Description of problem: EL5 boot hangs at "Starting udev" In PC of 16CPUs, it hangs according to the probability of 3%. Version-Release number of selected component (if applicable): kernel-2.6.18-8.1.3.EL How reproducible: 100％ Steps to Reproduce: 1. boot EL5 2. 3. Actual results: EL5 boot hangs at "Stating Udev". Expected results: EL5 boots. Additional info: I found the method reproducing this issue. It certainly reappears by the attached test program. 1. boot EL5 2. ./start_udev.sh 3.
Created attachment 154967 [details] It reappears 100%.
Created attachment 154968 [details] It will reappear 100% by multi-CPU.
Created attachment 154969 [details] intel rng driver update patch
(In reply to comment #3) > Created an attachment (id=154969) [edit] > intel rng driver update patch This issue is reported below. http://lkml.org/lkml/2007/2/27/117 I created the patch for EL5. In the environment of x86, it works well. Please check a patch.
As a patch is available and the request came from engineering I am proposing this as an exception for 5.1 and pm_acking. The severity needs to be reviewed. I am setting prio to high for now (is it 3% or 100%?, workarounds?) as this did not come through support.
(In reply to comment #6) > The severity needs to be reviewed. I am setting prio to high for now (is it 3% > or 100%?, workarounds?) as this did not come through support. I think that there is no workarounds. The probability of a hang is about 3%. Reproducibility is 100%.
This issue is considered as a show-stopper by Hitachi Linux division; they are actually starting to use RHEL5 from 5.1 but without this being fixed they are considering to put back all of their Linux business residing RHEL5.1. I'll have score/exception raised. This event sent from IssueTracker by tumeya issue 127015
Has the patch been submitted upstream or is it already in some upstream kernel? Wow! Per Jay, this is a MASSIVE patch which among other things redefines a type (unsigned to int) in mod_init which could cause "indian" issues. It also exports a new variable. This change is all in mainline code. We need to have Hitachi/Toshiba submit a SMALL patch that fixes the specific problem in order for us to consider accepting the fix in a RHEL5 minor release such as 5.1 or 5.2. Setting to NEEDINFO