linux-kernel - Re: [PATCH v5 4/4] mm: Mark pages in use for page tables

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Mon, 18 Jun 2018 00:09:31 +0900
From:   Stafford Horne <shorne@...il.com>
To:     Matthew Wilcox <willy@...radead.org>
Cc:     linux-mm@...ck.org, Matthew Wilcox <mawilcox@...rosoft.com>,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH v5 4/4] mm: Mark pages in use for page tables

On Wed, Mar 07, 2018 at 05:44:43AM -0800, Matthew Wilcox wrote:
> From: Matthew Wilcox <mawilcox@...rosoft.com>
> 
> Define a new PageTable bit in the page_type and use it to mark pages in
> use as page tables.  This can be helpful when debugging crashdumps or
> analysing memory fragmentation.  Add a KPF flag to report these pages
> to userspace and update page-types.c to interpret that flag.
> 
> Note that only pages currently accounted as NR_PAGETABLES are tracked
> as PageTable; this does not include pgd/p4d/pud/pmd pages.  Those will
> be the subject of a later patch.
> 
> Signed-off-by: Matthew Wilcox <mawilcox@...rosoft.com>
> Acked-by: Kirill A. Shutemov <kirill.shutemov@...ux.intel.com>
> ---
>  arch/tile/mm/pgtable.c                 | 3 +++
>  fs/proc/page.c                         | 2 ++
>  include/linux/mm.h                     | 2 ++
>  include/linux/page-flags.h             | 6 ++++++
>  include/uapi/linux/kernel-page-flags.h | 1 +
>  tools/vm/page-types.c                  | 1 +
>  6 files changed, 15 insertions(+)
> 


Helloi Matthew,

I have bisected a regression on OpenRISC in v4.18-rc1 to this commit.  Using
our defconfig after boot I am getting:

    BUG: Bad page state in process hostname  pfn:00b5c
    page:c1ff0b80 count:0 mapcount:-1024 mapping:00000000 index:0x0
    flags: 0x0()
    raw: 00000000 00000000 00000000 fffffbff 00000000 00000100 00000200 00000000
    page dumped because: nonzero mapcount
    Modules linked in:
    CPU: 1 PID: 38 Comm: hostname Tainted: G    B
    4.17.0-simple-smp-07461-g1d40a5ea01d5-dirty #993
    Call trace:
    [<(ptrval)>] show_stack+0x44/0x54
    [<(ptrval)>] dump_stack+0xb0/0xe8
    [<(ptrval)>] bad_page+0x138/0x174
    [<(ptrval)>] ? ipi_icache_page_inv+0x0/0x24
    [<(ptrval)>] ? cpumask_next+0x24/0x34
    [<(ptrval)>] free_pages_check_bad+0x6c/0xd0
    [<(ptrval)>] free_pcppages_bulk+0x174/0x42c
    [<(ptrval)>] free_unref_page_commit.isra.17+0xb8/0xc8
    [<(ptrval)>] free_unref_page_list+0x10c/0x190
    [<(ptrval)>] ? set_reset_devices+0x0/0x2c
    [<(ptrval)>] release_pages+0x3a0/0x414
    [<(ptrval)>] tlb_flush_mmu_free+0x5c/0x90
    [<(ptrval)>] tlb_flush_mmu+0x90/0xa4
    [<(ptrval)>] arch_tlb_finish_mmu+0x50/0x94
    [<(ptrval)>] tlb_finish_mmu+0x30/0x64
    [<(ptrval)>] exit_mmap+0x110/0x1e0
    [<(ptrval)>] mmput+0x50/0xf0
    [<(ptrval)>] do_exit+0x274/0xa94
    [<(ptrval)>] ? _raw_spin_unlock_irqrestore+0x1c/0x2c
    [<(ptrval)>] ? __up_read+0x70/0x88
    [<(ptrval)>] do_group_exit+0x50/0x110
    [<(ptrval)>] __wake_up_parent+0x0/0x38
    [<(ptrval)>] _syscall_return+0x0/0x4


In this series we are overloading mapcount with page_type, the above is caused
due to this check in mm/page_alloc.c (free_pages_check_bad):

        if (unlikely(atomic_read(&page->_mapcount) != -1))
                bad_reason = "nonzero mapcount";

We can see in the dump above that _mapcount is fffffbff, this corresponds to the
'PG_table' flag.  Which was added here.  But it seems for some case in openrisc
its not getting cleared during page free.

This is as far as I got tracing it.  It might be an issue with OpenRISC, but our
implementation is mostly generic.  I will look into it more in the next few days
but I figured you might be able to spot something more quickly.

-Stafford