Re: [Courier-imap] courier-imapd and OCFS2

[Dan G. <da...@ra...>]

Hi,
Thanks for your timing looking over what I'd found. You do make a very
good point about the error code being return and the behaviour of
Thunderbird. I'll do a little more digging in the source for ocfs2 and
Thunderbird, see if I can spot the issues and raise issues in their
upstreams.

Regards
Dan

On 11/02/15 00:18, Sam Varshavchik wrote:
> Dan Goodliffe writes:
>
>> OK, I probably should have done this 6 months ago.. but it seems that
>> OCFS2 simply doesn't allow hard links across different directories.
>
> And that happens to be the same story with AFS, hence the existing code.
>
>> defiant ~ $ mkdir 1 2
>> defiant ~ $ touch 1/a
>> defiant ~ $ ls 1/
>> a
>> defiant ~ $ ln 1/a 2/a
>> ln: failed to create hard link '2/a' => '1/a': No such file or directory
>
> So, what appears to be the case is that OCFS2's kernel returns an
> ENOENT to the client, which is absurd. EXDEV is the right error code
> for this situation.
>
> Two observations:
>
> 1) Encountering this error should not result in any actual lost mail.
> The server should return an error the the IMAP client, if it's an IMAP
> client that tries to add a new message or move messages between
> folders. If an IMAP client assumes that the operation will always
> succeed, and ignores the error message, it's the client's fault. If
> the message is being delivered to a mailbox, via deliverquota or
> maildrop, this should also result in an error and a non-zero exit
> code, which the sending mail server should handle by keeping the mail
> in its mail queue.
>
> 2) There are reasons why link() is used, but that's besides the point.
>
>
>> >         if (errno != EXDEV)
>
>
> If you change this to
>
>  if (errno != EXDEV && errno != ENOENT)
>
> then this should also work.
>
>> > Any thoughts very welcome. I'm planning on raising this upstream, but
>
> Tell upstream to give you a logical explanation why ENOENT is a better
> error code than EXDEV, whose POSIX definition is:
>
> EXDEV           Improper link (POSIX.1)
>
> I'd like to see someone make an argument that, in this situation
>
> ENOENT          No such file or directory (POSIX.1)
>
> makes more sense. Someone explain to me which "file or directory" does
> not exist here. As your example showed, everything exists just fine.
>
> The reason why you can't link across multiple directories is probably
> because with OCFS2, like AFS, a sysadmin can migrate any directory
> hierarchy between different physical servers that make up the virtual
> filesystem hierarchy. So, if you link a file across multiple
> directories, you wouldn't be able to migrate any intermediate part of
> the virtual filesystem hierarchy.
>
> The issue here is not really that you can't do it, it's just that the
> error code returned to the application is confusing.
>
>
>
> ------------------------------------------------------------------------------
> Dive into the World of Parallel Programming. The Go Parallel Website,
> sponsored by Intel and developed in partnership with Slashdot Media, is your
> hub for all things parallel software development, from weekly thought
> leadership blogs to news, videos, case studies, tutorials and more. Take a
> look and join the conversation now. http://goparallel.sourceforge.net/
>
>
> _______________________________________________
> Courier-imap mailing list
> Cou...@li...
> Unsubscribe: https://lists.sourceforge.net/lists/listinfo/courier-imap