[Xen-devel] [PATCH] Fix showing of CPU Affinity by xm vcpu-list

[Masaki Kanno <kanno.masaki@xxxxxxxxxxxxxx>]

Hi,

I saw strange information about CPU Affinity as follows.
My machine has two physical CPUs.  When I set CPU Affinity with both 
existing physical CPU numbers and non-existing physical CPU numbers, 
the information of CPU Affinity occasionally was "any cpu".

# xm info | grep nr_cpus
nr_cpus                : 2
# xm vcpu-pin vm1 0 1
# xm vcpu-list vm1
Name                                ID  VCPU   CPU State   Time(s) CPU Affinity
vm1                                  6     0     1   -b-      33.6 1
# xm vcpu-pin vm1 0 1-2
# xm vcpu-list vm1
Name                                ID  VCPU   CPU State   Time(s) CPU Affinity
vm1                                  6     0     1   -b-      33.6 any cpu  <-- 
Why?
# xm debug-key q
# xm dmesg
(snip)
(XEN) VCPU information and callbacks for domain 6:
(XEN)     VCPU0: CPU1 [has=F] flags=1 upcall_pend = 00, upcall_mask = 00 
dirty_cpus={} cpu_affinity={1-2}
(XEN)     100 Hz periodic timer (period 10 ms)
(XEN)     Notifying guest (virq 1, port 0, stat 0/-1/0)
# xm vcpu-pin vm1 0 1,3
# xm vcpu-list vm1
Name                                ID  VCPU   CPU State   Time(s) CPU Affinity
vm1                                  6     0     1   -b-      33.6 1
# xm debug-key q
# xm dmesg
(snip)
(XEN) VCPU information and callbacks for domain 6:
(XEN)     VCPU0: CPU1 [has=T] flags=0 upcall_pend = 00, upcall_mask = 00 
dirty_cpus={1} cpu_affinity={1,3}
(XEN)     100 Hz periodic timer (period 10 ms)
(XEN)     Notifying guest (virq 1, port 0, stat 0/-1/0)
# xm vcpu-pin vm1 0 1,4
# xm vcpu-list vm1
Name                                ID  VCPU   CPU State   Time(s) CPU Affinity
vm1                                  6     0     1   -b-      33.6 any cpu  <-- 
Why?
# xm debug-key q
# xm dmesg
(snip)
(XEN) VCPU information and callbacks for domain 6:
(XEN)     VCPU0: CPU1 [has=T] flags=0 upcall_pend = 00, upcall_mask = 00 
dirty_cpus={1} cpu_affinity={1,4}
(XEN)     100 Hz periodic timer (period 10 ms)
(XEN)     Notifying guest (virq 1, port 0, stat 0/-1/0)


I investigated a cause why strange information was shown.  The following 
line had the cause. 

diff -r d0817f08599a tools/python/xen/xm/main.py
--- a/tools/python/xen/xm/main.py       Fri May 16 09:37:19 2008 +0100
+++ b/tools/python/xen/xm/main.py       Mon May 19 10:03:37 2008 +0900
@@ -1096,7 +1096,7 @@ def xm_vcpu_list(args):
 
             # normalize cpumap by modulus nr_cpus, and drop duplicates
             cpumap = dict.fromkeys(
-                       map(lambda x: x % nr_cpus, cpumap)).keys()
+                       filter(lambda x: x < nr_cpus, cpumap)).keys()
             if len(cpumap) == nr_cpus:
                 return "any cpu"
 
Because the line makes cpumap by using results of "x % nr_cpus", 
non-existing physical CPU numbers are handled as existing physical CPU 
numbers. 

  - 1 % 2 --> Physical CPU number 1
  - 2 % 2 --> Physical CPU number 0
  - 3 % 2 --> Physical CPU number 1
  - 4 % 2 --> Physical CPU number 0

This patch removes non-existing physical CPU numbers from cpumap by 
filter(), then this problem is solved. 

Signed-off-by: Masaki Kanno <kanno.masaki@xxxxxxxxxxxxxx>

Best regards,
 Kan

Attachment: xm_vcpu_list.patch
Description: Binary data

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel