Re: [fuse-devel] Problems with "rm -rf foo" on my filesystem

[Goswin v. B. <gos...@we...>]

Miklos Szeredi <mi...@sz...> writes:

> On Thu, 11 Dec 2008, Goswin von Brederlow wrote:
>> In my filesystem I collect a list of (name, type, inode) in opendir()
>> and slowly feed that out in readdir(). As such readdir() will always
>> reflect the state of the directory at the time of the opendir() call
>> and not neccessarily the current situation.
>> 
>> I'm wondering now if that is fundamentally flawed and violates normal
>> filesystem semantics. Has anyone else experienced such a problem?
>
> That's OK, but rewinding a directory (setting the offset to zero)
> should have the same effect as closing/opening it:
>
>   The rewinddir() function shall reset the position of the directory
>   stream to which dirp refers to the beginning of the directory. It
>   shall also cause the directory stream to refer to the current state
>   of the corresponding directory, as a call to opendir() would have
>   done. If dirp does not refer to a directory stream, the effect is
>   undefined.
>
> Miklos

And to the fuse filesystem that looks like what? The next readdir call
will have offset == 0? I don't se a rewinddir callback.

Currently I generate the list of entries in opendir and store it in
the filehandle. In readdir() I check for offset==NULL and then copy
the list from the filehandle to the offset. After that readdir just
adds new entries till it hits a marker (name == NULL) or the buffer is
full, each with offset pointing to the next list entry.

I guess then I have to add an indirection

struct dir_file_handle {
   struct file_list *first;
}

so I can modify fi->fh->first when readdir is called with offset == 0
or can I modify fi->fh during readdir calls?

MfG
        Goswin