<!DOCTYPE html><html><head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
  </head>
  <body>
    <p><br>
    </p>
    <div class="moz-cite-prefix">On 4/29/2024 18:42, Lucas De Marchi
      wrote:<br>
    </div>
    <blockquote type="cite" cite="mid:a3c4pszai2ysc74yod3yj7ulridwqrztotyu6lrtm3n2f3ghgu@y6lh66mlz2xe">On
      Mon, Apr 29, 2024 at 09:14:30AM GMT, Chauhan, Shekhar wrote:
      <br>
      <blockquote type="cite">
        <br>
        On 4/27/2024 19:23, Lucas De Marchi wrote:
        <br>
        <blockquote type="cite">In order to detect duplicate
          implementations for the same workaround,
          <br>
          early in the implementation of RTP it was decided to error out
          even if
          <br>
          the values set are exactly the same. With the introduction of
          18034896535
          <br>
          in commit 74671d23ca18 ("drm/xe/xe2: Add workaround
          18034896535"), LNL
          <br>
          stepping with graphics stepping A1 now gives the following
          error on
          <br>
          module load:
          <br>
          <br>
              xe 0000:00:02.0: [drm] *ERROR* GT0: [GT OTHER] \
          <br>
              discarding save-restore reg e48c (clear: 00000200, set:
          00000200,\
          <br>
              masked: yes, mcr: yes): ret=-22
          <br>
          <br>
          RTP may be improved in the future, but for now simply join the
          entries
          <br>
          like done with e.g. "1607297627, 1607030317, 1607186500".
          <br>
          <br>
          Fixes: 74671d23ca18 ("drm/xe/xe2: Add workaround 18034896535")
          <br>
          Cc: Bommu Krishnaiah <a class="moz-txt-link-rfc2396E" href="mailto:krishnaiah.bommu@intel.com"><krishnaiah.bommu@intel.com></a>
          <br>
          Cc: Tejas Upadhyay <a class="moz-txt-link-rfc2396E" href="mailto:tejas.upadhyay@intel.com"><tejas.upadhyay@intel.com></a>
          <br>
          Cc: Matt Roper <a class="moz-txt-link-rfc2396E" href="mailto:matthew.d.roper@intel.com"><matthew.d.roper@intel.com></a>
          <br>
          Signed-off-by: Lucas De Marchi
          <a class="moz-txt-link-rfc2396E" href="mailto:lucas.demarchi@intel.com"><lucas.demarchi@intel.com></a>
          <br>
          ---
          <br>
           drivers/gpu/drm/xe/xe_wa.c | 7 +------
          <br>
           1 file changed, 1 insertion(+), 6 deletions(-)
          <br>
          <br>
          diff --git a/drivers/gpu/drm/xe/xe_wa.c
          b/drivers/gpu/drm/xe/xe_wa.c
          <br>
          index 9d9b7fa7a8f0..db7c7c7875c5 100644
          <br>
          --- a/drivers/gpu/drm/xe/xe_wa.c
          <br>
          +++ b/drivers/gpu/drm/xe/xe_wa.c
          <br>
          @@ -449,12 +449,7 @@ static const struct xe_rtp_entry_sr
          engine_was[] = {
          <br>
                 XE_RTP_RULES(GRAPHICS_VERSION(2004),
          FUNC(xe_rtp_match_first_render_or_compute)),
          <br>
                 XE_RTP_ACTIONS(SET(ROW_CHICKEN3,
          XE2_EUPEND_CHK_FLUSH_DIS))
          <br>
               },
          <br>
          -    { XE_RTP_NAME("16021540221"),
          <br>
          -      XE_RTP_RULES(GRAPHICS_VERSION(2004), GRAPHICS_STEP(A0,
          B0),
          <br>
          -               FUNC(xe_rtp_match_first_render_or_compute)),
          <br>
          -      XE_RTP_ACTIONS(SET(ROW_CHICKEN4, DISABLE_TDL_PUSH))
          <br>
          -    },
          <br>
          -    { XE_RTP_NAME("18034896535"),
          <br>
          +    { XE_RTP_NAME("18034896535, 16021540221"), /*
          16021540221: GRAPHICS_STEP(A0, B0) */
          <br>
                 XE_RTP_RULES(GRAPHICS_VERSION_RANGE(2001, 2004),
          <br>
        </blockquote>
        <br>
        w/a 16021540221 is applicable to only Steppings A0 to B0 (B0 not
        inclusive). With your change, I think the w/a will be applied to
        all the steppings. Wouldn't this be incorrect? Please provide
        some guidance.
        <br>
      </blockquote>
      <br>
      both 18034896535 and 16021540221 have the same action:
      <br>
      <br>
          SET(ROW_CHICKEN4, DISABLE_TDL_PUSH)
      <br>
      <br>
      In other words, 18034896535 applies to a superset of platforms
      <br>
      16021540221 applies to. So even when we remove support for
      steppings A*,
      <br>
      we want to keep the WA (possibly removing the comment about
      16021540221
      <br>
      and renaming it).
      <br>
    </blockquote>
    <p>Okay, this makes sense. Thanks. <br>
    </p>
    <p><i>-shekhar</i><br>
    </p>
    <blockquote type="cite" cite="mid:a3c4pszai2ysc74yod3yj7ulridwqrztotyu6lrtm3n2f3ghgu@y6lh66mlz2xe">
      <br>
      Lucas De Marchi
      <br>
      <br>
      <blockquote type="cite">
        <br>
        <blockquote type="cite">               
          FUNC(xe_rtp_match_first_render_or_compute)),
          <br>
                 XE_RTP_ACTIONS(SET(ROW_CHICKEN4, DISABLE_TDL_PUSH))
          <br>
        </blockquote>
        <br>
        -- <br>
        -shekhar
        <br>
        <br>
      </blockquote>
    </blockquote>
    <pre class="moz-signature" cols="72">-- 
-shekhar</pre>
  </body>
</html>