Releases: dm-vdo/kvdo
Releases · dm-vdo/kvdo
6.2.0.273
Version 6.2.0.273 Note: This is a pre-release version, future versions of VDO may not support VDO devices created with this version. - Fixed more error path memory leaks in the uds and kvdo modules. - Fixed module loading issues with the spec file on Fedora. - Removed the read cache. - Fixed error handling in preresume. - Converted table line parsing to use existing DM functions. - Fixed a bug which prevented parsing of version 0 table lines. - In order to properly handle version 0 table lines, made no-op physical growth not an error. - Limited the number of logical zones to 60. - Converted to use the kernel's bio zeroing method instead of a VDO specific one. - Added a missing call to flush_cache_page() after writing pages which may be owned by the page cache or a user as required by the kernel. - Added a version 2 table line which uses DM-style optional parameters. - Fixed a bug in the statistics tracking partial I/Os. - Added a maximum discard size table line parameter and removed the corresponding sysfs parameter which applied to all VDO devices.
6.2.0.239
Version 6.2.0.239 Note: This is a pre-release version, future versions of VDO may not support VDO devices created with this version. - Fixed error path memory leaks in the uds and kvdo modules. - Modified the physical and logical growth procedures to be consistent with other device mapper targets.
6.1.0.187
6.2.0.219
Version 6.2.0.219 Note: This is a pre-release version, future versions of VDO may not support VDO devices created with this version. - Fixed error path memory leaks in the uds and kvdo modules. - Fixed conflict errors when installing RPMs via dnf. - Fixed a hang when recovering a VDO volume with a physical size larger than 16TB. - Fixed some potential initialization issues in the UDS module. - Fixed a use-after-free bug in a UDS error path. - Added missing va_end() calls. - Modified Makefile to take build flags from rpmbuild.
6.1.1.125
6.1.0.185
6.2.0.197
Version 6.2.0.197 Note: This is a pre-release version, future versions of VDO may not support VDO devices created with this version. - Enabled the setting of max_discard_sectors for VDO devices via sysfs. This allows users stacking dm-thin devices on top of VDO to set a value which is large enough that dm-thin will send discards to VDO.
6.1.1.120
Version 6.1.1.120
- Improved memory allocation by not using the incorrect __GFP_NORETRY flag
and by using the memalloc_noio_save mechanism. - Fixed a potential deadlock in the UDS index by using the kernel supplied
struct callback instead of our own implementation of synchronous
callbacks. - Fixed a potential stack overflow when reaping the recovery journal.
6.2.0.187
Version 6.2.0.187 Note: This is a pre-release version, future versions of VDO may not support VDO devices created with this version. - Fixed a potential deadlock in the UDS index by using the kernel supplied struct callback instead of our own implementation of synchronous callbacks. - Eliminated obsolete code and fields from UDS. - Converted the VDO module to use numeric.h from the UDS module instead of having its own version. - Fixed a bug which would cause incorrect encoding of VDO data structures on disk. - Removed or modified logging which prints pointers from the kernel since newer kernels obfuscate the pointer values for security reasons. - Eliminated confusing and spurious error messages when rebuilding a UDS index. - Improved memory allocation by not using the incorrect __GFP_NORETRY flag and by using the memalloc_noio_save mechanism. - Finished conversion of the encoding and decoding of the VDO's on-disk structures to be platform independent. - Converted VDO to use the atomic API from the UDS module instead of its own. - Fixed memory leaks in UDS error paths. - Fixed a potential stack overflow when reaping the recovery journal.
6.1.0.181
Version 6.1.0.181 - Fixed a bug which would cause kernel panics when a VDO device is stacked on a RAID50 device. - Fixed a bug which could cause data loss when discarding unused portions of a VDO's logical space. - Modified grow physical to fail in the prepare step if the size isn't changing, avoiding a suspend-and-resume cycle.