Re: puc(4) device not attaching issue in -current: WORKING

(saved for context, see comments below).

John Baldwin wrote:
> On Monday 29 March 2004 10:39 am, othermark wrote:
>> John Baldwin wrote:
>> > On Tuesday 23 March 2004 05:58 pm, othermark wrote:
>> >> Doug White wrote:
>> >> > On Tue, 23 Mar 2004, othermark wrote:
>> >> >> none0_at_pci0:6:1: class=0x070002 card=0x00000000 chip=0x01811407
>> >> >> rev=0x00 hdr=0x00
>> >> >>     vendor   = 'Lava Computer Manufacturing Inc'
>> >> >>     device   = 'Lava Octopus PCI Ports 5-8'
>> >> >>     class    = simple comms
>> >> >>     subclass = UART
>> >> >>
>> >> >> device 6.1 has chip 0x018111407=and pucdata.c has only has an entry
>> >> >> for 0x0180, so I'm not sure how it was picking up 0x0181 in earlier
>> >> >> releases, or that it still does and something else is blocking it
>> >> >> from happening...
>> >> >
>> >> > What happens if you add that PCI ID to the driver?
>> >>
>> >> If I add a duplicate the entry in pucdata.c so it looks like this:
>> >>
>> >>
>> >>         /* Lava Computers Octopus-550 serial ports */
>> >>         {   "Lava Computers Octopus-550 8-port serial part B",
>> >>             NULL,
>> >>             {   0x1407, 0x0181, 0,      0       },
>> >>             {   0xffff, 0xfffc, 0,      0       },
>> >
>> > The 0xfffc is a mask meaning that this matches 0x0180, 0x0181, 0x0182,
>> > and 0x0183 by the way, so you don't need an extra entry.  Can you do a
>> > boot -v? I'm curious if the second device has its BAR's setup.
>>
>> I've attached a verbose boot to this message.   How do you tell if the
>> BAR is setup or not?
> 
> The map lines for the dump of PCI config registers.  Here is your working
> device:
> 
> $PIR: 0:5 INTA routed to irq 5
> found->       vendor=0x1407, dev=0x0180, revid=0x00
> bus=0, slot=5, func=0
> class=07-00-02, hdrtype=0x00, mfdev=1
> cmdreg=0x0001, statreg=0x0480, cachelnsz=0 (dwords)
> lattimer=0x00 (0 ns), mingnt=0x00 (0 ns), maxlat=0x00 (0 ns)
> intpin=a, irq=5
> map[10]: type 4, range 32, base 0000f888, size  3, enabled
> map[14]: type 4, range 32, base 0000f890, size  3, enabled
> map[18]: type 4, range 32, base 0000f898, size  3, enabled
> map[1c]: type 4, range 32, base 0000f8a0, size  3, enabled
> 
> And below it is the one that isn't working:
> 
> $PIR: 0:5 INTA routed to irq 5
> found->       vendor=0x1407, dev=0x0181, revid=0x00
> bus=0, slot=5, func=1
> class=07-00-02, hdrtype=0x00, mfdev=1
> cmdreg=0x0001, statreg=0x0480, cachelnsz=0 (dwords)
> lattimer=0x00 (0 ns), mingnt=0x00 (0 ns), maxlat=0x00 (0 ns)
> intpin=a, irq=5
> map[10]: type 1, range 32, base feaff000, size 12, enabled
> map[14]: type 4, range 32, base 0000fc40, size  6, enabled
> map[18]: type 1, range 32, base fe700000, size 20, enabled
> 
> Note that the working one has 4 BARs (10, 14, 18, and 1c) that are all I/O
> ports (type 4) and 8 bytes long (2 ^ size).  The second one only has 3
> bars (10, 14, and 18), of which only the second is an I/O port (the other
> two are
> memory) and which have very different sizes.  There must be some other way
> to talk to the other 4 ports or perhaps the board needs to be configured
> somehow in order to work.

I believe the recent PCI Heads UP commit fixed my other 4 BARs since it now
detects it and attaches properly.  I can now get FASTINTR to work with APIC
on this card too.  Unfortunately APIC breaks my USB mouse, so I have to
revert to using it as PS/2.

-- 
othermark
atkin901 at nospam dot yahoo dot com
(!wired)?(coffee++):(wired);