[linux-cifs-client] 1.08

[Steven French]

I think that the performance issue that I noticed on one of my three test 
systems was a red herring, caused by excessive logging of an unrelated 
error (nls module not built and installed on that system).  Performance 
seems fine of 1.08.

There are various minor adhoc things I would like to test - e.g. a very 
large file create and file copy (I did not have enough disk space to do 
the suggested dd of a large file but will do that this weekend).

FYI - the ipv6 client cifs enablement is getting closer to reality - I 
need to implement parsing of ASCII ipv6 (colon ':' as separator and 
convert to binary and copy into sin_server6 struct) ala inet_pton 

Also can anyone think of a good argument for whether we should accept and 
use file_mode as a parm when the server support the Unix Extensions and 
the client has left enabled unix extensions in /proc/fs/cifs (thus the 
mode is considered valid by the cifs vfs and vfs_permission is not invoked 
on the client - just on the server).    Currently the cifs vfs uses the 
file_mode and dir_mode specified on mount for the default when mounting to 
windows servers, but not for mounts to Samba (with Unix extensions 
enabled) and vfs_permission (local perm check) is only done on those 
mounts to servers whose file mode's we don't trust (non-Unix extensions 
ones).   I thought that calling vfs_permission on the client and server 
seemed redundant to Samba since the admin could always just simply change 
the mode on the server or from the client but this may be a Unix 
convention (allowing mount time mode override even locally) that I don't 
have the history to know about


Steve French
Senior Software Engineer
Linux Technology Center - IBM Austin
phone: 512-838-2294
email: sfrench at-sign us dot ibm dot com
-------------- next part --------------
HTML attachment scrubbed and removed