Q: kernel log not serialized (SLES10 SP3)?

[Ulrich Windl]

Hi!

I just found some "interesting" kernel boot messages from a two Dual-Core Opteron machine (SunFire X4100) running SLES10 SP3 (2.6.16.60-0.54.5-xen):
<6>Security Framework v1.0.0 initialized
<4>Mount-cache hash table entries: 256
<6>CPU: L1 I Cache: 64K (64 bytes/line), D cache 64K (64 bytes/line)
<6>CPU: L2 Cache: 1024K (64 bytes/line)
<6>checking if image is initramfs... it is
<4>Freeing initrd memory: 8553k freed
<4>Initializing CPU#1
<4>Initializing CPU#2
<6>CPU: L1 I Cache: 64K (64 bytes/line), D cache 64K (64 bytes/line)
<6>CPU: L2 Cache: 1024K (64 bytes/line)
<6>Brought up 4 CPUs
<4>Initializing CPU#3
<6>CPU: L1 I Cache: 64K (64 bytes/line), D cache 64K (64 bytes/line)
<6>CPU: L2 Cache: 1024K (64 bytes/line)
<6>CPU: L1 I Cache: 64K (64 bytes/line), D cache 64K (64 bytes/line)
<6>CPU: L2 Cache: 1024K (64 bytes/line)
<4>migration_cost=523
<6>NET: Registered protocol family 16

The entries seem to be out of order, specifically initializing a CPU after "Brought up 4 CPUs". If the race is intended that way, I'd add the CPU number to the lines that print cache information.

And on "migration cost": It seems to be a penalty for a vector leading from on CPU to another, most likely in a symmetrical fashion. On a NUMA architecture like the Opteron the CPUs could be added to the message. Sample:
CPU1-CPU2: migration cost=523

Regards,
Ulrich
P.S.:Please keep CC: as I'm not subscribed.


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/