[linux-cifs-client] Re: Initial try at a fix for truncate problem

[Steven French]

Yes this looks like a reasonable approach to handling this problem, and
easier to work than going to the NFS approach of keeping the search entries
around for longer in their own buffers distinct from the dentry cache.

I have been bogged down the last week since testing the cifs client at
Connectathon but will try to testing this later today.

>The bug happens on long directories that have many entries in the
>last FindNext slurp of entries.
>
>cifs_readdir is called on the directory
>
>FindNext( CLOSE_ON_END ) returns lots of directory entries
>at the end of a directory. Server closes directory handle.
>
>filldir( ) is only able to place some of them in the userland
> buffer the resume key is set, end of directory is not set
>
>cifs_readdir is called again on the directory
>
>FindNext( CLOSE_ON_END ) is called on the same directory --> Error
> from server (cuz directory handle is closed)
>
> cifs_readdir returns end of directory
>
> -Costa

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