Am 22.05.2012 17:01, schrieb Eric Blake:
On 05/22/2012 08:45 AM, Kevin Wolf wrote:
>> I understand that open("/dev/fd/42") would be the same as dup(42), but
>> I'm not sure that I'm entirely clear on how this would work. Could you
>> give an example?
>
> With your approach you open the file outside qemu, pass the fd to qemu
> along with a file name that it's supposed to replace and then you use
> that fake file name:
>
> (qemu) getfd_file abc
> (qemu) drive_add 0 file=abc,...
>
> Instead you could use the existing getfd command and avoid the translation:
>
> (qemu) getfd
> 42
> (qemu) drive_add 0 file=/dev/fd/42,...
>
> Er, well. Just that getfd doesn't return the assigned fd today, so the
> management tool doesn't know it. We would have to add that.
That actually sounds workable. As long as management knows _what_ fd
qemu recieved (that is, 'getfd' is enhanced to tell libvirt the
associated fd number), and as long as qemu makes the magic naming of
/dev/fd/ work everywhere (even if it isn't normally part of the host
OS), then libvirt could indeed reuse existing file mechanisms to open a
file using an fd that it knows qemu should already own, without needing
to invent a new 'getfd_file' monitor command.
Which OSes are you thinking of? I'm not sure if we need to support FD
passing on all OSes in all places where you can pass a file name.
The specific use case we have in mind is for bypassing a SELinux
limitation, so that would be useful only for Linux anyway.
I guess in this instance,
libvirt would have to unconditionally use 'closefd' after the command
that reused the fd, since using file=/dev/fd/42 dups the fd rather than
consuming it (this is different from commands that use fd:nnn to consume
an fd).
I actually liked the fd: protocol approach, but Anthony doesn't seem to
want it any more. But yes, I think if we use /dev/fd/42 you need to
closefd unconditionally.
Kevin