Can not see/access devices on Marvell 88SE9485 + SiI 3726 PMP
Hi,
I've got a box which recently got a Marvell 88SE9485 PCIe card (Supermicro AOC-SAS2LP-MV8), which is connected by a mSAS->4x SATA-cable to some 5 HDD backplanes (CFI-B53PM, chip is a SiI 3726).
The HBA's BIOS can see and access the HDDs through the backplanes, which various Linux distributions could not.
I tried Ubuntu 12.04.2, Ubuntu 13.04, Debian 7.0 and a current Arch Linux. All logs in this mail are from a fresh Ubuntu 12.04.2 install.
lspci shows the controller as:
01:00.0 RAID bus controller: Marvell Technology Group Ltd. Device 9485 (rev 03)
lshw does not show the attached disks, nor are there device nodes in /dev/ for those disks.
Unplugging a disk results in this (/var/log/kern.log):
kernel: [ 276.995910] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2062:phy 4 ctrl sts=0x00000000.
kernel: [ 276.995914] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2064:phy 4 irq sts = 0x01201001
kernel: [ 278.950547] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2062:phy 4 ctrl sts=0x00111000.
kernel: [ 278.950551] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2064:phy 4 irq sts = 0x00000081
kernel: [ 279.502757] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_94xx.c 838:get all reg link rate is 0x111000
kernel: [ 279.502760] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_94xx.c 843:get link rate is 9
kernel: [ 279.582118] mvsas 0000:01:00.0: Phy0 : No sig fis
kernel: [ 279.582121] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2006:phy0 Attached Device
kernel: [ 283.953502] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2048:Get signature time out, reset phy 4
kernel: [ 283.953517] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2062:phy 4 ctrl sts=0x00000000.
kernel: [ 283.953519] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2064:phy 4 irq sts = 0x00001001
kernel: [ 284.449485] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2062:phy 4 ctrl sts=0x00111000.
kernel: [ 284.449488] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2064:phy 4 irq sts = 0x00000081
kernel: [ 286.445635] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_94xx.c 838:get all reg link rate is 0x111000
kernel: [ 286.445638] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_94xx.c 843:get link rate is 9
kernel: [ 286.524997] mvsas 0000:01:00.0: Phy0 : No sig fis
kernel: [ 286.525000] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2006:phy0 Attached Device
Replugging the disk:
kernel: [ 310.738784] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2062:phy 4 ctrl sts=0x00111000.
kernel: [ 310.738788] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2064:phy 4 irq sts = 0x00010000
kernel: [ 310.738791] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2113:notify plug in on phy[0]
kernel: [ 310.758632] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_94xx.c 838:get all reg link rate is 0x111000
kernel: [ 310.758634] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_94xx.c 843:get link rate is 9
kernel: [ 310.877669] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 1175:phy 4 attach dev info is 20001
kernel: [ 310.877671] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 1177:phy 4 attach sas addr is 4
kernel: [ 310.877678] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 277:phy 4 byte dmaded.
kernel: [ 310.877689] sas: sas_form_port: phy0 belongs to port0 already(1)!
According to Marvell's product brief the controller has "Built-in support for SATA Port Multipliers with FIS-based switching", which its BIOS seems to confirm.
The PMPs used to work when connected via a Syba SY-PEX40008 (SiI3124) controller, which has bugs on its own when more than three drives per PMP are concurrently active.
Any help is greatly appreciated,
Hajo
James Bottomley
maintainerQuoted 98 lines · Hajo Möller
Hi,
I've got a box which recently got a Marvell 88SE9485 PCIe card (Supermicro AOC-SAS2LP-MV8), which is connected by a mSAS->4x SATA-cable to some 5 HDD backplanes (CFI-B53PM, chip is a SiI 3726).
The HBA's BIOS can see and access the HDDs through the backplanes, which various Linux distributions could not.
I tried Ubuntu 12.04.2, Ubuntu 13.04, Debian 7.0 and a current Arch Linux. All logs in this mail are from a fresh Ubuntu 12.04.2 install.lspci shows the controller as:
01:00.0 RAID bus controller: Marvell Technology Group Ltd. Device 9485 (rev 03)lshw does not show the attached disks, nor are there device nodes in
/dev/for those disks.Unplugging a disk results in this (
/var/log/kern.log):
kernel: [ 276.995910] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2062:phy 4 ctrl sts=0x00000000.
kernel: [ 276.995914] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2064:phy 4 irq sts =0x01201001
kernel: [ 278.950547] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2062:phy 4 ctrl sts=0x00111000.
kernel: [ 278.950551] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2064:phy 4 irq sts =0x00000081
kernel: [ 279.502757] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_94xx.c 838:get all reg link rate is0x111000
kernel: [ 279.502760] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_94xx.c 843:get link rate is 9
kernel: [ 279.582118] mvsas 0000:01:00.0: Phy0 : No sig fis
kernel: [ 279.582121] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2006:phy0 Attached Device
kernel: [ 283.953502] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2048:Get signature time out, reset phy 4
kernel: [ 283.953517] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2062:phy 4 ctrl sts=0x00000000.
kernel: [ 283.953519] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2064:phy 4 irq sts =0x00001001
kernel: [ 284.449485] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2062:phy 4 ctrl sts=0x00111000.
kernel: [ 284.449488] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2064:phy 4 irq sts =0x00000081
kernel: [ 286.445635] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_94xx.c 838:get all reg link rate is0x111000
kernel: [ 286.445638] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_94xx.c 843:get link rate is 9
kernel: [ 286.524997] mvsas 0000:01:00.0: Phy0 : No sig fis
kernel: [ 286.525000] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2006:phy0 Attached DeviceReplugging the disk:
kernel: [ 310.738784] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2062:phy 4 ctrl sts=0x00111000.
kernel: [ 310.738788] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2064:phy 4 irq sts =0x00010000
kernel: [ 310.738791] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 2113:notify plug in on phy[0]
kernel: [ 310.758632] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_94xx.c 838:get all reg link rate is0x111000
kernel: [ 310.758634] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_94xx.c 843:get link rate is 9
kernel: [ 310.877669] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 1175:phy 4 attach dev info is 20001
kernel: [ 310.877671] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 1177:phy 4 attach sas addr is 4
kernel: [ 310.877678] /build/buildd/linux-lts-quantal-3.5.0/drivers/scsi/mvsas/mv_sas.c 277:phy 4 byte dmaded.
kernel: [ 310.877689] sas:sas_form_port: phy0 belongs to port0 already(1)!According to Marvell's product brief the controller has "Built-in support for SATA Port Multipliers with FIS-based switching", which its BIOS seems to confirm.
Quoted 5 lines
The PMPs used to work when connected via a Syba SY-PEX40008 (SiI3124) controller, which has bugs on its own when more than three drives per PMP are concurrently active.
Any help is greatly appreciated,
mv_sas is a libsas based driver. libsas doesn't have any support for SATA PMPs. When it was added they were left as a todo item but then in the field everyone deployed enterprise type SATA devices in SAS expander chassis, so PMP support just got forgotten.
James
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Quoted 4 lines · James Bottomley
mv_sasis a libsas based driver. libsas doesn't have any support for SATA PMPs. When it was added they were left as a todo item but then in the field everyone deployed enterprise type SATA devices in SAS expander chassis, so PMP support just got forgotten.
That's a pity, so it's unlikely libsas will get support for PMPs in the near future or at all? Sadly I'm not able to add the support myself.
Do you know of any SAS controllers which do support SATA PMPs? I'm stuck with the PMPs, so I'm trying a SATA controller with an 88SX7042 next, although having a good SAS controller would be better in the long run.
Greetings,
Hajo
James Bottomley
maintainerQuoted 6 lines · Hajo Möller
Quoted 4 lines · James Bottomley
mv_sasis a libsas based driver. libsas doesn't have any support for SATA PMPs. When it was added they were left as a todo item but then in the field everyone deployed enterprise type SATA devices in SAS expander chassis, so PMP support just got forgotten.
Quoted 7 lines
That's a pity, so it's unlikely libsas will get support for PMPs in the near future or at all? Sadly I'm not able to add the support myself.
Do you know of any SAS controllers which do support SATA PMPs? I'm stuck with the PMPs, so I'm trying a SATA controller with an 88SX7042 next, although having a good SAS controller would be better in the long run.
No, I'm afraid not. Libsas has been released for nearly seven years and no-one has noticed in all that time, which means it's a really rare thing to do. Everyone who uses SAS controllers usually has either a direct connection or an expander. There's no real reason to buy expensive kit and then plug it into a cheap port switch. That said, it's possible that some of the firmware SAT blob based controllers, like the MPT ones might have PMP support ... but because of the rarity it's likely to be buggy.
James
--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html