Re: kern/51251: ACPI device probe raises NMI

[David Young <dyoung%pobox.com@localhost>]

The following reply was made to PR kern/51251; it has been noted by GNATS.

From: David Young <dyoung%pobox.com@localhost>
To: gnats-bugs%NetBSD.org@localhost, hannken%eis.cs.tu-bs.de@localhost
Cc: 
Subject: Re: kern/51251: ACPI device probe raises NMI
Date: Sun, 3 Jul 2016 13:45:01 -0500

 On Sun, Jul 03, 2016 at 01:05:01PM +0000, J. Hannken-Illjes wrote:
 > The following reply was made to PR kern/51251; it has been noted by GNATS.
 > 
 > From: "J. Hannken-Illjes" <hannken%eis.cs.tu-bs.de@localhost>
 > To: gnats-bugs%NetBSD.org@localhost
 > Cc: 
 > Subject: Re: kern/51251: ACPI device probe raises NMI
 > Date: Sun, 3 Jul 2016 15:03:25 +0200
 > 
 >  On Sun, Jul 03, 2016 at 12:10:01AM +0000, David Young wrote:
 >  [snip]
 >  >  ISTR there is a bit you can clear in the bus bridge to stop SERR from
 >  >  being forwarded upstream?  PCI_COMMAND_SERR_ENABLE, is it?
 >  >  
 >  >  It would be interesting to see the PCI Status Register on pci0 before
 >  >  and after the probe of 000:29:0.  Likewise, I'm curious what the PCI
 >  >  Command & Status Registers on 000:29:0 contain.
 >  
 >  With my ugly patch and this
 >  
 >  	if (bus == 0 && dev == 29 && func == 0) {
 >  		pcitag_t rtag;
 >  		pcireg_t rval;
 >  
 >  		rtag = pci_make_tag(pc, bus, 0, 0);
 >  		rval = pci_conf_read(pc, rtag, PCI_COMMAND_STATUS_REG);
 >  		printf("[ - 0:0:0 %#8x", rval);
 >  		rval = pci_conf_read(pc, tag, PCI_COMMAND_STATUS_REG);
 >  		printf(" 0:29:0 %#8x ]\n", rval);
 >  	}
 >  
 >  before and after "Probe extended configuration space." I get:
 >  
 >  [ - 0:0:0 0x00900146 0:29:0 0x02800005 ]
 >  acpi0: MCFG: 000:29:0: invalid config space (cfg[0x100]=0x24d28086, alias=true)
 >  [ + 0:0:0 0x00900146 0:29:0 0x02800005 ]
 
 IIUC, extended configuration space is memory-mapped?  Looks like
 memory-mapped I/O is not enabled on 0:29:0.  Perhaps that explains
 the NMI.  Does ACPI expect for NetBSD to set the enables?  If you set
 PCI_COMMAND_MEM_ENABLE on the CSR for 0:29:0 before ACPI probes it,
 maybe that will also prevent the NMI?
 
 Dave
 
 -- 
 David Young         //\ Trestle Technology Consulting
 (217) 721-9981      Urbana, IL   http://trestle.tech/