2008/8/2 Tejun Heo [off-list ref]:

Mark Lord wrote:

quoted

I have exactly zero documentation on those chips,
but they are supposedly AHCI compliant.

Looking in drivers/ata/ahci.c, one can see that the chips are indeed
supported,
and there the AHCI_HFLAG_NO_PMP is *not* set, so one might expect them
to work
with a port-multiplier already.  ?

I bet it's more like nobody tried yet.  :-P

 Thanks for your explaining, i had tried it already.

 Port multiplier of Silicon image 3726 is not supported from 2.6.24 to
2.6.26 with 6121 and 6145 which using ahci.c  : (

 Here is the test envirement:

  Port Multipiler : Silicon image 3726. (Marvell 4140 is not fully
tested, so i don't listed, but it is the same in arm arch.)

 arch/kernel   2.6.24        2.6.25       2.6.26

  arm            Not Work   Not tested   Not tested

  powerpc     Not Work   Not tested   Not tested

  x86_64       Not Work   Not Work     Not Work


 Should we need disable 6121/6145 pmp support in ahci.c until it is
fully supported?

Thanks a lot.
--------------------

  Temporal disable pmp support in 6121/6145 until it is fully supported.

Signed-off-by: KueiHuan Chen <redacted>
---

diff --git a/drivers/ata/ahci.c b/drivers/ata/ahci.c
index ef3e552..ef65ac0 100644
--- a/drivers/ata/ahci.c
+++ b/drivers/ata/ahci.c

@@ -420,7 +420,7 @@ static const struct ata_port_info ahci_port_info[] = {
        /* board_ahci_mv */
        {
                AHCI_HFLAGS     (AHCI_HFLAG_NO_NCQ | AHCI_HFLAG_NO_MSI |
-                                AHCI_HFLAG_MV_PATA),
+                                AHCI_HFLAG_MV_PATA | AHCI_HFLAG_NO_PMP),
                .flags          = ATA_FLAG_SATA | ATA_FLAG_NO_LEGACY |
                                  ATA_FLAG_MMIO | ATA_FLAG_PIO_DMA,
                .pio_mask       = 0x1f, /* pio0-4 */

---
Thanks,
KueiHuan-Chen