Am Montag, 16. Februar 2004 11:07 schrieb Stephan Linz:
--snip--
@@ -1174,7 +1175,11 @@ static void smc_write_phy_register (byte #ifndef CONFIG_SMC91111_EXT_PHY static void smc_wait_ms(unsigned int ms) {
- /* B E W A R N E D :
* DON'T EXCEED YOUR WATCHDOG TIMOUT WITH ARGUMENT 'ms' */- WATCHDOG_RESET (); udelay(ms*1000);
- WATCHDOG_RESET ();
This should never be necessary. The udelay() function itself is expected to trigger the waitchdog as necessary.
If this is not the case on NIOS, then please fix this first. It might make some more of your modifications redundand, too.
Hu, you're right. It isn't the case on NIOS. OK, I'll fix it up and try out this afternoon. You'll get a new patch tomorrow.
OK, here is the correct patch. Sorry for the delay :-)
Note: In my opinion this is the first stage. In second stage we should move the WATCHDOG_RESET macro from within udelay() C function into the underlayed assembly function dly_clks(). I've tried it out but without success, so I try again.
Best regards, Stephan Linz