Re: [PATCH 12/22] gpu: host1x: Correct host1x_job_pin() error handling

On Tue, May 23, 2017 at 2:14 AM, Dmitry Osipenko digetx@gmail.com wrote:

In case of relocations / waitchecks patching failure the jobs pins stay referenced till DRM file get closed, wasting memory. Add the missed unpinning.

Signed-off-by: Dmitry Osipenko digetx@gmail.com

drivers/gpu/host1x/job.c | 16 +++++++--------- 1 file changed, 7 insertions(+), 9 deletions(-)

diff --git a/drivers/gpu/host1x/job.c b/drivers/gpu/host1x/job.c index d9933828fe87..14f3f957ffab 100644 --- a/drivers/gpu/host1x/job.c +++ b/drivers/gpu/host1x/job.c @@ -592,22 +592,20 @@ int host1x_job_pin(struct host1x_job *job, struct device *dev)

            err = do_relocs(job, g->bo);
            if (err)

•                   break;
  

•                   goto out;
  
            err = do_waitchks(job, host, g->bo);
            if (err)
  

•                   break;
  

•                   goto out;
    }
  
  

•   if (IS_ENABLED(CONFIG_TEGRA_HOST1X_FIREWALL) && !err) {
  

•           err = copy_gathers(job, dev);
  

•           if (err) {
  

•                   host1x_job_unpin(job);
  

•                   return err;
  

•           }
  

•   }
  

•   if (!IS_ENABLED(CONFIG_TEGRA_HOST1X_FIREWALL))
  

•           goto out;
  
  

•   err = copy_gathers(job, dev);
  

out:

•   if (err)
  

•           host1x_job_unpin(job);
    wmb();
  
    return err;
  

-- 2.13.0

One subtle undocumented change here, is that wmb() now gets called in the copy_gathers()-error case (just like it already does for the other error-cases). That's seems like an OK change, so:

Reviewed-by: Erik Faye-Lund kusmabite@gmail.com