Re: ahcisata & pmp (was: Re: amd64-current (still?) won't boot (another one))

[David Holland <dholland-current%netbsd.org@localhost>]

On Mon, Dec 10, 2012 at 08:32:05PM +0100, Manuel Bouyer wrote:
 > >  > >>>>> "Manuel" == Manuel Bouyer <bouyer%antioche.eu.org@localhost> 
 > > writes:
 > >  > 
 > >  > I've just rebuilt an actual -current kernel yesterday and the problem
 > >  > came back, now with a different drive number (now '0', formerly '15':
 > >  > 
 > >  >   [...]
 > >  >   ahcisata0 channel 0: clearing WDCTL_RST failed for drive 0
 > >  >   ahcisata0 channel 2: clearing WDCTL_RST failed for drive 0
 > >  > 
 > >  > ... and no disk found.
 > > 
 > > I have run into this same problem, on today's -current, with a machine
 > > that hadn't been updated in a while.
 > > 
 > > Controller is an NVIDIA MCP65 ahcisata, which already has
 > > AHCI_QUIRK_BADPMP. I tried adding AHCI_QUIRK_BADPMPRESET as well,
 > > which of course didn't do anything.
 > > 
 > > Any other suggestions?
 > 
 > Do you have the drive model from an older dmesg ?

<WDC WD5000AAJS-22YFA0>
<WDC WD1502FAEX-007BA0>

(there are two, on channels 0 and 1, neither works)

 > also, It' really "clearing WDCTL_RST failed", not "setting WDCTL_RST failed"
 > (just to be sure) ?

Yes. It is the same failure path that has the AHCI_QUIRK_BADPMPRESET
check in it.

 > there is a aprint_debug() in ahci_exec_fis() with some values that
 > should give details about what failed; can you post it ?

It will be probably this weekend before I can try another test kernel.
(I did try booting more verbosely at the time, but this caused the
ahcisata stuff to be scrolled off by a ton of crypto messages.)

 > Does this controller have CLO in its capabilities ?

ahcisata0 at pci0 dev 10 function 0: NVIDIA nForce MCP65 AHCI Controller (rev. 
0xa3)
ahcisata0: interrupting at ioapic0 pin 5
ahcisata0: 64-bit DMA
ahcisata0: AHCI revision 1.10, 4 ports, 32 slots, CAP 
0xe7229f83<CCCS,PMD,SPM,ISS=0x2=Gen2,SCLO,SAL,SALP,SSNTF,SNCQ,S64A>

so, apparently not, but this readout is from a kernel without the port
multiplier code. (If that matters.)

-- 
David A. Holland
dholland%netbsd.org@localhost