[PATCH v14 7/7] ARM: kprobes: enable OPTPROBES for ARM 32

On 2014/12/8 21:22, Jon Medhurst (Tixy) wrote:

On Mon, 2014-12-08 at 20:31 +0800, Wang Nan wrote:

quoted

On 2014/12/8 20:06, Wang Nan wrote:

quoted

On 2014/12/8 19:50, Jon Medhurst (Tixy) wrote:

quoted

On Mon, 2014-12-08 at 19:15 +0800, Wang Nan wrote:

quoted

On 2014/12/8 19:04, Jon Medhurst (Tixy) wrote:

quoted

On Mon, 2014-12-08 at 14:28 +0800, Wang Nan wrote:

[...]

quoted

quoted

so another CPU could find and delete next before this one has finished
doing so. Would the list end up in a consistent state where no loops
develop and no probes are missed? I don't know the answer and a full
analysis would be complicated, but my gut feeling is that if a cpu can
observe the links in the list in an inconsistent state then only bad
things can result.

I see the problem.

I'm thinking about making core.c and opt-arm.c to share stop_machine() code.
stop_machine() is required when removing breakpoint, so I'd like to define
a "remove_breakpoint" function in core.c and make opt-arm.c to call it.
Do you think it is a good idea?

What I mean is something like this:

Yes, that should work, though as remove_breakpoint is a globally visible
symbol, I suggest a less generic name for it, perhaps
remove_kprobe_breakpoint ?

I don't think it is globally visible. Only files in arm/probes/kprobes
can include "core.h". However, I do agree that remove_breakpoint() is not
a good name. In my v15 patch, I'd like to rename it to kprobe_remove_breakpoint(),
due to may of names defined in core.h are called kprobes_xxx.

Thank you!