Re: IDE device timeout when writing

[David Brownlee <abs%netbsd.org@localhost>]

On Tue, 12 Oct 2021 at 15:58, Carlos Milán Figueredo
<cmilanf%hispamsx.org@localhost> wrote:
>
> I have noticed that something odd is happening when I do disk intensive operations, like downloading or gunziping David Browlee kernel with wsfb [1]. It looks like the gunziping finish correctly, but then I get this kernel messages:
>
> [   486.702655] autoconfiguration error: wdc0:0:0: lost interrupt
> [   486.714186]         type: ata tc_bcount: 41472 tc_skip: 24064
> [   486.728095] wd0a: device timeout writing fsbn 5611343 of 5611296-5611423 (wd0 bn 7420703; cn 7361 tn 12 sn 59), xfer 1f30, retry 0
> [   499.221228] wdc0:0:0: timeout waiting for DRQ, st=0x50, err=0x00
> [   499.241306] wd0a: device timeout writing fsbn 5611296 of 5611296-5611423 (wd0 bn 7420656; cn 7361 tn 12 sn 12), xfer 1f30, retry 1
> [   509.830048] wdc0:0:0: timeout waiting for DRQ, st=0x50, err=0x00
> [   509.830048] wd0a: device timeout writing fsbn 5611296 of 5611296-5611423 (wd0 bn 7420656; cn 7361 tn 12 sn 12), xfer 1f30, retry 2
> [   509.830048] wdc0:0:0: timeout waiting for DRQ, st=0x50, err=0x00
> [   509.830048] wd0a: device timeout writing fsbn 5611296 of 5611296-5611423 (wd0 bn 7420656; cn 7361 tn 12 sn 12), xfer 1f30, retry 3
> [   512.041365] wd0: soft error (corrected) xfer 1f30
>
> The pattern repeated 3 times and it looks the kernel was able to correct the problem. I get the feeling that the system could be trying to write too fast to the device. Is there any way to limit it? Or maybe the problem could be something else? I am using Amiga 1200 built-in IDE with a CF card.

Hi Carlos,

OK, so this is happening with a kernel based on the latest netbsd-9
branch, with Jukka's wsfb changes from current.

To try to track this down would you be willing to test a couple of things?
- For the kernel which is not showing an issue, is this a WSCONS
netbsd-9 kernel that you built yourself?  - from what source did you
build it?
- Can you reproduce this issue without starting X11 on the machine
with the wsfb kernel (to eliminate some interrupt/timing issue with
X11 running)
- If it does happen outside X11 does it happen with the latest
netbsd-9 kernel from
http://nyftp.netbsd.org/pub/NetBSD-daily/netbsd-9/latest/amiga/binary/kernel/
(to eliminate the pulled up changes)
- Does it happen on the latest -current kernel
http://nyftp.netbsd.org/pub/NetBSD-daily/HEAD/latest/amiga/binary/kernel/

Thanks