Bug 520814

Summary: GFS2: [RFE] Better verification of block read from journal during recovery
Product: [Fedora] Fedora Reporter: Steve Whitehouse <swhiteho>
Component: kernelAssignee: Steve Whitehouse <swhiteho>
Status: NEW --- QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: medium Docs Contact:
Priority: low    
Version: rawhideCC: adas, bmarzins, itamar, kernel-maint, rpeterso
Target Milestone: ---Keywords: FutureFeature
Target Release: ---   
Hardware: All   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: Enhancement
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 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: 437149    
Bug Blocks:    

Description Steve Whitehouse 2009-09-02 14:37:05 UTC 
It should be possible to do much better checking of the block being read from the journal during recovery. This would ensure that we do not write bad data over good data in the case that there is corruption of the journal blocks during recovery.

Lots of blocks have pointers in them (some inodes, indirect blocks, directory leaf blocks, etc) and we can verify that these point to valid blocks by checking the resource group data (i.e. ensure that the block number is higher than the super block, is less than the end of the filesystem and also does not refer to one of the rgrp blocks). We will need to ensure that the block to rgrp mapping is more efficient first though.

There are a number of other things which could be checked, particularly in the case of inode where we can check that the inode is of a valid type, and that the flags make sense for that type, etc.

Since recovery is an occasional activity it would be reasonable to do more tests that we can ordinarily do during normal fs operation (due to performance concerns).