Bug 6766 - read(0,..) after fork() can cause the system to hang
Summary: read(0,..) after fork() can cause the system to hang
Status: REJECTED UNREPRODUCIBLE
Alias: None
Product: IO/Storage
Classification: Unclassified
Component: Other (show other bugs)
Hardware: i386 Linux
: P2 normal
Assignee: io_other
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2006-06-29 07:20 UTC by Yair
Modified: 2007-02-01 09:04 UTC (History)
1 user (show)

See Also:
Kernel Version: 2.6.17
Subsystem:
Regression: ---
Bisected commit-id:

Attachments
causes system hang (472 bytes, text/plain)
2006-06-29 07:29 UTC, Yair 		Details
config file (43.80 KB, text/plain)
2006-07-02 05:06 UTC, Yair 		Details
Add an attachment (proposed patch, testcase, etc.)

Description Yair 2006-06-29 07:20:58 UTC 
Most recent kernel where this bug did not occur: 2.6.14
Distribution: debian testing (vanilla kernel)
Hardware Environment: N/A
Software Environment:
Problem Description:
After forking, if the child process calls read(0,...), and then the parent
exits, the child might not respond to KILL signal and stay opened consuming more
and more resourses, eventually freezing the system which forces a reboot.

This happens in about 1 out of 7 runs (in average).
Tested on several hardwares.

Steps to reproduce:
Attached c code.
Comment 1 Yair 2006-06-29 07:29:58 UTC 
Created attachment 8453 [details]
causes system hang
Comment 2 Andrew Morton 2006-06-29 14:09:54 UTC 
I'm unable to reproduce this.

Some system info and a copy of your .config might help.
Comment 3 Yair 2006-07-02 05:06:58 UTC 
Created attachment 8474 [details]
config file
Comment 4 Andrew Morton 2007-01-31 01:10:52 UTC 
Is this problem still present in 2.6.20-rc7?

It's rather weird.
Comment 5 Yair 2007-02-01 07:03:13 UTC 
As I recall, after updating some nfs utilities, the problem was gone.
Currently (2.6.19), I didn't succeed in reproducing it.
Comment 6 Andrew Morton 2007-02-01 09:04:42 UTC 
OK, thanks - I'll close the bug off, mark it unreproducible.
Note You need to log in before you can comment on or make changes to this bug.