Re: [RFC 03/16] kgr: initial code

[Aravinda Prasad]

On Wednesday 14 May 2014 03:42 PM, Jiri Slaby wrote:
> On 05/14/2014 11:28 AM, Aravinda Prasad wrote:
>> On Wednesday 30 April 2014 08:00 PM, Jiri Slaby wrote:
>>> From: Jiri Kosina <jkosina@xxxxxxx>
>>>
>>> Provide initial implementation. We are now able to do ftrace-based
>>> runtime patching of the kernel code.
>>>
>>> In addition to that, we will provide a kgr_patcher module in the next
>>> patch to test the functionality.
>>
>> Hi Jiri,
>>
>> Interesting! I have couple of comments:
>>
>> I think with kgraft (also with kpatch, though have not looked into
>> it yet), the patched function cannot be dynamically ftraced.
>> Though dynamic ftrace can be enabled on the new code, the user is
>> required to know the function label of the new code. This could
>> potentially break existing scripts. I think this should be documented.
> 
> Hi,
> 
> of course that the functions can be traced. Look, I turned on tracing
> for capable, then patched, then turned on tracing for new_capable (which
> is the patched function). So now, trace shows:
>   console-kit-dae-535   [001] ...1   181.729698: capable <-vt_ioctl
>  console-kit-dae-539   [001] ...1   181.729741: capable <-vt_ioctl
>  console-kit-dae-541   [000] .N.1   181.906014: capable <-vt_ioctl
>          systemd-1     [001] ...1   181.937328: capable <-SyS_epoll_ctl
>             sshd-662   [001] ...1   246.437561: capable <-sock_setsockopt
>             sshd-662   [001] ...1   246.437564: new_capable
> <-sock_setsockopt
>             sshd-662   [001] ...1   246.444790: capable <-sock_setsockopt
>             sshd-662   [001] ...1   246.444793: new_capable
> <-sock_setsockopt
>      dbus-daemon-128   [000] .N.1   246.456307: capable <-SyS_epoll_ctl
>      dbus-daemon-128   [000] ...1   246.456611: new_capable <-SyS_epoll_ctl
> 
> 
> There is no limitation thanks to the use of the ftrace subsystem. We are
> just another user, i.e. another piece of code called in a loop for a
> particular fentry location.

Yes true. What I intended to mention is that: the trace is turned on
for "capable" then the function is patched. Eventually, once the patch
is finalized, there will be no trace log for "capable". Someone tracing
the function "capable", not aware of patching, may think that it has not
been invoked. The user, hence, is expected to start tracing
"new_capable". I think this should be documented.

What if someone turns on tracing for "capable" after it is patched?
Will it overwrite the slow/fast stub?


> 
>>> +/*
>>> + * The stub needs to modify the RIP value stored in struct pt_regs
>>> + * so that ftrace redirects the execution properly.
>>> + */
>>> +#define KGR_STUB_ARCH_SLOW(_name, _new_function)			\
>>> +static void _new_function ##_stub_slow (unsigned long ip, unsigned long parent_ip,	\
>>> +		struct ftrace_ops *ops, struct pt_regs *regs)		\
>>> +{									\
>>> +	struct kgr_loc_caches *c = ops->private;			\
>>> +									\
>>> +	if (task_thread_info(current)->kgr_in_progress && current->mm) {\
>>
>> Is there a race here? The per task kgr_in_progress is set after
>> the slow stub is registered in register_ftrace_function(). If the
>> patched function is called in between it will be redirected to new code.
> 
> Hmm, that looks strange. I will look into that and the other comments
> later (and comment separately). Thanks.
> 

-- 
Regards,
Aravinda

--
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/