Re: arch/m68k/include/asm/mmu_context.h:164 load_ksp_mmu() warn: unsigned 'mmuar' is never less than zero.

[Hugh Dickins <hughd@google.com>]

Hi Geert,

On Thu, 6 Jul 2023, kernel test robot wrote:

> tree:   https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master
> head:   c17414a273b81fe4e34e11d69fc30cc8b1431614
> commit: e67b37c368b7cc24b8c0fe5ab6c44422312eab37 m68k: allow pte_offset_map[_lock]() to fail
> date:   2 weeks ago
> config: m68k-randconfig-m031-20230706 (https://download.01.org/0day-ci/archive/20230706/202307061849.rJKwVbXb-lkp@intel.com/config)
> compiler: m68k-linux-gcc (GCC) 12.3.0
> reproduce: (https://download.01.org/0day-ci/archive/20230706/202307061849.rJKwVbXb-lkp@intel.com/reproduce)
> 
> If you fix the issue in a separate patch/commit (i.e. not just a new version of
> the same patch/commit), kindly add following tags
> | Reported-by: kernel test robot <lkp@intel.com>
> | Closes: https://lore.kernel.org/oe-kbuild-all/202307061849.rJKwVbXb-lkp@intel.com/
> 
> New smatch warnings:
> arch/m68k/include/asm/mmu_context.h:164 load_ksp_mmu() warn: unsigned 'mmuar' is never less than zero.
> 
> Old smatch warnings:
> arch/m68k/include/asm/mmu_context.h:114 load_ksp_mmu() warn: always true condition '(mmuar >= (0)) => (0-u32max >= 0)'
> arch/m68k/include/asm/mmu_context.h:140 load_ksp_mmu() warn: always true condition '(mmuar >= (0)) => (0-u32max >= 0)'
> 
> vim +/mmuar +164 arch/m68k/include/asm/mmu_context.h
> 
>     92	
>     93	static inline void load_ksp_mmu(struct task_struct *task)
>     94	{
>     95		unsigned long flags;
>     96		struct mm_struct *mm;
>     97		int asid;
>     98		pgd_t *pgd;
>     99		p4d_t *p4d;
>    100		pud_t *pud;
>    101		pmd_t *pmd;
>    102		pte_t *pte = NULL;
>    103		unsigned long mmuar;
>    104	
>    105		local_irq_save(flags);
>    106		mmuar = task->thread.ksp;
>    107	
>    108		/* Search for a valid TLB entry, if one is found, don't remap */
>    109		mmu_write(MMUAR, mmuar);
>    110		mmu_write(MMUOR, MMUOR_STLB | MMUOR_ADR);
>    111		if (mmu_read(MMUSR) & MMUSR_HIT)
>    112			goto end;
>    113	
>    114		if (mmuar >= PAGE_OFFSET) {
>    115			mm = &init_mm;
>    116		} else {
>    117			pr_info("load_ksp_mmu: non-kernel mm found: 0x%p\n", task->mm);
>    118			mm = task->mm;
>    119		}
>    120	
>    121		if (!mm)
>    122			goto bug;
>    123	
>    124		pgd = pgd_offset(mm, mmuar);
>    125		if (pgd_none(*pgd))
>    126			goto bug;
>    127	
>    128		p4d = p4d_offset(pgd, mmuar);
>    129		if (p4d_none(*p4d))
>    130			goto bug;
>    131	
>    132		pud = pud_offset(p4d, mmuar);
>    133		if (pud_none(*pud))
>    134			goto bug;
>    135	
>    136		pmd = pmd_offset(pud, mmuar);
>    137		if (pmd_none(*pmd))
>    138			goto bug;
>    139	
>    140		pte = (mmuar >= PAGE_OFFSET) ? pte_offset_kernel(pmd, mmuar)
>    141					     : pte_offset_map(pmd, mmuar);
>    142		if (!pte || pte_none(*pte) || !pte_present(*pte))
>    143			goto bug;
>    144	
>    145		set_pte(pte, pte_mkyoung(*pte));
>    146		asid = mm->context & 0xff;
>    147		if (!pte_dirty(*pte) && mmuar <= PAGE_OFFSET)
>    148			set_pte(pte, pte_wrprotect(*pte));
>    149	
>    150		mmu_write(MMUTR, (mmuar & PAGE_MASK) | (asid << MMUTR_IDN) |
>    151			(((int)(pte->pte) & (int)CF_PAGE_MMUTR_MASK)
>    152			>> CF_PAGE_MMUTR_SHIFT) | MMUTR_V);
>    153	
>    154		mmu_write(MMUDR, (pte_val(*pte) & PAGE_MASK) |
>    155			((pte->pte) & CF_PAGE_MMUDR_MASK) | MMUDR_SZ_8KB | MMUDR_X);
>    156	
>    157		mmu_write(MMUOR, MMUOR_ACC | MMUOR_UAA);
>    158	
>    159		goto end;
>    160	
>    161	bug:
>    162		pr_info("ksp load failed: mm=0x%p ksp=0x08%lx\n", mm, mmuar);
>    163	end:
>  > 164		if (pte && mmuar < PAGE_OFFSET)
>    165			pte_unmap(pte);
>    166		local_irq_restore(flags);
>    167	}
>    168	
> 
> -- 
> 0-DAY CI Kernel Test Service
> https://github.com/intel/lkp-tests/wiki

I'm sorry, it appears that I've increased the number of smatch warnings on
m68k: but since it was already complaining about "mmuar >= PAGE_OFFSET"s
there, I'm not going to feel very guilty about my "mmuar < PAGE_OFFSET";
and I've no idea of the significance of PAGE_OFFSET 0 on m68k.  Over to
you - but I expect ignoring this will continue to be the right answer :)

Thanks,
Hugh