On Wed, Feb 12, 2025 at 7:21 AM Dan Carpenter <dan.carpenter@xxxxxxxxxx> wrote:
>
> Smatch complains that "next" could be uninitialized. The "next"
> assignment was accidentally left out when we moved these lines to earlier
> in the function.
>
> Fixes: bdadaec1526d ("x86/mm: make register_page_bootmem_memmap handle PTE mappings")
> Signed-off-by: Dan Carpenter <dan.carpenter@xxxxxxxxxx>
> ---
> This goes through the -mm tree.
>
> arch/x86/mm/init_64.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/arch/x86/mm/init_64.c b/arch/x86/mm/init_64.c
> index e7572af639a4..6e8e4ef5312a 100644
> --- a/arch/x86/mm/init_64.c
> +++ b/arch/x86/mm/init_64.c
> @@ -1600,8 +1600,10 @@ void register_page_bootmem_memmap(unsigned long section_nr,
> get_page_bootmem(section_nr, pud_page(*pud), MIX_SECTION_INFO);
>
> pmd = pmd_offset(pud, addr);
> - if (pmd_none(*pmd))
> + if (pmd_none(*pmd)) {
> + next = (addr + PAGE_SIZE) & PAGE_MASK;
> continue;
> + }
>
> if (!boot_cpu_has(X86_FEATURE_PSE) || !pmd_leaf(*pmd)) {
> next = (addr + PAGE_SIZE) & PAGE_MASK;
> --
> 2.47.2
>
Thanks for catching that Dan. I believe Andrew took the series out of
mm-unstable because of some conflicts, and asked me to do a v4 for
mm-unstable. Would you mind if I just folded your change in to the v4
series?
As an aside, it seems that this function could use some cleanup. It
seems wrong to only advance by PAGE_SIZE when you encounter an
unpopulated p4d/pgd/pud/pmd. It should advance to the end of that
p4d/pgd/pud/pmd. I suppose that case won't happen in practice, though,
which is also why this hasn't caused me any issues.
- Frank
