3.10
3.9
Btrfs:
native RAID 5 and 6 support.
LVM:
dm-cache(SSD caching) (DOC: cache-policies.txt cache.txt) <-- user mode: flashcache and bcache
Net:
SO_REUSEPORT
(multiple web server processes or threads to open individual sockets to listen on port 80)
3.8
加入新的 FS: F2FS (Flash-Friendly File System)
Btrfs
新加 "replace" feature
Ext4
to store files that only consist of a few bytes together with the inode
SEEK_DATA and SEEK_HOLE (backup or copy tools to detect, and omit, empty areas in sparse files)
Networking:
The Berkeley packet filter (BPF), used by sniffer tools such as tcpdump, can now be used to filter VLAN tags.
Virtualisation
balloon driver for Hyper-V
kmem extension
3.7
btrfs
- up to 65,536 hardlinks to one file
- Btrfs can now deallocate memory areas within files
- improves fsync performance
-
In ATA devices, the cache_type sysfs device file can now be used to switch between
the write-through and write-back cache behaviours. - The Libata subsystem supports "Aggressive SATA device sleep"
- The block layer offers the "WRITE SAME" command
3.6
- kvm: VFIO( Virtual Function I/O ) (passing through of PCI and PCIe devices)
- kvm: Qemu: storage migration (allows blocks to be copied from one running QEMU image to another)
- kvm: Qemu: block commits (snapshot deletion)
- Btrfs: now supports quotas for subvolumes
- Btrfs: userspace programs to determine the difference between two snapshots
- ext4: no longer stores quota information in visible files
- symlink and hardlink restrictions to the Linux VFS
- save swap files to NFS shares
-
virtio-scsi: now supports hotplugging
3.5
- ext4 <--- checksums to metadata (mkfs -O metadata_csum) 可惜沒有修復能力 ...
- Uprobes - user-space probes(performance probes) (no need to restart the program or modify the binaries. )
-
TCP connection repair (Putting the socket in a "repair" mode)
(relocate a entire network connection from one host to another transparently for the remote end) - TCP Early Retransmit(/proc/sys/net/ipv4/tcp_early_retrans)
- btrfs <--- CRC & I/O failure statistics (btrfs device stats)
3.4
- Btrfs <--- btrfs-restore, file system within a VM
- mdadm <--- hot-replace
- read-only support for the QNX4
- kvm: virtio-scsi
3.3
- Open vSwitch
- RDMA (SCSI Remote Direct Memory Access) Target driver
- Adaptive RED (Random Early Detection)
- teaming device (802.1AX)
- cgroups <--- TCP buffer limits, Network priority control group
- ext4 <--- online resize
- EFI boot support (Both BIOS and EFI boot loaders can still load and run the same bzImage)
- PMU v2 (Performance Monitoring Unit) for KVM
- MD: hot replace
3.2
- ext4 <--- block sizes bigger than 4KB and up to 1MB
-
btrfs <--- Manual inspection of the filesystem (btrfs inspect <inode/logical> <n> /mnt)
<--- mount option: "t -o subvol=path/to/subvol/you/want"
<--- mount option: "-o recovery" - DM layer <---thin provisioning
- CPU bandwidth control
- I/O-less dirty throttling
- TCP Proportional Rate Reduction
- perf top <--- explore the assembly code arbitrarily deep
- syscalls - Cross memory attach <---allow to r/w f/t another processes address space)
- B.A.T.M.A.N. ad hoc networking
3.1
- Dynamic writeback throttling
- ext3 mout opt: barrier=0 <--- disks provide a "cache flush" instruction that the OS uses when it needs (barrier)
- cpupowerutils
- MD layer(software raid) Bad block management
- ipv4 microoptimization: save cpu cycles from check_leaf()
- Strict CPU affinity, by writing the value 2 to /sys/block/<bdev>/queue/rq_affinity
- KVM: Nested VMX (Intel virtualization) support
3.0
- btrfs <--- Automatic defragmentation (-o autodefrag)
- btrfs <---- Scrub
- btrfs <---- delayed b+ tree insertions or deletions
- Wake on WLAN (ACPI S3 suspend)
- ext4 <--- Punching holes (portion of the file can be marked as unwanted)
- Net <--- pkt_sched: QFQ - quick fair queue scheduler
- Cache xattr security drop check for write
-
cgroup - Memory <--- Reclaim memory from nodes in round-robin order
<--- pagefault count - syscall <--- sendmmsg(): batching of sendmsg() calls
詳見:
http://kernelnewbies.org/Linux_3.2