Re: isp driver issues in SMP on NetBSD (fwd)

On 15-Nov-04, at 1:24 PM, Matthew Jacob wrote:

>
> I believe that we can drop tech-smp from this list- I don't believe
> that this is an SMP issue. The change I suggested just ducks around
> what is apparently some issue with config_interrupts not working
> correctly for some configurations- the config_interrupts hook turned on
> using interrupts for the rest of  mailbox (housekeeping type) commands-
> like fibre channel SAN evaluation. Apparently the interrupts weren't
> actually getting through after the config_interrupts hook was executed.


Hmm.  Maybe you missed my earlier post to current-users describing how 
the isp(4)
driver fails to probe the attached devices with an SMP kernel
(http://mail-index.netbsd.org/current-users/2004/11/12/0008.html) but 
works fine
with a UP kernel. I also tried a SMP kernel with options MPACPI in 
addition to option
MPBIOS and I get an "instant" command timeout:

[Mon Nov 15 15:50:18 2004]: isp0 at pci0 dev 8 function 0: QLogic FC-AL 
and Fabric HBA
[Mon Nov 15 15:50:18 2004]: isp0: interrupting at ioapic0 pin 16 (irq 
10)
[Mon Nov 15 15:50:18 2004]: isp0: Polled Mailbox Command (0x60) Timeout
[Mon Nov 15 15:50:18 2004]: isp1 at pci0 dev 9 function 0: QLogic FC-AL 
and Fabric HBA
[Mon Nov 15 15:50:18 2004]: isp1: interrupting at ioapic0 pin 17 (irq 5)
[Mon Nov 15 15:50:18 2004]: isp1: Polled Mailbox Command (0x60) Timeout



Just to recap, with a MPBIOS only (ie no acpi), the isp driver attaches 
the scsibus
correctly..
[Sat Nov 13 03:21:40 2004]: isp0 at pci0 dev 8 function 0: QLogic FC-AL 
and Fabric HBA
[Sat Nov 13 03:21:40 2004]: isp0: interrupting at ioapic0 pin 16 (irq 
10)
[Sat Nov 13 03:21:40 2004]: scsibus1 at isp0: 256 targets, 8 luns per 
target
[Sat Nov 13 03:22:00 2004]: isp1 at pci0 dev 9 function 0: QLogic FC-AL 
and Fabric HBA
[Sat Nov 13 03:22:00 2004]: isp1: interrupting at ioapic0 pin 17 (irq 5)
[Sat Nov 13 03:22:00 2004]: scsibus2 at isp1: 256 targets, 8 luns per 
target

..and fails to probe the devices on the busses:

[Sat Nov 13 03:22:35 2004]: isp0: Interrupting Mailbox Command (0x69) 
Timeout
[Sat Nov 13 03:22:39 2004]: isp0: Mailbox Command 'GET FW STATE' failed 
(TIMEOUT)
[Sat Nov 13 03:22:39 2004]: isp1: Interrupting Mailbox Command (0x69) 
Timeout
[Sat Nov 13 03:22:39 2004]: isp1: Mailbox Command 'GET FW STATE' failed 
(TIMEOUT)


With a UP kernel the busses and devices on the busses are all found:

[Mon Nov 15 15:54:23 2004]: isp0 at pci0 dev 8 function 0: QLogic FC-AL 
and Fabric HBA
[Mon Nov 15 15:54:23 2004]: isp0: interrupting at irq 10
[Mon Nov 15 15:54:23 2004]: scsibus1 at isp0: 256 targets, 8 luns per 
target
[Mon Nov 15 15:54:43 2004]: isp1 at pci0 dev 9 function 0: QLogic FC-AL 
and Fabric HBA
[Mon Nov 15 15:54:43 2004]: isp1: interrupting at irq 5
[Mon Nov 15 15:54:43 2004]: scsibus2 at isp1: 256 targets, 8 luns per 
target
[SNIP]
[Mon Nov 15 15:55:19 2004]: sd2 at scsibus1 target 0 lun 0: <APPLE, 
Xserve RAID, 1.21> disk fixed
[Mon Nov 15 15:55:21 2004]: sd2: 1035 GB, 132522 cyl, 128 head, 128 
sec, 512 bytes/sect x 2171240448 sectors
[Mon Nov 15 15:55:21 2004]: sd3 at scsibus2 target 0 lun 0: <APPLE, 
Xserve RAID, 1.21> disk fixed
[Mon Nov 15 15:55:21 2004]: sd3: 1035 GB, 132522 cyl, 128 head, 128 
sec, 512 bytes/sect x 2171240448 sectors

(I added current-users back in, where this thread originated)
-- 
	aew