+++ This bug was initially created as a clone of Bug #170341 +++ Description of problem: With the simple code shown below, gettimeofday() is going backward by a random amount of time (between 10 us and 2 seconds). #include "stdio.h" #include <sys/time.h> #include <memory.h> #include "errno.h" typedef long long long64; int main(int, char**) { long64 issues = 0; long64 total = 0; while(1) { bool bError = false; timeval tv; memset(&tv, 0, sizeof(tv)); if(::gettimeofday (&tv, NULL) == -1) { printf("returned error:%d\n", errno); bError = true; continue; } long64 ticks = tv.tv_sec*1000000ll+tv.tv_usec; memset(&tv, 0, sizeof(tv)); if(::gettimeofday (&tv, NULL) == -1) { printf("returned error:%d\n", errno); bError = true; continue; } long64 ticks2 = tv.tv_sec*1000000ll+tv.tv_usec; if(ticks2 < ticks) { printf("Delta time:%lld us percentage:%f\n", ticks - ticks2, 100.*double(issues)/total); issues++; } total++; } return 1; } Version-Release number of selected component (if applicable): kernel version: 2.6.9-11.ELsmp How reproducible: Always Steps to Reproduce: 1.Compile and run the above code 2.Program will print "Delta time: xx us percentage: yy" when gettimeofday() reports going backward 3.The frequency of error seems similar regardless if ntpd is running and synchronized or not. Actual Results: Delta time:200 us percentage:0.000003 Delta time:375 us percentage:0.000003 Delta time:910 us percentage:0.000003 Delta time:177 us percentage:0.000003 Delta time:177 us percentage:0.000003 Delta time:177 us percentage:0.000003 Delta time:139 us percentage:0.000003 Delta time:177 us percentage:0.000003 Delta time:841 us percentage:0.000003 Delta time:179 us percentage:0.000003 Delta time:199 us percentage:0.000003 Delta time:178 us percentage:0.000003 Delta time:177 us percentage:0.000003 Delta time:552 us percentage:0.000003 Delta time:176 us percentage:0.000003 Delta time:176 us percentage:0.000003 Delta time:84 us percentage:0.000003 Delta time:258 us percentage:0.000003 Delta time:436 us percentage:0.000003 Delta time:726 us percentage:0.000003 Delta time:60 us percentage:0.000003 Delta time:585 us percentage:0.000003 Delta time:758 us percentage:0.000003 Delta time:717 us percentage:0.000003 Delta time:546 us percentage:0.000002 Delta time:372 us percentage:0.000002 Delta time:444 us percentage:0.000002 Delta time:30 us percentage:0.000002 Delta time:858 us percentage:0.000002 Delta time:392 us percentage:0.000002 Delta time:571 us percentage:0.000002 Delta time:759 us percentage:0.000002 Delta time:746 us percentage:0.000002 Delta time:99 us percentage:0.000002 Delta time:274 us percentage:0.000002 Delta time:760 us percentage:0.000002 Delta time:156 us percentage:0.000002 Delta time:759 us percentage:0.000002 Delta time:754916 us percentage:0.000002 Delta time:808 us percentage:0.000002 Delta time:846 us percentage:0.000002 Delta time:1788404 us percentage:0.000002 Delta time:1787652 us percentage:0.000002 Delta time:1396667 us percentage:0.000002 Delta time:390090 us percentage:0.000002 Delta time:1788633 us percentage:0.000002 Delta time:858 us percentage:0.000002 Delta time:877 us percentage:0.000002 Delta time:860 us percentage:0.000002 Delta time:860 us percentage:0.000002 Delta time:858 us percentage:0.000002 Delta time:858 us percentage:0.000002 Delta time:494 us percentage:0.000002 Delta time:673 us percentage:0.000002 Delta time:909 us percentage:0.000002 Delta time:438 us percentage:0.000002 Delta time:1394161 us percentage:0.000002 Delta time:758667 us percentage:0.000002 Delta time:757164 us percentage:0.000002 Delta time:758494 us percentage:0.000002 Expected Results: I should not see the "Delta time: xx yy" message Additional info: This problem happens on a dual-core bi-processor Opteron box. >cat /proc/cpuinfo processor : 0 vendor_id : AuthenticAMD cpu family : 15 model : 33 model name : 02/21 stepping : 2 cpu MHz : 1994.147 cache size : 1024 KB fdiv_bug : no hlt_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 1 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext lm 3dnowext 3dnow pni bogomips : 3915.77 processor : 1 vendor_id : AuthenticAMD cpu family : 15 model : 33 model name : 02/21 stepping : 2 cpu MHz : 1994.147 cache size : 1024 KB fdiv_bug : no hlt_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 1 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext lm 3dnowext 3dnow pni bogomips : 3981.31 processor : 2 vendor_id : AuthenticAMD cpu family : 15 model : 33 model name : 02/21 stepping : 2 cpu MHz : 1994.147 cache size : 1024 KB fdiv_bug : no hlt_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 1 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext lm 3dnowext 3dnow pni bogomips : 3981.31 processor : 3 vendor_id : AuthenticAMD cpu family : 15 model : 33 model name : 02/21 stepping : 2 cpu MHz : 1994.147 cache size : 1024 KB fdiv_bug : no hlt_bug : no f00f_bug : no coma_bug : no fpu : yes fpu_exception : yes cpuid level : 1 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext lm 3dnowext 3dnow pni bogomips : 3981.31 I also tried to disable TSC but this doesn't work since the kernel is compiled with CONFIG_X86_TSC. >dmesg Linux version 2.6.9-11.ELsmp (bhcompile.redhat.com) (gcc version 3.4.3 20050227 (Red Hat 3.4.3-22)) #1 SMP Fri May 20 18:26:27 EDT 2005 ............ Kernel command line: ro root=LABEL=/ rhgb quiet notsc notsc: Kernel compiled with CONFIG_X86_TSC, cannot disable TSC. Initializing CPU#0 CPU 0 irqstacks, hard=c03db000 soft=c03bb000 PID hash table entries: 4096 (order: 12, 65536 bytes) Detected 1994.147 MHz processor. Using tsc for high-res timesource
*** This bug has been marked as a duplicate of 174390 ***
Reopening/reclosing to get off of kernel-maint list.
TSC's between processors are not syncronized on AMD based systems. As a result the kernel might read from different TSC's each time and introduce error into the timekeeping math. Use "notsc" at boot to prevent this problem. This is a known issue that will be fixed in future releases.