ahci, SActive flag, and the HD activity LED

[Martin Wilck <martin.wilck@fujitsu-siemens.com>]

[-- Attachment #1: Type: text/plain, Size: 1310 bytes --]

Hello Jeff, hello Jens, hello everybody,

I am referring to the debate about whether or not setting the SActive 
bit for non-NCQ ATA commands (e.g. http://lkml.org/lkml/2005/5/26/142).

In our machines, this behavior of the Linux AHCI driver causes the HD 
activity LED to stay on all the time. If I apply the attached trivial 
patch (this is for the RedHat EL4.0-U1 kernel), the LED behaves nicely.

Jeff has stated in the above thread that "SActive is intentionally used 
for non-NCQ devices". However I find clear indication in the specs that 
the SActive flag should be set if and only if tagged queuing is being 
used, and only for a specified subset of commands that support queuing 
(http://www.t13.org/docs2005/D1699r1e-ATA8-ACS.pdf, secs. 4.19 and 
4.20). The current mainline driver doesn't use queuing.

If I am reading the specs correctly, that'd mean the ahci driver is 
wrong in setting the SActive bit. Could you please comment? Jeff, in 
particular, could you please give more detail why you say this flag is 
"intentionally used"?

Regards
Martin

-- 
Martin Wilck                Phone: +49 5251 8 15113
Fujitsu Siemens Computers   Fax:   +49 5251 8 20409
Heinz-Nixdorf-Ring 1        mailto:Martin.Wilck@Fujitsu-Siemens.com
D-33106 Paderborn           http://www.fujitsu-siemens.com/primergy


[-- Attachment #2: ahci_sactive.diff --]
[-- Type: text/x-patch, Size: 383 bytes --]

--- ahci.c.orig	2005-06-13 11:39:26.000000000 +0200
+++ ahci.c	2005-08-02 10:48:47.000000000 +0200
@@ -691,9 +703,6 @@
 	struct ata_port *ap = qc->ap;
 	void *port_mmio = (void *) ap->ioaddr.cmd_addr;
 
-	writel(1, port_mmio + PORT_SCR_ACT);
-	readl(port_mmio + PORT_SCR_ACT);	/* flush */
-
 	writel(1, port_mmio + PORT_CMD_ISSUE);
 	readl(port_mmio + PORT_CMD_ISSUE);	/* flush */