]> git.kernelconcepts.de Git - karo-tx-linux.git/commit
tracing/uprobes: Revert "Support mix of ftrace and perf"
authorOleg Nesterov <oleg@redhat.com>
Fri, 27 Jun 2014 17:01:36 +0000 (19:01 +0200)
committerSteven Rostedt <rostedt@goodmis.org>
Mon, 30 Jun 2014 17:21:58 +0000 (13:21 -0400)
commit48212542067a7ff6cbe829dbae279c2ff7557b44
treecddf954fff810ca3dd1e03b817e263357d2ee686
parent4d4c9cc839a308be3289a361ccba4447ee140552
tracing/uprobes: Revert "Support mix of ftrace and perf"

This reverts commit 43fe98913c9f67e3b523615ee3316f9520a623e0.

This patch is very wrong. Firstly, this change leads to unbalanced
uprobe_unregister(). Just for example,

# perf probe -x /lib/libc.so.6 syscall
# echo 1 >> /sys/kernel/debug/tracing/events/probe_libc/enable
# perf record -e probe_libc:syscall whatever

after that uprobe is dead (unregistered) but the user of ftrace/perf
can't know this, and it looks as if nobody hits this probe.

This would be easy to fix, but there are other reasons why it is not
simple to mix ftrace and perf. If nothing else, they can't share the
same ->consumer.filter. This is fixable too, but probably we need to
fix the poorly designed uprobe_register() interface first. At least
"register" and "apply" should be clearly separated.

Link: http://lkml.kernel.org/p/20140627170136.GA18319@redhat.com
Cc: Tom Zanussi <tom.zanussi@linux.intel.com>
Cc: "zhangwei(Jovi)" <jovi.zhangwei@huawei.com>
Cc: stable@vger.kernel.org # v3.14
Acked-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Reviewed-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
kernel/trace/trace_uprobe.c