Re: [PATCH v4 2/3] blk-cgroup: store blkcg in bio instead of blkg

[Nilay Shroff]

On 9/20/26 2:53 PM, Yu Kuai wrote:
> From: Yu Kuai<yukuai@xxxxxxx>
>
> A bio currently stores and pins a queue-local blkg. This forces bio
> association and remap paths to look up or create a blkg even when no
> blkcg policy will use the bio, and ties the stored state to the current
> block device.
>
> Store and reference the queue-independent blkcg in the bio instead. Add
> helpers that lazily look up or create the queue-local blkg when a policy
> needs it, and pin the result until the bio changes devices or releases
> its cgroup state. Look up and pin existing blkgs under RCU, taking the
> queue lock only when a blkg may need to be created.
>
> Keep completion and accounting users lookup-only when they do not create
> a blkg themselves. In particular, blk_cgroup_bio_start() only accounts a
> bio when an online blkg already exists. This avoids creating a blkg
> from the unconditional accounting hook when no policy is enabled.
>
> Protect every rhashtable lookup with an RCU read-side critical section.
> Annotate the lookup helpers and recursive wrappers with
> __must_hold_shared(RCU) so Clang can verify that contract.
>
> Before entering a bio-based driver's ->submit_bio(), drop the pinned blkg
> after blk_cgroup_bio_start() has completed bio accounting. blk-throttle is
> the only policy supported by bio-based queues, so no queue-local blkg
> state needs to enter stacking drivers. The queue-independent blkcg
> association remains attached to the bio.
>
> If blkg creation fails while walking down the hierarchy, use the closest
> available ancestor and update the bio's blkcg association before
> recording the blkg reference. This keeps later CSS ID hash lookups
> matched with the pinned blkg.
>
> Signed-off-by: Yu Kuai<yukuai@xxxxxxx>

Looks good to me.

Reviewed-by: Nilay Shroff <nilay@xxxxxxxxxxxxx>