Re: [PATCH 1/2] This patch fixes GDBServer's run control for single stepping

[Antoine Tremblay <antoine.tremblay@ericsson.com>]

Antoine Tremblay writes:

> Yao Qi writes:
>
>> On 17-02-17 19:17:56, Antoine Tremblay wrote:
>>> > In ARM ARM, we have the pseudo code,
>>> >
>>> > boolean InITBlock()
>>> > return (ITSTATE.IT<3:0> != ‘0000’);
>>> >
>>> > ITSTATE can be got from CPSR.
>>>
>>> Yes that's good if you're inserting a breakpoint at current PC but
>>> otherwise you will need something else...
>>
>> In software single step, we calculate the next pcs, and select
>> breakpoint kinds of them, according to current pc.  If current
>> pc is not within IT block (!InITBlock ()) or the last instruction
>> in IT block (LastInITBlock ()), we can safely use 16-bit thumb
>> breakpoint for any thumb instruction.  That is, in
>> gdbserver/linux-aarch32-low.c:arm_breakpoint_kind_from_current_state,
>> we can return ARM_BP_KIND_THUMB if (!InITBlock () || LastInITBlock ()).
>>
>> Then, in some level, when installing software single step breakpoints,
>> if one breakpoint type is ARM_BP_KIND_THUMB2 and installed
>> address is 2-byte aligned, stop all threads.
>
> Yes that would make sense but I think we can be calling get_next_pc
> to insert a software single step breakpoint at an address different then
> the current PCs next address.
>
> See: https://sourceware.org/ml/gdb-patches/2016-06/msg00268.html
>
> "> If we only remove before reporting an event to gdb, then I don't
>> understand this.  We already insert single-step breakpoints when
>> we process the resume request from gdb, no?
>
> We insert single-step breakpoints when we process the resume requests
> and threads are about to be resumed.  If threads still have pending
> status, single-step breakpoints are not installed, so we need to install
> them in proceed_all_lwps."
>
> This is still true AFAIK so GDBServer may be at any PC stopped on a
> pending event and need to insert a single step breakpoint at an address
> unrelated to that event so in that case CPSR can't be used...
>
> Thanks,
> Antoine

Oops looks like it had been a while since I checked the get_next_pc
code, so we switch the current thread and use that PC so it should work.

Disregard my last mail :)

Looks like this solution is a good way forward.