Re: [PATCH 04/12] mips: Reserve memory for the kernel image resources

From: Geert Uytterhoeven <geert@linux-m68k.org>
To: Serge Semin <fancer.lancer@gmail.com>
Cc: Ralf Baechle <ralf@linux-mips.org>,
	Paul Burton <paul.burton@mips.com>,
	James Hogan <jhogan@kernel.org>,
	Matt Redfearn <matt.redfearn@mips.com>,
	Mike Rapoport <rppt@linux.ibm.com>,
	Andrew Morton <akpm@linux-foundation.org>,
	Michal Hocko <mhocko@suse.com>,
	Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	Thomas Bogendoerfer <tbogendoerfer@suse.de>,
	Huacai Chen <chenhc@lemote.com>, Stefan Agner <stefan@agner.ch>,
	Stephen Rothwell <sfr@canb.auug.org.au>,
	Alexandre Belloni <alexandre.belloni@bootlin.com>,
	Juergen Gross <jgross@suse.com>,
	linux-mips@vger.kernel.org,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Subject: Re: [PATCH 04/12] mips: Reserve memory for the kernel image resources
Date: Tue, 21 May 2019 16:56:39 +0200	[thread overview]
Message-ID: <CAMuHMdWPmL5Z86cgJ4N-U-3XKr4ys8Y7U85okDcXYEu7z4ybaw@mail.gmail.com> (raw)
In-Reply-To: <20190423224748.3765-5-fancer.lancer@gmail.com>

Hi Serge,

On Wed, Apr 24, 2019 at 12:50 AM Serge Semin <fancer.lancer@gmail.com> wrote:
> The reserved_end variable had been used by the bootmem_init() code
> to find a lowest limit of memory available for memmap blob. The original
> code just tried to find a free memory space higher than kernel was placed.
> This limitation seems justified for the memmap ragion search process, but
> I can't see any obvious reason to reserve the unused space below kernel
> seeing some platforms place it much higher than standard 1MB. Moreover
> the RELOCATION config enables it to be loaded at any memory address.
> So lets reserve the memory occupied by the kernel only, leaving the region
> below being free for allocations. After doing this we can now discard the
> code freeing a space between kernel _text and VMLINUX_LOAD_ADDRESS symbols
> since it's going to be free anyway (unless marked as reserved by
> platforms).
>
> Signed-off-by: Serge Semin <fancer.lancer@gmail.com>

This is now commit b93ddc4f9156205e ("mips: Reserve memory for the kernel
image resources") in v5.2-rc1, which causes rbtx4927 to crash during boot:

    VFS: Mounted root (nfs filesystem) on device 0:13.
    devtmpfs: mounted
    BUG: Bad page state in process swapper  pfn:00001
    page:804b7820 refcount:0 mapcount:-128 mapping:00000000 index:0x1
    flags: 0x0()
    raw: 00000000 00000100 00000200 00000000 00000001 00000000 ffffff7f 00000000
    page dumped because: nonzero mapcount
    Modules linked in:
    CPU: 0 PID: 1 Comm: swapper Not tainted
5.2.0-rc1-rbtx4927-00468-g3c05ea3d4077b756-dirty #137
    Stack : 00000000 10008400 8040dd2c 87c1b974 8044af63 8040dd2c
00000001 804a3490
            00000001 81000000 0030f231 80148558 00000003 10008400
87c1dd80 7599ee13
            00000000 00000000 804b0000 00000000 00000007 00000000
00000085 00000000
            62722d31 00000084 804b0000 39347874 00000000 804b7820
8040cef8 81000010
            00000001 00000007 00000001 81000000 00000008 8021de24
00000000 804a0000
            ...
    Call Trace:
    [<8010adec>] show_stack+0x74/0x104
    [<801a5e44>] bad_page+0x130/0x138
    [<801a654c>] free_pcppages_bulk+0x17c/0x3b0
    [<801a789c>] free_unref_page+0x40/0x68
    [<801120f4>] free_init_pages+0xec/0x104
    [<803bdde8>] free_initmem+0x10/0x58
    [<803bdb8c>] kernel_init+0x20/0x100
    [<801057c8>] ret_from_kernel_thread+0x14/0x1c
    Disabling lock debugging due to kernel taint
    BUG: Bad page state in process swapper  pfn:00002
    [...]

CONFIG_RELOCATABLE is not set, so the only relevant part is the
change quoted below.

> --- a/arch/mips/kernel/setup.c
> +++ b/arch/mips/kernel/setup.c
> @@ -371,7 +371,6 @@ static void __init bootmem_init(void)
>
>  static void __init bootmem_init(void)
>  {
> -       unsigned long reserved_end;
>         phys_addr_t ramstart = PHYS_ADDR_MAX;
>         int i;
>
> @@ -382,10 +381,10 @@ static void __init bootmem_init(void)
>          * will reserve the area used for the initrd.
>          */
>         init_initrd();
> -       reserved_end = (unsigned long) PFN_UP(__pa_symbol(&_end));
>
> -       memblock_reserve(PHYS_OFFSET,
> -                        (reserved_end << PAGE_SHIFT) - PHYS_OFFSET);
> +       /* Reserve memory occupied by kernel. */
> +       memblock_reserve(__pa_symbol(&_text),
> +                       __pa_symbol(&_end) - __pa_symbol(&_text));
>
>         /*
>          * max_low_pfn is not a number of pages. The number of pages

With some debug code added:

    Determined physical RAM map:
     memory: 08000000 @ 00000000 (usable)
    bootmem_init:390: PHYS_OFFSET = 0x0
    bootmem_init:391: __pa_symbol(&_text) = 0x100000
    bootmem_init:392: __pa_symbol(&_end) = 0x4b77c8
    bootmem_init:393: PFN_UP(__pa_symbol(&_end)) = 0x4b8

Hence the old code reserved 1 MiB extra at the beginning.

Note that the new code also dropped the rounding up of the memory block
size to a multiple of PAGE_SIZE. I'm not sure the latter actually
matters or not.

Do you have a clue? Thanks!

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds