This is the mail archive of the cygwin@cygwin.com mailing list for the Cygwin project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Processor time bug in fhandler_proc.cc



I think there is a bug in fhandler_proc.cc in the way the results of
calls to

  NTSTATUS ret = NtQuerySystemInformation (SystemProcessorTimes, (PVOID) &spt,
                                           sizeof spt, NULL);

are handled. Specifically the value of spt.IdleTime.QuadPart is already
included in spt.KernelTime.QuadPart but the code to compute the uptime
doesn't take that into account and adds in the idle time again as shown below.


      uptime = (spt.InterruptTime.QuadPart + spt.KernelTime.QuadPart +
                spt.IdleTime.QuadPart + spt.UserTime.QuadPart +
                spt.DpcTime.QuadPart) / 100000ULL;

Because the Idle time is already included in the KernelTime the computed
uptime is always too large. In the case of a lightly loaded system it's
about a factor of two and is easy to verify by just letting the system sit
for a while and then checking the reported value from /proc/uptime against 
wall clock time.

This also affects /proc/stat entry for CPU usage where you can see that
the system value is always larger than the idle value.

Regards
Emil


--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Bug reporting:         http://cygwin.com/bugs.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]