Patchwork [#upstream-fixes,3/3] ahci: implement AHCI_HFLAG_NO_FPDMA_AA and update NV quirks

login
register
mail settings
Submitter Tejun Heo
Date March 11, 2010, 2:40 a.m.
Message ID <4B98580F.8010908@kernel.org>
Download mbox | patch
Permalink /patch/47298/
State Not Applicable
Delegated to: David Miller
Headers show

Comments

Tejun Heo - March 11, 2010, 2:40 a.m.
It turns out different generations of MCPs have differing quirks.

* MCP 65-73 : FPDMA AA broken, lies about PMP support, forgets to report NCQ
* MCP 77-79 : FPDMA AA broken, lies about PMP support
* MCP 89    : FPDMA AA broken

Instead of turngin off FPDMA AA on all NVIDIAs, implement
HFLAG_NO_FPDMA_AA, define additional board IDs and apply necessary
quirks.

This fixes bko#15481 and the list of quirks is verified by Peer Chen.

  http://bugzilla.kernel.org/show_bug.cgi?id=15481

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Peer Chen <pchen@nvidia.com>
---
 drivers/ata/ahci.c |   39 +++++++++++++++++++++++++++++----------
 1 file changed, 29 insertions(+), 10 deletions(-)

--
To unsubscribe from this list: send the line "unsubscribe linux-ide" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Robert Hancock - March 12, 2010, 12:08 a.m.
On 03/10/2010 08:40 PM, Tejun Heo wrote:
> It turns out different generations of MCPs have differing quirks.
>
> * MCP 65-73 : FPDMA AA broken, lies about PMP support, forgets to report NCQ
> * MCP 77-79 : FPDMA AA broken, lies about PMP support
> * MCP 89    : FPDMA AA broken
>
> Instead of turngin off FPDMA AA on all NVIDIAs, implement
> HFLAG_NO_FPDMA_AA, define additional board IDs and apply necessary
> quirks.
>
> This fixes bko#15481 and the list of quirks is verified by Peer Chen.
>
>    http://bugzilla.kernel.org/show_bug.cgi?id=15481
>
> Signed-off-by: Tejun Heo<tj@kernel.org>
> Cc: Peer Chen<pchen@nvidia.com>

Is the FPDMA AA problem going to be fixed in whatever follows MCP89? 
This change is only safe if it will be, since a new chipset will fall 
into the generic case and it'll get turned back on.
--
To unsubscribe from this list: send the line "unsubscribe linux-ide" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Tejun Heo - March 12, 2010, 1:16 a.m.
Hello,

On 03/12/2010 09:08 AM, Robert Hancock wrote:
> On 03/10/2010 08:40 PM, Tejun Heo wrote:
>> It turns out different generations of MCPs have differing quirks.
>>
>> * MCP 65-73 : FPDMA AA broken, lies about PMP support, forgets to
>> report NCQ
>> * MCP 77-79 : FPDMA AA broken, lies about PMP support
>> * MCP 89    : FPDMA AA broken
>>
>> Instead of turngin off FPDMA AA on all NVIDIAs, implement
>> HFLAG_NO_FPDMA_AA, define additional board IDs and apply necessary
>> quirks.
>>
>> This fixes bko#15481 and the list of quirks is verified by Peer Chen.
>>
>>    http://bugzilla.kernel.org/show_bug.cgi?id=15481
>>
>> Signed-off-by: Tejun Heo<tj@kernel.org>
>> Cc: Peer Chen<pchen@nvidia.com>
> 
> Is the FPDMA AA problem going to be fixed in whatever follows MCP89?
> This change is only safe if it will be, since a new chipset will fall
> into the generic case and it'll get turned back on.

Yeap, it would be great if NVIDIA fixes it but that said blacklisting
by default usually isn't a very good idea because it often falls
through the crack and after a while not many people remember what
happened when.  If NVIDIA doesn't fix the next revision, we'll know
soon enough.

Thanks.

Patch

Index: ata/drivers/ata/ahci.c
===================================================================
--- ata.orig/drivers/ata/ahci.c
+++ ata/drivers/ata/ahci.c
@@ -239,6 +239,7 @@  enum {
 	AHCI_HFLAG_SRST_TOUT_IS_OFFLINE	= (1 << 11), /* treat SRST timeout as
 							link offline */
 	AHCI_HFLAG_NO_SNTF		= (1 << 12), /* no sntf */
+	AHCI_HFLAG_NO_FPDMA_AA		= (1 << 13), /* no FPDMA AA */
 
 	/* ap->flags bits */
 
@@ -267,6 +268,8 @@  enum board_ids {
 
 	/* board IDs for specific chipsets in alphabetical order */
 	board_ahci_mcp65,
+	board_ahci_mcp77,
+	board_ahci_mcp89,
 	board_ahci_mv,
 	board_ahci_sb600,
 	board_ahci_sb700,	/* for SB700 and SB800 */
@@ -276,9 +279,7 @@  enum board_ids {
 	board_ahci_mcp_linux	= board_ahci_mcp65,
 	board_ahci_mcp67	= board_ahci_mcp65,
 	board_ahci_mcp73	= board_ahci_mcp65,
-	board_ahci_mcp77	= board_ahci,
-	board_ahci_mcp79	= board_ahci,
-	board_ahci_mcp89	= board_ahci,
+	board_ahci_mcp79	= board_ahci_mcp77,
 };
 
 struct ahci_cmd_hdr {
@@ -500,7 +501,24 @@  static const struct ata_port_info ahci_p
 	/* by chipsets */
 	[board_ahci_mcp65] =
 	{
-		AHCI_HFLAGS	(AHCI_HFLAG_YES_NCQ),
+		AHCI_HFLAGS	(AHCI_HFLAG_NO_FPDMA_AA | AHCI_HFLAG_NO_PMP |
+				 AHCI_HFLAG_YES_NCQ),
+		.flags		= AHCI_FLAG_COMMON,
+		.pio_mask	= ATA_PIO4,
+		.udma_mask	= ATA_UDMA6,
+		.port_ops	= &ahci_ops,
+	},
+	[board_ahci_mcp77] =
+	{
+		AHCI_HFLAGS	(AHCI_HFLAG_NO_FPDMA_AA | AHCI_HFLAG_NO_PMP),
+		.flags		= AHCI_FLAG_COMMON,
+		.pio_mask	= ATA_PIO4,
+		.udma_mask	= ATA_UDMA6,
+		.port_ops	= &ahci_ops,
+	},
+	[board_ahci_mcp89] =
+	{
+		AHCI_HFLAGS	(AHCI_HFLAG_NO_FPDMA_AA),
 		.flags		= AHCI_FLAG_COMMON,
 		.pio_mask	= ATA_PIO4,
 		.udma_mask	= ATA_UDMA6,
@@ -3287,12 +3305,13 @@  static int ahci_init_one(struct pci_dev
 	/* prepare host */
 	if (hpriv->cap & HOST_CAP_NCQ) {
 		pi.flags |= ATA_FLAG_NCQ;
-		/* Auto-activate optimization is supposed to be supported on
-		   all AHCI controllers indicating NCQ support, but it seems
-		   to be broken at least on some NVIDIA MCP79 chipsets.
-		   Until we get info on which NVIDIA chipsets don't have this
-		   issue, if any, disable AA on all NVIDIA AHCIs. */
-		if (pdev->vendor != PCI_VENDOR_ID_NVIDIA)
+		/*
+		 * Auto-activate optimization is supposed to be
+		 * supported on all AHCI controllers indicating NCQ
+		 * capability, but it seems to be broken on some
+		 * chipsets including NVIDIAs.
+		 */
+		if (!(hpriv->flags & AHCI_HFLAG_NO_FPDMA_AA))
 			pi.flags |= ATA_FLAG_FPDMA_AA;
 	}