mbox series

[0/2] ublk: improve ublk device deletion

Message ID 20240223075539.89945-1-ming.lei@redhat.com (mailing list archive)
Headers show
Series ublk: improve ublk device deletion | expand

Message

Ming Lei Feb. 23, 2024, 7:55 a.m. UTC 
Hello,

The 1st patch cleans up get/put device, and annotate them
via noline for trace purpose.

The 2nd patch adds UBLK_U_CMD_DEL_DEV_ASYNC so userspace device
deletion can be implemented easier.

Ming Lei (2):
  ublk: improve getting & putting ublk device
  ublk: add UBLK_CMD_DEL_DEV_ASYNC

 drivers/block/ublk_drv.c      | 21 +++++++++++++--------
 include/uapi/linux/ublk_cmd.h |  2 ++
 2 files changed, 15 insertions(+), 8 deletions(-)

Comments

Ming Lei Feb. 29, 2024, 1:27 a.m. UTC | #1 
On Fri, Feb 23, 2024 at 03:55:37PM +0800, Ming Lei wrote:
> Hello,
> 
> The 1st patch cleans up get/put device, and annotate them
> via noline for trace purpose.
> 
> The 2nd patch adds UBLK_U_CMD_DEL_DEV_ASYNC so userspace device
> deletion can be implemented easier.

Hello Jens,


All APIs in lib/ are built as notrace, so the 1st patch adds
noline for ublk get/put device(slow path), and it is helpful for
investigating reference issue by existed trace utilities.

The 2nd one makes userspace happy to not consider complicated dependency
issue in removing code path.

Can you queue the two patches for 6.9 if you are fine?

Thanks,
Ming
Jens Axboe Feb. 29, 2024, 1:35 a.m. UTC | #2 
On 2/28/24 6:27 PM, Ming Lei wrote:
> On Fri, Feb 23, 2024 at 03:55:37PM +0800, Ming Lei wrote:
>> Hello,
>>
>> The 1st patch cleans up get/put device, and annotate them
>> via noline for trace purpose.
>>
>> The 2nd patch adds UBLK_U_CMD_DEL_DEV_ASYNC so userspace device
>> deletion can be implemented easier.
> 
> Hello Jens,
> 
> 
> All APIs in lib/ are built as notrace, so the 1st patch adds
> noline for ublk get/put device(slow path), and it is helpful for
> investigating reference issue by existed trace utilities.
> 
> The 2nd one makes userspace happy to not consider complicated dependency
> issue in removing code path.
> 
> Can you queue the two patches for 6.9 if you are fine?

Yep they look fine to me, will get them queued up. Thanks!
Jens Axboe Feb. 29, 2024, 1:47 a.m. UTC | #3 
On Fri, 23 Feb 2024 15:55:37 +0800, Ming Lei wrote:
> The 1st patch cleans up get/put device, and annotate them
> via noline for trace purpose.
> 
> The 2nd patch adds UBLK_U_CMD_DEL_DEV_ASYNC so userspace device
> deletion can be implemented easier.
> 
> Ming Lei (2):
>   ublk: improve getting & putting ublk device
>   ublk: add UBLK_CMD_DEL_DEV_ASYNC
> 
> [...]

Applied, thanks!

[1/2] ublk: improve getting & putting ublk device
      commit: 1221b9e982e181f1c37789c46fe5bfe32d97bec4
[2/2] ublk: add UBLK_CMD_DEL_DEV_ASYNC
      commit: 13fe8e6825e44129b6cbeee41d3012554bf8d687

Best regards,