Hi Iuliana,
First and foremost, you were correct to remind me of this patch - it had slipped
through. I got mixed up with your other patch[1], which has the same title
preprend and the same revision. That one is on my list of patches to review and
I should get to it later this week or early next week.
Please see below for comments on this patch.
[1]. [PATCH v4] remoteproc: imx_dsp_rproc: add custom memory copy implementation for i.MX DSP Cores
On Tue, Feb 14, 2023 at 06:37:44PM +0200, Iuliana Prodan (OSS) wrote:
> From: Iuliana Prodan <iuliana.prodan@xxxxxxx>
>
> There are cases when we want to test a simple "hello world"
> application on the DSP and we don't have IPC between the cores.
> Therefore, do not wait for a confirmation from the remote processor
> at start.
>
> Added "ignore_dsp_ready" flag while inserting the module to ignore
> remote processor reply after start.
> By default, this is off - do not ignore reply from rproc.
>
> Signed-off-by: Iuliana Prodan <iuliana.prodan@xxxxxxx>
> Reviewed-by: Daniel Baluta <daniel.baluta@xxxxxxx>
> ---
> Changes since v3
> - do not instantiate static var to 0, this is done by default
> - do not initialize mailbox if not IPC between the core
>
> Changes since v2
> - s/ignoreready/ignore_dsp_ready
>
> Changes since v1
> - change BIT(31) to BIT(1) for REMOTE_SKIP_WAIT
>
> ---
> drivers/remoteproc/imx_dsp_rproc.c | 27 +++++++++++++++++++++++++++
> 1 file changed, 27 insertions(+)
>
> diff --git a/drivers/remoteproc/imx_dsp_rproc.c b/drivers/remoteproc/imx_dsp_rproc.c
> index 95da1cbefacf..fb69f4e8ee96 100644
> --- a/drivers/remoteproc/imx_dsp_rproc.c
> +++ b/drivers/remoteproc/imx_dsp_rproc.c
> @@ -26,9 +26,18 @@
> #include "remoteproc_elf_helpers.h"
> #include "remoteproc_internal.h"
>
> +/*
> + * Module parameters
> + */
> +static unsigned int imx_dsp_rproc_ignore_ready;
> +module_param_named(ignore_dsp_ready, imx_dsp_rproc_ignore_ready, int, 0644);
This patch is about introducing a mode where the mailboxes aren't used... Why
not simply name the parameter "no_mailboxes"?
> +MODULE_PARM_DESC(ignore_dsp_ready,
> + "Ignore remote proc reply after start, default is 0 (off).");
> +
> #define DSP_RPROC_CLK_MAX 5
>
> #define REMOTE_IS_READY BIT(0)
> +#define REMOTE_SKIP_WAIT BIT(1)
> #define REMOTE_READY_WAIT_MAX_RETRIES 500
>
> /* att flags */
> @@ -282,6 +291,10 @@ static int imx_dsp_rproc_ready(struct rproc *rproc)
> struct imx_dsp_rproc *priv = rproc->priv;
> int i;
>
> + /* No IPC between the cores */
> + if (priv->flags & REMOTE_SKIP_WAIT)
> + return 0;
> +
This isn't needed since priv->rxdb_ch is NULL when mailboxes have not been
initialized.
> if (!priv->rxdb_ch)
> return 0;
>
> @@ -503,6 +516,13 @@ static int imx_dsp_rproc_mbox_init(struct imx_dsp_rproc *priv)
> struct mbox_client *cl;
> int ret;
I suggest to rename imx_dsp_rproc_mbox_init() to imx_dsp_rproc_mbox_alloc(),
introduce a new function called imx_dsp_rproc_mbox_no_alloc() that simply
returns 0 and make imx_dsp_rproc_mbox_init() a function pointer.
See imx_dsp_rproc_probe() for the rest of the solution...
>
> + /*
> + * If there is no IPC between the cores,
> + * then no need to initialize mailbox.
> + */
> + if (priv->flags & REMOTE_SKIP_WAIT)
> + return 0;
Remove this.
> +
> if (!of_get_property(dev->of_node, "mbox-names", NULL))
> return 0;
>
> @@ -562,6 +582,10 @@ static int imx_dsp_rproc_mbox_init(struct imx_dsp_rproc *priv)
>
> static void imx_dsp_rproc_free_mbox(struct imx_dsp_rproc *priv)
> {
> + /* No IPC between the cores */
> + if (priv->flags & REMOTE_SKIP_WAIT)
> + return;
> +
This isn't needed since mbox_free_channel() is able to handle a NULL parameter,
which is the case when imx_dsp_rproc_mbox_init() hasn't been called.
> mbox_free_channel(priv->tx_ch);
> mbox_free_channel(priv->rx_ch);
> mbox_free_channel(priv->rxdb_ch);
> @@ -903,6 +927,9 @@ static int imx_dsp_rproc_probe(struct platform_device *pdev)
> priv->rproc = rproc;
> priv->dsp_dcfg = dsp_dcfg;
>
> + if (imx_dsp_rproc_ignore_ready)
> + priv->flags |= REMOTE_SKIP_WAIT;
> +
if (no_mailboxes)
imx_dsp_rproc_mbox_init = imx_dsp_rproc_mbox_no_alloc;
else
imx_dsp_rproc_mbox_init = imx_dsp_rproc_mbox_alloc;
That way we don't introduce a new flag, there is no new conditionals peppered
throughout the code and calls to imx_dsp_rproc_mbox_init() remain unchainged.
Thanks,
Mathieu
> dev_set_drvdata(dev, rproc);
>
> INIT_WORK(&priv->rproc_work, imx_dsp_rproc_vq_work);
> --
> 2.17.1
>
