[PATCH 13/20] drm/amd/display: Add disable timeout option
Alex Hung
alex.hung at amd.com
Wed Nov 8 18:44:28 UTC 2023
From: Duncan Ma <duncan.ma at amd.com>
[WHY]
Driver continues running whenever there is
is timeout from smu or dmcub.
It is difficult to track failure state
when dcn, dc or dmcub changes on root failure.
[HOW]
Add disable_timeout option to halt driver
whenever there is a failure in response.
Reviewed-by: Charlene Liu <charlene.liu at amd.com>
Acked-by: Alex Hung <alex.hung at amd.com>
Signed-off-by: Duncan Ma <duncan.ma at amd.com>
---
.../amd/display/dc/clk_mgr/dcn35/dcn35_smu.c | 3 ++
drivers/gpu/drm/amd/display/dc/dc.h | 1 +
drivers/gpu/drm/amd/display/dc/dc_dmub_srv.c | 29 +++++++++++++++----
3 files changed, 27 insertions(+), 6 deletions(-)
diff --git a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn35/dcn35_smu.c b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn35/dcn35_smu.c
index b6b8c3ca1572..af0a0f292595 100644
--- a/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn35/dcn35_smu.c
+++ b/drivers/gpu/drm/amd/display/dc/clk_mgr/dcn35/dcn35_smu.c
@@ -116,6 +116,9 @@ static uint32_t dcn35_smu_wait_for_response(struct clk_mgr_internal *clk_mgr, un
msleep(delay_us/1000);
else if (delay_us > 0)
udelay(delay_us);
+
+ if (clk_mgr->base.ctx->dc->debug.disable_timeout)
+ max_retries++;
} while (max_retries--);
return res_val;
diff --git a/drivers/gpu/drm/amd/display/dc/dc.h b/drivers/gpu/drm/amd/display/dc/dc.h
index c5e7da302acb..18ab453bfdb1 100644
--- a/drivers/gpu/drm/amd/display/dc/dc.h
+++ b/drivers/gpu/drm/amd/display/dc/dc.h
@@ -978,6 +978,7 @@ struct dc_debug_options {
bool psp_disabled_wa;
unsigned int ips2_eval_delay_us;
unsigned int ips2_entry_delay_us;
+ bool disable_timeout;
};
struct gpu_info_soc_bounding_box_v1_0;
diff --git a/drivers/gpu/drm/amd/display/dc/dc_dmub_srv.c b/drivers/gpu/drm/amd/display/dc/dc_dmub_srv.c
index 0e07699c1e83..53400cc05b5b 100644
--- a/drivers/gpu/drm/amd/display/dc/dc_dmub_srv.c
+++ b/drivers/gpu/drm/amd/display/dc/dc_dmub_srv.c
@@ -241,7 +241,12 @@ bool dc_dmub_srv_cmd_run_list(struct dc_dmub_srv *dc_dmub_srv, unsigned int coun
// Wait for DMUB to process command
if (wait_type != DM_DMUB_WAIT_TYPE_NO_WAIT) {
- status = dmub_srv_wait_for_idle(dmub, 100000);
+ if (dc_dmub_srv->ctx->dc->debug.disable_timeout) {
+ do {
+ status = dmub_srv_wait_for_idle(dmub, 100000);
+ } while (status != DMUB_STATUS_OK);
+ } else
+ status = dmub_srv_wait_for_idle(dmub, 100000);
if (status != DMUB_STATUS_OK) {
DC_LOG_DEBUG("No reply for DMUB command: status=%d\n", status);
@@ -1147,10 +1152,16 @@ bool dc_dmub_srv_is_hw_pwr_up(struct dc_dmub_srv *dc_dmub_srv, bool wait)
return true;
if (wait) {
- status = dmub_srv_wait_for_hw_pwr_up(dc_dmub_srv->dmub, 500000);
- if (status != DMUB_STATUS_OK) {
- DC_ERROR("Error querying DMUB hw power up status: error=%d\n", status);
- return false;
+ if (dc_dmub_srv->ctx->dc->debug.disable_timeout) {
+ do {
+ status = dmub_srv_wait_for_hw_pwr_up(dc_dmub_srv->dmub, 500000);
+ } while (status != DMUB_STATUS_OK);
+ } else {
+ status = dmub_srv_wait_for_hw_pwr_up(dc_dmub_srv->dmub, 500000);
+ if (status != DMUB_STATUS_OK) {
+ DC_ERROR("Error querying DMUB hw power up status: error=%d\n", status);
+ return false;
+ }
}
} else
return dmub_srv_is_hw_pwr_up(dc_dmub_srv->dmub);
@@ -1187,7 +1198,7 @@ void dc_dmub_srv_exit_low_power_state(const struct dc *dc)
const uint32_t max_num_polls = 10000;
uint32_t allow_state = 0;
uint32_t commit_state = 0;
- uint32_t i;
+ int i;
if (dc->debug.dmcub_emulation)
return;
@@ -1220,6 +1231,9 @@ void dc_dmub_srv_exit_low_power_state(const struct dc *dc)
break;
udelay(1);
+
+ if (dc->debug.disable_timeout)
+ i--;
}
ASSERT(i < max_num_polls);
@@ -1242,6 +1256,9 @@ void dc_dmub_srv_exit_low_power_state(const struct dc *dc)
break;
udelay(1);
+
+ if (dc->debug.disable_timeout)
+ i--;
}
ASSERT(i < max_num_polls);
}
--
2.42.0
More information about the amd-gfx
mailing list