When using an ssh://host/path repo location an ssh connection is created to the host and 'pkg ssh' is executed. If the sshd process is killed/timeouts/brokenpipe the pkg process on the host is not terminated and consumes 100% CPU. The below was tested by manually killing the sshd process however I've seen the same results when the connection was broken with a broken pipe message on the client. ################################################# Reproduce: Client: 1) Execute a pkg command that connects to the host and prompts the user for input (pkg upgrade) 2) On the host kill the sshd process related to the 'pkg ssh' process 3) See 'pkg ssh' didn't die and now consumes 100% CPU client$ pkg -v 1.2.7_3 client# pkg upgrade Updating repository catalogue .... The upgrade will require 4 MB more space 24 MB to be downloaded Proceed with upgrading packages [y/N]: ------------ Host: host$ pkg -v 1.2.7_3 host$ ps aux | grep pkgrepo ... pkgrepo 83886 0.0 0.0 65484 7336 - I 1:56PM 0:00.01 sshd: pkgrepo@notty (sshd) pkgrepo 83887 0.0 0.0 53032 7600 - Is 1:56PM 0:00.02 /usr/local/sbin/pkg ssh ... host# kill 83886 output from top: 83887 pkgrepo 1 94 0 53032K 7600K CPU0 0 0:14 75.98% /usr/local/sbin/pkg ssh ... 83887 pkgrepo 1 103 0 53032K 7600K CPU2 2 0:51 100.00% /usr/local/sbin/pkg ssh ... 83887 pkgrepo 1 103 0 53032K 7600K CPU2 2 1:03 100.00% /usr/local/sbin/pkg ssh ################################################# Expected: 'pkg ssh' process would terminate if the ssh connection is lost.
over to maintainer
Fixed