[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAK8P3a1E1vP-vS=-aH-XP=5taSTy8EUTfNSpQk9x47WLRMCOzQ@mail.gmail.com>
Date: Wed, 27 Apr 2022 20:20:30 +0200
From: Arnd Bergmann <arnd@...db.de>
To: Kefeng Wang <wangkefeng.wang@...wei.com>
Cc: Catalin Marinas <catalin.marinas@....com>,
Will Deacon <will@...nel.org>,
Andrew Morton <akpm@...ux-foundation.org>,
Linux ARM <linux-arm-kernel@...ts.infradead.org>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
Linux-MM <linux-mm@...ck.org>
Subject: Re: [PATCH 2/4] mm: ioremap: Add arch_ioremap/iounmap_check()
On Wed, Apr 27, 2022 at 2:14 PM Kefeng Wang <wangkefeng.wang@...wei.com> wrote:
> @@ -964,6 +964,9 @@ static inline void iounmap(volatile void __iomem *addr)
> #elif defined(CONFIG_GENERIC_IOREMAP)
> #include <linux/pgtable.h>
>
> +bool arch_ioremap_check(phys_addr_t addr, size_t size, unsigned long prot);
> +bool arch_iounmap_check(void __iomem *addr);
> +
> void __iomem *ioremap_prot(phys_addr_t addr, size_t size, unsigned long prot);
> void iounmap(volatile void __iomem *addr);
>
> diff --git a/mm/ioremap.c b/mm/ioremap.c
> index 522ef899c35f..d1117005dcc7 100644
> --- a/mm/ioremap.c
> +++ b/mm/ioremap.c
> @@ -11,6 +11,16 @@
> #include <linux/io.h>
> #include <linux/export.h>
>
> +bool __weak arch_ioremap_check(phys_addr_t addr, size_t size, unsigned long prot)
> +{
> + return true;
> +}
> +
> +bool __weak arch_iounmap_check(void __iomem *addr)
> +{
> + return true;
> +}
> +
I don't really like the weak functions. The normal way to do this in
asm-generic headers
is to have something like
#ifndef arch_ioremap_check
static inline bool arch_ioremap_check(phys_addr_t addr, size_t size,
unsigned long prot)
{
return true;
}
#endif
and then in architectures that actually do some checking, have these
bits in asm/io.h
bool arch_ioremap_check(phys_addr_t addr, size_t size, unsigned long prot);
#define arch_ioremap_check arch_ioremap_check
(or alternatively an extern declaration, if the implementation is nontrivial)
It may be worth pointing out that either way requires including
asm-generic/io.h,
which most architectures don't. This is probably fine, as only csky, riscv and
now arm64 use CONFIG_GENERIC_IOREMAP, and we can probably require
that any further architectures using this symbol also have to use
asm-generic/io.h.
Arnd
Powered by blists - more mailing lists