gpu: nvgpu: nvlink: Add HAL to get link_mask

VBIOS link_disable_mask should be sufficient to find the connected
links. As VBIOS is not updated with correct mask, we parse the DT
node where we hardcode the link_id. DT method is not scalable as same
DT node is used for different dGPUs connected over PCIE. Remove the
DT parsing of link id and use HAL to get link_mask based on the GPU.

JIRA NVLINK-162

Change-Id: Idb7b639962928ce48711a0d7fc277c4c324bee91
Signed-off-by: Tejal Kudav <tkudav@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1738967
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>

3 files changed, 24 insertions, 3 deletions

diff --git a/drivers/gpu/nvgpu/gv100/hal_gv100.c b/drivers/gpu/nvgpu/gv100/hal_gv100.c
index 5e42ba9f..92900421 100644
--- a/drivers/gpu/nvgpu/gv100/hal_gv100.c
+++ b/drivers/gpu/nvgpu/gv100/hal_gv100.c
@@ -843,6 +843,7 @@ static const struct gpu_ops gv100_ops = {
                 .rxdet = NULL,
                 .setup_pll = gv100_nvlink_setup_pll,
                 .minion_data_ready_en = gv100_nvlink_minion_data_ready_en,
+                .get_connected_link_mask = gv100_nvlink_get_connected_link_mask,
                 /* API */
                 .link_early_init = gv100_nvlink_link_early_init,
                 .link_get_state = gv100_nvlink_link_get_state,
diff --git a/drivers/gpu/nvgpu/gv100/nvlink_gv100.c b/drivers/gpu/nvgpu/gv100/nvlink_gv100.c
index cae4f9bd..e85b5a93 100644
--- a/drivers/gpu/nvgpu/gv100/nvlink_gv100.c
+++ b/drivers/gpu/nvgpu/gv100/nvlink_gv100.c
@@ -2696,6 +2696,13 @@ u32 gv100_nvlink_link_get_rx_sublink_state(struct gk20a *g, u32 link_id)
         return nvl_sl1_slsm_status_rx_primary_state_v(reg);
 }
 
+/* Hardcode the link_mask while we wait for VBIOS link_disable_mask field
+ * to be updated.
+ */
+void gv100_nvlink_get_connected_link_mask(u32 *link_mask)
+{
+        *link_mask = GV100_CONNECTED_LINK_MASK;
+}
 /*
  * Performs nvlink device level initialization by discovering the topology
  * taking device out of reset, boot minion, set clocks up and common interrupts
@@ -2735,10 +2742,20 @@ int gv100_nvlink_early_init(struct gk20a *g)
         /* Links in reset should be removed from initialized link sw state */
         g->nvlink.initialized_links &= __gv100_nvlink_get_link_reset_mask(g);
 
-        nvgpu_log(g, gpu_dbg_nvlink, "connected_links = 0x%08x (from DT)",
-                g->nvlink.connected_links);
+        /* VBIOS link_disable_mask should be sufficient to find the connected
+         * links. As VBIOS is not updated with correct mask, we parse the DT
+         * node where we hardcode the link_id. DT method is not scalable as same
+         * DT node is used for different dGPUs connected over PCIE.
+         * Remove the DT parsing of link id and use HAL to get link_mask based
+         * on the GPU. This is temporary WAR while we get the VBIOS updated with
+         * correct mask.
+         */
+        g->ops.nvlink.get_connected_link_mask(&(g->nvlink.connected_links));
+
+        nvgpu_log(g, gpu_dbg_nvlink, "connected_links = 0x%08x",
+                                                g->nvlink.connected_links);
 
-        /* Track unconnected links */
+        /* Track only connected links */
         g->nvlink.discovered_links &= g->nvlink.connected_links;
 
         nvgpu_log(g, gpu_dbg_nvlink, "discovered_links = 0x%08x (combination)",
diff --git a/drivers/gpu/nvgpu/gv100/nvlink_gv100.h b/drivers/gpu/nvgpu/gv100/nvlink_gv100.h
index bf923d4b..d9a4b073 100644
--- a/drivers/gpu/nvgpu/gv100/nvlink_gv100.h
+++ b/drivers/gpu/nvgpu/gv100/nvlink_gv100.h
@@ -23,6 +23,8 @@
 #ifndef NVGPU_NVLINK_GV100_H
 #define NVGPU_NVLINK_GV100_H
 
+#define GV100_CONNECTED_LINK_MASK               0x8
+
 struct gk20a;
 
 int gv100_nvlink_discover_ioctrl(struct gk20a *g);
@@ -34,6 +36,7 @@ int gv100_nvlink_minion_send_command(struct gk20a *g, u32 link_id, u32 command,
 int gv100_nvlink_setup_pll(struct gk20a *g, unsigned long link_mask);
 int gv100_nvlink_minion_data_ready_en(struct gk20a *g,
                                         unsigned long link_mask, bool sync);
+void gv100_nvlink_get_connected_link_mask(u32 *link_mask);
 /* API */
 int gv100_nvlink_link_early_init(struct gk20a *g, unsigned long mask);
 u32 gv100_nvlink_link_get_mode(struct gk20a *g, u32 link_id);