Skip to contentSkip to conversation

libata: rework legacy handling to remove much of the cruft

PATCH David Woodhouse 2 participants 2 messages Sep 25, 2006 1 message not in the archive

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

#1 lore raw

Review

Quoted 8 lines · Linux Kernel Mailing List
(quoted diff) +7 -0
+#define ATA_PRIMARY_CMD                0x1F0
+#define ATA_PRIMARY_CTL                0x3F6
+#define ATA_PRIMARY_IRQ                14
+
+#define ATA_SECONDARY_CMD      0x170
+#define ATA_SECONDARY_CTL      0x376
+#define ATA_SECONDARY_IRQ      15

Please, don't do this. We've only just cleaned up the serial driver to get rid of crap like this -- we _don't_ want to do it like this.

We should register the non-discoverable devices as platform devices (or of_devices, or something), and not just hardcode stuff like this in asm/foo.h headers.

Signature · 1 line
dwmw2
#2 lore raw

Ar Llu, 2006-09-25 am 11:14 +0100, ysgrifennodd David Woodhouse:

Quoted 3 lines

We should register the non-discoverable devices as platform devices (or of_devices, or something), and not just hardcode stuff like this in asm/foo.h headers

That would be too late. Those "constants"[1] are part of the actual PCI bus setup. I have some patches to push most of them into drivers/pci/quirks.c. We cannot use platform devices for this in most cases because we already have a device for it - the PCI one.

In the non PCI case we can use platform devices and we do, but the platform device itself has to know what I/O ports are being used so you end up back with the "constants".

The old IDE layer works this way too except that it doesn't use platform devices at all.

Alan
[1] as in "constants aren't, variables won't"