Re: Stability issues with recent NetBSD 9.0_BETA amd64

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

On Mon, 16 Sep 2019 at 18:38, matthew green <mrg%eterna.com.au@localhost> wrote:
>
> > I've been playing with NetBSD 9.0_BETA amd64 from nyftp on a laptop for a
> > little while and noticed a significant drop in stability sometime after Aug
> > 18 (I have an Aug 18 kernel which seems solid, but Sep 2, 7 and 10 also
> > panic)
>
> i am hoping this problem is already fixed in -current and on
> netbsd-9 with this pullup:
>
>    http://mail-index.netbsd.org/source-changes/2019/09/13/msg108959.html
>
> can you at least see if the ure issue is resolved now?

The 20190915 nyftp build identifies as "NetBSD 9.0_BETA (GENERIC) #0:
Fri Sep 13 07:02:22 UTC 2019", so
I'm not certain if its picked that change up - I notice there is now a
201909170430Z build so I'll download and
switch to that this evening -- thanks.

> > Interestingly savecore always seems to fail with:
> > savecore: dump_exists:471: kvm_read newdumpmag:
> > _kvm_kvatop(ffffffff8180030c)
>
> this message is caused by some other changes i made last year,
> i need to figure out a way to make this not happen, but it's
> a little tricky -- there's a partial core dump, but it has
> issues getting basic info, but the warning itself comes from
> generic read code gets an error, and seeing those in other
> cases is very useful.

OK. would there be any sense in adding some more verbiage to suggest
to users what might have happened? ("may be partial dump")

> > A sample panic (captured from dmesg after reboot) is below. The stack trace
> > includes usbnet on a ure Ethernet interface on a USB port, but the panics
> > and firefox crash loops have occurred without this attached.
> [ .. ]
> > fatal page fault in supervisor mode
> > trap type 6 code 0 rip 0xffffffff80e5663f cs 0x8 rflags 0x10246 cr2
> > 0xffffb18137c07000 ilevel 0x5 rsp 0xffffb181343f2e58
> > curlwp 0xffff826d464394a0 pid 0.6 lowest kstack 0xffffb181343ef2c0
> > panic: trap
> > cpu0: Begin traceback...
> > vpanic() at netbsd:vpanic+0x160
> > snprintf() at netbsd:snprintf
> > startlwp() at netbsd:startlwp
> > alltraps() at netbsd:alltraps+0xc3
> > ure_rx_loop() at netbsd:ure_rx_loop+0x10f
> > usbnet_rxeof() at netbsd:usbnet_rxeof+0xb5
> > usb_transfer_complete() at netbsd:usb_transfer_complete+0x270
> > xhci_softintr() at netbsd:xhci_softintr+0x5c4
> > usb_soft_intr() at netbspiwmi0 not configured
>
> what's the crash when ure(4) isn't present?  the trace you've
> sent clearly needs ure(4) present...

I (annoyingly) don't have one to hand. I'll test the latest nyftp
build and see if I can trigger again both with and without ure. Last
panic I had managed to munch my openjdk pkgsrc install, which was an
unexpected bonus :-p

Thanks again

David