Skip to contentSkip to conversation

parport_pc.c PnP BIOS sanity check

PATCH Thomas Hood 5 participants 6 messages Sep 21, 2001 – Sep 22, 2001 Parallel port

lore
Share & tools
lore
#0 lore raw

I'm still wondering why this function in parport_pc.c rejects dma values of zero. Is DMA0 not usable by the parallel port for some reason? I should think that if the PnP BIOS returns a dma of zero then it means that the parallel port is using DMA0. Sorry if I'm
being obtuse. // Thomas Hood

parport_pc.c line 2799:
int init_pnp040x(struct pci_dev *dev)
{ int io,iohi,irq,dma;

        io=dev->resource[0].start;
        iohi=dev->resource[1].start;
        irq=dev->irq_resource[0].start;
        dma=dev->dma_resource[0].start;

        if(dma==0) dma=-1;             <--- Why?

        printk(KERN_INFO "PnPBIOS: Parport found %s %s at io=%04x,%04x irq=%d dma=%d\n",
                dev->name,dev->slot_name,io,iohi,irq,dma);
        if (parport_pc_probe_port(io,iohi,irq,dma,NULL))
                return 1;
        return 0;
}
drivers/parport/parport_pc.c renamed from drivers/parport/parport_pc.c_ORIG +0 -2
@@ -2797,8 +2797,6 @@
 	irq=dev->irq_resource[0].start;
 	dma=dev->dma_resource[0].start;
 
-	if (dma==0) dma=-1;
-
 	printk(KERN_INFO "PnPBIOS: Parport found %s %s at io=%04x,%04x irq=%d dma=%d\n",
 		dev->name,dev->slot_name,io,iohi,irq,dma);
 	if (parport_pc_probe_port(io,iohi,irq,dma,NULL))
drivers/parport/parport_pc.c renamed from drivers/parport/parport_pc.c_ORIG +1 -1
@@ -2797,7 +2797,7 @@
 	irq=dev->irq_resource[0].start;
 	dma=dev->dma_resource[0].start;
 
-	if (dma==0) dma=-1;
+	if (dma==0) dma=PARPORT_DMA_NONE;
 
 	printk(KERN_INFO "PnPBIOS: Parport found %s %s at io=%04x,%04x irq=%d dma=%d\n",
 		dev->name,dev->slot_name,io,iohi,irq,dma);
#1 lore raw
Quoted 5 lines · Thomas Hood

I'm still wondering why this function in parport_pc.c rejects dma values of zero. Is DMA0 not usable by the parallel port for some reason? I should think that if the PnP BIOS returns a dma of zero then it means that the parallel port is using DMA0. Sorry if I'm
being obtuse. // Thomas Hood

DMA0 is reserved for memory refresh. It _can't_ be used for anything else, therefore a value of 0 is representative of no value whatsoever.

Signature · 1 line
Ignacio Vazquez-Abrams  <[email protected]>
#2 lore raw
Quoted 1 line

therefore a value of 0 is representative of no value whatsoever.

This has been unsafe since about 1995, when DMA 0 became available as PC's stopped using the ISA DMA engine for memory refresh (a very neat original PC hack)

Alan

#5 lore raw
In reply to Thomas Hood
Quoted 6 lines · Thomas Hood

I'm still wondering why this function in parport_pc.c rejects dma values of zero. Is DMA0 not usable by the parallel port for some reason? I should think that if the PnP BIOS returns a dma of zero then it means that the parallel port is using DMA0. Sorry if I'm
being obtuse. // Thomas Hood

1)
I think I saw some BIOS report DMA0 for "none" (could even have been ACPI which is returning PNP formatted legacy resource data).

2)
I have never seen DMA0 for parport configured by a BIOS.

3)
Try "lssuperio" if you want the real hardware thing.

This qualifies the code as it is as a sanity check.

#3 lore raw
In reply to Ignacio Vazquez-Abrams

Thanks for the replies.

I'd like to underline Alan Cox's reply that DMA0 _is_ usable. In fact, my sound chip is configured to use DMA0.

If (and only if) parport _cannot_ use DMA0 for some reason then the sanity check is justified. In that case I'd just ask that a short comment be added to the code that gives the reason.

Cheers
Thomas

Quoted 2 lines · Ignacio Vazquez-Abrams

DMA0 is reserved for memory refresh. It _can't_ be used for anything else, therefore a value of 0 is representative of no value whatsoever.

Quoted 3 lines · Alan Cox

This has been unsafe since about 1995, when DMA 0 became available as PC's stopped using the ISA DMA engine for memory refresh (a very neat original PC hack)

Quoted 9 lines · Gunther Mayer

1)
I think I saw some BIOS report DMA0 for "none" (could even have been ACPI which is returning PNP formatted legacy resource data).
2)
I have never seen DMA0 for parport configured by a BIOS.
3)
Try "lssuperio" if you want the real hardware thing.

This qualifies the code as it is as a sanity check.

#4 lore raw
In reply to Ignacio Vazquez-Abrams
Quoted 2 lines

DMA0 is reserved for memory refresh. It _can't_ be used for anything else, therefore a value of 0 is representative of no value whatsoever.

On a PC/XT, perhaps. On any other computer, nothing of the sort is guaranteed.

p.