diff options
author | Damien Miller <djm@cvs.openbsd.org> | 2019-03-01 02:32:40 +0000 |
---|---|---|
committer | Damien Miller <djm@cvs.openbsd.org> | 2019-03-01 02:32:40 +0000 |
commit | b14b67e3224c846702f9c044cdaa19a88c7df71e (patch) | |
tree | 4f2c4afe9b476d5ae6ba819d7792b6ee445873d8 /usr.bin/ssh/sshd.c | |
parent | 9db9d3d4c716ff8d949d88d16e1290c7746582db (diff) |
Fix two race conditions in sshd relating to SIGHUP:
1. Recently-forked child processes will briefly remain listening to
listen_socks. If the main server sshd process completes its restart
via execv() before these sockets are closed by the child processes
then it can fail to listen at the desired addresses/ports and/or
fail to restart.
2. When a SIGHUP is received, there may be forked child processes that
are awaiting their reexecution state. If the main server sshd
process restarts before passing this state, these child processes
will yield errors and use a fallback path of reading the current
sshd_config from the filesystem rather than use the one that sshd
was started with.
To fix both of these cases, we reuse the startup_pipes that are shared
between the main server sshd and forked children. Previously this was
used solely to implement tracking of pre-auth child processes for
MaxStartups, but this extends the messaging over these pipes to include
a child->parent message that the parent process is safe to restart. This
message is sent from the child after it has completed its preliminaries:
closing listen_socks and receiving its reexec state.
bz#2953, reported by Michal Koutný; ok markus@ dtucker@
Diffstat (limited to 'usr.bin/ssh/sshd.c')
-rw-r--r-- | usr.bin/ssh/sshd.c | 114 |
1 files changed, 86 insertions, 28 deletions
diff --git a/usr.bin/ssh/sshd.c b/usr.bin/ssh/sshd.c index 21c8889e15f..cfa66911fe3 100644 --- a/usr.bin/ssh/sshd.c +++ b/usr.bin/ssh/sshd.c @@ -1,4 +1,4 @@ -/* $OpenBSD: sshd.c,v 1.532 2019/01/21 10:38:54 djm Exp $ */ +/* $OpenBSD: sshd.c,v 1.533 2019/03/01 02:32:39 djm Exp $ */ /* * Author: Tatu Ylonen <ylo@cs.hut.fi> * Copyright (c) 1995 Tatu Ylonen <ylo@cs.hut.fi>, Espoo, Finland @@ -194,9 +194,26 @@ u_int session_id2_len = 0; /* record remote hostname or ip */ u_int utmp_len = HOST_NAME_MAX+1; -/* options.max_startup sized array of fd ints */ +/* + * startup_pipes/flags are used for tracking children of the listening sshd + * process early in their lifespans. This tracking is needed for three things: + * + * 1) Implementing the MaxStartups limit of concurrent unauthenticated + * connections. + * 2) Avoiding a race condition for SIGHUP processing, where child processes + * may have listen_socks open that could collide with main listener process + * after it restarts. + * 3) Ensuring that rexec'd sshd processes have received their initial state + * from the parent listen process before handling SIGHUP. + * + * Child processes signal that they have completed closure of the listen_socks + * and (if applicable) received their rexec state by sending a char over their + * sock. Child processes signal that authentication has completed by closing + * the sock (or by exiting). + */ static int *startup_pipes = NULL; -static int startup_pipe; /* in child */ +static int *startup_flags = NULL; /* Indicates child closed listener */ +static int startup_pipe = -1; /* in child */ /* variables used for privilege separation */ int use_privsep = -1; @@ -850,14 +867,9 @@ server_accept_inetd(int *sock_in, int *sock_out) { int fd; - startup_pipe = -1; if (rexeced_flag) { close(REEXEC_CONFIG_PASS_FD); *sock_in = *sock_out = dup(STDIN_FILENO); - if (!debug_flag) { - startup_pipe = dup(REEXEC_STARTUP_PIPE_FD); - close(REEXEC_STARTUP_PIPE_FD); - } } else { *sock_in = dup(STDIN_FILENO); *sock_out = dup(STDOUT_FILENO); @@ -978,8 +990,9 @@ server_accept_loop(int *sock_in, int *sock_out, int *newsock, int *config_s) { fd_set *fdset; int i, j, ret, maxfd; - int startups = 0; + int startups = 0, listening = 0, lameduck = 0; int startup_p[2] = { -1 , -1 }; + char c = 0; struct sockaddr_storage from; socklen_t fromlen; pid_t pid; @@ -992,6 +1005,7 @@ server_accept_loop(int *sock_in, int *sock_out, int *newsock, int *config_s) maxfd = listen_socks[i]; /* pipes connected to unauthenticated childs */ startup_pipes = xcalloc(options.max_startups, sizeof(int)); + startup_flags = xcalloc(options.max_startups, sizeof(int)); for (i = 0; i < options.max_startups; i++) startup_pipes[i] = -1; @@ -1000,8 +1014,15 @@ server_accept_loop(int *sock_in, int *sock_out, int *newsock, int *config_s) * the daemon is killed with a signal. */ for (;;) { - if (received_sighup) - sighup_restart(); + if (received_sighup) { + if (!lameduck) { + debug("Received SIGHUP; waiting for children"); + close_listen_socks(); + lameduck = 1; + } + if (listening <= 0) + sighup_restart(); + } free(fdset); fdset = xcalloc(howmany(maxfd + 1, NFDBITS), sizeof(fd_mask)); @@ -1027,19 +1048,37 @@ server_accept_loop(int *sock_in, int *sock_out, int *newsock, int *config_s) if (ret < 0) continue; - for (i = 0; i < options.max_startups; i++) - if (startup_pipes[i] != -1 && - FD_ISSET(startup_pipes[i], fdset)) { - /* - * the read end of the pipe is ready - * if the child has closed the pipe - * after successful authentication - * or if the child has died - */ + for (i = 0; i < options.max_startups; i++) { + if (startup_pipes[i] == -1 || + !FD_ISSET(startup_pipes[i], fdset)) + continue; + switch (read(startup_pipes[i], &c, sizeof(c))) { + case -1: + if (errno == EINTR || errno == EAGAIN) + continue; + if (errno != EPIPE) { + error("%s: startup pipe %d (fd=%d): " + "read %s", __func__, i, + startup_pipes[i], strerror(errno)); + } + /* FALLTHROUGH */ + case 0: + /* child exited or completed auth */ close(startup_pipes[i]); startup_pipes[i] = -1; startups--; + if (startup_flags[i]) + listening--; + break; + case 1: + /* child has finished preliminaries */ + if (startup_flags[i]) { + listening--; + startup_flags[i] = 0; + } + break; } + } for (i = 0; i < num_listen_socks; i++) { if (!FD_ISSET(listen_socks[i], fdset)) continue; @@ -1093,6 +1132,7 @@ server_accept_loop(int *sock_in, int *sock_out, int *newsock, int *config_s) if (maxfd < startup_p[0]) maxfd = startup_p[0]; startups++; + startup_flags[j] = 1; break; } @@ -1118,7 +1158,7 @@ server_accept_loop(int *sock_in, int *sock_out, int *newsock, int *config_s) send_rexec_state(config_s[0], cfg); close(config_s[0]); } - break; + return; } /* @@ -1126,13 +1166,14 @@ server_accept_loop(int *sock_in, int *sock_out, int *newsock, int *config_s) * the child process the connection. The * parent continues listening. */ + listening++; if ((pid = fork()) == 0) { /* * Child. Close the listening and * max_startup sockets. Start using * the accepted socket. Reinitialize * logging (since our pid has changed). - * We break out of the loop to handle + * We return from this function to handle * the connection. */ startup_pipe = startup_p[1]; @@ -1146,7 +1187,18 @@ server_accept_loop(int *sock_in, int *sock_out, int *newsock, int *config_s) log_stderr); if (rexec_flag) close(config_s[0]); - break; + else { + /* + * Signal parent that the preliminaries + * for this child are complete. For the + * re-exec case, this happens after the + * child has received the rexec state + * from the server. + */ + (void)atomicio(vwrite, startup_pipe, + "\0", 1); + } + return; } /* Parent. Stay in the loop. */ @@ -1164,10 +1216,6 @@ server_accept_loop(int *sock_in, int *sock_out, int *newsock, int *config_s) } close(*newsock); } - - /* child process check (or debug mode) */ - if (num_listen_socks < 0) - break; } } @@ -1456,8 +1504,18 @@ main(int ac, char **av) /* Fetch our configuration */ if ((cfg = sshbuf_new()) == NULL) fatal("%s: sshbuf_new failed", __func__); - if (rexeced_flag) + if (rexeced_flag) { recv_rexec_state(REEXEC_CONFIG_PASS_FD, cfg); + if (!debug_flag) { + startup_pipe = dup(REEXEC_STARTUP_PIPE_FD); + close(REEXEC_STARTUP_PIPE_FD); + /* + * Signal parent that this child is at a point where + * they can go away if they have a SIGHUP pending. + */ + (void)atomicio(vwrite, startup_pipe, "\0", 1); + } + } else if (strcasecmp(config_file_name, "none") != 0) load_server_config(config_file_name, cfg); |