Re: [PATCH] Fix pointer cast for 32 bits arch

[Geert Uytterhoeven]

On Sun, Apr 12, 2015 at 3:48 PM, Peter Senna Tschudin
<peter.senna@xxxxxxxxx> wrote:
> On Sun, Apr 12, 2015 at 3:05 PM, Geert Uytterhoeven
> <geert@xxxxxxxxxxxxxx> wrote:
>> On Sun, Apr 12, 2015 at 11:26 AM, Peter Senna Tschudin
>> <peter.senna@xxxxxxxxx> wrote:
>>> Sparse compalins about casting void * to u64 on i386.
>>> Change the cast to resource_size_t.
>>>
>>> Signed-off-by: Peter Senna Tschudin <peter.senna@xxxxxxxxx>
>>> ---
>>>
>>> Tested by compilation only. Tested for x86 and x86_64.
>>>
>>>  drivers/staging/goldfish/goldfish_nand.c | 3 ++-
>>>  include/linux/goldfish.h                 | 2 +-
>>>  2 files changed, 3 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/drivers/staging/goldfish/goldfish_nand.c b/drivers/staging/goldfish/goldfish_nand.c
>>> index d68f216..738fdc4 100644
>>> --- a/drivers/staging/goldfish/goldfish_nand.c
>>> +++ b/drivers/staging/goldfish/goldfish_nand.c
>>> @@ -87,7 +87,8 @@ static u32 goldfish_nand_cmd(struct mtd_info *mtd, enum nand_cmd cmd,
>>>                 writel((u32)(addr >> 32), base + NAND_ADDR_HIGH);
>>>                 writel((u32)addr, base + NAND_ADDR_LOW);
>>>                 writel(len, base + NAND_TRANSFER_SIZE);
>>> -               gf_write64((u64)ptr, base + NAND_DATA, base + NAND_DATA_HIGH);
>>> +               gf_write64((resource_size_t)ptr, base + NAND_DATA,
>>> +                          base + NAND_DATA_HIGH);
>>
>> I guess sparse still complains if CONFIG_X86_PAE=y, which makes
>> resource_size_t u64?
>
> Yes, when CONFIG_X86_PAE=y the patch doesn't fix the warning(gcc
> warning). What is the correct/portable way of fixing this?

As gf_write64() takes "unsigned long data" for its first argument, why not
simply cast ptr to "unsigned long"? "void *" and "unsigned long" have the same
size on all Linux platforms.

Seems like there a several silly casting tricks being done in the goldfish
drivers:

$ git grep -w gf_write64drivers/platform/goldfish/goldfish_pipe.c:
 gf_write64((u64)(unsigned long)p
drivers/platform/goldfish/goldfish_pipe.c:      gf_write64((u64)(unsigned long)p
drivers/platform/goldfish/goldfish_pipe.c:                      gf_write64((u64)
drivers/platform/goldfish/goldfish_pipe.c:                      gf_write64(addre
drivers/staging/goldfish/goldfish_audio.c:      (gf_write64((u64)(x), data->reg_
drivers/staging/goldfish/goldfish_nand.c:               gf_write64((u64)ptr, bas
drivers/tty/goldfish.c: gf_write64((u64)buf, base + GOLDFISH_TTY_DATA_PTR,
drivers/tty/goldfish.c: gf_write64((u64)buf, base + GOLDFISH_TTY_DATA_PTR,
include/linux/goldfish.h:static inline void gf_write64(unsigned long data,

with interesting history showing one wrong cast being "fixed" using another
wrong cast:

commit f4e131dc38d34469b9fee1840ad237324831bce3
Author: Octavian Purdila <octavian.purdila@xxxxxxxxx>
Date:   Fri May 16 08:24:59 2014 +0300

    goldfish: pipe: fix warnings for 32bit builds

-       gf_write64((u64)pipe, dev->base + PIPE_REG_CHANNEL,
+       gf_write64((u64)(unsigned long)pipe, dev->base + PIPE_REG_CHANNEL,

As in general casts don't belong in drivers[*], what about creating an inline
helper gf_set_pipe() that converts the pointer to unsigned long and calls
gf_write64()?

Perhaps checkpatch should complain about casts outside header files?

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/