Re: [PATCH] uio: Provide UIO_MEM_PHYS_WC for write-combine mappings

[Mike Looijmans <mike.looijmans@topic.nl>]

On 1/14/26 12:24, Greg KH wrote:
> On Wed, Jan 14, 2026 at 11:51:49AM +0100, Mike Looijmans wrote:
>> Introduce UIO_MEM_PHYS_WC which is identical to UIO_MEM_PHYS except
>> that the mapping will enable write-combining. This allows the system to
>> coalesce writes and greatly improves performance if the underlying
>> hardware allows that. Typically used for video buffers, but also very
>> well suited for filling FPGA (block)RAMs with reference data.
>>
>> Signed-off-by: Mike Looijmans <mike.looijmans@topic.nl>
>>
>> ---
>>
>>   drivers/uio/uio.c          | 3 +++
>>   include/linux/uio_driver.h | 1 +
>>   2 files changed, 4 insertions(+)
>>
>> diff --git a/drivers/uio/uio.c b/drivers/uio/uio.c
>> index fa0d4e6aee16..c33db04d2004 100644
>> --- a/drivers/uio/uio.c
>> +++ b/drivers/uio/uio.c
>> @@ -757,6 +757,8 @@ static int uio_mmap_physical(struct vm_area_struct *vma)
>>   	vma->vm_ops = &uio_physical_vm_ops;
>>   	if (idev->info->mem[mi].memtype == UIO_MEM_PHYS)
>>   		vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
>> +	else if (idev->info->mem[mi].memtype == UIO_MEM_PHYS_WC)
>> +		vma->vm_page_prot = pgprot_writecombine(vma->vm_page_prot);
>>   
>>   	/*
>>   	 * We cannot use the vm_iomap_memory() helper here,
>> @@ -858,6 +860,7 @@ static int uio_mmap(struct file *filep, struct vm_area_struct *vma)
>>   	switch (idev->info->mem[mi].memtype) {
>>   	case UIO_MEM_IOVA:
>>   	case UIO_MEM_PHYS:
>> +	case UIO_MEM_PHYS_WC:
>>   		ret = uio_mmap_physical(vma);
>>   		break;
>>   	case UIO_MEM_LOGICAL:
>> diff --git a/include/linux/uio_driver.h b/include/linux/uio_driver.h
>> index 334641e20fb1..b925b8c17214 100644
>> --- a/include/linux/uio_driver.h
>> +++ b/include/linux/uio_driver.h
>> @@ -171,6 +171,7 @@ extern int __must_check
>>    * Do not use in new drivers.
>>    */
>>   #define UIO_MEM_DMA_COHERENT	5
>> +#define UIO_MEM_PHYS_WC	6
>>   
>>   /* defines for uio_port->porttype */
>>   #define UIO_PORT_NONE	0
> But you have no in-kernel user for this, so why add it?  We really can't
> take this unless there is such a user, sorry.

Understandable. I was planning to add some flag to uio_pdrv_genirq, but there 
don't appear to be existing bindings or interfaces to add a write-combine flag 
("ramoops" seems to be the only one that does that).


>
> thanks,
>
> greg k-h



Met vriendelijke groet / kind regards,

Mike Looijmans
System Expert


TOPIC Embedded Products B.V.
Materiaalweg 4, 5681 RJ Best
The Netherlands

T: +31 (0) 499 33 69 69
E: mike.looijmans@topic.nl
W: www.topic.nl

Please consider the environment before printing this e-mail