On Thu, Mar 17, 2022 at 08:36:17PM +0200, Jani Nikula wrote:
Register the DMC debugfs file only on platforms that support
DMC. There's no point in having a no-op debugfs file.

It seems this would not change much the behavior (fail on open vs fail
on read). But the code in igt is suspicious:


        bool igt_pm_dmc_loaded(int debugfs)
        {
                char buf[15];
                int len;

                len = igt_sysfs_read(debugfs, "i915_dmc_info", buf, sizeof(buf) 
- 1);
                if (len < 0)
                        return true; /* no CSR support, no DMC requirement */

From a quick inspection of igt_sysfs_read() it seems it would just
return 0 if there's nothing to be read. And it would return < 0 on
failure to open the file.

These would be the affected tests:

tests/i915/i915_pm_rpm.c:
tests/i915/i915_pm_lpsp.c:
tests/i915/i915_pm_dc.c:
        igt_require(igt_pm_dmc_loaded(data.debugfs_fd));


Lucas De Marchi


Signed-off-by: Jani Nikula <[email protected]>
---
drivers/gpu/drm/i915/display/intel_dmc.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/i915/display/intel_dmc.c 
b/drivers/gpu/drm/i915/display/intel_dmc.c
index 5de13f978e57..8dfa2aa9f8bd 100644
--- a/drivers/gpu/drm/i915/display/intel_dmc.c
+++ b/drivers/gpu/drm/i915/display/intel_dmc.c
@@ -818,9 +818,6 @@ static int intel_dmc_debugfs_status_show(struct seq_file 
*m, void *unused)
        struct intel_dmc *dmc;
        i915_reg_t dc5_reg, dc6_reg = INVALID_MMIO_REG;

-       if (!HAS_DMC(i915))
-               return -ENODEV;
-
        dmc = &i915->dmc;

        wakeref = intel_runtime_pm_get(&i915->runtime_pm);
@@ -890,6 +887,9 @@ void intel_dmc_debugfs_register(struct drm_i915_private 
*i915)
{
        struct drm_minor *minor = i915->drm.primary;

+       if (!HAS_DMC(i915))
+               return;
+
        debugfs_create_file("i915_dmc_info", 0444, minor->debugfs_root,
                            i915, &intel_dmc_debugfs_status_fops);
}
--
2.30.2

Reply via email to