Re: [PATCH] block: break circular locks in blk_request_module

[Desmond Cheong Zhi Xi]

On 17/6/21 7:51 pm, Christoph Hellwig wrote:
> On Thu, Jun 17, 2021 at 05:20:16PM +0800, Desmond Cheong Zhi Xi wrote:
> >   	mutex_lock(&major_names_lock);
> >   	for (n = &major_names[major_to_index(major)]; *n; n = &(*n)->next) {
> >   		if ((*n)->major == major && (*n)->probe) {
> > -			(*n)->probe(devt);
> > +			probe = (*n)->probe;
> >   			mutex_unlock(&major_names_lock);
> > +			probe(devt);
>
> And now you can all probe after it has been freed and/or the module has
> been unloaded. The obviously correct fix is to only hold mtd_table_mutex
> for the actually required critical section:

Thank you for the correction, Christoph. I hadn't thought of the 
scenario where the module is unloaded. I'll be more conscientious in the 
future.

> diff --git a/drivers/mtd/mtd_blkdevs.c b/drivers/mtd/mtd_blkdevs.c
> index fb8e12d590a1..065d94f9b1fb 100644
> --- a/drivers/mtd/mtd_blkdevs.c
> +++ b/drivers/mtd/mtd_blkdevs.c
> @@ -529,13 +529,11 @@ int register_mtd_blktrans(struct mtd_blktrans_ops *tr)
>   		register_mtd_user(&blktrans_notifier);
>   
>   
> -	mutex_lock(&mtd_table_mutex);
>   
>   	ret = register_blkdev(tr->major, tr->name);
>   	if (ret < 0) {
>   		printk(KERN_WARNING "Unable to register %s block device on major %d: %d\n",
>   		       tr->name, tr->major, ret);
> -		mutex_unlock(&mtd_table_mutex);
>   		return ret;
>   	}
>   
> @@ -545,12 +543,12 @@ int register_mtd_blktrans(struct mtd_blktrans_ops *tr)
>   	tr->blkshift = ffs(tr->blksize) - 1;
>   
>   	INIT_LIST_HEAD(&tr->devs);
> -	list_add(&tr->list, &blktrans_majors);
>   
> +	mutex_lock(&mtd_table_mutex);
> +	list_add(&tr->list, &blktrans_majors);
>   	mtd_for_each_device(mtd)
>   		if (mtd->type != MTD_ABSENT)
>   			tr->add_mtd(tr, mtd);
> -
>   	mutex_unlock(&mtd_table_mutex);
>   	return 0;
>   }

This fix passes the Syzkaller repro test on my local machine and on 
Syzbot. I can prepare a v2 patch for this. May I include you with the 
Co-developed-by: and Signed-off-by: tags? If another tag would be more 
appropriate, or if you want to submit the patch yourself, please let me 
know.

Best wishes,
Desmond