Re: [linux-next:master 1617/1848] include/linux/compiler_types.h:630:45: error: call to '__compiletime_assert_612' declared with attribute error: min(((pos + len - 1) >> 12) - (pos >> 12) + 1, max_pages) signedness error

[David Laight <david.laight.linux@gmail.com>]

On Tue, 16 Dec 2025 09:05:42 +0100
kernel test robot <lkp@intel.com> wrote:

> tree:   https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git master
> head:   563c8dd425b59e44470e28519107b1efc99f4c7b
> commit: 0f5bb0cfb0b40a31d2fe146ecbef5727690fa547 [1617/1848] fs: use min() or umin() instead of min_t()
> config: i386-randconfig-2006-20250825 (https://download.01.org/0day-ci/archive/20251216/202512160948.O7QqxHj2-lkp@intel.com/config)
> compiler: gcc-14 (Debian 14.2.0-19) 14.2.0
> reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20251216/202512160948.O7QqxHj2-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/202512160948.O7QqxHj2-lkp@intel.com/
> 
> All errors (new ones prefixed by >>):
> 
>    In file included from <command-line>:
>    In function 'fuse_wr_pages',
>        inlined from 'fuse_perform_write' at fs/fuse/file.c:1347:27:
> >> include/linux/compiler_types.h:630:45: error: call to '__compiletime_assert_612' declared with attribute error: min(((pos + len - 1) >> 12) - (pos >> 12) + 1, max_pages) signedness error  
...

The definitions are:

loff_t pos  - 64bit signed.
size_t len  - unsigned long, 32bit for this i386 build.
unsigned int max_pages - 32 bit unsigned.

On a 64bit build 'len' is 64bits unsigned which promotes the LHS to
   unsigned and min() is happy.
On a 32bit build this doesn't happen, the LHS remains signed and
   min() is unhappy.

I'm not sure why file offsets are signed (apart from relative lseek()),
causes signedness issues in a few places.

In any case there are a few ways to fix this.
Clearly a cast could be used somewhere, the most subtle would be changing
the prototype to 'u64 len'.

Perhaps better is rewriting the conditional as:
	min(((len + (pos & (PAGE_SIZE - 1)) - 1) >> PAGE_SHIFT) + 1,
		max_pages);
Although the LHS is still signed 64bit, the compiler knows it can't
be negative.

This still contains a lot of 64bit maths, adding a cast:
	min(((len + (size_t)(pos & (PAGE_SIZE - 1)) - 1) >> PAGE_SHIFT) + 1,
makes it 32bit and the generated code much, much better.

In theory the expression can overflow (fixable by masking off the high bit
of 'len' (eg (len << 1 >> 1)), but normal IO are limited to MAX_RW_COUNT
(INT_MAX & PAGE_MASK) even on 64bit.
I'm not sure io_uring enforces that limit (it should, a lot of drivers are
broken if the requested size exceeds 4G).

	David