Hi Rusty, On 08/19/2014 03:57 PM, Rusty Russell wrote: > POSIX says: "POLLOUT Normal data may be written without blocking.". > This is "may" is misleading, see the POSIX write page: > > Write requests to a pipe or FIFO shall be handled in the same way as > a regular file with the following exceptions: ... If the O_NONBLOCK > flag is clear, a write request may cause the thread to block, but on > normal completion it shall return nbyte. > > ... When attempting to write to a file descriptor (other than a pipe > or FIFO) that supports non-blocking writes and cannot accept the data > immediately: > > If the O_NONBLOCK flag is clear, write() shall block the calling > thread until the data can be accepted. > > If the O_NONBLOCK flag is set, write() shall not block the thread. If > some data can be written without blocking the thread, write() shall > write what it can and return the number of bytes written. Otherwise, > it shall return -1 and set errno to [EAGAIN].> > The net result is that write() of more than 1 byte on a socket, pipe or FIFO > which is "ready" may block: write() (unlike read!) will attempt to write > the entire buffer and only return a short write under exceptional > circumstances. > > Indeed, this is the behaviour we see in Linux: > > https://github.com/rustyrussell/ccan/commit/897626152d12d7fd13a8feb36989eb5c8c1f3485 > https://plus.google.com/103188246877163594460/posts/BkTGTMHDFgZ > > Signed-off-by: Rusty Russell <rusty@xxxxxxxxxxxxxxx> Thanks for the detailed log message. This all makes sense and I've applied the patch. One minor correction your point above. On pipes/FIFOs at least, select() and poll() return ready only if at least PIPE_BUF bytes of space are available. (On some other implementations that I've tested, they do however return true even if there is just one byte of space available. Cheers, Michael > diff --git a/man2/poll.2 b/man2/poll.2 > index 53aec82..9167472 100644 > --- a/man2/poll.2 > +++ b/man2/poll.2 > @@ -167,7 +167,10 @@ There is urgent data to read (e.g., out-of-band data on TCP socket; > pseudoterminal master in packet mode has seen state change in slave). > .TP > .B POLLOUT > -Writing now will not block. > +Writing is now possible, though a write larger that the available space > +in a socket or pipe will still block (unless > +.B O_NONBLOCK > +is set). > .TP > .BR POLLRDHUP " (since Linux 2.6.17)" > Stream socket peer closed connection, > diff --git a/man2/select.2 b/man2/select.2 > index ac70b85..ea35986 100644 > --- a/man2/select.2 > +++ b/man2/select.2 > @@ -86,9 +86,10 @@ allow a program to monitor multiple file descriptors, > waiting until one or more of the file descriptors become "ready" > for some class of I/O operation (e.g., input possible). > A file descriptor is considered ready if it is possible to > -perform the corresponding I/O operation (e.g., > -.BR read (2)) > -without blocking. > +perform a corresponding I/O operation (e.g., > +.BR read (2) > +without blocking, or a sufficiently small > +.BR write (2)). > .PP > The operation of > .BR select () > @@ -131,8 +132,8 @@ available for reading (more precisely, to see if a read will not > block; in particular, a file descriptor is also ready on end-of-file), > those in > .I writefds > -will be watched to see if a write will not block, and > -those in > +will be watched to see if space is available for write (though a large > +write may still block), and those in > .I exceptfds > will be watched for exceptions. > On exit, the sets are modified in place > -- Michael Kerrisk Linux man-pages maintainer; http://www.kernel.org/doc/man-pages/ Linux/UNIX System Programming Training: http://man7.org/training/ -- To unsubscribe from this list: send the line "unsubscribe linux-man" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
