Re: [PATCH v2 09/16] powerpc/watchpoint: Convert thread_struct->hw_brk to an array

From: Ravi Bangoria <ravi.bangoria@linux.ibm.com>
To: Christophe Leroy <christophe.leroy@c-s.fr>
Cc: mpe@ellerman.id.au, mikey@neuling.org, apopple@linux.ibm.com,
	paulus@samba.org, npiggin@gmail.com,
	naveen.n.rao@linux.vnet.ibm.com, peterz@infradead.org,
	jolsa@kernel.org, oleg@redhat.com, fweisbec@gmail.com,
	mingo@kernel.org, linuxppc-dev@lists.ozlabs.org,
	linux-kernel@vger.kernel.org,
	Ravi Bangoria <ravi.bangoria@linux.ibm.com>
Subject: Re: [PATCH v2 09/16] powerpc/watchpoint: Convert thread_struct->hw_brk to an array
Date: Wed, 1 Apr 2020 14:36:21 +0530	[thread overview]
Message-ID: <a1bd4536-9313-120e-b747-1a4a31b8da73@linux.ibm.com> (raw)
In-Reply-To: <e5af5ed7-b9df-e334-1bdb-e7f82ae32697@c-s.fr>

>>   static void set_debug_reg_defaults(struct thread_struct *thread)
>>   {
>> -    thread->hw_brk.address = 0;
>> -    thread->hw_brk.type = 0;
>> -    thread->hw_brk.len = 0;
>> -    thread->hw_brk.hw_len = 0;
>> -    if (ppc_breakpoint_available())
>> -        set_breakpoint(&thread->hw_brk);
>> +    int i;
>> +
>> +    for (i = 0; i < nr_wp_slots(); i++) {
> 
> Maybe you could add the following that you added other places:
> 
>      struct arch_hw_breakpoint null_brk = {0};
> 
> Then do
> 
>      thread->hw_brk[i] = null_brk;

Yes that's better.

[...]

>> +static void switch_hw_breakpoint(struct task_struct *new)
>> +{
>> +    int i;
>> +
>> +    for (i = 0; i < nr_wp_slots(); i++) {
>> +        if (unlikely(!hw_brk_match(this_cpu_ptr(&current_brk[i]),
>> +                       &new->thread.hw_brk[i]))) {
>> +            __set_breakpoint(&new->thread.hw_brk[i], i);
>> +        }
> 
> Or could be:
> 
>          if (likely(hw_brk_match(this_cpu_ptr(&current_brk[i]),
>                      &new->thread.hw_brk[i])))
>              continue;
>          __set_breakpoint(&new->thread.hw_brk[i], i);
> 

Sure.

[...]

>> @@ -128,8 +131,10 @@ static void do_signal(struct task_struct *tsk)
>>        * user space. The DABR will have been cleared if it
>>        * triggered inside the kernel.
>>        */
>> -    if (tsk->thread.hw_brk.address && tsk->thread.hw_brk.type)
>> -        __set_breakpoint(&tsk->thread.hw_brk, 0);
>> +    for (i = 0; i < nr_wp_slots(); i++) {
>> +        if (tsk->thread.hw_brk[i].address && tsk->thread.hw_brk[i].type)
>> +            __set_breakpoint(&tsk->thread.hw_brk[i], i);
>> +    }
> 
> thread.hwbrk also exists when CONFIG_PPC_ADV_DEBUG_REGS is selected.
> 
> You could replace the #ifndef CONFIG_PPC_ADV_DEBUG_REGS by an if (!IS_ENABLED(CONFIG_PPC_ADV_DEBUG_REGS)) and then no need of an ifdef when declaring the int i;

Makes sense. Will change it.

Ravi