Red Hat Bugzilla – Bug 77844
netstat uses signed int for socket inodes - should be unsigned
Last modified: 2007-04-18 12:48:25 EDT
From Bugzilla Helper: User-Agent: Mozilla/5.0 (compatible; Konqueror/3; Linux) Description of problem: netstat uses signed int for socket inodes - should be unsigned. When enough sockets have been used, the inode number exceeds 2^31 and netstat treats this as negative This causes a crash when processing /proc/net/unix and prevents sockets being matched to processes with the -p option Version-Release number of selected component (if applicable): How reproducible: Always Steps to Reproduce: 1. Use enough sockets (you don't have to keep them open) so that new sockets are allocated inodes > 2^31 (this may take some time - months in our case when running large nightly tests) 2. Run netstat -a -p Actual Results: processes cannot be matched to sockets (-p) segfault when processing UNIX sockets Expected Results: processes should be matched to sockets (-p) UNIX sockets are printed OK Additional info: I have produced a patch that makes netstat treat socket inodes as unsigned ints. Also sent to net-tools maintainer (philb@gnu.org and bernd.eckenfels@inka.de
Created attachment 85005 [details] Treat inodes as unsigned
Also seen on 7.1 According to http://www.ussg.iu.edu/hypermail/linux/kernel/0202.2/0834.html netstat > 1.60 is fixed.
Can anyone help me how to reproduce this bug?