Re: [PATCH v2] ACPI/IORT: Reject platform dev creation when dev set to wrong numa node

[Kefeng Wang <wangkefeng.wang@huawei.com>]

On 2019/3/28 21:59, Robin Murphy wrote:
> On 28/03/2019 14:00, Kefeng Wang wrote:
>> If there is only node 0 in system, but smmuv3 device is set to offline
>> node 1, parsed from proximity domain in SMMUv3 IORT table, it will lead
>> to following crash,
>>
>> [   47.492451] Unable to handle kernel paging request at virtual address 0000000000001388
>> [   47.500361] Mem abort info:
>> [   47.503143]   ESR = 0x96000004
>> [   47.506189]   Exception class = DABT (current EL), IL = 32 bits
>> [   47.512099]   SET = 0, FnV = 0
>> [   47.515140]   EA = 0, S1PTW = 0
>> [   47.518272] Data abort info:
>> [   47.521144]   ISV = 0, ISS = 0x00000004
>> [   47.524970]   CM = 0, WnR = 0
>> [   47.527929] [0000000000001388] user address but active_mm is swapper
>> [   47.534285] Internal error: Oops: 96000004 [#1] SMP
>> [   47.539151] Modules linked in:
>> [   47.542194] CPU: 5 PID: 1 Comm: swapper/0 Not tainted 5.0.0 #15
>> [   47.549490] pstate: 80c00009 (Nzcv daif +PAN +UAO)
>> [   47.554272] pc : __alloc_pages_nodemask+0x13c/0x1068
>> [   47.559224] lr : __alloc_pages_nodemask+0xdc/0x1068
>> ...
>> [   47.646873] Process swapper/0 (pid: 1, stack limit = 0x(____ptrval____))
>> [   47.653560] Call trace:
>> [   47.655994]  __alloc_pages_nodemask+0x13c/0x1068
>> [   47.660600]  new_slab+0xec/0x570
>> [   47.663816]  ___slab_alloc+0x3e0/0x4f8
>> [   47.667553]  __slab_alloc+0x60/0x80
>> [   47.671029]  __kmalloc_node_track_caller+0x10c/0x478
>> [   47.675984]  devm_kmalloc+0x44/0xb0
>> [   47.679460]  pinctrl_bind_pins+0x4c/0x188
>> [   47.683457]  really_probe+0x78/0x2b8
>> [   47.687019]  driver_probe_device+0x64/0x110
>> [   47.691189]  device_driver_attach+0x74/0x98
>> [   47.695360]  __driver_attach+0x9c/0xe8
>> [   47.699095]  bus_for_each_dev+0x84/0xd8
>> [   47.702919]  driver_attach+0x30/0x40
>> [   47.706481]  bus_add_driver+0x170/0x218
>> [   47.710304]  driver_register+0x64/0x118
>> [   47.714128]  __platform_driver_register+0x54/0x60
>> [   47.718820]  arm_smmu_driver_init+0x24/0x2c
>> [   47.722991]  do_one_initcall+0xbc/0x328
>> [   47.726816]  kernel_init_freeable+0x304/0x3ac
>> [   47.731162]  kernel_init+0x18/0x110
>> [   47.734638]  ret_from_fork+0x10/0x1c
>> [   47.738202] Code: f90013b5 b9410fa1 1a9f0694 b50014c2 (b9400804)
>> [   47.744307] ---[ end trace dfeaed4c373a32da ]--
>>
>> This could be triggered by firmware bug with bad IORT configuration,
>> or a NUMA node has no memory attaching to it, also with NR_CPUS less
>> than CPUs presented in MADT.
>>
>> Make dev_set_proximity() with a return value, terminating device creation
>> if it return failure.
>>
>> Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
>> ---
>>   drivers/acpi/arm64/iort.c | 24 ++++++++++++++++++------
>>   1 file changed, 18 insertions(+), 6 deletions(-)
>>
>> diff --git a/drivers/acpi/arm64/iort.c b/drivers/acpi/arm64/iort.c
>> index e48894e002ba..c294c3490e66 100644
>> --- a/drivers/acpi/arm64/iort.c
>> +++ b/drivers/acpi/arm64/iort.c
>> @@ -1232,21 +1232,30 @@ static bool __init arm_smmu_v3_is_coherent(struct acpi_iort_node *node)
>>   /*
>>    * set numa proximity domain for smmuv3 device
>>    */
>> -static void  __init arm_smmu_v3_set_proximity(struct device *dev,
>> +static int  __init arm_smmu_v3_set_proximity(struct device *dev,
>>                             struct acpi_iort_node *node)
>>   {
>>       struct acpi_iort_smmu_v3 *smmu;
>>         smmu = (struct acpi_iort_smmu_v3 *)node->node_data;
>>       if (smmu->flags & ACPI_IORT_SMMU_V3_PXM_VALID) {
>> -        set_dev_node(dev, acpi_map_pxm_to_node(smmu->pxm));
>> +        int node = acpi_map_pxm_to_node(smmu->pxm);
>> +        if (node != NUMA_NO_NODE && !node_online(node))
>> +            return -EINVAL;
>> +
>> +        set_dev_node(dev, node);
>>           pr_info("SMMU-v3[%llx] Mapped to Proximity domain %d\n",
>>               smmu->base_address,
>>               smmu->pxm);
>>       }
>> +    return 0;
>>   }
>>   #else
>> -#define arm_smmu_v3_set_proximity NULL
>> +static int  __init arm_smmu_v3_set_proximity(struct device *dev,
>> +                          struct acpi_iort_node *node)
>> +{
>> +    return 0;
>> +}
>
> Doesn't this end up having the same effect as just leaving the callback assigned with NULL? Not sure why that would need to change :/

Oops, should not change this part  ; (

if no other issue, will resend

Thanks.


>
> Robin.
>


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel