From 869feb33881ac0ee6f95fb3baa7eeb870c429c64 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Zbigniew=20J=C4=99drzejewski-Szmek?= Date: Wed, 2 Nov 2016 11:58:18 -0400 Subject: [PATCH 1/4] analyze: add syscall-filter verb This should make it easier for users to understand what each filter means as the list of syscalls is updated in subsequent systemd versions. --- man/systemd-analyze.xml | 11 +++ man/systemd.exec.xml | 10 ++- shell-completion/bash/systemd-analyze | 6 ++ shell-completion/zsh/_systemd-analyze | 1 + src/analyze/analyze.c | 98 ++++++++++++++++++++------- 5 files changed, 101 insertions(+), 25 deletions(-) diff --git a/man/systemd-analyze.xml b/man/systemd-analyze.xml index 8fa7cd3329..634e16b5f4 100644 --- a/man/systemd-analyze.xml +++ b/man/systemd-analyze.xml @@ -101,6 +101,12 @@ set-log-target TARGET + + systemd-analyze + OPTIONS + syscall-filter + SET... + systemd-analyze OPTIONS @@ -181,6 +187,11 @@ , described in systemd1). + systemd-analyze syscall-filter SET... + will list system calls contained in the specified system call set SET, + or all known sets if no sets are specified. Argument SET must include + the @ prefix. + systemd-analyze verify will load unit files and print warnings if any errors are detected. Files specified on the command line will be loaded, but also any other units referenced by them. The full unit search path is diff --git a/man/systemd.exec.xml b/man/systemd.exec.xml index 3c350df11f..0de263cb07 100644 --- a/man/systemd.exec.xml +++ b/man/systemd.exec.xml @@ -1373,8 +1373,13 @@ - Note that as new system calls are added to the kernel, additional system calls might be added to the groups - above, so the contents of the sets may change between systemd versions. + Note, that as new system calls are added to the kernel, additional system calls might be + added to the groups above. Contents of the sets may also change between systemd + versions. In addition, the list of system calls depends on the kernel version and + architecture for which systemd was compiled. Use + systemd-analyze syscall-filter to list the actual list of system calls in + each filter. + It is recommended to combine the file system namespacing related options with SystemCallFilter=~@mount, in order to prohibit the unit's processes to undo the @@ -1844,6 +1849,7 @@ systemd1, systemctl1, + systemd-analyze1, journalctl8, systemd.unit5, systemd.service5, diff --git a/shell-completion/bash/systemd-analyze b/shell-completion/bash/systemd-analyze index 7a5f46ba1d..92ff13d316 100644 --- a/shell-completion/bash/systemd-analyze +++ b/shell-completion/bash/systemd-analyze @@ -45,6 +45,7 @@ _systemd_analyze() { [DOT]='dot' [LOG_LEVEL]='set-log-level' [VERIFY]='verify' + [SECCOMP_FILTER]='syscall-filter' ) _init_completion || return @@ -100,6 +101,11 @@ _systemd_analyze() { comps='debug info notice warning err crit alert emerg' fi + elif __contains_word "$verb" ${VERBS[SECCOMP_FILTER]}; then + if [[ $cur = -* ]]; then + comps='--help --version' + fi + elif __contains_word "$verb" ${VERBS[VERIFY]}; then if [[ $cur = -* ]]; then comps='--help --version --system --user --man' diff --git a/shell-completion/zsh/_systemd-analyze b/shell-completion/zsh/_systemd-analyze index efafddc686..0e670034c1 100644 --- a/shell-completion/zsh/_systemd-analyze +++ b/shell-completion/zsh/_systemd-analyze @@ -21,6 +21,7 @@ _systemd_analyze_command(){ 'dot:Dump dependency graph (in dot(1) format)' 'dump:Dump server status' 'set-log-level:Set systemd log threshold' + 'syscall-filter:List syscalls in seccomp filter' 'verify:Check unit files for correctness' ) diff --git a/src/analyze/analyze.c b/src/analyze/analyze.c index cbf9354a7a..0495fca17d 100644 --- a/src/analyze/analyze.c +++ b/src/analyze/analyze.c @@ -36,6 +36,7 @@ #include "log.h" #include "pager.h" #include "parse-util.h" +#include "seccomp-util.h" #include "special.h" #include "strv.h" #include "strxcpyx.h" @@ -1275,36 +1276,85 @@ static int set_log_target(sd_bus *bus, char **args) { return 0; } +static void dump_syscall_filter(const SyscallFilterSet *set) { + const char *syscall; + + printf("%s\n", set->name); + NULSTR_FOREACH(syscall, set->value) + printf(" %s\n", syscall); +} + +static int dump_syscall_filters(char** names) { + bool first = true; + + pager_open(arg_no_pager, false); + + if (strv_isempty(names)) { + int i; + + for (i = 0; i < _SYSCALL_FILTER_SET_MAX; i++) { + if (!first) + puts(""); + dump_syscall_filter(syscall_filter_sets + i); + first = false; + } + } else { + char **name; + + STRV_FOREACH(name, names) { + const SyscallFilterSet *set; + + if (!first) + puts(""); + + set = syscall_filter_set_find(*name); + if (!set) { + /* make sure the error appears below normal output */ + fflush(stdout); + + log_error("Filter set \"%s\" not found.", *name); + return -ENOENT; + } + + dump_syscall_filter(set); + first = false; + } + } + + return 0; +} + static void help(void) { pager_open(arg_no_pager, false); printf("%s [OPTIONS...] {COMMAND} ...\n\n" "Profile systemd, show unit dependencies, check unit files.\n\n" - " -h --help Show this help\n" - " --version Show package version\n" - " --no-pager Do not pipe output into a pager\n" - " --system Operate on system systemd instance\n" - " --user Operate on user systemd instance\n" - " -H --host=[USER@]HOST Operate on remote host\n" - " -M --machine=CONTAINER Operate on local container\n" - " --order Show only order in the graph\n" - " --require Show only requirement in the graph\n" - " --from-pattern=GLOB Show only origins in the graph\n" - " --to-pattern=GLOB Show only destinations in the graph\n" - " --fuzz=SECONDS Also print also services which finished SECONDS\n" - " earlier than the latest in the branch\n" - " --man[=BOOL] Do [not] check for existence of man pages\n\n" + " -h --help Show this help\n" + " --version Show package version\n" + " --no-pager Do not pipe output into a pager\n" + " --system Operate on system systemd instance\n" + " --user Operate on user systemd instance\n" + " -H --host=[USER@]HOST Operate on remote host\n" + " -M --machine=CONTAINER Operate on local container\n" + " --order Show only order in the graph\n" + " --require Show only requirement in the graph\n" + " --from-pattern=GLOB Show only origins in the graph\n" + " --to-pattern=GLOB Show only destinations in the graph\n" + " --fuzz=SECONDS Also print also services which finished SECONDS\n" + " earlier than the latest in the branch\n" + " --man[=BOOL] Do [not] check for existence of man pages\n\n" "Commands:\n" - " time Print time spent in the kernel\n" - " blame Print list of running units ordered by time to init\n" - " critical-chain Print a tree of the time critical chain of units\n" - " plot Output SVG graphic showing service initialization\n" - " dot Output dependency graph in dot(1) format\n" - " set-log-level LEVEL Set logging threshold for manager\n" - " set-log-target TARGET Set logging target for manager\n" - " dump Output state serialization of service manager\n" - " verify FILE... Check unit files for correctness\n" + " time Print time spent in the kernel\n" + " blame Print list of running units ordered by time to init\n" + " critical-chain Print a tree of the time critical chain of units\n" + " plot Output SVG graphic showing service initialization\n" + " dot Output dependency graph in dot(1) format\n" + " set-log-level LEVEL Set logging threshold for manager\n" + " set-log-target TARGET Set logging target for manager\n" + " dump Output state serialization of service manager\n" + " syscall-filter [NAME...] Print list of syscalls in seccomp filter\n" + " verify FILE... Check unit files for correctness\n" , program_invocation_short_name); /* When updating this list, including descriptions, apply @@ -1471,6 +1521,8 @@ int main(int argc, char *argv[]) { r = set_log_level(bus, argv+optind+1); else if (streq(argv[optind], "set-log-target")) r = set_log_target(bus, argv+optind+1); + else if (streq(argv[optind], "syscall-filter")) + r = dump_syscall_filters(argv+optind+1); else log_error("Unknown operation '%s'.", argv[optind]); } From 40eb6a801427e28a237d32869639fffc8436c930 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Zbigniew=20J=C4=99drzejewski-Szmek?= Date: Wed, 2 Nov 2016 12:01:04 -0400 Subject: [PATCH 2/4] seccomp-util: move @default to the first position Now that the list is user-visible, @default should be first. --- src/shared/seccomp-util.c | 36 ++++++++++++++++++------------------ src/shared/seccomp-util.h | 3 ++- 2 files changed, 20 insertions(+), 19 deletions(-) diff --git a/src/shared/seccomp-util.c b/src/shared/seccomp-util.c index c9b24f1065..325dcc866e 100644 --- a/src/shared/seccomp-util.c +++ b/src/shared/seccomp-util.c @@ -217,6 +217,24 @@ bool is_seccomp_available(void) { } const SyscallFilterSet syscall_filter_sets[_SYSCALL_FILTER_SET_MAX] = { + [SYSCALL_FILTER_SET_DEFAULT] = { + /* Default list: the most basic of operations */ + .name = "@default", + .value = + "clock_getres\0" + "clock_gettime\0" + "clock_nanosleep\0" + "execve\0" + "exit\0" + "exit_group\0" + "getrlimit\0" /* make sure processes can query stack size and such */ + "gettimeofday\0" + "nanosleep\0" + "pause\0" + "rt_sigreturn\0" + "sigreturn\0" + "time\0" + }, [SYSCALL_FILTER_SET_BASIC_IO] = { /* Basic IO */ .name = "@basic-io", @@ -270,24 +288,6 @@ const SyscallFilterSet syscall_filter_sets[_SYSCALL_FILTER_SET_MAX] = { #endif "sys_debug_setcontext\0" }, - [SYSCALL_FILTER_SET_DEFAULT] = { - /* Default list: the most basic of operations */ - .name = "@default", - .value = - "clock_getres\0" - "clock_gettime\0" - "clock_nanosleep\0" - "execve\0" - "exit\0" - "exit_group\0" - "getrlimit\0" /* make sure processes can query stack size and such */ - "gettimeofday\0" - "nanosleep\0" - "pause\0" - "rt_sigreturn\0" - "sigreturn\0" - "time\0" - }, [SYSCALL_FILTER_SET_IO_EVENT] = { /* Event loop use */ .name = "@io-event", diff --git a/src/shared/seccomp-util.h b/src/shared/seccomp-util.h index 8e209efef2..ce7417b0ba 100644 --- a/src/shared/seccomp-util.h +++ b/src/shared/seccomp-util.h @@ -38,11 +38,12 @@ typedef struct SyscallFilterSet { } SyscallFilterSet; enum { + /* Please leave DEFAULT first, but sort the rest alphabetically */ + SYSCALL_FILTER_SET_DEFAULT, SYSCALL_FILTER_SET_BASIC_IO, SYSCALL_FILTER_SET_CLOCK, SYSCALL_FILTER_SET_CPU_EMULATION, SYSCALL_FILTER_SET_DEBUG, - SYSCALL_FILTER_SET_DEFAULT, SYSCALL_FILTER_SET_IO_EVENT, SYSCALL_FILTER_SET_IPC, SYSCALL_FILTER_SET_KEYRING, From d5efc18b609ad8cfe63325c8b4f82264e56eadf3 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Zbigniew=20J=C4=99drzejewski-Szmek?= Date: Wed, 2 Nov 2016 12:24:34 -0400 Subject: [PATCH 3/4] seccomp-util, analyze: export comments as a help string Just to make the whole thing easier for users. --- man/systemd.exec.xml | 6 +++--- src/analyze/analyze.c | 1 + src/shared/seccomp-util.c | 34 ++++++++++++++++++---------------- src/shared/seccomp-util.h | 1 + 4 files changed, 23 insertions(+), 19 deletions(-) diff --git a/man/systemd.exec.xml b/man/systemd.exec.xml index 0de263cb07..dd574df82a 100644 --- a/man/systemd.exec.xml +++ b/man/systemd.exec.xml @@ -1339,11 +1339,11 @@ @module - Kernel module control (init_module2, delete_module2 and related calls) + Loading and unloading of kernel modules (init_module2, delete_module2 and related calls) @mount - File system mounting and unmounting (mount2, chroot2, and related calls) + Mounting and unmounting of file systems (mount2, chroot2, and related calls) @network-io @@ -1359,7 +1359,7 @@ @process - Process control, execution, namespaces (clone2, kill2, namespaces7, … + Process control, execution, namespaceing operations (clone2, kill2, namespaces7, … @raw-io diff --git a/src/analyze/analyze.c b/src/analyze/analyze.c index 0495fca17d..83e1cdf11d 100644 --- a/src/analyze/analyze.c +++ b/src/analyze/analyze.c @@ -1280,6 +1280,7 @@ static void dump_syscall_filter(const SyscallFilterSet *set) { const char *syscall; printf("%s\n", set->name); + printf(" # %s\n", set->help); NULSTR_FOREACH(syscall, set->value) printf(" %s\n", syscall); } diff --git a/src/shared/seccomp-util.c b/src/shared/seccomp-util.c index 325dcc866e..fc1f6b68f2 100644 --- a/src/shared/seccomp-util.c +++ b/src/shared/seccomp-util.c @@ -218,8 +218,8 @@ bool is_seccomp_available(void) { const SyscallFilterSet syscall_filter_sets[_SYSCALL_FILTER_SET_MAX] = { [SYSCALL_FILTER_SET_DEFAULT] = { - /* Default list: the most basic of operations */ .name = "@default", + .help = "System calls that are always permitted", .value = "clock_getres\0" "clock_gettime\0" @@ -236,8 +236,8 @@ const SyscallFilterSet syscall_filter_sets[_SYSCALL_FILTER_SET_MAX] = { "time\0" }, [SYSCALL_FILTER_SET_BASIC_IO] = { - /* Basic IO */ .name = "@basic-io", + .help = "Basic IO", .value = "close\0" "dup2\0" @@ -254,8 +254,8 @@ const SyscallFilterSet syscall_filter_sets[_SYSCALL_FILTER_SET_MAX] = { "writev\0" }, [SYSCALL_FILTER_SET_CLOCK] = { - /* Clock */ .name = "@clock", + .help = "Change the system time", .value = "adjtimex\0" "clock_adjtime\0" @@ -264,8 +264,8 @@ const SyscallFilterSet syscall_filter_sets[_SYSCALL_FILTER_SET_MAX] = { "stime\0" }, [SYSCALL_FILTER_SET_CPU_EMULATION] = { - /* CPU emulation calls */ .name = "@cpu-emulation", + .help = "System calls for CPU emulation functionality", .value = "modify_ldt\0" "subpage_prot\0" @@ -274,8 +274,8 @@ const SyscallFilterSet syscall_filter_sets[_SYSCALL_FILTER_SET_MAX] = { "vm86old\0" }, [SYSCALL_FILTER_SET_DEBUG] = { - /* Debugging/Performance Monitoring/Tracing */ .name = "@debug", + .help = "Debugging, performance monitoring and tracing functionality", .value = "lookup_dcookie\0" "perf_event_open\0" @@ -289,8 +289,8 @@ const SyscallFilterSet syscall_filter_sets[_SYSCALL_FILTER_SET_MAX] = { "sys_debug_setcontext\0" }, [SYSCALL_FILTER_SET_IO_EVENT] = { - /* Event loop use */ .name = "@io-event", + .help = "Event loop system calls", .value = "_newselect\0" "epoll_create1\0" @@ -308,9 +308,10 @@ const SyscallFilterSet syscall_filter_sets[_SYSCALL_FILTER_SET_MAX] = { "select\0" }, [SYSCALL_FILTER_SET_IPC] = { - /* Message queues, SYSV IPC or other IPC */ .name = "@ipc", - .value = "ipc\0" + .help = "SysV IPC, POSIX Message Queues or other IPC", + .value = + "ipc\0" "memfd_create\0" "mq_getsetattr\0" "mq_notify\0" @@ -336,24 +337,24 @@ const SyscallFilterSet syscall_filter_sets[_SYSCALL_FILTER_SET_MAX] = { "shmget\0" }, [SYSCALL_FILTER_SET_KEYRING] = { - /* Keyring */ .name = "@keyring", + .help = "Kernel keyring access", .value = "add_key\0" "keyctl\0" "request_key\0" }, [SYSCALL_FILTER_SET_MODULE] = { - /* Kernel module control */ .name = "@module", + .help = "Loading and unloading of kernel modules", .value = "delete_module\0" "finit_module\0" "init_module\0" }, [SYSCALL_FILTER_SET_MOUNT] = { - /* Mounting */ .name = "@mount", + .help = "Mounting and unmounting of file systems", .value = "chroot\0" "mount\0" @@ -362,8 +363,8 @@ const SyscallFilterSet syscall_filter_sets[_SYSCALL_FILTER_SET_MAX] = { "umount\0" }, [SYSCALL_FILTER_SET_NETWORK_IO] = { - /* Network or Unix socket IO, should not be needed if not network facing */ .name = "@network-io", + .help = "Network or Unix socket IO, should not be needed if not network facing", .value = "accept4\0" "accept\0" @@ -388,8 +389,9 @@ const SyscallFilterSet syscall_filter_sets[_SYSCALL_FILTER_SET_MAX] = { "socketpair\0" }, [SYSCALL_FILTER_SET_OBSOLETE] = { - /* Unusual, obsolete or unimplemented, some unknown even to libseccomp */ + /* some unknown even to libseccomp */ .name = "@obsolete", + .help = "Unusual, obsolete or unimplemented system calls", .value = "_sysctl\0" "afs_syscall\0" @@ -417,8 +419,8 @@ const SyscallFilterSet syscall_filter_sets[_SYSCALL_FILTER_SET_MAX] = { "vserver\0" }, [SYSCALL_FILTER_SET_PRIVILEGED] = { - /* Nice grab-bag of all system calls which need superuser capabilities */ .name = "@privileged", + .help = "All system calls which need super-user capabilities", .value = "@clock\0" "@module\0" @@ -459,8 +461,8 @@ const SyscallFilterSet syscall_filter_sets[_SYSCALL_FILTER_SET_MAX] = { "vhangup\0" }, [SYSCALL_FILTER_SET_PROCESS] = { - /* Process control, execution, namespaces */ .name = "@process", + .help = "Process control, execution, namespaceing operations", .value = "arch_prctl\0" "clone\0" @@ -475,8 +477,8 @@ const SyscallFilterSet syscall_filter_sets[_SYSCALL_FILTER_SET_MAX] = { "vfork\0" }, [SYSCALL_FILTER_SET_RAW_IO] = { - /* Raw I/O ports */ .name = "@raw-io", + .help = "Raw I/O port access", .value = "ioperm\0" "iopl\0" diff --git a/src/shared/seccomp-util.h b/src/shared/seccomp-util.h index ce7417b0ba..f0b9f455ab 100644 --- a/src/shared/seccomp-util.h +++ b/src/shared/seccomp-util.h @@ -34,6 +34,7 @@ bool is_seccomp_available(void); typedef struct SyscallFilterSet { const char *name; + const char *help; const char *value; } SyscallFilterSet; From 1720590bfd7618efa10891f956bf4b265311be04 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Zbigniew=20J=C4=99drzejewski-Szmek?= Date: Thu, 3 Nov 2016 14:33:49 -0400 Subject: [PATCH 4/4] analyze: fix build w/o seccomp --- src/analyze/analyze.c | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/src/analyze/analyze.c b/src/analyze/analyze.c index 83e1cdf11d..f744a84501 100644 --- a/src/analyze/analyze.c +++ b/src/analyze/analyze.c @@ -1276,6 +1276,7 @@ static int set_log_target(sd_bus *bus, char **args) { return 0; } +#ifdef HAVE_SECCOMP static void dump_syscall_filter(const SyscallFilterSet *set) { const char *syscall; @@ -1325,6 +1326,13 @@ static int dump_syscall_filters(char** names) { return 0; } +#else +static int dump_syscall_filters(char** names) { + log_error("Not compiled with syscall filters, sorry."); + return -EOPNOTSUPP; +} +#endif + static void help(void) { pager_open(arg_no_pager, false);