Re: CVS commit: src/sys/kern

[Joerg Sonnenberger <joerg%bec.de@localhost>]

On Wed, Mar 07, 2018 at 11:10:52AM +0100, Manuel Bouyer wrote:
> On Tue, Mar 06, 2018 at 03:43:24PM +0100, Joerg Sonnenberger wrote:
> > > > This change is responsible for an instant trap of init(8) on evbppc, at
> > > > least for the TWR1025.
> > > 
> > > Do you have more details ?
> > > I guess it's because we now call pcu_state_release() with a possibly-NULL
> > > l->l_pcu_cpu[id]. 
> > > I had a quick look at the powerpc fpu code and couldn't find where this could
> > > happen.
> > 
> > Well, this is on a e500, so there is no FPU.
> 
> In this case pcu_valid should be 0, unless pcu is used for something else
> than FPU.
> What is happening exactly ? Is it a kernel panic ? do you have a stack
> trace ?

OK, so there are two different error modes depending on the specific
kernel version. For latest current, boot the install kernel and run
newfs:

# newfs ld0a
/dev/rld0a: 7443.8MB (15244992 sectors) block size 16384, fragment size
2048 using 41 cylinder groups of 181.56MB, 11620 blks, 22912 inodes.
Stopped in pid 9.1 (newfs) at   8797f0: illegal instruction 879810
db> bt
0x01304be0: at mi_switch+0x218
0x01304c20: at sleepq_block+0xd8
0x01304c50: at cv_wait+0xfc
0x01304c80: at physio_wait+0x50
0x01304ca0: at physio+0x35c
0x01304d20: at spec_write+0xd8
0x01304d90: at VOP_WRITE+0x3c
0x01304dc0: at vn_write+0x140
0x01304df0: at dofilewrite+0xbc
0x01304e60: at sys_pwrite+0xe0
0x01304ea0: at syscall+0x2e0
0x01304f20: user SC trap #174 by 0x1969628: srr1=0x2d230 r1=0xffff5e40 cr=0x24004882 xer=0 ctr=0

Joerg