On 20050115, Roland commited a number of fixes to 2.6-bk in order to correct races. We've not investigated the impact of these issues but they look from the descriptions like they could lead to and unprivileged user deadlocking a system (DoS, crash). Please let us know if this is incorrect. 1. fix coredump_wait deadlock with ptracer & tracee on shared mm fixed=2.6-bk (20050115 cset@41e9a97cuQ7FWekabtf12Orvpfbp1w) 2. fix race between core dumping and exec with shared mm fixed=2.6-bk (20050115 cset@41e9a98dJKJjUrTCrKsyak1DOQxQug) 3. fix exec deadlock when ptrace used inside the thread group fixed=2.6-bk (20050115 cset@41e9a99fugWWHLAKeLYyNRl6h9RHlg) 4. clear false pending signal indication in core dump fixed=2.6-bk (20050115 cset@41e9a9bcwYGXPK_HRWKybNR060H80g) (Prefix cset@ with http://linux.bkbits.net:8080/linux-2.6/ for a direct link to each fix)
Agreed, this can be deferred.
Patches confirmed in 2.6.9-6.37.EL. Moving to PROD_READY.
An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on the solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHSA-2005-420.html
Created attachment 120583 [details] test case for case #3 (exec deadlock) For regression testing, this is my test case for the #3 scenario, an exec that leaves an unkillable stopped process.