Re: [PATCH 1/5] powerpc/spufs: simplify spufs core dumping

[Al Viro]

On Mon, Apr 27, 2020 at 10:06:21PM +0200, Christoph Hellwig wrote:

> @@ -1988,7 +1984,12 @@ static ssize_t spufs_mbox_info_read(struct file *file, char __user *buf,
>  	if (ret)
>  		return ret;
>  	spin_lock(&ctx->csa.register_lock);
> -	ret = __spufs_mbox_info_read(ctx, buf, len, pos);
> +	/* EOF if there's no entry in the mbox */
> +	if (ctx->csa.prob.mb_stat_R & 0x0000ff) {
> +		ret = simple_read_from_buffer(buf, len, pos,
> +				&ctx->csa.prob.pu_mb_R,
> +				sizeof(ctx->csa.prob.pu_mb_R));
> +	}
>  	spin_unlock(&ctx->csa.register_lock);
>  	spu_release_saved(ctx);

Again, this really needs fixing.  Preferably - as a separate commit preceding
this series, so that it could be backported.  simple_read_from_buffer() is
a blocking operation.  Yes, I understand that mainline has the same bug;
it really does need to be fixed and having to backport this series is not
a good idea, for obvious reasons.