[PATCH] iommu/arm-smmu-v3: fixed check process for disable_bypass module parameter

[Robin Murphy]

On 2022-08-03 13:45, ishii.shuuichir at fujitsu.com wrote:
> Hi, Robin,
> Thank you for your comments.
> 
>>>    	/* Enable the SMMU interface, or ensure bypass */
>>> -	if (!bypass || disable_bypass) {
>>> +	if (!bypass && disable_bypass) {
>>
>> This change looks obviously wrong - if bypass is false here then we definitely
>> want to enable the SMMU, so disable_bypass is irrelevant. It shouldn't even be
>> possible to get here with bypass==true under ACPI, since
>> arm_smmu_device_acpi_probe() cannot fail :/
> 
> Sorry, my understanding of the meaning of the disable_bypass module parameter
> and the process of setting GBPA_ABORT was insufficient.
> 
> I misunderstood that the disable_bypass module parameter is used to simply
> bypass (disable) SMMU (SMMU_CR0.SMMUEN == 0 and SMMU_GBPA.ABORT == 0).
> Forget about the fixes in this patch.
> 
> Although our understanding was lacking,
> we thought it would be a good idea to have a module parameter that simply disables SMMU,
> so we were considering a fix.

Right, disable_bypass is a security/robustness feature for when the 
driver *is* in use. If for some reason you want to disable the SMMU 
drivers completely, they are regular driver model drivers, so just don't 
load the module in the first place (or use initcall_blacklist if it's 
built-in).

Thanks,
Robin.

> 
> Best regards,
> Shuuichirou.
> 
>> -----Original Message-----
>> From: Robin Murphy <robin.murphy at arm.com>
>> Sent: Wednesday, August 3, 2022 6:26 PM
>> To: Ishii, Shuuichirou/石井 周一郎 <ishii.shuuichir at fujitsu.com>;
>> will at kernel.org; joro at 8bytes.org; thunder.leizhen at huawei.com; jgg at ziepe.ca;
>> tglx at linutronix.de; chenxiang66 at hisilicon.com; christophe.jaillet at wanadoo.fr;
>> john.garry at huawei.com; baolu.lu at linux.intel.com;
>> linux-arm-kernel at lists.infradead.org; iommu at lists.linux.dev;
>> linux-kernel at vger.kernel.org
>> Subject: Re: [PATCH] iommu/arm-smmu-v3: fixed check process for
>> disable_bypass module parameter
>>
>> On 2022-08-03 00:42, Shuuichirou Ishii wrote:
>>> The current process does not enable the bypass setting regardless of
>>> the value of the disable_bypass module parameter when ACPI is enabled,
>>> so the value of the disable_bypass module parameter has been corrected
>>> so that it is handled correctly.
>>>
>>> Signed-off-by: Shuuichirou Ishii <ishii.shuuichir at fujitsu.com>
>>> ---
>>>    drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c | 2 +-
>>>    1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
>>> b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
>>> index 88817a3376ef..256d7b2a83a7 100644
>>> --- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
>>> +++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.c
>>> @@ -3396,7 +3396,7 @@ static int arm_smmu_device_reset(struct
>> arm_smmu_device *smmu, bool bypass)
>>>    		enables &= ~(CR0_EVTQEN | CR0_PRIQEN);
>>>
>>>    	/* Enable the SMMU interface, or ensure bypass */
>>> -	if (!bypass || disable_bypass) {
>>> +	if (!bypass && disable_bypass) {
>>
>> This change looks obviously wrong - if bypass is false here then we definitely
>> want to enable the SMMU, so disable_bypass is irrelevant. It shouldn't even be
>> possible to get here with bypass==true under ACPI, since
>> arm_smmu_device_acpi_probe() cannot fail :/
>> Robin.
>>
>>>    		enables |= CR0_SMMUEN;
>>>    	} else {
>>>    		ret = arm_smmu_update_gbpa(smmu, 0, GBPA_ABORT);