port-amiga/52415: lost interrupt and device timeout messages in dmesg

[charlie%amigaspirit.hu@localhost]

>Number:         52415
>Category:       port-amiga
>Synopsis:       lost interrupt and device timeout messages in dmesg
>Confidential:   no
>Severity:       non-critical
>Priority:       medium
>Responsible:    port-amiga-maintainer
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Mon Jul 17 13:25:00 +0000 2017
>Originator:     Karoly Balogh
>Release:        7.1
>Organization:
amigaspirit.hu
>Environment:
NetBSD bsdmiga 7.1 NetBSD 7.1 (GENERIC.201703111743Z) amiga

Boot dmesg: http://dmesgd.nycbug.org/index.cgi?do=view&id=3222
>Description:
I have NetBSD 7.1 installed on my Amiga 1200 with a Blizzard 1260 CPU card. Stock GENERIC kernel.

Under heavy I/O load (example, compiling from pkgsrc) lines like this appear in dmesg:

wdc0:0:0: lost interrupt
        type: ata tc_bcount: 512 tc_skip: 15872
wd0a: device timeout writing fsbn 3748895 of 3748864-3748895 (wd0 bn 6370955; cn 25281 tn 2 sn 17), retrying
wd0: soft error (corrected)
wdc0:0:0: lost interrupt
        type: ata tc_bcount: 512 tc_skip: 5632
wd0a: device timeout writing fsbn 890667 of 890656-890667 (wd0 bn 3512727; cn 13939 tn 1 sn 36), retrying
wd0: soft error (corrected)
wdc0:0:0: lost interrupt
        type: ata tc_bcount: 512 tc_skip: 5632
wd0a: device timeout writing fsbn 897867 of 897856-897867 (wd0 bn 3519927; cn 13967 tn 3 sn 54), retrying
wd0: soft error (corrected)
wdc0:0:0: lost interrupt
        type: ata tc_bcount: 512 tc_skip: 15872
wd0a: device timeout writing fsbn 1128671 of 1128640-1128671 (wd0 bn 3750731; cn 14883 tn 3 sn 26), retrying
wd0: soft error (corrected)

The disk in question is the single disk connected to the internal IDE port of the A1200, it's an SD card on an SD2IDE44pin adapter, with a 8GB class 10 card. Using a different SD card doesn't fix the problem, the hardware doesn't report/experiences any problems with AmigaOS 3.1 at least. The Blizzard 1260 is not overclocked, runs at 50Mhz.

The machine is very stable otherwise, runs with week+ uptimes.
>How-To-Repeat:
I can provide remote SSH access to the box if necessary for debugging.
>Fix: