Re: [PATCH v3 3/5] locking/qspinlock: Introduce CNA into the slow path of qspinlock

[Peter Zijlstra <peterz@xxxxxxxxxxxxx>]

On Tue, Jul 16, 2019 at 10:16:29PM -0400, Waiman Long wrote:
>  A simple graphic to illustrate those queues will help too, for example

Very much yes!

> /*
>  * MCS lock holder
>  * ===============
>  *    mcs_node
>  *   +--------+      +----+         +----+
>  *   | next   | ---> |next| -> ...  |next| -> NULL  [Main queue]
>  *   | locked | -+   +----+         +----+
>  *   +--------+  |
>  *               |   +----+         +----+
>  *               +-> |next| -> ...  |next| -> X     [Secondary queue]
>  *    cna_node       +----+         +----+
>  *   +--------*                       ^
>  *   | tail   | ----------------------+
>  *   +--------*   

Almost; IIUC that cna_node is the same as the one from locked, so you
end up with something like:

>  *    mcs_node
>  *   +--------+      +----+         +----+
>  *   | next   | ---> |next| -> ...  |next| -> NULL  [Main queue]
>  *   | locked | -+   +----+         +----+
>  *   +--------+  |
>  *               |   +---------+         +----+
>  *               +-> |mcs::next| -> ...  |next| -> NULL     [Secondary queue]
>  *                   |cna::tail| -+      +----+
>  *                   +---------+  |        ^
>  *                                +--------+
>  *
>  * N.B. locked = 1 if secondary queue is absent.
>  */