Messages in this thread | | | Date | Mon, 18 Mar 1996 11:09:41 +0100 (MET) | From | Ingo Molnar <> | Subject | 1.3.75 BUG REPORT: new method to reproduce double lock on socket |
| |
how to reproduce it: (new method):
1) "rlogin localhost" on the same VC, till it starts swapping bad. The echo of your input has to be "slow", even when there is no swapping after some time. (i have reached this point after 20 rlogins)
2) switch to another VC, and start many progs, this swaps out all the rlogins
3) now switch back to the VC with the rlogins, and push Enter, and dont release the button: after some swapping activity the messages show up
Note: the messages dont show up on any keypress, you have to "saturate" the system somehow. Maybe it's related to rlogin sending "out of bound data"? The messages show up too when you hit Ctrl-C after you have hit alot of keys, in this case "rlogin" sends the "Ctrl-C" out of bound, and here i can picture quite a few race conditions along those 20 rlogin processes piping alot of keypresses and an out of bound Ctrl-C. Just a wild shot into the darkness.
here is where the messages are printed:
001444c0 t wait_for_tcp_memory 00144570 t do_tcp_sendmsg 00145030 t tcp_sendmsg <------------- here 00145150 T tcp_read_wakeup 00145410 t tcp_recv_urg
-- mingo
| |