Re: [RFC][PATCH -tip 8/9] kprobes: support respawn probes for moduleprobing

From: Masami Hiramatsu
Date: Thu Mar 19 2009 - 23:52:49 EST


Frederic Weisbecker wrote:
> On Thu, Mar 19, 2009 at 05:10:44PM -0400, Masami Hiramatsu wrote:
>> Add module_*probe API's to respawn probes on kernel modules.
>> kprobes which have been registered through register_module_*probe are
>> activated when the target module is loaded, and deactivated when unloading.
>> register_module_*probe require an activate_handler which is called right
>> before activating the probe, and if it returns !0, kprobes will be activated.
>>
>> Thus, users can check whether the target module is true target or not
>> (e.g. checking build-id) in activate_handler.
>>
>> Signed-off-by: Masami Hiramatsu <mhiramat@xxxxxxxxxx>
>> Cc: Ananth N Mavinakayanahalli <ananth@xxxxxxxxxx>
>> ---
>> include/linux/kprobes.h | 39 ++++++++
>> kernel/kprobes.c | 250 +++++++++++++++++++++++++++++++++++++++++++++++
>> 2 files changed, 289 insertions(+), 0 deletions(-)
>>
>> diff --git a/include/linux/kprobes.h b/include/linux/kprobes.h
>> index 2ec6cc1..9d47d08 100644
>> --- a/include/linux/kprobes.h
>> +++ b/include/linux/kprobes.h
>> @@ -37,6 +37,7 @@
>> #include <linux/spinlock.h>
>> #include <linux/rcupdate.h>
>> #include <linux/mutex.h>
>> +#include <linux/module.h>
>>
>> #ifdef CONFIG_KPROBES
>> #include <asm/kprobes.h>
>> @@ -69,6 +70,7 @@ typedef int (*kprobe_fault_handler_t) (struct kprobe *, struct pt_regs *,
>> int trapnr);
>> typedef int (*kretprobe_handler_t) (struct kretprobe_instance *,
>> struct pt_regs *);
>> +typedef int (*probe_activate_handler_t)(void *, struct module *);
>
>
> I guess it could also be useful to call the handler when the module
> is unloaded. Be using another parameter with a LOAD/UNLOAD enum value.

Hmm, when unloading module, kprobes becomes "gone". So I think we don't
need to check whether it can be deactivated or not when unloading.



>> struct kprobe {
>> struct hlist_node hlist;
>> @@ -279,6 +281,16 @@ void unregister_kretprobes(struct kretprobe **rps, int num);
>> void kprobe_flush_task(struct task_struct *tk);
>> void recycle_rp_inst(struct kretprobe_instance *ri, struct hlist_head *head);
>>
>> +int register_module_kprobe(struct kprobe *kp,
>> + probe_activate_handler_t handler, void *data);
>> +int register_module_kretprobe(struct kretprobe *rp,
>> + probe_activate_handler_t handler, void *data);
>> +int register_module_jprobe(struct jprobe *jp,
>> + probe_activate_handler_t handler, void *data);
>> +void unregister_module_kprobe(struct kprobe *kp);
>> +void unregister_module_kretprobe(struct kretprobe *rp);
>> +void unregister_module_jprobe(struct jprobe *jp);
>> +
>> #else /* !CONFIG_KPROBES: */
>>
>> static inline int kprobes_built_in(void)
>> @@ -345,5 +357,32 @@ static inline void unregister_kretprobes(struct kretprobe **rps, int num)
>> static inline void kprobe_flush_task(struct task_struct *tk)
>> {
>> }
>> +static inline int register_module_kprobe(struct kprobe *kp,
>> + probe_activate_handler_t handler,
>> + void *data)
>> +{
>> + return -ENOSYS;
>> +}
>> +static inline int register_module_kretprobe(struct kretprobe *rp,
>> + probe_activate_handler_t handler,
>> + void *data)
>> +{
>> + return -ENOSYS;
>> +}
>> +static inline int register_module_jprobe(struct jprobe *jp,
>> + probe_activate_handler_t handler,
>> + void *data)
>> +{
>> + return -ENOSYS;
>> +}
>> +void unregister_module_kprobe(struct kprobe *kp)
>> +{
>> +}
>> +void unregister_module_kretprobe(struct kretprobe *rp)
>> +{
>> +}
>> +void unregister_module_jprobe(struct jprobe *jp)
>> +{
>> +}
>
>
> Shouldn't the unregister_* be inlined too?
> I think they all could be static inlined too in case of !CONFIG_MODULE

Ah, right... it was my mistake.


>> #endif /* CONFIG_KPROBES */
>> #endif /* _LINUX_KPROBES_H */
>> diff --git a/kernel/kprobes.c b/kernel/kprobes.c
>> index 5016bfb..f16a54e 100644
>> --- a/kernel/kprobes.c
>> +++ b/kernel/kprobes.c
>> @@ -1416,6 +1416,256 @@ static int __kprobes debugfs_kprobe_init(void)
>> late_initcall(debugfs_kprobe_init);
>> #endif /* CONFIG_DEBUG_FS */
>>
>> +/* Kprobes module respawn support */
>> +enum probe_type {
>> + PROBE_TYPE_KPROBE,
>> + PROBE_TYPE_KRETPROBE,
>> + PROBE_TYPE_JPROBE,
>> +};
>> +
>> +struct module_probe_client {
>> + struct list_head list;
>> + const char *module; /* including symbol name */
>> + int active;
>> + void *data;
>> + probe_activate_handler_t handler;
>> + enum probe_type type;
>> + union {
>> + struct kprobe *kp;
>> + struct kretprobe *rp;
>> + struct jprobe *jp;
>> + };
>> +};
>
>
> #ifdef CONFIG_MODULE ?
>
> Frederic.

Yes.

Thank you,


--
Masami Hiramatsu

Software Engineer
Hitachi Computer Products (America) Inc.
Software Solutions Division

e-mail: mhiramat@xxxxxxxxxx

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/