Re: [PATCH i-g-t v2] lib/igt_psr: Fix Early Transport status handling

["Hogander, Jouni" <jouni.hogander@intel.com>]

On Wed, 2026-03-18 at 01:00 +0530, Naladala Ramanaidu wrote:
> Switch PSR status reads to the connector‑specific debugfs path
> to avoid ambiguity on dual eDP or multi‑output systems. This
> ensures the Early Transport state is checked for the correct
> connector.
> 
> Update the Early Transport debugfs handling in line with changes
> introduced in KMD (commit c390bf07961b, "drm/i915/dp: Fix panel
> replay when DSC is enabled").
> 
> v2: Addressed review comments. (jeevan)
> 
> Link:
> https://lore.kernel.org/all/20251015161934.262108-8-imre.deak@intel.com/
> 
> Signed-off-by: Naladala Ramanaidu <ramanaidu.naladala@intel.com>
> ---
>  lib/igt_psr.c             | 8 +++++---
>  lib/igt_psr.h             | 2 +-
>  tests/intel/kms_psr2_sf.c | 2 +-
>  3 files changed, 7 insertions(+), 5 deletions(-)
> 
> diff --git a/lib/igt_psr.c b/lib/igt_psr.c
> index 7e50c3dc6..09c34c97d 100644
> --- a/lib/igt_psr.c
> +++ b/lib/igt_psr.c
> @@ -69,14 +69,16 @@ static bool psr_active_sink_check(int debugfs_fd,
> igt_output_t *output)
>  /*
>   * Checks if Early Transport is enabled in PSR status by reading the
> debugfs.
>   */
> -bool early_transport_check(int debugfs_fd)
> +bool early_transport_check(int debugfs_fd, igt_output_t *output)
>  {
>  	char buf[PSR_STATUS_MAX_LEN];
> +	char debugfs_file[128] = {0};
>  
> -	igt_debugfs_simple_read(debugfs_fd, "i915_edp_psr_status",
> buf,
> +	SET_DEBUGFS_PATH(output, debugfs_file);
> +	igt_debugfs_simple_read(debugfs_fd, debugfs_file, buf,
>  		 		sizeof(buf));
>  
> -	return strstr(buf, "enabled (Early Transport)");
> +	return strstr(buf, "Early Transport");

Checked this and to me it looks like it may indicate improperly ET
being enabled. This is what I'm seeing in debugfs entry:

Sink support: PSR = yes [0x04] (Early Transport), Panel Replay = yes,
Panel Replay Selective Update = yes (Early Transport)
PSR mode: Panel Replay Selective Update enabled (Early Transport)
Source PSR/PanelReplay ctl: enabled [0x40000000]
PSR2_CTL: 0x08000000
Source PSR/PanelReplay status: SLEEP [0x30200011]
Busy frontbuffer bits: 0x00000000
Performance counter: 0
Frame:	PSR2 SU blocks:
0	33
1	256
2	512
3	0
4	0
5	0
6	0
7	0
PSR2 selective fetch: enabled

So I think panel stating ET support already returns true if using this
check?

BR,
Jouni Högander

>  }
>  
>  static bool sink_status_checks(void)
> diff --git a/lib/igt_psr.h b/lib/igt_psr.h
> index 7639f8d46..53ab04483 100644
> --- a/lib/igt_psr.h
> +++ b/lib/igt_psr.h
> @@ -49,7 +49,7 @@ enum fbc_mode {
>  
>  bool psr_disabled_check(int debugfs_fd);
>  bool selective_fetch_check(int debugfs_fd, igt_output_t *output);
> -bool early_transport_check(int debugfs_fd);
> +bool early_transport_check(int debugfs_fd, igt_output_t *output);
>  bool psr_wait_entry(int debugfs_fd, enum psr_mode mode, igt_output_t
> *output);
>  bool psr_wait_update(int debugfs_fd, enum psr_mode mode,
> igt_output_t *output);
>  bool psr_long_wait_update(int debugfs_fd, enum psr_mode mode,
> igt_output_t *output);
> diff --git a/tests/intel/kms_psr2_sf.c b/tests/intel/kms_psr2_sf.c
> index 1cc57ad35..7503c7682 100644
> --- a/tests/intel/kms_psr2_sf.c
> +++ b/tests/intel/kms_psr2_sf.c
> @@ -961,7 +961,7 @@ static void run(data_t *data)
>  			     "FBC still disabled\n");
>  
>  	if (is_et_check_needed(data))
> -		igt_assert_f(early_transport_check(data-
> >debugfs_fd),
> +		igt_assert_f(early_transport_check(data->debugfs_fd,
> data->output),
>  			     "Early Transport Disabled\n");
>  
>  	data->screen_changes = 0;