NFTL Performance

[David Woodhouse]

On Tue, 6 Feb 2001, Ian Eure <Ian Eure wrote:

> a more important problem that i see is that when blocks are being
> flushed to the DOC, certain parts of the system (mouse, maybe others) stop
> responding until the write is complete.

More likely to be the busy-wait in _DoC_WaitReady():

 while (!(ReadDOC(docptr, CDSNControl) & CDSN_CTRL_FR_B) && --c)
                ;

Try... 
	while (!(ReadDOC(docptr, CDSNControl) & CDSN_CTRL_FR_B) && --c) {
		if (current->need_resched)
			schedule();
		else
			udelay(1);
	}

But I think you have to protect against being re-entered by using a 
'state' variable to mark that you're currently writing to the chip, and 
probably also a spinlock to protect that state variable. The NOR flash 
drivers do all this if you're after an example.

-- 
dwmw2




To unsubscribe, send "unsubscribe mtd" to majordomo at infradead.org