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

[Ashok Raj]

On Tue, Jan 31, 2023 at 09:20:37PM +0100, Borislav Petkov wrote:
> On Tue, Jan 31, 2023 at 08:51:25AM -0800, Ashok Raj wrote:
> > remove ret = 0 during initialization since its cleared right below. (tglx)
> 
> Sure.
> 
> > Need to set ret explicitly to either -EINVAL, or size. Otherwise it will be
> > endlessly waiting for write to complete. (As Aubrey pointed out)
> 
> Then do:
> 
>         tmp_ret = microcode_ops->request_microcode_fw(bsp, &microcode_pdev->dev);
>         if (tmp_ret != UCODE_NEW)
>                 return size;
> 
> to signal what it is. It certainly ain't an error if it doesn't find new
> microcode.

It's not an error, only when request_microcode() returns UCODE_ERROR, should
it return -EINVAL, if its UCODE_NFOUND, or otherwise the code should treat
as success. 

The diff I attached was: https://lore.kernel.org/lkml/Y9lHDWjjnqdletL3@xxxxxxxxxxxxxxxxxxxxxxxxx/

 	if (tmp_ret != UCODE_NEW)
-		return ret;
+		return (tmp_ret == UCODE_ERROR ? -EINVAL : size);

Does the above look fine? 

> 
> > I think its safe to leave ret as is, since microcode_reload_late() only
> > returns -1, or 0.
> 
> No it doesn't. Hint: stop_machine_cpuslocked().
> 
> > Pull this into the ret == 0, so taint only if the update was successful? 
> 
> Ok.
> 
> > And add a message so its not silent?
> 
> You'd add a printk for every possible operation, wouldn't you?

:-) Not like that.. But looking through most of the cases that does
add_taint() either have some print, or there some big spalt message around
it.

This shouldn't be noisy, but if you think this isn't needed, it can go
away.

> 
> See, the world doesn't revolve around microcode loading. If that thing
> fails, then someone has done a bad job at the CPU vendor testing,
> provided the code does the right thing.
> 

When it fails due to current_rev < min_rev, Isn't it good to add indication
to user space that it didn't succeed? Thomas wanted these return codes, so
someone scripting can get a status after an attempt to load.

Otherwise agree, it shouldn't generally fail.