]>
Commit | Line | Data |
---|---|---|
00e5a55c BS |
1 | Date: Tue, 6 Jan 2009 10:27:41 -0600 |
2 | From: Dimitri Sivanich <sivanich@sgi.com> | |
3 | To: linux-ia64@vger.kernel.org, Tony Luck <tony.luck@intel.com>, | |
4 | Greg KH <greg@kroah.com> | |
5 | Cc: linux-kernel@vger.kernel.org, | |
6 | Peter Zijlstra <peterz@infradead.org>, | |
7 | Gregory Haskins <ghaskins@novell.com>, Nick Piggin <npiggin@suse.de>, | |
8 | Tony Luck <tony.luck@gmail.com>, Robin Holt <holt@sgi.com> | |
9 | Subject: configure HAVE_UNSTABLE_SCHED_CLOCK for SGI_SN systems | |
10 | ||
11 | Turn on CONFIG_HAVE_UNSTABLE_SCHED_CLOCK for SGI_SN. | |
12 | ||
13 | SGI Altix has unsynchronized itc clocks. This results in rq->clock | |
14 | occasionally being set to a time in the past by a remote cpu. | |
15 | ||
16 | Note that it is possible that this problem may exist for other ia64 | |
17 | machines as well, based on the following comment for sched_clock() in | |
18 | arch/ia64/kernel/head.S: | |
19 | ||
20 | * Return a CPU-local timestamp in nano-seconds. This timestamp is | |
21 | * NOT synchronized across CPUs its return value must never be | |
22 | * compared against the values returned on another CPU. The usage in | |
23 | * kernel/sched.c ensures that. | |
24 | ||
25 | ||
26 | Signed-off-by: Dimitri Sivanich <sivanich@sgi.com> | |
27 | Signed-off-by: Gregory Haskins <ghaskins@novell.com> | |
28 | ||
29 | --- | |
30 | ||
31 | arch/ia64/Kconfig | 1 + | |
32 | 1 file changed, 1 insertion(+) | |
33 | ||
34 | --- a/arch/ia64/Kconfig | |
35 | +++ b/arch/ia64/Kconfig | |
36 | @@ -504,6 +504,7 @@ config IA64_MC_ERR_INJECT | |
37 | ||
38 | config SGI_SN | |
39 | def_bool y if (IA64_SGI_SN2 || IA64_GENERIC) | |
40 | + select HAVE_UNSTABLE_SCHED_CLOCK | |
41 | ||
42 | config IA64_ESI | |
43 | bool "ESI (Extensible SAL Interface) support" |