diff mbox

[v3] kvm tools: Add QCOW level2 caching support

Message ID 1307044916-18109-1-git-send-email-prasadjoshi124@gmail.com (mailing list archive)
State New, archived
Headers show

Commit Message

Prasad Joshi June 2, 2011, 8:01 p.m. UTC
QCOW uses two tables level1 (L1) table and level2 (L2) table. The L1 table
points to offset of L2 table. When a QCOW image is probed, the L1 table is
cached in the memory to avoid reading it from disk on every access. This
caching improves the performance.

The similar performance improvement can be observed when L2 tables are cached.
It is impossible to cache all of the L2 tables because of the memory
constraint. The patch adds L2 table caching capability for up to 128 L2 tables,
it uses combination of RB tree and List to manage the L2 cached tables. The
link list implementation helps in building simple LRU structure and RB tree
helps to search cached table efficiently

The performance numbers are below, the machine was started with following
command line arguments

$ ./kvm run -d /home/prasad/VMDisks/Ubuntu10.10_64_cilk_qemu.qcow \
> --params "root=/dev/vda1" -m 1024

Without QCOW caching

Comments

Pekka Enberg June 2, 2011, 8:37 p.m. UTC | #1
On Thu, Jun 2, 2011 at 11:01 PM, Prasad Joshi <prasadjoshi124@gmail.com> wrote:
> QCOW uses two tables level1 (L1) table and level2 (L2) table. The L1 table
> points to offset of L2 table. When a QCOW image is probed, the L1 table is
> cached in the memory to avoid reading it from disk on every access. This
> caching improves the performance.
>
> The similar performance improvement can be observed when L2 tables are cached.
> It is impossible to cache all of the L2 tables because of the memory
> constraint. The patch adds L2 table caching capability for up to 128 L2 tables,
> it uses combination of RB tree and List to manage the L2 cached tables. The
> link list implementation helps in building simple LRU structure and RB tree
> helps to search cached table efficiently
>
> The performance numbers are below, the machine was started with following
> command line arguments
>
> $ ./kvm run -d /home/prasad/VMDisks/Ubuntu10.10_64_cilk_qemu.qcow \
>> --params "root=/dev/vda1" -m 1024
>
> Without QCOW caching
> ====================
> $ bonnie++ -d tmp/ -c 10 -s 2048
> Writing a byte at a time...done
> Writing intelligently...done
> Rewriting...done
> Reading a byte at a time...done
> Reading intelligently...done
> start 'em...done...done...done...done...done...
> Create files in sequential order...done.
> Stat files in sequential order...done.
> Delete files in sequential order...done.
> Create files in random order...done.
> Stat files in random order...done.
> Delete files in random order...done.
> Version  1.96       ------Sequential Output------ --Sequential Input- --Random-
> Concurrency  10     -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
> Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  /sec %CP
> prasad-virtual-m 2G  1043  99 555406  74 227605  55  5360  99 489080  68 +++++ +++
> Latency             24646us   48544us   57893us    6686us    3595us   21026us
> Version  1.96       ------Sequential Create------ --------Random Create--------
> prasad-virtual-mach -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
>              files  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP
>                 16 +++++ +++ +++++ +++ +++++ +++ +++++ +++ +++++ +++ +++++ +++
> Latency               343us    1175us     327us     555us      48us      82us
> 1.96,1.96,prasad-virtual-machine,10,1307043085,2G,,1043,99,555406,74,227605,55,
> 5360,99,489080,68,+++++,+++,16,,,,,+++++,+++,+++++,+++,+++++,+++,+++++,+++,
> +++++,+++,+++++,+++,24646us,48544us,57893us,6686us,3595us,21026us,343us,1175us,
> 327us,555us,48us,82us
>
> With QCOW caching
> =================
> $ bonnie++ -d tmp/ -c 10 -s 2048
> Writing a byte at a time...done
> Writing intelligently...done
> Rewriting...done
> Reading a byte at a time...done
> Reading intelligently...done
> start 'em...done...done...done...done...done...
> Create files in sequential order...done.
> Stat files in sequential order...done.
> Delete files in sequential order...done.
> Create files in random order...done.
> Stat files in random order...done.
> Delete files in random order...done.
> Version  1.96       ------Sequential Output------ --Sequential Input- --Random-
> Concurrency  10     -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
> Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  /sec %CP
> prasad-virtual-m 2G  1033  99 467899  64 182387  41  5422 100 338294  48 +++++ +++
> Latency             21549us   60585us   65723us    6331us   30014us   19994us
> Version  1.96       ------Sequential Create------ --------Random Create--------
> prasad-virtual-mach -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
>              files  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP
>                 16 +++++ +++ +++++ +++ +++++ +++ +++++ +++ +++++ +++ +++++ +++
> Latency               478us    1142us     344us     402us      72us      98us
> 1.96,1.96,prasad-virtual-machine,10,1307042839,2G,,1033,99,467899,64,182387,41,
> 5422,100,338294,48,+++++,+++,16,,,,,+++++,+++,+++++,+++,+++++,+++,+++++,+++,
> +++++,+++,+++++,+++,21549us,60585us,65723us,6331us,30014us,19994us,478us,1142us,
> 344us,402us,72us,98us
>
> Summary of performance numbers
> ==============================
> There is not much difference with sequential character operations are
> performed, the code with caching performed better by small margin. The caching
> code performance raised by 18% to 24% with sequential block output and by 44%
> for sequentail block input. Which is understandable as the Level2 table will
> always be cached after a write operation. Random seek operation worked slower
> with caching code.

I see performance _degradation_ in the raw data:

Before:

> Concurrency  10     -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
> Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  /sec %CP
> prasad-virtual-m 2G  1043  99 555406  74 227605  55  5360  99 489080  68 +++++ +++

After:

> prasad-virtual-m 2G  1033  99 467899  64 182387  41  5422 100 338294  48 +++++ +++

So that's drop from 555406 to 467899 K/sec for sequential block writes
and drop from 489080 K/sec to 338294 K/sec for sequential block reads.

Random seek latency shows _improvement_:

> Latency             24646us   48544us   57893us    6686us    3595us   21026us

> Latency             21549us   60585us   65723us    6331us   30014us   19994us

Am I reading the Bonnie report wrong or did you mix up the 'before'
and 'after' data?

Assuming the data is just mixed up, I'm not completely happy that
we're making random seeks almost 5% slower. Any ideas why that's
happening?

> +static int search_table(struct qcow *q, u64 **table, u64 offset)
> +{
> +       struct qcow_l2_cache *c;
> +
> +       *table = NULL;
> +
> +       c = search(&q->root, offset);
> +       if (!c)
> +               return -1;
> +
> +       /* Update the LRU state */
> +       list_del_init(&c->list);
> +       list_add_tail(&c->list, &q->lru_list);

Why not use list_move() here? The "update the LRU state" comment is
pretty useless. It would be more important to explain the reader how
the list is ordered (i.e. least recently used are at the head of the
list).

> @@ -17,6 +19,16 @@
>  #define QCOW2_OFLAG_COMPRESSED (1LL << 62)
>  #define QCOW2_OFLAG_MASK       (QCOW2_OFLAG_COPIED|QCOW2_OFLAG_COMPRESSED)
>
> +#define MAX_CACHE_NODES         128

Did you test the results with MAX_CACHE_NODES set to 1? Did you get
similar results than with no caching at all? I also wonder if we could
get away with even smaller cache than 128.

>  struct qcow_table {
>        u32                     table_size;
>        u64                     *l1_table;
> @@ -26,6 +38,11 @@ struct qcow {
>        void                    *header;
>        struct qcow_table       table;
>        int                     fd;
> +
> +       /* Level2 caching data structures */
> +       struct rb_root          root;
> +       struct list_head        lru_list;
> +       int                     no_cached;

I've said this many times: please don't invent strange new names.
Really, "no_cached" reads out like it's a boolean! Use 'nr_cached'
instead.

                        Pekka
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Prasad Joshi June 2, 2011, 8:57 p.m. UTC | #2
On Thu, Jun 2, 2011 at 9:37 PM, Pekka Enberg <penberg@kernel.org> wrote:
> On Thu, Jun 2, 2011 at 11:01 PM, Prasad Joshi <prasadjoshi124@gmail.com> wrote:
>> QCOW uses two tables level1 (L1) table and level2 (L2) table. The L1 table
>> points to offset of L2 table. When a QCOW image is probed, the L1 table is
>> cached in the memory to avoid reading it from disk on every access. This
>> caching improves the performance.
>>
>> The similar performance improvement can be observed when L2 tables are cached.
>> It is impossible to cache all of the L2 tables because of the memory
>> constraint. The patch adds L2 table caching capability for up to 128 L2 tables,
>> it uses combination of RB tree and List to manage the L2 cached tables. The
>> link list implementation helps in building simple LRU structure and RB tree
>> helps to search cached table efficiently
>>
>> The performance numbers are below, the machine was started with following
>> command line arguments
>>
>> $ ./kvm run -d /home/prasad/VMDisks/Ubuntu10.10_64_cilk_qemu.qcow \
>>> --params "root=/dev/vda1" -m 1024
>>
>> Without QCOW caching
>> ====================
>> $ bonnie++ -d tmp/ -c 10 -s 2048
>> Writing a byte at a time...done
>> Writing intelligently...done
>> Rewriting...done
>> Reading a byte at a time...done
>> Reading intelligently...done
>> start 'em...done...done...done...done...done...
>> Create files in sequential order...done.
>> Stat files in sequential order...done.
>> Delete files in sequential order...done.
>> Create files in random order...done.
>> Stat files in random order...done.
>> Delete files in random order...done.
>> Version  1.96       ------Sequential Output------ --Sequential Input- --Random-
>> Concurrency  10     -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
>> Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  /sec %CP
>> prasad-virtual-m 2G  1043  99 555406  74 227605  55  5360  99 489080  68 +++++ +++
>> Latency             24646us   48544us   57893us    6686us    3595us   21026us
>> Version  1.96       ------Sequential Create------ --------Random Create--------
>> prasad-virtual-mach -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
>>              files  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP
>>                 16 +++++ +++ +++++ +++ +++++ +++ +++++ +++ +++++ +++ +++++ +++
>> Latency               343us    1175us     327us     555us      48us      82us
>> 1.96,1.96,prasad-virtual-machine,10,1307043085,2G,,1043,99,555406,74,227605,55,
>> 5360,99,489080,68,+++++,+++,16,,,,,+++++,+++,+++++,+++,+++++,+++,+++++,+++,
>> +++++,+++,+++++,+++,24646us,48544us,57893us,6686us,3595us,21026us,343us,1175us,
>> 327us,555us,48us,82us
>>
>> With QCOW caching
>> =================
>> $ bonnie++ -d tmp/ -c 10 -s 2048
>> Writing a byte at a time...done
>> Writing intelligently...done
>> Rewriting...done
>> Reading a byte at a time...done
>> Reading intelligently...done
>> start 'em...done...done...done...done...done...
>> Create files in sequential order...done.
>> Stat files in sequential order...done.
>> Delete files in sequential order...done.
>> Create files in random order...done.
>> Stat files in random order...done.
>> Delete files in random order...done.
>> Version  1.96       ------Sequential Output------ --Sequential Input- --Random-
>> Concurrency  10     -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
>> Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  /sec %CP
>> prasad-virtual-m 2G  1033  99 467899  64 182387  41  5422 100 338294  48 +++++ +++
>> Latency             21549us   60585us   65723us    6331us   30014us   19994us
>> Version  1.96       ------Sequential Create------ --------Random Create--------
>> prasad-virtual-mach -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
>>              files  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP
>>                 16 +++++ +++ +++++ +++ +++++ +++ +++++ +++ +++++ +++ +++++ +++
>> Latency               478us    1142us     344us     402us      72us      98us
>> 1.96,1.96,prasad-virtual-machine,10,1307042839,2G,,1033,99,467899,64,182387,41,
>> 5422,100,338294,48,+++++,+++,16,,,,,+++++,+++,+++++,+++,+++++,+++,+++++,+++,
>> +++++,+++,+++++,+++,21549us,60585us,65723us,6331us,30014us,19994us,478us,1142us,
>> 344us,402us,72us,98us
>>
>> Summary of performance numbers
>> ==============================
>> There is not much difference with sequential character operations are
>> performed, the code with caching performed better by small margin. The caching
>> code performance raised by 18% to 24% with sequential block output and by 44%
>> for sequentail block input. Which is understandable as the Level2 table will
>> always be cached after a write operation. Random seek operation worked slower
>> with caching code.
>
> I see performance _degradation_ in the raw data:

I think I copied the wrong results in the commit log.

>
> Before:
>
>> Concurrency  10     -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
>> Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  /sec %CP
>> prasad-virtual-m 2G  1043  99 555406  74 227605  55  5360  99 489080  68 +++++ +++
>
> After:
>
>> prasad-virtual-m 2G  1033  99 467899  64 182387  41  5422 100 338294  48 +++++ +++
>
> So that's drop from 555406 to 467899 K/sec for sequential block writes
> and drop from 489080 K/sec to 338294 K/sec for sequential block reads.
>
> Random seek latency shows _improvement_:
>
>> Latency             24646us   48544us   57893us    6686us    3595us   21026us
>
>> Latency             21549us   60585us   65723us    6331us   30014us   19994us
>
> Am I reading the Bonnie report wrong or did you mix up the 'about:startpagebefore'
> and 'after' data?
>
> Assuming the data is just mixed up, I'm not completely happy that
> we're making random seeks almost 5% slower. Any ideas why that's
> happening?

Yes they are mixed.

I am assuming, probably delay while caching.

>
>> +static int search_table(struct qcow *q, u64 **table, u64 offset)
>> +{
>> +       struct qcow_l2_cache *c;
>> +
>> +       *table = NULL;
>> +
>> +       c = search(&q->root, offset);
>> +       if (!c)
>> +               return -1;
>> +
>> +       /* Update the LRU state */
>> +       list_del_init(&c->list);
>> +       list_add_tail(&c->list, &q->lru_list);
>
> Why not use list_move() here? The "update the LRU state" comment is
> pretty useless. It would be more important to explain the reader how
> the list is ordered (i.e. least recently used are at the head of the
> list).

Okay

>
>> @@ -17,6 +19,16 @@
>>  #define QCOW2_OFLAG_COMPRESSED (1LL << 62)
>>  #define QCOW2_OFLAG_MASK       (QCOW2_OFLAG_COPIED|QCOW2_OFLAG_COMPRESSED)
>>
>> +#define MAX_CACHE_NODES         128
>
> Did you test the results with MAX_CACHE_NODES set to 1? Did you get
> similar results than with no caching at all?

Never tested with MAX_CACHE_NODES with 1. I need to do that.

> I also wonder if we could
> get away with even smaller cache than 128.
>

What would be the right number?
If number of cache nodes are pretty much less for example 32 or
something. We can use arrays for caching rather than Link List and RB
Tree.

>>  struct qcow_table {
>>        u32                     table_size;
>>        u64                     *l1_table;
>> @@ -26,6 +38,11 @@ struct qcow {
>>        void                    *header;
>>        struct qcow_table       table;
>>        int                     fd;
>> +
>> +       /* Level2 caching data structures */
>> +       struct rb_root          root;
>> +       struct list_head        lru_list;
>> +       int                     no_cached;
>
> I've said this many times: please don't invent strange new names.
> Really, "no_cached" reads out like it's a boolean! Use 'nr_cached'
> instead.

okay.

>
>                        Pekka
>
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Pekka Enberg June 2, 2011, 9:07 p.m. UTC | #3
On Thu, 2011-06-02 at 21:57 +0100, Prasad Joshi wrote:
> > Assuming the data is just mixed up, I'm not completely happy that
> > we're making random seeks almost 5% slower. Any ideas why that's
> > happening?
> 
> Yes they are mixed.
> 
> I am assuming, probably delay while caching.

That doesn't make sense to me. Why would "caching delay" be
significantly slower than the non-cached version where we do full read
all the time? Maybe I didn't quite understand the code but that sounds
like a bug lurking somewhere in the caching code - maybe we cache too
much or something?

			Pekka

--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Pekka Enberg June 2, 2011, 9:11 p.m. UTC | #4
On Thu, Jun 2, 2011 at 11:01 PM, Prasad Joshi <prasadjoshi124@gmail.com> wrote:
> There is not much difference with sequential character operations are
> performed, the code with caching performed better by small margin. The caching
> code performance raised by 18% to 24% with sequential block output and by 44%
> for sequentail block input. Which is understandable as the Level2 table will
> always be cached after a write operation. Random seek operation worked slower
> with caching code.

Btw, please mention how much slower random seeks got in the next
version of the patch - unless you're able to make the slowdown go
away, of course. ;-)
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Sasha Levin June 2, 2011, 9:19 p.m. UTC | #5
On Thu, 2011-06-02 at 21:01 +0100, Prasad Joshi wrote:
> QCOW uses two tables level1 (L1) table and level2 (L2) table. The L1 table
> points to offset of L2 table. When a QCOW image is probed, the L1 table is
> cached in the memory to avoid reading it from disk on every access. This
> caching improves the performance.
> 
> The similar performance improvement can be observed when L2 tables are cached.
> It is impossible to cache all of the L2 tables because of the memory
> constraint. The patch adds L2 table caching capability for up to 128 L2 tables,
> it uses combination of RB tree and List to manage the L2 cached tables. The
> link list implementation helps in building simple LRU structure and RB tree
> helps to search cached table efficiently
> 
> The performance numbers are below, the machine was started with following
> command line arguments
> 
> $ ./kvm run -d /home/prasad/VMDisks/Ubuntu10.10_64_cilk_qemu.qcow \
> > --params "root=/dev/vda1" -m 1024
> 
> Without QCOW caching
> ====================
> $ bonnie++ -d tmp/ -c 10 -s 2048
> Writing a byte at a time...done
> Writing intelligently...done
> Rewriting...done
> Reading a byte at a time...done
> Reading intelligently...done
> start 'em...done...done...done...done...done...
> Create files in sequential order...done.
> Stat files in sequential order...done.
> Delete files in sequential order...done.
> Create files in random order...done.
> Stat files in random order...done.
> Delete files in random order...done.
> Version  1.96       ------Sequential Output------ --Sequential Input- --Random-
> Concurrency  10     -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
> Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  /sec %CP
> prasad-virtual-m 2G  1043  99 555406  74 227605  55  5360  99 489080  68 +++++ +++
> Latency             24646us   48544us   57893us    6686us    3595us   21026us
> Version  1.96       ------Sequential Create------ --------Random Create--------
> prasad-virtual-mach -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
>               files  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP
>                  16 +++++ +++ +++++ +++ +++++ +++ +++++ +++ +++++ +++ +++++ +++
> Latency               343us    1175us     327us     555us      48us      82us
> 1.96,1.96,prasad-virtual-machine,10,1307043085,2G,,1043,99,555406,74,227605,55,
> 5360,99,489080,68,+++++,+++,16,,,,,+++++,+++,+++++,+++,+++++,+++,+++++,+++,
> +++++,+++,+++++,+++,24646us,48544us,57893us,6686us,3595us,21026us,343us,1175us,
> 327us,555us,48us,82us
> 
> With QCOW caching
> =================
> $ bonnie++ -d tmp/ -c 10 -s 2048
> Writing a byte at a time...done
> Writing intelligently...done
> Rewriting...done
> Reading a byte at a time...done
> Reading intelligently...done
> start 'em...done...done...done...done...done...
> Create files in sequential order...done.
> Stat files in sequential order...done.
> Delete files in sequential order...done.
> Create files in random order...done.
> Stat files in random order...done.
> Delete files in random order...done.
> Version  1.96       ------Sequential Output------ --Sequential Input- --Random-
> Concurrency  10     -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
> Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  /sec %CP
> prasad-virtual-m 2G  1033  99 467899  64 182387  41  5422 100 338294  48 +++++ +++
> Latency             21549us   60585us   65723us    6331us   30014us   19994us
> Version  1.96       ------Sequential Create------ --------Random Create--------
> prasad-virtual-mach -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
>               files  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP
>                  16 +++++ +++ +++++ +++ +++++ +++ +++++ +++ +++++ +++ +++++ +++
> Latency               478us    1142us     344us     402us      72us      98us
> 1.96,1.96,prasad-virtual-machine,10,1307042839,2G,,1033,99,467899,64,182387,41,
> 5422,100,338294,48,+++++,+++,16,,,,,+++++,+++,+++++,+++,+++++,+++,+++++,+++,
> +++++,+++,+++++,+++,21549us,60585us,65723us,6331us,30014us,19994us,478us,1142us,
> 344us,402us,72us,98us
> 
> Summary of performance numbers
> ==============================
> There is not much difference with sequential character operations are
> performed, the code with caching performed better by small margin. The caching
> code performance raised by 18% to 24% with sequential block output and by 44%
> for sequentail block input. Which is understandable as the Level2 table will
> always be cached after a write operation. Random seek operation worked slower
> with caching code.
> 
> Signed-off-by: Prasad Joshi <prasadjoshi124@gmail.com>
> ---
>  tools/kvm/disk/qcow.c        |  216 ++++++++++++++++++++++++++++++++++++++----
>  tools/kvm/include/kvm/qcow.h |   17 ++++
>  2 files changed, 216 insertions(+), 17 deletions(-)
> 
> diff --git a/tools/kvm/disk/qcow.c b/tools/kvm/disk/qcow.c
> index b52b734..4fa3850 100644
> --- a/tools/kvm/disk/qcow.c
> +++ b/tools/kvm/disk/qcow.c
> @@ -16,6 +16,140 @@
>  #include <linux/kernel.h>
>  #include <linux/types.h>
>  
> +static int insert(struct rb_root *root, struct qcow_l2_cache *new)
> +{
> +	struct rb_node **link = &(root->rb_node), *parent = NULL;
> +	u64 offset = new->offset;
> +
> +	/* search the tree */
> +	while (*link) {
> +		struct qcow_l2_cache *t;
> +
> +		t = rb_entry(*link, struct qcow_l2_cache, node);
> +		if (!t)
> +			goto error;
> +
> +		parent = *link;
> +
> +		if (t->offset > offset)
> +			link = &(*link)->rb_left;
> +		else if (t->offset < offset)
> +			link = &(*link)->rb_right;
> +		else
> +			goto out;
> +	}
> +
> +	/* add new node */
> +	rb_link_node(&new->node, parent, link);
> +	rb_insert_color(&new->node, root);
> +out:
> +	return 0;
> +error:
> +	return -1;
> +}
> +
> +static struct qcow_l2_cache *search(struct rb_root *root, u64 offset)
> +{
> +	struct rb_node *link = root->rb_node;
> +
> +	while (link) {
> +		struct qcow_l2_cache *t;
> +
> +		t = rb_entry(link, struct qcow_l2_cache, node);
> +		if (!t)
> +			goto out;
> +
> +		if (t->offset > offset)
> +			link = link->rb_left;
> +		else if (t->offset < offset)
> +			link = link->rb_right;
> +		else
> +			return t;
> +	}
> +out:
> +	return NULL;
> +}
> +
> +static void free_cache(struct qcow *q)
> +{
> +	struct list_head *pos, *n;
> +	struct qcow_l2_cache *t;
> +	struct rb_root *r = &q->root;
> +
> +	list_for_each_safe(pos, n, &q->lru_list) {
> +		/* Remove cache table from the list and RB tree */
> +		list_del(pos);
> +		t = list_entry(pos, struct qcow_l2_cache, list);
> +		rb_erase(&t->node, r);
> +
> +		/* Free the cached node */
> +		free(t->table);
> +		free(t);
> +	}
> +}
> +
> +static int cache_table(struct qcow *q, u64 *table, u64 offset)
> +{
> +	struct qcow_l2_cache *n;
> +	struct rb_root *r = &q->root;
> +	struct qcow_l2_cache *lru;
> +
> +	n = calloc(1, sizeof(struct qcow_l2_cache));
			sizeof(*n)
sizeof() should use the variable name itself, not the data type. Check
out chapter 14 in 'Documentation/CodingStyle'.

> +	if (!n)
> +		goto error;
> +
> +	n->offset = offset;
> +	n->table  = table;
> +	n->qcow   = q;
> +	RB_CLEAR_NODE(&n->node);
> +	INIT_LIST_HEAD(&n->list);
> +
> +	if (q->no_cached == MAX_CACHE_NODES) {
> +		/* Find the node for LRU replacement */
> +		lru = list_first_entry(&q->lru_list, struct qcow_l2_cache, list);
> +
> +		/* Remove the node from the cache */
> +		rb_erase(&lru->node, r);
> +		list_del_init(&lru->list);
> +		q->no_cached--;
> +
> +		/* Free the LRUed node */
> +		free(lru->table);
> +		free(lru);
> +	}
> +
> +	/* Add new node in RB Tree: Helps in searching faster */
> +	if (insert(r, n) < 0)
> +		goto error;
> +
> +	/* Add in LRU replacement list */
> +	list_add_tail(&n->list, &q->lru_list);
> +	q->no_cached++;
> +
> +	return 0;
> +error:
> +	free(n);
> +	return -1;
> +}
> +
> +static int search_table(struct qcow *q, u64 **table, u64 offset)
> +{
> +	struct qcow_l2_cache *c;
> +
> +	*table = NULL;
> +
> +	c = search(&q->root, offset);
> +	if (!c)
> +		return -1;
> +
> +	/* Update the LRU state */
> +	list_del_init(&c->list);
> +	list_add_tail(&c->list, &q->lru_list);
> +
> +	*table = c->table;
> +	return 0;
> +}
> +
>  static inline u64 get_l1_index(struct qcow *q, u64 offset)
>  {
>  	struct qcow_header *header = q->header;
> @@ -37,6 +171,43 @@ static inline u64 get_cluster_offset(struct qcow *q, u64 offset)
>  	return offset & ((1 << header->cluster_bits)-1);
>  }
>  
> +static int qcow_read_l2_table(struct qcow *q, u64 **table, u64 offset)
> +{
> +	struct qcow_header *header = q->header;
> +	u64 size;
> +	u64 i;
> +	u64 *t;
> +
> +	*table = NULL;
> +	size   = 1 << header->l2_bits;
> +
> +	/* search an entry for offset in cache */
> +	if (search_table(q, table, offset) >= 0)
> +		return 0;
> +
> +	t = calloc(size, sizeof(*t));
> +	if (!t)
> +		goto error;
> +
> +	/* table not cached: read from the disk */
> +	if (pread_in_full(q->fd, t, size * sizeof(u64), offset) < 0)
> +		goto error;
> +
> +	/* cache the table */
> +	if (cache_table(q, t, offset) < 0)
> +		goto error;
> +
> +	/* change cached table to CPU's byte-order */
> +	for (i = 0; i < size; i++)
> +		be64_to_cpus(&t[i]);
> +
> +	*table = t;
> +	return 0;
> +error:
> +	free(t);
> +	return -1;
> +}
> +
>  static ssize_t qcow_read_cluster(struct qcow *q, u64 offset, void *dst, u32 dst_len)
>  {
>  	struct qcow_header *header = q->header;
> @@ -51,7 +222,6 @@ static ssize_t qcow_read_cluster(struct qcow *q, u64 offset, void *dst, u32 dst_
>  	u64 l1_idx;
>  	u64 l2_idx;
>  
> -	l2_table = NULL;
>  
>  	cluster_size = 1 << header->cluster_bits;
>  
> @@ -72,18 +242,16 @@ static ssize_t qcow_read_cluster(struct qcow *q, u64 offset, void *dst, u32 dst_
>  		goto zero_cluster;
>  
>  	l2_table_size = 1 << header->l2_bits;
> -	l2_table = calloc(l2_table_size, sizeof(u64));
> -	if (!l2_table)
> -		goto out_error;
>  
> -	if (pread_in_full(q->fd, l2_table, l2_table_size * sizeof(u64), l2_table_offset) < 0)
> +	/* read and cache level 2 table */
> +	if (qcow_read_l2_table(q, &l2_table, l2_table_offset) < 0)
>  		goto out_error;
>  
>  	l2_idx = get_l2_index(q, offset);
>  	if (l2_idx >= l2_table_size)
>  		goto out_error;
>  
> -	clust_start = be64_to_cpu(l2_table[l2_idx]) & ~header->oflag_mask;
> +	clust_start = l2_table[l2_idx] & ~header->oflag_mask;
>  	if (!clust_start)
>  		goto zero_cluster;
>  
> @@ -91,7 +259,6 @@ static ssize_t qcow_read_cluster(struct qcow *q, u64 offset, void *dst, u32 dst_
>  		goto out_error;
>  
>  out:
> -	free(l2_table);
>  	return length;
>  
>  zero_cluster:
> @@ -221,15 +388,16 @@ static ssize_t qcow_write_cluster(struct qcow *q, u64 offset, void *buf, u32 src
>  	if (len > src_len)
>  		len = src_len;
>  
> -	l2t = calloc(l2t_sz, sizeof(u64));
> -	if (!l2t)
> -		goto error;
> -
>  	l2t_off		= table->l1_table[l1t_idx] & ~header->oflag_mask;
>  	if (l2t_off) {
> -		if (pread_in_full(q->fd, l2t, l2t_sz * sizeof(u64), l2t_off) < 0)
> -			goto free_l2;
> +		/* cache level2 table */
> +		if (qcow_read_l2_table(q, &l2t, l2t_off) < 0)
> +			goto error;
>  	} else {
> +		l2t = calloc(l2t_sz, sizeof(*l2t));
> +		if (!l2t)
> +			goto error;
> +
>  		/* Capture the state of the consistent QCOW image */
>  		f_sz		= file_size(q->fd);
>  		if (!f_sz)
> @@ -251,6 +419,13 @@ static ssize_t qcow_write_cluster(struct qcow *q, u64 offset, void *buf, u32 src
>  			goto free_l2;
>  		}
>  
> +		if (cache_table(q, l2t, l2t_off) < 0) {
> +			if (ftruncate(q->fd, f_sz) < 0)
> +				goto free_l2;
> +
> +			goto free_l2;
> +		}
> +
>  		/* Update the in-core entry */
>  		table->l1_table[l1t_idx] = l2t_off;
>  	}
> @@ -258,17 +433,15 @@ static ssize_t qcow_write_cluster(struct qcow *q, u64 offset, void *buf, u32 src
>  	/* Capture the state of the consistent QCOW image */
>  	f_sz		= file_size(q->fd);
>  	if (!f_sz)
> -		goto free_l2;
> +		goto error;
>  
> -	clust_start	= be64_to_cpu(l2t[l2t_idx]) & ~header->oflag_mask;
> +	clust_start	= l2t[l2t_idx] & ~header->oflag_mask;
>  	if (!clust_start) {
>  		clust_start	= ALIGN(f_sz, clust_sz);
>  		update_meta	= true;
>  	} else
>  		update_meta	= false;
>  
> -	free(l2t);
> -
>  	/* Write actual data */
>  	if (pwrite_in_full(q->fd, buf, len, clust_start + clust_off) < 0)
>  		goto error;
> @@ -282,9 +455,13 @@ static ssize_t qcow_write_cluster(struct qcow *q, u64 offset, void *buf, u32 src
>  
>  			goto error;
>  		}
> +
> +		/* Update the cached level2 entry */
> +		l2t[l2t_idx] = clust_start;
>  	}
>  
>  	return len;
> +
>  free_l2:
>  	free(l2t);
>  error:
> @@ -336,6 +513,7 @@ static int qcow_disk_close(struct disk_image *disk)
>  
>  	q = disk->priv;
>  
> +	free_cache(q);
>  	free(q->table.l1_table);
>  	free(q->header);
>  	free(q);
> @@ -428,6 +606,8 @@ static struct disk_image *qcow2_probe(int fd, bool readonly)
>  		goto error;
>  
>  	q->fd = fd;
> +	q->root = RB_ROOT;
> +	INIT_LIST_HEAD(&q->lru_list);
>  
>  	h = q->header = qcow2_read_header(fd);
>  	if (!h)
> @@ -525,6 +705,8 @@ static struct disk_image *qcow1_probe(int fd, bool readonly)
>  		goto error;
>  
>  	q->fd = fd;
> +	q->root = RB_ROOT;
> +	INIT_LIST_HEAD(&q->lru_list);
>  
>  	h = q->header = qcow1_read_header(fd);
>  	if (!h)
> diff --git a/tools/kvm/include/kvm/qcow.h b/tools/kvm/include/kvm/qcow.h
> index b6e7493..1663819 100644
> --- a/tools/kvm/include/kvm/qcow.h
> +++ b/tools/kvm/include/kvm/qcow.h
> @@ -3,6 +3,8 @@
>  
>  #include <linux/types.h>
>  #include <stdbool.h>
> +#include <linux/rbtree.h>
> +#include <linux/list.h>
>  
>  #define QCOW_MAGIC		(('Q' << 24) | ('F' << 16) | ('I' << 8) | 0xfb)
>  
> @@ -17,6 +19,16 @@
>  #define QCOW2_OFLAG_COMPRESSED	(1LL << 62)
>  #define QCOW2_OFLAG_MASK	(QCOW2_OFLAG_COPIED|QCOW2_OFLAG_COMPRESSED)
>  
> +#define MAX_CACHE_NODES         128
> +
> +struct qcow_l2_cache {
> +	u64                     offset;
> +	u64                     *table;
> +	struct rb_node          node;
> +	struct list_head        list;
> +	struct qcow             *qcow;

I only see 'qcow' being initialized, but isn't used anywhere. Is it
really needed?

> +};
> +
>  struct qcow_table {
>  	u32			table_size;
>  	u64			*l1_table;
> @@ -26,6 +38,11 @@ struct qcow {
>  	void			*header;
>  	struct qcow_table	table;
>  	int			fd;
> +
> +	/* Level2 caching data structures */
> +	struct rb_root          root;
> +	struct list_head        lru_list;
> +	int                     no_cached;
>  };
>  
>  struct qcow_header {
Pekka Enberg June 3, 2011, 5:56 a.m. UTC | #6
On Fri, Jun 3, 2011 at 12:19 AM, Sasha Levin <levinsasha928@gmail.com> wrote:
>> +static int cache_table(struct qcow *q, u64 *table, u64 offset)
>> +{
>> +     struct qcow_l2_cache *n;
>> +     struct rb_root *r = &q->root;
>> +     struct qcow_l2_cache *lru;
>> +
>> +     n = calloc(1, sizeof(struct qcow_l2_cache));
>                        sizeof(*n)
> sizeof() should use the variable name itself, not the data type. Check
> out chapter 14 in 'Documentation/CodingStyle'.

Well, it doesn't matter that much, to be honest. 'n' could use a
better name, though - 'cache' or 'c'.

                        Pekka
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Ingo Molnar June 3, 2011, 7:15 a.m. UTC | #7
* Pekka Enberg <penberg@kernel.org> wrote:

> On Fri, Jun 3, 2011 at 12:19 AM, Sasha Levin <levinsasha928@gmail.com> wrote:
> >> +static int cache_table(struct qcow *q, u64 *table, u64 offset)
> >> +{
> >> +     struct qcow_l2_cache *n;
> >> +     struct rb_root *r = &q->root;
> >> +     struct qcow_l2_cache *lru;
> >> +
> >> +     n = calloc(1, sizeof(struct qcow_l2_cache));
> >                        sizeof(*n)
> > sizeof() should use the variable name itself, not the data type. Check
> > out chapter 14 in 'Documentation/CodingStyle'.
> 
> Well, it doesn't matter that much, to be honest. 'n' could use a 
> better name, though - 'cache' or 'c'.

I personally prefer the sizeof(*cache) variant for a subtle reason: 
because during review it's easier to match up local variable names 
than to match up types.

For example when i review code i only look at the types once: i just 
establish their main nature and attach any semantic meaning to the 
local variable name.

So if 'later in the code' i see "sizeof(struct qcow_l2_cache)" i wont 
know it intuitively whether it matches up with 'cache' or not. So for 
example i might not notice such a bug:

	cache = calloc(1, sizeof(struct qcow_l1_cache));

(trivia: how many seconds does it take for you to notice the bug in 
the above line? Note, you already have the advantage that you *know* 
that there's a bug in that line :-)

Even if i noticed the bug, i'd notice it by looking back at the local 
variables defininition block - which is a distraction from reading 
the code flow itself.

But if it's written differently i will notice this bug immediately:

	cache = calloc(1, sizeof(*r));

i will even notice this pattern:

	cache = calloc(1, sizeof(cache));

I guess we could introduce a type-safe zalloc variant, something 
like:

	cache = zalloc_t(*cache);

Which is a clever macro that allocates sizeof(struct qcow_l2_cache) 
bytes and gives back a 'struct qcow_l2_cache *' typed result. That 
way this:

	cache = zalloc_t(r);

or this:

	cache = zalloc_t(cache);

Will result in a compiler error.

Hm?

Thanks,

	Ingo
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Pekka Enberg June 3, 2011, 7:23 a.m. UTC | #8
On Fri, Jun 3, 2011 at 10:15 AM, Ingo Molnar <mingo@elte.hu> wrote:
> I personally prefer the sizeof(*cache) variant for a subtle reason:
> because during review it's easier to match up local variable names
> than to match up types.

[snip, snip]

Aye, aye, cap'n!

As long as we have good volunteers (hi Sasha, hi Ingo!) reviewing
patches and enforcing that, I'm completely fine with it.

                        Pekka
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
diff mbox

Patch

====================
$ bonnie++ -d tmp/ -c 10 -s 2048
Writing a byte at a time...done
Writing intelligently...done
Rewriting...done
Reading a byte at a time...done
Reading intelligently...done
start 'em...done...done...done...done...done...
Create files in sequential order...done.
Stat files in sequential order...done.
Delete files in sequential order...done.
Create files in random order...done.
Stat files in random order...done.
Delete files in random order...done.
Version  1.96       ------Sequential Output------ --Sequential Input- --Random-
Concurrency  10     -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  /sec %CP
prasad-virtual-m 2G  1043  99 555406  74 227605  55  5360  99 489080  68 +++++ +++
Latency             24646us   48544us   57893us    6686us    3595us   21026us
Version  1.96       ------Sequential Create------ --------Random Create--------
prasad-virtual-mach -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
              files  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP
                 16 +++++ +++ +++++ +++ +++++ +++ +++++ +++ +++++ +++ +++++ +++
Latency               343us    1175us     327us     555us      48us      82us
1.96,1.96,prasad-virtual-machine,10,1307043085,2G,,1043,99,555406,74,227605,55,
5360,99,489080,68,+++++,+++,16,,,,,+++++,+++,+++++,+++,+++++,+++,+++++,+++,
+++++,+++,+++++,+++,24646us,48544us,57893us,6686us,3595us,21026us,343us,1175us,
327us,555us,48us,82us

With QCOW caching
=================
$ bonnie++ -d tmp/ -c 10 -s 2048
Writing a byte at a time...done
Writing intelligently...done
Rewriting...done
Reading a byte at a time...done
Reading intelligently...done
start 'em...done...done...done...done...done...
Create files in sequential order...done.
Stat files in sequential order...done.
Delete files in sequential order...done.
Create files in random order...done.
Stat files in random order...done.
Delete files in random order...done.
Version  1.96       ------Sequential Output------ --Sequential Input- --Random-
Concurrency  10     -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine        Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP  /sec %CP
prasad-virtual-m 2G  1033  99 467899  64 182387  41  5422 100 338294  48 +++++ +++
Latency             21549us   60585us   65723us    6331us   30014us   19994us
Version  1.96       ------Sequential Create------ --------Random Create--------
prasad-virtual-mach -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete--
              files  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP  /sec %CP
                 16 +++++ +++ +++++ +++ +++++ +++ +++++ +++ +++++ +++ +++++ +++
Latency               478us    1142us     344us     402us      72us      98us
1.96,1.96,prasad-virtual-machine,10,1307042839,2G,,1033,99,467899,64,182387,41,
5422,100,338294,48,+++++,+++,16,,,,,+++++,+++,+++++,+++,+++++,+++,+++++,+++,
+++++,+++,+++++,+++,21549us,60585us,65723us,6331us,30014us,19994us,478us,1142us,
344us,402us,72us,98us

Summary of performance numbers
==============================
There is not much difference with sequential character operations are
performed, the code with caching performed better by small margin. The caching
code performance raised by 18% to 24% with sequential block output and by 44%
for sequentail block input. Which is understandable as the Level2 table will
always be cached after a write operation. Random seek operation worked slower
with caching code.

Signed-off-by: Prasad Joshi <prasadjoshi124@gmail.com>
---
 tools/kvm/disk/qcow.c        |  216 ++++++++++++++++++++++++++++++++++++++----
 tools/kvm/include/kvm/qcow.h |   17 ++++
 2 files changed, 216 insertions(+), 17 deletions(-)

diff --git a/tools/kvm/disk/qcow.c b/tools/kvm/disk/qcow.c
index b52b734..4fa3850 100644
--- a/tools/kvm/disk/qcow.c
+++ b/tools/kvm/disk/qcow.c
@@ -16,6 +16,140 @@ 
 #include <linux/kernel.h>
 #include <linux/types.h>
 
+static int insert(struct rb_root *root, struct qcow_l2_cache *new)
+{
+	struct rb_node **link = &(root->rb_node), *parent = NULL;
+	u64 offset = new->offset;
+
+	/* search the tree */
+	while (*link) {
+		struct qcow_l2_cache *t;
+
+		t = rb_entry(*link, struct qcow_l2_cache, node);
+		if (!t)
+			goto error;
+
+		parent = *link;
+
+		if (t->offset > offset)
+			link = &(*link)->rb_left;
+		else if (t->offset < offset)
+			link = &(*link)->rb_right;
+		else
+			goto out;
+	}
+
+	/* add new node */
+	rb_link_node(&new->node, parent, link);
+	rb_insert_color(&new->node, root);
+out:
+	return 0;
+error:
+	return -1;
+}
+
+static struct qcow_l2_cache *search(struct rb_root *root, u64 offset)
+{
+	struct rb_node *link = root->rb_node;
+
+	while (link) {
+		struct qcow_l2_cache *t;
+
+		t = rb_entry(link, struct qcow_l2_cache, node);
+		if (!t)
+			goto out;
+
+		if (t->offset > offset)
+			link = link->rb_left;
+		else if (t->offset < offset)
+			link = link->rb_right;
+		else
+			return t;
+	}
+out:
+	return NULL;
+}
+
+static void free_cache(struct qcow *q)
+{
+	struct list_head *pos, *n;
+	struct qcow_l2_cache *t;
+	struct rb_root *r = &q->root;
+
+	list_for_each_safe(pos, n, &q->lru_list) {
+		/* Remove cache table from the list and RB tree */
+		list_del(pos);
+		t = list_entry(pos, struct qcow_l2_cache, list);
+		rb_erase(&t->node, r);
+
+		/* Free the cached node */
+		free(t->table);
+		free(t);
+	}
+}
+
+static int cache_table(struct qcow *q, u64 *table, u64 offset)
+{
+	struct qcow_l2_cache *n;
+	struct rb_root *r = &q->root;
+	struct qcow_l2_cache *lru;
+
+	n = calloc(1, sizeof(struct qcow_l2_cache));
+	if (!n)
+		goto error;
+
+	n->offset = offset;
+	n->table  = table;
+	n->qcow   = q;
+	RB_CLEAR_NODE(&n->node);
+	INIT_LIST_HEAD(&n->list);
+
+	if (q->no_cached == MAX_CACHE_NODES) {
+		/* Find the node for LRU replacement */
+		lru = list_first_entry(&q->lru_list, struct qcow_l2_cache, list);
+
+		/* Remove the node from the cache */
+		rb_erase(&lru->node, r);
+		list_del_init(&lru->list);
+		q->no_cached--;
+
+		/* Free the LRUed node */
+		free(lru->table);
+		free(lru);
+	}
+
+	/* Add new node in RB Tree: Helps in searching faster */
+	if (insert(r, n) < 0)
+		goto error;
+
+	/* Add in LRU replacement list */
+	list_add_tail(&n->list, &q->lru_list);
+	q->no_cached++;
+
+	return 0;
+error:
+	free(n);
+	return -1;
+}
+
+static int search_table(struct qcow *q, u64 **table, u64 offset)
+{
+	struct qcow_l2_cache *c;
+
+	*table = NULL;
+
+	c = search(&q->root, offset);
+	if (!c)
+		return -1;
+
+	/* Update the LRU state */
+	list_del_init(&c->list);
+	list_add_tail(&c->list, &q->lru_list);
+
+	*table = c->table;
+	return 0;
+}
+
 static inline u64 get_l1_index(struct qcow *q, u64 offset)
 {
 	struct qcow_header *header = q->header;
@@ -37,6 +171,43 @@  static inline u64 get_cluster_offset(struct qcow *q, u64 offset)
 	return offset & ((1 << header->cluster_bits)-1);
 }
 
+static int qcow_read_l2_table(struct qcow *q, u64 **table, u64 offset)
+{
+	struct qcow_header *header = q->header;
+	u64 size;
+	u64 i;
+	u64 *t;
+
+	*table = NULL;
+	size   = 1 << header->l2_bits;
+
+	/* search an entry for offset in cache */
+	if (search_table(q, table, offset) >= 0)
+		return 0;
+
+	t = calloc(size, sizeof(*t));
+	if (!t)
+		goto error;
+
+	/* table not cached: read from the disk */
+	if (pread_in_full(q->fd, t, size * sizeof(u64), offset) < 0)
+		goto error;
+
+	/* cache the table */
+	if (cache_table(q, t, offset) < 0)
+		goto error;
+
+	/* change cached table to CPU's byte-order */
+	for (i = 0; i < size; i++)
+		be64_to_cpus(&t[i]);
+
+	*table = t;
+	return 0;
+error:
+	free(t);
+	return -1;
+}
+
 static ssize_t qcow_read_cluster(struct qcow *q, u64 offset, void *dst, u32 dst_len)
 {
 	struct qcow_header *header = q->header;
@@ -51,7 +222,6 @@  static ssize_t qcow_read_cluster(struct qcow *q, u64 offset, void *dst, u32 dst_
 	u64 l1_idx;
 	u64 l2_idx;
 
-	l2_table = NULL;
 
 	cluster_size = 1 << header->cluster_bits;
 
@@ -72,18 +242,16 @@  static ssize_t qcow_read_cluster(struct qcow *q, u64 offset, void *dst, u32 dst_
 		goto zero_cluster;
 
 	l2_table_size = 1 << header->l2_bits;
-	l2_table = calloc(l2_table_size, sizeof(u64));
-	if (!l2_table)
-		goto out_error;
 
-	if (pread_in_full(q->fd, l2_table, l2_table_size * sizeof(u64), l2_table_offset) < 0)
+	/* read and cache level 2 table */
+	if (qcow_read_l2_table(q, &l2_table, l2_table_offset) < 0)
 		goto out_error;
 
 	l2_idx = get_l2_index(q, offset);
 	if (l2_idx >= l2_table_size)
 		goto out_error;
 
-	clust_start = be64_to_cpu(l2_table[l2_idx]) & ~header->oflag_mask;
+	clust_start = l2_table[l2_idx] & ~header->oflag_mask;
 	if (!clust_start)
 		goto zero_cluster;
 
@@ -91,7 +259,6 @@  static ssize_t qcow_read_cluster(struct qcow *q, u64 offset, void *dst, u32 dst_
 		goto out_error;
 
 out:
-	free(l2_table);
 	return length;
 
 zero_cluster:
@@ -221,15 +388,16 @@  static ssize_t qcow_write_cluster(struct qcow *q, u64 offset, void *buf, u32 src
 	if (len > src_len)
 		len = src_len;
 
-	l2t = calloc(l2t_sz, sizeof(u64));
-	if (!l2t)
-		goto error;
-
 	l2t_off		= table->l1_table[l1t_idx] & ~header->oflag_mask;
 	if (l2t_off) {
-		if (pread_in_full(q->fd, l2t, l2t_sz * sizeof(u64), l2t_off) < 0)
-			goto free_l2;
+		/* cache level2 table */
+		if (qcow_read_l2_table(q, &l2t, l2t_off) < 0)
+			goto error;
 	} else {
+		l2t = calloc(l2t_sz, sizeof(*l2t));
+		if (!l2t)
+			goto error;
+
 		/* Capture the state of the consistent QCOW image */
 		f_sz		= file_size(q->fd);
 		if (!f_sz)
@@ -251,6 +419,13 @@  static ssize_t qcow_write_cluster(struct qcow *q, u64 offset, void *buf, u32 src
 			goto free_l2;
 		}
 
+		if (cache_table(q, l2t, l2t_off) < 0) {
+			if (ftruncate(q->fd, f_sz) < 0)
+				goto free_l2;
+
+			goto free_l2;
+		}
+
 		/* Update the in-core entry */
 		table->l1_table[l1t_idx] = l2t_off;
 	}
@@ -258,17 +433,15 @@  static ssize_t qcow_write_cluster(struct qcow *q, u64 offset, void *buf, u32 src
 	/* Capture the state of the consistent QCOW image */
 	f_sz		= file_size(q->fd);
 	if (!f_sz)
-		goto free_l2;
+		goto error;
 
-	clust_start	= be64_to_cpu(l2t[l2t_idx]) & ~header->oflag_mask;
+	clust_start	= l2t[l2t_idx] & ~header->oflag_mask;
 	if (!clust_start) {
 		clust_start	= ALIGN(f_sz, clust_sz);
 		update_meta	= true;
 	} else
 		update_meta	= false;
 
-	free(l2t);
-
 	/* Write actual data */
 	if (pwrite_in_full(q->fd, buf, len, clust_start + clust_off) < 0)
 		goto error;
@@ -282,9 +455,13 @@  static ssize_t qcow_write_cluster(struct qcow *q, u64 offset, void *buf, u32 src
 
 			goto error;
 		}
+
+		/* Update the cached level2 entry */
+		l2t[l2t_idx] = clust_start;
 	}
 
 	return len;
+
 free_l2:
 	free(l2t);
 error:
@@ -336,6 +513,7 @@  static int qcow_disk_close(struct disk_image *disk)
 
 	q = disk->priv;
 
+	free_cache(q);
 	free(q->table.l1_table);
 	free(q->header);
 	free(q);
@@ -428,6 +606,8 @@  static struct disk_image *qcow2_probe(int fd, bool readonly)
 		goto error;
 
 	q->fd = fd;
+	q->root = RB_ROOT;
+	INIT_LIST_HEAD(&q->lru_list);
 
 	h = q->header = qcow2_read_header(fd);
 	if (!h)
@@ -525,6 +705,8 @@  static struct disk_image *qcow1_probe(int fd, bool readonly)
 		goto error;
 
 	q->fd = fd;
+	q->root = RB_ROOT;
+	INIT_LIST_HEAD(&q->lru_list);
 
 	h = q->header = qcow1_read_header(fd);
 	if (!h)
diff --git a/tools/kvm/include/kvm/qcow.h b/tools/kvm/include/kvm/qcow.h
index b6e7493..1663819 100644
--- a/tools/kvm/include/kvm/qcow.h
+++ b/tools/kvm/include/kvm/qcow.h
@@ -3,6 +3,8 @@ 
 
 #include <linux/types.h>
 #include <stdbool.h>
+#include <linux/rbtree.h>
+#include <linux/list.h>
 
 #define QCOW_MAGIC		(('Q' << 24) | ('F' << 16) | ('I' << 8) | 0xfb)
 
@@ -17,6 +19,16 @@ 
 #define QCOW2_OFLAG_COMPRESSED	(1LL << 62)
 #define QCOW2_OFLAG_MASK	(QCOW2_OFLAG_COPIED|QCOW2_OFLAG_COMPRESSED)
 
+#define MAX_CACHE_NODES         128
+
+struct qcow_l2_cache {
+	u64                     offset;
+	u64                     *table;
+	struct rb_node          node;
+	struct list_head        list;
+	struct qcow             *qcow;
+};
+
 struct qcow_table {
 	u32			table_size;
 	u64			*l1_table;
@@ -26,6 +38,11 @@  struct qcow {
 	void			*header;
 	struct qcow_table	table;
 	int			fd;
+
+	/* Level2 caching data structures */
+	struct rb_root          root;
+	struct list_head        lru_list;
+	int                     no_cached;
 };
 
 struct qcow_header {