NetBSD Problem Report #9018

Received: (qmail 4026 invoked from network); 17 Dec 1999 19:59:44 -0000
Message-Id: <199912171959.LAA15991@lestat.nas.nasa.gov>
Date: Fri, 17 Dec 1999 11:59:42 -0800 (PST)
From: Jason R Thorpe <thorpej@nas.nasa.gov>
Reply-To: thorpej@nas.nasa.gov
To: gnats-bugs@gnats.netbsd.org
Subject: Qlogic ISP driver no longer works on AlphaServer 8200
X-Send-Pr-Version: 3.95

>Number:         9018
>Category:       kern
>Synopsis:       Qlogic ISP driver no longer works on AlphaServer 8200
>Confidential:   no
>Severity:       serious
>Priority:       high
>Responsible:    kern-bug-people
>State:          closed
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Fri Dec 17 12:00:01 +0000 1999
>Closed-Date:    Mon Dec 20 01:01:41 +0000 1999
>Last-Modified:  Mon Dec 20 01:23:52 +0000 1999
>Originator:     Jason R Thorpe
>Release:        Fri Dec 17 11:24:18 PST 1999
>Organization:
Numerical Aerospace Simulation Facility - NASA Ames
>Environment:


	AlphaServer 8200 5/440, 2G RAM, 2 CPUs

	(jules.nas.nasa.gov)

>Description:
	The Qlogic ISP driver does not function properly on the
	AlphaServer 8200.  Failed boot:

Jumping to entry point...

NetBSD/alpha 1.4 Secondary Bootstrap, Revision 1.10
(ross@netbsd.org, May 8 11:34:00 PDT 1999)

VMS PAL rev: 0x10000a00010114
OSF PAL rev: 0x10000500020116
Switch to OSF PAL code succeeded.

Boot file: netbsd.mp
Boot flags: S

Loading netbsd.mp...
3920128+332528 [85+251592+146715]

Entering netbsd.mp at 0xfffffc00003010f0...
[ preserving 399296 bytes of netbsd ELF symbol table ]
consinit: using prom console
Copyright (c) 1996, 1997, 1998, 1999
    The NetBSD Foundation, Inc.  All rights reserved.
Copyright (c) 1982, 1986, 1989, 1991, 1993
    The Regents of the University of California.  All rights reserved.
NetBSD 1.4P (GENERIC.MP) #74: Fri Dec 17 11:11:42 PST 1999
    thorpej@bishop:/amd/dracul/u2/netbsd/src/sys/arch/alpha/compile/GENERIC.MP
AlphaServer 8200 5/440, 437MHz
8192 byte page size, 2 processors.
total memory = 2048 MB
(2720 KB reserved for PROM, 2045 MB used by NetBSD)
avail memory = 1894 MB
using 13103 buffers containing 102 MB of memory
mainbus0 (root)
cpu0 at mainbus0: ID 8 (primary), 21164A-2 (pass 2)
cpu1 at mainbus0: ID 9, 21164A-2 (pass 2)
tlsb0 at mainbus0
tlsb0 node 4: Dual CPU, 4MB cache
gbus0 at tlsb0 node 4: Dual CPU, 4MB cache
zsc at gbus0 offset 0x10000000 not configured
zsc at gbus0 offset 0x11000000 not configured
mcclock0 at gbus0 offset 0x20000000: mc146818 or compatible
tlsbmem0 at tlsb0 node 5: MS7CC Memory Module
tlsb0 node 8: routing interrupts to cpu0
kft0 at tlsb0 node 8: KFTIA I/O interface
dwlpx0 at kft0 hose 0: PCIA rev. 0, STD I/O not present, 32K S/G entries
pci0 at dwlpx0 bus 0
pci0: i/o space, memory space enabled, rd/line, rd/mult, wr/inv ok
isp0 at pci0 dev 0 function 0
isp0: interrupting at kn8ae irq 14 vector 0x8400 PCI Interrupt Pin A
scsibus0 at isp0: 16 targets, 8 luns per target
isp1 at pci0 dev 1 function 0
isp1: interrupting at kn8ae irq 14 vector 0x8401 PCI Interrupt Pin A
scsibus1 at isp1: 16 targets, 8 luns per target
de0 at pci0 dev 2 function 0
de0: interrupting at kn8ae irq 14 vector 0x8402 PCI Interrupt Pin A
de0: DEC 21040 [10Mb/s] pass 2.4
de0: address 00:00:f8:24:ce:e1
de0: supplying EUI64: 00:00:f8:ff:fe:24:ce:e1
fpa0 at pci0 dev 3 function 0: DEC DEFPA PCI FDDI SAS Controller
fpa0: FDDI address 08:00:2b:b9:f7:5a, FW=3.10, HW=1, SMT V7.2
fpa0: FDDI Port = S (PMD = ANSI Multi-Mode)
fpa0: interrupting at kn8ae irq 14 vector 0x8403 PCI Interrupt Pin A
isp2 at pci0 dev 4 function 0
isp2: interrupting at kn8ae irq 14 vector 0x8404 PCI Interrupt Pin A
scsibus2 at isp2: 16 targets, 8 luns per target
isp3 at pci0 dev 5 function 0
isp3: interrupting at kn8ae irq 14 vector 0x8405 PCI Interrupt Pin A
scsibus3 at isp3: 16 targets, 8 luns per target
de1 at pci0 dev 6 function 0
de1: interrupting at kn8ae irq 14 vector 0x8406 PCI Interrupt Pin A
de1: DEC 21040 [10Mb/s] pass 2.4
de1: address 00:00:f8:24:9f:2f
dwlpx1 at kft0 hose 1: PCIA rev. 1, STD I/O not present, 32K S/G entries
pci1 at dwlpx1 bus 0
pci1: i/o space, memory space enabled, rd/line, rd/mult, wr/inv ok
Essential Communications RoadRunner HIPPI Interface (ethernet network) at pci1 d
ev 3 function 0 not configured
isp4 at pci1 dev 4 function 0
isp4: interrupting at kn8ae irq 14 vector 0x8414 PCI Interrupt Pin A
scsibus4 at isp4: 16 targets, 8 luns per target
ahc0 at pci1 dev 5 function 0
ahc0: interrupting at kn8ae irq 14 vector 0x8415 PCI Interrupt Pin A
ahc0: aic7880 Wide Channel, SCSI Id=7, 16 SCBs
scsibus5 at ahc0 channel 0: 16 targets, 8 luns per target
Essential Communications RoadRunner HIPPI Interface (ethernet network) at pci1 d
ev 8 function 0 not configured
tlsb0 node 7: routing interrupts to cpu0
kft1 at tlsb0 node 7: KFTHA I/O interface
dwlpx2 at kft1 hose 2: PCIA rev. 1, STD I/O not present, 32K S/G entries
pci2 at dwlpx2 bus 0
pci2: i/o space, memory space enabled, rd/line, rd/mult, wr/inv ok
de2 at pci2 dev 0 function 0
de2: interrupting at kn8ae irq 14 vector 0x8320 PCI Interrupt Pin A
de2: SMC 9332BDT 21140A [10-100Mb/s] pass 2.0
de2: address 00:00:c0:7f:a1:ef
de2: enabling 10baseT port
Essential Communications RoadRunner HIPPI Interface (ethernet network) at pci2 d
ev 4 function 0 not configured
Essential Communications RoadRunner HIPPI Interface (ethernet network) at pci2 d
ev 8 function 0 not configured
dwlpx3 at kft1 hose 3: PCIA rev. 1, STD I/O not present, 32K S/G entries
pci3 at dwlpx3 bus 0
pci3: i/o space, memory space enabled, rd/line, rd/mult, wr/inv ok
Essential Communications RoadRunner HIPPI Interface (ethernet network) at pci3 d
ev 0 function 0 not configured
Essential Communications RoadRunner HIPPI Interface (ethernet network) at pci3 d
ev 4 function 0 not configured
tlsb0 node 6: routing interrupts to cpu0
kft2 at tlsb0 node 6: KFTHA I/O interface
dwlpx0: node 8 hose 0 error interrupt
        HPC 0: ERR=0x0000c001; DMA read from Memory, Failing Address 0x0
               Illegal CSR Address Error
               Nonexistent PCI Address Error
scsibus0: waiting 2 seconds for devices to settle...
de0: enabling 10baseT port
scsibus1: waiting 2 seconds for devices to settle...
de1: autosense failed: cable problem?
scsibus2: waiting 2 seconds for devices to settle...
scsibus3: waiting 2 seconds for devices to settle...
isp3: command for target 0 lun 0 was aborted
isp3: command for target 0 lun 0 was aborted
isp3: command for target 0 lun 0 was aborted
isp3: command for target 0 lun 0 was aborted
isp3: command for target 0 lun 0 was aborted
.
.
.


A `successful' 1.4 boot looks as follows (trimmed for brevity):

isp3 at pci0 dev 5 function 0
isp3: interrupting at kn8ae irq 14 vector 0x8405 PCI Interrupt Pin A
isp3: Board Revision 1020, loaded F/W Revision 7.55
isp3: Last F/W revision was 5.57
isp3: driver initiated bus reset
scsibus3 at isp3: 16 targets, 8 luns per target
sd0 at scsibus3 targ 0 lun 0: <DEC, RZ29B    (C) DEC, 0016> SCSI2 0/direct fixed
sd0: 4091MB, 3708 cyl, 20 head, 113 sec, 512 bytes/sect x 8380080 sectors
sd1 at scsibus3 targ 1 lun 0: <DEC, RZ29B    (C) DEC, 0016> SCSI2 0/direct fixed
sd1: 4091MB, 3708 cyl, 20 head, 113 sec, 512 bytes/sect x 8380080 sectors
sd2 at scsibus3 targ 2 lun 0: <DEC, RZ29B    (C) DEC, 0016> SCSI2 0/direct fixed
sd2: 4091MB, 3708 cyl, 20 head, 113 sec, 512 bytes/sect x 8380080 sectors
cd0 at scsibus3 targ 4 lun 0: <DEC, RRD45   (C) DEC, 0436> SCSI2 5/cdrom removab
le
sd3 at scsibus3 targ 5 lun 0: <SEAGATE, ST32430N, 0250> SCSI2 0/direct fixed
sd3: 2049MB, 3992 cyl, 9 head, 116 sec, 512 bytes/sect x 4197405 sectors

>How-To-Repeat:
	See above.

>Fix:
	Not supplied.

	There was some discussion on port-alpha recently about the ISP
	driver incorrectly thinking it can do Ultra mode?  Maybe this
	is the issue.  I will rebuild the kernel with -DDEBUG to see
	what other information I can obtain.
>Release-Note:
>Audit-Trail:

From: Matthew Jacob <mjacob@feral.com>
To: Jason R Thorpe <thorpej@nas.nasa.gov>
Cc: gnats-bugs@gnats.netbsd.org
Subject: Re: kern/9018: Qlogic ISP driver no longer works on AlphaServer 8200
Date: Fri, 17 Dec 1999 14:40:43 -0800 (PST)

 > 	There was some discussion on port-alpha recently about the ISP
 > 	driver incorrectly thinking it can do Ultra mode?  Maybe this
 > 	is the issue.  I will rebuild the kernel with -DDEBUG to see
 > 	what other information I can obtain.

 This has the same symptoms, but this isn't it. The part has to be at least
 a 1040B before it'll be considered Ultra capable. 

 I've reproduced this on another 8200, so I'll do some debugging. The 8200
 has always been wierd for Qlogic support. The chips aren't on real PCI
 cards, and I've often wondered whether the 60 MHz clock setting is right.
 I also wonder really whether it's safe to negotiate wide mode since part
 of that bus isn't (and encapsulating *that* for the isp_netbsd code would
 be hell).

 At least it *does* get it right that the first three Qlogic channels are
 differential (isp3 is the SE one).

 -matt



From: Jason Thorpe <thorpej@nas.nasa.gov>
To: mjacob@feral.com
Cc: gnats-bugs@gnats.netbsd.org
Subject: Re: kern/9018: Qlogic ISP driver no longer works on AlphaServer 8200 
Date: Fri, 17 Dec 1999 14:52:05 -0800

 On Fri, 17 Dec 1999 14:40:43 -0800 (PST) 
  Matthew Jacob <mjacob@feral.com> wrote:

  > I've reproduced this on another 8200, so I'll do some debugging. The 8200
  > has always been wierd for Qlogic support. The chips aren't on real PCI
  > cards, and I've often wondered whether the 60 MHz clock setting is right.
  > I also wonder really whether it's safe to negotiate wide mode since part
  > of that bus isn't (and encapsulating *that* for the isp_netbsd code would
  > be hell).

 Yah.  Though, just because part of the bus isn't wide, it should still
 work, right?  I mean, the negotiation is supposed to happen in narrow,
 async mode...

  > At least it *does* get it right that the first three Qlogic channels are
  > differential (isp3 is the SE one).

 Yah.  BTW, how come you changed those informational messages to DEBUG-only?

 isp3 at pci0 dev 5 function 0
 isp3: interrupting at kn8ae irq 14 vector 0x8405 PCI Interrupt Pin A
 isp3: Board Revision 1020, loaded F/W Revision 7.65.0	<--- these
 isp3: Last F/W revision was 5.57.1			<--- these
 isp3: 124 max I/O commands supported			<--- these
 isp3: driver initiated bus reset of bus 0

 ...seems like those are good tidbits to have, and we have other drivers
 that display such information all the time...

         -- Jason R. Thorpe <thorpej@nas.nasa.gov>


From: Matthew Jacob <mjacob@feral.com>
To: Jason Thorpe <thorpej@nas.nasa.gov>
Cc: gnats-bugs@gnats.netbsd.org
Subject: Re: kern/9018: Qlogic ISP driver no longer works on AlphaServer 8200
Date: Fri, 17 Dec 1999 14:58:05 -0800 (PST)

 On Fri, 17 Dec 1999, Jason Thorpe wrote:

 > On Fri, 17 Dec 1999 14:40:43 -0800 (PST) 
 >  Matthew Jacob <mjacob@feral.com> wrote:
 > 
 >  > I've reproduced this on another 8200, so I'll do some debugging. The 8200
 >  > has always been wierd for Qlogic support. The chips aren't on real PCI
 >  > cards, and I've often wondered whether the 60 MHz clock setting is right.
 >  > I also wonder really whether it's safe to negotiate wide mode since part
 >  > of that bus isn't (and encapsulating *that* for the isp_netbsd code would
 >  > be hell).
 > 
 > Yah.  Though, just because part of the bus isn't wide, it should still
 > work, right?  I mean, the negotiation is supposed to happen in narrow,
 > async mode...

 That's correct, but if, for example, bad termination is provided at the
 transition point between the wide and narrow bus *AND* the f/w is as
 stupid as I sometimes think it is, there may be some floating lines
 connected to the SXP (the internal SCSI Executive processor).

 > 
 >  > At least it *does* get it right that the first three Qlogic channels are
 >  > differential (isp3 is the SE one).
 > 
 > Yah.  BTW, how come you changed those informational messages to DEBUG-only?
 > 
 > isp3 at pci0 dev 5 function 0
 > isp3: interrupting at kn8ae irq 14 vector 0x8405 PCI Interrupt Pin A
 > isp3: Board Revision 1020, loaded F/W Revision 7.65.0	<--- these
 > isp3: Last F/W revision was 5.57.1			<--- these
 > isp3: 124 max I/O commands supported			<--- these
 > isp3: driver initiated bus reset of bus 0
 > 
 > ...seems like those are good tidbits to have, and we have other drivers
 > that display such information all the time...

 Umm- I forget- somebody asked me to quiesce them. I always liked them.
 That makes it you and mr. doran to have asked for them back. Well, if I
 fix this foulup I'll put 'em back on, at least for DIAGNOSTIC.

 -matt



From: Jason Thorpe <thorpej@nas.nasa.gov>
To: mjacob@feral.com
Cc: gnats-bugs@gnats.netbsd.org
Subject: Re: kern/9018: Qlogic ISP driver no longer works on AlphaServer 8200 
Date: Fri, 17 Dec 1999 15:01:32 -0800

 On Fri, 17 Dec 1999 14:58:05 -0800 (PST) 
  Matthew Jacob <mjacob@feral.com> wrote:

  > Umm- I forget- somebody asked me to quiesce them. I always liked them.
  > That makes it you and mr. doran to have asked for them back. Well, if I
  > fix this foulup I'll put 'em back on, at least for DIAGNOSTIC.

 I say just on all the time, personally :-)

         -- Jason R. Thorpe <thorpej@nas.nasa.gov>


From: Matthew Jacob <mjacob@feral.com>
To: Jason Thorpe <thorpej@nas.nasa.gov>, Ross Harvey <ross@netbsd.org>
Cc: gnats-bugs@gnats.netbsd.org
Subject: Re: kern/9018: Qlogic ISP driver no longer works on AlphaServer 8200
Date: Sat, 18 Dec 1999 00:16:53 -0800 (PST)

 This turns out to be a f/w version issue. 7.65 F/W apparently does
 something bad with this Qlogic 1020 implementation. Configuring
 the f/w to *not* be loaded does the right thing. 1.4.X still has the 7.55
 f/w. I suppose you could count this as a case where *not* loading f/w
 (letting the SRM loaded f/w be used) is a good thing.

 I've lowered the priority to 'serious' as there's an easy temp workaround
 while I figure out the best thing to do. 

 Ross- it's pretty clear though that 1.4.2 should not go out with things as
 they are, and I haven't finished the pullups from -current yet and this is
 along, so please help me by allowing me to fix and get this stuff pulled
 up before 1.4.2 closes.

 -matt



From: Jason Thorpe <thorpej@nas.nasa.gov>
To: mjacob@feral.com
Cc: Ross Harvey <ross@netbsd.org>, gnats-bugs@gnats.netbsd.org
Subject: Re: kern/9018: Qlogic ISP driver no longer works on AlphaServer 8200 
Date: Sat, 18 Dec 1999 09:38:44 -0800

 On Sat, 18 Dec 1999 00:16:53 -0800 (PST) 
  Matthew Jacob <mjacob@feral.com> wrote:

  > This turns out to be a f/w version issue. 7.65 F/W apparently does
  > something bad with this Qlogic 1020 implementation. Configuring
  > the f/w to *not* be loaded does the right thing. 1.4.X still has the 7.55
  > f/w. I suppose you could count this as a case where *not* loading f/w
  > (letting the SRM loaded f/w be used) is a good thing.

 Oh dear... well, maybe the firmware for just the 1020 can be reverted
 for 7.55?

         -- Jason R. Thorpe <thorpej@nas.nasa.gov>


From: Matthew Jacob <mjacob@feral.com>
To: Jason Thorpe <thorpej@nas.nasa.gov>
Cc: Ross Harvey <ross@netbsd.org>, gnats-bugs@gnats.netbsd.org
Subject: Re: kern/9018: Qlogic ISP driver no longer works on AlphaServer 8200
Date: Sat, 18 Dec 1999 10:13:27 -0800 (PST)

 > On Sat, 18 Dec 1999 00:16:53 -0800 (PST) 
 >  Matthew Jacob <mjacob@feral.com> wrote:
 > 
 >  > This turns out to be a f/w version issue. 7.65 F/W apparently does
 >  > something bad with this Qlogic 1020 implementation. Configuring
 >  > the f/w to *not* be loaded does the right thing. 1.4.X still has the 7.55
 >  > f/w. I suppose you could count this as a case where *not* loading f/w
 >  > (letting the SRM loaded f/w be used) is a good thing.
 > 
 > Oh dear... well, maybe the firmware for just the 1020 can be reverted
 > for 7.55?

 There are 4 options I'm pursuing:

 a) Does > 7.65 f/w fixing this problem

 b) Going to/checking Initiator-only f/w. There's reasons to do this if
 TARGET_MODE isn't supported anyway, and you get a helluva lot more SRAM
 for commands on the vanilla 1020/1040 chips anyway. Maybe this fixes it.

 c) Raising this with Qlogic as a bug to fix. I doubt they will. The
 1020/1040 is effectively an EOL'd product- they're concentrating on (and
 rightly so) FC and the newer (1080/1280/12160) Ultra2/Ultra3 chipsets.

 d) Special casing/encapsulating somehow this 'onboard' 1020 version. I
 have a 1020 that works fine with 7.65 f/w, so it isn't all 1020s- just
 probably some circuit layout issues on the 8200 KFTIA that the newer f/w
 didn't manage. It's all very vague- kinda annoying because if I were back
 working for a real company on this, resources would be expended on really
 tracking it down. As it is, we have to guess a lot. I have a guess in that
 I believe that the 7.55 f/w thinks that it's connected to a differential
 bus- there's some SXP register settings that change based on this (about
 how you tristate some pins)- it's possible I could fool around with those,
 but frankly, I think that I should just special case this in in isp_pci as
 (bad, I know):

 #ifdef	__alpha__
 	if (chip_type < ISP_HA_SCSI_1040)
 		isp->isp_confopts |= ISP_CFG_NORELOAD;
 #endif


 On a related note: I'll check in a fix today to isp_pci/isp_sbus that
 restores snagging config 'flags', so doing

 isp*    at      pci? dev ? function ? flags 0x80

 will work again.


 On a less related note: I've been using mother.nas.nasa.gov with Ross's
 1204 snapshot and the fixed kernel. Things work well, but it still does
 hald with the 'kernel stack not valid' halt at varying points. If Harry
 approves the time, I'll try chase this down in the next week, or at least
 conclude that it can't be a DWLPX/BUSDMA problem.


 -matt



From: Matthew Jacob <mjacob@feral.com>
To: Jason Thorpe <thorpej@nas.nasa.gov>
Cc: Ross Harvey <ross@netbsd.org>, gnats-bugs@gnats.netbsd.org
Subject: Re: kern/9018: Qlogic ISP driver no longer works on AlphaServer 8200
Date: Sat, 18 Dec 1999 15:06:38 -0800 (PST)

 > b) Going to/checking Initiator-only f/w. There's reasons to do this if
 > TARGET_MODE isn't supported anyway, and you get a helluva lot more SRAM
 > for commands on the vanilla 1020/1040 chips anyway. Maybe this fixes it.

 4.65 seems to not have this problem set. I have to do some provisioning
 now in the core driver to not base features purely on a linear version
 tuple/hba type, but instead on a generated feature. When that is done,
 then I'll put back 4.65 as the default f/w level and note that 7.65 is a
 problem for Alpha 8200s when TARGET_MODE is enabled.

 -matt


State-Changed-From-To: open->closed 
State-Changed-By: mjacob 
State-Changed-When: Sun Dec 19 17:01:41 PST 1999 
State-Changed-Why:  
Problem solved by defaulting to 4.65 (initiator mode only) f/w. 
>Unformatted:

NetBSD Home
NetBSD PR Database Search

(Contact us) $NetBSD: query-full-pr,v 1.51 2026/08/10 02:28:17 riastradh Exp $
$NetBSD: gnats_config.sh,v 1.10 2026/05/13 22:00:09 riastradh Exp $
Copyright © 1994-2026 The NetBSD Foundation, Inc. ALL RIGHTS RESERVED.