Re: [Question] How to print size_t type variable?

From: Geert Uytterhoeven
Date: Tue Dec 16 2014 - 04:50:13 EST


Hi Yamada-san,

On Tue, Dec 16, 2014 at 4:05 AM, Masahiro Yamada
<yamada.m@xxxxxxxxxxxxxxxx> wrote:
> On Mon, 15 Dec 2014 12:00:25 +0100
> Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote:
>> On Mon, Dec 15, 2014 at 11:51 AM, Masahiro Yamada
>> <yamada.m@xxxxxxxxxxxxxxxx> wrote:
>> > Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote:
>> >> On Mon, Dec 15, 2014 at 10:32 AM, Masahiro Yamada
>> >> <yamada.m@xxxxxxxxxxxxxxxx> wrote:
>> >> > I read through Documentation/printk-formats.txt
>> >> >
>> >> > It clearly says to use "%zu" or "%zx" to print size_t variables,
>> >> > but I still have a question.
>> >> >
>> >> > Assume we have code something like:
>> >> >
>> >> > printk("%zx", (size_t)10);
>> >> >
>> >> > I think this code works fine as long as it includes
>> >> > the compiler-provided <stddef.h>.
>> >> >
>> >> > In the kernel space, however, <stddef.h> is never included.
>> >> > Instead, size_t is defined by include/linux/types.h
>> >> > and include/uapi/asm-generic/posix_types.h.
>> >> >
>> >> > That is, size_t is defined independently from the compiler you are using,
>> >> > although the compiler still decides which variable type is expected for the "%zx" format.
>> >>
>> >> That's correct.
>> >>
>> >> > This causes compiler warnings for some compilers.
>> >> >
>> >> > On bare-metal m68k toolchains, for example, size_t is "unsignd long",
>> >> > whearas it is "unsigned int" on kernel.org m68k toolchains.
>> >> >
>> >> > I see such warnings when I built the kernel with bare-metal m68k toolchains.
>> >> >
>> >> > $ git describe
>> >> > v3.18
>> >> > $ make ARCH=m68k CROSS_COMPILE=m68k-elf- defconfig all
>> >> > HOSTCC scripts/basic/fixdep
>> >> > HOSTCC scripts/kconfig/conf.o
>> >> > SHIPPED scripts/kconfig/zconf.tab.c
>> >> > SHIPPED scripts/kconfig/zconf.lex.c
>> >> > SHIPPED scripts/kconfig/zconf.hash.c
>> >> > HOSTCC scripts/kconfig/zconf.tab.o
>> >> > HOSTLD scripts/kconfig/conf
>> >> > *** Default configuration is based on 'multi_defconfig'
>> >> > kernel/time/Kconfig:163:warning: range is invalid
>> >> > #
>> >> > # configuration written to .config
>> >> > #
>> >> >
>> >> > [ snip ]
>> >> >
>> >> > LD init/mounts.o
>> >> > CC init/initramfs.o
>> >> > init/initramfs.c: In function 'populate_rootfs':
>> >> > init/initramfs.c:635:5: warning: format '%zd' expects argument of type 'signed size_t', but argument 2 has type 'ssize_t' [-Wformat]
>> >>
>> >> Please use a compiler configured for Linux, i.e. m68k-linux-*.
>> >
>> > Yes, I can use it, but I am still curious.
>> >
>> > Do we have a good reason to keep this limitation?
>> >
>> > (All the problem I could see for using GCC that was not configured for Linux
>> > was just the printk-related warnings.)
>> >
>> > Instead of hard-coding the size_t type,
>> > can we use compiler-provided __SIZE_TYPE__ (or include <stddef.h>) ??
>>
>> Note that cris and s390 used __SIZE_TYPE__ in kernel headers before,
>> but it caused other compiler warnings due to a mismatch in the base types
>> for size_t and ssize_t, cfr. https://lkml.org/lkml/2012/8/12/36.
>> AFAIK, there's no __SSIZE_TYPE__.
>
> I notice __kernel_(s)size_t has the same width as "long".
> (The kernel does not support LLP64. We just have to take LP64 into account.)

On 64-bit, __kernel_(s)size_t are indeed (unsigned) long
On 32-bit, __kernel_(s)size_t is usually (unsigned) int for historical
reasons.
Cfr. include/uapi/asm-generic/posix_types.h:

/*
* Most 32 bit architectures use "unsigned int" size_t,
* and all 64 bit architectures use "unsigned long" size_t.
*/
#ifndef __kernel_size_t
#if __BITS_PER_LONG != 64
typedef unsigned int __kernel_size_t;
typedef int __kernel_ssize_t;
typedef int __kernel_ptrdiff_t;
#else
typedef __kernel_ulong_t __kernel_size_t;
typedef __kernel_long_t __kernel_ssize_t;
typedef __kernel_long_t __kernel_ptrdiff_t;
#endif
#endif

The are a few exceptions (e.g. avr32 and blackfin)

> Perhaps, we should have hard-coded
>
> typedef __kernel_ulong_t __kernel_size_t;
> typedef __kernel_long_t __kernel_ssize_t;
>
> and used "%lx" to print (s)size_t type variables.
>
> I think we do not have a good reason to use "%zx",
> although it might be too late.

In hindsight, it would have been better for (s)size_t to match (unsigned) long,
removing the need for %z.
However, that was realized only after 64-bit architectures were introduced.
It's definitely to late to change this for existing 32-bit architectures.

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/