Re: [Patch v5 5/8] firmware: qcom: scm: Convert to streaming DMA APIS

From: Andy Gross <andy.gross@linaro.org>
To: Stephen Boyd <sboyd@codeaurora.org>
Cc: linux-arm-msm@vger.kernel.org,
	linux-arm-kernel@lists.infradead.org,
	linux-kernel@vger.kernel.org,
	Bjorn Andersson <bjorn.andersson@linaro.org>,
	devicetree@vger.kernel.org, jilai wang <jilaiw@codeaurora.org>
Subject: Re: [Patch v5 5/8] firmware: qcom: scm: Convert to streaming DMA APIS
Date: Thu, 2 Jun 2016 22:57:48 -0500	[thread overview]
Message-ID: <20160603035748.GC13357@hector.attlocal.net> (raw)
In-Reply-To: <20160602232602.GS28218@codeaurora.org>

On Thu, Jun 02, 2016 at 04:26:02PM -0700, Stephen Boyd wrote:
> On 05/12, Andy Gross wrote:
> > This patch converts the Qualcomm SCM driver to use the streaming DMA APIs
> > for communication buffers.
> 
> Yes, but why?

This was done so that we can remove the secure_flush mechanism, as we are the
only consumer (that I could find).  Using the DMA APIs would match the scm-64 so
it would be symmetric.  I'll add this to the commit message to give more reason
why the changes are being made.

> > 
> > Signed-off-by: Andy Gross <andy.gross@linaro.org>
> > ---
> >  drivers/firmware/qcom_scm-32.c | 189 +++++++++++------------------------------
> >  drivers/firmware/qcom_scm.c    |   6 +-
> >  drivers/firmware/qcom_scm.h    |  10 ++-
> >  3 files changed, 59 insertions(+), 146 deletions(-)
> > 
> > diff --git a/drivers/firmware/qcom_scm-32.c b/drivers/firmware/qcom_scm-32.c
> > index 4388d13..e92bf7a 100644
> > --- a/drivers/firmware/qcom_scm-32.c
> > +++ b/drivers/firmware/qcom_scm-32.c
> > @@ -64,11 +63,11 @@ static DEFINE_MUTEX(qcom_scm_lock);
> >   *
> >   *	------------------- <--- struct qcom_scm_command
> >   *	| command header  |
> > - *	------------------- <--- qcom_scm_get_command_buffer()
> > + *	------------------- <--- buf[0]
> >   *	| command buffer  |
> > - *	------------------- <--- struct qcom_scm_response and
> > - *	| response header |      qcom_scm_command_to_response()
> > - *	------------------- <--- qcom_scm_get_response_buffer()
> > + *	------------------- <--- struct qcom_scm_response
> > + *	| response header |
> > + *	-------------------
> 
> You don't like my convenience functions? :) I always thought
> qcom_scm_get_response_buffer() read better than
> 
> (void *)cmd->buf + le32_to_cpu(cmd->resp_hdr_offset);

If it was used in more than one function, yes.  But I felt that single use
functions were kind of a waste.  But this is in the eye of the beholder.

> 
> >   *	| response buffer |
> >   *	-------------------
> >   *
> > @@ -96,79 +95,7 @@ struct qcom_scm_response {
> >  	__le32 is_complete;
> >  };
> >  
> > -
> > -/**
> > - * free_qcom_scm_command() - Free an SCM command
> > - * @cmd: command to free
> > - *
> > - * Free an SCM command.
> > - */
> > -static inline void free_qcom_scm_command(struct qcom_scm_command *cmd)
> > -{
> > -	kfree(cmd);
> > -}
> > -
> > -/**
> > - * qcom_scm_command_to_response() - Get a pointer to a qcom_scm_response
> > - * @cmd: command
> > - *
> > - * Returns a pointer to a response for a command.
> > - */
> > -static inline struct qcom_scm_response *qcom_scm_command_to_response(
> > -		const struct qcom_scm_command *cmd)
> > -{
> > -	return (void *)cmd + le32_to_cpu(cmd->resp_hdr_offset);
> > -}
> > -
> > -/**
> > - * qcom_scm_get_command_buffer() - Get a pointer to a command buffer
> > - * @cmd: command
> > - *
> > - * Returns a pointer to the command buffer of a command.
> > - */
> > -static inline void *qcom_scm_get_command_buffer(const struct qcom_scm_command *cmd)
> > -{
> > -	return (void *)cmd->buf;
> > -}
> > -
> > -/**
> > - * qcom_scm_get_response_buffer() - Get a pointer to a response buffer
> > - * @rsp: response
> > - *
> > - * Returns a pointer to a response buffer of a response.
> > - */
> > -static inline void *qcom_scm_get_response_buffer(const struct qcom_scm_response *rsp)
> > -{
> > -	return (void *)rsp + le32_to_cpu(rsp->buf_offset);
> > -}
> 
> At the least the diff would be more concentrated on what really
> is happening in this patch if we left these functions as is.

Ok.  I'll run that and see how it turns out.  

> > -
> > -static u32 smc(u32 cmd_addr)
> > +static u32 smc(dma_addr_t cmd_addr)
> 
> Please leave this as u32, the interface doesn't support anything
> wider here so we shouldn't let this change on LPAE.

Fair point.  I hadn't considered that.  Thanks for catching that.

> >  {
> >  	int context_id;
> >  	register u32 r0 asm("r0") = 1;
> > @@ -192,45 +119,9 @@ static u32 smc(u32 cmd_addr)
> >  	return r0;
> >  }
> >  
> > -static int __qcom_scm_call(const struct qcom_scm_command *cmd)
> > -{
> > -	int ret;
> > -	u32 cmd_addr = virt_to_phys(cmd);
> > -
> > -	/*
> > -	 * Flush the command buffer so that the secure world sees
> > -	 * the correct data.
> > -	 */
> > -	secure_flush_area(cmd, cmd->len);
> 
> Yay we should delete secure_flush_area() too.

Yes I was going to do a follow on to remove that.  I opted to not add this to
the set of patches.

> > -
> > -	ret = smc(cmd_addr);
> > -	if (ret < 0)
> > -		ret = qcom_scm_remap_error(ret);
> > -
> > -	return ret;
> > -}
> > -
> > -
> > @@ -143,7 +143,7 @@ int qcom_scm_hdcp_req(struct qcom_scm_hdcp_req *req, u32 req_cnt, u32 *resp)
> >  	if (ret)
> >  		return ret;
> >  
> > -	ret = __qcom_scm_hdcp_req(req, req_cnt, resp);
> > +	ret = __qcom_scm_hdcp_req(__scm->dev, req, req_cnt, resp);
> 
> Hmm maybe we should pass __scm instead of dev? Is there any
> advantage to that?

If we needed access to something in the scm other than the device, yes.  If we
passed scm instead, we might be able to remove the singleton usage.  I'll look
and see if it makes sense.  Otherwise i'd leave the passing of the device.