Re: NFSv4 O_DIRECT with 3.16.2 kernel issue (qemu kvm cache=none / RHEL5 guest)

[Zdenek Kaspar <zkaspar82@gmail.com>]

On 09/12/2014 06:31 PM, Trond Myklebust wrote:
> On Fri, Sep 12, 2014 at 12:23 PM, Zdenek Kaspar <zkaspar82@gmail.com> wrote:
>> Hello everyone, got this specific problem with RHEL5 guest under KVM
>> hypervisor.
>>
>> client/server kernel version: 3.16.2
>>
>> /etc/exports: rw,async,no_subtree_check,no_root_squash
>> /proc/mounts: nfs4 rw,relatime,vers=4.0,rsize=524288,wsize=524288,
>>               namlen=255,hard,proto=tcp,port=0,
>>               timeo=600,retrans=2,sec=sys,clientaddr=xxx.xxx.xxx.xxx,
>>               local_lock=none,addr=xxx.xxx.xxx.yyy 0 0
>>
>> CONFIG_NFS_FS=m
>> # CONFIG_NFS_V2 is not set
>> CONFIG_NFS_V3=m
>> # CONFIG_NFS_V3_ACL is not set
>> CONFIG_NFS_V4=m
>> # CONFIG_NFS_SWAP is not set
>> # CONFIG_NFS_V4_1 is not set
>> # CONFIG_NFS_USE_LEGACY_DNS is not set
>> CONFIG_NFS_USE_KERNEL_DNS=y
>> CONFIG_NFSD=m
>> CONFIG_NFSD_V3=y
>> # CONFIG_NFSD_V3_ACL is not set
>> CONFIG_NFSD_V4=y
>> # CONFIG_NFSD_FAULT_INJECTION is not set
>> CONFIG_NFS_COMMON=y
>>
>> CentOS-5.10-i386-bin-1of8.iso -- did not test everything on centos,
>> behaved same as RHEL5. So let's use it as VM reproducer:
>>
>> dd if=/dev/zero of=kvm-test1.img bs=1M count=2048
>>
>> start VM using:
>> file=kvm-test1.img,cache=writeback,aio=native,if=virtio,boot=on
>>
>> create custom layout: /dev/vda1 / ext3 max size primary
>> unselect Desktop-Gnome, select Customize now
>> use only: Base System/Base and install it..
>>
>> [root@test1 ~]# find / >/dev/null
>> everything works as expected with writeback
>>
>> shutdown, backup kvm-test1.img (fs will explode)
>> start VM using:
>> file=kvm-test1.img,cache=none,aio=native,if=virtio,boot=on
>>
>> [root@test1 ~]# find / >/dev/null
>> and check dmesg, reboot etc, ie:
>> EXT3-fs error (device vda1): htree_dirblock_to_tree: bad entry in
>> directory #460317: rec_len is smaller than minimal - offset=0, inode=0,
>> rec_len=0, name_len=0
>> EXT3-fs error (device vda1): htree_dirblock_to_tree: bad entry in
>> directory #98503: rec_len is smaller than minimal - offset=0, inode=0,
>> rec_len=0, name_len=0
>>
>> Tested locally on ext4, xfs          - WORK
>> Tested on NFSv4 share with ext4, xfs - FAIL
>> Tested NFSv4 share sync, async       - FAIL
>> Tested on NFSv3 share with xfs       - WORK
>> KVM with qemu rhel-6.5, 2.1.1        - SAME RESULTS
>>
>> NOTE: it seems just the rhel5 guest triggers it easily..
>>
>> HTH, Z.
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 
> Don't use 3.16.x kernels with NFS yet. There are known bugs in the
> read/write code that need a stable backport of a set of fixes that are
> in 3.17-rc.
> 
> 

Hi Trond, noticed big nfs update that got into -stable, running 3.16.4
on both sides and problem is fixed, thx!

Z.