Device timeout from mfi(9) while booting 9.2-RELEASE

[Konstantin Belousov]

On Thu, Oct 03, 2013 at 11:50:15AM -0400, Ryan Stone wrote:
> We're having trouble booting 9.2-RELEASE on a system with an mfi
> (Drake Skinny) controller.  When it tries to mount root we get:
> 
> mfi0: COMMAND 0xffffff8000cb7220 TIMEOUT AFTER 59 SECONDS
> mfi0: COMMAND 0xffffff8000cb72a8 TIMEOUT AFTER 60 SECONDS
> mfi0: COMMAND 0xffffff8000cb7220 TIMEOUT AFTER 90 SECONDS
> 
> I also get the following when mfi attaches (with bootverbose):
> 
> mfi0: <Drake Skinny> port 0x5000-0x50ff mem 0xfbd60000-0xfbd63fff,0xfbd00000-0xf
> bd3ffff irq 16 at device 0.0 on pci2^M
> mfi0: attempting to allocate 1 MSI vectors (1 supported)^M
> msi: routing MSI IRQ 282 to local APIC 0 vector 77^M
> mfi0: using IRQ 282 for MSI^M
> mfi0: Using MSI^M
> mfi0: Megaraid SAS driver Ver 4.23 ^M
> mfi0: Frame 0xffffff85d394b280 timed out command 0x1010000^M
> error 60 in callback from mfi_send_frame^M
> mfi0: Frame 0xffffff85d394b500 timed out command 0x1040100^M
> error 60 in callback from mfi_send_frame^M
> mfi0: Frame 0xffffff85d394b780 timed out command 0x1040300^M
> error 60 in callback from mfi_send_frame^M
> mfi0: Error 255 fetching controller entries^M
> 
> I've put full bootverbose logs from boots with hw.mfi.msi=0 and
> hw.mfi.msi=1 here:
> http://people.freebsd.org/~rstone/mfi_timeout_irq.txt
> http://people.freebsd.org/~rstone/mfi_timeout_msi.txt
> 
> Neither boot was successful.  I don't have the ability to boot from
> alternative media so I have a limited ability to test changes (I can
> re-image from a full disk image, but that requires building a new disk
> image which takes some time).
> 
> I did see the following forum post and tried with cbb and pccard
> removed from GENERIC.  That didn't fix it.
> http://forums.freebsd.org/showpost.php?p=231135&postcount=4

I have two Drake Skinny cards, one four-ports, one with eight ports, both
not working with either version of driver in HEAD from February up to now.
I tried to talk about this both with Doug Ambrisko and
"Desai, Kashyap" <Kashyap.Desai at lsi.com>, and provided the debugging
data I could think of, but no avail.

There is a rumor that it might be related to the slot the card is
inserted in. I only have 8x or wider PCIe slot provided by CPU
north-bridge. It might be that card does not like it and only work in
the south-bridge PCIe lanes. I have no idea why could it be so, neither
I saw the confirmation of the speculation.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 834 bytes
Desc: not available
URL: <http://lists.freebsd.org/pipermail/freebsd-stable/attachments/20131003/6bc762c2/attachment.sig>