Re: [PATCH] hmm-tests: Fix migrate_dirty_page test

[David Hildenbrand <david@redhat.com>]

On 13.09.22 10:20, Alistair Popple wrote:
> 
> David Hildenbrand <david@redhat.com> writes:
> 
>> On 13.09.22 07:22, Alistair Popple wrote:
>>> As noted by John Hubbard the original test relied on side effects of the
>>> implementation of migrate_vma_setup() to detect if pages had been
>>> swapped to disk or not. This is subject to change in future so
>>> explicitly check for swap entries via pagemap instead. Fix a spelling
>>> mistake while we're at it.
>>> Signed-off-by: Alistair Popple <apopple@nvidia.com>
>>> Fixes: 5cc88e844e87 ("selftests/hmm-tests: add test for dirty bits")
>>> ---
>>>    tools/testing/selftests/vm/hmm-tests.c | 50 +++++++++++++++++++++++---
>>>    1 file changed, 46 insertions(+), 4 deletions(-)
>>> diff --git a/tools/testing/selftests/vm/hmm-tests.c
>>> b/tools/testing/selftests/vm/hmm-tests.c
>>> index 70fdb49b59ed..b5f6a7dc1f12 100644
>>> --- a/tools/testing/selftests/vm/hmm-tests.c
>>> +++ b/tools/testing/selftests/vm/hmm-tests.c
>>> @@ -1261,9 +1261,47 @@ static int destroy_cgroup(void)
>>>    	return 0;
>>>    }
>>>    +static uint64_t get_pfn(int fd, uint64_t ptr)
>>> +{
>>> +	uint64_t pfn;
>>> +	int ret;
>>> +
>>> +	ret = pread(fd, &pfn, sizeof(ptr),
>>> +		(uint64_t) ptr / getpagesize() * sizeof(ptr));
>>> +	if (ret != sizeof(ptr))
>>> +		return 0;
>>> +
>>> +	return pfn;
>>> +}
>>> +
>>> +#define PAGEMAP_SWAPPED (1ULL << 62)
>>> +
>>> +/* Returns true if at least one page in the range is on swap */
>>> +static bool pages_swapped(void *ptr, unsigned long pages)
>>> +{
>>> +	uint64_t pfn;
>>> +	int fd = open("/proc/self/pagemap", O_RDONLY);
>>> +	unsigned long i;
>>> +
>>> +	if (fd < 0)
>>> +		return false;
>>> +
>>> +	for (i = 0; i < pages; i++) {
>>> +		pfn = get_pfn(fd, (uint64_t) ptr + i * getpagesize());
>>> +
>>> +		if (pfn & PAGEMAP_SWAPPED) {
>>> +			close(fd);
>>> +			return true;
>>> +		}
>>
>> We do have pagemap_get_entry() in vm_util.c to query the pagemap entry.
> 
> Thanks. I'd missed that, although `grep pagemap
> tools/testing/selftests/vm` suggests I'm not the first to follow a
> tradition of open-coding this :-)
> 
> But there's no need to perpetuate that tradition, so will redo this to
> use vm_util.c instead.

Yeah, we just recently factored stuff out into there. I'll be factoring 
out more in my upcoming tests from the madv_populate tests.

-- 
Thanks,

David / dhildenb