88SE9172 additional PCI ID: 1b4b:9172
On a Gigabyte X79-UP4 motherboard, which is documented as having 3x 88SE9172, lspci -nn gives me:
05:00.0 SATA controller [0106]: Marvell Technology Group Ltd. 88SE9172 SATA 6Gb/s Controller [1b4b:9172] (rev 11) 06:00.0 SATA controller [0106]: Marvell Technology Group Ltd. 88SE9172 SATA 6Gb/s Controller [1b4b:9172] (rev 11) 07:00.0 SATA controller [0106]: Marvell Technology Group Ltd. 88SE9172 SATA 6Gb/s Controller [1b4b:9172] (rev 11) 08:00.0 SATA controller [0106]: Marvell Technology Group Ltd. 88SE9123 PCIe SATA 6.0 Gb/s controller [1b4b:9123] (rev 11)
(The fourth is an expansion card I have pluggied into a PCIE x1 slot.)
drivers/ata/ahci.c lists:
{ PCI_DEVICE(0x1b4b, 0x917a),
.driver_data = board_ahci_yes_fbs }, /* 88se9172 */
{ PCI_DEVICE(0x1b4b, 0x9192),
.driver_data = board_ahci_yes_fbs }, /* 88se9172 on some Gigabyte */Is that second one a typo, or do we need a third line for the 0x9172 PCI ID? For now, I have added the third line and it's mostly working.
(If I enable VT-d and the IOMMU, the devices are recognized but not the disks plugged into them, but I'm still working on figuring that out.)
Quoted 8 lines · George Spelvin
On a Gigabyte X79-UP4 motherboard, which is documented as having 3x 88SE9172, lspci -nn gives me:
05:00.0 SATA controller [0106]: Marvell Technology Group Ltd. 88SE9172 SATA 6Gb/s Controller [1b4b:9172] (rev 11) 06:00.0 SATA controller [0106]: Marvell Technology Group Ltd. 88SE9172 SATA 6Gb/s Controller [1b4b:9172] (rev 11) 07:00.0 SATA controller [0106]: Marvell Technology Group Ltd. 88SE9172 SATA 6Gb/s Controller [1b4b:9172] (rev 11) 08:00.0 SATA controller [0106]: Marvell Technology Group Ltd. 88SE9123 PCIe SATA 6.0 Gb/s controller [1b4b:9123] (rev 11)
Quoted 7 lines
(The fourth is an expansion card I have pluggied into a PCIE x1 slot.)
drivers/ata/ahci.clists:{ PCI_DEVICE(0x1b4b, 0x917a), .driver_data = board_ahci_yes_fbs }, /* 88se9172 */ { PCI_DEVICE(0x1b4b, 0x9192), .driver_data = board_ahci_yes_fbs }, /* 88se9172 on some Gigabyte */
I don't think it's a typo.
Quoted 2 lines
Is that second one a typo, or do we need a third line for the
0x9172PCI ID? For now, I have added the third line and it's mostly working.
And we'll need a separate entry
Quoted 2 lines
(If I enable VT-d and the IOMMU, the devices are recognized but not the disks plugged into them, but I'm still working on figuring that out.)
when the device gets working.
Thanks!
Signature · 1 line
tejun
Quoted 1 line
I don't think it's a typo.
Indeed, pci.ids has three entries.
Quoted 6 lines
And we'll need a separate entry
Quoted 2 lines
(If I enable VT-d and the IOMMU, the devices are recognized but not the disks plugged into them, but I'm still working on figuring that out.)
when the device gets working.
I don't quite understand this. With VT-d disabled in the BIOS, the devices *are* working. I even have a port multiplier plugged into one. (Which gets all confused by hdparm -m16, but whatever.)
When I get time, I'm going to play with various BIOS and kernel options to narrow down the problem. But that's for later.
Can you explain what you think is not working now?
Hello,
Quoted 10 lines · George Spelvin
Quoted 1 line
when the device gets working.
I don't quite understand this. With VT-d disabled in the BIOS, the devices *are* working. I even have a port multiplier plugged into one. (Which gets all confused by hdparm -m16, but whatever.)
When I get time, I'm going to play with various BIOS and kernel options to narrow down the problem. But that's for later.
Can you explain what you think is not working now?
I just misunderstood what you were saying in the previous email. Can you please post a proper patch with SOB?
Thanks.
Signature · 1 line
tejun
PATCH 1/1 Use observed PCI ID for Marvell 88se9172 SATA controller.
A third possible PCI ID, as personally observed, and found in the pci.ids list.
--- drivers/ata/ahci.c | 2 ++ 1 file changed, 2 insertions(+)
drivers/ata/ahci.c +2 -0
@@ -421,6 +421,8 @@ static const struct pci_device_id ahci_pci_tbl[] = { .driver_data = board_ahci_yes_fbs }, /* 88se9128 */ { PCI_DEVICE(0x1b4b, 0x9125), .driver_data = board_ahci_yes_fbs }, /* 88se9125 */ + { PCI_DEVICE(0x1b4b, 0x9172), + .driver_data = board_ahci_yes_fbs }, /* 88se9172 */ { PCI_DEVICE(0x1b4b, 0x917a), .driver_data = board_ahci_yes_fbs }, /* 88se9172 */ { PCI_DEVICE(0x1b4b, 0x9192),
Signature · 1 line
1.8.3.rc3
Tejun Heo
AppliedPATCH 1/1 ahci: add an observed PCI ID for Marvell 88se9172 SATA controller
Applied to libata/for-3.10-fixes. In the future, please base patches on top of git://git.kernel.org/pub/scm/linux/kernel/git/tj/libata.git
for-XXX branches.
Thanks.
Quoted 1 line
From
fcce9a35f8faaa1f52236c554ef1b15d99a7537eMon Sep 17 00:00:00 2001
From: George Spelvin <[email protected]>
Date: Wed, 29 May 2013 10:20:35 +0900
A third possible PCI ID, as personally observed, and found in the pci.ids list.
--- drivers/ata/ahci.c | 2 ++ 1 file changed, 2 insertions(+)
drivers/ata/ahci.c +2 -0
@@ -423,6 +423,8 @@ static const struct pci_device_id ahci_pci_tbl[] = { .driver_data = board_ahci_yes_fbs }, /* 88se9125 */ { PCI_DEVICE(PCI_VENDOR_ID_MARVELL_EXT, 0x917a), .driver_data = board_ahci_yes_fbs }, /* 88se9172 */ + { PCI_DEVICE(PCI_VENDOR_ID_MARVELL_EXT, 0x9172), + .driver_data = board_ahci_yes_fbs }, /* 88se9172 */ { PCI_DEVICE(PCI_VENDOR_ID_MARVELL_EXT, 0x9192), .driver_data = board_ahci_yes_fbs }, /* 88se9172 on some Gigabyte */ { PCI_DEVICE(PCI_VENDOR_ID_MARVELL_EXT, 0x91a3),
Signature · 1 line
1.8.1.4
Re: [PATCH] ahci: add an observed PCI ID for Marvell 88se9172 SATA controller
Quoted 3 lines
Applied to libata/for-3.10-fixes. In the future, please base patches on top of
git://git.kernel.org/pub/scm/linux/kernel/git/tj/libata.git
for-XXX branches.
As you saw, it didn't initially occur to me to prepare a patch at all; it seemed such a trivial copy/paste/one-character-edit exercise. After diagnosis, the most difficult part was writing a meaningful commit message.
Which I presumed would be easier for you; a lot of the difficulty for me was trying to figure out one line that would fit the ongoing narrative that is the kernel changelog.
I was actually surprised when you asked for a full patch, so I just sent you a copy of my private commit, with an improved commit message and S-o-b. (Should I have added "Cc: stable" to it?)
But since I now know you prefer even trivial changes like this in formal patch form, I'll try to do it right in future.
Re: [PATCH] ahci: add an observed PCI ID for Marvell 88se9172 SATA controller
Hello,
Quoted 9 lines · George Spelvin
As you saw, it didn't initially occur to me to prepare a patch at all; it seemed such a trivial copy/paste/one-character-edit exercise. After diagnosis, the most difficult part was writing a meaningful commit message.
Which I presumed would be easier for you; a lot of the difficulty for me was trying to figure out one line that would fit the ongoing narrative that is the kernel changelog.
Quoted 6 lines
I was actually surprised when you asked for a full patch, so I just sent you a copy of my private commit, with an improved commit message and S-o-b. (Should I have added "Cc: stable" to it?)
But since I now know you prefer even trivial changes like this in formal patch form, I'll try to do it right in future.
It's partly for convenience on my part but more to attribute changes correctly. This is something you did so if at all possible I'd like you to take the credit and (however minute it may be) the accompanying responsibility. It doesn't have to be perfect and I'm happy to edit the commit messages / patches as necessary and adding cc to stable. Well, at least unless you're gonna be submitting large volume of patches.
Thanks.
Signature · 1 line
tejun
Quoted 11 lines · George Spelvin
Quoted 1 line
I don't think it's a typo.
Indeed, pci.ids has three entries.
Quoted 6 lines
And we'll need a separate entry
Quoted 2 lines
(If I enable VT-d and the IOMMU, the devices are recognized but not the disks plugged into them, but I'm still working on figuring that out.)
when the device gets working.
Quoted 8 lines
I don't quite understand this. With VT-d disabled in the BIOS, the devices *are* working. I even have a port multiplier plugged into one. (Which gets all confused by hdparm -m16, but whatever.)
When I get time, I'm going to play with various BIOS and kernel options to narrow down the problem. But that's for later.
Can you explain what you think is not working now?
There are some known problems with some Marvell controllers and VT-d, apparently they issue PCIe requests containing the wrong device function causing IOMMU faults to occur. I believe a PCI quirk to work around the problem was either being worked on or finished - if it's in already, it might need this device ID added to it as well.