We don't try to allocate huge pages here so remove the huge word.
Signed-off-by: Oded Gabbay <[email protected]>
---
drivers/misc/habanalabs/common/memory.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/misc/habanalabs/common/memory.c b/drivers/misc/habanalabs/common/memory.c
index 3324332811bc..84227819e4d1 100644
--- a/drivers/misc/habanalabs/common/memory.c
+++ b/drivers/misc/habanalabs/common/memory.c
@@ -77,8 +77,8 @@ static int alloc_device_memory(struct hl_ctx *ctx, struct hl_mem_in *args,
paddr = (u64) gen_pool_alloc(vm->dram_pg_pool, total_size);
if (!paddr) {
dev_err(hdev->dev,
- "failed to allocate %llu huge contiguous pages\n",
- num_pgs);
+ "failed to allocate %llu contiguous pages with total size of %llu\n",
+ num_pgs, total_size);
return -ENOMEM;
}
}
--
2.17.1
Some engines use resources that belong to the kernel context (e.g. MMU
mappings). In case the halt-engines doesn't work properly due to H/W
restriction, we need to make sure the kernel context lives on until after
the hw_fini. The hw_fini resets the ASIC after that no engine is alive and
we can safely close the kernel context.
Signed-off-by: Oded Gabbay <[email protected]>
---
drivers/misc/habanalabs/common/device.c | 13 ++++++-------
1 file changed, 6 insertions(+), 7 deletions(-)
diff --git a/drivers/misc/habanalabs/common/device.c b/drivers/misc/habanalabs/common/device.c
index 196e35d71118..20572224099a 100644
--- a/drivers/misc/habanalabs/common/device.c
+++ b/drivers/misc/habanalabs/common/device.c
@@ -967,14 +967,13 @@ int hl_device_reset(struct hl_device *hdev, bool hard_reset,
flush_workqueue(hdev->eq_wq);
}
- /* Release kernel context */
- if ((hard_reset) && (hl_ctx_put(hdev->kernel_ctx) == 1))
- hdev->kernel_ctx = NULL;
-
/* Reset the H/W. It will be in idle state after this returns */
hdev->asic_funcs->hw_fini(hdev, hard_reset);
if (hard_reset) {
+ /* Release kernel context */
+ if (hl_ctx_put(hdev->kernel_ctx) == 1)
+ hdev->kernel_ctx = NULL;
hl_vm_fini(hdev);
hl_mmu_fini(hdev);
hl_eq_reset(hdev, &hdev->event_queue);
@@ -1465,13 +1464,13 @@ void hl_device_fini(struct hl_device *hdev)
hl_cb_pool_fini(hdev);
+ /* Reset the H/W. It will be in idle state after this returns */
+ hdev->asic_funcs->hw_fini(hdev, true);
+
/* Release kernel context */
if ((hdev->kernel_ctx) && (hl_ctx_put(hdev->kernel_ctx) != 1))
dev_err(hdev->dev, "kernel ctx is still alive\n");
- /* Reset the H/W. It will be in idle state after this returns */
- hdev->asic_funcs->hw_fini(hdev, true);
-
hl_vm_fini(hdev);
hl_mmu_fini(hdev);
--
2.17.1
The device should be idle after a context is closed. If not, print a
notice.
Signed-off-by: Oded Gabbay <[email protected]>
---
drivers/misc/habanalabs/common/context.c | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/drivers/misc/habanalabs/common/context.c b/drivers/misc/habanalabs/common/context.c
index bd03ef074eed..7a59dd7c6450 100644
--- a/drivers/misc/habanalabs/common/context.c
+++ b/drivers/misc/habanalabs/common/context.c
@@ -12,6 +12,7 @@
static void hl_ctx_fini(struct hl_ctx *ctx)
{
struct hl_device *hdev = ctx->hdev;
+ u64 idle_mask = 0;
int i;
/*
@@ -42,6 +43,13 @@ static void hl_ctx_fini(struct hl_ctx *ctx)
hl_cb_va_pool_fini(ctx);
hl_vm_ctx_fini(ctx);
hl_asid_free(hdev, ctx->asid);
+
+ if ((!hdev->pldm) && (hdev->pdev) &&
+ (!hdev->asic_funcs->is_device_idle(hdev,
+ &idle_mask, NULL)))
+ dev_notice(hdev->dev,
+ "device not idle after user context is closed (0x%llx)\n",
+ idle_mask);
} else {
dev_dbg(hdev->dev, "closing kernel context\n");
hl_mmu_ctx_fini(ctx);
--
2.17.1
During debugging of error we sometimes need to know whether the error
happened when a user context was open. Add debug prints when opening and
closing user contexts.
Signed-off-by: Oded Gabbay <[email protected]>
---
drivers/misc/habanalabs/common/context.c | 5 +++++
1 file changed, 5 insertions(+)
diff --git a/drivers/misc/habanalabs/common/context.c b/drivers/misc/habanalabs/common/context.c
index df8171a2226c..bd03ef074eed 100644
--- a/drivers/misc/habanalabs/common/context.c
+++ b/drivers/misc/habanalabs/common/context.c
@@ -28,6 +28,8 @@ static void hl_ctx_fini(struct hl_ctx *ctx)
kfree(ctx->cs_pending);
if (ctx->asid != HL_KERNEL_ASID_ID) {
+ dev_dbg(hdev->dev, "closing user context %d\n", ctx->asid);
+
/* The engines are stopped as there is no executing CS, but the
* Coresight might be still working by accessing addresses
* related to the stopped engines. Hence stop it explicitly.
@@ -41,6 +43,7 @@ static void hl_ctx_fini(struct hl_ctx *ctx)
hl_vm_ctx_fini(ctx);
hl_asid_free(hdev, ctx->asid);
} else {
+ dev_dbg(hdev->dev, "closing kernel context\n");
hl_mmu_ctx_fini(ctx);
}
}
@@ -168,6 +171,8 @@ int hl_ctx_init(struct hl_device *hdev, struct hl_ctx *ctx, bool is_kernel_ctx)
dev_err(hdev->dev, "ctx_init failed\n");
goto err_cb_va_pool_fini;
}
+
+ dev_dbg(hdev->dev, "create user context %d\n", ctx->asid);
}
return 0;
--
2.17.1
On Thu, Sep 24, 2020 at 10:03 AM Oded Gabbay <[email protected]> wrote:
> The device should be idle after a context is closed. If not, print a
> notice.
>
> Signed-off-by: Oded Gabbay <[email protected]>
This patch-set is:
Reviewed-by: Tomer Tayar <[email protected]>