sftp 傳輸大約有 1/5 次失敗，並報告client_loop: send disconnect: Broken pipe我的情況。 sshd 日誌顯示：forced close而不是closed檔案可能比預期小。大約（再次）1/5 次（如果 sftp 失敗並顯示上述訊息）檔案正常，這可能意味著連線在檔案傳輸後立即中斷。
伺服器1：OpenSSH_8.0p1，OpenSSL 1.1.1c FIPS 2019年5月28日（CentOS 8.2.2004） 伺服器
2：OpenSSH_8.3p1，OpenSSL 1.1.1g FIPS 202042121212（Fedora32）
OpenSSH_8.3p1，OpenSSL 1.1.1g FIPS 21 2020 年 4 月（Fedora 32）
客戶端 2：OpenSSH_6.6.1p1、OpenSSL 1.0.1e-fips 2013 年 2 月 11 日（R.HEL/CentOS R.HEL/Cent）
命令（類似）：（下面是長版本）f=file.pdf echo -e "-ls -l $f\nput $f\nls -l $f" | sftp -b- [email protected]:/upload
以下內容似乎無關緊要（儘管在其他地方報告並有時給出“解決方案”）：
ChrootDirectory伺服器端的選項（sshd）- 伺服器端的主目錄權限
-oIPQoS=throughput伺服器和/或客戶端上的選項- 伺服器網路介面上的 MTU 大小（我嘗試過 1200 和 1500）
- 伺服器端防火牆/網關上的流量整形/QoS 設定
- 透過 iptables/firewalld 設定 TCP MSS
但是，如果兩個選項
-B1500-oCiphers=aes128-cbc
使用後，錯誤率要低得多。 -B1000 OR -oCiphers=aes256-cbc OR -oCiphers=aes128-ctr 會提高錯誤等級。
我使用得到的最佳結果-oCiphers=aes128-cbc -oBatchMode=yes -oCompression=no -oIPQoS=throughput -oRequestTTY=no -B1500：
- 109kbyte 檔案的 1216 次傳輸中有 4 個錯誤（0.33% 錯誤）
- 6.81MB 檔案的 2695 次傳輸中有 10 個錯誤（0.37% 錯誤）
當嘗試 sftp/scp sthg 到 vmware 主機時，會發生相同（或類似？）的情況。在“put”命令之後幾乎立即報告“Broken pipeline”。事實證明，只有選項 -B 有幫助。就我而言 -B265 給出了穩定的結果。 -B266 工作過一次，然後總是（我沒有經常嘗試）「斷管」。我實際上是將作業系統映像傳輸到本地資料儲存。
伺服器：OpenSSH_7.9p1，OpenSSL 1.0.2r-fips 2019年2月26日（ESXi 6.7？）
客戶端：OpenSSH_7.4p1，OpenSSL 1.0.2k-fips 2017年1月26日（Fedora
也許有人可以弄清楚這裡發生了什麼事。感覺像是 openssl 的一些問題，可能與 openssh 結合使用 - 不一定與 sftp 本身有關。
我實際上正在使用
dd if=/dev/urandom of=x.rand bs=1021 count=6999
n=1; f=x.rand; while true; do
echo "=== $n ==="
date
echo -e "-ls -l $f\nput $f\nls -l $f" \
| sftp -b- -oPort=22 -B1500 -oCiphers=aes128-cbc -oBatchMode=yes -oCompression=no -oIPQoS=throughput -oRequestTTY=no [email protected]:/upload
n=`expr $n + 1`
sleep 0.1
done 2>&1 | tee output_x_rand.log
來測試這個。
sshd_配置：
Match User user
ForceCommand internal-sftp -m 640 -e -P remove,mkdir,rmdir,symlink,posix-rename,hardlink,readlink,setstat,fsetstat,lsetstat,statvfs,fstatvfs -l INFO
PasswordAuthentication no
GSSAPIAuthentication no
HostbasedAuthentication no
KbdInteractiveAuthentication no
KerberosAuthentication no
PubkeyAuthentication yes
ChrootDirectory /var/transfer/incoming
PermitTunnel no
AllowAgentForwarding no
AllowTcpForwarding no
X11Forwarding no
這就是發生的事情（strace）伺服器，如果連線中斷：
168224 read(5, 0x7fff0f71e350, 16384) = -1 ECONNRESET (Connection reset by peer)
168224 getpid() = 168224
168224 getuid() = 1000
168224 write(9, "\0\0\0{|", 5) = 5
168211 <... poll resumed>) = 1 ([{fd=12, revents=POLLIN}])
168224 write(9, "\0\0\0fSHA256:fc:5f:4c:91:1a:75:eb:"..., 122 <unfinished ...>
168211 read(12, <unfinished ...>
168224 <... write resumed>) = 122
168211 <... read resumed>"\0\0\0{", 4) = 4
168224 getuid( <unfinished ...>
168211 read(12, <unfinished ...>
168224 <... getuid resumed>) = 1000
168211 <... read resumed>"|\0\0\0fSHA256:fc:5f:4c:91:1a:75:eb"..., 123) = 123
168224 getpid( <unfinished ...>
168211 socket(AF_NETLINK, SOCK_RAW, NETLINK_AUDIT <unfinished ...>
168224 <... getpid resumed>) = 168224
168211 <... socket resumed>) = 9
168224 write(9, "\0\0\0\25z", 5 <unfinished ...>
168211 fcntl(9, F_SETFD, FD_CLOEXEC <unfinished ...>
168224 <... write resumed>) = 5
168211 <... fcntl resumed>) = 0
168224 write(9, "\0\0\0\2\0\0\0\0\0\2\221 \0\0\0\0\0\0\3\350", 20 <unfinished ...>
168211 ioctl(0, TCGETS <unfinished ...>
168224 <... write resumed>) = 20
168211 <... ioctl resumed>, 0x7fff0f71fbe0) = -1 ENOTTY (Inappropriate ioctl for device)
168224 read(9, <unfinished ...>
168211 ioctl(1, TCGETS, 0x7fff0f71fbe0) = -1 ENOTTY (Inappropriate ioctl for device)
168211 ioctl(2, TCGETS, 0x7fff0f71fbe0) = -1 ENOTTY (Inappropriate ioctl for device)
168211 sendto(9, {{len=244, type=0x964 /* NLMSG_??? */, flags=NLM_F_REQUEST|NLM_F_ACK, seq=15, pid=0}, "\x6f\x70\x3d\x64\x65\x73\x74\x72\x6f\x79\x20\x6b\x69\x6e\x64\x3d\x73\x65\x72\x76\x65\x72\x20\x66\x70\x3d\x53\x48\x41\x32\x35\x36"...}, 244, 0, {sa_family=AF_NETLINK, nl_pid=0, nl_groups=00000000}, 12) = 244
168211 poll([{fd=9, events=POLLIN}], 1, 500) = 1 ([{fd=9, revents=POLLIN}])
168211 recvfrom(9, {{len=36, type=NLMSG_ERROR, flags=NLM_F_CAPPED, seq=15, pid=168211}, {error=0, msg={len=244, type=0x964 /* AUDIT_??? */, flags=NLM_F_REQUEST|NLM_F_ACK, seq=15, pid=0}}}, 8988, MSG_PEEK|MSG_DONTWAIT, {sa_family=AF_NETLINK, nl_pid=0, nl_groups=00000000}, [12]) = 36
168211 recvfrom(9, {{len=36, type=NLMSG_ERROR, flags=NLM_F_CAPPED, seq=15, pid=168211}, {error=0, msg={len=244, type=0x964 /* AUDIT_??? */, flags=NLM_F_REQUEST|NLM_F_ACK, seq=15, pid=0}}}, 8988, MSG_DONTWAIT, {sa_family=AF_NETLINK, nl_pid=0, nl_groups=00000000}, [12]) = 36
168211 close(9) = 0
168211 poll([{fd=12, events=POLLIN}], 1, -1) = 1 ([{fd=12, revents=POLLIN}])
168211 read(12, "\0\0\0\25", 4) = 4
168211 read(12, "z\0\0\0\2\0\0\0\0\0\2\221 \0\0\0\0\0\0\3\350", 21) = 21
168211 getsockname(5, {sa_family=AF_INET, sin_port=htons(22), sin_addr=inet_addr("172.17.17.2")}, [128->16]) = 0
168211 socket(AF_NETLINK, SOCK_RAW, NETLINK_AUDIT) = 9
168211 fcntl(9, F_SETFD, FD_CLOEXEC) = 0
168211 ioctl(0, TCGETS, 0x7fff0f71fbb0) = -1 ENOTTY (Inappropriate ioctl for device)
168211 ioctl(1, TCGETS, 0x7fff0f71fbb0) = -1 ENOTTY (Inappropriate ioctl for device)
168211 ioctl(2, TCGETS, 0x7fff0f71fbb0) = -1 ENOTTY (Inappropriate ioctl for device)
168211 sendto(9, {{len=196, type=0x964 /* NLMSG_??? */, flags=NLM_F_REQUEST|NLM_F_ACK, seq=16, pid=0}, "\x6f\x70\x3d\x64\x65\x73\x74\x72\x6f\x79\x20\x6b\x69\x6e\x64\x3d\x73\x65\x73\x73\x69\x6f\x6e\x20\x66\x70\x3d\x3f\x20\x64\x69\x72"...}, 196, 0, {sa_family=AF_NETLINK, nl_pid=0, nl_groups=00000000}, 12) = 196
168211 poll([{fd=9, events=POLLIN}], 1, 500) = 1 ([{fd=9, revents=POLLIN}])
168211 recvfrom(9, {{len=36, type=NLMSG_ERROR, flags=NLM_F_CAPPED, seq=16, pid=168211}, {error=0, msg={len=196, type=0x964 /* AUDIT_??? */, flags=NLM_F_REQUEST|NLM_F_ACK, seq=16, pid=0}}}, 8988, MSG_PEEK|MSG_DONTWAIT, {sa_family=AF_NETLINK, nl_pid=0, nl_groups=00000000}, [12]) = 36
168211 recvfrom(9, {{len=36, type=NLMSG_ERROR, flags=NLM_F_CAPPED, seq=16, pid=168211}, {error=0, msg={len=196, type=0x964 /* AUDIT_??? */, flags=NLM_F_REQUEST|NLM_F_ACK, seq=16, pid=0}}}, 8988, MSG_DONTWAIT, {sa_family=AF_NETLINK, nl_pid=0, nl_groups=00000000}, [12]) = 36
168211 close(9) = 0
168211 write(12, "\0\0\0\1{", 5 <unfinished ...>
168224 <... read resumed>"\0\0\0\1", 4) = 4
168211 <... write resumed>) = 5
168224 read(9, <unfinished ...>
168211 poll([{fd=12, events=POLLIN}], 1, -1 <unfinished ...>
168224 <... read resumed>"{", 1) = 1
168224 exit_group(255) = ?
168225 <... select resumed>) = 1 (in [0])
168224 +++ exited with 255 +++
168225 read(0, <unfinished ...>
168211 <... poll resumed>) = ? ERESTART_RESTARTBLOCK (Interrupted by signal)
168225 <... read resumed>"", 16384) = 0
168211 --- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=168224, si_uid=1000, si_status=255, si_utime=0, si_stime=1} ---
168225 write(2, "forced close \"/upload/xx.rand\" b"..., 60 <unfinished ...>
168211 restart_syscall(<... resuming interrupted poll ...> <unfinished ...>
168225 <... write resumed>) = -1 EPIPE (Broken pipe)
168225 --- SIGPIPE {si_signo=SIGPIPE, si_code=SI_USER, si_pid=168225, si_uid=1000} ---
168225 +++ killed by SIGPIPE +++
這就是發生的事情（strace）客戶，如果連線中斷：
48403 select(7, [3 4], [3], NULL, {tv_sec=59, tv_usec=0}) = 1 (in [4], left {tv_sec=58, tv_usec=999997})
48403 read(4, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 16384) = 16384
48403 mmap(NULL, 2019328, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7f00c3731000
48403 munmap(0x7f00c391e000, 2002944) = 0
48403 select(7, [3 4], [3], NULL, {tv_sec=59, tv_usec=0}) = 1 (in [4], left {tv_sec=58, tv_usec=999997})
48403 read(4, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 16384) = 16384
48402 <... writev resumed>) = 32797
48403 mmap(NULL, 2035712, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0 <unfinished ...>
48402 read(3, <unfinished ...>
48403 <... mmap resumed>) = 0x7f00c3afb000
48402 <... read resumed>"\0\0\0\30", 4) = 4
48402 read(3, "e\0\0\0\f\0\0\0\0\0\0\0\7Success\0\0\0\0", 24) = 24
48402 read(4, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 32768) = 32768
48402 writev(3, [{iov_base="\0\0\200\31", iov_len=4}, {iov_base="\6\0\0\0M\0\0\0\4\0\0\0\0\0\0\0\0\0#\0\0\0\0\200\0\0\0\0\0\0\0\0"..., iov_len=32793}], 2) = 32797
48402 read(3, <unfinished ...>
48403 munmap(0x7f00c3731000, 2019328) = 0
48403 select(7, [3 4], [3], NULL, {tv_sec=59, tv_usec=0}) = 1 (in [4], left {tv_sec=58, tv_usec=999997})
48403 read(4, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 16384) = 16384
48403 mmap(NULL, 2052096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7f00c3906000
48403 munmap(0x7f00c3afb000, 2035712) = 0
48403 select(7, [3 4], [3], NULL, {tv_sec=59, tv_usec=0}) = 1 (in [4], left {tv_sec=58, tv_usec=999997})
48403 read(4, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 16384) = 16384
48403 mmap(NULL, 2068480, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7f00c370d000
48403 munmap(0x7f00c3906000, 2052096) = 0
48403 select(7, [3 4], [3], NULL, {tv_sec=59, tv_usec=0}) = 1 (in [4], left {tv_sec=58, tv_usec=999997})
48403 read(4, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 16384) = 16384
48403 mmap(NULL, 2084864, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7f00c3aef000
48403 munmap(0x7f00c370d000, 2068480) = 0
48403 select(7, [3 4], [3], NULL, {tv_sec=59, tv_usec=0}) = 1 (in [4], left {tv_sec=58, tv_usec=999997})
48403 read(4, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 16384) = 16384
48403 mmap(NULL, 2101248, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7f00c38ee000
48403 munmap(0x7f00c3aef000, 2084864) = 0
48403 select(7, [3], [3], NULL, {tv_sec=59, tv_usec=0}) = 2 (in [3], out [3], left {tv_sec=58, tv_usec=981048})
48403 read(3, 0x7ffc9bb9c500, 8192) = -1 ECONNRESET (Connection reset by peer)
48403 rt_sigaction(SIGWINCH, {sa_handler=SIG_DFL, sa_mask=~[RTMIN RT_1], sa_flags=SA_RESTORER|SA_RESTART, sa_restorer=0x7f00c3f83ab0}, {sa_handler=0x55ff7bcc73f0, sa_mask=~[KILL STOP RTMIN RT_1], sa_flags=SA_RESTORER|SA_RESTART, sa_restorer=0x7f00c3f83ab0}, 8) = 0
48403 getpid() = 48403
48403 write(3, "\364\3301X\302Y\34\361\310:\323\275\315!Hy\334;\312\201\231\273\217\17h\363)&\207bD7"..., 154924) = -1 EPIPE (Broken pipe)
48403 --- SIGPIPE {si_signo=SIGPIPE, si_code=SI_USER, si_pid=48403, si_uid=1000} ---
48403 write(2, "client_loop: send disconnect: Br"..., 43) = 43
48403 exit_group(255) = ?
48402 <... read resumed>0x5611a3af2810, 4) = -1 ECONNRESET (Connection reset by peer)
48403 +++ exited with 255 +++
48402 --- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=48403, si_uid=1000, si_status=255, si_utime=15, si_stime=24} ---
48402 wait4(48403, NULL, WNOHANG, NULL) = 48403
48402 write(2, "\rConnection closed. \n", 22) = 22
48402 rt_sigreturn({mask=[]}) = -1 ECONNRESET (Connection reset by peer)
48402 write(2, "Connection closed\r\n", 19) = 19
48402 exit_group(255) = ?
48402 +++ exited with 255 +++
答案1
好的。我已經追蹤到 meraki 威脅防護、入侵防禦。
- https://community.cisco.com/t5/network-security/can-sftp-packets-trigger-ssh-event-respoverflow-events-in-ips/td-p/2933423
- https://bst.cloudapps.cisco.com/bugsearch/bug/CSCze89434/?rfs=iqvred
- https://community.meraki.com/t5/Security-SD-WAN/SFTP-session-disconnecting/td-p/49244
如果 IPS 規則 SSH_EVENT_RESPOVERFLOW 被列入白名單，則錯誤對我來說就消失了。
儘管如此，我認為將 IPS 規則白化並不是正確的解決方案。 IPS 規則應該以更好的方式偵測入侵，以防止誤報。