Bug 694800

Summary: Anaconda gets EDD information wrong
Product: Red Hat Enterprise Linux 6 Reporter: Martin Wilck <martin.wilck>
Component: anacondaAssignee: Ales Kozumplik <akozumpl>
Status: CLOSED DUPLICATE QA Contact: Release Test Team <release-test-team-automation>
Severity: medium Docs Contact:
Priority: unspecified    
Version: 6.1CC: akozumpl, ddumas, gasmith, gleb, jzeleny
Target Milestone: rcKeywords: Reopened
Target Release: ---   
Hardware: All   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2011-05-24 15:41:17 UTC Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Bug Depends On: 668737    
Bug Blocks:    
Attachments:
Description Flags
anaconda install.log
none
install.log.syslog
none
sosreport none

Description Martin Wilck 2011-04-08 13:16:25 UTC 
Description of problem:
I have a system with 2 megaraid_sas controllers which is con figured to boot from the 2nd drive on the 1st controller (sdb). The BIOS edd information is correct. Anaconda misinterprets the EDD information and gets the grub disk numbers wrong.

hd1: /dev/sdb

correct would be:
hd0: /dev/sdb
hd1: /dev/sda

Version-Release number of selected component (if applicable):
RHEL6.1 beta snap shot 2

How reproducible:
always

Steps to Reproduce:
1. install RHEL6.1
  
Actual results:
System fails to boot after installation because grub disk ordering is wrong

Expected results:
System boots

Additional info:
The info in sysfs clearly shows that sda=int13_dev81, sdb=int13_dev80:

[root@RX600S5-B578 ~]# cat /sys/block/sd[abc]/size^M
70582272^M
285671424^M
284557312^M
[root@RX600S5-B578 ~]# cat /sys/firmware/edd/s^H^[[K*/sectors^M
285671424^M
70582272^M
[root@RX600S5-B578 ~]# cat /sys/block/sda/size^M
70582272^M
[root@RX600S5-B578 ~]# cat /sys/block/sdb/size^M
285671424^M
[root@RX600S5-B578 ~]# cat /sys/firmware/edd/int13_dev80/sectors^M
285671424^M
[root@RX600S5-B578 ~]# cat /sys/firmware/edd/int13_dev81/sectors^M
70582272^M
Comment 1 Martin Wilck 2011-04-08 13:17:24 UTC 
Created attachment 490789 [details]
anaconda install.log
Comment 2 Martin Wilck 2011-04-08 13:18:05 UTC 
Created attachment 490790 [details]
install.log.syslog
Comment 3 Martin Wilck 2011-04-08 13:18:35 UTC 
Created attachment 490791 [details]
sosreport
Comment 6 Ales Kozumplik 2011-04-11 15:16:13 UTC 
Until this is fixed in kernel (bug 668737) everything we can do in Anaconda is hack around this by guessing (which is what we do already and it fails on certain occasions).
Comment 8 Martin Wilck 2011-04-12 07:10:59 UTC 
Could you open up bug 668737 for me, please?
Comment 9 Martin Wilck 2011-04-12 08:34:20 UTC 
(In reply to comment #6)
> Until this is fixed in kernel (bug 668737) everything we can do in Anaconda is
> hack around this by guessing (which is what we do already and it fails on
> certain occasions).

Your guessing algorithm doesn't appear to be perfect. For example, you don't seem to use the device size in your heuristic. As I tried to point out in the bug description, the size heuristic would have worked fine in my case (Obviously, it won't work all the time, but in many cases).
Comment 10 Ales Kozumplik 2011-04-12 10:42:22 UTC 
(In reply to comment #9)
> Your guessing algorithm doesn't appear to be perfect. For example, you don't
> seem to use the device size in your heuristic. As I tried to point out in the
> bug description, the size heuristic would have worked fine in my case
> (Obviously, it won't work all the time, but in many cases).

We are not attempting to make the guessing perfect, if will always be a hack. We want kernel to expose the information to us instead.
Comment 11 Ales Kozumplik 2011-05-24 15:05:04 UTC 
This bug has the same root cause as bug 621175, that is Anaconda is unable to correctly guess bios device order from the edd information exposed by kernel. Closing as a dup.

Note this issue is in fact waiting on bug 668737 to be resolved. Once kernel provides the complete information there will be changes in Anaconda necessary to accommodate the kernel changes. Reporter, if you can't see the kernel bug 668737 please let me know and I'll add you there.

*** This bug has been marked as a duplicate of bug 621175 ***
Comment 12 Gleb Natapov 2011-05-24 15:21:57 UTC 
(In reply to comment #10)
> (In reply to comment #9)
> > Your guessing algorithm doesn't appear to be perfect. For example, you don't
> > seem to use the device size in your heuristic. As I tried to point out in the
> > bug description, the size heuristic would have worked fine in my case
> > (Obviously, it won't work all the time, but in many cases).
> 
> We are not attempting to make the guessing perfect, if will always be a hack.
> We want kernel to expose the information to us instead.

Kernel does not have more information that anaconda. There are cases when there is not way to know boot device for certain. Kernel already expose all information anaconda needs to make educated guess. Kernel is not a place to do guesses.
Comment 13 Martin Wilck 2011-05-24 15:37:01 UTC 
I conclude from the last remark that anaconda indeed needs to improve the guessing algorithm. The example in the bug description is pretty obvious - if all BIOS devices have different sizes, the disk size is a very good guessing heuristic that anaconda should use. Together with the information EDD provides about the PCI device and the MBR signature, a good guess should be available in many more situations than now.
Comment 14 Ales Kozumplik 2011-05-24 15:41:17 UTC 
Please keep this closed. In any case this still is the same issue as bug 621175 which already has more information than this one.

*** This bug has been marked as a duplicate of bug 621175 ***
Comment 15 Gleb Natapov 2011-05-24 15:46:00 UTC 
(In reply to comment #14)
> Please keep this closed. In any case this still is the same issue as bug 621175
> which already has more information than this one.
> 
> *** This bug has been marked as a duplicate of bug 621175 ***

FWIW I agree with the closing of the bug as a dup of 621175. 621175 should be promoted to be "improve anaconda heuristics for choosing of boot device" or something.