"config get repl-backlog-size huge-numbers" makes redis crash

[JeremyYu2018]

redis version 3.0.7 belows are key logs

33399:M 21 Apr 09:17:18.242 * Unable to partial resync with slave x.x.x.x:6379 for lack of backlog (Slave request was: 245114654245741).
33399:M 21 Apr 09:17:18.242 * Starting BGSAVE for SYNC with target: disk
33399:M 21 Apr 09:17:18.433 * Background saving started by pid 31246
33399:M 21 Apr 09:17:52.857 # Connection with slave x.x.x.x:6379 lost.
31246:C 21 Apr 09:18:10.796 * DB saved on disk
31246:C 21 Apr 09:18:11.058 * RDB: 1082 MB of memory used by copy-on-write
33399:M 21 Apr 09:18:11.585 * Background saving terminated with success
33399:M 21 Apr 09:24:20.958 # Out Of Memory allocating 545114654245741 bytes!
33399:M 21 Apr 09:24:20.958 # 

=== REDIS BUG REPORT START: Cut & paste starting from here ===
33399:M 21 Apr 09:24:20.958 # ------------------------------------------------
33399:M 21 Apr 09:24:20.958 # !!! Software Failure. Press left mouse button to continue
33399:M 21 Apr 09:24:20.958 # Guru Meditation: "Redis aborting for OUT OF MEMORY" #redis.c:3572
33399:M 21 Apr 09:24:20.958 # (forcing SIGSEGV in order to print the stack trace)
33399:M 21 Apr 09:24:20.958 # ------------------------------------------------
33399:M 21 Apr 09:24:20.958 #     Redis 3.0.7 crashed by signal: 11
33399:M 21 Apr 09:24:20.958 #     SIGSEGV caused by address: 0xffffffffffffffff
33399:M 21 Apr 09:24:20.958 #     Failed assertion: <no assertion failed> (<no file>:0)
33399:M 21 Apr 09:24:20.958 # --- STACK TRACE
/usr/local/bin/redis-server *:6379(logStackTrace+0x43)[0x450aa3]
/usr/local/bin/redis-server *:6379(_redisPanic+0x7f)[0x4505bf]
/lib64/libpthread.so.0[0x327dc0f7e0]
/usr/local/bin/redis-server *:6379(_redisPanic+0x7f)[0x4505bf]
/usr/local/bin/redis-server *:6379(redisOutOfMemoryHandler+0x2c)[0x4224ac]
/usr/local/bin/redis-server *:6379(zmalloc+0x71)[0x425a41]
/usr/local/bin/redis-server *:6379(resizeReplicationBacklog+0x41)[0x4356e1]
/usr/local/bin/redis-server *:6379(configSetCommand+0x12a9)[0x44a079]
/usr/local/bin/redis-server *:6379(call+0x72)[0x420292]

33399:M 21 Apr 09:24:21.039 # --- CURRENT CLIENT INFO
33399:M 21 Apr 09:24:21.039 # client: id=574559186 addr=127.0.0.1:57989 fd=859 name= age=105 idle=0 flags=N db=0 sub=0 psub=0 multi=-1 qbuf=0 qbuf-free=32768 obl=0 oll=0 omem=0 events=r cmd=config
33399:M 21 Apr 09:24:21.039 # argv[0]: 'config'
33399:M 21 Apr 09:24:21.039 # argv[1]: 'set'
33399:M 21 Apr 09:24:21.039 # argv[2]: 'repl-backlog-size'
33399:M 21 Apr 09:24:21.039 # argv[3]: '545114654245741'
33399:M 21 Apr 09:24:21.039 # --- REGISTERS
33399:M 21 Apr 09:24:21.039 # 
RAX:0000000000000000 RBX:00000000004bdab0
RCX:0000000000000000 RDX:0000000000014460
RDI:000000327db8fe80 RSI:0000000000000000
RBP:00000000004cc887 RSP:00007ffc94ff8010
R8 :000000327db8fe80 R9 :0000000000000000
R10:000000327db8fed0 R11:0000000000000206
R12:0000000000000df4 R13:0000000000004000
R14:000000000008f0f0 R15:000587003617cc08
RIP:00000000004505bf EFL:0000000000010202
CSGSFS:2d2d000000000033
33399:M 21 Apr 09:24:21.039 # (00007ffc94ff801f) -> 00007ffc94fff8bc
33399:M 21 Apr 09:24:21.039 # (00007ffc94ff801e) -> 00007ffc94ff80a0
33399:M 21 Apr 09:24:21.039 # (00007ffc94ff801d) -> 0000000000000000
33399:M 21 Apr 09:24:21.039 # (00007ffc94ff801c) -> 00007ffc94ff80c0
33399:M 21 Apr 09:24:21.040 # (00007ffc94ff801b) -> 000000000044a079
33399:M 21 Apr 09:24:21.040 # (00007ffc94ff801a) -> 000000000008f0f0
33399:M 21 Apr 09:24:21.040 # (00007ffc94ff8019) -> 00000000004356e1
33399:M 21 Apr 09:24:21.040 # (00007ffc94ff8018) -> 00007f3387f9f000
33399:M 21 Apr 09:24:21.040 # (00007ffc94ff8017) -> 00007f35a78b07c0
33399:M 21 Apr 09:24:21.040 # (00007ffc94ff8016) -> 00007f35aba40e68
33399:M 21 Apr 09:24:21.040 # (00007ffc94ff8015) -> 0000000000425a41
33399:M 21 Apr 09:24:21.040 # (00007ffc94ff8014) -> 000587003617cc08
33399:M 21 Apr 09:24:21.040 # (00007ffc94ff8013) -> 00000000004224ac

This could also happen to redis 4.0.8, and I think it's a dangerous synonym in some scenarios。 so can we just make it safer?

[AngusP]

If you ask Redis to allocate you a massive buffer, there's not a huge amount really wrong with it trying 😉 - however, failing in a slightly nicer manner would perhaps be good, rather than OOMing. Remember that CONFIG is a privileged command, there are many ways to cause breakage with it.