NFSv3 vs. CIFS - Unicode filenames getting mangled

[bepresseditor]

On our Data ONTap 7.3.3 box we're having issues.

For example a file created via CIFS named

"Delta δ Δ - XP CIFS.txt"

becomes

":KV7R100" on the nfs mount.

NFSv3 originated file

"Delta δ Δ - Ubuntu.txt"

shows up on SMB with substitution character 0094:

"Delta δ Î ” ␚  - Ubuntu.txt"

The settings on the volume are as follows:

Name:		cabinet_main
Type:		Flexible		Root Volume?		-
FlexClone ?		-		Containing Aggregate:		aggr0
Status:		online,raid_dp
Used Capacity:		42.6 GB		Space Guarantee		volume
% Used:		35%		Language:		C
Total Capacity:		120 GB		Total Size:		150 GB
Number of Files:		98.3 k		Max Directory Size:		8.96 MB
Max Files:		4.67 m
SNAP Mirror?		-		SNAP Directory?
SNAP?				Resync SNAP Time:		60
SVO Enable?		-		SVO Checksum?		-
Allow SVO RMAN?		-		SVO Reject Errors?		-
Create Unicode?				Convert Unicode?
Minimal Read Ahead?		-		NV Fail?		-
Fractional Reserve :		100		Extent?		-
FS Size Fixed?		-		Update Access time?
I2P?				Ignore Inconsistent??		-

See also  http://communities.netapp.com/thread/10393

Can anyone shed insight into this issue?

[reena]

Do you have the volume options “convert_ucode” and “create_ucode” turned on ?

Reena

[bepresseditor]

Reena: please look above under the checkboxes for "Create Uncode?" and "Convert Unicode?"

[reena]

Try to turn these options ON for that specific volume, where the files have been created from the Linux client. CIFS would freak out, if these options are not turned on.

Reena

[bepresseditor]

Please look above.

The options are on.

Were on.

Always were on.

I  can't turn them on when they are already... on.

[reena]

I’m sorry, the screenshot actually doesn’t show it, that’s why I asked.

Reena

[bepresseditor]

Look below the screenshot at:

Create Unicode?

Convert Unicode?

[aborzenkov · Accepted solution]

You probably want to set language to C.UTF-8 instead of plain C. File names that have been created before will remain mangled though.

[bepresseditor]

That was it.  Apparently there are three unicode controls on a NetApp filer, only two of which can be changed from the web interface on an existing volume. And you have to reboot the filer after changing the language!

The C.utf8 language worked to keep both our CIFS and NFS clients happily in Unicode filenames.  Thanks.

Data ONTAP 7.3 Storage Management Guide (210-04122_A0)

Protocols in use	Volume language
NFS Classic (v2 or v3) only	Language setting does not matter
NFS Classic (v2 or v3) and CIFS	Language of the clients
NFS v4, with or without CIFS	cl_lang.UTF-8, where cl_lang is the language of the clients. Note: If you use NFS v4, all NFS Classic clients must be configured to present file names using UTF-8.