Subject: RE: [PATCH V4 2/2] firmware: arm_scmi: add smc/hvc transport

Hi Sudeep,

quoted

Subject: Re: [PATCH V4 2/2] firmware: arm_scmi: add smc/hvc transport

On Tue, Mar 03, 2020 at 10:06:59AM +0800, peng.fan@nxp.com wrote:

quoted

From: Peng Fan <peng.fan@nxp.com>

Take arm,smc-id as the 1st arg, leave the other args as zero for now.
There is no Rx, only Tx because of smc/hvc not support Rx.

Signed-off-by: Peng Fan <peng.fan@nxp.com>

[...]

quoted

+static int smc_send_message(struct scmi_chan_info *cinfo,
+			    struct scmi_xfer *xfer)
+{
+	struct scmi_smc *scmi_info = cinfo->transport_info;
+	struct arm_smccc_res res;
+
+	shmem_tx_prepare(scmi_info->shmem, xfer);

How do we protect another thread/process on another CPU going and
modifying the same shmem with another request ? We may need notion of
channel with associated shmem and it is protected with some lock.

This is valid concern. But I think if shmem is shared bwteen protocols, the
access to shmem should be protected in
drivers/firmware/arm_scmi/driver.c: scmi_do_xfer, because send_message
and fetch_response both touches shmem

The mailbox transport also has the issue you mentioned, I think.

Ignore my upper comments. How do think the following diff based on current patch?

If ok, I'll squash it with current patch and send out v5.

diff --git a/drivers/firmware/arm_scmi/smc.c b/drivers/firmware/arm_scmi/smc.c
index 88f91b68f297..7d770112f339 100644
--- a/drivers/firmware/arm_scmi/smc.c
+++ b/drivers/firmware/arm_scmi/smc.c

@@ -29,6 +29,8 @@ struct scmi_smc {
        u32 func_id;
 };

+static DEFINE_MUTEX(smc_mutex);
+
 static bool smc_chan_available(struct device *dev, int idx)
 {
        return true;

@@ -99,11 +101,15 @@ static int smc_send_message(struct scmi_chan_info *cinfo,
        struct scmi_smc *scmi_info = cinfo->transport_info;
        struct arm_smccc_res res;

+       mutex_lock(&smc_mutex);
+
        shmem_tx_prepare(scmi_info->shmem, xfer);

        arm_smccc_1_1_invoke(scmi_info->func_id, 0, 0, 0, 0, 0, 0, 0, &res);
        scmi_rx_callback(scmi_info->cinfo, shmem_read_header(scmi_info->shmem));

+       mutex_unlock(&smc_mutex);
+
        return res.a0;
 }

Thanks,
Peng.

Thanks,
Peng.

quoted

--
Regards,
Sudeep