Re: [PATCH v3 1/3] mm/swapfile: unuse_pte can map random data if swap read fails

From: Miaohe Lin <linmiaohe@huawei.com>
To: David Hildenbrand <david@redhat.com>,
	"ying.huang@intel.com" <ying.huang@intel.com>
Cc: <willy@infradead.org>, <vbabka@suse.cz>, <dhowells@redhat.com>,
	<neilb@suse.de>, <apopple@nvidia.com>, <surenb@google.com>,
	<minchan@kernel.org>, <peterx@redhat.com>, <sfr@canb.auug.org.au>,
	<naoya.horiguchi@nec.com>, <linux-mm@kvack.org>,
	<linux-kernel@vger.kernel.org>, Tim C Chen <tim.c.chen@intel.com>,
	Andrew Morton <akpm@linux-foundation.org>
Subject: Re: [PATCH v3 1/3] mm/swapfile: unuse_pte can map random data if swap read fails
Date: Mon, 25 Apr 2022 16:51:02 +0800	[thread overview]
Message-ID: <63c96d54-f16b-4a39-a665-227cc194cf96@huawei.com> (raw)
In-Reply-To: <a64b1987-1f98-db78-c0e4-189690d7a45d@redhat.com>

On 2022/4/25 16:01, David Hildenbrand wrote:
> On 25.04.22 09:55, ying.huang@intel.com wrote:
>> On Mon, 2022-04-25 at 09:49 +0200, David Hildenbrand wrote:
>>> On 25.04.22 09:41, ying.huang@intel.com wrote:
>>>> Hi, Miaohe,
>>>>
>>>> On Sun, 2022-04-24 at 17:11 +0800, Miaohe Lin wrote:
>>>>> There is a bug in unuse_pte(): when swap page happens to be unreadable,
>>>>> page filled with random data is mapped into user address space.  In case
>>>>> of error, a special swap entry indicating swap read fails is set to the
>>>>> page table.  So the swapcache page can be freed and the user won't end up
>>>>> with a permanently mounted swap because a sector is bad.  And if the page
>>>>> is accessed later, the user process will be killed so that corrupted data
>>>>> is never consumed.  On the other hand, if the page is never accessed, the
>>>>> user won't even notice it.
>>>>>
>>>>> Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
>>>>> Acked-by: David Hildenbrand <david@redhat.com>
>>>>> ---
>>>>>  include/linux/swap.h    |  7 ++++++-
>>>>>  include/linux/swapops.h | 10 ++++++++++
>>>>>  mm/memory.c             |  5 ++++-
>>>>>  mm/swapfile.c           | 11 +++++++++++
>>>>>  4 files changed, 31 insertions(+), 2 deletions(-)
>>>>>
>>>>> diff --git a/include/linux/swap.h b/include/linux/swap.h
>>>>> index 5553189d0215..b82c196d8867 100644
>>>>> --- a/include/linux/swap.h
>>>>> +++ b/include/linux/swap.h
>>>>> @@ -55,6 +55,10 @@ static inline int current_is_kswapd(void)
>>>>>   * actions on faults.
>>>>>   */
>>>>>
>>>>> +#define SWP_SWAPIN_ERROR_NUM 1
>>>>> +#define SWP_SWAPIN_ERROR     (MAX_SWAPFILES + SWP_HWPOISON_NUM + \
>>>>> +			     SWP_MIGRATION_NUM + SWP_DEVICE_NUM + \
>>>>> +			     SWP_PTE_MARKER_NUM)
>>>>>
>>>>>
>>>>
>>>> It appears wasteful to use another swap device number. 
>>>
>>> Do we really care?
>>>
>>> We currently use 5 bits for swap types, so we have a total of 32.
>>>
>>> SWP_HWPOISON_NUM -> 1
>>> SWP_MIGRATION_NUM -> 3
>>> SWP_PTE_MARKER_NUM -> 1
>>> SWP_DEVICE_NUM -> 4
>>> SWP_SWAPIN_ERROR_NUM -> 1
>>>
>>> Which would leave us with 32 - 10 = 22 swap devices. IMHO that's plenty
>>> for real life scenarios.
>>
>> Creating multiple swap partitions on one disk can improve the
>> scalability of swap subsystem, although we usually don't have so many
>> disks for swap. 
> 
> Exactly, and IMHO if we have 22 or 23 doesn't make a real difference
> here ...

I tend to agree with David. Thanks both!

> 