Re: [PATCH] soc: qcom: smem: introduce qcom_smem_virt_to_phys()

[Alex Elder]

On 04/25/2018 06:29 PM, Chris Lew wrote:
> Hi Alex,
> 
> Minor comment.
> 
> On 4/25/2018 8:18 AM, Alex Elder wrote:
>> Create function qcom_smem_virt_to_phys(), which returns the physical
>> address corresponding to a given SMEM item's virtual address. This
>> feature is required for a driver that will soon be out for review.
>>
>> Signed-off-by: Alex Elder <elder@xxxxxxxxxx>
>> ---
>> Â drivers/soc/qcom/smem.cÂÂÂÂÂÂ | 27 +++++++++++++++++++++++++++
>> Â include/linux/soc/qcom/smem.h |Â 2 ++
>> Â 2 files changed, 29 insertions(+)
>>
>> diff --git a/drivers/soc/qcom/smem.c b/drivers/soc/qcom/smem.c
>> index 7d9a43da5084..70b2ee80d6bd 100644
>> --- a/drivers/soc/qcom/smem.c
>> +++ b/drivers/soc/qcom/smem.c
>> @@ -655,6 +655,33 @@ int qcom_smem_get_free_space(unsigned host)
>> Â }
>> Â EXPORT_SYMBOL(qcom_smem_get_free_space);
>> Â +/**
>> + * qcom_smem_virt_to_phys() - return the physical address associated
>> + * with an smem item pointer (previously returned by qcom_smem_get()
>> + * @p:ÂÂÂ the virtual address to convert
>> + *
>> + * Returns 0 if the pointer provided is not within any smem region.
>> + */
>> +phys_addr_t qcom_smem_virt_to_phys(void *p)
>> +{
>> +ÂÂÂ unsigned i;
>> +
> 
> We have a null pointer check for __smem here since it is called by
> external clients. This case should probably never happen though.

I think you're suggesting that we should verify __smem is non-null first?

I'll make a few statements about that.
- This function can only be called with a pointer that was returned by
  qcom_smem_get().  That function won't return a valid pointer unless
  __smem was non-null.
- The only other way __smem would be null is if this were called after
  qcom_smem_remove(), which is erroneous.
- I think putting a null pointer check suggests that it's a condition
  that might be expected to occur.  If anything, I'd put an assertion
  in there (e.g. BUG_ON(!__smem)) but I don't think it's warranted.

I do understand why you suggest this--and it's a relatively harmless
check.   But I think it's better without it.

					-Alex


>> +ÂÂÂ for (i = 0; i < __smem->num_regions; i++) {
>> +ÂÂÂÂÂÂÂ struct smem_region *region = &__smem->regions[i];
>> +
>> +ÂÂÂÂÂÂÂ if (p < region->virt_base)
>> +ÂÂÂÂÂÂÂÂÂÂÂ continue;
>> +ÂÂÂÂÂÂÂ if (p < region->virt_base + region->size) {
>> +ÂÂÂÂÂÂÂÂÂÂÂ u64 offset = p - region->virt_base;
>> +
>> +ÂÂÂÂÂÂÂÂÂÂÂ return (phys_addr_t)region->aux_base + offset;
>> +ÂÂÂÂÂÂÂ }
>> +ÂÂÂ }
>> +
>> +ÂÂÂ return 0;
>> +}
>> +EXPORT_SYMBOL(qcom_smem_virt_to_phys);
> 
> Thanks,
> Chris
>