Re: [PATCH v5 00/21] block: remove aio_disable_external() API

[Kevin Wolf]

Am 04.05.2023 um 21:53 hat Stefan Hajnoczi geschrieben:
> v5:
> - Use atomic accesses for in_flight counter in vhost-user-server.c [Kevin]
> - Stash SCSIDevice id/lun values for VIRTIO_SCSI_T_TRANSPORT_RESET event
>   before unrealizing the SCSIDevice [Kevin]
> - Keep vhost-user-blk export .detach() callback so ctx is set to NULL [Kevin]
> - Narrow BdrvChildClass and BlockDriver drained_{begin/end/poll} callbacks 
> from
>   IO_OR_GS_CODE() to GLOBAL_STATE_CODE() [Kevin]
> - Include Kevin's "block: Fix use after free in blockdev_mark_auto_del()" to
>   fix a latent bug that was exposed by this series
> 
> v4:
> - Remove external_disable_cnt variable [Philippe]
> - Add Patch 1 to fix assertion failure in .drained_end() -> 
> blk_get_aio_context()
> v3:
> - Resend full patch series. v2 was sent in the middle of a git rebase and was
>   missing patches. [Eric]
> - Apply Reviewed-by tags.
> v2:
> - Do not rely on BlockBackend request queuing, implement .drained_begin/end()
>   instead in xen-block, virtio-blk, and virtio-scsi [Paolo]
> - Add qdev_is_realized() API [Philippe]
> - Add patch to avoid AioContext lock around blk_exp_ref/unref() [Paolo]
> - Add patch to call .drained_begin/end() from main loop thread to simplify
>   callback implementations
> 
> The aio_disable_external() API temporarily suspends file descriptor monitoring
> in the event loop. The block layer uses this to prevent new I/O requests being
> submitted from the guest and elsewhere between bdrv_drained_begin() and
> bdrv_drained_end().
> 
> While the block layer still needs to prevent new I/O requests in drained
> sections, the aio_disable_external() API can be replaced with
> .drained_begin/end/poll() callbacks that have been added to BdrvChildClass and
> BlockDevOps.
> 
> This newer .bdrained_begin/end/poll() approach is attractive because it works
> without specifying a specific AioContext. The block layer is moving towards
> multi-queue and that means multiple AioContexts may be processing I/O
> simultaneously.
> 
> The aio_disable_external() was always somewhat hacky. It suspends all file
> descriptors that were registered with is_external=true, even if they have
> nothing to do with the BlockDriverState graph nodes that are being drained.
> It's better to solve a block layer problem in the block layer than to have an
> odd event loop API solution.
> 
> The approach in this patch series is to implement BlockDevOps
> .drained_begin/end() callbacks that temporarily stop file descriptor handlers.
> This ensures that new I/O requests are not submitted in drained sections.

Patches 2-16: Reviewed-by: Kevin Wolf <kwolf@redhat.com>