Re: [PATCH] Resource: fix region_intersects() for CXL memory

[David Hildenbrand <david@redhat.com>]

On 16.08.24 07:07, Dan Williams wrote:
> [ add David ]
> 
> Andrew Morton wrote:
>> On Fri, 16 Aug 2024 10:07:23 +0800 Huang Ying <ying.huang@intel.com> wrote:
>>
>>> On a system with CXL memory installed, the resource tree (/proc/iomem)
>>> related to CXL memory looks like something as follows.
>>>
>>> 490000000-50fffffff : CXL Window 0
>>>    490000000-50fffffff : region0
>>>      490000000-50fffffff : dax0.0
>>>        490000000-50fffffff : System RAM (kmem)
>>>
>>> When the following command line is run to try writing some memory in
>>> CXL memory range,
>>>
>>>   $ dd if=data of=/dev/mem bs=1k seek=19136512 count=1
>>>   dd: error writing '/dev/mem': Bad address
>>>   1+0 records in
>>>   0+0 records out
>>>   0 bytes copied, 0.0283507 s, 0.0 kB/s
>>>
>>> the command fails as expected.  However, the error code is wrong.  It
>>> should be "Operation not permitted" instead of "Bad address".  And,
>>> the following warning is reported in kernel log.
>>>
>>>   ioremap on RAM at 0x0000000490000000 - 0x0000000490000fff
>>>   WARNING: CPU: 2 PID: 416 at arch/x86/mm/ioremap.c:216 __ioremap_caller.constprop.0+0x131/0x35d

But we should definitely fix the warning.

>>> ...
>>>
>>
>> Presumably we want to fix earlier kernels?  If so, are you able to
>> identify a suitable Fixes: target?  Possibly 974854ab0728 ("cxl/acpi:
>> Track CXL resources in iomem_resource")?
> 
> At least that commit, but I think this problem potentially goes back
> farther to:
> 
> c221c0b0308f device-dax: "Hotplug" persistent memory for use like normal RAM
> 
> ...because that started the era of "System RAM" as a non-top-level
> resource.
> 
> David did a bunch of work to fix this back in:
> 
> 97f61c8f44ec kernel/resource: make walk_system_ram_res() find all busy IORESOURCE_SYSTEM_RAM resources
> 
> ..but the fallout in region_intersects() was missed.

Sounds reasonable.

For virtio-mem we set IORESOURCE_SYSTEM_RAM|IORESOURCE_EXCLUSIVE on our 
(highest) parent resource (to make any /dev/mem access attempts of that 
memory fail). So the problem is likely specific to other
add_memory_driver_managed() users.

I have a faint recollection that at some point we had code that would 
set IORESOURCE_SYSTEM_RAM on parent resources up the tree, but either my 
memory is wrong or that code was ripped out long ago.

Fix idea is reasonable: check if anything in that range is 
IORESOURCE_SYSTEM_RAM.

-- 
Cheers,

David / dhildenb