[9fans] Detecting EOF vs. Interrupt across 9P

[9fans] Detecting EOF vs. Interrupt across 9P

[Lyndon Nerenberg (VE6BBM/VE7TFX)]

Given a foofs which serves the writable file /mnt/foo, is there
any reliable way to distinguish between

	% cat > /mnt/foo
	type some
	text and quit
	^D
	%

and

	% cat > /mnt/foo
	type some
	text, then change your mind and hit <DEL>
	%

at the server end?  I know I read something about this, somewhere, but
I can't find it now.  It could very well have been buried in some
source I was reading (about Tflush vs. 0 length Twrites or some
such?).

--lyndon

Re: [9fans] Detecting EOF vs. Interrupt across 9P

[Charles Forsyth]

in the first case, you'll see the last Twrite and then a Tclunk.
in the second, you'll see the first Twrite and then a Tflush (followed by
a Tclunk when cat exits).

Re: [9fans] Detecting EOF vs. Interrupt across 9P

[Russ Cox]

On Fri, Jan 15, 2010 at 2:45 PM, Charles Forsyth <forsyth@terzarima.net> wrote:
> in the first case, you'll see the last Twrite and then a Tclunk.
> in the second, you'll see the first Twrite and then a Tflush (followed by
> a Tclunk when cat exits).

i don't believe these two cases can be distinguished.
in particular i think you'd only see the Tflush if the first
Twrite was still in flight when you typed DEL.  assuming
the first write had completed before DEL, the two scenarios
are indistinguishable other than the different values being
written.

> at the server end?  I know I read something about this, somewhere, but
> I can't find it now.  It could very well have been buried in some
> source I was reading (about Tflush vs. 0 length Twrites or some
> such?).

wikifs uses a trick to convey errors back at the end of
the file: clients can send a zero-length write before closing
the file, and the zero-length write will fail if the page
cannot be committed.  but i don't recommend this in
general.  it's pretty gross and means, among other things,
that cat is not as useful as a custom client.

russ

Re: [9fans] Detecting EOF vs. Interrupt across 9P

[Lyndon Nerenberg (VE6BBM/VE7TFX)]

Russ says:
> i don't believe these two cases can be distinguished.
> in particular i think you'd only see the Tflush if the first
> Twrite was still in flight when you typed DEL.  assuming
> the first write had completed before DEL, the two scenarios
> are indistinguishable other than the different values being
> written.

And after more digging through the docs and kernel this seems to
be the case. (And it describes the behaviour I'm seeing.)

--lyndon

Re: [9fans] Detecting EOF vs. Interrupt across 9P

[Charles Forsyth]

>i don't believe these two cases can be distinguished.

you're right, and indeed don't believe it either.

Re: [9fans] Detecting EOF vs. Interrupt across 9P

[Charles Forsyth]

>you're right, and indeed don't believe it either.

i meant: and indeed I don't believe it either.