Must really be safe_read(), not full_read()? (was: [PATCH] fix httpd lockup in cgi POSTs)

[Denis Vlasenko]

On Monday 12 February 2007 23:21, Matthias Reichl wrote:
> On Mon, Feb 12, 2007 at 10:01:13PM +0100, Denis Vlasenko wrote:
> > On Monday 12 February 2007 18:09, Matthias Reichl wrote:
> > > if (firstline)
> > >     count = full_read(inFd, rbuf, 4);
> > >     /* read 4 bytes so we can check if the line begins with "HTTP" */
> > 
> > So far I am happy with "bbox httpd doesn't support insane cgis
> > which split their HTTP response" way.
> 
> Would be OK for me, too. OTOH: it's a simple patch and it assures
> that the current httpd could also deal with insane cgis.

NO, IT CAN'T.

If cgi will output "HTT" and then block in read() from fd#0
while httpd is also blocked in full_read() trying to get at least
four bytes, we will deadlock.

> > > Better yet, do a full line read for the first line or completely
> > > switch to line buffered input, as you suggested.
> > 
> > Are you suggesting using stdio?
> 4> Can't do that, or POSTDATA will break again.
> > 
> > You basically need to _open-code_ buffering here. Than will work.
> 
> No, I was thinking about using a function like getLine() when
> reading the first line from the cgi. But that's not needed ATM
> and would only be useful for future versions of httpd if they'd like
> to parse the full first line.
--
vda