Re: port-alpha/40604

[Hans Rosenfeld <rosenfeld%grumpf.hope-2000.org@localhost>]

The following reply was made to PR port-alpha/40604; it has been noted by GNATS.

From: Hans Rosenfeld <rosenfeld%grumpf.hope-2000.org@localhost>
To: gnats-bugs%netbsd.org@localhost
Cc: 
Subject: Re: port-alpha/40604
Date: Thu, 7 Oct 2010 22:01:18 +0200

 I just committed code to enable and decode the peripheral machine checks
 on this kind of systems. This clearly shows that this not bad memory or
 anything else:
 
 ...
 mlx0 at pci0 dev 9 function 0: Mylex RAID (v2 interface)
 mlx0: interrupting at dec 6600 irq 23
 sgmap_load: ----- buf = 0xfffffe0000031300 -----
 sgmap_load: dmaoffset = 0x1300, buflen = 0x80
 sgmap_load: va:endva = 0xfffffe0000030000:0xfffffe0000032000
 sgmap_load: sgvalen = 0x2000, boundary = 0x0
 sgmap_load: sgva = 0x0, pteidx = 0, pte = 0xfffffc00002d0000 (pt = 
0xfffffc00002d0000)
 sgmap_load: wbase = 0x800000, vpage = 0x0, DMA addr = 0x801300
 sgmap_load:     pa = 0xfffa6000, pte = 0xfffffc00002d0000, *pte = 0xfffa7
 
 System Machine Check (660): Rev 0x1, Code 0x202, Flags 0x0
 
     Software Error Summary Flags   = 0x0000000000000001
     CPU Device Interrupt Requests  = 0x4000000000000000
         DIR = 0x4000000000000000<Pchip 0 error>
     Cchip Miscellaneous Register   = 0x0000000100000030
     Pchip 0 Error Register         = 0x0070fffa73700041
         error    = 0x41<Error lost,Target abort>
         address  = 0xfffa7370, 0x0<No DAC>
         command  = 0x7<PCI memory write>
     Pchip 1 Error Register         = 0x0000000000000000
 
 unexpected machine check:
 
     mces    = 0x1
     vector  = 0x660
     param   = 0xfffffc0000006080
     pc      = 0xfffffc00009db220
     ra      = 0xfffffc00009db1e8
     code    = 0x100000202
     curlwp = 0xfffffc0001234a60
         pid = 0.1, comm = system
 
 panic: machine check
 ...
 
 This only happens when using SG DMA, which is done for address
 everything > 1GB. I have no idea why it is failing, the code looks ok.
 Maybe I get something wrong here, but why is the Pchip doing a PCI
 memory write when it should be doing a DMA transfer to host memory?
 
 
 Hans
 
 
 -- 
 %SYSTEM-F-ANARCHISM, The operating system has been overthrown