| Summary: | freeze on centos5.3+preempt x86_64 for most scripts | ||
|---|---|---|---|
| Product: | systemtap | Reporter: | Steve Fink <sphink> |
| Component: | kprobes | Assignee: | Unassigned <systemtap> |
| Status: | RESOLVED INVALID | ||
| Severity: | critical | ||
| Priority: | P2 | ||
| Version: | unspecified | ||
| Target Milestone: | --- | ||
| Host: | Target: | ||
| Build: | Last reconfirmed: | ||
Actually, it's kind of random how far the VM makes it. I rebooted and ran the same thing, and this time only the 1st line made it out. I suspect we need to get hold of a suchly configured kernel in order to diagnose the situation. (In reply to comment #0) > I am running a preemptive x86_64 kernel (pretty much just the CentOS 5.3 kernel > with preemption enabled). Does this happen on upstream kernel too? (with preemption enabled) Whoops! My kernel does have one local change applied: the atop accounting patches. Removing those fixes the problem. I'll go try to file a bug on atop instead. (Fortunately, I really don't need those patches.) |
I am running a preemptive x86_64 kernel (pretty much just the CentOS 5.3 kernel with preemption enabled). Most systemtap scripts hang the machine. I have also tried running it from within a VirtualBox VM, and get similar results. The version I am using on actual hardware and in the VM is systemtap-0.9.7-5.el5. Within the VM, I have also compiled the snapshot from 20091114. It managed to run for a little while before hanging. Specific example: this script probe kernel.function("*@net/socket.c") { printf ("%s -> %s\n", thread_indent(1), probefunc()) } probe kernel.function("*@net/socket.c").return { printf ("%s <- %s\n", thread_indent(-1), probefunc()) } run as stap -v socket-tree.stap On the hardware, it just hangs immediately after it starts running. On the VM, it does the same for 0.9.7, but with the snapshot version it manages to print out (after "Pass 5: starting run."): 0 ntpd(1903): -> do_sock_read 1000834 ntpd(1902): <- do_sock_read and then hangs the VM. The same thing on my 32-bit desktop prints out a bunch of stuff and then exits cleanly after 1.2 seconds. (Without me pressing Ctrl-C, which I don't understand yet, but that's different.) Earlier, I also got an oops on the VM with 0.9.7, but I can't remember what exactly I was doing. (It was something simple like stap -v -e 'probe syscall.* { printf("I am alive\n"); exit(); }') I can post my .config. I didn't change a whole lot in it. It is not built to be relocatable.