sparc64/151404: Kernel hangs before mounting root fs

[Marius Strobl]

The following reply was made to PR sparc64/151404; it has been noted by GNATS.

From: Marius Strobl <marius at alchemy.franken.de>
To: Pyun YongHyeon <pyunyh at gmail.com>
Cc: Dmitry Afanasiev <KOT at MATPOCKuH.Ru>, yongari at freebsd.org,
        bug-followup at freebsd.org
Subject: Re: sparc64/151404: Kernel hangs before mounting root fs
Date: Sat, 16 Oct 2010 13:30:13 +0200

 On Fri, Oct 15, 2010 at 05:09:24PM -0700, Pyun YongHyeon wrote:
 > On Sat, Oct 16, 2010 at 01:25:17AM +0200, Marius Strobl wrote:
 > > 
 > > No, this most likely is fallout from the recent changes to bge(4), I've
 > > already got bitten by two regressions they caused and apparently there's
 > > at least a third one. As of r213890 I can confirm that the interfaces get
 > > no carrier when booting the kernel from disk, however when booting via
 > > net at least that interface gets up. Yongari, could you please look into
 > > this? The interfaces in question are:
 > > bge0: <Broadcom BCM5715 A3, ASIC rev. 0x009003> mem 0x4200000-0x420ffff,0x4010000-0x401ffff at device 4.0 on pci7
 > > bge0: CHIP ID 0x00009003; ASIC REV 0x09; CHIP REV 0x90; PCI-X 0.0 on pci0
 > > miibus0: <MII bus> on bge0
 > > brgphy0: <BCM5714 10/100/1000baseTX PHY> PHY 1 on miibus0
 > > brgphy0:  10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, 1000baseT, 1000baseT-FDX, auto
 > > bge0: Ethernet address: 00:14:4f:c3:df:7a
 > >  
 > > bge0 at pci0:9:4:0:        class=0x020000 card=0x1668108e chip=0x166814e4 rev=0xa3hdr=0x00
 > >     vendor     = 'Broadcom Corporation'
 > >     device     = 'NetXtreme BCM5714 Gigabit Ethernet'
 > >     class      = network
 > >     subclass   = ethernet
 > >     cap 07[40] = PCI-X 64-bit supports 133MHz, 2048 burst read, 1 split transaction
 > >     cap 01[48] = powerspec 2  supports D0 D3  current D0
 > >     cap 03[50] = VPD
 > >     cap 05[58] = MSI supports 8 messages, 64 bit
 > > 
 > > In case it matters they are using MSIs.
 > > 
 > 
 > It seems BCM5714 also requires some special handling in PHY.
 > auto-polling seemed to hide the it. Would you try attached patch?
 > 
 > > Marius
 > > 
 
 > Index: sys/dev/bge/if_bge.c
 > ===================================================================
 > --- sys/dev/bge/if_bge.c	(revision 213900)
 > +++ sys/dev/bge/if_bge.c	(working copy)
 > @@ -2677,8 +2677,10 @@
 >  		sc->bge_mi_mode = BGE_MIMODE_500KHZ_CONST;
 >  	else
 >  		sc->bge_mi_mode = BGE_MIMODE_BASE;
 > -	/* Enable auto polling for BCM570[0-5]. */
 > -	if (BGE_IS_5700_FAMILY(sc) || sc->bge_asicrev == BGE_ASICREV_BCM5705)
 > +	/* Enable auto polling for BCM570[0-5], BCM5714. */
 > +	if (BGE_IS_5700_FAMILY(sc) ||
 > +	    sc->bge_asicrev == BGE_ASICREV_BCM5705 ||
 > +	    sc->bge_asicrev == BGE_ASICREV_BCM5714)
 >  		sc->bge_mi_mode |= BGE_MIMODE_AUTOPOLL;
 >  
 >  	/*
 
 Unfortunately, the above patch makes no difference to the problem
 (the netboot interface works, but otherwise the link doesn't come
 up). I guess it's time to dig out all of my bge(4) controllers :)
 
 Marius