Joachim Worringen wrote:
> Greetings,
>
> I have two kernel modules (brought over from Linux), one being a
> hardware driver ("D") and another a plain module "S" (actually, a socket
> module, see project volo-gate).
>
> S depends on D as it calls its functions. This is expressed in the link
> step via "-dy -ND". Things load fine, but problems are with D
> attaching/detaching:
> - D does not load or even attach when S is loaded, but only when I i.e.
> read from its /dev entry.
> - as soon as D attaches, things work well (S calling D's funcions get
> the expected result)
> - however, D detaches after some time (the kernel seems to trigger this,
> not D itself!), and if S calls D's functions now, it gets results
> indicating "now hardware found" (internal state of D when not being
> attached).
> - once I manually make D attach again, thinks work well. Both D and S
> remain loaded all the time.
>
> This should be a pretty common problem - how is this handled properly?
>
> Joachim
>
>
We have this same issue with RDS (reliable datagram sockets) that runs
over IB. Most system do not have IB hardware, So just like you, we
also have two drivers. When a socket is opened, the socket driver checks
if the HW driver has been loaded, if not it loads the driver by opening
the device.
see rds_open()/rds_open_transport_driver()/rds_close_transport_driver()
rds_open()
{
<..........>
/* Open the transport driver if IB HW is present */
rw_enter(&rds_transport_lock, RW_READER);
if (rds_transport_handle == NULL) {
rw_exit(&rds_transport_lock);
ret = rds_open_transport_driver();
rw_enter(&rds_transport_lock, RW_READER);
if (ret != 0) {
/* Transport driver failed to load */
rw_exit(&rds_transport_lock);
return (ret);
}
<.....>
Rao.
_______________________________________________
opensolaris-code mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
