[thirdparty/systemd.git] / src / journal / journald-native.c

/***
  This file is part of systemd.

  Copyright 2011 Lennart Poettering

  systemd is free software; you can redistribute it and/or modify it
  under the terms of the GNU Lesser General Public License as published by
  the Free Software Foundation; either version 2.1 of the License, or
  (at your option) any later version.

  systemd is distributed in the hope that it will be useful, but
  WITHOUT ANY WARRANTY; without even the implied warranty of
  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  Lesser General Public License for more details.

  You should have received a copy of the GNU Lesser General Public License
  along with systemd; If not, see <http://www.gnu.org/licenses/>.
***/

#include <stddef.h>
#include <sys/epoll.h>
#include <sys/mman.h>
#include <sys/statvfs.h>
#include <unistd.h>

#include "alloc-util.h"
#include "fd-util.h"
#include "fs-util.h"
#include "io-util.h"
#include "journal-importer.h"
#include "journald-console.h"
#include "journald-kmsg.h"
#include "journald-native.h"
#include "journald-server.h"
#include "journald-syslog.h"
#include "journald-wall.h"
#include "memfd-util.h"
#include "parse-util.h"
#include "path-util.h"
#include "selinux-util.h"
#include "socket-util.h"
#include "string-util.h"

bool valid_user_field(const char *p, size_t l, bool allow_protected) {
        const char *a;

        /* We kinda enforce POSIX syntax recommendations for
           environment variables here, but make a couple of additional
           requirements.

           http://pubs.opengroup.org/onlinepubs/000095399/basedefs/xbd_chap08.html */

        /* No empty field names */
        if (l <= 0)
                return false;

        /* Don't allow names longer than 64 chars */
        if (l > 64)
                return false;

        /* Variables starting with an underscore are protected */
        if (!allow_protected && p[0] == '_')
                return false;

        /* Don't allow digits as first character */
        if (p[0] >= '0' && p[0] <= '9')
                return false;

        /* Only allow A-Z0-9 and '_' */
        for (a = p; a < p + l; a++)
                if ((*a < 'A' || *a > 'Z') &&
                    (*a < '0' || *a > '9') &&
                    *a != '_')
                        return false;

        return true;
}

static bool allow_object_pid(const struct ucred *ucred) {
        return ucred && ucred->uid == 0;
}

void server_process_native_message(
                Server *s,
                const void *buffer, size_t buffer_size,
                const struct ucred *ucred,
                const struct timeval *tv,
                const char *label, size_t label_len) {

        struct iovec *iovec = NULL;
        unsigned n = 0, j, tn = (unsigned) -1;
        const char *p;
        size_t remaining, m = 0, entry_size = 0;
        int priority = LOG_INFO;
        char *identifier = NULL, *message = NULL;
        pid_t object_pid = 0;

        assert(s);
        assert(buffer || buffer_size == 0);

        p = buffer;
        remaining = buffer_size;

        while (remaining > 0) {
                const char *e, *q;

                e = memchr(p, '\n', remaining);

                if (!e) {
                        /* Trailing noise, let's ignore it, and flush what we collected */
                        log_debug("Received message with trailing noise, ignoring.");
                        break;
                }

                if (e == p) {
                        /* Entry separator */

                        if (entry_size + n + 1 > ENTRY_SIZE_MAX) { /* data + separators + trailer */
                                log_debug("Entry is too big with %u properties and %zu bytes, ignoring.", n, entry_size);
                                continue;
                        }

                        server_dispatch_message(s, iovec, n, m, ucred, tv, label, label_len, NULL, priority, object_pid);
                        n = 0;
                        priority = LOG_INFO;
                        entry_size = 0;

                        p++;
                        remaining--;
                        continue;
                }

                if (*p == '.' || *p == '#') {
                        /* Ignore control commands for now, and
                         * comments too. */
                        remaining -= (e - p) + 1;
                        p = e + 1;
                        continue;
                }

                /* A property follows */

                /* n existing properties, 1 new, +1 for _TRANSPORT */
                if (!GREEDY_REALLOC(iovec, m, n + 2 + N_IOVEC_META_FIELDS + N_IOVEC_OBJECT_FIELDS)) {
                        log_oom();
                        break;
                }

                q = memchr(p, '=', e - p);
                if (q) {
                        if (valid_user_field(p, q - p, false)) {
                                size_t l;

                                l = e - p;

                                /* If the field name starts with an
                                 * underscore, skip the variable,
                                 * since that indidates a trusted
                                 * field */
                                iovec[n].iov_base = (char*) p;
                                iovec[n].iov_len = l;
                                entry_size += iovec[n].iov_len;
                                n++;

                                /* We need to determine the priority
                                 * of this entry for the rate limiting
                                 * logic */
                                if (l == 10 &&
                                    startswith(p, "PRIORITY=") &&
                                    p[9] >= '0' && p[9] <= '9')
                                        priority = (priority & LOG_FACMASK) | (p[9] - '0');

                                else if (l == 17 &&
                                         startswith(p, "SYSLOG_FACILITY=") &&
                                         p[16] >= '0' && p[16] <= '9')
                                        priority = (priority & LOG_PRIMASK) | ((p[16] - '0') << 3);

                                else if (l == 18 &&
                                         startswith(p, "SYSLOG_FACILITY=") &&
                                         p[16] >= '0' && p[16] <= '9' &&
                                         p[17] >= '0' && p[17] <= '9')
                                        priority = (priority & LOG_PRIMASK) | (((p[16] - '0')*10 + (p[17] - '0')) << 3);

                                else if (l >= 19 &&
                                         startswith(p, "SYSLOG_IDENTIFIER=")) {
                                        char *t;

                                        t = strndup(p + 18, l - 18);
                                        if (t) {
                                                free(identifier);
                                                identifier = t;
                                        }

                                } else if (l >= 8 &&
                                           startswith(p, "MESSAGE=")) {
                                        char *t;

                                        t = strndup(p + 8, l - 8);
                                        if (t) {
                                                free(message);
                                                message = t;
                                        }

                                } else if (l > strlen("OBJECT_PID=") &&
                                           l < strlen("OBJECT_PID=")  + DECIMAL_STR_MAX(pid_t) &&
                                           startswith(p, "OBJECT_PID=") &&
                                           allow_object_pid(ucred)) {
                                        char buf[DECIMAL_STR_MAX(pid_t)];
                                        memcpy(buf, p + strlen("OBJECT_PID="), l - strlen("OBJECT_PID="));
                                        buf[l-strlen("OBJECT_PID=")] = '\0';

                                        /* ignore error */
                                        parse_pid(buf, &object_pid);
                                }
                        }

                        remaining -= (e - p) + 1;
                        p = e + 1;
                        continue;
                } else {
                        le64_t l_le;
                        uint64_t l;
                        char *k;

                        if (remaining < e - p + 1 + sizeof(uint64_t) + 1) {
                                log_debug("Failed to parse message, ignoring.");
                                break;
                        }

                        memcpy(&l_le, e + 1, sizeof(uint64_t));
                        l = le64toh(l_le);

                        if (l > DATA_SIZE_MAX) {
                                log_debug("Received binary data block of %"PRIu64" bytes is too large, ignoring.", l);
                                break;
                        }

                        if ((uint64_t) remaining < e - p + 1 + sizeof(uint64_t) + l + 1 ||
                            e[1+sizeof(uint64_t)+l] != '\n') {
                                log_debug("Failed to parse message, ignoring.");
                                break;
                        }

                        k = malloc((e - p) + 1 + l);
                        if (!k) {
                                log_oom();
                                break;
                        }

                        memcpy(k, p, e - p);
                        k[e - p] = '=';
                        memcpy(k + (e - p) + 1, e + 1 + sizeof(uint64_t), l);

                        if (valid_user_field(p, e - p, false)) {
                                iovec[n].iov_base = k;
                                iovec[n].iov_len = (e - p) + 1 + l;
                                entry_size += iovec[n].iov_len;
                                n++;
                        } else
                                free(k);

                        remaining -= (e - p) + 1 + sizeof(uint64_t) + l + 1;
                        p = e + 1 + sizeof(uint64_t) + l + 1;
                }
        }

        if (n <= 0)
                goto finish;

        tn = n++;
        IOVEC_SET_STRING(iovec[tn], "_TRANSPORT=journal");
        entry_size += strlen("_TRANSPORT=journal");

        if (entry_size + n + 1 > ENTRY_SIZE_MAX) { /* data + separators + trailer */
                log_debug("Entry is too big with %u properties and %zu bytes, ignoring.",
                          n, entry_size);
                goto finish;
        }

        if (message) {
                if (s->forward_to_syslog)
                        server_forward_syslog(s, syslog_fixup_facility(priority), identifier, message, ucred, tv);

                if (s->forward_to_kmsg)
                        server_forward_kmsg(s, priority, identifier, message, ucred);

                if (s->forward_to_console)
                        server_forward_console(s, priority, identifier, message, ucred);

                if (s->forward_to_wall)
                        server_forward_wall(s, priority, identifier, message, ucred);
        }

        server_dispatch_message(s, iovec, n, m, ucred, tv, label, label_len, NULL, priority, object_pid);

finish:
        for (j = 0; j < n; j++)  {
                if (j == tn)
                        continue;

                if (iovec[j].iov_base < buffer ||
                    (const uint8_t*) iovec[j].iov_base >= (const uint8_t*) buffer + buffer_size)
                        free(iovec[j].iov_base);
        }

        free(iovec);
        free(identifier);
        free(message);
}

void server_process_native_file(
                Server *s,
                int fd,
                const struct ucred *ucred,
                const struct timeval *tv,
                const char *label, size_t label_len) {

        struct stat st;
        bool sealed;
        int r;

        /* Data is in the passed fd, since it didn't fit in a
         * datagram. */

        assert(s);
        assert(fd >= 0);

        /* If it's a memfd, check if it is sealed. If so, we can just
         * use map it and use it, and do not need to copy the data
         * out. */
        sealed = memfd_get_sealed(fd) > 0;

        if (!sealed && (!ucred || ucred->uid != 0)) {
                _cleanup_free_ char *sl = NULL, *k = NULL;
                const char *e;

                /* If this is not a sealed memfd, and the peer is unknown or
                 * unprivileged, then verify the path. */

                if (asprintf(&sl, "/proc/self/fd/%i", fd) < 0) {
                        log_oom();
                        return;
                }

                r = readlink_malloc(sl, &k);
                if (r < 0) {
                        log_error_errno(r, "readlink(%s) failed: %m", sl);
                        return;
                }

                e = path_startswith(k, "/dev/shm/");
                if (!e)
                        e = path_startswith(k, "/tmp/");
                if (!e)
                        e = path_startswith(k, "/var/tmp/");
                if (!e) {
                        log_error("Received file outside of allowed directories. Refusing.");
                        return;
                }

                if (!filename_is_valid(e)) {
                        log_error("Received file in subdirectory of allowed directories. Refusing.");
                        return;
                }
        }

        if (fstat(fd, &st) < 0) {
                log_error_errno(errno, "Failed to stat passed file, ignoring: %m");
                return;
        }

        if (!S_ISREG(st.st_mode)) {
                log_error("File passed is not regular. Ignoring.");
                return;
        }

        if (st.st_size <= 0)
                return;

        if (st.st_size > ENTRY_SIZE_MAX) {
                log_error("File passed too large. Ignoring.");
                return;
        }

        if (sealed) {
                void *p;
                size_t ps;

                /* The file is sealed, we can just map it and use it. */

                ps = PAGE_ALIGN(st.st_size);
                p = mmap(NULL, ps, PROT_READ, MAP_PRIVATE, fd, 0);
                if (p == MAP_FAILED) {
                        log_error_errno(errno, "Failed to map memfd, ignoring: %m");
                        return;
                }

                server_process_native_message(s, p, st.st_size, ucred, tv, label, label_len);
                assert_se(munmap(p, ps) >= 0);
        } else {
                _cleanup_free_ void *p = NULL;
                struct statvfs vfs;
                ssize_t n;

                if (fstatvfs(fd, &vfs) < 0) {
                        log_error_errno(errno, "Failed to stat file system of passed file, ignoring: %m");
                        return;
                }

                /* Refuse operating on file systems that have
                 * mandatory locking enabled, see:
                 *
                 * https://github.com/systemd/systemd/issues/1822
                 */
                if (vfs.f_flag & ST_MANDLOCK) {
                        log_error("Received file descriptor from file system with mandatory locking enable, refusing.");
                        return;
                }

                /* Make the fd non-blocking. On regular files this has
                 * the effect of bypassing mandatory locking. Of
                 * course, this should normally not be necessary given
                 * the check above, but let's better be safe than
                 * sorry, after all NFS is pretty confusing regarding
                 * file system flags, and we better don't trust it,
                 * and so is SMB. */
                r = fd_nonblock(fd, true);
                if (r < 0) {
                        log_error_errno(r, "Failed to make fd non-blocking, ignoring: %m");
                        return;
                }

                /* The file is not sealed, we can't map the file here, since
                 * clients might then truncate it and trigger a SIGBUS for
                 * us. So let's stupidly read it */

                p = malloc(st.st_size);
                if (!p) {
                        log_oom();
                        return;
                }

                n = pread(fd, p, st.st_size, 0);
                if (n < 0)
                        log_error_errno(errno, "Failed to read file, ignoring: %m");
                else if (n > 0)
                        server_process_native_message(s, p, n, ucred, tv, label, label_len);
        }
}

int server_open_native_socket(Server*s) {

        static const union sockaddr_union sa = {
                .un.sun_family = AF_UNIX,
                .un.sun_path = "/run/systemd/journal/socket",
        };
        static const int one = 1;
        int r;

        assert(s);

        if (s->native_fd < 0) {
                s->native_fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC|SOCK_NONBLOCK, 0);
                if (s->native_fd < 0)
                        return log_error_errno(errno, "socket() failed: %m");

                (void) unlink(sa.un.sun_path);

                r = bind(s->native_fd, &sa.sa, SOCKADDR_UN_LEN(sa.un));
                if (r < 0)
                        return log_error_errno(errno, "bind(%s) failed: %m", sa.un.sun_path);

                (void) chmod(sa.un.sun_path, 0666);
        } else
                fd_nonblock(s->native_fd, 1);

        r = setsockopt(s->native_fd, SOL_SOCKET, SO_PASSCRED, &one, sizeof(one));
        if (r < 0)
                return log_error_errno(errno, "SO_PASSCRED failed: %m");

#ifdef HAVE_SELINUX
        if (mac_selinux_have()) {
                r = setsockopt(s->native_fd, SOL_SOCKET, SO_PASSSEC, &one, sizeof(one));
                if (r < 0)
                        log_warning_errno(errno, "SO_PASSSEC failed: %m");
        }
#endif

        r = setsockopt(s->native_fd, SOL_SOCKET, SO_TIMESTAMP, &one, sizeof(one));
        if (r < 0)
                return log_error_errno(errno, "SO_TIMESTAMP failed: %m");

        r = sd_event_add_io(s->event, &s->native_event_source, s->native_fd, EPOLLIN, server_process_datagram, s);
        if (r < 0)
                return log_error_errno(r, "Failed to add native server fd to event loop: %m");

        r = sd_event_source_set_priority(s->native_event_source, SD_EVENT_PRIORITY_NORMAL+5);
        if (r < 0)
                return log_error_errno(r, "Failed to adjust native event source priority: %m");

        return 0;
}
Commit	Line	Data
0153028a LP	1	/***
	2	This file is part of systemd.
	3
	4	Copyright 2011 Lennart Poettering
	5
	6	systemd is free software; you can redistribute it and/or modify it
	7	under the terms of the GNU Lesser General Public License as published by
	8	the Free Software Foundation; either version 2.1 of the License, or
	9	(at your option) any later version.
	10
	11	systemd is distributed in the hope that it will be useful, but
	12	WITHOUT ANY WARRANTY; without even the implied warranty of
	13	MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	14	Lesser General Public License for more details.
	15
	16	You should have received a copy of the GNU Lesser General Public License
	17	along with systemd; If not, see <http://www.gnu.org/licenses/>.
	18	***/
	19
4871690d	20	#include <stddef.h>
0153028a	21	#include <sys/epoll.h>
c79e98ea	22	#include <sys/mman.h>
1e603a48	23	#include <sys/statvfs.h>
07630cea	24	#include <unistd.h>
0153028a	25
b5efdb8a	26	#include "alloc-util.h"
3ffd4af2	27	#include "fd-util.h"
f4f15635	28	#include "fs-util.h"
afc5dbf3	29	#include "io-util.h"
b18453ed	30	#include "journal-importer.h"
0153028a	31	#include "journald-console.h"
07630cea	32	#include "journald-kmsg.h"
3ffd4af2	33	#include "journald-native.h"
07630cea	34	#include "journald-server.h"
0153028a	35	#include "journald-syslog.h"
40b71e89	36	#include "journald-wall.h"
a09abc4a	37	#include "memfd-util.h"
6bedfcbb	38	#include "parse-util.h"
07630cea LP	39	#include "path-util.h"
	40	#include "selinux-util.h"
	41	#include "socket-util.h"
	42	#include "string-util.h"
0153028a	43
d18d46ec	44	bool valid_user_field(const char *p, size_t l, bool allow_protected) {
0153028a LP	45	const char *a;
	46
	47	/* We kinda enforce POSIX syntax recommendations for
	48	environment variables here, but make a couple of additional
	49	requirements.
	50
	51	http://pubs.opengroup.org/onlinepubs/000095399/basedefs/xbd_chap08.html */
	52
	53	/* No empty field names */
	54	if (l <= 0)
	55	return false;
	56
	57	/* Don't allow names longer than 64 chars */
	58	if (l > 64)
	59	return false;
	60
	61	/* Variables starting with an underscore are protected */
d18d46ec	62	if (!allow_protected && p[0] == '_')
0153028a LP	63	return false;
	64
	65	/* Don't allow digits as first character */
	66	if (p[0] >= '0' && p[0] <= '9')
	67	return false;
	68
	69	/* Only allow A-Z0-9 and '_' */
	70	for (a = p; a < p + l; a++)
d18d46ec ZJS	71	if ((a < 'A' \|\| a > 'Z') &&
	72	(a < '0' \|\| a > '9') &&
	73	*a != '_')
0153028a LP	74	return false;
	75
	76	return true;
	77	}
	78
3b3154df	79	static bool allow_object_pid(const struct ucred *ucred) {
968f3196 ZJS	80	return ucred && ucred->uid == 0;
	81	}
	82
0153028a LP	83	void server_process_native_message(
	84	Server *s,
	85	const void *buffer, size_t buffer_size,
3b3154df LP	86	const struct ucred *ucred,
3b3154df LP	87	const struct timeval *tv,
0153028a LP	88	const char *label, size_t label_len) {
	89
	90	struct iovec *iovec = NULL;
968f3196	91	unsigned n = 0, j, tn = (unsigned) -1;
0153028a	92	const char *p;
874bc134	93	size_t remaining, m = 0, entry_size = 0;
0153028a LP	94	int priority = LOG_INFO;
0153028a LP	95	char identifier = NULL, message = NULL;
968f3196	96	pid_t object_pid = 0;
0153028a LP	97
	98	assert(s);
	99	assert(buffer \|\| buffer_size == 0);
	100
	101	p = buffer;
	102	remaining = buffer_size;
	103
	104	while (remaining > 0) {
	105	const char e, q;
	106
	107	e = memchr(p, '\n', remaining);
	108
	109	if (!e) {
	110	/* Trailing noise, let's ignore it, and flush what we collected */
	111	log_debug("Received message with trailing noise, ignoring.");
	112	break;
	113	}
	114
	115	if (e == p) {
	116	/* Entry separator */
874bc134 ZJS	117
874bc134 ZJS	118	if (entry_size + n + 1 > ENTRY_SIZE_MAX) { /* data + separators + trailer */
8c0b803b	119	log_debug("Entry is too big with %u properties and %zu bytes, ignoring.", n, entry_size);
874bc134 ZJS	120	continue;
	121	}
	122
968f3196	123	server_dispatch_message(s, iovec, n, m, ucred, tv, label, label_len, NULL, priority, object_pid);
0153028a LP	124	n = 0;
0153028a LP	125	priority = LOG_INFO;
874bc134	126	entry_size = 0;
0153028a LP	127
	128	p++;
	129	remaining--;
	130	continue;
	131	}
	132
	133	if (p == '.' \|\| p == '#') {
	134	/* Ignore control commands for now, and
	135	* comments too. */
	136	remaining -= (e - p) + 1;
	137	p = e + 1;
	138	continue;
	139	}
	140
	141	/* A property follows */
	142
12a717f8	143	/* n existing properties, 1 new, +1 for _TRANSPORT */
92ee6447	144	if (!GREEDY_REALLOC(iovec, m, n + 2 + N_IOVEC_META_FIELDS + N_IOVEC_OBJECT_FIELDS)) {
968f3196 ZJS	145	log_oom();
968f3196 ZJS	146	break;
0153028a LP	147	}
	148
	149	q = memchr(p, '=', e - p);
	150	if (q) {
d18d46ec	151	if (valid_user_field(p, q - p, false)) {
0153028a LP	152	size_t l;
	153
	154	l = e - p;
	155
	156	/* If the field name starts with an
	157	* underscore, skip the variable,
	158	* since that indidates a trusted
	159	* field */
	160	iovec[n].iov_base = (char*) p;
	161	iovec[n].iov_len = l;
874bc134	162	entry_size += iovec[n].iov_len;
a174f94d	163	n++;
0153028a LP	164
	165	/* We need to determine the priority
	166	* of this entry for the rate limiting
	167	* logic */
	168	if (l == 10 &&
2a0e0692	169	startswith(p, "PRIORITY=") &&
0153028a LP	170	p[9] >= '0' && p[9] <= '9')
	171	priority = (priority & LOG_FACMASK) \| (p[9] - '0');
	172
	173	else if (l == 17 &&
2a0e0692	174	startswith(p, "SYSLOG_FACILITY=") &&
0153028a LP	175	p[16] >= '0' && p[16] <= '9')
	176	priority = (priority & LOG_PRIMASK) \| ((p[16] - '0') << 3);
	177
	178	else if (l == 18 &&
2a0e0692	179	startswith(p, "SYSLOG_FACILITY=") &&
0153028a LP	180	p[16] >= '0' && p[16] <= '9' &&
	181	p[17] >= '0' && p[17] <= '9')
	182	priority = (priority & LOG_PRIMASK) \| (((p[16] - '0')*10 + (p[17] - '0')) << 3);
	183
	184	else if (l >= 19 &&
2a0e0692	185	startswith(p, "SYSLOG_IDENTIFIER=")) {
0153028a LP	186	char *t;
	187
	188	t = strndup(p + 18, l - 18);
	189	if (t) {
	190	free(identifier);
	191	identifier = t;
	192	}
1f6b4113	193
0153028a	194	} else if (l >= 8 &&
2a0e0692	195	startswith(p, "MESSAGE=")) {
0153028a LP	196	char *t;
	197
	198	t = strndup(p + 8, l - 8);
	199	if (t) {
	200	free(message);
	201	message = t;
	202	}
1f6b4113	203
968f3196 ZJS	204	} else if (l > strlen("OBJECT_PID=") &&
968f3196 ZJS	205	l < strlen("OBJECT_PID=") + DECIMAL_STR_MAX(pid_t) &&
2a0e0692	206	startswith(p, "OBJECT_PID=") &&
968f3196 ZJS	207	allow_object_pid(ucred)) {
	208	char buf[DECIMAL_STR_MAX(pid_t)];
	209	memcpy(buf, p + strlen("OBJECT_PID="), l - strlen("OBJECT_PID="));
daf535a3	210	buf[l-strlen("OBJECT_PID=")] = '\0';
968f3196 ZJS	211
	212	/* ignore error */
	213	parse_pid(buf, &object_pid);
0153028a LP	214	}
	215	}
	216
	217	remaining -= (e - p) + 1;
	218	p = e + 1;
	219	continue;
	220	} else {
	221	le64_t l_le;
	222	uint64_t l;
	223	char *k;
	224
	225	if (remaining < e - p + 1 + sizeof(uint64_t) + 1) {
	226	log_debug("Failed to parse message, ignoring.");
	227	break;
	228	}
	229
	230	memcpy(&l_le, e + 1, sizeof(uint64_t));
	231	l = le64toh(l_le);
	232
505b6a61	233	if (l > DATA_SIZE_MAX) {
fa1c4b51	234	log_debug("Received binary data block of %"PRIu64" bytes is too large, ignoring.", l);
505b6a61 LP	235	break;
	236	}
	237
	238	if ((uint64_t) remaining < e - p + 1 + sizeof(uint64_t) + l + 1 \|\|
0153028a LP	239	e[1+sizeof(uint64_t)+l] != '\n') {
	240	log_debug("Failed to parse message, ignoring.");
	241	break;
	242	}
	243
	244	k = malloc((e - p) + 1 + l);
	245	if (!k) {
	246	log_oom();
	247	break;
	248	}
	249
	250	memcpy(k, p, e - p);
	251	k[e - p] = '=';
	252	memcpy(k + (e - p) + 1, e + 1 + sizeof(uint64_t), l);
	253
d18d46ec	254	if (valid_user_field(p, e - p, false)) {
0153028a LP	255	iovec[n].iov_base = k;
0153028a LP	256	iovec[n].iov_len = (e - p) + 1 + l;
874bc134	257	entry_size += iovec[n].iov_len;
a174f94d	258	n++;
0153028a LP	259	} else
	260	free(k);
	261
	262	remaining -= (e - p) + 1 + sizeof(uint64_t) + l + 1;
	263	p = e + 1 + sizeof(uint64_t) + l + 1;
	264	}
	265	}
	266
	267	if (n <= 0)
	268	goto finish;
	269
	270	tn = n++;
	271	IOVEC_SET_STRING(iovec[tn], "_TRANSPORT=journal");
874bc134 ZJS	272	entry_size += strlen("_TRANSPORT=journal");
	273
	274	if (entry_size + n + 1 > ENTRY_SIZE_MAX) { /* data + separators + trailer */
	275	log_debug("Entry is too big with %u properties and %zu bytes, ignoring.",
	276	n, entry_size);
	277	goto finish;
	278	}
0153028a LP	279
	280	if (message) {
	281	if (s->forward_to_syslog)
b6a20306	282	server_forward_syslog(s, syslog_fixup_facility(priority), identifier, message, ucred, tv);
0153028a LP	283
	284	if (s->forward_to_kmsg)
	285	server_forward_kmsg(s, priority, identifier, message, ucred);
	286
	287	if (s->forward_to_console)
	288	server_forward_console(s, priority, identifier, message, ucred);
40b71e89 ST	289
	290	if (s->forward_to_wall)
	291	server_forward_wall(s, priority, identifier, message, ucred);
0153028a LP	292	}
0153028a LP	293
968f3196	294	server_dispatch_message(s, iovec, n, m, ucred, tv, label, label_len, NULL, priority, object_pid);
0153028a LP	295
	296	finish:
	297	for (j = 0; j < n; j++) {
	298	if (j == tn)
	299	continue;
	300
	301	if (iovec[j].iov_base < buffer \|\|
	302	(const uint8_t) iovec[j].iov_base >= (const uint8_t) buffer + buffer_size)
	303	free(iovec[j].iov_base);
	304	}
	305
	306	free(iovec);
	307	free(identifier);
	308	free(message);
	309	}
	310
	311	void server_process_native_file(
	312	Server *s,
	313	int fd,
3b3154df LP	314	const struct ucred *ucred,
3b3154df LP	315	const struct timeval *tv,
0153028a LP	316	const char *label, size_t label_len) {
	317
	318	struct stat st;
c79e98ea	319	bool sealed;
1dfa7e79	320	int r;
0153028a	321
c79e98ea LP	322	/* Data is in the passed fd, since it didn't fit in a
	323	* datagram. */
	324
0153028a LP	325	assert(s);
	326	assert(fd >= 0);
	327
c79e98ea LP	328	/* If it's a memfd, check if it is sealed. If so, we can just
	329	* use map it and use it, and do not need to copy the data
	330	* out. */
73843b52	331	sealed = memfd_get_sealed(fd) > 0;
c79e98ea LP	332
c79e98ea LP	333	if (!sealed && (!ucred \|\| ucred->uid != 0)) {
1dfa7e79 LP	334	_cleanup_free_ char sl = NULL, k = NULL;
	335	const char *e;
	336
c79e98ea LP	337	/* If this is not a sealed memfd, and the peer is unknown or
	338	* unprivileged, then verify the path. */
	339
1dfa7e79 LP	340	if (asprintf(&sl, "/proc/self/fd/%i", fd) < 0) {
	341	log_oom();
	342	return;
	343	}
	344
	345	r = readlink_malloc(sl, &k);
	346	if (r < 0) {
709f6e46	347	log_error_errno(r, "readlink(%s) failed: %m", sl);
1dfa7e79 LP	348	return;
	349	}
	350
	351	e = path_startswith(k, "/dev/shm/");
	352	if (!e)
	353	e = path_startswith(k, "/tmp/");
	354	if (!e)
	355	e = path_startswith(k, "/var/tmp/");
	356	if (!e) {
	357	log_error("Received file outside of allowed directories. Refusing.");
	358	return;
	359	}
	360
ae6c3cc0	361	if (!filename_is_valid(e)) {
1dfa7e79 LP	362	log_error("Received file in subdirectory of allowed directories. Refusing.");
	363	return;
	364	}
	365	}
	366
0153028a	367	if (fstat(fd, &st) < 0) {
56f64d95	368	log_error_errno(errno, "Failed to stat passed file, ignoring: %m");
0153028a LP	369	return;
	370	}
	371
	372	if (!S_ISREG(st.st_mode)) {
	373	log_error("File passed is not regular. Ignoring.");
	374	return;
	375	}
	376
	377	if (st.st_size <= 0)
	378	return;
	379
	380	if (st.st_size > ENTRY_SIZE_MAX) {
	381	log_error("File passed too large. Ignoring.");
	382	return;
	383	}
	384
c79e98ea LP	385	if (sealed) {
	386	void *p;
	387	size_t ps;
	388
	389	/* The file is sealed, we can just map it and use it. */
0153028a	390
c79e98ea LP	391	ps = PAGE_ALIGN(st.st_size);
	392	p = mmap(NULL, ps, PROT_READ, MAP_PRIVATE, fd, 0);
	393	if (p == MAP_FAILED) {
56f64d95	394	log_error_errno(errno, "Failed to map memfd, ignoring: %m");
c79e98ea LP	395	return;
	396	}
	397
	398	server_process_native_message(s, p, st.st_size, ucred, tv, label, label_len);
	399	assert_se(munmap(p, ps) >= 0);
	400	} else {
	401	_cleanup_free_ void *p = NULL;
1e603a48	402	struct statvfs vfs;
c79e98ea LP	403	ssize_t n;
c79e98ea LP	404
1e603a48 LP	405	if (fstatvfs(fd, &vfs) < 0) {
	406	log_error_errno(errno, "Failed to stat file system of passed file, ignoring: %m");
	407	return;
	408	}
	409
	410	/* Refuse operating on file systems that have
	411	* mandatory locking enabled, see:
	412	*
	413	* https://github.com/systemd/systemd/issues/1822
	414	*/
	415	if (vfs.f_flag & ST_MANDLOCK) {
	416	log_error("Received file descriptor from file system with mandatory locking enable, refusing.");
	417	return;
	418	}
	419
	420	/* Make the fd non-blocking. On regular files this has
	421	* the effect of bypassing mandatory locking. Of
	422	* course, this should normally not be necessary given
	423	* the check above, but let's better be safe than
	424	* sorry, after all NFS is pretty confusing regarding
	425	* file system flags, and we better don't trust it,
	426	* and so is SMB. */
	427	r = fd_nonblock(fd, true);
	428	if (r < 0) {
	429	log_error_errno(r, "Failed to make fd non-blocking, ignoring: %m");
	430	return;
	431	}
	432
c79e98ea LP	433	/* The file is not sealed, we can't map the file here, since
	434	* clients might then truncate it and trigger a SIGBUS for
	435	* us. So let's stupidly read it */
	436
	437	p = malloc(st.st_size);
	438	if (!p) {
	439	log_oom();
	440	return;
	441	}
	442
	443	n = pread(fd, p, st.st_size, 0);
	444	if (n < 0)
c3753458	445	log_error_errno(errno, "Failed to read file, ignoring: %m");
c79e98ea LP	446	else if (n > 0)
	447	server_process_native_message(s, p, n, ucred, tv, label, label_len);
	448	}
0153028a LP	449	}
	450
	451	int server_open_native_socket(Server*s) {
fc2fffe7 LP	452
	453	static const union sockaddr_union sa = {
	454	.un.sun_family = AF_UNIX,
	455	.un.sun_path = "/run/systemd/journal/socket",
	456	};
3b3154df LP	457	static const int one = 1;
3b3154df LP	458	int r;
0153028a LP	459
	460	assert(s);
	461
	462	if (s->native_fd < 0) {
0153028a	463	s->native_fd = socket(AF_UNIX, SOCK_DGRAM\|SOCK_CLOEXEC\|SOCK_NONBLOCK, 0);
4a62c710 MS	464	if (s->native_fd < 0)
4a62c710 MS	465	return log_error_errno(errno, "socket() failed: %m");
0153028a	466
fc2fffe7	467	(void) unlink(sa.un.sun_path);
0153028a	468
fc2fffe7	469	r = bind(s->native_fd, &sa.sa, SOCKADDR_UN_LEN(sa.un));
4a62c710 MS	470	if (r < 0)
4a62c710 MS	471	return log_error_errno(errno, "bind(%s) failed: %m", sa.un.sun_path);
0153028a	472
4a61c3e5	473	(void) chmod(sa.un.sun_path, 0666);
0153028a LP	474	} else
	475	fd_nonblock(s->native_fd, 1);
	476
0153028a	477	r = setsockopt(s->native_fd, SOL_SOCKET, SO_PASSCRED, &one, sizeof(one));
4a62c710 MS	478	if (r < 0)
4a62c710 MS	479	return log_error_errno(errno, "SO_PASSCRED failed: %m");
0153028a LP	480
0153028a LP	481	#ifdef HAVE_SELINUX
6355e756	482	if (mac_selinux_have()) {
d682b3a7 LP	483	r = setsockopt(s->native_fd, SOL_SOCKET, SO_PASSSEC, &one, sizeof(one));
d682b3a7 LP	484	if (r < 0)
56f64d95	485	log_warning_errno(errno, "SO_PASSSEC failed: %m");
d682b3a7	486	}
0153028a LP	487	#endif
0153028a LP	488
0153028a	489	r = setsockopt(s->native_fd, SOL_SOCKET, SO_TIMESTAMP, &one, sizeof(one));
4a62c710 MS	490	if (r < 0)
4a62c710 MS	491	return log_error_errno(errno, "SO_TIMESTAMP failed: %m");
0153028a	492
8531ae70	493	r = sd_event_add_io(s->event, &s->native_event_source, s->native_fd, EPOLLIN, server_process_datagram, s);
23bbb0de MS	494	if (r < 0)
23bbb0de MS	495	return log_error_errno(r, "Failed to add native server fd to event loop: %m");
0153028a	496
48cef295 VC	497	r = sd_event_source_set_priority(s->native_event_source, SD_EVENT_PRIORITY_NORMAL+5);
	498	if (r < 0)
	499	return log_error_errno(r, "Failed to adjust native event source priority: %m");
	500
0153028a LP	501	return 0;
0153028a LP	502	}