coredump: do not allow user to access coredumps with changed uid/gid/capabilities

When the user starts a program which elevates its permissions via setuid,
setgid, or capabilities set on the file, it may access additional information
which would then be visible in the coredump. We shouldn't make the the coredump
visible to the user in such cases.

Reported-by: Matthias Gerstner <mgerstner@suse.de>

This reads the /proc/<pid>/auxv file and attaches it to the process metadata as
PROC_AUXV. Before the coredump is submitted, it is parsed and if either
at_secure was set (which the kernel will do for processes that are setuid,
setgid, or setcap), or if the effective uid/gid don't match uid/gid, the file
is not made accessible to the user. If we can't access this data, we assume the
file should not be made accessible either. In principle we could also access
the auxv data from a note in the core file, but that is much more complex and
it seems better to use the stand-alone file that is provided by the kernel.

Attaching auxv is both convient for this patch (because this way it's passed
between the stages along with other fields), but I think it makes sense to save
it in general.

We use the information early in the core file to figure out if the program was
32-bit or 64-bit and its endianness. This way we don't need heuristics to guess
whether the format of the auxv structure. This test might reject some cases on
fringe architecutes. But the impact would be limited: we just won't grant the
user permissions to view the coredump file. If people report that we're missing
some cases, we can always enhance this to support more architectures.

I tested auxv parsing on amd64, 32-bit program on amd64, arm64, arm32, and
ppc64el, but not the whole coredump handling.

(cherry picked from commit 3e4d0f6cf9)

src/basic/io-util.h

@@ -91,7 +91,16 @@ struct iovec_wrapper *iovw_new(void);
 struct iovec_wrapper *iovw_free(struct iovec_wrapper *iovw);
 struct iovec_wrapper *iovw_free_free(struct iovec_wrapper *iovw);
 void iovw_free_contents(struct iovec_wrapper *iovw, bool free_vectors);
+
 int iovw_put(struct iovec_wrapper *iovw, void *data, size_t len);
+static inline int iovw_consume(struct iovec_wrapper *iovw, void *data, size_t len) {
+        /* Move data into iovw or free on error */
+        int r = iovw_put(iovw, data, len);
+        if (r < 0)
+                free(data);
+        return r;
+}
+
 int iovw_put_string_field(struct iovec_wrapper *iovw, const char *field, const char *value);
 int iovw_put_string_field_free(struct iovec_wrapper *iovw, const char *field, char *value);
 void iovw_rebase(struct iovec_wrapper *iovw, char *old, char *new);

src/coredump/coredump.c

@@ -4,6 +4,7 @@
 #include <stdio.h>
 #include <sys/prctl.h>
 #include <sys/statvfs.h>
+#include <sys/auxv.h>
 #include <sys/xattr.h>
 #include <unistd.h>
 
@@ -106,6 +107,7 @@ enum {
 
         META_EXE = _META_MANDATORY_MAX,
         META_UNIT,
+        META_PROC_AUXV,
         _META_MAX
 };
 
@@ -120,10 +122,12 @@ static const char * const meta_field_names[_META_MAX] = {
         [META_COMM]           = "COREDUMP_COMM=",
         [META_EXE]            = "COREDUMP_EXE=",
         [META_UNIT]           = "COREDUMP_UNIT=",
+        [META_PROC_AUXV]      = "COREDUMP_PROC_AUXV=",
 };
 
 typedef struct Context {
         const char *meta[_META_MAX];
+        size_t meta_size[_META_MAX];
         pid_t pid;
         bool is_pid1;
         bool is_journald;
@@ -185,13 +189,16 @@ static uint64_t storage_size_max(void) {
         return 0;
 }
 
-static int fix_acl(int fd, uid_t uid) {
+static int fix_acl(int fd, uid_t uid, bool allow_user) {
+        assert(fd >= 0);
+        assert(uid_is_valid(uid));
 
 #if HAVE_ACL
         int r;
 
-        assert(fd >= 0);
-        assert(uid_is_valid(uid));
+        /* We don't allow users to read coredumps if the uid or capabilities were changed. */
+        if (!allow_user)
+                return 0;
 
         if (uid_is_system(uid) || uid_is_dynamic(uid) || uid == UID_NOBODY)
                 return 0;
@@ -251,7 +258,8 @@ static int fix_permissions(
                 const char *filename,
                 const char *target,
                 const Context *context,
-                uid_t uid) {
+                uid_t uid,
+                bool allow_user) {
 
         int r;
 
@@ -261,7 +269,7 @@ static int fix_permissions(
 
         /* Ignore errors on these */
         (void) fchmod(fd, 0640);
-        (void) fix_acl(fd, uid);
+        (void) fix_acl(fd, uid, allow_user);
         (void) fix_xattr(fd, context);
 
         r = fsync_full(fd);
@@ -331,6 +339,153 @@ static int make_filename(const Context *context, char **ret) {
         return 0;
 }
 
+static int parse_auxv64(
+                const uint64_t *auxv,
+                size_t size_bytes,
+                int *at_secure,
+                uid_t *uid,
+                uid_t *euid,
+                gid_t *gid,
+                gid_t *egid) {
+
+        assert(auxv || size_bytes == 0);
+
+        if (size_bytes % (2 * sizeof(uint64_t)) != 0)
+                return log_warning_errno(SYNTHETIC_ERRNO(EIO), "Incomplete auxv structure (%zu bytes).", size_bytes);
+
+        size_t words = size_bytes / sizeof(uint64_t);
+
+        /* Note that we set output variables even on error. */
+
+        for (size_t i = 0; i + 1 < words; i += 2)
+                switch (auxv[i]) {
+                case AT_SECURE:
+                        *at_secure = auxv[i + 1] != 0;
+                        break;
+                case AT_UID:
+                        *uid = auxv[i + 1];
+                        break;
+                case AT_EUID:
+                        *euid = auxv[i + 1];
+                        break;
+                case AT_GID:
+                        *gid = auxv[i + 1];
+                        break;
+                case AT_EGID:
+                        *egid = auxv[i + 1];
+                        break;
+                case AT_NULL:
+                        if (auxv[i + 1] != 0)
+                                goto error;
+                        return 0;
+                }
+ error:
+        return log_warning_errno(SYNTHETIC_ERRNO(ENODATA),
+                                 "AT_NULL terminator not found, cannot parse auxv structure.");
+}
+
+static int parse_auxv32(
+                const uint32_t *auxv,
+                size_t size_bytes,
+                int *at_secure,
+                uid_t *uid,
+                uid_t *euid,
+                gid_t *gid,
+                gid_t *egid) {
+
+        assert(auxv || size_bytes == 0);
+
+        size_t words = size_bytes / sizeof(uint32_t);
+
+        if (size_bytes % (2 * sizeof(uint32_t)) != 0)
+                return log_warning_errno(SYNTHETIC_ERRNO(EIO), "Incomplete auxv structure (%zu bytes).", size_bytes);
+
+        /* Note that we set output variables even on error. */
+
+        for (size_t i = 0; i + 1 < words; i += 2)
+                switch (auxv[i]) {
+                case AT_SECURE:
+                        *at_secure = auxv[i + 1] != 0;
+                        break;
+                case AT_UID:
+                        *uid = auxv[i + 1];
+                        break;
+                case AT_EUID:
+                        *euid = auxv[i + 1];
+                        break;
+                case AT_GID:
+                        *gid = auxv[i + 1];
+                        break;
+                case AT_EGID:
+                        *egid = auxv[i + 1];
+                        break;
+                case AT_NULL:
+                        if (auxv[i + 1] != 0)
+                                goto error;
+                        return 0;
+                }
+ error:
+        return log_warning_errno(SYNTHETIC_ERRNO(ENODATA),
+                                 "AT_NULL terminator not found, cannot parse auxv structure.");
+}
+
+static int grant_user_access(int core_fd, const Context *context) {
+        int at_secure = -1;
+        uid_t uid = UID_INVALID, euid = UID_INVALID;
+        uid_t gid = GID_INVALID, egid = GID_INVALID;
+        int r;
+
+        assert(core_fd >= 0);
+        assert(context);
+
+        if (!context->meta[META_PROC_AUXV])
+                return log_warning_errno(SYNTHETIC_ERRNO(ENODATA), "No auxv data, not adjusting permissions.");
+
+        uint8_t elf[EI_NIDENT];
+        errno = 0;
+        if (pread(core_fd, &elf, sizeof(elf), 0) != sizeof(elf))
+                return log_warning_errno(errno_or_else(EIO),
+                                         "Failed to pread from coredump fd: %s", STRERROR_OR_EOF(errno));
+
+        if (elf[EI_MAG0] != ELFMAG0 ||
+            elf[EI_MAG1] != ELFMAG1 ||
+            elf[EI_MAG2] != ELFMAG2 ||
+            elf[EI_MAG3] != ELFMAG3 ||
+            elf[EI_VERSION] != EV_CURRENT)
+                return log_info_errno(SYNTHETIC_ERRNO(EUCLEAN),
+                                      "Core file does not have ELF header, not adjusting permissions.");
+        if (!IN_SET(elf[EI_CLASS], ELFCLASS32, ELFCLASS64) ||
+            !IN_SET(elf[EI_DATA], ELFDATA2LSB, ELFDATA2MSB))
+                return log_info_errno(SYNTHETIC_ERRNO(EUCLEAN),
+                                      "Core file has strange ELF class, not adjusting permissions.");
+
+        if ((elf[EI_DATA] == ELFDATA2LSB) != (__BYTE_ORDER == __LITTLE_ENDIAN))
+                return log_info_errno(SYNTHETIC_ERRNO(EUCLEAN),
+                                      "Core file has non-native endianness, not adjusting permissions.");
+
+        if (elf[EI_CLASS] == ELFCLASS64)
+                r = parse_auxv64((const uint64_t*) context->meta[META_PROC_AUXV],
+                                 context->meta_size[META_PROC_AUXV],
+                                 &at_secure, &uid, &euid, &gid, &egid);
+        else
+                r = parse_auxv32((const uint32_t*) context->meta[META_PROC_AUXV],
+                                 context->meta_size[META_PROC_AUXV],
+                                 &at_secure, &uid, &euid, &gid, &egid);
+        if (r < 0)
+                return r;
+
+        /* We allow access if we got all the data and at_secure is not set and
+         * the uid/gid matches euid/egid. */
+        bool ret =
+                at_secure == 0 &&
+                uid != UID_INVALID && euid != UID_INVALID && uid == euid &&
+                gid != GID_INVALID && egid != GID_INVALID && gid == egid;
+        log_debug("Will %s access (uid="UID_FMT " euid="UID_FMT " gid="GID_FMT " egid="GID_FMT " at_secure=%s)",
+                  ret ? "permit" : "restrict",
+                  uid, euid, gid, egid, yes_no(at_secure));
+        return ret;
+}
+
 static int save_external_coredump(
                 const Context *context,
                 int input_fd,
@@ -453,6 +608,8 @@ static int save_external_coredump(
                                 context->meta[META_ARGV_PID], context->meta[META_COMM]);
         truncated = r == 1;
 
+        bool allow_user = grant_user_access(fd, context) > 0;
+
 #if HAVE_COMPRESSION
         if (arg_compress) {
                 _cleanup_(unlink_and_freep) char *tmp_compressed = NULL;
@@ -490,7 +647,7 @@ static int save_external_coredump(
                         uncompressed_size += partial_uncompressed_size;
                 }
 
-                r = fix_permissions(fd_compressed, tmp_compressed, fn_compressed, context, uid);
+                r = fix_permissions(fd_compressed, tmp_compressed, fn_compressed, context, uid, allow_user);
                 if (r < 0)
                         return r;
 
@@ -517,7 +674,7 @@ static int save_external_coredump(
                            "SIZE_LIMIT=%"PRIu64, max_size,
                            "MESSAGE_ID=" SD_MESSAGE_TRUNCATED_CORE_STR);
 
-        r = fix_permissions(fd, tmp, fn, context, uid);
+        r = fix_permissions(fd, tmp, fn, context, uid, allow_user);
         if (r < 0)
                 return log_error_errno(r, "Failed to fix permissions and finalize coredump %s into %s: %m", coredump_tmpfile_name(tmp), fn);
 
@@ -765,7 +922,7 @@ static int change_uid_gid(const Context *context) {
 }
 
 static int submit_coredump(
-                Context *context,
+                const Context *context,
                 struct iovec_wrapper *iovw,
                 int input_fd) {
 
@@ -944,16 +1101,15 @@ static int save_context(Context *context, const struct iovec_wrapper *iovw) {
                 struct iovec *iovec = iovw->iovec + n;
 
                 for (size_t i = 0; i < ELEMENTSOF(meta_field_names); i++) {
-                        char *p;
-
                         /* Note that these strings are NUL terminated, because we made sure that a
                          * trailing NUL byte is in the buffer, though not included in the iov_len
                          * count (see process_socket() and gather_pid_metadata_*()) */
                         assert(((char*) iovec->iov_base)[iovec->iov_len] == 0);
 
-                        p = startswith(iovec->iov_base, meta_field_names[i]);
+                        const char *p = startswith(iovec->iov_base, meta_field_names[i]);
                         if (p) {
                                 context->meta[i] = p;
+                                context->meta_size[i] = iovec->iov_len - strlen(meta_field_names[i]);
                                 break;
                         }
                 }
@@ -1190,6 +1346,7 @@ static int gather_pid_metadata(struct iovec_wrapper *iovw, Context *context) {
         uid_t owner_uid;
         pid_t pid;
         char *t;
+        size_t size;
         const char *p;
         int r;
 
@@ -1254,13 +1411,26 @@ static int gather_pid_metadata(struct iovec_wrapper *iovw, Context *context) {
                 (void) iovw_put_string_field_free(iovw, "COREDUMP_PROC_LIMITS=", t);
 
         p = procfs_file_alloca(pid, "cgroup");
-        if (read_full_virtual_file(p, &t, NULL) >=0)
+        if (read_full_virtual_file(p, &t, NULL) >= 0)
                 (void) iovw_put_string_field_free(iovw, "COREDUMP_PROC_CGROUP=", t);
 
         p = procfs_file_alloca(pid, "mountinfo");
-        if (read_full_virtual_file(p, &t, NULL) >=0)
+        if (read_full_virtual_file(p, &t, NULL) >= 0)
                 (void) iovw_put_string_field_free(iovw, "COREDUMP_PROC_MOUNTINFO=", t);
 
+        /* We attach /proc/auxv here. ELF coredumps also contain a note for this (NT_AUXV), see elf(5). */
+        p = procfs_file_alloca(pid, "auxv");
+        if (read_full_virtual_file(p, &t, &size) >= 0) {
+                char *buf = malloc(strlen("COREDUMP_PROC_AUXV=") + size + 1);
+                if (buf) {
+                        /* Add a dummy terminator to make save_context() happy. */
+                        *((uint8_t*) mempcpy(stpcpy(buf, "COREDUMP_PROC_AUXV="), t, size)) = '\0';
+                        (void) iovw_consume(iovw, buf, size + strlen("COREDUMP_PROC_AUXV="));
+                }
+
+                free(t);
+        }
+
         if (get_process_cwd(pid, &t) >= 0)
                 (void) iovw_put_string_field_free(iovw, "COREDUMP_CWD=", t);