Roland, With the current utrace/master tree, I am seeing that utrace_attach_task() never returns when invoked from the clone callback. The same module works fine with prior utrace (rcu as well as with my embed version).
The testcase is simple: a. attach an engine to attachstop-mt (from the gdb testsuite) _before_ it calls pthread_create. b. Watch for CLONE_THREAD and try to attach a utrace engine to the new thread. The utrace_attach_task() call never returns. If the utrace module is unloaded, the kernel barfs with the following innocuous information: BUG: unable to handle kernel paging request at fffffffffffffdff IP: [<ffffffffa012009a>] 0xffffffffa012009a PGD 203067 PUD 204067 PMD 0 Oops: 0002 [#1] SMP last sysfs file: /sys/devices/pci0000:01/0000:01:01.1/irq CPU 6 Modules linked in: <big list> [last unloaded: utrace_quiesce_threads] Pid: 6203, comm: attachstop-mt Not tainted 2.6.29-rc7-ut #1 eserver xSeries 366-[88632RA]- RIP: 0010:[<ffffffffa012009a>] [<ffffffffa012009a>] 0xffffffffa012009a RSP: 0018:ffff8801d34ebe10 EFLAGS: 00010246 RAX: fffffffffffffdff RBX: ffff8801f11a36c0 RCX: 00000000c0000100 RDX: 0000000000000000 RSI: ffff8801dd0507f8 RDI: ffff88022daf4500 RBP: 00000000fffffff4 R08: ffff8801d34ea000 R09: ffff88022f2596a0 R10: ffff8800280b1600 R11: 0000000000000018 R12: ffff8801d34f1860 R13: ffff8802210dd300 R14: ffff8801dd07e2c0 R15: 00000000003d0f00 FS: 00007f58c8d286e0(0000) GS:ffff88022f18e5c0(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b CR2: fffffffffffffdff CR3: 00000002029bd000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process attachstop-mt (pid: 6203, threadinfo ffff8801d34ea000, task ffff8801d3512440) Stack: 00000000003d0f00 ffff8801d34f1860 ffff8802210dd300 ffff8801d3512440 ffff8801d34ebe70 ffffffffa012028d ffff8801dd050618 ffff8801d35129e0 ffff8801d35129d8 ffffffff80260480 0000000000000000 ffff8801d34f1860 Call Trace: [<ffffffff80260480>] ? utrace_report_clone+0x95/0xfc [<ffffffff80239120>] ? do_fork+0x20b/0x2f3 [<ffffffff804a4035>] ? do_page_fault+0x3c7/0x74e [<ffffffff8020c243>] ? stub_clone+0x13/0x20 [<ffffffff8020bedb>] ? system_call_fastpath+0x16/0x1b Code: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 <00> 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 RIP [<ffffffffa012009a>] 0xffffffffa012009a RSP <ffff8801d34ebe10> CR2: fffffffffffffdff ---[ end trace 96bb7eb644ab73a4 ]--- I have verified that the earlier version of utrace works just fine. In the earlier case, the engine would go directly on to the attached list if the calling task was the creator of the new thread. This has changed with the new implementation. I haven't yet zeroed in on what exact change caused this problem. Ananth
