Bug 100815
| Summary: | FBSD 5.5 broke nanosleep in libc_r | ||
|---|---|---|---|
| Product: | Base System | Reporter: | Michael Scheidell <scheidell> |
| Component: | threads | Assignee: | freebsd-threads (Nobody) <threads> |
| Status: | Closed FIXED | ||
| Severity: | Affects Only Me | CC: | hackers |
| Priority: | Normal | ||
| Version: | 5.5-RELEASE | ||
| Hardware: | Any | ||
| OS: | Any | ||
State Changed From-To: open->suspended In RELENG_5,6 and HEAD libc_r is deprecated in favour of libpthread and libthr. Nobody is working on libc_r bugs so mark this PR as suspended. Linked same program with lib_thr and it exits, signal 10. (make clean, make) Linked same program with lib_pthread and it actually reboots the computer. I will attempt to either identify the problem to a repeatable POC, or might just remote threads. The first one I will work on is the lib_pthread. Nothing (well, almost nothing) run in userland should reboot the computer. -- Michael Scheidell, CTO SECNAP Network Security / www.secnap.com scheidell@secnap.net / 1+561-999-5000, x 1131 State Changed From-To: suspended->feedback We need to check, if this behavior also occurs in recent FreeBSD versions and with libthread. State Changed From-To: feedback->closed I was not able to reproduce any issues described in this PR with a recent version of HEAD. As you can see A201102231549.22 B201102231549.23 A201102231549.23 B201102231549.23 A201102231549.23 B201102231548.00 A201102231548.00 B201102231548.00 A201102231548.00 B201102231548.01 A201102231548.01 B201102231548.01 A201102231548.01 B201102231548.02 A201102231548.02 B201102231548.02 A201102231548.02 the date got set back ~1.5 minutes, however no delays or even reboots occured. |
nanosleep returns after time+delta if wallclock changed with libc_r Fix: compile with: cc -g -c nanotest.c cc -g -o nanotest nanotest.o or compile on 5.4 system. (I have only tried compiling and running on 5.4 system. I have not yet tried compiling on 5.4 and running on 5.5 or visa versa) I used ntpdate -su to change clock. (I have an internal system that is 45 seconds behind, just for this test) output with 5.4 or without libc_r shows wallclock 'stepping' when you change time no delay in display. 5.5 with libc_r shows stepping if moving clock head, but if you move wallclock back 45 seconds, display stops for 45 seconds. How-To-Repeat: 5.4 works, not including libc_r in 5.5 works Given this POC: int main( int argc, char **argv ) { char execs1[256] = "/bin/date +A%Y%m%d%H%M.%S"; char execs2[256] = "/bin/date +B%Y%m%d%H%M.%S"; int a; while ( 1) { system( execs1 ); usleep( 500*1000); system( execs2 ); } return 0; } /* main */ compile with: cc -g -c nanotest.c cc -g -o nanotest nanotest.o -lc_r run it, set clock back 45 seconds, observe 45 second delay