From: Alex Hung <alex.hung@amd.com>
To: James Lin <PingLei.Lin@amd.com>, igt-dev@lists.freedesktop.org
Cc: sunpeng.li@amd.com, chiahsuan.chung@amd.com
Subject: Re: [PATCH i-g-t] tests/amdgpu: Resolve DRM card index dynamically instead of hardcoding card0
Date: Tue, 23 Jun 2026 11:18:02 -0600 [thread overview]
Message-ID: <18e2630f-dffc-4759-be3d-158ae33dee64@amd.com> (raw)
In-Reply-To: <20260623074749.13734-1-PingLei.Lin@amd.com>
LGTM.
Reviewed-by: Alex Hung <alex.hung@amd.com>
On 6/23/26 01:47, James Lin wrote:
> [Why]
> amd_assr and amd_abm build connector sysfs paths assuming the amdgpu
> device is card0. On systems where simpledrm registers the boot
> framebuffer first, it takes card0 and amdgpu comes up as card1.
> Since DRM minor numbers are never reclaimed, amdgpu stays card1 even
> after simpledrm goes away, so the hardcoded "card0-<connector>" paths
> do not exist and find_aux_dev() and set_abm_level() fail, preventing
> the tests from running.
>
> [How]
> Open the connector's sysfs directory via igt_connector_sysfs_open(),
> which derives the real DRM card index, and access the AUX device and
> panel_power_savings relative to that directory instead of building a
> card0 path by hand.
>
> Reviewed-by: Tom Chung <chiahsuan.chung@amd.com>
> Signed-off-by: James Lin <PingLei.Lin@amd.com>
> ---
> tests/amdgpu/amd_abm.c | 27 ++++++++++++++++++---------
> tests/amdgpu/amd_assr.c | 26 ++++++++++++++------------
> 2 files changed, 32 insertions(+), 21 deletions(-)
>
> diff --git a/tests/amdgpu/amd_abm.c b/tests/amdgpu/amd_abm.c
> index 207d075b0..ba90344d3 100644
> --- a/tests/amdgpu/amd_abm.c
> +++ b/tests/amdgpu/amd_abm.c
> @@ -35,7 +35,7 @@
> #define DEBUGFS_CURRENT_BACKLIGHT_PWM "amdgpu_current_backlight_pwm"
> #define DEBUGFS_TARGET_BACKLIGHT_PWM "amdgpu_target_backlight_pwm"
> #define BACKLIGHT_PATH "/sys/class/backlight/amdgpu_bl0"
> -#define PANEL_POWER_SAVINGS_PATH "/sys/class/drm/card0-%s/amdgpu/panel_power_savings"
> +#define PANEL_POWER_SAVINGS_ATTR "amdgpu/panel_power_savings"
> #define MK_COLOR(r, g, b) ((0 << 24) | (r << 16) | (g << 8) | b)
>
> typedef struct data {
> @@ -252,14 +252,23 @@ static int backlight_write_brightness(int value)
> static void set_abm_level(data_t *data, igt_output_t *output, int level)
> {
> char buf[PATH_MAX];
> - int fd;
> -
> - igt_assert(snprintf(buf, PATH_MAX, PANEL_POWER_SAVINGS_PATH,
> - output->name) < PATH_MAX);
> -
> - fd = open(buf, O_WRONLY);
> -
> - igt_skip_on_f(fd == -1, "Cannot find %s. Is it an OLED?\n", buf);
> + int fd, conn_fd;
> +
> + /* Open the connector's sysfs directory via the IGT helper, which
> + * derives the real DRM card index instead of assuming card0. On
> + * systems where the amdgpu device is not card0 (e.g. another DRM
> + * node grabbed card0), the previous hardcoded "card0-<conn>" path
> + * did not exist and panel_power_savings could never be found.
> + */
> + conn_fd = igt_connector_sysfs_open(data->drm_fd, output->config.connector);
> + igt_skip_on_f(conn_fd < 0, "Cannot open sysfs dir for connector %s\n",
> + output->name);
> +
> + fd = openat(conn_fd, PANEL_POWER_SAVINGS_ATTR, O_WRONLY);
> + close(conn_fd);
> +
> + igt_skip_on_f(fd == -1, "Cannot find %s for %s. Is it an OLED?\n",
> + PANEL_POWER_SAVINGS_ATTR, output->name);
>
> igt_assert_eq(snprintf(buf, sizeof(buf), "%d", level),
> write(fd, buf, 1));
> diff --git a/tests/amdgpu/amd_assr.c b/tests/amdgpu/amd_assr.c
> index 71209e79f..ae8420183 100644
> --- a/tests/amdgpu/amd_assr.c
> +++ b/tests/amdgpu/amd_assr.c
> @@ -63,25 +63,27 @@ static void test_fini(data_t *data)
> static char *find_aux_dev(data_t *data, igt_output_t *output,
> char *aux_dev, size_t max_aux_dev_len)
> {
> - char sysfs_name[PATH_MAX] = { 0 };
> - /* +7 only to get rid of snprintf_chk warning.
> - * Path name cannot exceed the size of PATH_MAX anyway.
> - */
> - char conn_dir_name[PATH_MAX+7] = { 0 };
> DIR *dir;
> struct dirent *dirent;
> + int conn_fd;
>
> aux_dev[0] = 0;
>
> - if(igt_sysfs_path(data->fd, sysfs_name, sizeof(sysfs_name))) {
> - snprintf(conn_dir_name, sizeof(conn_dir_name),
> - "%s%scard0%s%s",
> - sysfs_name, "/", "-", output->name);
> - }
> + /* Open the connector's sysfs directory via the IGT helper, which
> + * derives the real DRM card index instead of assuming card0. On
> + * systems where the amdgpu device is not card0 (e.g. another DRM
> + * node grabbed card0), the previous hardcoded "card0-<conn>" path
> + * did not exist and the AUX device could never be found.
> + */
> + conn_fd = igt_connector_sysfs_open(data->fd, output->config.connector);
> + if (conn_fd < 0)
> + return NULL;
>
> - dir = opendir(conn_dir_name);
> - if (!dir)
> + dir = fdopendir(conn_fd);
> + if (!dir) {
> + close(conn_fd);
> return NULL;
> + }
>
> while((dirent = readdir(dir))) {
> if (strncmp(dirent->d_name, "drm_dp_aux", sizeof("drm_dp_aux")-1))
prev parent reply other threads:[~2026-06-23 17:18 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-06-23 7:47 [PATCH i-g-t] tests/amdgpu: Resolve DRM card index dynamically instead of hardcoding card0 James Lin
2026-06-23 8:33 ` ✓ Xe.CI.BAT: success for " Patchwork
2026-06-23 8:57 ` ✓ i915.CI.BAT: " Patchwork
2026-06-23 10:56 ` ✗ Xe.CI.FULL: failure " Patchwork
2026-06-23 11:08 ` Patchwork
2026-06-23 15:12 ` ✗ i915.CI.Full: " Patchwork
2026-06-23 17:18 ` Alex Hung [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=18e2630f-dffc-4759-be3d-158ae33dee64@amd.com \
--to=alex.hung@amd.com \
--cc=PingLei.Lin@amd.com \
--cc=chiahsuan.chung@amd.com \
--cc=igt-dev@lists.freedesktop.org \
--cc=sunpeng.li@amd.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox