cygwin 3.3.x: another problem that may be related to pipes

Takashi Yano takashi.yano@nifty.ne.jp
Tue Nov 16 09:45:15 GMT 2021


On Mon, 15 Nov 2021 18:49:34 -0500
Ken Brown wrote:
> On 11/15/2021 6:35 PM, Takashi Yano wrote:
> > On Mon, 15 Nov 2021 13:47:44 -0500
> > Ken Brown wrote:
> >> On 11/15/2021 11:52 AM, Takashi Yano wrote:
> >>> So, currently we have three options.
> >>>
> >>> 1) Call CancelIo() immediately after STATUS_PENDING like my patch.
> >>> 2) Wait for pipe handle after STATUS_PENDING for nonblocking mode
> >>>      like Ken's patch.
> >>> 3) Create evt event even for nonblocking mode as Corinna mentioned.
> >>>
> >>> Which is the best solution, do you think?
> >>
> >> I'm completely unbiased, of course, but I like option 2.
> > 
> > OK. Shall I make a patch? Or would you like to do that?
> 
> If you don't mind doing it, I'd appreciate it.  I have a lot of Real Life things 
> going on at the moment and might not get to it for a couple days.

Hi Ken,

I noticed the following while fixing this issue.

We now noticed that NtWriteFile() may return STATUS_PENDING
even in nonblocking mode.

If NtWriteFile() returns STATUS_PENDING in nonblocking mode,
does not the following 'if' block in raw_write() refer to 
uninitialized io.Information?

      while (len1 > 0)
        {
          status = NtWriteFile (get_handle (), evt, NULL, NULL, &io,
                                (PVOID) ptr, len1, NULL, NULL);
          if (evt || !NT_SUCCESS (status) || io.Information > 0
              || len <= PIPE_BUF)
            break;
          len1 >>= 1;
        }

'evt' is false if we are in nonblocking mode.
'!NT_SUCCESS(status)' is false if status == STATUS_PENDING.
Then io.Information would be referred I think.

Isn't this another bug in raw_write()?

What should we do in this case? Should we do like:

          if (evt || !NT_SUCCESS (status) || status == STATUS_PENDING
              || io.Information > 0 || len <= PIPE_BUF)

?

-- 
Takashi Yano <takashi.yano@nifty.ne.jp>


More information about the Cygwin-developers mailing list