Select

Select

[David Feuer]

In the discussion on "select bug", some people noted that select does not 
wake up a process until the buffer is half full (or all full, or 
whatever).  Does this mean that if a small amount is written to the 
device/pipe the process may never be woken?  Or is there a time limit that 
wakes up the process after a certain amount of time if there are _any_ 
bytes in the pipe/dev?

--
This message has been brought to you by the letter alpha and the number pi.
David Feuer
David_Feuer@brown.edu

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: Select

[Andi Kleen]

On Sun, Nov 05, 2000 at 01:46:19AM -0500, David Feuer wrote:
> In the discussion on "select bug", some people noted that select does not 
> wake up a process until the buffer is half full (or all full, or 
> whatever).  Does this mean that if a small amount is written to the 
> device/pipe the process may never be woken?  Or is there a time limit that 
> wakes up the process after a certain amount of time if there are _any_ 
> bytes in the pipe/dev?

This only applies to POLLOUT. Yes, if an send buffer stays 90%
full forever you may be never woken up.

In case of a TCP socket you would be at worst waken up after considerable
time with a ETIMEDOUT. 

Pipes do not signal writable until the pipe is empty. 


-Andi
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: Select

[Alan Cox]

> wake up a process until the buffer is half full (or all full, or 
> whatever).  Does this mean that if a small amount is written to the 

Writer not reader
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/