Re: TCP sockets not flagged as writable?

[Michal Ostrowski <mostrows@xxxxxxxxxx>]

kuznet@xxxxxxxxxxxxx writes:
 > Hello!
 > 
 > > Is this wrong or is this just my imagination?
 > 
 > Not this and not this, the third alternative is "simply right". 8)
 > 
 > Wakeup happens as soon as there is _enough_ space to wake up.
 > 1 byte is not an enough.


But my concern is that when you have just a little bit of space
available a write operation may succeed without blocking, even though
poll() says it won't.

I understand why the criteria for poll() are as strict as they are,
I'm just puzzled as to why the criteria are relaxed in tcp_sendmsg().
Should tcp_sendmsg() not also rely on the 1/3 free requirement?

Michal Ostrowski
mostrows@xxxxxxxxxxxxx