Bug 6273

Summary: rpm hangs forever consuming entire cpu
Product: [Retired] Red Hat Linux Reporter: lyndond
Component: rpmAssignee: Jeff Johnson <jbj>
Status: CLOSED WORKSFORME QA Contact:
Severity: high Docs Contact:
Priority: medium    
Version: 6.0   
Target Milestone: ---   
Target Release: ---   
Hardware: i386   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 1999-10-25 20:48:14 UTC Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:

Description lyndond 1999-10-22 21:47:51 UTC 
I am writing a program that queries packages on a remote
ftp server. I found that rpm command was sitting forever and
never returning and eating the entire cpu about 10% of the
time.
To get more debugging info I recompiled rpm with Debug=1
in lib/ftp.c

The command and the corresponding output is:

[rpmproj@sentinet rpm-3.0.3]$ ./rpm --rcfile ./rpmrc -vv -q
-p ftp://sunsite.org.uk/Mirrors/ftp.calderasys
tems.com/pub/OpenLinux/1.3/col/contrib/RPMS/howto-pl-sgml-19
980729-1.i386.rpm
D: logging into sunsite.org.uk as ftp, pw (username)
++ connect 193.63.255.4:21 on fd 3

and there it sits, forever eating all the cpu.
At this time netstat shows:

tcp        0      0 sentinet.co.uk:3763
swallow.doc.ic.ac.u:ftp CLOSE_WAIT

The output of tcpdump during this time is:

21:27:48.930079 sentinet.co.uk.3766 >
swallow.doc.ic.ac.uk.ftp: S 3007349526:3007349526(0) win
32120 <mss
1460,sackOK,timestamp 427876587 0,nop,wscale 0> (DF) (ttl
64, id 34262)
                         4500 003c 85d6 4000 4006 b19d c152
81b1
                         c13f ff04 0eb6 0015 b340 8316 0000
0000
                         a002 7d78 87b8 0000 0204 05b4 0402
080a
                         1980 e0eb 0000 0000 0103 0300
21:27:49.033345 swallow.doc.ic.ac.uk.ftp >
sentinet.co.uk.3766: S 130578483:130578483(0) ack 3007349527
wi
n 10136 <nop,nop,timestamp 804265412 427876587,nop,wscale
0,nop,nop,sackOK,mss 1460> (DF) (ttl 247, id 106
65)
                         4500 0040 29a9 4000 f706 56c6 c13f
ff04
                         c152 81b1 0015 0eb6 07c8 7833 b340
8317
                         b012 2798 fdd1 0000 0101 080a 2ff0
1dc4
                         1980 e0eb 0103 0300 0101 0402 0204
05b4
21:27:49.033380 sentinet.co.uk.3766 >
swallow.doc.ic.ac.uk.ftp: . ack 1 win 32120
<nop,nop,timestamp 42787
6597 804265412> (DF) (ttl 64, id 34265)
                         4500 0034 85d9 4000 4006 b1a2 c152
81b1
                         c13f ff04 0eb6 0015 b340 8317 07c8
7834
                         8010 7d78 e8b2 0000 0101 080a 1980
e0f5
                         2ff0 1dc4
21:27:49.107570 swallow.doc.ic.ac.uk.ftp >
sentinet.co.uk.3766: F 1:1(0) ack 1 win 10136
<nop,nop,timestam
p 804265420 427876597> (DF) (ttl 247, id 10666)
                         4500 0034 29aa 4000 f706 56d1 c13f
ff04
                         c152 81b1 0015 0eb6 07c8 7834 b340
8317
                         8011 2798 3e8a 0000 0101 080a 2ff0
1dcc
                         1980 e0f5
21:27:49.107592 sentinet.co.uk.3766 >
swallow.doc.ic.ac.uk.ftp: . ack 2 win 32120
<nop,nop,timestamp 42787
6604 804265420> (DF) (ttl 64, id 34267)
                         4500 0034 85db 4000 4006 b1a0 c152
81b1
                         c13f ff04 0eb6 0015 b340 8317 07c8
7835
                         8010 7d78 e8a2 0000 0101 080a 1980
e0fc
                         2ff0 1dcc


if I kill the rpm command the netstat CLOSE_WAIT disapears
immediately and we get the following packet exchange


21:28:09.879625 sentinet.co.uk.3766 >
swallow.doc.ic.ac.uk.ftp: F 3007349527:3007349527(0) ack
130578485 w
in 32120 <nop,nop,timestamp 427878682 804265420> (DF) (ttl
64, id 34309)
                         4500 0034 8605 4000 4006 b176 c152
81b1
                         c13f ff04 0eb6 0015 b340 8317 07c8
7835
                         8011 7d78 e083 0000 0101 080a 1980
e91a
                         2ff0 1dcc
21:28:09.911831 swallow.doc.ic.ac.uk.ftp >
sentinet.co.uk.3766: . ack 1 win 10136 <nop,nop,timestamp
80426
7500 427878682> (DF) (ttl 247, id 31555)
                         4500 0034 7b43 4000 f706 0538 c13f
ff04
                         c152 81b1 0015 0eb6 07c8 7835 b340
8318
                         8010 2798 2e44 0000 0101 080a 2ff0
25ec
                         1980 e91a
Comment 1 lyndond 1999-10-22 21:52:59 UTC 
This output was generated with rpm version 3.0.3. It is identical with
the version 3.0 supplied with RedHat 6.0 and version rpm version 3.0.2
Comment 2 lyndond 1999-10-22 21:55:59 UTC 
It is repeatable across multiple machines running RedHat 6.0
Comment 3 Jeff Johnson 1999-10-25 20:48:59 UTC 
Your command works for me running ~rpm-3.0.3 on Red Hat 6.1/sparc.

The tcpdump indicates that the ftp connection initiated on the local
machine (sentinet?) is being immediately closed by the remote machine
(swallow?). This is not an rpm problem, but rather an FTP
configuration problem.

Several other things:

1) --rcfile ./rpmrc
	rpm-3.0.x no longer reads /usr/lib/rpm/rpmrc by default. That
	means that either ./rpmrc must do
		include: /usr/lib/rpm/rpmrc
	or you should invoke as
		--rcfile /usr/lib/rpm/rpmrc:./rpmrc
	(Add /etc/rpmrc and/or ~/.rpmrc if those are required)

2) rpm uses passive FTP for all transfers.