Re: kern/38035: repeatable kernel lock spinout

[Andrew Doran <ad%NetBSD.org@localhost>]

The following reply was made to PR kern/38035; it has been noted by GNATS.

From: Andrew Doran <ad%NetBSD.org@localhost>
To: Martin Husemann <martin%duskware.de@localhost>
Cc: YAMAMOTO Takashi <yamt%mwd.biglobe.ne.jp@localhost>, 
gnats-bugs%NetBSD.org@localhost
Subject: Re: kern/38035: repeatable kernel lock spinout
Date: Fri, 15 Feb 2008 17:12:23 +0000

 On Fri, Feb 15, 2008 at 06:08:47PM +0100, Martin Husemann wrote:
 
 > slightly different this time: first cpu:
 > 
 > _kernel_lock()
 > sleepq_block()
 > iic_smbus_intr_thread()
 > 
 > second cpu:
 > 
 > lapic_gettick()
 > lapic_delay(2)
 > fxp_intr()+0x162
 > intr_biglock_wrapper
 > 
 > and the interrupted lwp:
 > 
 > uvm_unmap_remove()
 > uvmspace_free()
 > exit1()
 > sys_exit()
 > syscall()
 > 
 > I still have the machine in ddb - let me know what you want to know off-list.
 
 I think this is a bug with recent changes to the fxp driver or the lapic
 code. I see it with two different etherexpress pro chipsets, on two
 different machines. With a third chipset (and a third machine!) there is
 no problem.
 
 Andrew