Re: [PATCH] ARM: mm: Facilitate debugging CONFIG_KUSER_HELPERS disabled

From: Florian Fainelli
Date: Thu Oct 25 2018 - 14:48:40 EST


On 10/25/18 11:30 AM, Souptick Joarder wrote:
> On Thu, Oct 25, 2018 at 11:40 PM Florian Fainelli <f.fainelli@xxxxxxxxx> wrote:
>>
>> Some software such as perf makes unconditional use of the special
>> [vectors] page which is only provided when CONFIG_KUSER_HELPERS is
>> enabled in the kernel.
>>
>> Facilitate the debugging of such situations by printing a debug message
>> to the kernel log showing the task name and the faulting address.
>>
>> Suggested-by: Russell King <rmk+kernel@xxxxxxxxxxxxxxx>
>> Signed-off-by: Florian Fainelli <f.fainelli@xxxxxxxxx>
>> ---
>> arch/arm/mm/fault.c | 5 +++++
>> 1 file changed, 5 insertions(+)
>>
>> diff --git a/arch/arm/mm/fault.c b/arch/arm/mm/fault.c
>> index f4ea4c62c613..f17471fbc1c4 100644
>> --- a/arch/arm/mm/fault.c
>> +++ b/arch/arm/mm/fault.c
>> @@ -173,6 +173,11 @@ __do_user_fault(struct task_struct *tsk, unsigned long addr,
>> show_regs(regs);
>> }
>> #endif
>> +#ifndef CONFIG_KUSER_HELPERS
>
> Just have one doubt, if the condition is "#ifdef CONFIG_KUSER_HELPER"
> as commit message suggests the scenario is valid when CONFIG_KUSER_HELPER
> is enabled ? No ?

#ifndef CONFIG_KUSER_HELPERS is what is intended here, when that option
is not enabled, there is no [vectors] page provided in a program's
virtual address space, so accesses to that virtual address will cause a
fault which we are catching here.

When CONFIG_KUSER_HELPERS is enabled, every program gets a valid vectors
page in its virtual address space, and accessing that address would not
cause a fault, since the page is there.

Does this clarify the intent of this commit?

>
>> + if ((sig == SIGSEGV) && ((addr & PAGE_MASK) == 0xffff0000))
>> + printk(KERN_DEBUG "%s: CONFIG_KUSER_HELPERS disabled at 0x%08lx\n",
>> + tsk->comm, addr);
>> +#endif
>>
>> tsk->thread.address = addr;
>> tsk->thread.error_code = fsr;
>> --
>> 2.17.1
>>


--
Florian