Red Hat Bugzilla – Bug 110507
scp corrupts files in transit
Last modified: 2007-11-30 17:06:59 EST
From Bugzilla Helper:
User-Agent: Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1)
Description of problem:
occasional bytes get corrupted when using scp
Version-Release number of selected component (if applicable):
Copying from RedHat 9 (openssl-0.9.6-3) to Enterprise 3 (openssl-
Steps to Reproduce:
1. Run "md5sum" on a big (800 megs) file
2. use "scp" to copy it from RH9 th RHEL3 (other combinations may
3. Run "md5sum" on the copy: it's different.
(scping it back again and using "diff" reveals the occasional byte
I have my SSH running on port 52263, so the command I use is:-
scp -P 52266 bigfile.tgz email@example.com:
Actual Results: the new file is corrupted
Expected Results: no corruption
I've had the same results using several different machines.
Also - same result using different ethernet cards between same
machines (I have 2 subnets and 2 cards in my test PCs)
Searching google reveals other people experiencing similar problems.
I am unable to reproduce this, at least while using machines with the
default port. If I specify the default port, I can't reproduce.
Even if I change the port, and specify it, I can't reproduce it. Note
that I changed the port on the RHEL3 machine, and scp'd while on and
from the RHL9 machine.
Can you give more detail on how you make this happen? Eg, how often it
happens? If it happens on specific machines with specific set ups, or
on all machines of the appropriate releases that you've tried it on?
I have 4 physical machines.
A 1 in California (RHEL3)
3 on my office LAN (Australia)
B 1 RH9
C 1 RH9
D 1 Quad-boot: RHEL2.1, RHEL3.0, Windows XP, RHEL3.0
The problem occurs with SCP between
B and A (B(eth0) -> DSL504 -> internet -> (eth0)A)
B and D (B(eth0) -> (eth0)D) [both RHEL3's and the RHEL2.1 XP
B and D (B(eth1) -> (eth1)D) [both RHEL3's and the RHEL2.1 XP
I will attempt the "scp" commands again using "C" instead of "B" and
let you know the results.
The same file, copied using
A: Samba, and
B: HTTP, and
C: using "sz" from within an SSH terminal window (nonstandard SSH
does NOT get corrupted.
The corruption that appears when using "scp" seems to be random -
it's not the same bytes that get messed up every time - it's
False alarm, something else seems to be the cause:-
[cnd]$ cp D1 D2;cp D1 D3;cp D1 D4;cp D1 D5;cp D1 D6;cp D1 D7;cp D1
D8;cp D1 D9
[cnd]$ md5sum D*
Sorry guys - my original test SCP file was smaller, so got corrupted
less often than the above, and by fluke of unlucky coincidence, only
did it when I was using "scp" and not when I used anything else -
until I noticed it occur once without scp - hence my idea to try the
If this bug is still "open", someone else will need to close it - I
don't know how.