Description of problem:

Do kvm block regresson test w/ (1Gb link) nfs backend between rhel6.2GA and rhel6.3 beta1 hosts .there is 35% performance regression in large file sequential writes.

Detailed result:
================================================================
raw format in LargeFile Creates(256KB)
    
  w/kernel-257 guest       threads|       IOPS|   Thro(MBps)|  
rhel6.2GA                        1|     195.35|         48.8|  
r63 kernel-267 qemu-272          1|     126.29|         31.6|   
diff %                            |      -35.4|        -35.2|   

qcow2 format in LargeFile Creates(256KB)
  w/kernel-257 guest       threads|       IOPS|   Thro(MBps)|  
rhel6.2GA                        1|     199.77|         49.9|
r63 kernel-267 qemu-272          1|     130.61|         32.7|
diff %                            |      -34.6|        -34.5| 

================================================================

Version-Release number of selected component (if applicable):

(host)kernel-2.6_32-262
qemu-kvm-0.12.1.2-257
(guest)kernel-2.6_32-257

How reproducible:

always 

Steps to Reproduce:
1.mount 1gb netapp server in client. 
cat /proc/mounts

192.168.0.113:/vol/s2wquan116171nfs /home/kvm_autotest_root/images nfs rw,sync,relatime,vers=3,rsize=65536,wsize=65536,namlen=255,acregmin=0,acregmax=0,acdirmin=0,acdirmax=0,hard,noac,proto=tcp,timeo=600,retrans=2,sec=sys,mountaddr=192.168.0.113,mountvers=3,mountport=4046,mountproto=tcp,local_lock=none,addr=192.168.0.113 0 0

3. create 20G raw/qcow2 images.
qemu-img create -f raw /home/kvm_autotest_root/images/storage2.raw 20G
qemu-img create -f qcow2 /home/kvm_autotest_root/images/storage2.qcow2 110G -o preallocation=metadata

4.Pass this raw/qcow2 file image to the KVM guest as a block device and use
cache=none aio=threads for the KVM guest.
/usr/libexec/qemu-kvm -name vm1 -nodefaults -vga std -chardev socket,id=qmp_monitor_id_qmpmonitor1,path=/tmp/monitor-qmpmonitor1-20120418-021037-GUK9,server,nowait -mon chardev=qmp_monitor_id_qmpmonitor1,mode=control -chardev socket,id=serial_id_20120418-021037-GUK9,path=/tmp/serial-20120418-021037-GUK9,server,nowait -device isa-serial,chardev=serial_id_20120418-021037-GUK9 -device ich9-usb-uhci1,id=usb1,bus=pci.0,addr=0x4 -drive file=/usr/local/autotest/tests/kvm/images/RHEL-Server-6.3-64-virtio.raw,index=0,if=none,id=drive-virtio-disk1,media=disk,cache=none,boot=off,snapshot=off,readonly=off,format=raw,aio=threads -device virtio-blk-pci,bus=pci.0,addr=0x5,drive=drive-virtio-disk1,id=virtio-disk1 -drive file=/usr/local/autotest/tests/kvm/images/test.raw,index=2,if=none,id=drive-virtio-disk2,media=disk,cache=none,boot=off,snapshot=off,readonly=off,format=raw,aio=threads -device virtio-blk-pci,bus=pci.0,addr=0x6,drive=drive-virtio-disk2,id=virtio-disk2 -device virtio-net-pci,netdev=idExu1N2,mac=9a:ff:15:58:09:ce,id=ndev00idExu1N2,bus=pci.0,addr=0x3 -netdev tap,id=idExu1N2,vhost=on  -m 4096 -smp 2,cores=1,threads=1,sockets=2 -cpu Westmere -device usb-tablet,id=usb-tablet1,bus=usb1.0 -vnc :1 -vga qxl -rtc base=utc,clock=host,driftfix=slew -M rhel6.3.0 -boot order=cdn,once=c,menu=off -no-kvm-pit-reinjection -enable-kvm

4)In the KVM guest (running RHEL6.3-257), write filed in second disk (dev/vdb). ,then format the virtual disk with ext4 and mount the virtual disk in the guest.
  mkfs.ext4 /dev/vdb; mount /dev/vdb /mnt

5)run ffsb test.define direct io is on in configuration of ffsb test
# ffsb large_file_creates_256k_1.ffsb

Actual results:


Expected results:


Additional info:
this bug may caused by linearized QEMU hack removal patch in bug #767606.