Re: [PATCH] 2/2 mtd: Add support for the Dreamcast VMU flash

[Adrian McMenamin <adrian@newgolddream.dyndns.info>]

On Thu, 2008-03-20 at 14:10 -0700, Andrew Morton wrote:

> If this error is taken, vmu_flash_write() (at least) will not call
> maple_vmu_write_block() and hence mdev->mq->sem never gets up()ed.  As far
> as I can tell.
> 
> Also, when this function is called from vmu_flash_read_char() I can't see
> where mdev->mq->sem gets up()ed ever.  But I didn't look too hard.
> 

It's not an issue - the bus will report "a no reply" at the end of the
bus read and that will result in an automatic up.


<snip>
> > +
> > +	for (x = 0; x < card->writecnt; x++) {
> > +		/* take the lock to protect the contents of sendbuf */
> > +		locking = down_interruptible(&mdev->mq->sem);
> > +		if (locking) {
> > +			error = -EBUSY;
> > +			goto fail_nolock;
> > +		}
> 
> Confused.  Where within this loop does the up(&mdev->mq->sem) happen?
> 

It's in the bus code.


> > +		((unsigned long *)sendbuf)[1] > > +			cpu_to_be32(partition << 24 | x << 16 | num);
> > +		memcpy(sendbuf + 8, buf + phaselen * x, phaselen);
> > +		mdev->mq->sendbuf = sendbuf;
> > +		/* wait for the mutex to be available */
> > +		maple_add_packet(mdev->mq);
> 
> Within here, I guess, via a workqueue or timer function?


It's handled in a workqueue in the main bus code, yes.