ahci: Detect Marvell 88SE9172 SATA controller
PATCH 1/1 ahci: Detect Marvell 88SE9172 SATA controller
The Marvell 88SE9172 SATA controller (PCI ID 1b4b 917a) already worked once it was detected, but was missing an ahci_pci_tbl entry.
Boot tested on a Gigabyte Z68X-UD3H-B3 motherboard.
--- drivers/ata/ahci.c | 2 ++ 1 file changed, 2 insertions(+)
drivers/ata/ahci.c +0 -0
@@ -394,6 +394,8 @@ 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, 0x917a),
+ .driver_data = board_ahci_yes_fbs }, /* 88se9172 */
{ PCI_DEVICE(0x1b4b, 0x91a3),
.driver_data = board_ahci_yes_fbs },
Signature · 1 line
1.7.9.5
For a bit more context, the Z68X-UD3H-B3 has the SATA controllers provided by the Z68 chipset plus an extra 2 SATA3 ports from a Marvell 88SE9172-NNX2. The 4 lines on the chip on my board read:
88SE9172-NNX2
PAY4850.3
1116 A1P
TW
It seems that this chip is used on other boards as well
(see http://www.techpowerup.com/reviews/Gigabyte/G1_Sniper2/5.html).
I'm not sure whether 88SE9172-NNX2 is a distinct part from the 88SE9172, or whether the 'NNX2' is manufacturing metadata.
In any case, drives connected to these 2 ports are not recognized at boot time in the latest kernel. I put /boot and the bootloader on a drive attached to one of the Marvell ports. Grub worked fine, but I couldn't boot the kernel; it just dumped me to busybox. Once in busybox, I could mount the drive with 'echo 1b4b 917a > /sys/bus/pci/drivers/ahci/new_id', and Ctrl+d would exit busybox and boot the kernel successfully.
Others have run into similar issues (see
http://theangryangel.co.uk/blog/marvell-88se9172-sata3-under-linux-as-of-320 , which has been copy-pasted onto the Marvell chipset Wikipedia page
http://en.wikipedia.org/wiki/List_of_Marvell_Technology_Group_chipsets ). The PCI ID suggested in the linked page (1b4b 9192) didn't work for me, which suggests that either the poster's chip was not a 88se9172, there are multiple revs of 88se9172 with different PCI IDs, or the -NNX2 is a distinct model with its own PCI ID.
In any case, I think accepting this patch will make boards like mine work in the near term, and we can add additional PCI IDs and clarify the comments for similar chips as people report them or as we get clarification from Marvell.
With this patch, the vanilla and Ubuntu kernels both boot flawlessly with the boot partition hanging off the Marvell controller.
Thanks,
Matt
Quoted 26 lines · Matt Johnson
The Marvell 88SE9172 SATA controller (PCI ID 1b4b 917a) already worked once it was detected, but was missing an
ahci_pci_tblentry.Boot tested on a Gigabyte Z68X-UD3H-B3 motherboard.
Signed-off-by Matt Johnson <[email protected]>--- drivers/ata/ahci.c | 2 ++ 1 file changed, 2 insertions(+)drivers/ata/ahci.c +0
-0@@ -394,6 +394,8 @@ 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 */drivers/ata/ahci.c +2
-0+ { PCI_DEVICE(0x1b4b, 0x917a), + .driver_data = board_ahci_yes_fbs }, /*88se9172 */
{PCI_DEVICE(0x1b4b, 0x91a3),
.driver_data =board_ahci_yes_fbs},
Ping? This patch has been rock-solid on this machine for the last 3 weeks, and the machine won't boot without it. I'm happy to provide any additional context, but otherwise, can this be committed? Thanks!
-Matt
Quoted 68 lines · Matt Johnson
For a bit more context, the Z68X-UD3H-B3 has the SATA controllers provided by the Z68 chipset plus an extra 2 SATA3 ports from a Marvell 88SE9172-NNX2.
The 4 lines on the chip on my board read:
88SE9172-NNX2
PAY4850.3
1116 A1P
TWIt seems that this chip is used on other boards as well
(see http://www.techpowerup.com/reviews/Gigabyte/G1_Sniper2/5.html).
I'm not sure whether 88SE9172-NNX2 is a distinct part from the 88SE9172, or whether the 'NNX2' is manufacturing metadata.
In any case, drives connected to these 2 ports are not recognized at boot time
in the latest kernel. I put /boot and the bootloader on a drive attached to one of the Marvell ports. Grub worked fine, but I couldn't boot the kernel; it just dumped me to busybox. Once in busybox, I could mount the drive with 'echo 1b4b 917a >/sys/bus/pci/drivers/ahci/new_id', and Ctrl+d would exit busybox and boot the kernel successfully.Others have run into similar issues (see
http://theangryangel.co.uk/blog/marvell-88se9172-sata3-under-linux-as-of-320
,
which has been copy-pasted onto the Marvell chipset Wikipedia page
http://en.wikipedia.org/wiki/List_of_Marvell_Technology_Group_chipsets ). The PCI ID suggested in the linked page (1b4b 9192) didn't work for me, which suggests that either the poster's chip was not a 88se9172, there are multiple revs of 88se9172 with different PCI IDs, or the -NNX2 is a distinct model with its own PCI ID.In any case, I think accepting this patch will make boards like mine work in
the near term, and we can add additional PCI IDs and clarify the comments for similar chips as people report them or as we get clarification from Marvell.With this patch, the vanilla and Ubuntu kernels both boot flawlessly with the boot partition hanging off the Marvell controller.
Thanks,
MattQuoted 23 lines · Matt Johnson
The Marvell 88SE9172 SATA controller (PCI ID 1b4b 917a) already worked once it was detected, but was missing an
ahci_pci_tblentry.Boot tested on a Gigabyte Z68X-UD3H-B3 motherboard.
Signed-off-by Matt Johnson <[email protected]>---
drivers/ata/ahci.c| 2 ++1 file changed, 2 insertions(+)drivers/ata/ahci.c +0
-0@@ -394,6 +394,8 @@ 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, 0x917a),
+ .driver_data =board_ahci_yes_fbs}, /* 88se9172 */ {PCI_DEVICE(0x1b4b, 0x91a3),
.driver_data =board_ahci_yes_fbs},