On (Tue) Mar 23 2010 [12:04:28], Gerd Hoffmann wrote: > On 03/23/10 11:52, Amit Shah wrote: > >>> With poll support done right (don't signal "writable" when not >>> connected) this shouldn't be a issue. Apps retrying over and over >>> without waiting for the filehandle becoming writable are broken by >>> design. >> >> Yeah; poll returns POLLOUT only when the host connection is open. This >> is for apps that write() without consulting poll(). > > Well, /me just remembers this also depends on the file mode. Returning > -EAGAIN instantly is fine for non-blocking file handles. For file > handles in blocking mode you would have to block here, waiting for the > host to connect. Ah, right. I didn't really look at blocking / non-blocking support in write() because of the cpu_relax(), which will always block the app. I'll add this. Amit _______________________________________________ Virtualization mailing list Virtualization@xxxxxxxxxxxxxxxxxxxxxxxxxx https://lists.linux-foundation.org/mailman/listinfo/virtualization
