This patch series fix a few bugs in mm/zswap based on Linux-3.11.
v2 --> v3
- keep GFP_KERNEL flag
v1 --> v2
- free memory in zswap_frontswap_invalidate_area(in patch 1)
- fix whitespace corruption (line wrapping)
Corresponding mail thread: https://lkml.org/lkml/2013/8/18/59
These issues fixed/optimized are:
1. memory leaks when re-swapon
2. memory leaks when invalidate and reclaim occur concurrently
3. avoid unnecessary page scanning
Issues discussed in that mail thread NOT fixed as it happens rarely or
not a big problem or controversial:
1. a "theoretical race condition" when reclaim page
When a handle alloced from zbud, zbud considers this handle is used
validly by upper(zswap) and can be a candidate for reclaim. But zswap has
to initialize it such as setting swapentry and adding it to rbtree.
so there is a race condition, such as:
thread 0: obtain handle x from zbud_alloc
thread 1: zbud_reclaim_page is called
thread 1: callback zswap_writeback_entry to reclaim handle x
thread 1: get swpentry from handle x (it is random value now)
thread 1: bad thing may happen
thread 0: initialize handle x with swapentry
2. frontswap_map bitmap not cleared after zswap reclaim
Frontswap uses frontswap_map bitmap to track page in "backend" implementation,
when zswap reclaim a page, the corresponding bitmap record is not cleared.
3. the potential that zswap store and reclaim functions called recursively
mm/zswap.c | 28 ++++++++++++++++++++--------
1 file changed, 20 insertions(+), 8 deletions(-)
On Mon, Sep 23, 2013 at 04:19:36PM +0800, Weijie Yang wrote:
> This patch series fix a few bugs in mm/zswap based on Linux-3.11.
>
> v2 --> v3
> - keep GFP_KERNEL flag
Why do you drop this?
It's plain BUG. I read Bob's reply but it couldn't justify to let the pain
remain. First of all, let's fix it and better idea could come later.
>
> v1 --> v2
> - free memory in zswap_frontswap_invalidate_area(in patch 1)
> - fix whitespace corruption (line wrapping)
>
> Corresponding mail thread: https://lkml.org/lkml/2013/8/18/59
>
> These issues fixed/optimized are:
>
> 1. memory leaks when re-swapon
>
> 2. memory leaks when invalidate and reclaim occur concurrently
>
> 3. avoid unnecessary page scanning
>
>
> Issues discussed in that mail thread NOT fixed as it happens rarely or
> not a big problem or controversial:
>
> 1. a "theoretical race condition" when reclaim page
> When a handle alloced from zbud, zbud considers this handle is used
> validly by upper(zswap) and can be a candidate for reclaim. But zswap has
> to initialize it such as setting swapentry and adding it to rbtree.
> so there is a race condition, such as:
> thread 0: obtain handle x from zbud_alloc
> thread 1: zbud_reclaim_page is called
> thread 1: callback zswap_writeback_entry to reclaim handle x
> thread 1: get swpentry from handle x (it is random value now)
> thread 1: bad thing may happen
> thread 0: initialize handle x with swapentry
>
> 2. frontswap_map bitmap not cleared after zswap reclaim
> Frontswap uses frontswap_map bitmap to track page in "backend" implementation,
> when zswap reclaim a page, the corresponding bitmap record is not cleared.
>
> 3. the potential that zswap store and reclaim functions called recursively
>
>
> mm/zswap.c | 28 ++++++++++++++++++++--------
> 1 file changed, 20 insertions(+), 8 deletions(-)
>
> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to [email protected]. For more info on Linux MM,
> see: http://www.linux-mm.org/ .
> Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
--
Kind regards,
Minchan Kim
On Tue, Sep 24, 2013 at 9:11 AM, Minchan Kim <[email protected]> wrote:
> On Mon, Sep 23, 2013 at 04:19:36PM +0800, Weijie Yang wrote:
>> This patch series fix a few bugs in mm/zswap based on Linux-3.11.
>>
>> v2 --> v3
>> - keep GFP_KERNEL flag
>
> Why do you drop this?
>
> It's plain BUG. I read Bob's reply but it couldn't justify to let the pain
> remain. First of all, let's fix it and better idea could come later.
Hi, Minchan. Thanks for review
What I thought is that better idea could come sooner if I keep this
flag rather than fixing it
Yes, you are right. Fixing a bug has a higher priority than waiting
for better idea.
I will add it again in the next version patch set.
>>
>> v1 --> v2
>> - free memory in zswap_frontswap_invalidate_area(in patch 1)
>> - fix whitespace corruption (line wrapping)
>>
>> Corresponding mail thread: https://lkml.org/lkml/2013/8/18/59
>>
>> These issues fixed/optimized are:
>>
>> 1. memory leaks when re-swapon
>>
>> 2. memory leaks when invalidate and reclaim occur concurrently
>>
>> 3. avoid unnecessary page scanning
>>
>>
>> Issues discussed in that mail thread NOT fixed as it happens rarely or
>> not a big problem or controversial:
>>
>> 1. a "theoretical race condition" when reclaim page
>> When a handle alloced from zbud, zbud considers this handle is used
>> validly by upper(zswap) and can be a candidate for reclaim. But zswap has
>> to initialize it such as setting swapentry and adding it to rbtree.
>> so there is a race condition, such as:
>> thread 0: obtain handle x from zbud_alloc
>> thread 1: zbud_reclaim_page is called
>> thread 1: callback zswap_writeback_entry to reclaim handle x
>> thread 1: get swpentry from handle x (it is random value now)
>> thread 1: bad thing may happen
>> thread 0: initialize handle x with swapentry
>>
>> 2. frontswap_map bitmap not cleared after zswap reclaim
>> Frontswap uses frontswap_map bitmap to track page in "backend" implementation,
>> when zswap reclaim a page, the corresponding bitmap record is not cleared.
>>
>> 3. the potential that zswap store and reclaim functions called recursively
>>
>>
>> mm/zswap.c | 28 ++++++++++++++++++++--------
>> 1 file changed, 20 insertions(+), 8 deletions(-)
>>
>> --
>> To unsubscribe, send a message with 'unsubscribe linux-mm' in
>> the body to [email protected]. For more info on Linux MM,
>> see: http://www.linux-mm.org/ .
>> Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
>
> --
> Kind regards,
> Minchan Kim