Re: CURRENT broken on Raspberry Pi 2?

["Herbert J. Skuhra" <herbert%gojira.at@localhost>]

On Tue, 26 Feb 2019 15:00:36 +0100, Nick Hudson wrote:
> On 26/02/2019 07:00, Herbert J. Skuhra wrote:
> > On Tue, Feb 26, 2019 at 12:19:34PM +0900, Masanobu SAITOH wrote:
> >> Hi.
> >> 
> >> On 2019/02/26 10:07, Herbert J. Skuhra wrote:
> >>> On Mon, 25 Feb 2019 15:09:21 +0100, Nick Hudson wrote:
> >>>> 
> >>>> On 24/02/2019 22:26, Herbert J. Skuhra wrote:
> >>>>> On Thu, 14 Feb 2019 20:26:15 +0100, "Herbert J. Skuhra" wrote:
> >>>>>> 
> >>>>>> On Sun, 10 Feb 2019 16:44:33 +0100, Nick Hudson wrote:
> >>>>>>> RapsberryPI 2 is earmv7hf
> >>>>>> 
> >>>>>> The following happens when I boot an earmv7hf image on my Raspberry Pi 2:
> >>>>>> 
> >>>>>> Starting syslogd.
> >>>>>> [  33.4972146] Fatal kernel mode prefetch abort at 0x00000000
> >>>>>> [  33.5120546] trapframe: 0xba9cdef0, spsr=a0090013
> >>>>>> [  33.5120546] r0 =5dde7d88, r1 =f2a00e00, r2 =ba9cdfb0, r3 =00000010
> >>>>>> [  33.5228162] r4 =807dfd1c, r5 =00000000, r6 =ba9cdfb0, r7 =5dde7d88
> >>>>>> [  33.5228162] r8 =00000010, r9 =f2a00e00, r10=baaa79a0, r11=ba9cdfac
> >>>>>> [  33.5351457] r12=ba9cdf40, ssp=ba9cdf40, slr=80024114, pc =00000000
> >>>>>> 
> >>>>>> Stopped in pid 316.1 (syslogd) at       0:      address 0x0 is invalid
> >>>>>> andeq   r0, r0, r0
> >>>>>> db{0}> bt
> >>>>>> 0xba9cdfac: netbsd:undefinedinstruction+0xc
> >>>>>> db{0}> show mbuf
> >>>>>> MBUF(0) 0x0
> >>>>>> 
> >>>>>> [  33.5351457] uvm_fault(0xbafbeca8, 0, 1) -> e
> >>>>>> [  33.5351457] Fatal kernel mode data abort: 'Translation Fault (S)'
> >>>>>> [  33.5351457] trapframe: 0xba9cd9f0
> >>>>>> [  33.5351457] FSR=00000005, FAR=00000014, spsr=60090113
> >>>>>> [  33.5351457] r0 =0000000c, r1 =f3201000, r2 =00000000, r3 =00000000
> >>>>>> [  33.5351457] r4 =00000000, r5 =00000000, r6 =80374f5c, r7 =00000000
> >>>>>> [  33.5351457] r8 =8060c10c, r9 =ba9cdef0, r10=ba9cda70, r11=ba9cdc9c
> >>>>>> [  33.5351457] r12=ba9cd8e0, ssp=ba9cda40, slr=80374c98, pc =8039fa20
> >>>>> 
> >>>>> Obviously a commit between 8.99.29 and 8.99.30 must be causing this
> >>>>> issue:
> >>>>> 
> >>>>> 8.99.29 2018-12-24 16:58:54 GMT => OK
> >>>>> 8.99.30 2019-01-01 14:01:46 GMT => Not OK
> >>>> 
> >>>> Can you narrow it down further?
> >>> 
> >>> +---
> >>> |Module Name:    src
> >>> |Committed By:   msaitoh
> >>> |Date:           Sun Dec 30 06:33:30 UTC 2018
> >>> |
> >>> |Modified Files:
> >>> |        src/sys/dev/mii: makphy.c
> >>> |
> >>> |Log Message:
> >>> |- Control BMCR_PDOWN for IFM_NONE. Some chips still don't work as expected.
> >>> |  It would be required to modify PSCR and/or other register.
> >>> |- Set mii_media_active correctly on non-autonego mode.
> >>> |
> >>> |
> >>> |To generate a diff of this commit:
> >>> |cvs rdiff -u -r1.46 -r1.47 src/sys/dev/mii/makphy.c
> >>> +---
> >>> 
> >>> No crash if I disable "makphy*" line in sys/arch/evbarm/conf/RPI.
> >>> I'll test 8.99.34 later today.
> >> 
> >>   IMHO, this change doesn't cause such type of panic...
> >> 
> >> Does RPI2 use makphy?
> > 
> > dmesg shows only ukphy.
> > 
> > Current from today works if I disable makphy* in RPI config. The
> > daily evbarm-earmv7hf snapshot from nycdn.netbsd.org (e.g:
> > 201902240830Z) panics.
> > 
> 
> Can you provide the netbsd.map from each kernel build, please?

Sure.

1. https://bsd4all.net/tmp/NetBSD/netbsd_1.map.gz

CVS date: 2018-12-30 05:05:30 GMT
http://mail-index.netbsd.org/source-changes/2018/12/30/msg101863.html

Result: OK

2. https://bsd4all.net/tmp/NetBSD/netbsd_2.map.gz

CVS date: 2018-12-30 06:34:00 GMT
http://mail-index.netbsd.org/source-changes/2018/12/30/msg101868.html

Result: Panic

3. https://bsd4all.net/tmp/NetBSD/netbsd_3.map.gz

CVS date same as in #2 but makphy line disabled in RPI kernel config.

Result: OK

--
Herbert