Re: [PATCH] RISC-V: Handle R_RISCV_32 in modules

[Andreas Schwab]

On Jun 08 2018, Palmer Dabbelt <palmer@xxxxxxxxxx> wrote:

> On Thu, 07 Jun 2018 03:27:27 PDT (-0700), schwab@xxxxxxx wrote:
>> With CONFIG_MODVERSIONS=y the R_RISCV_32 relocation is used by the
>> __kcrctab section.
>>
>> Signed-off-by: Andreas Schwab <schwab@xxxxxxx>
>> ---
>>  arch/riscv/kernel/module.c | 12 ++++++++++++
>>  1 file changed, 12 insertions(+)
>>
>> diff --git a/arch/riscv/kernel/module.c b/arch/riscv/kernel/module.c
>> index 5dddba301d..1d5e9b934b 100644
>> --- a/arch/riscv/kernel/module.c
>> +++ b/arch/riscv/kernel/module.c
>> @@ -17,6 +17,17 @@
>>  #include <linux/errno.h>
>>  #include <linux/moduleloader.h>
>>
>> +static int apply_r_riscv_32_rela(struct module *me, u32 *location, Elf_Addr v)
>> +{
>> +	if (v != (u32)v) {
>
> My worry with this kind of check is that it relies on some sort of
> undefined behavior in C and that at some point in the future GCC will just
> go decide the check can never fail.  I checked and GCC doesn't elide these
> checks now, so I might be wrong.
>
> Is this defined to do what it looks like it's doing?

This is unsigned arithmetic, thus fully defined.

Andreas.

-- 
Andreas Schwab, SUSE Labs, schwab@xxxxxxx
GPG Key fingerprint = 0196 BAD8 1CE9 1970 F4BE  1748 E4D4 88E3 0EEA B9D7
"And now for something completely different."