Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Fri, 28 Jul 2023 14:22:17 +0100
From:   Will Deacon <will@kernel.org>
To:     Besar Wicaksono <bwicaksono@nvidia.com>, suzuki.poulose@arm.com
Cc:     robin.murphy@arm.com, ilkka@os.amperecomputing.com,
        catalin.marinas@arm.com, mark.rutland@arm.com,
        linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org,
        linux-tegra@vger.kernel.org, treding@nvidia.com,
        jonathanh@nvidia.com, vsethi@nvidia.com, rwiley@nvidia.com,
        efunsten@nvidia.com
Subject: Re: [PATCH v5] perf: arm_cspmu: Separate Arm and vendor module
Message-ID: <20230728132216.GA21394@willie-the-truck>
References: <20230705104745.52255-1-bwicaksono@nvidia.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20230705104745.52255-1-bwicaksono@nvidia.com>
User-Agent: Mutt/1.10.1 (2018-07-13)
Precedence: bulk

On Wed, Jul 05, 2023 at 05:47:45AM -0500, Besar Wicaksono wrote:
> Arm Coresight PMU driver consists of main standard code and
> vendor backend code. Both are currently built as a single module.
> This patch adds vendor registration API to separate the two to
> keep things modular. The main driver requests each known backend
> module during initialization and defer device binding process.
> The backend module then registers an init callback to the main
> driver and continue the device driver binding process.
> 
> Signed-off-by: Besar Wicaksono <bwicaksono@nvidia.com>
> ---
> 
> Changes from v4:
>  * Fix warning reported by kernel test robot
> v4: https://lore.kernel.org/linux-arm-kernel/20230620041438.32514-1-bwicaksono@nvidia.com/T/#u

One minor comment below, but this mostly looks good to me. I'd like Suzuki's
Ack before I queue it, though.

> +	/* Load implementer module and initialize the callbacks. */
> +	if (match) {
> +		mutex_lock(&arm_cspmu_lock);
> +
> +		if (match->impl_init_ops) {
> +			if (try_module_get(match->module)) {
> +				cspmu->impl.match = match;
> +				ret = match->impl_init_ops(cspmu);
> +				module_put(match->module);

Why is it safe to drop the module reference here? If I'm understanding the
flow correctly, ->impl_init_ops() will populate more function pointers
in the cspmu->impl.ops structure, and we don't appear to take a module
reference when calling those.

What happens if the backend module is unloaded while the core module
is executed those functions?

Cheers,

Will