Two new releases are available: 2.02.182 (from the stable branch) https://sourceware.org/git/?p=lvm2.git;a=tag;h=d64377b5bbd2d170f1d6365157e2245d90a63c1a 2.03.01 (from the development branch) https://sourceware.org/git/?p=lvm2.git;a=tag;h=1326762567c49c7305d83ac59dff819d7de964fd These releases fix a bug that can potentially cause data corruption in a rare situation described below. The bug first appeared in version 2.02.178 (June 2018). The subject of the fix is "metadata: prevent writing beyond metadata area". Bug details: Most VGs will not be affected by this bug. It can only happen when the first physical extent (PE) in the VG is unaligned with 128KB. By default the first PE is at 1MB, so any VG using the default will be immune. To check the first PE offset, run: $ vgs -o+pe_start --units k <vgname> If the 1st PE field is 1024 KB, or another multiple of 128KB, then the VG will be unaffected. In cases where a command line or config option was used to create a VG with an unaligned first PE, then the bug can potentially arise when: 1. An lvm command is being run and is updating the VG metadata, e.g. lvcreate, lvextend, vgextend. 2. The VG metadata is being updated at the tail end of the circular metadata buffer on disk. 3. A program (or file system) is writing changes to the 128KB of data that follows the lvm metadata, at the same time that lvm is writing its metadata. Given this specific combination of updates, lvm could "undo" the change made by the program/fs (lvm reads X, program writes X+1, lvm writes X.) Apart from updating to a fixed version of lvm, the best way to avoid any possible problem (in an unaligned VG) is to avoid running lvm commands that change metadata while LVs in the VG are being used. _______________________________________________ linux-lvm mailing list linux-lvm@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/linux-lvm read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/
