[Why] While disabling mst topology manager in drm_dp_mst_topology_mgr_set_mst(), now just reset the mgr->payloads but doesn't handle the mgr->proposed_vcpis.
[How] Remove mgr->proposed_vcpis to NULL.
Signed-off-by: Wayne Lin Wayne.Lin@amd.com --- drivers/gpu/drm/drm_dp_mst_topology.c | 12 ++++++++++++ 1 file changed, 12 insertions(+)
diff --git a/drivers/gpu/drm/drm_dp_mst_topology.c b/drivers/gpu/drm/drm_dp_mst_topology.c index ae5809a1f19a..81e92b260d7a 100644 --- a/drivers/gpu/drm/drm_dp_mst_topology.c +++ b/drivers/gpu/drm/drm_dp_mst_topology.c @@ -3386,6 +3386,7 @@ static int drm_dp_get_vc_payload_bw(u8 dp_link_bw, u8 dp_link_count) int drm_dp_mst_topology_mgr_set_mst(struct drm_dp_mst_topology_mgr *mgr, bool mst_state) { int ret = 0; + int i = 0; struct drm_dp_mst_branch *mstb = NULL;
mutex_lock(&mgr->lock); @@ -3446,10 +3447,21 @@ int drm_dp_mst_topology_mgr_set_mst(struct drm_dp_mst_topology_mgr *mgr, bool ms /* this can fail if the device is gone */ drm_dp_dpcd_writeb(mgr->aux, DP_MSTM_CTRL, 0); ret = 0; + mutex_lock(&mgr->payload_lock); memset(mgr->payloads, 0, mgr->max_payloads * sizeof(struct drm_dp_payload)); mgr->payload_mask = 0; set_bit(0, &mgr->payload_mask); + for (i = 0; i < mgr->max_payloads; i++) { + struct drm_dp_vcpi *tmp_vcpi = mgr->proposed_vcpis[i]; + + if (tmp_vcpi) { + tmp_vcpi->vcpi = 0; + tmp_vcpi->num_slots = 0; + } + mgr->proposed_vcpis[i] = NULL; + } mgr->vcpi_mask = 0; + mutex_unlock(&mgr->payload_lock); }
out_unlock:
I'm, not entirely sure what this patch is trying to accomplish. I'm guessing maybe we're leaving stale VCPI allocations from the previous topology enablement and then somehow trying to use those again when allocating payloads? The patch looks correct at least.
If this patch is fixing an issue, such as displays not turning on with amdgpu, I'd definitely mention it in more detail here and Cc to stable if applicable. Also, one nitpick below:
On Mon, 2019-12-02 at 11:57 +0800, Wayne Lin wrote:
[Why] While disabling mst topology manager in drm_dp_mst_topology_mgr_set_mst(), now just reset the mgr->payloads but doesn't handle the mgr->proposed_vcpis.
[How] Remove mgr->proposed_vcpis to NULL.
Signed-off-by: Wayne Lin Wayne.Lin@amd.com
drivers/gpu/drm/drm_dp_mst_topology.c | 12 ++++++++++++ 1 file changed, 12 insertions(+)
diff --git a/drivers/gpu/drm/drm_dp_mst_topology.c b/drivers/gpu/drm/drm_dp_mst_topology.c index ae5809a1f19a..81e92b260d7a 100644 --- a/drivers/gpu/drm/drm_dp_mst_topology.c +++ b/drivers/gpu/drm/drm_dp_mst_topology.c @@ -3386,6 +3386,7 @@ static int drm_dp_get_vc_payload_bw(u8 dp_link_bw, u8 dp_link_count) int drm_dp_mst_topology_mgr_set_mst(struct drm_dp_mst_topology_mgr *mgr, bool mst_state) { int ret = 0;
int i = 0; struct drm_dp_mst_branch *mstb = NULL;
mutex_lock(&mgr->lock);
@@ -3446,10 +3447,21 @@ int drm_dp_mst_topology_mgr_set_mst(struct drm_dp_mst_topology_mgr *mgr, bool ms /* this can fail if the device is gone */ drm_dp_dpcd_writeb(mgr->aux, DP_MSTM_CTRL, 0); ret = 0;
mutex_lock(&mgr->payload_lock);drm_dp_payload)); mgr->payload_mask = 0; set_bit(0, &mgr->payload_mask);
for (i = 0; i < mgr->max_payloads; i++) {struct drm_dp_vcpi *tmp_vcpi = mgr->proposed_vcpis[i];if (tmp_vcpi) {tmp_vcpi->vcpi = 0;tmp_vcpi->num_slots = 0;}mgr->proposed_vcpis[i] = NULL;}mutex_unlock(&mgr->payload_lock);
bikeshed: I'd just rename tmp_vcpi here to vcpi, but I'll leave that up to you
}
out_unlock:
-----Original Message----- From: Lyude Paul lyude@redhat.com Sent: Tuesday, December 3, 2019 8:03 AM To: Lin, Wayne Wayne.Lin@amd.com; dri-devel@lists.freedesktop.org; amd-gfx@lists.freedesktop.org Cc: Kazlauskas, Nicholas Nicholas.Kazlauskas@amd.com; Wentland, Harry Harry.Wentland@amd.com; Zuo, Jerry Jerry.Zuo@amd.com Subject: Re: [PATCH] drm/dp_mst: Remove VCPI while disabling topology mgr
I'm, not entirely sure what this patch is trying to accomplish. I'm guessing maybe we're leaving stale VCPI allocations from the previous topology enablement and then somehow trying to use those again when allocating payloads? The patch looks correct at least.
Thanks for your time and the comment.
Yes, this patch is trying to address the problem you mentioned. Once unplug a DP MST capable device, driver will call drm_dp_mst_topology_mgr_set_mst() to reset mgr->payloads but it doesn't reset the mgr->proposed_vcpis. If it doesn't reset the proposed_vcpi, code will fail at checking port validation once plug in MST device later. Once MST capable device plug in again and try to allocate payloads by calling drm_dp_update_payload_part1(), this function will iterate over all proposed virtual channels and check each port validation to see if the specified port is still in the topology. Since there are stale VCPI allocations from the previous topology enablement in proposed_vcpi[], code flow will fail and reurn EINVAL.
If this patch is fixing an issue, such as displays not turning on with amdgpu, I'd definitely mention it in more detail here and Cc to stable if applicable.
Thanks for your comment. I will Cc to stable@vger.kernel.org and amend the message in more detail in next version.
Also, one nitpick below:
Thanks, I'll modify it.
On Mon, 2019-12-02 at 11:57 +0800, Wayne Lin wrote:
[Why] While disabling mst topology manager in drm_dp_mst_topology_mgr_set_mst(), now just reset the mgr->payloads but doesn't handle the mgr->proposed_vcpis.
[How] Remove mgr->proposed_vcpis to NULL.
Signed-off-by: Wayne Lin Wayne.Lin@amd.com
drivers/gpu/drm/drm_dp_mst_topology.c | 12 ++++++++++++ 1 file changed, 12 insertions(+)
diff --git a/drivers/gpu/drm/drm_dp_mst_topology.c b/drivers/gpu/drm/drm_dp_mst_topology.c index ae5809a1f19a..81e92b260d7a 100644 --- a/drivers/gpu/drm/drm_dp_mst_topology.c +++ b/drivers/gpu/drm/drm_dp_mst_topology.c @@ -3386,6 +3386,7 @@ static int drm_dp_get_vc_payload_bw(u8 dp_link_bw, u8 dp_link_count) int drm_dp_mst_topology_mgr_set_mst(struct
drm_dp_mst_topology_mgr
*mgr, bool mst_state) { int ret = 0;
int i = 0; struct drm_dp_mst_branch *mstb = NULL;
mutex_lock(&mgr->lock);
@@ -3446,10 +3447,21 @@ int drm_dp_mst_topology_mgr_set_mst(struct drm_dp_mst_topology_mgr *mgr, bool ms /* this can fail if the device is gone */ drm_dp_dpcd_writeb(mgr->aux, DP_MSTM_CTRL, 0); ret = 0;
mutex_lock(&mgr->payload_lock);drm_dp_payload)); mgr->payload_mask = 0; set_bit(0, &mgr->payload_mask);
for (i = 0; i < mgr->max_payloads; i++) {struct drm_dp_vcpi *tmp_vcpi = mgr->proposed_vcpis[i];if (tmp_vcpi) {tmp_vcpi->vcpi = 0;tmp_vcpi->num_slots = 0;}mgr->proposed_vcpis[i] = NULL;}mutex_unlock(&mgr->payload_lock);bikeshed: I'd just rename tmp_vcpi here to vcpi, but I'll leave that up to you
}
out_unlock:
-- Cheers, Lyude Paul
-- BR, Wayne Lin
dri-devel@lists.freedesktop.org
-
Lin, Wayne -
Lyude Paul -
Wayne Lin