<html>
  <head>
    <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
  </head>
  <body text="#000000" bgcolor="#FFFFFF">
    <p>Are you referring to insert_work->smp_mb(); ?</p>
    <p>Andrey<br>
    </p>
    <br>
    <div class="moz-cite-prefix">On 06/20/2018 10:50 AM, Christian König
      wrote:<br>
    </div>
    <blockquote type="cite"
      cite="mid:d96fc78e-ef3d-fb3a-8548-7a6a7cc3d52f@gmail.com">
      <blockquote type="cite" style="color: #000000;">+     * Try to
        submit work so SQ_EDC_INFO can be accessed from
        <br>
        +     * BH. If previous work submission hasn't finished yet
        <br>
        +     * just print whatever info is possible directly from the
        ISR.
        <br>
        +     */
        <br>
        +    if (work_pending(&adev->gfx.sq_work.work)) {
        <br>
        +        gfx_v8_0_parse_sq_irq(adev, ih_data);
        <br>
        +    } else {
        <br>
        +        adev->gfx.sq_work.ih_data = ih_data;
        <br>
        +        /* Verify the new value visible in BH handler */
        <br>
        +        smp_wmb();
        <br>
      </blockquote>
      <br>
      You can drop the barrier here and in gfx_v8_0_sq_irq_work_func(),
      the schedule_work() function is a barrier itself for both reads
      and writes anyway.
      <br>
      <br>
      Apart from that the both patches are Acked-by: Christian König <a
        class="moz-txt-link-rfc2396E"
        href="mailto:christian.koenig@amd.com" moz-do-not-send="true"><christian.koenig@amd.com></a>.
      <br>
      <br>
      Regards,
      <br>
      Christian.
    </blockquote>
    <br>
  </body>
</html>