Re: [Patch v3 Part2 1/9] x86/microcode: Taint kernel only if microcode loading was successful

[Ashok Raj <ashok.raj@intel.com>]

On Tue, Jan 31, 2023 at 12:50:44PM +0100, Borislav Petkov wrote:
> On Mon, Jan 30, 2023 at 01:39:47PM -0800, Ashok Raj wrote:
> >  arch/x86/kernel/cpu/microcode/core.c | 19 +++++++++++++------
> >  1 file changed, 13 insertions(+), 6 deletions(-)
> 
> Why all this hoopla and unrelated changes?
> 
> Why don't you simply hoist the call to ->request_microcode_fw outside of
> the locked region as it doesn't have to be there and then do the usual
> pattern?

Makes total sense, and seems to make the code more readable. Thanks!

Just some minor changes below.

remove ret = 0 during initialization since its cleared right below. (tglx)

Some more below, updated patch at the end.

I have tested with the modified patch below.

> 
> ---
> diff --git a/arch/x86/kernel/cpu/microcode/core.c b/arch/x86/kernel/cpu/microcode/core.c
> index 14a2280fdcd2..23f4f22df581 100644
> --- a/arch/x86/kernel/cpu/microcode/core.c
> +++ b/arch/x86/kernel/cpu/microcode/core.c
> @@ -481,28 +481,28 @@ static ssize_t reload_store(struct device *dev,
>  	if (val != 1)
>  		return size;
>  
> +	tmp_ret = microcode_ops->request_microcode_fw(bsp, &microcode_pdev->dev);
> +	if (tmp_ret != UCODE_NEW)
> +		return ret;
> +
>  	cpus_read_lock();
>  
>  	ret = check_online_cpus();
>  	if (ret)
> -		goto put;
> -
> -	tmp_ret = microcode_ops->request_microcode_fw(bsp, &microcode_pdev->dev);
> -	if (tmp_ret != UCODE_NEW)
> -		goto put;
> +		goto unlock;

Need to set ret explicitly to either -EINVAL, or size. Otherwise it will be
endlessly waiting for write to complete. (As Aubrey pointed out)

>  
>  	mutex_lock(&microcode_mutex);
>  	ret = microcode_reload_late();

I think its safe to leave ret as is, since microcode_reload_late() only
returns -1, or 0.

>  	mutex_unlock(&microcode_mutex);
>  
> -put:
> -	cpus_read_unlock();
> -
>  	if (ret == 0)
>  		ret = size;
>  
>  	add_taint(TAINT_CPU_OUT_OF_SPEC, LOCKDEP_STILL_OK);

Pull this into the ret == 0, so taint only if the update was successful? 
And add a message so its not silent?

>  
> +unlock:
> +	cpus_read_unlock();
> +
>  	return ret;
>  }
>  
> 

diff --git a/arch/x86/kernel/cpu/microcode/core.c b/arch/x86/kernel/cpu/microcode/core.c
index 94d942c1bf2c..550b7c566311 100644
--- a/arch/x86/kernel/cpu/microcode/core.c
+++ b/arch/x86/kernel/cpu/microcode/core.c
@@ -472,7 +472,7 @@ static ssize_t reload_store(struct device *dev,
 	enum ucode_state tmp_ret = UCODE_OK;
 	int bsp = boot_cpu_data.cpu_index;
 	unsigned long val;
-	ssize_t ret = 0;
+	ssize_t ret;
 
 	ret = kstrtoul(buf, 0, &val);
 	if (ret)
@@ -483,7 +483,7 @@ static ssize_t reload_store(struct device *dev,
 
 	tmp_ret = microcode_ops->request_microcode_fw(bsp, &microcode_pdev->dev);
 	if (tmp_ret != UCODE_NEW)
-		return ret;
+		return (tmp_ret == UCODE_ERROR ? -EINVAL : size);
 
 	cpus_read_lock();
 
@@ -495,10 +495,11 @@ static ssize_t reload_store(struct device *dev,
 	ret = microcode_reload_late();
 	mutex_unlock(&microcode_mutex);
 
-	if (ret == 0)
+	if (ret == 0) {
 		ret = size;
-
-	add_taint(TAINT_CPU_OUT_OF_SPEC, LOCKDEP_STILL_OK);
+		add_taint(TAINT_CPU_OUT_OF_SPEC, LOCKDEP_STILL_OK);
+		pr_warn("Microcode late loading tainted the kernel\n");
+	}
 
 unlock:
 	cpus_read_unlock();