Re: [PATCH v4 2/2] RISC-V: Implement sparsemem

From: Greentime Hu <green.hu@gmail.com>
To: Logan Gunthorpe <logang@deltatee.com>
Cc: Paul Walmsley <paul.walmsley@sifive.com>,
	Rob Herring <robh@kernel.org>, Albert Ou <aou@eecs.berkeley.edu>,
	Andrew Waterman <andrew@sifive.com>,
	Palmer Dabbelt <palmer@sifive.com>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	Stephen Bates <sbates@raithlin.com>,
	Olof Johansson <olof@lixom.net>,
	greentime.hu@sifive.com, linux-riscv@lists.infradead.org,
	Michael Clark <michaeljclark@mac.com>,
	Christoph Hellwig <hch@lst.de>,
	linux-mm@vger.kernel.org
Subject: Re: [PATCH v4 2/2] RISC-V: Implement sparsemem
Date: Wed, 14 Aug 2019 21:35:55 +0800	[thread overview]
Message-ID: <CAEbi=3f+JDywuHYspfYKuC8z2wm8inRenBz+3DYbKK3ixFjU_g@mail.gmail.com> (raw)
In-Reply-To: <e1f78a33-18bb-bd6e-eede-e5e86758a4d0@deltatee.com>

Logan Gunthorpe <logang@deltatee.com> 於 2019年8月14日 週三 上午12:50寫道：
>
> On 2019-08-13 10:39 a.m., Paul Walmsley wrote:
> > On Tue, 13 Aug 2019, Logan Gunthorpe wrote:
> >
> >> On 2019-08-13 12:04 a.m., Greentime Hu wrote:
> >>
> >>> Every architecture with mmu defines their own pfn_valid().
> >>
> >> Not true. Arm64, for example just uses the generic implementation in
> >> mmzone.h.
> >
> > arm64 seems to define their own:
> >
> > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/arm64/Kconfig#n899
>
> Oh, yup. My mistake.
>
> > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/arm64/mm/init.c#n235
> >
> > While there are many architectures which have their own pfn_valid();
> > oddly, almost none of them set HAVE_ARCH_PFN_VALID ?
>
> Yes, much of this is super confusing. Seems HAVE_ARCH_PFN_VALID only
> matters if SPARSEMEM is set. So risc-v probably doesn't need to set it
> and we just need a #ifdef !CONFIG_FLATMEM around the pfn_valid
> definition like other arches.
>

Maybe this commit explains why it used HAVE_ARCH_PFN_VALID instead of SPARSEMEM.
https://github.com/torvalds/linux/commit/7b7bf499f79de3f6c85a340c8453a78789523f85

BTW, I found another issue here.
#define FIXADDR_TOP            (VMALLOC_START)
#define FIXADDR_START           (FIXADDR_TOP - FIXADDR_SIZE)
#define VMEMMAP_END    (VMALLOC_START - 1)
#define VMEMMAP_START  (VMALLOC_START - VMEMMAP_SIZE)
These 2 regions are overlapped.

How about this fix? Not sure if it is good for everyone.

diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
index 3f12b069af1d..3c4d394679d0 100644
--- a/arch/riscv/Kconfig
+++ b/arch/riscv/Kconfig
@@ -115,9 +115,6 @@ config PGTABLE_LEVELS
        default 3 if 64BIT
        default 2

-config HAVE_ARCH_PFN_VALID
-       def_bool y
-
 menu "Platform type"

 choice
diff --git a/arch/riscv/include/asm/fixmap.h b/arch/riscv/include/asm/fixmap.h
index c207f6634b91..72e106b60bc5 100644
--- a/arch/riscv/include/asm/fixmap.h
+++ b/arch/riscv/include/asm/fixmap.h
@@ -26,7 +26,7 @@ enum fixed_addresses {
 };

 #define FIXADDR_SIZE           (__end_of_fixed_addresses * PAGE_SIZE)
-#define FIXADDR_TOP            (VMALLOC_START)
+#define FIXADDR_TOP            (VMEMMAP_START)
 #define FIXADDR_START          (FIXADDR_TOP - FIXADDR_SIZE)

 #define FIXMAP_PAGE_IO         PAGE_KERNEL
diff --git a/arch/riscv/include/asm/page.h b/arch/riscv/include/asm/page.h
index 8ddb6c7fedac..83830997dce6 100644
--- a/arch/riscv/include/asm/page.h
+++ b/arch/riscv/include/asm/page.h
@@ -100,8 +100,10 @@ extern unsigned long min_low_pfn;
 #define page_to_bus(page)      (page_to_phys(page))
 #define phys_to_page(paddr)    (pfn_to_page(phys_to_pfn(paddr)))

+#if defined(CONFIG_FLATMEM)
 #define pfn_valid(pfn) \
        (((pfn) >= pfn_base) && (((pfn)-pfn_base) < max_mapnr))
+#endif

 #define ARCH_PFN_OFFSET                (pfn_base)
