Re: [PATCH v2 8/8] selftests/mm: report SKIP in pfnmap if a check fails

["David Hildenbrand (Red Hat)" <david@kernel.org>]

On 1/19/26 15:26, Ryan Roberts wrote:
> On 19/01/2026 11:16, David Hildenbrand (Red Hat) wrote:
>> On 1/7/26 17:48, Kevin Brodsky wrote:
>>> pfnmap currently checks the target file in FIXTURE_SETUP(pfnmap),
>>> meaning once for every test, and skips the test if any check fails.
>>>
>>> The target file is the same for every test so this is a little
>>> overkill. More importantly, this approach means that the whole suite
>>> will report PASS even if all the tests are skipped because kernel
>>> configuration (e.g. CONFIG_STRICT_DEVMEM=y) prevented /dev/mem from
>>> being mapped, for instance.
>>>
>>> Let's ensure that KSFT_SKIP is returned as exit code if any check
>>> fails by performing the checks in pfnmap_init(), run once. That
>>> function also takes care of finding the offset of the pages to be
>>> mapped and saves it in a global. The file is still mapped/unmapped
>>> for every test, as some of them modify the mapping.
>>>
>>> Signed-off-by: Kevin Brodsky <kevin.brodsky@arm.com>
>>> ---
>>>    tools/testing/selftests/mm/pfnmap.c | 81 ++++++++++++++++++++---------
>>>    1 file changed, 55 insertions(+), 26 deletions(-)
>>>
>>> diff --git a/tools/testing/selftests/mm/pfnmap.c b/tools/testing/selftests/mm/
>>> pfnmap.c
>>> index 35b0e3ed54cd..e41d5464130b 100644
>>> --- a/tools/testing/selftests/mm/pfnmap.c
>>> +++ b/tools/testing/selftests/mm/pfnmap.c
>>> @@ -25,8 +25,11 @@
>>>    #include "kselftest_harness.h"
>>>    #include "vm_util.h"
>>>    +#define DEV_MEM_NPAGES    2
>>> +
>>>    static sigjmp_buf sigjmp_buf_env;
>>>    static char *file = "/dev/mem";
>>> +static off_t file_offset;
>>>      static void signal_handler(int sig)
>>>    {
>>> @@ -88,7 +91,7 @@ static int find_ram_target(off_t *offset,
>>>                break;
>>>              /* We need two pages. */
>>> -        if (end > start + 2 * pagesize) {
>>> +        if (end > start + DEV_MEM_NPAGES * pagesize) {
>>>                fclose(file);
>>>                *offset = start;
>>>                return 0;
>>> @@ -97,9 +100,49 @@ static int find_ram_target(off_t *offset,
>>>        return -ENOENT;
>>>    }
>>>    +static void pfnmap_init(void)
>>> +{
>>> +    size_t pagesize = getpagesize();
>>> +    size_t size = DEV_MEM_NPAGES * pagesize;
>>> +    int fd;
>>> +    void *addr;
>>> +
>>> +    if (strncmp(file, "/dev/mem", strlen("/dev/mem")) == 0) {
>>> +        int err = find_ram_target(&file_offset, pagesize);
>>> +
>>> +        if (err)
>>> +            ksft_exit_skip("Cannot find ram target in '/proc/iomem': %s\n",
>>> +                       strerror(-err));
>>> +    } else {
>>> +        file_offset = 0;
>>> +    }
>>> +
>>> +    /*
>>> +     * Make sure we can open and map the file, and perform some basic
>>> +     * checks; skip the whole suite if anything goes wrong.
>>> +     * A fresh mapping is then created for every test case by
>>> +     * FIXTURE_SETUP(pfnmap).
>>> +     */
>>> +    fd = open(file, O_RDONLY);
>>> +    if (fd < 0)
>>> +        ksft_exit_skip("Cannot open '%s': %s\n", file, strerror(errno));
>>> +
>>> +    addr = mmap(NULL, size, PROT_READ, MAP_SHARED, fd, file_offset);
>>> +    if (addr == MAP_FAILED)
>>> +        ksft_exit_skip("Cannot mmap '%s': %s\n", file, strerror(errno));
>>> +
>>> +    if (!check_vmflag_pfnmap(addr))
>>> +        ksft_exit_skip("Invalid file: '%s'. Not pfnmap'ed\n", file);
>>> +
>>> +    if (test_read_access(addr, size))
>>> +        ksft_exit_skip("Cannot read-access mmap'ed '%s'\n", file);
>>> +
>>> +    munmap(addr, size);
>>
>> Why not keep the fd open then and supply that to all tests without the need for
>> them to open/close?
>>
>> Then, also the file cannot change etc.
> 
> I had a private conversation with Kevin about this before he posted; my very
> minor, theorectical concern about that was that it's possible to pass in a
> custom file to be pfnmapped and I wondered if such a file could map a device
> region that has read side effects? In that case I think you'd want to open it
> fresh for each test to ensure consistent starting state?

Are we aware of devices where we would actually require a new open, and 
not just a new mmap()?

The reason we added support for other files was "other pfnmap'ed memory 
like NVIDIA's EGM". I'd assume that people rather should not pass in 
something that has any side-effects.

> 
> But if you think that concern is unfounded, certainly just opening it once and
> reusing will simplify.

I would just keep it simple here, yes. If this ever becomes a real 
problem, my intuition would tell me that probably the caller is doing 
something unsupported that we just cannot easily identify+reject.

-- 
Cheers

David