Re: unifdef MODULAR for powerpc

[Rin Okuyama <rokuyama.rk%gmail.com@localhost>]

# The previous message seems to be caught by SPAM filter.
# Resend with CC field dropped.

On 2020/06/29 15:31, Rin Okuyama wrote:
> At the moment, unfortunately, something is still missing; nfs module,
> e.g., compiled for oea causes kernel panic on ibm4xx. But it should
> be worth trying to fix it!

Ugh, this is not a problem of compatibility between different
sub-architectures, but nfs module for powerpc itself; nfs modules
built for oea and ibm4xx do not work on macppc and evbppc-ibm4xx,
respectively.

If nfs is statically linked to kernel, it works of course. Also,
nfs module works fine on i386.

I've carried out somewhat exhaustive check under arch/powerpc, and
fixed several obvious bugs related to module. However, unfortunately,
it still causes panic:

panic: kernel diagnostic assertion "l->l_blcnt == 0" failed: file "../../../../kern/kern_synch.c", line 584 kernel_lock leaked
cpu0: Begin traceback...
0x00753a60: at vpanic+0x12c
0x00753a90: at kern_assert+0x60
0x00753ad0: at mi_switch+0x614
0x00753b10: at sleepq_block+0x19c
0x00753b30: at cv_timedwait+0x104
0x00753b50: at sbwait+0x54
0x00753b70: at soreceive+0xc74
0x00753c20: at nfs_request+0x1558
0x00753d80: at nfs_readdirrpc+0x244
0x00753e10: at nfs_doio+0xa04
0x00753ed0: at nfssvc_iod+0x300
0x00753f20: at cpu_lwp_bootstrap+0xc
saved LR(0x3332302f) is invalid.cpu0: End traceback...

LOCKDEBUG reveals kernel_lock, sleepq_locks, and spc_mutex are held.

I will examine further, but no ideas at the moment...

Thanks,
rin