Received-SPF: pass (google.com: domain of linux-kernel+bounces-92876-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.199.223 as permitted sender) client-ip=147.75.199.223;
Precedence: bulk
MIME-Version: 1.0
References: <20240228-multi_waitq-v1-0-ccb096419af0@quicinc.com> <20240228-multi_waitq-v1-1-ccb096419af0@quicinc.com>
In-Reply-To: <20240228-multi_waitq-v1-1-ccb096419af0@quicinc.com>
From: Bartosz Golaszewski <brgl@bgdev.pl>
Date: Tue, 5 Mar 2024 20:11:17 +0100
Message-ID: <CAMRc=McSbTmActdZ_JR+sFuW65TT=mfM+xgBFFs_31dMAO+4kA@mail.gmail.com>
Subject: Re: [PATCH 1/3] firmware: qcom-scm: Initialize waitq before setting
 global __scm
To: Unnathi Chalicheemala <quic_uchalich@quicinc.com>
Cc: Bjorn Andersson <andersson@kernel.org>, Konrad Dybcio <konrad.dybcio@linaro.org>, 
	Sibi Sankar <quic_sibis@quicinc.com>, linux-arm-msm@vger.kernel.org, 
	linux-kernel@vger.kernel.org, kernel@quicinc.com, 
	Prasad Sodagudi <quic_psdoagud@quicinc.com>, Murali Nalajala <quic_mnalajal@quicinc.com>, 
	Satya Durga Srinivasu Prabhala <quic_satyap@quicinc.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Wed, Feb 28, 2024 at 7:50=E2=80=AFPM Unnathi Chalicheemala
<quic_uchalich@quicinc.com> wrote:
>
> Suppose there's another device probing at the same time as SCM driver.
> It calls into SCM and triggers WAITQ sleep but waitq completion
> structure may not be initialized yet.
> Fix this by moving initialization of waitq before setting global __scm.
>
> Fixes: 6bf325992236 ("firmware: qcom: scm: Add wait-queue handling logic"=
)
> Signed-off-by: Unnathi Chalicheemala <quic_uchalich@quicinc.com>
> ---
>  drivers/firmware/qcom/qcom_scm.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/firmware/qcom/qcom_scm.c b/drivers/firmware/qcom/qco=
m_scm.c
> index 520de9b5633a..c1be8270ead1 100644
> --- a/drivers/firmware/qcom/qcom_scm.c
> +++ b/drivers/firmware/qcom/qcom_scm.c
> @@ -1854,11 +1854,11 @@ static int qcom_scm_probe(struct platform_device =
*pdev)
>         if (ret)
>                 return ret;
>
> +       init_completion(&scm->waitq_comp);
> +
>         __scm =3D scm;
>         __scm->dev =3D &pdev->dev;

If another driver can actually probe at the same time then this needs
to be moved to the end of this function, the `__scm->dev` assignment
must be done before the `__scm =3D scm` one and it must use
WRITE_ONCE().

Bart

>
> -       init_completion(&__scm->waitq_comp);
> -
>         irq =3D platform_get_irq_optional(pdev, 0);
>         if (irq < 0) {
>                 if (irq !=3D -ENXIO)
>
> --
> 2.25.1
>
>