Re: [PATCH 4.14 00/57] 4.14.324-rc1 review

[Nathan Lynch]

Daniel Díaz <daniel.diaz@xxxxxxxxxx> writes:
>
> We see this build regression on PowerPC with GCC-8 and GCC-12:
> -----8<-----
>    /builds/linux/arch/powerpc/kernel/rtas_flash.c: In function 'rtas_flash_init':
>    /builds/linux/arch/powerpc/kernel/rtas_flash.c:717:22: error: implicit declaration of function 'kmem_cache_create_usercopy'; did you mean 'kmem_cache_create'? [-Werror=implicit-function-declaration]
>      flash_block_cache = kmem_cache_create_usercopy("rtas_flash_cache",
>                          ^~~~~~~~~~~~~~~~~~~~~~~~~~
>                          kmem_cache_create
>    /builds/linux/arch/powerpc/kernel/rtas_flash.c:717:20: error: assignment to 'struct kmem_cache *' from 'int' makes pointer from integer without a cast [-Werror=int-conversion]
>      flash_block_cache = kmem_cache_create_usercopy("rtas_flash_cache",
>                        ^
> ----->8-----
>
> That's on defconfig and cell_defconfig.
>
> Bisection points to "powerpc/rtas_flash: allow user copy to flash
> block cache objects" (5190538c66e5). Reverting that patch makes the
> build pass again.

That change should not be applied to 4.14.x. The problem it fixes is not
present in that version.