Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20190402055902.14017-1-anup.patel@wdc.com> <20190402083456.GA2153@rapoport-lnx>
In-Reply-To: <20190402083456.GA2153@rapoport-lnx>
From:   Anup Patel <anup@brainfault.org>
Date:   Tue, 2 Apr 2019 15:00:02 +0530
Message-ID: <CAAhSdy1ZFW1zAqc-7904SOse-tWKOmpY3dCHb5Xe6U6KNno_JQ@mail.gmail.com>
Subject: Re: [PATCH] RISC-V: Fix Maximum Physical Memory 2GiB option for 64bit systems
To:     Mike Rapoport <rppt@linux.ibm.com>
Cc:     Anup Patel <Anup.Patel@wdc.com>,
        Palmer Dabbelt <palmer@sifive.com>,
        Albert Ou <aou@eecs.berkeley.edu>,
        Atish Patra <Atish.Patra@wdc.com>,
        Christoph Hellwig <hch@infradead.org>,
        Paul Walmsley <paul.walmsley@sifive.com>,
        "linux-riscv@lists.infradead.org" <linux-riscv@lists.infradead.org>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Tue, Apr 2, 2019 at 2:05 PM Mike Rapoport <rppt@linux.ibm.com> wrote:
>
> On Tue, Apr 02, 2019 at 06:02:38AM +0000, Anup Patel wrote:
> > The Maximum Physical Memory 2GiB option for 64bit systems is currently
> > broken because kernel hangs at boot-time when this option is enabled
> > and the underlying system has more than 2GiB memory.
> >
> > This issue can be easily reproduced on SiFive Unleashed board where
> > we have 8GiB of memory.
> >
> > This patch fixes above issue by reserving unusable memory region in
> > setup_bootmem().
> >
> > Signed-off-by: Anup Patel <anup.patel@wdc.com>
> > ---
> >  arch/riscv/mm/init.c | 8 ++++++++
> >  1 file changed, 8 insertions(+)
> >
> > diff --git a/arch/riscv/mm/init.c b/arch/riscv/mm/init.c
> > index 5fd8c922e1c2..6b063f20a9d0 100644
> > --- a/arch/riscv/mm/init.c
> > +++ b/arch/riscv/mm/init.c
> > @@ -121,6 +121,14 @@ void __init setup_bootmem(void)
> >                        */
> >                       memblock_reserve(reg->base, vmlinux_end - reg->base);
> >                       mem_size = min(reg->size, (phys_addr_t)-PAGE_OFFSET);
> > +
> > +                     /*
> > +                      * Reserve from the end of usable area to the end of
> > +                      * region
> > +                      */
> > +                     if ((reg->base + mem_size) < end)
> > +                             memblock_reserve(reg->base + mem_size,
> > +                                              end - reg->base - mem_size);
>
> The memory above MAXPHYSMEM should not be reserved. It should be either
> removed from memblock with memblock_remove or not added at the first place.
>
> Frankly, I fail to understand the logic behind setting PAGE_OFFSET to
> MAXPHYSMEM and then using PAGE_OFFSET as the limit for accessible physical
> memory. Still, as it is there, you can set MAX_MEMBLOCK_ADDR=PAGE_OFFSET in
> arch/riscv/include/page.h and then early_init_dt_add_memory_arch() will
> simply ignore the memory above PAGE_OFFSET.

Little explanation about current code ...

The current code, assumes PAGE_OFFSET to have value such that
upper-bits are 1s lower bits are 0s.
For example,
0xc0000000 (32bit),
0xffffffff80000000 (64bit), and
0xffffffe000000000 (64bit)

For above PAGE_OFFSET values, -PAGE_OFFSET is size of virtual
address space and maximum supported physical memory is also
-PAGE_OFFSET hence MAXPHYMEM is tied with -PAGE_OFFSET.

If we try to force some arbitrary PAGE_OFFSET then things break.

Regards,
Anup