mbox series

[0/2] XArray multi-index tests

Message ID 20231104005747.1389762-1-da.gomez@samsung.com
Headers
Series XArray multi-index tests |

Message

Daniel Gomez Nov. 4, 2023, 12:57 a.m. UTC 
  Add multi-index XArray tests.

Tests were first introduced as part of '[RFC PATCH 00/11] shmem: high
order folios support in write path' [1]. I've created this new series with
Matthew Wilcox's feedback [2] regarding test 'XArray: add cmpxchg order test'.

[1] https://lore.kernel.org/all/20231028211518.3424020-1-da.gomez@samsung.com/
[2] https://lore.kernel.org/all/ZT68dBiJKNLXLRZA@casper.infradead.org/

The full node is lost (not only the order) when using xa_cmpxchg with NULL
entry. Adding another (FIVE) entry at '1 << order' with the order information
does not keep the original node but duplicates it. I'm not sure if that's
the intention from the proposal you mentioned in [2] but please, let me know
your comments.

Changes since RFC:
* Update cmpxchg test to include another entry at 1 << order that
'keeps' the node around and order information.
* Update cmpxchg test to verify the entries and order in all tied
indexes.
* Drop previous Luis Chamberlain's review as changes are significant
from the RFC.

Daniel

Daniel Gomez (1):
  XArray: add cmpxchg order test

Luis Chamberlain (1):
  test_xarray: add tests for advanced multi-index use

 lib/test_xarray.c | 188 ++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 188 insertions(+)

-- 
2.39.2

Comments

Daniel Gomez Nov. 4, 2023, 1 a.m. UTC | #1 
On Sat, Nov 04, 2023 at 01:57:45AM +0100, Daniel Gomez wrote:
> Add multi-index XArray tests.
>
> Tests were first introduced as part of '[RFC PATCH 00/11] shmem: high
> order folios support in write path' [1]. I've created this new series with
> Matthew Wilcox's feedback [2] regarding test 'XArray: add cmpxchg order test'.
>
> [1] https://lore.kernel.org/all/20231028211518.3424020-1-da.gomez@samsung.com/
> [2] https://lore.kernel.org/all/ZT68dBiJKNLXLRZA@casper.infradead.org/
>
> The full node is lost (not only the order) when using xa_cmpxchg with NULL
> entry. Adding another (FIVE) entry at '1 << order' with the order information
> does not keep the original node but duplicates it. I'm not sure if that's
> the intention from the proposal you mentioned in [2] but please, let me know
> your comments.
>
> Changes since RFC:
> * Update cmpxchg test to include another entry at 1 << order that
> 'keeps' the node around and order information.
> * Update cmpxchg test to verify the entries and order in all tied
> indexes.
> * Drop previous Luis Chamberlain's review as changes are significant
> from the RFC.
* Make cmpxchg_order test conditional to CONFIG_XARRAY_MULTI.

>
> Daniel
>
> Daniel Gomez (1):
>   XArray: add cmpxchg order test
>
> Luis Chamberlain (1):
>   test_xarray: add tests for advanced multi-index use
>
>  lib/test_xarray.c | 188 ++++++++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 188 insertions(+)
>
> --
> 2.39.2