[PATCH v3 4/5] powerpc/fault: Avoid heavy search_exception_tables() verification

[Christophe Leroy]

Le 08/12/2020 à 16:07, Christophe Leroy a écrit :
> 
> 
> Le 08/12/2020 à 15:52, Aneesh Kumar K.V a écrit :
>> Christophe Leroy <christophe.leroy at csgroup.eu> writes:
>>
>>> search_exception_tables() is an heavy operation, we have to avoid it.
>>> When KUAP is selected, we'll know the fault has been blocked by KUAP.
>>> Otherwise, it behaves just as if the address was already in the TLBs
>>> and no fault was generated.
>>>
>>> Signed-off-by: Christophe Leroy <christophe.leroy at csgroup.eu>
>>> Reviewed-by: Nicholas Piggin <npiggin at gmail.com>
>>> ---
>>> v3: rebased
>>> v2: Squashed with the preceeding patch which was re-ordering tests that get removed in this patch.
>>> ---
>>>   arch/powerpc/mm/fault.c | 23 +++++++----------------
>>>   1 file changed, 7 insertions(+), 16 deletions(-)
>>>
>>> diff --git a/arch/powerpc/mm/fault.c b/arch/powerpc/mm/fault.c
>>> index 3fcd34c28e10..1770b41e4730 100644
>>> --- a/arch/powerpc/mm/fault.c
>>> +++ b/arch/powerpc/mm/fault.c
>>> @@ -210,28 +210,19 @@ static bool bad_kernel_fault(struct pt_regs *regs, unsigned long error_code,
>>>           return true;
>>>       }
>>> -    if (!is_exec && address < TASK_SIZE && (error_code & (DSISR_PROTFAULT | DSISR_KEYFAULT)) &&
>>> -        !search_exception_tables(regs->nip)) {
>>> -        pr_crit_ratelimited("Kernel attempted to access user page (%lx) - exploit attempt? (uid: 
>>> %d)\n",
>>> -                    address,
>>> -                    from_kuid(&init_user_ns, current_uid()));
>>> -    }
>>> -
>>>       // Kernel fault on kernel address is bad
>>>       if (address >= TASK_SIZE)
>>>           return true;
>>> -    // Fault on user outside of certain regions (eg. copy_tofrom_user()) is bad
>>> -    if (!search_exception_tables(regs->nip))
>>> -        return true;
>>> -
>>> -    // Read/write fault in a valid region (the exception table search passed
>>> -    // above), but blocked by KUAP is bad, it can never succeed.
>>> -    if (bad_kuap_fault(regs, address, is_write))
>>> +    // Read/write fault blocked by KUAP is bad, it can never succeed.
>>> +    if (bad_kuap_fault(regs, address, is_write)) {
>>> +        pr_crit_ratelimited("Kernel attempted to %s user page (%lx) - exploit attempt? (uid: 
>>> %d)\n",
>>> +                    is_write ? "write" : "read", address,
>>> +                    from_kuid(&init_user_ns, current_uid()));
>>>           return true;
>>> +    }
>>
>>
>> With this I am wondering whether the WARN() in bad_kuap_fault() is
>> needed. A direct access of userspace address will trigger this, whereas
>> previously we used bad_kuap_fault() only to identify incorrect restore
>> of AMR register (ie, to identify kernel bugs). Hence a WARN() there was
>> useful. We loose that differentiation now?
> 
> Yes, I wanted to remove the WARN(), see 
> https://patchwork.ozlabs.org/project/linuxppc-dev/patch/cc9129bdda1dbc2f0a09cf45fece7d0b0e690784.1605541983.git.christophe.leroy@csgroup.eu/ 
> 
> but I understood from Michael that maybe it was not a good idea, so I left it aside for now when 
> rebasing to v3.
> 
> Yes previously we were able to differentiate between a direct access of userspace and a valid access 
> triggering a KUAP fault, but at the cost of the heavy search_exception_tables().
> The issue was reported by Nick through https://github.com/linuxppc/issues/issues/317
> 
> Should be perform the search_exception_tables() once we have hit the KUAP fault and WARN() only in 
> that case ?

I sent out v4 which does that: only emit the warning once we know it is a KUAP fault within an 
uaccess routine. With that, we should be back more or less as before: warning only if we hit KUAP 
fault AND it is a place where a userspace access should be granted.
We are not anymore in the fast hot path, so calling search_exception_tables() there should be a 
performance issue.

Christophe


> 
> I was wondering also if we should keep the WARN() only when CONFIG_PPC_KUAP_DEBUG is set ?
>