Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 31 Aug 2013 08:32:50 -0600
From:      Ian Lepore <ian@FreeBSD.org>
To:        Mattia Rossi <mattia.rossi.mate@gmail.com>
Cc:        freebsd-arm@FreeBSD.org
Subject:   Re: Alignment Fault 1 - still/again
Message-ID:  <1377959570.1111.345.camel@revolution.hippie.lan>
In-Reply-To: <5221BA13.5040309@gmail.com>
References:  <5221BA13.5040309@gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, 2013-08-31 at 11:40 +0200, Mattia Rossi wrote:
> I've already posted my alignment issues once here, where I got stuck in 
> something like this:
> 
> mvsch1: <Marvell SATA channel> at channel 1 on mvs0
> cryptosoft0: <software crypto>
> Timecounters tick every 10.000 msec
> IPsec: Initialized Security Association Processing.
> ipfw2 (+ipv6) initialized, divert enabled, nat enabled, default to 
> accept, loggd
> DUMMYNET 0 with IPv6 initialized (100409)
> load_dn_sched dn_sched FIFO loaded
> load_dn_sched dn_sched PRIO loaded
> load_dn_sched dn_sched QFQ loaded
> load_dn_sched dn_sched RR loaded
> load_dn_sched dn_sched WF2Q+ loaded
> Fatal kernel mode data abort: 'Alignment Fault 1'
> trapframe: 0xc0f1acf0
> FSR=00000001, FAR=c0f1aed4, spsr=a00000d3
> r0 =c0f05160, r1 =c3b4cc40, r2 =c0f1aebc, r3 =00000001
> r4 =c0f05160, r5 =c3b4cc40, r6 =c0ef9244, r7 =c0f0b180
> r8 =00000000, r9 =c0edf5f0, r10=00000104, r11=c0f1ad68
> r12=00000000, ssp=c0f1ad40, slr=c0ad1690, pc =c0d9d8c4
> 
> [ thread pid 0 tid 100036 ]
> Stopped at      cpu_switch+0x28:        und     0xe1c281f8
> db>
> 
> I've managed to get around this for a while by adding
> 
> options SCTP
> 
> to the kernel. But then usually got stuck at "entropy harvesting", like 
> Ian and Ronald. (See WITHOUT_ARM_EABI) thread.
> Compiling without EABI doesn't help anymore, as that only changes things 
> for userspace, where I got a Signal 11 for init instead of getting stuck 
> at entropy harvesting.
> 
> Now I've updated to r255074M and options SCTP doesn't help anymore, and 
> I'm stuck at the alignment failure above.
> So I activated witness and here's the output:
> 
> ipfw2 (+ipv6) initialized, divert enabled, nat enabled, default to 
> accept, loggd
> DUMMYNET 0 with IPv6 initialized (100409)
> load_dn_sched dn_sched FIFO loaded
> load_dn_sched dn_sched PRIO loaded
> load_dn_sched dn_sched QFQ loaded
> load_dn_sched dn_sched RR loaded
> load_dn_sched dn_sched WF2Q+ loaded
> lock order reversal:
>   1st 0xc1094e2c entropy harvest mutex (entropy harvest mutex) @ 
> /usr/devel/drea2
>   2nd 0xc3896420 uart_hwmtx (uart_hwmtx) @ 
> /usr/devel/dreamplug/sys/dev/uart/uar2
> KDB: stack backtrace:
> [backtrace removed]
> Unable to unwind further
> KDB: enter: witness_checkorder
> [ thread pid 13 tid 100012 ]
> Stopped at      kdb_enter+0x4c: ldrb    r15, [r15, r15, ror r15]!
> db>
> 
> Any hints?
> 
> Cheers,
> 
> Mat

That LOR is harmless (at least, I've been seeing it for a while).

It's not correct that EABI only changes things for userspace -- quite
the opposite, you can't run a mismatched kernel and userspace (as the
init failure points out).  

I still can't get EABI to work on my dreamplug, every time I try I get
more confusing symptoms (the latest --  trying to view a manpage gave a
"too many symbols" error trying to launch man).

-- Ian




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1377959570.1111.345.camel>