openssh/sftp-client.c

2979 lines
78 KiB
C

/* $OpenBSD: sftp-client.c,v 1.171 2023/04/30 22:54:22 djm Exp $ */
/*
* Copyright (c) 2001-2004 Damien Miller <djm@openbsd.org>
*
* Permission to use, copy, modify, and distribute this software for any
* purpose with or without fee is hereby granted, provided that the above
* copyright notice and this permission notice appear in all copies.
*
* THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
* WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
* ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
* ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
* OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
*/
/* XXX: memleaks */
/* XXX: signed vs unsigned */
/* XXX: remove all logging, only return status codes */
/* XXX: copy between two remote sites */
#include "includes.h"
#include <sys/types.h>
#ifdef HAVE_SYS_STATVFS_H
#include <sys/statvfs.h>
#endif
#include "openbsd-compat/sys-queue.h"
#ifdef HAVE_SYS_STAT_H
# include <sys/stat.h>
#endif
#ifdef HAVE_SYS_TIME_H
# include <sys/time.h>
#endif
#include <sys/uio.h>
#include <dirent.h>
#include <errno.h>
#ifdef HAVE_POLL_H
#include <poll.h>
#else
# ifdef HAVE_SYS_POLL_H
# include <sys/poll.h>
# endif
#endif
#include <fcntl.h>
#include <signal.h>
#include <stdarg.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include "xmalloc.h"
#include "ssherr.h"
#include "sshbuf.h"
#include "log.h"
#include "atomicio.h"
#include "progressmeter.h"
#include "misc.h"
#include "utf8.h"
#include "sftp.h"
#include "sftp-common.h"
#include "sftp-client.h"
extern volatile sig_atomic_t interrupted;
extern int showprogress;
/* Default size of buffer for up/download (fix sftp.1 scp.1 if changed) */
#define DEFAULT_COPY_BUFLEN 32768
/* Default number of concurrent xfer requests (fix sftp.1 scp.1 if changed) */
#define DEFAULT_NUM_REQUESTS 64
/* Minimum amount of data to read at a time */
#define MIN_READ_SIZE 512
/* Maximum depth to descend in directory trees */
#define MAX_DIR_DEPTH 64
/* Directory separator characters */
#ifdef HAVE_CYGWIN
# define SFTP_DIRECTORY_CHARS "/\\"
#else /* HAVE_CYGWIN */
# define SFTP_DIRECTORY_CHARS "/"
#endif /* HAVE_CYGWIN */
struct sftp_conn {
int fd_in;
int fd_out;
u_int download_buflen;
u_int upload_buflen;
u_int num_requests;
u_int version;
u_int msg_id;
#define SFTP_EXT_POSIX_RENAME 0x00000001
#define SFTP_EXT_STATVFS 0x00000002
#define SFTP_EXT_FSTATVFS 0x00000004
#define SFTP_EXT_HARDLINK 0x00000008
#define SFTP_EXT_FSYNC 0x00000010
#define SFTP_EXT_LSETSTAT 0x00000020
#define SFTP_EXT_LIMITS 0x00000040
#define SFTP_EXT_PATH_EXPAND 0x00000080
#define SFTP_EXT_COPY_DATA 0x00000100
#define SFTP_EXT_GETUSERSGROUPS_BY_ID 0x00000200
u_int exts;
u_int64_t limit_kbps;
struct bwlimit bwlimit_in, bwlimit_out;
};
/* Tracks in-progress requests during file transfers */
struct request {
u_int id;
size_t len;
u_int64_t offset;
TAILQ_ENTRY(request) tq;
};
TAILQ_HEAD(requests, request);
static u_char *
get_handle(struct sftp_conn *conn, u_int expected_id, size_t *len,
const char *errfmt, ...) __attribute__((format(printf, 4, 5)));
static struct request *
request_enqueue(struct requests *requests, u_int id, size_t len,
uint64_t offset)
{
struct request *req;
req = xcalloc(1, sizeof(*req));
req->id = id;
req->len = len;
req->offset = offset;
TAILQ_INSERT_TAIL(requests, req, tq);
return req;
}
static struct request *
request_find(struct requests *requests, u_int id)
{
struct request *req;
for (req = TAILQ_FIRST(requests);
req != NULL && req->id != id;
req = TAILQ_NEXT(req, tq))
;
return req;
}
static int
sftpio(void *_bwlimit, size_t amount)
{
struct bwlimit *bwlimit = (struct bwlimit *)_bwlimit;
refresh_progress_meter(0);
if (bwlimit != NULL)
bandwidth_limit(bwlimit, amount);
return 0;
}
static void
send_msg(struct sftp_conn *conn, struct sshbuf *m)
{
u_char mlen[4];
struct iovec iov[2];
if (sshbuf_len(m) > SFTP_MAX_MSG_LENGTH)
fatal("Outbound message too long %zu", sshbuf_len(m));
/* Send length first */
put_u32(mlen, sshbuf_len(m));
iov[0].iov_base = mlen;
iov[0].iov_len = sizeof(mlen);
iov[1].iov_base = (u_char *)sshbuf_ptr(m);
iov[1].iov_len = sshbuf_len(m);
if (atomiciov6(writev, conn->fd_out, iov, 2, sftpio,
conn->limit_kbps > 0 ? &conn->bwlimit_out : NULL) !=
sshbuf_len(m) + sizeof(mlen))
fatal("Couldn't send packet: %s", strerror(errno));
sshbuf_reset(m);
}
static void
get_msg_extended(struct sftp_conn *conn, struct sshbuf *m, int initial)
{
u_int msg_len;
u_char *p;
int r;
sshbuf_reset(m);
if ((r = sshbuf_reserve(m, 4, &p)) != 0)
fatal_fr(r, "reserve");
if (atomicio6(read, conn->fd_in, p, 4, sftpio,
conn->limit_kbps > 0 ? &conn->bwlimit_in : NULL) != 4) {
if (errno == EPIPE || errno == ECONNRESET)
fatal("Connection closed");
else
fatal("Couldn't read packet: %s", strerror(errno));
}
if ((r = sshbuf_get_u32(m, &msg_len)) != 0)
fatal_fr(r, "sshbuf_get_u32");
if (msg_len > SFTP_MAX_MSG_LENGTH) {
do_log2(initial ? SYSLOG_LEVEL_ERROR : SYSLOG_LEVEL_FATAL,
"Received message too long %u", msg_len);
fatal("Ensure the remote shell produces no output "
"for non-interactive sessions.");
}
if ((r = sshbuf_reserve(m, msg_len, &p)) != 0)
fatal_fr(r, "reserve");
if (atomicio6(read, conn->fd_in, p, msg_len, sftpio,
conn->limit_kbps > 0 ? &conn->bwlimit_in : NULL)
!= msg_len) {
if (errno == EPIPE)
fatal("Connection closed");
else
fatal("Read packet: %s", strerror(errno));
}
}
static void
get_msg(struct sftp_conn *conn, struct sshbuf *m)
{
get_msg_extended(conn, m, 0);
}
static void
send_string_request(struct sftp_conn *conn, u_int id, u_int code, const char *s,
u_int len)
{
struct sshbuf *msg;
int r;
if ((msg = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
if ((r = sshbuf_put_u8(msg, code)) != 0 ||
(r = sshbuf_put_u32(msg, id)) != 0 ||
(r = sshbuf_put_string(msg, s, len)) != 0)
fatal_fr(r, "compose");
send_msg(conn, msg);
debug3("Sent message fd %d T:%u I:%u", conn->fd_out, code, id);
sshbuf_free(msg);
}
static void
send_string_attrs_request(struct sftp_conn *conn, u_int id, u_int code,
const void *s, u_int len, Attrib *a)
{
struct sshbuf *msg;
int r;
if ((msg = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
if ((r = sshbuf_put_u8(msg, code)) != 0 ||
(r = sshbuf_put_u32(msg, id)) != 0 ||
(r = sshbuf_put_string(msg, s, len)) != 0 ||
(r = encode_attrib(msg, a)) != 0)
fatal_fr(r, "compose");
send_msg(conn, msg);
debug3("Sent message fd %d T:%u I:%u F:0x%04x M:%05o",
conn->fd_out, code, id, a->flags, a->perm);
sshbuf_free(msg);
}
static u_int
get_status(struct sftp_conn *conn, u_int expected_id)
{
struct sshbuf *msg;
u_char type;
u_int id, status;
int r;
if ((msg = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
get_msg(conn, msg);
if ((r = sshbuf_get_u8(msg, &type)) != 0 ||
(r = sshbuf_get_u32(msg, &id)) != 0)
fatal_fr(r, "compose");
if (id != expected_id)
fatal("ID mismatch (%u != %u)", id, expected_id);
if (type != SSH2_FXP_STATUS)
fatal("Expected SSH2_FXP_STATUS(%u) packet, got %u",
SSH2_FXP_STATUS, type);
if ((r = sshbuf_get_u32(msg, &status)) != 0)
fatal_fr(r, "parse");
sshbuf_free(msg);
debug3("SSH2_FXP_STATUS %u", status);
return status;
}
static u_char *
get_handle(struct sftp_conn *conn, u_int expected_id, size_t *len,
const char *errfmt, ...)
{
struct sshbuf *msg;
u_int id, status;
u_char type;
u_char *handle;
char errmsg[256];
va_list args;
int r;
va_start(args, errfmt);
if (errfmt != NULL)
vsnprintf(errmsg, sizeof(errmsg), errfmt, args);
va_end(args);
if ((msg = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
get_msg(conn, msg);
if ((r = sshbuf_get_u8(msg, &type)) != 0 ||
(r = sshbuf_get_u32(msg, &id)) != 0)
fatal_fr(r, "parse");
if (id != expected_id)
fatal("%s: ID mismatch (%u != %u)",
errfmt == NULL ? __func__ : errmsg, id, expected_id);
if (type == SSH2_FXP_STATUS) {
if ((r = sshbuf_get_u32(msg, &status)) != 0)
fatal_fr(r, "parse status");
if (errfmt != NULL)
error("%s: %s", errmsg, fx2txt(status));
sshbuf_free(msg);
return(NULL);
} else if (type != SSH2_FXP_HANDLE)
fatal("%s: Expected SSH2_FXP_HANDLE(%u) packet, got %u",
errfmt == NULL ? __func__ : errmsg, SSH2_FXP_HANDLE, type);
if ((r = sshbuf_get_string(msg, &handle, len)) != 0)
fatal_fr(r, "parse handle");
sshbuf_free(msg);
return handle;
}
/* XXX returning &static is error-prone. Refactor to fill *Attrib argument */
static Attrib *
get_decode_stat(struct sftp_conn *conn, u_int expected_id, int quiet)
{
struct sshbuf *msg;
u_int id;
u_char type;
int r;
static Attrib a;
if ((msg = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
get_msg(conn, msg);
if ((r = sshbuf_get_u8(msg, &type)) != 0 ||
(r = sshbuf_get_u32(msg, &id)) != 0)
fatal_fr(r, "parse");
if (id != expected_id)
fatal("ID mismatch (%u != %u)", id, expected_id);
if (type == SSH2_FXP_STATUS) {
u_int status;
if ((r = sshbuf_get_u32(msg, &status)) != 0)
fatal_fr(r, "parse status");
if (quiet)
debug("stat remote: %s", fx2txt(status));
else
error("stat remote: %s", fx2txt(status));
sshbuf_free(msg);
return(NULL);
} else if (type != SSH2_FXP_ATTRS) {
fatal("Expected SSH2_FXP_ATTRS(%u) packet, got %u",
SSH2_FXP_ATTRS, type);
}
if ((r = decode_attrib(msg, &a)) != 0) {
error_fr(r, "decode_attrib");
sshbuf_free(msg);
return NULL;
}
debug3("Received stat reply T:%u I:%u F:0x%04x M:%05o",
type, id, a.flags, a.perm);
sshbuf_free(msg);
return &a;
}
static int
get_decode_statvfs(struct sftp_conn *conn, struct sftp_statvfs *st,
u_int expected_id, int quiet)
{
struct sshbuf *msg;
u_char type;
u_int id;
u_int64_t flag;
int r;
if ((msg = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
get_msg(conn, msg);
if ((r = sshbuf_get_u8(msg, &type)) != 0 ||
(r = sshbuf_get_u32(msg, &id)) != 0)
fatal_fr(r, "parse");
debug3("Received statvfs reply T:%u I:%u", type, id);
if (id != expected_id)
fatal("ID mismatch (%u != %u)", id, expected_id);
if (type == SSH2_FXP_STATUS) {
u_int status;
if ((r = sshbuf_get_u32(msg, &status)) != 0)
fatal_fr(r, "parse status");
if (quiet)
debug("remote statvfs: %s", fx2txt(status));
else
error("remote statvfs: %s", fx2txt(status));
sshbuf_free(msg);
return -1;
} else if (type != SSH2_FXP_EXTENDED_REPLY) {
fatal("Expected SSH2_FXP_EXTENDED_REPLY(%u) packet, got %u",
SSH2_FXP_EXTENDED_REPLY, type);
}
memset(st, 0, sizeof(*st));
if ((r = sshbuf_get_u64(msg, &st->f_bsize)) != 0 ||
(r = sshbuf_get_u64(msg, &st->f_frsize)) != 0 ||
(r = sshbuf_get_u64(msg, &st->f_blocks)) != 0 ||
(r = sshbuf_get_u64(msg, &st->f_bfree)) != 0 ||
(r = sshbuf_get_u64(msg, &st->f_bavail)) != 0 ||
(r = sshbuf_get_u64(msg, &st->f_files)) != 0 ||
(r = sshbuf_get_u64(msg, &st->f_ffree)) != 0 ||
(r = sshbuf_get_u64(msg, &st->f_favail)) != 0 ||
(r = sshbuf_get_u64(msg, &st->f_fsid)) != 0 ||
(r = sshbuf_get_u64(msg, &flag)) != 0 ||
(r = sshbuf_get_u64(msg, &st->f_namemax)) != 0)
fatal_fr(r, "parse statvfs");
st->f_flag = (flag & SSH2_FXE_STATVFS_ST_RDONLY) ? ST_RDONLY : 0;
st->f_flag |= (flag & SSH2_FXE_STATVFS_ST_NOSUID) ? ST_NOSUID : 0;
sshbuf_free(msg);
return 0;
}
struct sftp_conn *
do_init(int fd_in, int fd_out, u_int transfer_buflen, u_int num_requests,
u_int64_t limit_kbps)
{
u_char type;
struct sshbuf *msg;
struct sftp_conn *ret;
int r;
ret = xcalloc(1, sizeof(*ret));
ret->msg_id = 1;
ret->fd_in = fd_in;
ret->fd_out = fd_out;
ret->download_buflen = ret->upload_buflen =
transfer_buflen ? transfer_buflen : DEFAULT_COPY_BUFLEN;
ret->num_requests =
num_requests ? num_requests : DEFAULT_NUM_REQUESTS;
ret->exts = 0;
ret->limit_kbps = 0;
if ((msg = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
if ((r = sshbuf_put_u8(msg, SSH2_FXP_INIT)) != 0 ||
(r = sshbuf_put_u32(msg, SSH2_FILEXFER_VERSION)) != 0)
fatal_fr(r, "parse");
send_msg(ret, msg);
get_msg_extended(ret, msg, 1);
/* Expecting a VERSION reply */
if ((r = sshbuf_get_u8(msg, &type)) != 0)
fatal_fr(r, "parse type");
if (type != SSH2_FXP_VERSION) {
error("Invalid packet back from SSH2_FXP_INIT (type %u)",
type);
sshbuf_free(msg);
free(ret);
return(NULL);
}
if ((r = sshbuf_get_u32(msg, &ret->version)) != 0)
fatal_fr(r, "parse version");
debug2("Remote version: %u", ret->version);
/* Check for extensions */
while (sshbuf_len(msg) > 0) {
char *name;
u_char *value;
size_t vlen;
int known = 0;
if ((r = sshbuf_get_cstring(msg, &name, NULL)) != 0 ||
(r = sshbuf_get_string(msg, &value, &vlen)) != 0)
fatal_fr(r, "parse extension");
if (strcmp(name, "posix-rename@openssh.com") == 0 &&
strcmp((char *)value, "1") == 0) {
ret->exts |= SFTP_EXT_POSIX_RENAME;
known = 1;
} else if (strcmp(name, "statvfs@openssh.com") == 0 &&
strcmp((char *)value, "2") == 0) {
ret->exts |= SFTP_EXT_STATVFS;
known = 1;
} else if (strcmp(name, "fstatvfs@openssh.com") == 0 &&
strcmp((char *)value, "2") == 0) {
ret->exts |= SFTP_EXT_FSTATVFS;
known = 1;
} else if (strcmp(name, "hardlink@openssh.com") == 0 &&
strcmp((char *)value, "1") == 0) {
ret->exts |= SFTP_EXT_HARDLINK;
known = 1;
} else if (strcmp(name, "fsync@openssh.com") == 0 &&
strcmp((char *)value, "1") == 0) {
ret->exts |= SFTP_EXT_FSYNC;
known = 1;
} else if (strcmp(name, "lsetstat@openssh.com") == 0 &&
strcmp((char *)value, "1") == 0) {
ret->exts |= SFTP_EXT_LSETSTAT;
known = 1;
} else if (strcmp(name, "limits@openssh.com") == 0 &&
strcmp((char *)value, "1") == 0) {
ret->exts |= SFTP_EXT_LIMITS;
known = 1;
} else if (strcmp(name, "expand-path@openssh.com") == 0 &&
strcmp((char *)value, "1") == 0) {
ret->exts |= SFTP_EXT_PATH_EXPAND;
known = 1;
} else if (strcmp(name, "copy-data") == 0 &&
strcmp((char *)value, "1") == 0) {
ret->exts |= SFTP_EXT_COPY_DATA;
known = 1;
} else if (strcmp(name,
"users-groups-by-id@openssh.com") == 0 &&
strcmp((char *)value, "1") == 0) {
ret->exts |= SFTP_EXT_GETUSERSGROUPS_BY_ID;
known = 1;
}
if (known) {
debug2("Server supports extension \"%s\" revision %s",
name, value);
} else {
debug2("Unrecognised server extension \"%s\"", name);
}
free(name);
free(value);
}
sshbuf_free(msg);
/* Query the server for its limits */
if (ret->exts & SFTP_EXT_LIMITS) {
struct sftp_limits limits;
if (do_limits(ret, &limits) != 0)
fatal_f("limits failed");
/* If the caller did not specify, find a good value */
if (transfer_buflen == 0) {
ret->download_buflen = MINIMUM(limits.read_length,
SFTP_MAX_MSG_LENGTH - 1024);
ret->upload_buflen = MINIMUM(limits.write_length,
SFTP_MAX_MSG_LENGTH - 1024);
ret->download_buflen = MAXIMUM(ret->download_buflen, 64);
ret->upload_buflen = MAXIMUM(ret->upload_buflen, 64);
debug3("server upload/download buffer sizes "
"%llu / %llu; using %u / %u",
(unsigned long long)limits.write_length,
(unsigned long long)limits.read_length,
ret->upload_buflen, ret->download_buflen);
}
/* Use the server limit to scale down our value only */
if (num_requests == 0 && limits.open_handles) {
ret->num_requests =
MINIMUM(DEFAULT_NUM_REQUESTS, limits.open_handles);
if (ret->num_requests == 0)
ret->num_requests = 1;
debug3("server handle limit %llu; using %u",
(unsigned long long)limits.open_handles,
ret->num_requests);
}
}
/* Some filexfer v.0 servers don't support large packets */
if (ret->version == 0) {
ret->download_buflen = MINIMUM(ret->download_buflen, 20480);
ret->upload_buflen = MINIMUM(ret->upload_buflen, 20480);
}
ret->limit_kbps = limit_kbps;
if (ret->limit_kbps > 0) {
bandwidth_limit_init(&ret->bwlimit_in, ret->limit_kbps,
ret->download_buflen);
bandwidth_limit_init(&ret->bwlimit_out, ret->limit_kbps,
ret->upload_buflen);
}
return ret;
}
u_int
sftp_proto_version(struct sftp_conn *conn)
{
return conn->version;
}
int
do_limits(struct sftp_conn *conn, struct sftp_limits *limits)
{
u_int id, msg_id;
u_char type;
struct sshbuf *msg;
int r;
if ((conn->exts & SFTP_EXT_LIMITS) == 0) {
error("Server does not support limits@openssh.com extension");
return -1;
}
if ((msg = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
id = conn->msg_id++;
if ((r = sshbuf_put_u8(msg, SSH2_FXP_EXTENDED)) != 0 ||
(r = sshbuf_put_u32(msg, id)) != 0 ||
(r = sshbuf_put_cstring(msg, "limits@openssh.com")) != 0)
fatal_fr(r, "compose");
send_msg(conn, msg);
debug3("Sent message limits@openssh.com I:%u", id);
get_msg(conn, msg);
if ((r = sshbuf_get_u8(msg, &type)) != 0 ||
(r = sshbuf_get_u32(msg, &msg_id)) != 0)
fatal_fr(r, "parse");
debug3("Received limits reply T:%u I:%u", type, msg_id);
if (id != msg_id)
fatal("ID mismatch (%u != %u)", msg_id, id);
if (type != SSH2_FXP_EXTENDED_REPLY) {
debug_f("expected SSH2_FXP_EXTENDED_REPLY(%u) packet, got %u",
SSH2_FXP_EXTENDED_REPLY, type);
/* Disable the limits extension */
conn->exts &= ~SFTP_EXT_LIMITS;
sshbuf_free(msg);
return 0;
}
memset(limits, 0, sizeof(*limits));
if ((r = sshbuf_get_u64(msg, &limits->packet_length)) != 0 ||
(r = sshbuf_get_u64(msg, &limits->read_length)) != 0 ||
(r = sshbuf_get_u64(msg, &limits->write_length)) != 0 ||
(r = sshbuf_get_u64(msg, &limits->open_handles)) != 0)
fatal_fr(r, "parse limits");
sshbuf_free(msg);
return 0;
}
int
do_close(struct sftp_conn *conn, const u_char *handle, u_int handle_len)
{
u_int id, status;
struct sshbuf *msg;
int r;
if ((msg = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
id = conn->msg_id++;
if ((r = sshbuf_put_u8(msg, SSH2_FXP_CLOSE)) != 0 ||
(r = sshbuf_put_u32(msg, id)) != 0 ||
(r = sshbuf_put_string(msg, handle, handle_len)) != 0)
fatal_fr(r, "parse");
send_msg(conn, msg);
debug3("Sent message SSH2_FXP_CLOSE I:%u", id);
status = get_status(conn, id);
if (status != SSH2_FX_OK)
error("close remote: %s", fx2txt(status));
sshbuf_free(msg);
return status == SSH2_FX_OK ? 0 : -1;
}
static int
do_lsreaddir(struct sftp_conn *conn, const char *path, int print_flag,
SFTP_DIRENT ***dir)
{
struct sshbuf *msg;
u_int count, id, i, expected_id, ents = 0;
size_t handle_len;
u_char type, *handle;
int status = SSH2_FX_FAILURE;
int r;
if (dir)
*dir = NULL;
id = conn->msg_id++;
if ((msg = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
if ((r = sshbuf_put_u8(msg, SSH2_FXP_OPENDIR)) != 0 ||
(r = sshbuf_put_u32(msg, id)) != 0 ||
(r = sshbuf_put_cstring(msg, path)) != 0)
fatal_fr(r, "compose OPENDIR");
send_msg(conn, msg);
handle = get_handle(conn, id, &handle_len,
"remote readdir(\"%s\")", path);
if (handle == NULL) {
sshbuf_free(msg);
return -1;
}
if (dir) {
ents = 0;
*dir = xcalloc(1, sizeof(**dir));
(*dir)[0] = NULL;
}
for (; !interrupted;) {
id = expected_id = conn->msg_id++;
debug3("Sending SSH2_FXP_READDIR I:%u", id);
sshbuf_reset(msg);
if ((r = sshbuf_put_u8(msg, SSH2_FXP_READDIR)) != 0 ||
(r = sshbuf_put_u32(msg, id)) != 0 ||
(r = sshbuf_put_string(msg, handle, handle_len)) != 0)
fatal_fr(r, "compose READDIR");
send_msg(conn, msg);
sshbuf_reset(msg);
get_msg(conn, msg);
if ((r = sshbuf_get_u8(msg, &type)) != 0 ||
(r = sshbuf_get_u32(msg, &id)) != 0)
fatal_fr(r, "parse");
debug3("Received reply T:%u I:%u", type, id);
if (id != expected_id)
fatal("ID mismatch (%u != %u)", id, expected_id);
if (type == SSH2_FXP_STATUS) {
u_int rstatus;
if ((r = sshbuf_get_u32(msg, &rstatus)) != 0)
fatal_fr(r, "parse status");
debug3("Received SSH2_FXP_STATUS %d", rstatus);
if (rstatus == SSH2_FX_EOF)
break;
error("Couldn't read directory: %s", fx2txt(rstatus));
goto out;
} else if (type != SSH2_FXP_NAME)
fatal("Expected SSH2_FXP_NAME(%u) packet, got %u",
SSH2_FXP_NAME, type);
if ((r = sshbuf_get_u32(msg, &count)) != 0)
fatal_fr(r, "parse count");
if (count > SSHBUF_SIZE_MAX)
fatal_f("nonsensical number of entries");
if (count == 0)
break;
debug3("Received %d SSH2_FXP_NAME responses", count);
for (i = 0; i < count; i++) {
char *filename, *longname;
Attrib a;
if ((r = sshbuf_get_cstring(msg, &filename,
NULL)) != 0 ||
(r = sshbuf_get_cstring(msg, &longname,
NULL)) != 0)
fatal_fr(r, "parse filenames");
if ((r = decode_attrib(msg, &a)) != 0) {
error_fr(r, "couldn't decode attrib");
free(filename);
free(longname);
goto out;
}
if (print_flag)
mprintf("%s\n", longname);
/*
* Directory entries should never contain '/'
* These can be used to attack recursive ops
* (e.g. send '../../../../etc/passwd')
*/
if (strpbrk(filename, SFTP_DIRECTORY_CHARS) != NULL) {
error("Server sent suspect path \"%s\" "
"during readdir of \"%s\"", filename, path);
} else if (dir) {
*dir = xreallocarray(*dir, ents + 2, sizeof(**dir));
(*dir)[ents] = xcalloc(1, sizeof(***dir));
(*dir)[ents]->filename = xstrdup(filename);
(*dir)[ents]->longname = xstrdup(longname);
memcpy(&(*dir)[ents]->a, &a, sizeof(a));
(*dir)[++ents] = NULL;
}
free(filename);
free(longname);
}
}
status = 0;
out:
sshbuf_free(msg);
do_close(conn, handle, handle_len);
free(handle);
if (status != 0 && dir != NULL) {
/* Don't return results on error */
free_sftp_dirents(*dir);
*dir = NULL;
} else if (interrupted && dir != NULL && *dir != NULL) {
/* Don't return partial matches on interrupt */
free_sftp_dirents(*dir);
*dir = xcalloc(1, sizeof(**dir));
**dir = NULL;
}
return status == SSH2_FX_OK ? 0 : -1;
}
int
do_readdir(struct sftp_conn *conn, const char *path, SFTP_DIRENT ***dir)
{
return(do_lsreaddir(conn, path, 0, dir));
}
void free_sftp_dirents(SFTP_DIRENT **s)
{
int i;
if (s == NULL)
return;
for (i = 0; s[i]; i++) {
free(s[i]->filename);
free(s[i]->longname);
free(s[i]);
}
free(s);
}
int
do_rm(struct sftp_conn *conn, const char *path)
{
u_int status, id;
debug2("Sending SSH2_FXP_REMOVE \"%s\"", path);
id = conn->msg_id++;
send_string_request(conn, id, SSH2_FXP_REMOVE, path, strlen(path));
status = get_status(conn, id);
if (status != SSH2_FX_OK)
error("remote delete %s: %s", path, fx2txt(status));
return status == SSH2_FX_OK ? 0 : -1;
}
int
do_mkdir(struct sftp_conn *conn, const char *path, Attrib *a, int print_flag)
{
u_int status, id;
debug2("Sending SSH2_FXP_MKDIR \"%s\"", path);
id = conn->msg_id++;
send_string_attrs_request(conn, id, SSH2_FXP_MKDIR, path,
strlen(path), a);
status = get_status(conn, id);
if (status != SSH2_FX_OK && print_flag)
error("remote mkdir \"%s\": %s", path, fx2txt(status));
return status == SSH2_FX_OK ? 0 : -1;
}
int
do_rmdir(struct sftp_conn *conn, const char *path)
{
u_int status, id;
debug2("Sending SSH2_FXP_RMDIR \"%s\"", path);
id = conn->msg_id++;
send_string_request(conn, id, SSH2_FXP_RMDIR, path,
strlen(path));
status = get_status(conn, id);
if (status != SSH2_FX_OK)
error("remote rmdir \"%s\": %s", path, fx2txt(status));
return status == SSH2_FX_OK ? 0 : -1;
}
Attrib *
do_stat(struct sftp_conn *conn, const char *path, int quiet)
{
u_int id;
debug2("Sending SSH2_FXP_STAT \"%s\"", path);
id = conn->msg_id++;
send_string_request(conn, id,
conn->version == 0 ? SSH2_FXP_STAT_VERSION_0 : SSH2_FXP_STAT,
path, strlen(path));
return(get_decode_stat(conn, id, quiet));
}
Attrib *
do_lstat(struct sftp_conn *conn, const char *path, int quiet)
{
u_int id;
if (conn->version == 0) {
if (quiet)
debug("Server version does not support lstat operation");
else
logit("Server version does not support lstat operation");
return(do_stat(conn, path, quiet));
}
id = conn->msg_id++;
send_string_request(conn, id, SSH2_FXP_LSTAT, path,
strlen(path));
return(get_decode_stat(conn, id, quiet));
}
#ifdef notyet
Attrib *
do_fstat(struct sftp_conn *conn, const u_char *handle, u_int handle_len,
int quiet)
{
u_int id;
debug2("Sending SSH2_FXP_FSTAT \"%s\"");
id = conn->msg_id++;
send_string_request(conn, id, SSH2_FXP_FSTAT, handle,
handle_len);
return(get_decode_stat(conn, id, quiet));
}
#endif
int
do_setstat(struct sftp_conn *conn, const char *path, Attrib *a)
{
u_int status, id;
debug2("Sending SSH2_FXP_SETSTAT \"%s\"", path);
id = conn->msg_id++;
send_string_attrs_request(conn, id, SSH2_FXP_SETSTAT, path,
strlen(path), a);
status = get_status(conn, id);
if (status != SSH2_FX_OK)
error("remote setstat \"%s\": %s", path, fx2txt(status));
return status == SSH2_FX_OK ? 0 : -1;
}
int
do_fsetstat(struct sftp_conn *conn, const u_char *handle, u_int handle_len,
Attrib *a)
{
u_int status, id;
debug2("Sending SSH2_FXP_FSETSTAT");
id = conn->msg_id++;
send_string_attrs_request(conn, id, SSH2_FXP_FSETSTAT, handle,
handle_len, a);
status = get_status(conn, id);
if (status != SSH2_FX_OK)
error("remote fsetstat: %s", fx2txt(status));
return status == SSH2_FX_OK ? 0 : -1;
}
/* Implements both the realpath and expand-path operations */
static char *
do_realpath_expand(struct sftp_conn *conn, const char *path, int expand)
{
struct sshbuf *msg;
u_int expected_id, count, id;
char *filename, *longname;
Attrib a;
u_char type;
int r;
const char *what = "SSH2_FXP_REALPATH";
if (expand)
what = "expand-path@openssh.com";
if ((msg = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
expected_id = id = conn->msg_id++;
if (expand) {
debug2("Sending SSH2_FXP_EXTENDED(expand-path@openssh.com) "
"\"%s\"", path);
if ((r = sshbuf_put_u8(msg, SSH2_FXP_EXTENDED)) != 0 ||
(r = sshbuf_put_u32(msg, id)) != 0 ||
(r = sshbuf_put_cstring(msg,
"expand-path@openssh.com")) != 0 ||
(r = sshbuf_put_cstring(msg, path)) != 0)
fatal_fr(r, "compose %s", what);
send_msg(conn, msg);
} else {
debug2("Sending SSH2_FXP_REALPATH \"%s\"", path);
send_string_request(conn, id, SSH2_FXP_REALPATH,
path, strlen(path));
}
get_msg(conn, msg);
if ((r = sshbuf_get_u8(msg, &type)) != 0 ||
(r = sshbuf_get_u32(msg, &id)) != 0)
fatal_fr(r, "parse");
if (id != expected_id)
fatal("ID mismatch (%u != %u)", id, expected_id);
if (type == SSH2_FXP_STATUS) {
u_int status;
char *errmsg;
if ((r = sshbuf_get_u32(msg, &status)) != 0 ||
(r = sshbuf_get_cstring(msg, &errmsg, NULL)) != 0)
fatal_fr(r, "parse status");
error("%s %s: %s", expand ? "expand" : "realpath",
path, *errmsg == '\0' ? fx2txt(status) : errmsg);
free(errmsg);
sshbuf_free(msg);
return NULL;
} else if (type != SSH2_FXP_NAME)
fatal("Expected SSH2_FXP_NAME(%u) packet, got %u",
SSH2_FXP_NAME, type);
if ((r = sshbuf_get_u32(msg, &count)) != 0)
fatal_fr(r, "parse count");
if (count != 1)
fatal("Got multiple names (%d) from %s", count, what);
if ((r = sshbuf_get_cstring(msg, &filename, NULL)) != 0 ||
(r = sshbuf_get_cstring(msg, &longname, NULL)) != 0 ||
(r = decode_attrib(msg, &a)) != 0)
fatal_fr(r, "parse filename/attrib");
debug3("%s %s -> %s", what, path, filename);
free(longname);
sshbuf_free(msg);
return(filename);
}
char *
do_realpath(struct sftp_conn *conn, const char *path)
{
return do_realpath_expand(conn, path, 0);
}
int
can_expand_path(struct sftp_conn *conn)
{
return (conn->exts & SFTP_EXT_PATH_EXPAND) != 0;
}
char *
do_expand_path(struct sftp_conn *conn, const char *path)
{
if (!can_expand_path(conn)) {
debug3_f("no server support, fallback to realpath");
return do_realpath_expand(conn, path, 0);
}
return do_realpath_expand(conn, path, 1);
}
int
do_copy(struct sftp_conn *conn, const char *oldpath, const char *newpath)
{
Attrib junk, *a;
struct sshbuf *msg;
u_char *old_handle, *new_handle;
u_int mode, status, id;
size_t old_handle_len, new_handle_len;
int r;
/* Return if the extension is not supported */
if ((conn->exts & SFTP_EXT_COPY_DATA) == 0) {
error("Server does not support copy-data extension");
return -1;
}
/* Make sure the file exists, and we can copy its perms */
if ((a = do_stat(conn, oldpath, 0)) == NULL)
return -1;
/* Do not preserve set[ug]id here, as we do not preserve ownership */
if (a->flags & SSH2_FILEXFER_ATTR_PERMISSIONS) {
mode = a->perm & 0777;
if (!S_ISREG(a->perm)) {
error("Cannot copy non-regular file: %s", oldpath);
return -1;
}
} else {
/* NB: The user's umask will apply to this */
mode = 0666;
}
/* Set up the new perms for the new file */
attrib_clear(a);
a->perm = mode;
a->flags |= SSH2_FILEXFER_ATTR_PERMISSIONS;
if ((msg = sshbuf_new()) == NULL)
fatal("%s: sshbuf_new failed", __func__);
attrib_clear(&junk); /* Send empty attributes */
/* Open the old file for reading */
id = conn->msg_id++;
if ((r = sshbuf_put_u8(msg, SSH2_FXP_OPEN)) != 0 ||
(r = sshbuf_put_u32(msg, id)) != 0 ||
(r = sshbuf_put_cstring(msg, oldpath)) != 0 ||
(r = sshbuf_put_u32(msg, SSH2_FXF_READ)) != 0 ||
(r = encode_attrib(msg, &junk)) != 0)
fatal("%s: buffer error: %s", __func__, ssh_err(r));
send_msg(conn, msg);
debug3("Sent message SSH2_FXP_OPEN I:%u P:%s", id, oldpath);
sshbuf_reset(msg);
old_handle = get_handle(conn, id, &old_handle_len,
"remote open(\"%s\")", oldpath);
if (old_handle == NULL) {
sshbuf_free(msg);
return -1;
}
/* Open the new file for writing */
id = conn->msg_id++;
if ((r = sshbuf_put_u8(msg, SSH2_FXP_OPEN)) != 0 ||
(r = sshbuf_put_u32(msg, id)) != 0 ||
(r = sshbuf_put_cstring(msg, newpath)) != 0 ||
(r = sshbuf_put_u32(msg, SSH2_FXF_WRITE|SSH2_FXF_CREAT|
SSH2_FXF_TRUNC)) != 0 ||
(r = encode_attrib(msg, a)) != 0)
fatal("%s: buffer error: %s", __func__, ssh_err(r));
send_msg(conn, msg);
debug3("Sent message SSH2_FXP_OPEN I:%u P:%s", id, newpath);
sshbuf_reset(msg);
new_handle = get_handle(conn, id, &new_handle_len,
"remote open(\"%s\")", newpath);
if (new_handle == NULL) {
sshbuf_free(msg);
free(old_handle);
return -1;
}
/* Copy the file data */
id = conn->msg_id++;
if ((r = sshbuf_put_u8(msg, SSH2_FXP_EXTENDED)) != 0 ||
(r = sshbuf_put_u32(msg, id)) != 0 ||
(r = sshbuf_put_cstring(msg, "copy-data")) != 0 ||
(r = sshbuf_put_string(msg, old_handle, old_handle_len)) != 0 ||
(r = sshbuf_put_u64(msg, 0)) != 0 ||
(r = sshbuf_put_u64(msg, 0)) != 0 ||
(r = sshbuf_put_string(msg, new_handle, new_handle_len)) != 0 ||
(r = sshbuf_put_u64(msg, 0)) != 0)
fatal("%s: buffer error: %s", __func__, ssh_err(r));
send_msg(conn, msg);
debug3("Sent message copy-data \"%s\" 0 0 -> \"%s\" 0",
oldpath, newpath);
status = get_status(conn, id);
if (status != SSH2_FX_OK)
error("Couldn't copy file \"%s\" to \"%s\": %s", oldpath,
newpath, fx2txt(status));
/* Clean up everything */
sshbuf_free(msg);
do_close(conn, old_handle, old_handle_len);
do_close(conn, new_handle, new_handle_len);
free(old_handle);
free(new_handle);
return status == SSH2_FX_OK ? 0 : -1;
}
int
do_rename(struct sftp_conn *conn, const char *oldpath, const char *newpath,
int force_legacy)
{
struct sshbuf *msg;
u_int status, id;
int r, use_ext = (conn->exts & SFTP_EXT_POSIX_RENAME) && !force_legacy;
if ((msg = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
/* Send rename request */
id = conn->msg_id++;
if (use_ext) {
debug2("Sending SSH2_FXP_EXTENDED(posix-rename@openssh.com) "
"\"%s\" to \"%s\"", oldpath, newpath);
if ((r = sshbuf_put_u8(msg, SSH2_FXP_EXTENDED)) != 0 ||
(r = sshbuf_put_u32(msg, id)) != 0 ||
(r = sshbuf_put_cstring(msg,
"posix-rename@openssh.com")) != 0)
fatal_fr(r, "compose posix-rename");
} else {
debug2("Sending SSH2_FXP_RENAME \"%s\" to \"%s\"",
oldpath, newpath);
if ((r = sshbuf_put_u8(msg, SSH2_FXP_RENAME)) != 0 ||
(r = sshbuf_put_u32(msg, id)) != 0)
fatal_fr(r, "compose rename");
}
if ((r = sshbuf_put_cstring(msg, oldpath)) != 0 ||
(r = sshbuf_put_cstring(msg, newpath)) != 0)
fatal_fr(r, "compose paths");
send_msg(conn, msg);
debug3("Sent message %s \"%s\" -> \"%s\"",
use_ext ? "posix-rename@openssh.com" :
"SSH2_FXP_RENAME", oldpath, newpath);
sshbuf_free(msg);
status = get_status(conn, id);
if (status != SSH2_FX_OK)
error("remote rename \"%s\" to \"%s\": %s", oldpath,
newpath, fx2txt(status));
return status == SSH2_FX_OK ? 0 : -1;
}
int
do_hardlink(struct sftp_conn *conn, const char *oldpath, const char *newpath)
{
struct sshbuf *msg;
u_int status, id;
int r;
if ((conn->exts & SFTP_EXT_HARDLINK) == 0) {
error("Server does not support hardlink@openssh.com extension");
return -1;
}
debug2("Sending SSH2_FXP_EXTENDED(hardlink@openssh.com) "
"\"%s\" to \"%s\"", oldpath, newpath);
if ((msg = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
/* Send link request */
id = conn->msg_id++;
if ((r = sshbuf_put_u8(msg, SSH2_FXP_EXTENDED)) != 0 ||
(r = sshbuf_put_u32(msg, id)) != 0 ||
(r = sshbuf_put_cstring(msg, "hardlink@openssh.com")) != 0 ||
(r = sshbuf_put_cstring(msg, oldpath)) != 0 ||
(r = sshbuf_put_cstring(msg, newpath)) != 0)
fatal_fr(r, "compose");
send_msg(conn, msg);
debug3("Sent message hardlink@openssh.com \"%s\" -> \"%s\"",
oldpath, newpath);
sshbuf_free(msg);
status = get_status(conn, id);
if (status != SSH2_FX_OK)
error("remote link \"%s\" to \"%s\": %s", oldpath,
newpath, fx2txt(status));
return status == SSH2_FX_OK ? 0 : -1;
}
int
do_symlink(struct sftp_conn *conn, const char *oldpath, const char *newpath)
{
struct sshbuf *msg;
u_int status, id;
int r;
if (conn->version < 3) {
error("This server does not support the symlink operation");
return(SSH2_FX_OP_UNSUPPORTED);
}
debug2("Sending SSH2_FXP_SYMLINK \"%s\" to \"%s\"", oldpath, newpath);
if ((msg = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
/* Send symlink request */
id = conn->msg_id++;
if ((r = sshbuf_put_u8(msg, SSH2_FXP_SYMLINK)) != 0 ||
(r = sshbuf_put_u32(msg, id)) != 0 ||
(r = sshbuf_put_cstring(msg, oldpath)) != 0 ||
(r = sshbuf_put_cstring(msg, newpath)) != 0)
fatal_fr(r, "compose");
send_msg(conn, msg);
debug3("Sent message SSH2_FXP_SYMLINK \"%s\" -> \"%s\"", oldpath,
newpath);
sshbuf_free(msg);
status = get_status(conn, id);
if (status != SSH2_FX_OK)
error("remote symlink file \"%s\" to \"%s\": %s", oldpath,
newpath, fx2txt(status));
return status == SSH2_FX_OK ? 0 : -1;
}
int
do_fsync(struct sftp_conn *conn, u_char *handle, u_int handle_len)
{
struct sshbuf *msg;
u_int status, id;
int r;
/* Silently return if the extension is not supported */
if ((conn->exts & SFTP_EXT_FSYNC) == 0)
return -1;
debug2("Sending SSH2_FXP_EXTENDED(fsync@openssh.com)");
/* Send fsync request */
if ((msg = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
id = conn->msg_id++;
if ((r = sshbuf_put_u8(msg, SSH2_FXP_EXTENDED)) != 0 ||
(r = sshbuf_put_u32(msg, id)) != 0 ||
(r = sshbuf_put_cstring(msg, "fsync@openssh.com")) != 0 ||
(r = sshbuf_put_string(msg, handle, handle_len)) != 0)
fatal_fr(r, "compose");
send_msg(conn, msg);
debug3("Sent message fsync@openssh.com I:%u", id);
sshbuf_free(msg);
status = get_status(conn, id);
if (status != SSH2_FX_OK)
error("remote fsync: %s", fx2txt(status));
return status == SSH2_FX_OK ? 0 : -1;
}
#ifdef notyet
char *
do_readlink(struct sftp_conn *conn, const char *path)
{
struct sshbuf *msg;
u_int expected_id, count, id;
char *filename, *longname;
Attrib a;
u_char type;
int r;
debug2("Sending SSH2_FXP_READLINK \"%s\"", path);
expected_id = id = conn->msg_id++;
send_string_request(conn, id, SSH2_FXP_READLINK, path, strlen(path));
if ((msg = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
get_msg(conn, msg);
if ((r = sshbuf_get_u8(msg, &type)) != 0 ||
(r = sshbuf_get_u32(msg, &id)) != 0)
fatal_fr(r, "parse");
if (id != expected_id)
fatal("ID mismatch (%u != %u)", id, expected_id);
if (type == SSH2_FXP_STATUS) {
u_int status;
if ((r = sshbuf_get_u32(msg, &status)) != 0)
fatal_fr(r, "parse status");
error("Couldn't readlink: %s", fx2txt(status));
sshbuf_free(msg);
return(NULL);
} else if (type != SSH2_FXP_NAME)
fatal("Expected SSH2_FXP_NAME(%u) packet, got %u",
SSH2_FXP_NAME, type);
if ((r = sshbuf_get_u32(msg, &count)) != 0)
fatal_fr(r, "parse count");
if (count != 1)
fatal("Got multiple names (%d) from SSH_FXP_READLINK", count);
if ((r = sshbuf_get_cstring(msg, &filename, NULL)) != 0 ||
(r = sshbuf_get_cstring(msg, &longname, NULL)) != 0 ||
(r = decode_attrib(msg, &a)) != 0)
fatal_fr(r, "parse filenames/attrib");
debug3("SSH_FXP_READLINK %s -> %s", path, filename);
free(longname);
sshbuf_free(msg);
return filename;
}
#endif
int
do_statvfs(struct sftp_conn *conn, const char *path, struct sftp_statvfs *st,
int quiet)
{
struct sshbuf *msg;
u_int id;
int r;
if ((conn->exts & SFTP_EXT_STATVFS) == 0) {
error("Server does not support statvfs@openssh.com extension");
return -1;
}
debug2("Sending SSH2_FXP_EXTENDED(statvfs@openssh.com) \"%s\"", path);
id = conn->msg_id++;
if ((msg = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
if ((r = sshbuf_put_u8(msg, SSH2_FXP_EXTENDED)) != 0 ||
(r = sshbuf_put_u32(msg, id)) != 0 ||
(r = sshbuf_put_cstring(msg, "statvfs@openssh.com")) != 0 ||
(r = sshbuf_put_cstring(msg, path)) != 0)
fatal_fr(r, "compose");
send_msg(conn, msg);
sshbuf_free(msg);
return get_decode_statvfs(conn, st, id, quiet);
}
#ifdef notyet
int
do_fstatvfs(struct sftp_conn *conn, const u_char *handle, u_int handle_len,
struct sftp_statvfs *st, int quiet)
{
struct sshbuf *msg;
u_int id;
if ((conn->exts & SFTP_EXT_FSTATVFS) == 0) {
error("Server does not support fstatvfs@openssh.com extension");
return -1;
}
debug2("Sending SSH2_FXP_EXTENDED(fstatvfs@openssh.com)");
id = conn->msg_id++;
if ((msg = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
if ((r = sshbuf_put_u8(msg, SSH2_FXP_EXTENDED)) != 0 ||
(r = sshbuf_put_u32(msg, id)) != 0 ||
(r = sshbuf_put_cstring(msg, "fstatvfs@openssh.com")) != 0 ||
(r = sshbuf_put_string(msg, handle, handle_len)) != 0)
fatal_fr(r, "compose");
send_msg(conn, msg);
sshbuf_free(msg);
return get_decode_statvfs(conn, st, id, quiet);
}
#endif
int
do_lsetstat(struct sftp_conn *conn, const char *path, Attrib *a)
{
struct sshbuf *msg;
u_int status, id;
int r;
if ((conn->exts & SFTP_EXT_LSETSTAT) == 0) {
error("Server does not support lsetstat@openssh.com extension");
return -1;
}
debug2("Sending SSH2_FXP_EXTENDED(lsetstat@openssh.com) \"%s\"", path);
id = conn->msg_id++;
if ((msg = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
if ((r = sshbuf_put_u8(msg, SSH2_FXP_EXTENDED)) != 0 ||
(r = sshbuf_put_u32(msg, id)) != 0 ||
(r = sshbuf_put_cstring(msg, "lsetstat@openssh.com")) != 0 ||
(r = sshbuf_put_cstring(msg, path)) != 0 ||
(r = encode_attrib(msg, a)) != 0)
fatal_fr(r, "compose");
send_msg(conn, msg);
sshbuf_free(msg);
status = get_status(conn, id);
if (status != SSH2_FX_OK)
error("remote lsetstat \"%s\": %s", path, fx2txt(status));
return status == SSH2_FX_OK ? 0 : -1;
}
static void
send_read_request(struct sftp_conn *conn, u_int id, u_int64_t offset,
u_int len, const u_char *handle, u_int handle_len)
{
struct sshbuf *msg;
int r;
if ((msg = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
if ((r = sshbuf_put_u8(msg, SSH2_FXP_READ)) != 0 ||
(r = sshbuf_put_u32(msg, id)) != 0 ||
(r = sshbuf_put_string(msg, handle, handle_len)) != 0 ||
(r = sshbuf_put_u64(msg, offset)) != 0 ||
(r = sshbuf_put_u32(msg, len)) != 0)
fatal_fr(r, "compose");
send_msg(conn, msg);
sshbuf_free(msg);
}
static int
send_open(struct sftp_conn *conn, const char *path, const char *tag,
u_int openmode, Attrib *a, u_char **handlep, size_t *handle_lenp)
{
Attrib junk;
u_char *handle;
size_t handle_len;
struct sshbuf *msg;
int r;
u_int id;
debug2("Sending SSH2_FXP_OPEN \"%s\"", path);
*handlep = NULL;
*handle_lenp = 0;
if (a == NULL) {
attrib_clear(&junk); /* Send empty attributes */
a = &junk;
}
/* Send open request */
if ((msg = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
id = conn->msg_id++;
if ((r = sshbuf_put_u8(msg, SSH2_FXP_OPEN)) != 0 ||
(r = sshbuf_put_u32(msg, id)) != 0 ||
(r = sshbuf_put_cstring(msg, path)) != 0 ||
(r = sshbuf_put_u32(msg, openmode)) != 0 ||
(r = encode_attrib(msg, a)) != 0)
fatal_fr(r, "compose %s open", tag);
send_msg(conn, msg);
sshbuf_free(msg);
debug3("Sent %s message SSH2_FXP_OPEN I:%u P:%s M:0x%04x",
tag, id, path, openmode);
if ((handle = get_handle(conn, id, &handle_len,
"%s open \"%s\"", tag, path)) == NULL)
return -1;
/* success */
*handlep = handle;
*handle_lenp = handle_len;
return 0;
}
static const char *
progress_meter_path(const char *path)
{
const char *progresspath;
if ((progresspath = strrchr(path, '/')) == NULL)
return path;
progresspath++;
if (*progresspath == '\0')
return path;
return progresspath;
}
int
do_download(struct sftp_conn *conn, const char *remote_path,
const char *local_path, Attrib *a, int preserve_flag, int resume_flag,
int fsync_flag, int inplace_flag)
{
struct sshbuf *msg;
u_char *handle;
int local_fd = -1, write_error;
int read_error, write_errno, lmodified = 0, reordered = 0, r;
u_int64_t offset = 0, size, highwater = 0, maxack = 0;
u_int mode, id, buflen, num_req, max_req, status = SSH2_FX_OK;
off_t progress_counter;
size_t handle_len;
struct stat st;
struct requests requests;
struct request *req;
u_char type;
debug2_f("download remote \"%s\" to local \"%s\"",
remote_path, local_path);
TAILQ_INIT(&requests);
if (a == NULL && (a = do_stat(conn, remote_path, 0)) == NULL)
return -1;
/* Do not preserve set[ug]id here, as we do not preserve ownership */
if (a->flags & SSH2_FILEXFER_ATTR_PERMISSIONS)
mode = a->perm & 0777;
else
mode = 0666;
if ((a->flags & SSH2_FILEXFER_ATTR_PERMISSIONS) &&
(!S_ISREG(a->perm))) {
error("download %s: not a regular file", remote_path);
return(-1);
}
if (a->flags & SSH2_FILEXFER_ATTR_SIZE)
size = a->size;
else
size = 0;
buflen = conn->download_buflen;
/* Send open request */
if (send_open(conn, remote_path, "remote", SSH2_FXF_READ, NULL,
&handle, &handle_len) != 0)
return -1;
local_fd = open(local_path, O_WRONLY | O_CREAT |
((resume_flag || inplace_flag) ? 0 : O_TRUNC), mode | S_IWUSR);
if (local_fd == -1) {
error("open local \"%s\": %s", local_path, strerror(errno));
goto fail;
}
if (resume_flag) {
if (fstat(local_fd, &st) == -1) {
error("stat local \"%s\": %s",
local_path, strerror(errno));
goto fail;
}
if (st.st_size < 0) {
error("\"%s\" has negative size", local_path);
goto fail;
}
if ((u_int64_t)st.st_size > size) {
error("Unable to resume download of \"%s\": "
"local file is larger than remote", local_path);
fail:
do_close(conn, handle, handle_len);
free(handle);
if (local_fd != -1)
close(local_fd);
return -1;
}
offset = highwater = maxack = st.st_size;
}
/* Read from remote and write to local */
write_error = read_error = write_errno = num_req = 0;
max_req = 1;
progress_counter = offset;
if (showprogress && size != 0) {
start_progress_meter(progress_meter_path(remote_path),
size, &progress_counter);
}
if ((msg = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
while (num_req > 0 || max_req > 0) {
u_char *data;
size_t len;
/*
* Simulate EOF on interrupt: stop sending new requests and
* allow outstanding requests to drain gracefully
*/
if (interrupted) {
if (num_req == 0) /* If we haven't started yet... */
break;
max_req = 0;
}
/* Send some more requests */
while (num_req < max_req) {
debug3("Request range %llu -> %llu (%d/%d)",
(unsigned long long)offset,
(unsigned long long)offset + buflen - 1,
num_req, max_req);
req = request_enqueue(&requests, conn->msg_id++,
buflen, offset);
offset += buflen;
num_req++;
send_read_request(conn, req->id, req->offset,
req->len, handle, handle_len);
}
sshbuf_reset(msg);
get_msg(conn, msg);
if ((r = sshbuf_get_u8(msg, &type)) != 0 ||
(r = sshbuf_get_u32(msg, &id)) != 0)
fatal_fr(r, "parse");
debug3("Received reply T:%u I:%u R:%d", type, id, max_req);
/* Find the request in our queue */
if ((req = request_find(&requests, id)) == NULL)
fatal("Unexpected reply %u", id);
switch (type) {
case SSH2_FXP_STATUS:
if ((r = sshbuf_get_u32(msg, &status)) != 0)
fatal_fr(r, "parse status");
if (status != SSH2_FX_EOF)
read_error = 1;
max_req = 0;
TAILQ_REMOVE(&requests, req, tq);
free(req);
num_req--;
break;
case SSH2_FXP_DATA:
if ((r = sshbuf_get_string(msg, &data, &len)) != 0)
fatal_fr(r, "parse data");
debug3("Received data %llu -> %llu",
(unsigned long long)req->offset,
(unsigned long long)req->offset + len - 1);
if (len > req->len)
fatal("Received more data than asked for "
"%zu > %zu", len, req->len);
lmodified = 1;
if ((lseek(local_fd, req->offset, SEEK_SET) == -1 ||
atomicio(vwrite, local_fd, data, len) != len) &&
!write_error) {
write_errno = errno;
write_error = 1;
max_req = 0;
} else {
/*
* Track both the highest offset acknowledged
* and the highest *contiguous* offset
* acknowledged.
* We'll need the latter for ftruncate()ing
* interrupted transfers.
*/
if (maxack < req->offset + len)
maxack = req->offset + len;
if (!reordered && req->offset <= highwater)
highwater = maxack;
else if (!reordered && req->offset > highwater)
reordered = 1;
}
progress_counter += len;
free(data);
if (len == req->len) {
TAILQ_REMOVE(&requests, req, tq);
free(req);
num_req--;
} else {
/* Resend the request for the missing data */
debug3("Short data block, re-requesting "
"%llu -> %llu (%2d)",
(unsigned long long)req->offset + len,
(unsigned long long)req->offset +
req->len - 1, num_req);
req->id = conn->msg_id++;
req->len -= len;
req->offset += len;
send_read_request(conn, req->id,
req->offset, req->len, handle, handle_len);
/* Reduce the request size */
if (len < buflen)
buflen = MAXIMUM(MIN_READ_SIZE, len);
}
if (max_req > 0) { /* max_req = 0 iff EOF received */
if (size > 0 && offset > size) {
/* Only one request at a time
* after the expected EOF */
debug3("Finish at %llu (%2d)",
(unsigned long long)offset,
num_req);
max_req = 1;
} else if (max_req < conn->num_requests) {
++max_req;
}
}
break;
default:
fatal("Expected SSH2_FXP_DATA(%u) packet, got %u",
SSH2_FXP_DATA, type);
}
}
if (showprogress && size)
stop_progress_meter();
/* Sanity check */
if (TAILQ_FIRST(&requests) != NULL)
fatal("Transfer complete, but requests still in queue");
if (!read_error && !write_error && !interrupted) {
/* we got everything */
highwater = maxack;
}
/*
* Truncate at highest contiguous point to avoid holes on interrupt,
* or unconditionally if writing in place.
*/
if (inplace_flag || read_error || write_error || interrupted) {
if (reordered && resume_flag &&
(read_error || write_error || interrupted)) {
error("Unable to resume download of \"%s\": "
"server reordered requests", local_path);
}
debug("truncating at %llu", (unsigned long long)highwater);
if (ftruncate(local_fd, highwater) == -1)
error("local ftruncate \"%s\": %s", local_path,
strerror(errno));
}
if (read_error) {
error("read remote \"%s\" : %s", remote_path, fx2txt(status));
status = -1;
do_close(conn, handle, handle_len);
} else if (write_error) {
error("write local \"%s\": %s", local_path,
strerror(write_errno));
status = SSH2_FX_FAILURE;
do_close(conn, handle, handle_len);
} else {
if (do_close(conn, handle, handle_len) != 0 || interrupted)
status = SSH2_FX_FAILURE;
else
status = SSH2_FX_OK;
/* Override umask and utimes if asked */
#ifdef HAVE_FCHMOD
if (preserve_flag && fchmod(local_fd, mode) == -1)
#else
if (preserve_flag && chmod(local_path, mode) == -1)
#endif /* HAVE_FCHMOD */
error("local chmod \"%s\": %s", local_path,
strerror(errno));
if (preserve_flag &&
(a->flags & SSH2_FILEXFER_ATTR_ACMODTIME)) {
struct timeval tv[2];
tv[0].tv_sec = a->atime;
tv[1].tv_sec = a->mtime;
tv[0].tv_usec = tv[1].tv_usec = 0;
if (utimes(local_path, tv) == -1)
error("local set times \"%s\": %s",
local_path, strerror(errno));
}
if (resume_flag && !lmodified)
logit("File \"%s\" was not modified", local_path);
else if (fsync_flag) {
debug("syncing \"%s\"", local_path);
if (fsync(local_fd) == -1)
error("local sync \"%s\": %s",
local_path, strerror(errno));
}
}
close(local_fd);
sshbuf_free(msg);
free(handle);
return status == SSH2_FX_OK ? 0 : -1;
}
static int
download_dir_internal(struct sftp_conn *conn, const char *src, const char *dst,
int depth, Attrib *dirattrib, int preserve_flag, int print_flag,
int resume_flag, int fsync_flag, int follow_link_flag, int inplace_flag)
{
int i, ret = 0;
SFTP_DIRENT **dir_entries;
char *filename, *new_src = NULL, *new_dst = NULL;
mode_t mode = 0777, tmpmode = mode;
if (depth >= MAX_DIR_DEPTH) {
error("Maximum directory depth exceeded: %d levels", depth);
return -1;
}
debug2_f("download dir remote \"%s\" to local \"%s\"", src, dst);
if (dirattrib == NULL &&
(dirattrib = do_stat(conn, src, 1)) == NULL) {
error("stat remote \"%s\" directory failed", src);
return -1;
}
if (!S_ISDIR(dirattrib->perm)) {
error("\"%s\" is not a directory", src);
return -1;
}
if (print_flag && print_flag != SFTP_PROGRESS_ONLY)
mprintf("Retrieving %s\n", src);
if (dirattrib->flags & SSH2_FILEXFER_ATTR_PERMISSIONS) {
mode = dirattrib->perm & 01777;
tmpmode = mode | (S_IWUSR|S_IXUSR);
} else {
debug("download remote \"%s\": server "
"did not send permissions", dst);
}
if (mkdir(dst, tmpmode) == -1 && errno != EEXIST) {
error("mkdir %s: %s", dst, strerror(errno));
return -1;
}
if (do_readdir(conn, src, &dir_entries) == -1) {
error("remote readdir \"%s\" failed", src);
return -1;
}
for (i = 0; dir_entries[i] != NULL && !interrupted; i++) {
free(new_dst);
free(new_src);
filename = dir_entries[i]->filename;
new_dst = path_append(dst, filename);
new_src = path_append(src, filename);
if (S_ISDIR(dir_entries[i]->a.perm)) {
if (strcmp(filename, ".") == 0 ||
strcmp(filename, "..") == 0)
continue;
if (download_dir_internal(conn, new_src, new_dst,
depth + 1, &(dir_entries[i]->a), preserve_flag,
print_flag, resume_flag,
fsync_flag, follow_link_flag, inplace_flag) == -1)
ret = -1;
} else if (S_ISREG(dir_entries[i]->a.perm) ||
(follow_link_flag && S_ISLNK(dir_entries[i]->a.perm))) {
/*
* If this is a symlink then don't send the link's
* Attrib. do_download() will do a FXP_STAT operation
* and get the link target's attributes.
*/
if (do_download(conn, new_src, new_dst,
S_ISLNK(dir_entries[i]->a.perm) ? NULL :
&(dir_entries[i]->a),
preserve_flag, resume_flag, fsync_flag,
inplace_flag) == -1) {
error("Download of file %s to %s failed",
new_src, new_dst);
ret = -1;
}
} else
logit("download \"%s\": not a regular file", new_src);
}
free(new_dst);
free(new_src);
if (preserve_flag) {
if (dirattrib->flags & SSH2_FILEXFER_ATTR_ACMODTIME) {
struct timeval tv[2];
tv[0].tv_sec = dirattrib->atime;
tv[1].tv_sec = dirattrib->mtime;
tv[0].tv_usec = tv[1].tv_usec = 0;
if (utimes(dst, tv) == -1)
error("local set times on \"%s\": %s",
dst, strerror(errno));
} else
debug("Server did not send times for directory "
"\"%s\"", dst);
}
if (mode != tmpmode && chmod(dst, mode) == -1)
error("local chmod directory \"%s\": %s", dst,
strerror(errno));
free_sftp_dirents(dir_entries);
return ret;
}
int
download_dir(struct sftp_conn *conn, const char *src, const char *dst,
Attrib *dirattrib, int preserve_flag, int print_flag, int resume_flag,
int fsync_flag, int follow_link_flag, int inplace_flag)
{
char *src_canon;
int ret;
if ((src_canon = do_realpath(conn, src)) == NULL) {
error("download \"%s\": path canonicalization failed", src);
return -1;
}
ret = download_dir_internal(conn, src_canon, dst, 0,
dirattrib, preserve_flag, print_flag, resume_flag, fsync_flag,
follow_link_flag, inplace_flag);
free(src_canon);
return ret;
}
int
do_upload(struct sftp_conn *conn, const char *local_path,
const char *remote_path, int preserve_flag, int resume,
int fsync_flag, int inplace_flag)
{
int r, local_fd;
u_int openmode, id, status = SSH2_FX_OK, reordered = 0;
off_t offset, progress_counter;
u_char type, *handle, *data;
struct sshbuf *msg;
struct stat sb;
Attrib a, t, *c = NULL;
u_int32_t startid, ackid;
u_int64_t highwater = 0, maxack = 0;
struct request *ack = NULL;
struct requests acks;
size_t handle_len;
debug2_f("upload local \"%s\" to remote \"%s\"",
local_path, remote_path);
TAILQ_INIT(&acks);
if ((local_fd = open(local_path, O_RDONLY)) == -1) {
error("open local \"%s\": %s", local_path, strerror(errno));
return(-1);
}
if (fstat(local_fd, &sb) == -1) {
error("fstat local \"%s\": %s", local_path, strerror(errno));
close(local_fd);
return(-1);
}
if (!S_ISREG(sb.st_mode)) {
error("local \"%s\" is not a regular file", local_path);
close(local_fd);
return(-1);
}
stat_to_attrib(&sb, &a);
a.flags &= ~SSH2_FILEXFER_ATTR_SIZE;
a.flags &= ~SSH2_FILEXFER_ATTR_UIDGID;
a.perm &= 0777;
if (!preserve_flag)
a.flags &= ~SSH2_FILEXFER_ATTR_ACMODTIME;
if (resume) {
/* Get remote file size if it exists */
if ((c = do_stat(conn, remote_path, 0)) == NULL) {
close(local_fd);
return -1;
}
if ((off_t)c->size >= sb.st_size) {
error("resume \"%s\": destination file "
"same size or larger", local_path);
close(local_fd);
return -1;
}
if (lseek(local_fd, (off_t)c->size, SEEK_SET) == -1) {
close(local_fd);
return -1;
}
}
openmode = SSH2_FXF_WRITE|SSH2_FXF_CREAT;
if (resume)
openmode |= SSH2_FXF_APPEND;
else if (!inplace_flag)
openmode |= SSH2_FXF_TRUNC;
/* Send open request */
if (send_open(conn, remote_path, "dest", openmode, &a,
&handle, &handle_len) != 0) {
close(local_fd);
return -1;
}
id = conn->msg_id;
startid = ackid = id + 1;
data = xmalloc(conn->upload_buflen);
/* Read from local and write to remote */
offset = progress_counter = (resume ? c->size : 0);
if (showprogress) {
start_progress_meter(progress_meter_path(local_path),
sb.st_size, &progress_counter);
}
if ((msg = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
for (;;) {
int len;
/*
* Can't use atomicio here because it returns 0 on EOF,
* thus losing the last block of the file.
* Simulate an EOF on interrupt, allowing ACKs from the
* server to drain.
*/
if (interrupted || status != SSH2_FX_OK)
len = 0;
else do
len = read(local_fd, data, conn->upload_buflen);
while ((len == -1) &&
(errno == EINTR || errno == EAGAIN || errno == EWOULDBLOCK));
if (len == -1) {
fatal("read local \"%s\": %s",
local_path, strerror(errno));
} else if (len != 0) {
ack = request_enqueue(&acks, ++id, len, offset);
sshbuf_reset(msg);
if ((r = sshbuf_put_u8(msg, SSH2_FXP_WRITE)) != 0 ||
(r = sshbuf_put_u32(msg, ack->id)) != 0 ||
(r = sshbuf_put_string(msg, handle,
handle_len)) != 0 ||
(r = sshbuf_put_u64(msg, offset)) != 0 ||
(r = sshbuf_put_string(msg, data, len)) != 0)
fatal_fr(r, "compose");
send_msg(conn, msg);
debug3("Sent message SSH2_FXP_WRITE I:%u O:%llu S:%u",
id, (unsigned long long)offset, len);
} else if (TAILQ_FIRST(&acks) == NULL)
break;
if (ack == NULL)
fatal("Unexpected ACK %u", id);
if (id == startid || len == 0 ||
id - ackid >= conn->num_requests) {
u_int rid;
sshbuf_reset(msg);
get_msg(conn, msg);
if ((r = sshbuf_get_u8(msg, &type)) != 0 ||
(r = sshbuf_get_u32(msg, &rid)) != 0)
fatal_fr(r, "parse");
if (type != SSH2_FXP_STATUS)
fatal("Expected SSH2_FXP_STATUS(%d) packet, "
"got %d", SSH2_FXP_STATUS, type);
if ((r = sshbuf_get_u32(msg, &status)) != 0)
fatal_fr(r, "parse status");
debug3("SSH2_FXP_STATUS %u", status);
/* Find the request in our queue */
if ((ack = request_find(&acks, rid)) == NULL)
fatal("Can't find request for ID %u", rid);
TAILQ_REMOVE(&acks, ack, tq);
debug3("In write loop, ack for %u %zu bytes at %lld",
ack->id, ack->len, (unsigned long long)ack->offset);
++ackid;
progress_counter += ack->len;
/*
* Track both the highest offset acknowledged and the
* highest *contiguous* offset acknowledged.
* We'll need the latter for ftruncate()ing
* interrupted transfers.
*/
if (maxack < ack->offset + ack->len)
maxack = ack->offset + ack->len;
if (!reordered && ack->offset <= highwater)
highwater = maxack;
else if (!reordered && ack->offset > highwater) {
debug3_f("server reordered ACKs");
reordered = 1;
}
free(ack);
}
offset += len;
if (offset < 0)
fatal_f("offset < 0");
}
sshbuf_free(msg);
if (showprogress)
stop_progress_meter();
free(data);
if (status == SSH2_FX_OK && !interrupted) {
/* we got everything */
highwater = maxack;
}
if (status != SSH2_FX_OK) {
error("write remote \"%s\": %s", remote_path, fx2txt(status));
status = SSH2_FX_FAILURE;
}
if (inplace_flag || (resume && (status != SSH2_FX_OK || interrupted))) {
debug("truncating at %llu", (unsigned long long)highwater);
attrib_clear(&t);
t.flags = SSH2_FILEXFER_ATTR_SIZE;
t.size = highwater;
do_fsetstat(conn, handle, handle_len, &t);
}
if (close(local_fd) == -1) {
error("close local \"%s\": %s", local_path, strerror(errno));
status = SSH2_FX_FAILURE;
}
/* Override umask and utimes if asked */
if (preserve_flag)
do_fsetstat(conn, handle, handle_len, &a);
if (fsync_flag)
(void)do_fsync(conn, handle, handle_len);
if (do_close(conn, handle, handle_len) != 0)
status = SSH2_FX_FAILURE;
free(handle);
return status == SSH2_FX_OK ? 0 : -1;
}
static int
upload_dir_internal(struct sftp_conn *conn, const char *src, const char *dst,
int depth, int preserve_flag, int print_flag, int resume, int fsync_flag,
int follow_link_flag, int inplace_flag)
{
int ret = 0;
DIR *dirp;
struct dirent *dp;
char *filename, *new_src = NULL, *new_dst = NULL;
struct stat sb;
Attrib a, *dirattrib;
u_int32_t saved_perm;
debug2_f("upload local dir \"%s\" to remote \"%s\"", src, dst);
if (depth >= MAX_DIR_DEPTH) {
error("Maximum directory depth exceeded: %d levels", depth);
return -1;
}
if (stat(src, &sb) == -1) {
error("stat local \"%s\": %s", src, strerror(errno));
return -1;
}
if (!S_ISDIR(sb.st_mode)) {
error("\"%s\" is not a directory", src);
return -1;
}
if (print_flag && print_flag != SFTP_PROGRESS_ONLY)
mprintf("Entering %s\n", src);
stat_to_attrib(&sb, &a);
a.flags &= ~SSH2_FILEXFER_ATTR_SIZE;
a.flags &= ~SSH2_FILEXFER_ATTR_UIDGID;
a.perm &= 01777;
if (!preserve_flag)
a.flags &= ~SSH2_FILEXFER_ATTR_ACMODTIME;
/*
* sftp lacks a portable status value to match errno EEXIST,
* so if we get a failure back then we must check whether
* the path already existed and is a directory. Ensure we can
* write to the directory we create for the duration of the transfer.
*/
saved_perm = a.perm;
a.perm |= (S_IWUSR|S_IXUSR);
if (do_mkdir(conn, dst, &a, 0) != 0) {
if ((dirattrib = do_stat(conn, dst, 0)) == NULL)
return -1;
if (!S_ISDIR(dirattrib->perm)) {
error("\"%s\" exists but is not a directory", dst);
return -1;
}
}
a.perm = saved_perm;
if ((dirp = opendir(src)) == NULL) {
error("local opendir \"%s\": %s", src, strerror(errno));
return -1;
}
while (((dp = readdir(dirp)) != NULL) && !interrupted) {
if (dp->d_ino == 0)
continue;
free(new_dst);
free(new_src);
filename = dp->d_name;
new_dst = path_append(dst, filename);
new_src = path_append(src, filename);
if (lstat(new_src, &sb) == -1) {
logit("local lstat \"%s\": %s", filename,
strerror(errno));
ret = -1;
} else if (S_ISDIR(sb.st_mode)) {
if (strcmp(filename, ".") == 0 ||
strcmp(filename, "..") == 0)
continue;
if (upload_dir_internal(conn, new_src, new_dst,
depth + 1, preserve_flag, print_flag, resume,
fsync_flag, follow_link_flag, inplace_flag) == -1)
ret = -1;
} else if (S_ISREG(sb.st_mode) ||
(follow_link_flag && S_ISLNK(sb.st_mode))) {
if (do_upload(conn, new_src, new_dst,
preserve_flag, resume, fsync_flag,
inplace_flag) == -1) {
error("upload \"%s\" to \"%s\" failed",
new_src, new_dst);
ret = -1;
}
} else
logit("%s: not a regular file", filename);
}
free(new_dst);
free(new_src);
do_setstat(conn, dst, &a);
(void) closedir(dirp);
return ret;
}
int
upload_dir(struct sftp_conn *conn, const char *src, const char *dst,
int preserve_flag, int print_flag, int resume, int fsync_flag,
int follow_link_flag, int inplace_flag)
{
char *dst_canon;
int ret;
if ((dst_canon = do_realpath(conn, dst)) == NULL) {
error("upload \"%s\": path canonicalization failed", dst);
return -1;
}
ret = upload_dir_internal(conn, src, dst_canon, 0, preserve_flag,
print_flag, resume, fsync_flag, follow_link_flag, inplace_flag);
free(dst_canon);
return ret;
}
static void
handle_dest_replies(struct sftp_conn *to, const char *to_path, int synchronous,
u_int *nreqsp, u_int *write_errorp)
{
struct sshbuf *msg;
u_char type;
u_int id, status;
int r;
struct pollfd pfd;
if ((msg = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
/* Try to eat replies from the upload side */
while (*nreqsp > 0) {
debug3_f("%u outstanding replies", *nreqsp);
if (!synchronous) {
/* Bail out if no data is ready to be read */
pfd.fd = to->fd_in;
pfd.events = POLLIN;
if ((r = poll(&pfd, 1, 0)) == -1) {
if (errno == EINTR)
break;
fatal_f("poll: %s", strerror(errno));
} else if (r == 0)
break; /* fd not ready */
}
sshbuf_reset(msg);
get_msg(to, msg);
if ((r = sshbuf_get_u8(msg, &type)) != 0 ||
(r = sshbuf_get_u32(msg, &id)) != 0)
fatal_fr(r, "dest parse");
debug3("Received dest reply T:%u I:%u R:%u", type, id, *nreqsp);
if (type != SSH2_FXP_STATUS) {
fatal_f("Expected SSH2_FXP_STATUS(%d) packet, got %d",
SSH2_FXP_STATUS, type);
}
if ((r = sshbuf_get_u32(msg, &status)) != 0)
fatal_fr(r, "parse dest status");
debug3("dest SSH2_FXP_STATUS %u", status);
if (status != SSH2_FX_OK) {
/* record first error */
if (*write_errorp == 0)
*write_errorp = status;
}
/*
* XXX this doesn't do full reply matching like do_upload and
* so cannot gracefully truncate terminated uploads at a
* high-water mark. ATM the only caller of this function (scp)
* doesn't support transfer resumption, so this doesn't matter
* a whole lot.
*
* To be safe, do_crossload truncates the destination file to
* zero length on upload failure, since we can't trust the
* server not to have reordered replies that could have
* inserted holes where none existed in the source file.
*
* XXX we could get a more accutate progress bar if we updated
* the counter based on the reply from the destination...
*/
(*nreqsp)--;
}
debug3_f("done: %u outstanding replies", *nreqsp);
sshbuf_free(msg);
}
int
do_crossload(struct sftp_conn *from, struct sftp_conn *to,
const char *from_path, const char *to_path,
Attrib *a, int preserve_flag)
{
struct sshbuf *msg;
int write_error, read_error, r;
u_int64_t offset = 0, size;
u_int id, buflen, num_req, max_req, status = SSH2_FX_OK;
u_int num_upload_req;
off_t progress_counter;
u_char *from_handle, *to_handle;
size_t from_handle_len, to_handle_len;
struct requests requests;
struct request *req;
u_char type;
debug2_f("crossload src \"%s\" to dst \"%s\"", from_path, to_path);
TAILQ_INIT(&requests);
if (a == NULL && (a = do_stat(from, from_path, 0)) == NULL)
return -1;
if ((a->flags & SSH2_FILEXFER_ATTR_PERMISSIONS) &&
(!S_ISREG(a->perm))) {
error("download \"%s\": not a regular file", from_path);
return(-1);
}
if (a->flags & SSH2_FILEXFER_ATTR_SIZE)
size = a->size;
else
size = 0;
buflen = from->download_buflen;
if (buflen > to->upload_buflen)
buflen = to->upload_buflen;
/* Send open request to read side */
if (send_open(from, from_path, "origin", SSH2_FXF_READ, NULL,
&from_handle, &from_handle_len) != 0)
return -1;
/* Send open request to write side */
a->flags &= ~SSH2_FILEXFER_ATTR_SIZE;
a->flags &= ~SSH2_FILEXFER_ATTR_UIDGID;
a->perm &= 0777;
if (!preserve_flag)
a->flags &= ~SSH2_FILEXFER_ATTR_ACMODTIME;
if (send_open(to, to_path, "dest",
SSH2_FXF_WRITE|SSH2_FXF_CREAT|SSH2_FXF_TRUNC, a,
&to_handle, &to_handle_len) != 0) {
do_close(from, from_handle, from_handle_len);
return -1;
}
/* Read from remote "from" and write to remote "to" */
offset = 0;
write_error = read_error = num_req = num_upload_req = 0;
max_req = 1;
progress_counter = 0;
if (showprogress && size != 0) {
start_progress_meter(progress_meter_path(from_path),
size, &progress_counter);
}
if ((msg = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
while (num_req > 0 || max_req > 0) {
u_char *data;
size_t len;
/*
* Simulate EOF on interrupt: stop sending new requests and
* allow outstanding requests to drain gracefully
*/
if (interrupted) {
if (num_req == 0) /* If we haven't started yet... */
break;
max_req = 0;
}
/* Send some more requests */
while (num_req < max_req) {
debug3("Request range %llu -> %llu (%d/%d)",
(unsigned long long)offset,
(unsigned long long)offset + buflen - 1,
num_req, max_req);
req = request_enqueue(&requests, from->msg_id++,
buflen, offset);
offset += buflen;
num_req++;
send_read_request(from, req->id, req->offset,
req->len, from_handle, from_handle_len);
}
/* Try to eat replies from the upload side (nonblocking) */
handle_dest_replies(to, to_path, 0,
&num_upload_req, &write_error);
sshbuf_reset(msg);
get_msg(from, msg);
if ((r = sshbuf_get_u8(msg, &type)) != 0 ||
(r = sshbuf_get_u32(msg, &id)) != 0)
fatal_fr(r, "parse");
debug3("Received origin reply T:%u I:%u R:%d",
type, id, max_req);
/* Find the request in our queue */
if ((req = request_find(&requests, id)) == NULL)
fatal("Unexpected reply %u", id);
switch (type) {
case SSH2_FXP_STATUS:
if ((r = sshbuf_get_u32(msg, &status)) != 0)
fatal_fr(r, "parse status");
if (status != SSH2_FX_EOF)
read_error = 1;
max_req = 0;
TAILQ_REMOVE(&requests, req, tq);
free(req);
num_req--;
break;
case SSH2_FXP_DATA:
if ((r = sshbuf_get_string(msg, &data, &len)) != 0)
fatal_fr(r, "parse data");
debug3("Received data %llu -> %llu",
(unsigned long long)req->offset,
(unsigned long long)req->offset + len - 1);
if (len > req->len)
fatal("Received more data than asked for "
"%zu > %zu", len, req->len);
/* Write this chunk out to the destination */
sshbuf_reset(msg);
if ((r = sshbuf_put_u8(msg, SSH2_FXP_WRITE)) != 0 ||
(r = sshbuf_put_u32(msg, to->msg_id++)) != 0 ||
(r = sshbuf_put_string(msg, to_handle,
to_handle_len)) != 0 ||
(r = sshbuf_put_u64(msg, req->offset)) != 0 ||
(r = sshbuf_put_string(msg, data, len)) != 0)
fatal_fr(r, "compose write");
send_msg(to, msg);
debug3("Sent message SSH2_FXP_WRITE I:%u O:%llu S:%zu",
id, (unsigned long long)offset, len);
num_upload_req++;
progress_counter += len;
free(data);
if (len == req->len) {
TAILQ_REMOVE(&requests, req, tq);
free(req);
num_req--;
} else {
/* Resend the request for the missing data */
debug3("Short data block, re-requesting "
"%llu -> %llu (%2d)",
(unsigned long long)req->offset + len,
(unsigned long long)req->offset +
req->len - 1, num_req);
req->id = from->msg_id++;
req->len -= len;
req->offset += len;
send_read_request(from, req->id,
req->offset, req->len,
from_handle, from_handle_len);
/* Reduce the request size */
if (len < buflen)
buflen = MAXIMUM(MIN_READ_SIZE, len);
}
if (max_req > 0) { /* max_req = 0 iff EOF received */
if (size > 0 && offset > size) {
/* Only one request at a time
* after the expected EOF */
debug3("Finish at %llu (%2d)",
(unsigned long long)offset,
num_req);
max_req = 1;
} else if (max_req < from->num_requests) {
++max_req;
}
}
break;
default:
fatal("Expected SSH2_FXP_DATA(%u) packet, got %u",
SSH2_FXP_DATA, type);
}
}
if (showprogress && size)
stop_progress_meter();
/* Drain replies from the server (blocking) */
debug3_f("waiting for %u replies from destination", num_upload_req);
handle_dest_replies(to, to_path, 1, &num_upload_req, &write_error);
/* Sanity check */
if (TAILQ_FIRST(&requests) != NULL)
fatal("Transfer complete, but requests still in queue");
/* Truncate at 0 length on interrupt or error to avoid holes at dest */
if (read_error || write_error || interrupted) {
debug("truncating \"%s\" at 0", to_path);
do_close(to, to_handle, to_handle_len);
free(to_handle);
if (send_open(to, to_path, "dest",
SSH2_FXF_WRITE|SSH2_FXF_CREAT|SSH2_FXF_TRUNC, a,
&to_handle, &to_handle_len) != 0) {
error("dest truncate \"%s\" failed", to_path);
to_handle = NULL;
}
}
if (read_error) {
error("read origin \"%s\": %s", from_path, fx2txt(status));
status = -1;
do_close(from, from_handle, from_handle_len);
if (to_handle != NULL)
do_close(to, to_handle, to_handle_len);
} else if (write_error) {
error("write dest \"%s\": %s", to_path, fx2txt(write_error));
status = SSH2_FX_FAILURE;
do_close(from, from_handle, from_handle_len);
if (to_handle != NULL)
do_close(to, to_handle, to_handle_len);
} else {
if (do_close(from, from_handle, from_handle_len) != 0 ||
interrupted)
status = -1;
else
status = SSH2_FX_OK;
if (to_handle != NULL) {
/* Need to resend utimes after write */
if (preserve_flag)
do_fsetstat(to, to_handle, to_handle_len, a);
do_close(to, to_handle, to_handle_len);
}
}
sshbuf_free(msg);
free(from_handle);
free(to_handle);
return status == SSH2_FX_OK ? 0 : -1;
}
static int
crossload_dir_internal(struct sftp_conn *from, struct sftp_conn *to,
const char *from_path, const char *to_path,
int depth, Attrib *dirattrib, int preserve_flag, int print_flag,
int follow_link_flag)
{
int i, ret = 0;
SFTP_DIRENT **dir_entries;
char *filename, *new_from_path = NULL, *new_to_path = NULL;
mode_t mode = 0777;
Attrib curdir;
debug2_f("crossload dir src \"%s\" to dst \"%s\"", from_path, to_path);
if (depth >= MAX_DIR_DEPTH) {
error("Maximum directory depth exceeded: %d levels", depth);
return -1;
}
if (dirattrib == NULL &&
(dirattrib = do_stat(from, from_path, 1)) == NULL) {
error("stat remote \"%s\" failed", from_path);
return -1;
}
if (!S_ISDIR(dirattrib->perm)) {
error("\"%s\" is not a directory", from_path);
return -1;
}
if (print_flag && print_flag != SFTP_PROGRESS_ONLY)
mprintf("Retrieving %s\n", from_path);
curdir = *dirattrib; /* dirattrib will be clobbered */
curdir.flags &= ~SSH2_FILEXFER_ATTR_SIZE;
curdir.flags &= ~SSH2_FILEXFER_ATTR_UIDGID;
if ((curdir.flags & SSH2_FILEXFER_ATTR_PERMISSIONS) == 0) {
debug("Origin did not send permissions for "
"directory \"%s\"", to_path);
curdir.perm = S_IWUSR|S_IXUSR;
curdir.flags |= SSH2_FILEXFER_ATTR_PERMISSIONS;
}
/* We need to be able to write to the directory while we transfer it */
mode = curdir.perm & 01777;
curdir.perm = mode | (S_IWUSR|S_IXUSR);
/*
* sftp lacks a portable status value to match errno EEXIST,
* so if we get a failure back then we must check whether
* the path already existed and is a directory. Ensure we can
* write to the directory we create for the duration of the transfer.
*/
if (do_mkdir(to, to_path, &curdir, 0) != 0) {
if ((dirattrib = do_stat(to, to_path, 0)) == NULL)
return -1;
if (!S_ISDIR(dirattrib->perm)) {
error("\"%s\" exists but is not a directory", to_path);
return -1;
}
}
curdir.perm = mode;
if (do_readdir(from, from_path, &dir_entries) == -1) {
error("origin readdir \"%s\" failed", from_path);
return -1;
}
for (i = 0; dir_entries[i] != NULL && !interrupted; i++) {
free(new_from_path);
free(new_to_path);
filename = dir_entries[i]->filename;
new_from_path = path_append(from_path, filename);
new_to_path = path_append(to_path, filename);
if (S_ISDIR(dir_entries[i]->a.perm)) {
if (strcmp(filename, ".") == 0 ||
strcmp(filename, "..") == 0)
continue;
if (crossload_dir_internal(from, to,
new_from_path, new_to_path,
depth + 1, &(dir_entries[i]->a), preserve_flag,
print_flag, follow_link_flag) == -1)
ret = -1;
} else if (S_ISREG(dir_entries[i]->a.perm) ||
(follow_link_flag && S_ISLNK(dir_entries[i]->a.perm))) {
/*
* If this is a symlink then don't send the link's
* Attrib. do_download() will do a FXP_STAT operation
* and get the link target's attributes.
*/
if (do_crossload(from, to, new_from_path, new_to_path,
S_ISLNK(dir_entries[i]->a.perm) ? NULL :
&(dir_entries[i]->a), preserve_flag) == -1) {
error("crossload \"%s\" to \"%s\" failed",
new_from_path, new_to_path);
ret = -1;
}
} else {
logit("origin \"%s\": not a regular file",
new_from_path);
}
}
free(new_to_path);
free(new_from_path);
do_setstat(to, to_path, &curdir);
free_sftp_dirents(dir_entries);
return ret;
}
int
crossload_dir(struct sftp_conn *from, struct sftp_conn *to,
const char *from_path, const char *to_path,
Attrib *dirattrib, int preserve_flag, int print_flag, int follow_link_flag)
{
char *from_path_canon;
int ret;
if ((from_path_canon = do_realpath(from, from_path)) == NULL) {
error("crossload \"%s\": path canonicalization failed",
from_path);
return -1;
}
ret = crossload_dir_internal(from, to, from_path_canon, to_path, 0,
dirattrib, preserve_flag, print_flag, follow_link_flag);
free(from_path_canon);
return ret;
}
int
can_get_users_groups_by_id(struct sftp_conn *conn)
{
return (conn->exts & SFTP_EXT_GETUSERSGROUPS_BY_ID) != 0;
}
int
do_get_users_groups_by_id(struct sftp_conn *conn,
const u_int *uids, u_int nuids,
const u_int *gids, u_int ngids,
char ***usernamesp, char ***groupnamesp)
{
struct sshbuf *msg, *uidbuf, *gidbuf;
u_int i, expected_id, id;
char *name, **usernames = NULL, **groupnames = NULL;
u_char type;
int r;
*usernamesp = *groupnamesp = NULL;
if (!can_get_users_groups_by_id(conn))
return SSH_ERR_FEATURE_UNSUPPORTED;
if ((msg = sshbuf_new()) == NULL ||
(uidbuf = sshbuf_new()) == NULL ||
(gidbuf = sshbuf_new()) == NULL)
fatal_f("sshbuf_new failed");
expected_id = id = conn->msg_id++;
debug2("Sending SSH2_FXP_EXTENDED(users-groups-by-id@openssh.com)");
for (i = 0; i < nuids; i++) {
if ((r = sshbuf_put_u32(uidbuf, uids[i])) != 0)
fatal_fr(r, "compose uids");
}
for (i = 0; i < ngids; i++) {
if ((r = sshbuf_put_u32(gidbuf, gids[i])) != 0)
fatal_fr(r, "compose gids");
}
if ((r = sshbuf_put_u8(msg, SSH2_FXP_EXTENDED)) != 0 ||
(r = sshbuf_put_u32(msg, id)) != 0 ||
(r = sshbuf_put_cstring(msg,
"users-groups-by-id@openssh.com")) != 0 ||
(r = sshbuf_put_stringb(msg, uidbuf)) != 0 ||
(r = sshbuf_put_stringb(msg, gidbuf)) != 0)
fatal_fr(r, "compose");
send_msg(conn, msg);
get_msg(conn, msg);
if ((r = sshbuf_get_u8(msg, &type)) != 0 ||
(r = sshbuf_get_u32(msg, &id)) != 0)
fatal_fr(r, "parse");
if (id != expected_id)
fatal("ID mismatch (%u != %u)", id, expected_id);
if (type == SSH2_FXP_STATUS) {
u_int status;
char *errmsg;
if ((r = sshbuf_get_u32(msg, &status)) != 0 ||
(r = sshbuf_get_cstring(msg, &errmsg, NULL)) != 0)
fatal_fr(r, "parse status");
error("users-groups-by-id %s",
*errmsg == '\0' ? fx2txt(status) : errmsg);
free(errmsg);
sshbuf_free(msg);
sshbuf_free(uidbuf);
sshbuf_free(gidbuf);
return -1;
} else if (type != SSH2_FXP_EXTENDED_REPLY)
fatal("Expected SSH2_FXP_EXTENDED_REPLY(%u) packet, got %u",
SSH2_FXP_EXTENDED_REPLY, type);
/* reuse */
sshbuf_free(uidbuf);
sshbuf_free(gidbuf);
uidbuf = gidbuf = NULL;
if ((r = sshbuf_froms(msg, &uidbuf)) != 0 ||
(r = sshbuf_froms(msg, &gidbuf)) != 0)
fatal_fr(r, "parse response");
if (nuids > 0) {
usernames = xcalloc(nuids, sizeof(*usernames));
for (i = 0; i < nuids; i++) {
if ((r = sshbuf_get_cstring(uidbuf, &name, NULL)) != 0)
fatal_fr(r, "parse user name");
/* Handle unresolved names */
if (*name == '\0') {
free(name);
name = NULL;
}
usernames[i] = name;
}
}
if (ngids > 0) {
groupnames = xcalloc(ngids, sizeof(*groupnames));
for (i = 0; i < ngids; i++) {
if ((r = sshbuf_get_cstring(gidbuf, &name, NULL)) != 0)
fatal_fr(r, "parse user name");
/* Handle unresolved names */
if (*name == '\0') {
free(name);
name = NULL;
}
groupnames[i] = name;
}
}
if (sshbuf_len(uidbuf) != 0)
fatal_f("unexpected extra username data");
if (sshbuf_len(gidbuf) != 0)
fatal_f("unexpected extra groupname data");
sshbuf_free(uidbuf);
sshbuf_free(gidbuf);
sshbuf_free(msg);
/* success */
*usernamesp = usernames;
*groupnamesp = groupnames;
return 0;
}
char *
path_append(const char *p1, const char *p2)
{
char *ret;
size_t len = strlen(p1) + strlen(p2) + 2;
ret = xmalloc(len);
strlcpy(ret, p1, len);
if (p1[0] != '\0' && p1[strlen(p1) - 1] != '/')
strlcat(ret, "/", len);
strlcat(ret, p2, len);
return(ret);
}
/*
* Arg p must be dynamically allocated. It will either be returned or
* freed and a replacement allocated. Caller must free returned string.
*/
char *
make_absolute(char *p, const char *pwd)
{
char *abs_str;
/* Derelativise */
if (p && !path_absolute(p)) {
abs_str = path_append(pwd, p);
free(p);
return(abs_str);
} else
return(p);
}
int
remote_is_dir(struct sftp_conn *conn, const char *path)
{
Attrib *a;
/* XXX: report errors? */
if ((a = do_stat(conn, path, 1)) == NULL)
return(0);
if (!(a->flags & SSH2_FILEXFER_ATTR_PERMISSIONS))
return(0);
return(S_ISDIR(a->perm));
}
int
local_is_dir(const char *path)
{
struct stat sb;
/* XXX: report errors? */
if (stat(path, &sb) == -1)
return(0);
return(S_ISDIR(sb.st_mode));
}
/* Check whether path returned from glob(..., GLOB_MARK, ...) is a directory */
int
globpath_is_dir(const char *pathname)
{
size_t l = strlen(pathname);
return l > 0 && pathname[l - 1] == '/';
}