Re: [squid-users] squid source problem

On 21 Aug 2001 23:19:31 -0700, fatemeh bishemar wrote:
> hi,
> I read squid source,but I have some problems.
> my understandig is that:

nearly right.

Replace

> "when squid finished initialization(that includes
> regestering handler for
> each fd for i/o events or read/write from/to socket)
> ,if it run as a daemon in
> background it fork and create a child process that
> continue the rest of the
> process and parent process wait until child process
> exit,and then fork
> another child and its go on.

with

"when squid is executed it closes all fds, so as to run as a daemon in
the background. It then forks and creates a child process that continues
the rest of the process and the parent process waits until the child
process exists, and then forks another child and so on.

The child initialises the software including registering a handler for
each fd for io events or read/write from/to socket.

and the rest is correct.

> after that child process call eventrun and in taht
> functions E link list of the
> event will be scanned and the event that its the time
> to run will be run and
> they remove from link list.until arriving to an event
> that its time feild is
> greater than current time. then its called nextevent
> function that return the
> time of the next event then squid pass the time - as
> timeout- to the
> comm_select function. in this function select system
> call will be activated
> on all fds(for read write or close)if one fd is ready
> its related handler will
> be called after time out ocurred its call eventrun
> function and go on
> until an error occured or shutdown requeted or ... ,
> and then the child
> exit."

> first,Is it true ?. if no, what is the machanism?
Nearly. see above,

> second,if yes,What happens if during execution of the
> child process
> another request arrives from a client? Which of the
> parent and child will
> reply the request?

The child - the parent never calls bind().
Rob
Received on Wed Aug 22 2001 - 00:52:58 MDT