Re: [PATCH] updated version of Jens' SATA suspend-to-ram patch

[Jens Axboe <axboe@suse.de>]

On Fri, Sep 23 2005, Jeff Garzik wrote:
> >--- linux-2.6.13/drivers/scsi/libata-core.c~	2005-09-01 
> >12:22:19.000000000 +0200
> >+++ linux-2.6.13/drivers/scsi/libata-core.c	2005-09-01 
> >12:24:38.000000000 +0200
> >@@ -3738,8 +3738,8 @@
> > 	unsigned long flags;
> > 	int rc;
> > 
> >-	qc = ata_qc_new_init(ap, dev);
> >-	BUG_ON(qc == NULL);
> >+	while ((qc = ata_qc_new_init(ap, dev)) == NULL)
> >+		msleep(10);
> > 
> > 	qc->tf.command = cmd;
> > 	qc->tf.flags |= ATA_TFLAG_DEVICE;
> 
> Worried now!
> 
> If this patch is needed, something VERY VERY WRONG is going on.  This 
> patch indicates that the queueing state machine has been violated, and 
> something is trying to IGNORE the command synchronization :(

I haven't diagnosed this further and it only ever happened in the SUSE
kernel to my knowledge (no one has reported it to me for the vanilla
kernels + suspend patch).

So lets just keep this patch out of the equation for now, it could be
that other SUSE patches broke something in this area :/

> Further, you cannot always assume that msleep() is valid in that 
> context.  It should be the caller that waits (libata suspend code), not 
> ata_do_simple_cmd() itself.

ata_do_simple_cmd() always blocks anyways, so I don't see the point.
Perhaps rename the function to ata_execute_and_wait_simple_cmd().

-- 
Jens Axboe