DragonFly BSD 6.4 released: hardware support for type-2 hypervisors

DragonFly is an operating system and application environment designed as a continuation of the FreeBSD-4.x family. These operating systems and Linux can be classified as the same category, because they are based on UNIX concepts and application programming interface. DragonFly is a branch of this road of development, it can be said, is a BSD to a different from the FreeBSD-5 series of new directions and the development of the opportunity.

DragonFly 6.4 releases.

Changelog:

Security Issues

• Recommend updating all systems, a locally-exploitable kernel vulnerability was fixed for this release.

Kernel

• Fix execvPe()s environment passing.
• Fix ktrace’s handling of ‘long’ system call return values.
• Fix a sysv semaphore panic related to an exit race.
• Fix a namecache bloating issue related to dead entries that could slow systems down over time.
• Fix several namecache eviction issues which were interfering with nlookup*() operation.
• Fix a namecache issues that evicted too many resolved entries, causing unnecessary re-lookups.
• Critical section count mismatch panics are more verbose.
• Fix multiple issues with vnode recyclement which could result in a long-running system slowing down. Generally speaking, there were leaks in vnode state counts which could prevent vnlru from either running or from finding recycleable vnodes.
• Implement mlockall()’s MCL_CURRENT feature and generally match with linux-like expectations.
• Fix a lock order reversal and deadlock in cache_resolve_mp() which can get hit when concurrent path lookups cross mount boundaries.
• Add the gtaskqueue API from FreeBSD.
• nlookup*() now retries lookups on elements which only undergo modification races, and does not retry lookups on elements which are merely locked for access and then unlocked. Improves performance.
• KERN_PROC – Change behavior for process list retrievals to not include pure kernel threads by default.
• Do not panic if an IP packet with a non-atomic header in the first mbuf makes it to ip*_input().

Filesystems

• hammer2 – Fix run-away kernel memory allocations during bulkfree when operating on deep trees.
• hammer2 – Fix chain allocation tracking counter calculation.
• hammer2 – Fix panic related to USB stick pull of mounted H2 filesystem.
• hammer2 – Report critical bulkfree transitions that are not supposed to happen.
• hammer2 – Fix CHECK FAIL paths that might mangle an inode in-memory.
• hammer2 – Validate that inode number on-media against the inode number being looked up.
• hammer2 – FIX CRITICAL BULKFREE BUG when multiple PFSs are mounted that could corrupt a H2 filesystem.
• hammer2 – FIX ISSUE WHERE DELETED FILES SOMETIMES LINGER UNTIL UNMOUNT, which can leak dead inodes (leave them on the media), slowly filling up the filesystem under certain circumstances.
• hammer2 – Lots of syntax cleanups.
• hammer2 – Properly set the read-only flag for read-only mounts of PFSs.
• hammer2 – Fail the mount if the root volume is not specified.
• hammer2 – Fix a panic related to a PFS not being found on-mount.
• tmpfs – Fix heavily threaded readdir() races against deletions.
• tmpfs – Improve the performance if cookie seeks during directory scans. O(n) to n log n.
• msdosfs – Fix a case where a non-dot lookup returns the current directory during a path lookup.
• msdosfs – Generate an error instead of a panic when inusemap is inconsistent.
• msdosfs – Sanity-check sector count from BPB.
• msdosfs – Fix mounting when the device sector size is larger than 512 bytes.
• msdosfs – Add /dev prefix if necessary when looking up mount device.
• ext2fs – Do a more accurate check for the root inode and for directory entries.

Networking

• ipfw – Fix broken mixed network and host IP specifications in ip tables. The network prefix was improperly inherited from prior entries.
• pf – Fix ‘set skip on’ for interface groups.
• pf – Make “:0” (noalias) also ignore link-local IPV6 addresses.
• IPV6_V6ONLY now silently accepts any value for this socketopt, fixing certain bits of third-party code.
• if_bridge – Fix a broadcast issues that could sometimes cause ARPs to be lost.
• if_bridge – Distinguish between interfaces that are part of the same bonded set to allow explicit priorities to be set for backup interfaces. Round-robin operation may not be desired on all interfaces in the bonding set. See the manual page.
• urtwn – Add support for the Edimax EW-7811Un V2 (N150)
• jail – sysctl net_raw_sockets renamed to allow_raw_sockets.

Graphics

• drm – Retry page fault handler on buffer data in-transit, fixing an Xorg crash on certain monitor connect/disconnect events.
• evdev – Make adjustments for better compatibility with the latest X server code.

More…

Download