On Thu, 14 Jun 2001, Derek Martin wrote:
>> Maybe I phrased the question poorly. I know I can't nfs mount a
>> remote /dev but if I created a /nfs/exportstuff directory and made
>> symlinks to the devices I wanted to share, and then machine B nfs
>> mounted /nfs/exportstuff on /dev/remdevs could machine B play a wav
>> on /dev/remdevs/audio ?
>
> I'd say not. A device file, even though it is treated much like a regular
> file in many respects, is very different. It's sort of like a symbolic
> link to a kernel driver.
In terms of how they are implemented in Linux: When you invoke a syscall
(like open(2), read(2), or write(2)) on a device node, the call is actually
mapped onto functions provided by the device driver for the major number of
the node. There is a structure ("struct file_operations") which contains a
number of function pointers, each corresponding to the usual file I/O
operations. The appropriate driver simply places pointers in a structure, and
registers that structure in the kernel.
Thus, when you call write(2) on a file descriptor opened on the "/dev/null"
device, you are actually calling write_null() in <drivers/char/mem.c>.
This is not supported over standard NFS. I do not think it could be, as
many device drivers access memory directly (buffers and such), expecting to
hand those buffers back to userspace, and "the PC down the hall" is a bit
outside your computer's address space. ;-)
> My best guess is that if you NFS mounted a device file, say, /dev/dsp,
> that if you wrote to that NFS-mounted device file, the sound would play on
> the local machine, just as if you'd sent it to the local filesystem's
> /dev/dsp. It would do that because the major and minor number of the
> device file is the same as the local one, and that (as I understand it) is
> really all that the kernel cares about.
I believe you are correct. A quick glance at the NFS code shows that the
standard kernel open function is used for NFS inodes, and the standard open
function should do what it always does. However, I haven't actually tried
this. And, of course, if the other system did not use the exact same
major/minor numbers, serious havoc would likely follow.
--
Ben Scott <[EMAIL PROTECTED]>
| The opinions expressed in this message are those of the author and do not |
| necessarily represent the views or policy of any other person, entity or |
| organization. All information is provided without warranty of any kind. |
**********************************************************
To unsubscribe from this list, send mail to
[EMAIL PROTECTED] with the following text in the
*body* (*not* the subject line) of the letter:
unsubscribe gnhlug
**********************************************************
