[uClinux-dev] Kernel 2.6.19: More proc problems

Atle Nissestad atle at nissestad.no
Tue Jan 30 04:38:00 EST 2007


When updating from kernel 2.6.18.6 to 2.6.19.1 on the nios2nommu architecture 
I in addition to the proc cmdline-issue (which is fixed) noticed strange 
behaviour of the top utility. This is understandable when looking at the 
output of /proc/<pid>/stat for the same process running on 2.6.18 vs 2.6.19:

2.6.18:
# cat /proc/66/stat && sleep 5 && cat /proc/66/stat
66 (testprog) S 1 14 14 0 -1 0 0 0 0 0 1542 19758 0 0 15 0 1 0 8701 139264 0 
4294967295 61472832 61519536 61603604 61602948 61496560 0 0 6 0 0 0 0 17 0 0 
0 0
66 (testprog) S 1 14 14 0 -1 0 0 0 0 0 1616 20569 0 0 15 0 1 0 8701 139264 0 
4294967295 61472832 61519536 61603604 61602972 61495400 0 0 6 0 0 0 0 17 0 0 
0 0

2.6.19:
# cat /proc/67/stat && sleep 5 && cat /proc/67/stat
67 (testprog) S 1 14 14 0 -1 0 0 0 0 0 10018 0 0 0 15 0 1 0 7063 139264 0 
4294967295 61079616 61126320 61210388 61209732 61103344 0 0 6 0 0 0 0 17 0 0 
0 0
67 (testprog) S 1 14 14 0 -1 0 0 0 0 0 10995 0 0 0 15 0 1 0 7063 139264 0 
4294967295 61079616 61126320 61210388 61209732 61103344 0 0 6 0 0 0 0 17 0 0 
0 0

The stime (kernel jiffies) entry (no 15) is 0 for the 2.6.19 kernel, and the 
behaviour of the utime (usermode jiffies) entry (no 14) is more like stime on 
2.6.18.
Has there been any format changes of proc stat, or is this a bug 
(arch-specific perhaps) ?


- Atle



More information about the uClinux-dev mailing list