Re: [PATCH] x86/microcode: Remove microcode_mutex.

[Sohil Mehta]

Nit: The full stop at the end is not needed.

On 8/3/2023 1:32 AM, Sebastian Andrzej Siewior wrote:
> microcode_mutex is only used by reload_store().  It has a comment saying
> "to synchronize with each other". This probably means the sysfs
> interface vs the legacy interface which was removed in commit
> 181b6f40e9ea8 ("x86/microcode: Rip out the OLD_INTERFACE").
> 

There is also commit b6f86689d5b7 ("x86/microcode: Rip out the subsys
interface gunk") which last year removed another usage of microcode_mutex.

> The sysfs interface does not need additional synchronisation vs itself
> because it is provided as kernfs_ops::mutex which is acquired in
> kernfs_fop_write_iter().
> 
> Remove superfluous microcode_mutex.

I agree, the current usage does look unnecessary.

Maybe reword the commit message to say that after these two Rip outs
there are no of *other* usages of microcode_mutex to synchronize with?

> 
> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@xxxxxxxxxxxxx>
> ---
> This poped up as "defined but not used" on RT builds without
> CONFIG_MICROCODE_LATE_LOADING enabled.

This issue has been raised a couple of times recently but the
justification has been deemed insufficient since it can't be reproduced
with a .config file.

See:
https://lore.kernel.org/lkml/20230324114720.1756466-1-john.ogness@xxxxxxxxxxxxx/

and

https://lore.kernel.org/lkml/20230522062713.427998-1-christian.gmeiner@xxxxxxxxx/

However, your current justification of not needing the mutex itself
seems reasonable to me.

> 
>  arch/x86/kernel/cpu/microcode/core.c | 6 ------
>  1 file changed, 6 deletions(-)
> 
> diff --git a/arch/x86/kernel/cpu/microcode/core.c b/arch/x86/kernel/cpu/microcode/core.c
> index 3afcf3de0dd49..2f9d35744bc41 100644
> --- a/arch/x86/kernel/cpu/microcode/core.c
> +++ b/arch/x86/kernel/cpu/microcode/core.c
> @@ -54,15 +54,12 @@ LIST_HEAD(microcode_cache);
>   *
>   * All non cpu-hotplug-callback call sites use:
>   *
> - * - microcode_mutex to synchronize with each other;
>   * - cpus_read_lock/unlock() to synchronize with
>   *   the cpu-hotplug-callback call sites.
>   *
>   * We guarantee that only a single cpu is being
>   * updated at any particular moment of time.
>   */
> -static DEFINE_MUTEX(microcode_mutex);
> -
>  struct ucode_cpu_info		ucode_cpu_info[NR_CPUS];
>  
>  struct cpu_info_ctx {
> @@ -488,10 +485,7 @@ static ssize_t reload_store(struct device *dev,
>  	if (tmp_ret != UCODE_NEW)
>  		goto put;
>  
> -	mutex_lock(&microcode_mutex);
>  	ret = microcode_reload_late();
> -	mutex_unlock(&microcode_mutex);
> -
>  put:
>  	cpus_read_unlock();
>  

The code changes look fine to me.

You can also add below to the patch.

> diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
> index 52683fddafaf..777340724ec3 100644
> --- a/arch/x86/kernel/cpu/common.c
> +++ b/arch/x86/kernel/cpu/common.c
> @@ -2276,8 +2276,7 @@ void store_cpu_caps(struct cpuinfo_x86 *curr_info)
>   * @prev_info: CPU capabilities stored before an update.
>   *
>   * The microcode loader calls this upon late microcode load to recheck features,
> - * only when microcode has been updated. Caller holds microcode_mutex and CPU
> - * hotplug lock.
> + * only when microcode has been updated. Caller holds CPU hotplug lock.
>   *
>   * Return: None
>   */