		Second mode is the "offline" mode, where you can inspect a non-running kernel.
 		You'll need to specify the location of the vmlinux file, and if possible, the corresponding config and System.map files:
@@ -33,13 +33,15 @@ show_usage()
 
 	Options:
 		--no-color			Don't use color codes
-		-v, --verbose			Increase verbosity level
+		--verbose, -v			Increase verbosity level
+		--no-sysfs			Don't use the /sys interface even if present
 		--batch text			Produce machine readable output, this is the default if --batch is specified alone
 		--batch json			Produce JSON output formatted for Puppet, Ansible, Chef...
 		--batch nrpe			Produce machine readable output formatted for NRPE
 		--variant [1,2,3]		Specify which variant you'd like to check, by default all variants are checked
 						Can be specified multiple times (e.g. --variant 2 --variant 3)
 
+
 	IMPORTANT:
 	A false sense of security is worse than no security at all.
 	Please use the --disclaimer option to understand exactly what this script does.
@@ -65,7 +67,7 @@ in which it runs.
 
 The nature of the discovered vulnerabilities being quite new, the landscape of vulnerable processors can be expected
 to change over time, which is why this script makes the assumption that all CPUs are vulnerable, except if the manufacturer
-explicitely stated otherwise in a verifiable public announcement.
+explicitly stated otherwise in a verifiable public announcement.
 
 This tool has been released in the hope that it'll be useful, but don't use it to jump to conclusions about your security.
 
@@ -86,6 +88,7 @@ opt_variant1=0
 opt_variant2=0
 opt_variant3=0
 opt_allvariants=1
+opt_no_sysfs=0
 
 nrpe_critical=0
 nrpe_unknown=0
@@ -95,13 +98,13 @@ __echo()
 {
 	opt="$1"
 	shift
-	msg="$@"
+	_msg="$@"
 	if [ "$opt_no_color" = 1 ] ; then
 		# strip ANSI color codes
-		msg=$(/bin/echo -e  "$msg" | sed -r "s/\x1B\[([0-9]{1,2}(;[0-9]{1,2})?)?[m|K]//g")
+		_msg=$(/bin/echo -e  "$_msg" | sed -r "s/\x1B\[([0-9]{1,2}(;[0-9]{1,2})?)?[m|K]//g")
 	fi
 	# explicitely call /bin/echo to avoid shell builtins that might not take options
-	/bin/echo $opt -e "$msg"
+	/bin/echo $opt -e "$_msg"
 }
 
 _echo()
@@ -122,7 +125,7 @@ _echo_nol()
 
 _warn()
 {
-	_echo 0 "\033[31m${@}\033[0m"
+	_echo 0 "\033[31m${@}\033[0m" >&2
 }
 
 _info()
@@ -142,7 +145,7 @@ _verbose()
 
 _debug()
 {
-	_echo 3 "(debug) $@"
+	_echo 3 "\033[34m(debug) $@\033[0m"
 }
 
 is_cpu_vulnerable()
@@ -250,6 +253,9 @@ while [ -n "$1" ]; do
 	elif [ "$1" = "--no-color" ]; then
 		opt_no_color=1
 		shift
+	elif [ "$1" = "--no-sysfs" ]; then
+		opt_no_sysfs=1
+		shift
 	elif [ "$1" = "--batch" ]; then
 		opt_batch=1
 		opt_verbose=0
@@ -285,6 +291,10 @@ while [ -n "$1" ]; do
 		show_header
 		show_usage
 		exit 0
+	elif [ "$1" = "--version" ]; then
+		opt_no_color=1
+		show_header
+		exit 1
 	elif [ "$1" = "--disclaimer" ]; then
 		show_header
 		show_disclaimer
@@ -452,7 +462,8 @@ if [ "$opt_live" = 1 ]; then
 		_warn "To run it as root, you can try the following command: sudo $0"
 		_warn
 	fi
-	_info "Checking for vulnerabilities against live running kernel \033[35m"$(uname -s) $(uname -r) $(uname -v) $(uname -m)"\033[0m"
+	_info "Checking for vulnerabilities against running kernel \033[35m"$(uname -s) $(uname -r) $(uname -v) $(uname -m)"\033[0m"
+	_info "CPU is\033[35m"$(grep '^model name' /proc/cpuinfo | cut -d: -f2 | head -1)"\033[0m"
 
 	# try to find the image of the current running kernel
 	# first, look for the BOOT_IMAGE hint in the kernel cmdline
@@ -548,46 +559,84 @@ umount_debugfs()
 	fi
 }
 
+sys_interface_check()
+{
+	[ "$opt_live" = 1 -a "$opt_no_sysfs" = 0 -a -r "$1" ] || return 1
+	_info_nol "* Checking whether we're safe according to the /sys interface: "
+	if grep -qi '^not affected' "$1"; then
+		# Not affected
+		status=OK
+		pstatus green YES "kernel confirms that your CPU is unaffected"
+	elif grep -qi '^mitigation' "$1"; then
+		# Mitigation: PTI
+		status=OK
+		pstatus green YES "kernel confirms that the mitigation is active"
+	elif grep -qi '^vulnerable' "$1"; then
+		# Vulnerable
+		status=VULN
+		pstatus red NO "kernel confirms your system is vulnerable"
+	else
+		status=UNK
+		pstatus yellow UNKNOWN "unknown value reported by kernel"
+	fi
+	msg=$(cat "$1")
+	_debug "sys_interface_check: $1=$msg"
+	return 0
+}
+
 ###################
 # SPECTRE VARIANT 1
 check_variant1()
 {
 	_info "\033[1;34mCVE-2017-5753 [bounds check bypass] aka 'Spectre Variant 1'\033[0m"
-	_info_nol "* Checking count of LFENCE opcodes in kernel: "
 
-	status=0
+	status=UNK
+	sys_interface_available=0
+	msg=''
+	if sys_interface_check "/sys/devices/system/cpu/vulnerabilities/spectre_v1"; then
+		# this kernel has the /sys interface, trust it over everything
+		sys_interface_available=1
+	else
+		# no /sys interface (or offline mode), fallback to our own ways
+		_info_nol "* Checking count of LFENCE opcodes in kernel: "
 		if [ -n "$vmlinux_err" ]; then
-		pstatus yellow UNKNOWN "$vmlinux_err"
+			msg="couldn't check ($vmlinux_err)"
+			status=UNK
+			pstatus yellow UNKNOWN
 		else
 			if ! which objdump >/dev/null 2>&1; then
-			pstatus yellow UNKNOWN "missing 'objdump' tool, please install it, usually it's in the binutils package"
+				msg="missing 'objdump' tool, please install it, usually it's in the binutils package"
+				status=UNK
+				pstatus yellow UNKNOWN
 			else
 				# here we disassemble the kernel and count the number of occurences of the LFENCE opcode
 				# in non-patched kernels, this has been empirically determined as being around 40-50
 				# in patched kernels, this is more around 70-80, sometimes way higher (100+)
 				# v0.13: 68 found in a 3.10.23-xxxx-std-ipv6-64 (with lots of modules compiled-in directly), which doesn't have the LFENCE patches,
 				# so let's push the threshold to 70.
-			# TODO LKML patch is starting to dump LFENCE in favor of the PAUSE opcode, we might need to check that (patch not stabilized yet)
+				nb_lfence=$(objdump -d "$vmlinux" | grep -wc lfence)
-			nb_lfence=$(objdump -D "$vmlinux" | grep -wc lfence)
 				if [ "$nb_lfence" -lt 70 ]; then
-				pstatus red NO "only $nb_lfence opcodes found, should be >= 70"
+					msg="only $nb_lfence opcodes found, should be >= 70, heuristic to be improved when official patches become available"
-				status=1
+					status=VULN
+					pstatus red NO
 				else
-				pstatus green YES "$nb_lfence opcodes found, which is >= 70"
+					msg="$nb_lfence opcodes found, which is >= 70, heuristic to be improved when official patches become available"
-				status=2
+					status=OK
+					pstatus green YES
+				fi
 			fi
 		fi
 	fi
 
-	if ! is_cpu_vulnerable 1; then
+	# if we have the /sys interface, don't even check is_cpu_vulnerable ourselves, the kernel already does it
-		pvulnstatus CVE-2017-5753 OK "your CPU vendor reported your CPU model as not vulnerable"
+	if [ "$sys_interface_available" = 0 ] && ! is_cpu_vulnerable 1; then
-	else
+		# override status & msg in case CPU is not vulnerable after all
-		case "$status" in
+		msg="your CPU vendor reported your CPU model as not vulnerable"
-			0) pvulnstatus CVE-2017-5753 UNK "impossible to check ${vmlinux}";;
+		status=OK
-			1) pvulnstatus CVE-2017-5753 VULN 'heuristic to be improved when official patches become available';;
-			2) pvulnstatus CVE-2017-5753 OK 'heuristic to be improved when official patches become available';;
-		esac
 	fi
+
+	# report status
+	pvulnstatus CVE-2017-5753 "$status" "$msg"
 }
 
 ###################
@@ -595,11 +644,20 @@ check_variant1()
 check_variant2()
 {
 	_info "\033[1;34mCVE-2017-5715 [branch target injection] aka 'Spectre Variant 2'\033[0m"
+
+	status=UNK
+	sys_interface_available=0
+	msg=''
+	if sys_interface_check "/sys/devices/system/cpu/vulnerabilities/spectre_v2"; then
+		# this kernel has the /sys interface, trust it over everything
+		sys_interface_available=1
+	else
 		_info "* Mitigation 1"
 		_info_nol "*   Hardware (CPU microcode) support for mitigation: "
 		if [ ! -e /dev/cpu/0/msr ]; then
 			# try to load the module ourselves (and remember it so we can rmmod it afterwards)
 			modprobe msr 2>/dev/null && insmod_msr=1
+			_debug "attempted to load module msr, ret=$insmod_msr"
 		fi
 		if [ ! -e /dev/cpu/0/msr ]; then
 			pstatus yellow UNKNOWN "couldn't read /dev/cpu/0/msr, is msr support enabled in your kernel?"
@@ -618,6 +676,7 @@ check_variant2()
 		if [ "$insmod_msr" = 1 ]; then
 			# if we used modprobe ourselves, rmmod the module
 			rmmod msr 2>/dev/null
+			_debug "attempted to unload module msr, ret=$?"
 		fi
 
 		_info_nol "*   Kernel support for IBRS: "
@@ -635,7 +694,10 @@ check_variant2()
 					pstatus green YES
 					ibrs_supported=1
 					ibrs_enabled=$(cat "$ibrs_file" 2>/dev/null)
+					_debug "ibrs: found $ibrs_file=$ibrs_enabled"
 					break
+				else
+					_debug "ibrs: file $ibrs_file doesn't exist"
 				fi
 			done
 		fi
@@ -643,6 +705,7 @@ check_variant2()
 			if grep -q spec_ctrl "$opt_map"; then
 				pstatus green YES
 				ibrs_supported=1
+				_debug "ibrs: found '*spec_ctrl*' symbol in $opt_map"
 			fi
 		fi
 		if [ "$ibrs_supported" != 1 ]; then
@@ -683,6 +746,7 @@ check_variant2()
 			if grep -q '^CONFIG_RETPOLINE=y' "$opt_config"; then
 				pstatus green YES
 				retpoline=1
+				_debug "retpoline: found "$(grep '^CONFIG_RETPOLINE' "$opt_config")" in $opt_config"
 			else
 				pstatus red NO
 			fi
@@ -725,10 +789,15 @@ check_variant2()
 		else
 			pstatus yellow UNKNOWN "couldn't find your kernel image or System.map"
 		fi
+	fi
 
-	if ! is_cpu_vulnerable 2; then
+	# if we have the /sys interface, don't even check is_cpu_vulnerable ourselves, the kernel already does it
+	if [ "$sys_interface_available" = 0 ] && ! is_cpu_vulnerable 2; then
+		# override status & msg in case CPU is not vulnerable after all
 		pvulnstatus CVE-2017-5715 OK "your CPU vendor reported your CPU model as not vulnerable"
-	elif [ "$retpoline" = 1 -a "$retpoline_compiler" = 1 ]; then
+	elif [ -z "$msg" ]; then
+		# if msg is empty, sysfs check didn't fill it, rely on our own test
+		if [ "$retpoline" = 1 -a "$retpoline_compiler" = 1 ]; then
 			pvulnstatus CVE-2017-5715 OK "retpoline mitigate the vulnerability"
 		elif [ "$opt_live" = 1 ]; then
 			if [ "$ibrs_enabled" = 1 -o "$ibrs_enabled" = 2 ]; then
@@ -743,6 +812,9 @@ check_variant2()
 				pvulnstatus CVE-2017-5715 VULN "IBRS hardware + kernel support OR kernel with retpoline are needed to mitigate the vulnerability"
 			fi
 		fi
+	else
+		pvulnstatus CVE-2017-5715 "$status" "$msg"
+	fi
 }
 
 ########################
@@ -750,12 +822,21 @@ check_variant2()
 check_variant3()
 {
 	_info "\033[1;34mCVE-2017-5754 [rogue data cache load] aka 'Meltdown' aka 'Variant 3'\033[0m"
+
+	status=UNK
+	sys_interface_available=0
+	msg=''
+	if sys_interface_check "/sys/devices/system/cpu/vulnerabilities/meltdown"; then
+		# this kernel has the /sys interface, trust it over everything
+		sys_interface_available=1
+	else
 		_info_nol "* Kernel supports Page Table Isolation (PTI): "
 		kpti_support=0
 		kpti_can_tell=0
 		if [ -n "$opt_config" ]; then
 			kpti_can_tell=1
 			if grep -Eq '^(CONFIG_PAGE_TABLE_ISOLATION|CONFIG_KAISER)=y' "$opt_config"; then
+				_debug "kpti_support: found option "$(grep -E '^(CONFIG_PAGE_TABLE_ISOLATION|CONFIG_KAISER)=y' "$opt_config")" in $opt_config"
 				kpti_support=1
 			fi
 		fi
@@ -764,6 +845,7 @@ check_variant3()
 			# so we try to find an exported symbol that is part of the PTI patch in System.map
 			kpti_can_tell=1
 			if grep -qw kpti_force_enabled "$opt_map"; then
+				_debug "kpti_support: found kpti_force_enabled in $opt_map"
 				kpti_support=1
 			fi
 		fi
@@ -775,6 +857,7 @@ check_variant3()
 				pstatus yellow UNKNOWN "missing 'strings' tool, please install it, usually it's in the binutils package"
 			else
 				if strings "$vmlinux" | grep -qw nopti; then
+					_debug "kpti_support: found nopti string in $vmlinux"
 					kpti_support=1
 				fi
 			fi
@@ -791,22 +874,31 @@ check_variant3()
 		mount_debugfs
 		_info_nol "* PTI enabled and active: "
 		if [ "$opt_live" = 1 ]; then
+			dmesg_grep="Kernel/User page tables isolation: enabled"
+			dmesg_grep="$dmesg_grep|Kernel page table isolation enabled"
+			dmesg_grep="$dmesg_grep|x86/pti: Unmapping kernel while in userspace"
 			if grep ^flags /proc/cpuinfo | grep -qw pti; then
 				# vanilla PTI patch sets the 'pti' flag in cpuinfo
+				_debug "kpti_enabled: found 'pti' flag in /proc/cpuinfo"
 				kpti_enabled=1
 			elif grep ^flags /proc/cpuinfo | grep -qw kaiser; then
 				# kernel line 4.9 sets the 'kaiser' flag in cpuinfo
+				_debug "kpti_enabled: found 'kaiser' flag in /proc/cpuinfo"
 				kpti_enabled=1
 			elif [ -e /sys/kernel/debug/x86/pti_enabled ]; then
 				# RedHat Backport creates a dedicated file, see https://access.redhat.com/articles/3311301
 				kpti_enabled=$(cat /sys/kernel/debug/x86/pti_enabled 2>/dev/null)
-		elif dmesg | grep -Eq 'Kernel/User page tables isolation: enabled|Kernel page table isolation enabled'; then
+				_debug "kpti_enabled: file /sys/kernel/debug/x86/pti_enabled exists and says: $kpti_enabled"
+			elif dmesg | grep -Eq "$dmesg_grep"; then
 				# if we can't find the flag, grep dmesg output
+				_debug "kpti_enabled: found hint in dmesg: "$(dmesg | grep -E "$dmesg_grep")
 				kpti_enabled=1
-		elif [ -r /var/log/dmesg ] && grep -Eq 'Kernel/User page tables isolation: enabled|Kernel page table isolation enabled' /var/log/dmesg; then
+			elif [ -r /var/log/dmesg ] && grep -Eq "$dmesg_grep" /var/log/dmesg; then
 				# if we can't find the flag in dmesg output, grep in /var/log/dmesg when readable
+				_debug "kpti_enabled: found hint in /var/log/dmesg: "$(grep -E "$dmesg_grep" /var/log/dmesg)
 				kpti_enabled=1
 			else
+				_debug "kpti_enabled: couldn't find any hint that PTI is enabled"
 				kpti_enabled=0
 			fi
 			if [ "$kpti_enabled" = 1 ]; then
@@ -817,22 +909,31 @@ check_variant3()
 		else
 			pstatus blue N/A "can't verify if PTI is enabled in offline mode"
 		fi
+	fi
 
-	if ! is_cpu_vulnerable 3; then
+	# if we have the /sys interface, don't even check is_cpu_vulnerable ourselves, the kernel already does it
-		pvulnstatus CVE-2017-5754 OK "your CPU vendor reported your CPU model as not vulnerable"
+	cve='CVE-2017-5754'
-	elif [ "$opt_live" = 1 ]; then
+	if [ "$sys_interface_available" = 0 ] && ! is_cpu_vulnerable 3; then
+		# override status & msg in case CPU is not vulnerable after all
+		pvulnstatus $cve OK "your CPU vendor reported your CPU model as not vulnerable"
+	elif [ -z "$msg" ]; then
+		# if msg is empty, sysfs check didn't fill it, rely on our own test
+		if [ "$opt_live" = 1 ]; then
 			if [ "$kpti_enabled" = 1 ]; then
-			pvulnstatus CVE-2017-5754 OK "PTI mitigates the vulnerability"
+				pvulnstatus $cve OK "PTI mitigates the vulnerability"
 			else
-			pvulnstatus CVE-2017-5754 VULN "PTI is needed to mitigate the vulnerability"
+				pvulnstatus $cve VULN "PTI is needed to mitigate the vulnerability"
 			fi
 		else
 			if [ "$kpti_support" = 1 ]; then
-			pvulnstatus CVE-2017-5754 OK "offline mode: PTI will mitigate the vulnerability if enabled at runtime"
+				pvulnstatus $cve OK "offline mode: PTI will mitigate the vulnerability if enabled at runtime"
 			else
-			pvulnstatus CVE-2017-5754 VULN "PTI is needed to mitigate the vulnerability"
+				pvulnstatus $cve VULN "PTI is needed to mitigate the vulnerability"
 			fi
 		fi
+	else
+		pvulnstatus $cve "$status" "$msg"
+	fi
 }
 
 # now run the checks the user asked for