kern/55115: siisata(4) broken after enabling MSI support

[jdbaker%consolidated.net@localhost]

>Number:         55115
>Category:       kern
>Synopsis:       siisata(4) broken after enabling MSI support
>Confidential:   no
>Severity:       serious
>Priority:       high
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Sat Mar 28 06:00:00 +0000 2020
>Originator:     John D. Baker
>Release:        NetBSD/amd64-8.99.25 and later
>Organization:
>Environment:
NetBSD 8.99.25 (GENERIC_R5RS) #1: Fri Mar 27 08:07:24 CDT 2020 sysop%yggdrasil.technoskunk.fur@localhost:/r0/build/nbsd-tst/obj/amd64/sys/arch/amd64/compile/GENERIC_R5RS

>Description:
Begining with "src/sys/dev/pci/siisata_pci.c" r1.17, MSI is enabled.
While this appears to be fine for everyone else, on my siisata(4)-
equipped machine, it causes the disks attached to it to fail to be
identified/attached.

This was reviously logged in kern/54289, but is now seen to be a
different issue.

dmesg excerpt from failure:

[...]
EMACHINES W3609 (400)
  (OEM Ontel desktop board D945GCL -- JDB)
[...]
pci5 at ppb4 bus 5
pci5: i/o space, memory space enabled, rd/line, wr/inv ok
siisata0 at pci5 dev 4 function 0: CMD Technology SiI3124 SATALink (rev. 0x02)
siisata0: interrupting at msi1 vec 0
siisata0: SiI3124, 3.0Gb/s
siisata0: 32-bit 66MHz PCI
atabus0 at siisata0 channel 0
atabus1 at siisata0 channel 1
atabus2 at siisata0 channel 2
atabus3 at siisata0 channel 3
[...]
siisata0 port 0: device present, speed: 3.0Gb/s
siisata0 port 1: device present, speed: 3.0Gb/s
siisata0 port 2: device present, speed: 3.0Gb/s
siisata0 port 3: device present, speed: 3.0Gb/s
[...]
wd0 at atabus0 drive 0
wd0: <Hitachi HDS721010CLA332>
wd0: drive supports 16-sector PIO transfers, LBA48 addressing
wd0: 931 GB, 1938021 cyl, 16 head, 63 sec, 512 bytes/sect x 1953525168 sectors
[...]
wd0: autoconfiguration error: IDENTIFY failed
autoconfiguration error: wd0: unable to open device, error = 5
[...]
wd0(siisata0:0:0): using PIO mode 0
wd1 at atabus1 drive 0
wd1: autoconfiguration error: IDENTIFY failed
wd1: fixing 0 sector size
wd1: secperunit and ncylinders are zero
wd1(siisata0:1:0): using PIO mode 0
wd2 at atabus2 drive 0
wd2: autoconfiguration error: IDENTIFY failed
wd2: fixing 0 sector size
wd2: secperunit and ncylinders are zero
wd2(siisata0:2:0): using PIO mode 0
wd3 at atabus3 drive 0
wd3: autoconfiguration error: IDENTIFY failed
wd3: fixing 0 sector size
wd3: secperunit and ncylinders are zero
wd3(siisata0:3:0): using PIO mode 0
[...]
wd0: autoconfiguration error: IDENTIFY failed
opendisk: can't open dev wd0 (5)
wd0: autoconfiguration error: IDENTIFY failed
opendisk: can't open dev wd0 (5)
wd0: autoconfiguration error: IDENTIFY failed
[...]


The relevant exerpt from a success:

siisata0 at pci5 dev 4 function 0: CMD Technology SiI3124 SATALink (rev. 0x02)
siisata0: interrupting at ioapic0 pin 19
siisata0: SiI3124, 3.0Gb/s
siisata0: 32-bit 66MHz PCI
>How-To-Repeat:
Build a kernel from sources of 22-Oct-2018 21:00 UTC or later (includes
r1.17 or later of "src/sys/dev/pci/siisata_pci.c") and boot on machine
equipped with SiI3124 card.

May depend on host system, as there has been a report in kern/54289
that it works fine for at least one person.
>Fix:
Workaround:

Revert "src/sys/dev/pci/siisata_pci.c" to r1.16 and rebuild.

With this, I can finally boot netbsd-9 on my file server and have it
see the first four disks in my RAID.