[PATCH 0/2] AMD PMF Smart PC error handling cleanups

[PATCH 0/2] AMD PMF Smart PC error handling cleanups

[Mario Limonciello]

While debugging the suspend issue for amd-pmf the initial bisect result
pointed at red herrings of cleanup flow problems for
amd_pmf_init_smart_pc().  The actual issue wasn't in this code, but still
a lot of memory is allocated and not immediately released if any of the
error branches are taken.

This series cleans that up so that every step is cleaned up. I believe
this actually fixes driver bugs that "could" occur if a BIOS advertisd
Smart PC as well as ITS auto or CNQF but didn't include a policy in the
BIOS.

Mario Limonciello (2):
  platform/x86/amd/pmf: Add debugging message for missing policy data
  platform/x86/amd/pmf: Fixup error handling for amd_pmf_init_smart_pc()

 drivers/platform/x86/amd/pmf/tee-if.c | 71 ++++++++++++++++++---------
 1 file changed, 47 insertions(+), 24 deletions(-)

-- 
2.34.1

[PATCH 1/2] platform/x86/amd/pmf: Add debugging message for missing policy data

[Mario Limonciello]

If a machine advertises Smart PC support but is missing policy data
show a debugging message to help clarify why Smart PC wasn't enabled.

Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
---
 drivers/platform/x86/amd/pmf/tee-if.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/platform/x86/amd/pmf/tee-if.c b/drivers/platform/x86/amd/pmf/tee-if.c
index 8b7e3f87702e..1359ab340f7c 100644
--- a/drivers/platform/x86/amd/pmf/tee-if.c
+++ b/drivers/platform/x86/amd/pmf/tee-if.c
@@ -252,8 +252,10 @@ static int amd_pmf_start_policy_engine(struct amd_pmf_dev *dev)
 	cookie = readl(dev->policy_buf + POLICY_COOKIE_OFFSET);
 	length = readl(dev->policy_buf + POLICY_COOKIE_LEN);
 
-	if (cookie != POLICY_SIGN_COOKIE || !length)
+	if (cookie != POLICY_SIGN_COOKIE || !length) {
+		dev_dbg(dev->dev, "cookie doesn't match\n");
 		return -EINVAL;
+	}
 
 	/* Update the actual length */
 	dev->policy_sz = length + 512;
-- 
2.34.1

Re: [PATCH 1/2] platform/x86/amd/pmf: Add debugging message for missing policy data

[Hans de Goede]

Hi,

On 2/17/24 02:22, Mario Limonciello wrote:
> If a machine advertises Smart PC support but is missing policy data
> show a debugging message to help clarify why Smart PC wasn't enabled.
> 
> Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>

Thanks, patch looks good to me:

Reviewed-by: Hans de Goede <hdegoede@redhat.com>

Regards,

Hans



> ---
>  drivers/platform/x86/amd/pmf/tee-if.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/platform/x86/amd/pmf/tee-if.c b/drivers/platform/x86/amd/pmf/tee-if.c
> index 8b7e3f87702e..1359ab340f7c 100644
> --- a/drivers/platform/x86/amd/pmf/tee-if.c
> +++ b/drivers/platform/x86/amd/pmf/tee-if.c
> @@ -252,8 +252,10 @@ static int amd_pmf_start_policy_engine(struct amd_pmf_dev *dev)
>  	cookie = readl(dev->policy_buf + POLICY_COOKIE_OFFSET);
>  	length = readl(dev->policy_buf + POLICY_COOKIE_LEN);
>  
> -	if (cookie != POLICY_SIGN_COOKIE || !length)
> +	if (cookie != POLICY_SIGN_COOKIE || !length) {
> +		dev_dbg(dev->dev, "cookie doesn't match\n");
>  		return -EINVAL;
> +	}
>  
>  	/* Update the actual length */
>  	dev->policy_sz = length + 512;

[PATCH 2/2] platform/x86/amd/pmf: Fixup error handling for amd_pmf_init_smart_pc()

[Mario Limonciello]

amd_pmf_init_smart_pc() calls out to amd_pmf_get_bios_buffer() but
the error handling flow doesn't clean everything up all allocated
memory.

As amd_pmf_get_bios_buffer() is only called by amd_pmf_init_smart_pc(),
fold it into the function and add labels to clean up any step that
can fail along the way.

Fixes: 7c45534afa44 ("platform/x86/amd/pmf: Add support for PMF Policy Binary")
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
---
 drivers/platform/x86/amd/pmf/tee-if.c | 67 ++++++++++++++++++---------
 1 file changed, 44 insertions(+), 23 deletions(-)

diff --git a/drivers/platform/x86/amd/pmf/tee-if.c b/drivers/platform/x86/amd/pmf/tee-if.c
index 1359ab340f7c..feb9dfafea30 100644
--- a/drivers/platform/x86/amd/pmf/tee-if.c
+++ b/drivers/platform/x86/amd/pmf/tee-if.c
@@ -338,25 +338,6 @@ static void amd_pmf_remove_pb(struct amd_pmf_dev *dev) {}
 static void amd_pmf_hex_dump_pb(struct amd_pmf_dev *dev) {}
 #endif
 
-static int amd_pmf_get_bios_buffer(struct amd_pmf_dev *dev)
-{
-	dev->policy_buf = kzalloc(dev->policy_sz, GFP_KERNEL);
-	if (!dev->policy_buf)
-		return -ENOMEM;
-
-	dev->policy_base = devm_ioremap(dev->dev, dev->policy_addr, dev->policy_sz);
-	if (!dev->policy_base)
-		return -ENOMEM;
-
-	memcpy(dev->policy_buf, dev->policy_base, dev->policy_sz);
-
-	amd_pmf_hex_dump_pb(dev);
-	if (pb_side_load)
-		amd_pmf_open_pb(dev, dev->dbgfs_dir);
-
-	return amd_pmf_start_policy_engine(dev);
-}
-
 static int amd_pmf_amdtee_ta_match(struct tee_ioctl_version_data *ver, const void *data)
 {
 	return ver->impl_id == TEE_IMPL_ID_AMDTEE;
@@ -454,14 +435,54 @@ int amd_pmf_init_smart_pc(struct amd_pmf_dev *dev)
 	if (ret)
 		return ret;
 
+	ret = amd_pmf_set_dram_addr(dev, true);
+	if (ret)
+		goto out_dram;
+
+	dev->policy_base = devm_ioremap(dev->dev, dev->policy_addr, dev->policy_sz);
+	if (!dev->policy_base) {
+		ret = -ENOMEM;
+		goto out_policy_base;
+	}
+
+	dev->policy_buf = kzalloc(dev->policy_sz, GFP_KERNEL);
+	if (!dev->policy_buf) {
+		ret = -ENOMEM;
+		goto out_policy_buf;
+	}
+
+	memcpy(dev->policy_buf, dev->policy_base, dev->policy_sz);
+
+	amd_pmf_hex_dump_pb(dev);
+	if (pb_side_load)
+		amd_pmf_open_pb(dev, dev->dbgfs_dir);
+
 	INIT_DELAYED_WORK(&dev->pb_work, amd_pmf_invoke_cmd);
-	amd_pmf_set_dram_addr(dev, true);
-	amd_pmf_get_bios_buffer(dev);
+
+	ret = amd_pmf_start_policy_engine(dev);
+	if (ret)
+		goto out_start_engine;
+
 	dev->prev_data = kzalloc(sizeof(*dev->prev_data), GFP_KERNEL);
 	if (!dev->prev_data)
-		return -ENOMEM;
+		goto out_prev_data;
+
+	return 0;
 
-	return dev->smart_pc_enabled;
+out_prev_data:
+	cancel_delayed_work_sync(&dev->pb_work);
+
+out_start_engine:
+	kfree(dev->policy_buf);
+
+out_policy_buf:
+out_policy_base:
+	kfree(dev->buf);
+
+out_dram:
+	amd_pmf_tee_deinit(dev);
+
+	return ret;
 }
 
 void amd_pmf_deinit_smart_pc(struct amd_pmf_dev *dev)
-- 
2.34.1

Re: [PATCH 2/2] platform/x86/amd/pmf: Fixup error handling for amd_pmf_init_smart_pc()

[Hans de Goede]

Hi Mario,

Thank you for your patches.

On 2/17/24 02:22, Mario Limonciello wrote:
> amd_pmf_init_smart_pc() calls out to amd_pmf_get_bios_buffer() but
> the error handling flow doesn't clean everything up all allocated
> memory.
> 
> As amd_pmf_get_bios_buffer() is only called by amd_pmf_init_smart_pc(),
> fold it into the function and add labels to clean up any step that
> can fail along the way.
> 
> Fixes: 7c45534afa44 ("platform/x86/amd/pmf: Add support for PMF Policy Binary")
> Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
> ---
>  drivers/platform/x86/amd/pmf/tee-if.c | 67 ++++++++++++++++++---------
>  1 file changed, 44 insertions(+), 23 deletions(-)
> 
> diff --git a/drivers/platform/x86/amd/pmf/tee-if.c b/drivers/platform/x86/amd/pmf/tee-if.c
> index 1359ab340f7c..feb9dfafea30 100644
> --- a/drivers/platform/x86/amd/pmf/tee-if.c
> +++ b/drivers/platform/x86/amd/pmf/tee-if.c
> @@ -338,25 +338,6 @@ static void amd_pmf_remove_pb(struct amd_pmf_dev *dev) {}
>  static void amd_pmf_hex_dump_pb(struct amd_pmf_dev *dev) {}
>  #endif
>  
> -static int amd_pmf_get_bios_buffer(struct amd_pmf_dev *dev)
> -{
> -	dev->policy_buf = kzalloc(dev->policy_sz, GFP_KERNEL);
> -	if (!dev->policy_buf)
> -		return -ENOMEM;
> -
> -	dev->policy_base = devm_ioremap(dev->dev, dev->policy_addr, dev->policy_sz);
> -	if (!dev->policy_base)
> -		return -ENOMEM;
> -
> -	memcpy(dev->policy_buf, dev->policy_base, dev->policy_sz);
> -
> -	amd_pmf_hex_dump_pb(dev);
> -	if (pb_side_load)
> -		amd_pmf_open_pb(dev, dev->dbgfs_dir);
> -
> -	return amd_pmf_start_policy_engine(dev);
> -}
> -
>  static int amd_pmf_amdtee_ta_match(struct tee_ioctl_version_data *ver, const void *data)
>  {
>  	return ver->impl_id == TEE_IMPL_ID_AMDTEE;
> @@ -454,14 +435,54 @@ int amd_pmf_init_smart_pc(struct amd_pmf_dev *dev)
>  	if (ret)
>  		return ret;
>  
> +	ret = amd_pmf_set_dram_addr(dev, true);
> +	if (ret)
> +		goto out_dram;
> +
> +	dev->policy_base = devm_ioremap(dev->dev, dev->policy_addr, dev->policy_sz);
> +	if (!dev->policy_base) {
> +		ret = -ENOMEM;
> +		goto out_policy_base;
> +	}
> +
> +	dev->policy_buf = kzalloc(dev->policy_sz, GFP_KERNEL);
> +	if (!dev->policy_buf) {
> +		ret = -ENOMEM;
> +		goto out_policy_buf;
> +	}
> +
> +	memcpy(dev->policy_buf, dev->policy_base, dev->policy_sz);
> +
> +	amd_pmf_hex_dump_pb(dev);
> +	if (pb_side_load)
> +		amd_pmf_open_pb(dev, dev->dbgfs_dir);
> +
>  	INIT_DELAYED_WORK(&dev->pb_work, amd_pmf_invoke_cmd);
> -	amd_pmf_set_dram_addr(dev, true);
> -	amd_pmf_get_bios_buffer(dev);
> +
> +	ret = amd_pmf_start_policy_engine(dev);
> +	if (ret)
> +		goto out_start_engine;
> +
>  	dev->prev_data = kzalloc(sizeof(*dev->prev_data), GFP_KERNEL);
>  	if (!dev->prev_data)

I just checked and dev->prev_data gets used by dev->pb_work
which gets queued by amd_pmf_start_policy_engine() so there
is a (pre-existing) race here and dev->prev_data should be allocated
before amd_pmf_start_policy_engine().

Note kfree(NULL) is a no-op as is cancel_delayed_work_sync()
on a non-queued work.

So I think you can just use a single error_exit: label 
(or any other label-name you like) and do:

error_exit:
	cancel_delayed_work_sync(&dev->pb_work);
	kfree(dev->prev_data);
	dev->prev_data = NULL;
	kfree(dev->policy_buf);
	dev->policy_buf = NULL;
	kfree(dev->buf);
	dev->buf = NULL;
	amd_pmf_tee_deinit(dev);
	return ret;

There as long as you do the INIT_DELAYED_WORK() before
any of the code which may fail with a goto error_exit.

Note I also added clearing of the pointers after freeing
them, at least for dev->buf this is important because
that also gets used in non smart-pc paths and those
count on it either being NULL or a valid pointer.

Regards,

Hans















> -		return -ENOMEM;
> +		goto out_prev_data;
> +
> +	return 0;
>  
> -	return dev->smart_pc_enabled;
> +out_prev_data:
> +	cancel_delayed_work_sync(&dev->pb_work);
> +
> +out_start_engine:
> +	kfree(dev->policy_buf);
> +
> +out_policy_buf:
> +out_policy_base:
> +	kfree(dev->buf);
> +
> +out_dram:
> +	amd_pmf_tee_deinit(dev);
> +
> +	return ret;
>  }
>  
>  void amd_pmf_deinit_smart_pc(struct amd_pmf_dev *dev)