Re: dmesg and GIANT-LOCK

2007-03-13 Thread Kris Kennaway
On Tue, Mar 13, 2007 at 12:12:26PM -0700, James Long wrote:
> On Tue, Mar 13, 2007 at 01:50:34PM -0400, Kris Kennaway wrote:
> > On Tue, Mar 13, 2007 at 01:06:12AM -0700, James Long wrote:
> > > With regard to the recent thread about looking for GIANT-LOCKs in
> > > dmesg, why would one system say:
> > > 
> > > ns : 00:56:29 /home/james> uname -v ;dmesg | grep fxp
> > > FreeBSD 6.2-STABLE #0: Tue Feb 20 15:47:09 PST 2007
> > > fxp0:  port 0x2400-0x243f mem 
> > > 0xc4fff000-0xc4ff,0xc4e0-0xc4ef irq 10 at device 2.0 on pci0
> > > miibus0:  on fxp0
> > > fxp0: Ethernet address: 00:02:a5:0a:57:73
> > > 
> > > 
> > > while a more recent build says:
> > > 
> > > t30 : 00:56:19 /home/james> uname -v ;dmesg | grep fxp
> > > FreeBSD 6.2-STABLE #2: Thu Mar  8 08:23:11 PST 2007
> > > fxp0:  port 0x7400-0x743f mem 
> > > 0xd020-0xd0200fff irq 11 at device 8.0 on pci2
> > > miibus0:  on fxp0
> > > fxp0: Ethernet address: 00:09:6b:86:82:a6
> > > fxp0: [GIANT-LOCKED]
> > 
> > fxp is not giant locked, you can check the source for the INTR_MPSAFE
> > flag in sys/dev/fxp/if_fxp.c.  I'm not sure how you are seeing this,
> > please describe the configuration of this system further (kernel
> > config, loader.conf).
> > 
> > Kris
> 
> It just dawned on me when you said "INTR_MPSAFE", would having
> 
> options   IPSEC
> 
> in the kernel config cause fxp to use GIANT?
> 
> dmesg says in part:
> 
> FreeBSD 6.2-STABLE #2: Thu Mar  8 08:23:11 PST 2007
> [EMAIL PROTECTED]:/usr/obj/usr/src/sys/T30
> WARNING: debug.mpsafenet forced to 0 as ipsec requires Giant
> WARNING: MPSAFE network stack disabled, expect reduced performance.

Yes.  Use FAST_IPSEC instead, it's also faster in other ways than just
having better SMP scaling properties.

Kris
___
freebsd-questions@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-questions
To unsubscribe, send any mail to "[EMAIL PROTECTED]"


Re: dmesg and GIANT-LOCK

2007-03-13 Thread James Long
On Tue, Mar 13, 2007 at 01:50:34PM -0400, Kris Kennaway wrote:
> On Tue, Mar 13, 2007 at 01:06:12AM -0700, James Long wrote:
> > With regard to the recent thread about looking for GIANT-LOCKs in
> > dmesg, why would one system say:
> > 
> > ns : 00:56:29 /home/james> uname -v ;dmesg | grep fxp
> > FreeBSD 6.2-STABLE #0: Tue Feb 20 15:47:09 PST 2007
> > fxp0:  port 0x2400-0x243f mem 
> > 0xc4fff000-0xc4ff,0xc4e0-0xc4ef irq 10 at device 2.0 on pci0
> > miibus0:  on fxp0
> > fxp0: Ethernet address: 00:02:a5:0a:57:73
> > 
> > 
> > while a more recent build says:
> > 
> > t30 : 00:56:19 /home/james> uname -v ;dmesg | grep fxp
> > FreeBSD 6.2-STABLE #2: Thu Mar  8 08:23:11 PST 2007
> > fxp0:  port 0x7400-0x743f mem 
> > 0xd020-0xd0200fff irq 11 at device 8.0 on pci2
> > miibus0:  on fxp0
> > fxp0: Ethernet address: 00:09:6b:86:82:a6
> > fxp0: [GIANT-LOCKED]
> 
> fxp is not giant locked, you can check the source for the INTR_MPSAFE
> flag in sys/dev/fxp/if_fxp.c.  I'm not sure how you are seeing this,
> please describe the configuration of this system further (kernel
> config, loader.conf).
> 
> Kris

It just dawned on me when you said "INTR_MPSAFE", would having

options IPSEC

in the kernel config cause fxp to use GIANT?

dmesg says in part:

FreeBSD 6.2-STABLE #2: Thu Mar  8 08:23:11 PST 2007
[EMAIL PROTECTED]:/usr/obj/usr/src/sys/T30
WARNING: debug.mpsafenet forced to 0 as ipsec requires Giant
WARNING: MPSAFE network stack disabled, expect reduced performance.

___
freebsd-questions@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-questions
To unsubscribe, send any mail to "[EMAIL PROTECTED]"


Re: dmesg and GIANT-LOCK

2007-03-13 Thread Kris Kennaway
On Tue, Mar 13, 2007 at 01:06:12AM -0700, James Long wrote:
> With regard to the recent thread about looking for GIANT-LOCKs in
> dmesg, why would one system say:
> 
> ns : 00:56:29 /home/james> uname -v ;dmesg | grep fxp
> FreeBSD 6.2-STABLE #0: Tue Feb 20 15:47:09 PST 2007
> fxp0:  port 0x2400-0x243f mem 
> 0xc4fff000-0xc4ff,0xc4e0-0xc4ef irq 10 at device 2.0 on pci0
> miibus0:  on fxp0
> fxp0: Ethernet address: 00:02:a5:0a:57:73
> 
> 
> while a more recent build says:
> 
> t30 : 00:56:19 /home/james> uname -v ;dmesg | grep fxp
> FreeBSD 6.2-STABLE #2: Thu Mar  8 08:23:11 PST 2007
> fxp0:  port 0x7400-0x743f mem 
> 0xd020-0xd0200fff irq 11 at device 8.0 on pci2
> miibus0:  on fxp0
> fxp0: Ethernet address: 00:09:6b:86:82:a6
> fxp0: [GIANT-LOCKED]

fxp is not giant locked, you can check the source for the INTR_MPSAFE
flag in sys/dev/fxp/if_fxp.c.  I'm not sure how you are seeing this,
please describe the configuration of this system further (kernel
config, loader.conf).

Kris


pgpkzr1YXlNlZ.pgp
Description: PGP signature


Re: dmesg and GIANT-LOCK

2007-03-13 Thread Ivan Voras
James Long wrote:

> Is it just the difference in chipset/controller type that requires
> the fxp driver to use GIANT on the second machine, but not the first? 
> I also note that on the first machine, irq 10 is solely assigned to
> fxp0.  On t30, irq 11 is shared with a number of other devices.  Dunno
> if that matters.

It's probably only the matter of output, this has changed with time to
start visibly marking which devices need GIANT.

Look for NEEDSGIANT flag in ifconfig output.

___
freebsd-questions@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-questions
To unsubscribe, send any mail to "[EMAIL PROTECTED]"


dmesg and GIANT-LOCK

2007-03-13 Thread James Long
With regard to the recent thread about looking for GIANT-LOCKs in
dmesg, why would one system say:

ns : 00:56:29 /home/james> uname -v ;dmesg | grep fxp
FreeBSD 6.2-STABLE #0: Tue Feb 20 15:47:09 PST 2007
fxp0:  port 0x2400-0x243f mem 
0xc4fff000-0xc4ff,0xc4e0-0xc4ef irq 10 at device 2.0 on pci0
miibus0:  on fxp0
fxp0: Ethernet address: 00:02:a5:0a:57:73


while a more recent build says:

t30 : 00:56:19 /home/james> uname -v ;dmesg | grep fxp
FreeBSD 6.2-STABLE #2: Thu Mar  8 08:23:11 PST 2007
fxp0:  port 0x7400-0x743f mem 
0xd020-0xd0200fff irq 11 at device 8.0 on pci2
miibus0:  on fxp0
fxp0: Ethernet address: 00:09:6b:86:82:a6
fxp0: [GIANT-LOCKED]



Is it just the difference in chipset/controller type that requires
the fxp driver to use GIANT on the second machine, but not the first? 
I also note that on the first machine, irq 10 is solely assigned to
fxp0.  On t30, irq 11 is shared with a number of other devices.  Dunno
if that matters.


Thanks!
___
freebsd-questions@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-questions
To unsubscribe, send any mail to "[EMAIL PROTECTED]"