Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Tue, 19 May 2020 16:59:44 -0700
From:   Bjorn Andersson <bjorn.andersson@linaro.org>
To:     Rishabh Bhatnagar <rishabhb@codeaurora.org>
Cc:     linux-remoteproc@vger.kernel.org, linux-kernel@vger.kernel.org,
        mathieu.poirier@linaro.org, tsoni@codeaurora.org,
        psodagud@codeaurora.org, sidgup@codeaurora.org
Subject: Re: [PATCH v3 2/3] remoteproc: Add inline coredump functionality
Message-ID: <20200519235944.GF408178@builder.lan>
References: <1589486856-23440-1-git-send-email-rishabhb@codeaurora.org>
 <1589486856-23440-3-git-send-email-rishabhb@codeaurora.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <1589486856-23440-3-git-send-email-rishabhb@codeaurora.org>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Thu 14 May 13:07 PDT 2020, Rishabh Bhatnagar wrote:

> The current coredump implementation uses vmalloc area to copy
> all the segments. But this might put strain on low memory targets
> as the firmware size sometimes is in tens of MBs. The situation
> becomes worse if there are multiple remote processors undergoing
> recovery at the same time. This patch adds inline coredump
> functionality that avoids extra memory usage. This requires
> recovery to be halted until data is read by userspace and free
> function is called.
> 

Overall I think this looks really good now, but I spotted an issue with
INLINE dumps not using segment->dump().

Also there's 3 checkpatch --strict warnings, please fix those.

> Signed-off-by: Rishabh Bhatnagar <rishabhb@codeaurora.org>
> ---
>  drivers/remoteproc/remoteproc_coredump.c | 129 +++++++++++++++++++++++++++++--
>  include/linux/remoteproc.h               |  15 ++++
>  2 files changed, 139 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/remoteproc/remoteproc_coredump.c b/drivers/remoteproc/remoteproc_coredump.c
[..]
> +static ssize_t rproc_coredump_read(char *buffer, loff_t offset, size_t count,
> +				    void *data, size_t header_sz)
> +{
> +	void *device_mem;
> +	size_t seg_data;
> +	size_t copy_sz, bytes_left = count;
> +	unsigned long addr;
> +	struct rproc_coredump_state *dump_state = data;
> +	struct rproc *rproc = dump_state->rproc;
> +	void *elfcore = dump_state->header;
> +
> +	/* Copy the vmalloc'ed header first. */
> +	if (offset < header_sz) {
> +		copy_sz = memory_read_from_buffer(buffer, count, &offset,
> +						  elfcore, header_sz);
> +		if (copy_sz < 0)
> +			return -EINVAL;
> +
> +		return copy_sz;
> +	}
> +
> +	/* Find out the segment memory chunk to be copied based on offset.
> +	 * Keep copying data until count bytes are read.
> +	 */

	/*
	 * Multiline comments start on the second line throughout
	 * remoteproc, please follow this.
	 */

> +	while (bytes_left) {
> +		addr = rproc_coredump_find_segment(offset - header_sz,
> +						   &rproc->dump_segments,
> +						   &seg_data);
> +		/* EOF check */
> +		if (seg_data == 0) {
> +			dev_info(&rproc->dev, "Ramdump done, %lld bytes read",
> +				 offset);
> +			break;
> +		}
> +
> +		copy_sz = min_t(size_t, bytes_left, seg_data);
> +
> +		device_mem = rproc_da_to_va(rproc, addr, copy_sz);
> +		if (!device_mem) {
> +			dev_err(&rproc->dev, "Coredump: %lx with size %zd out of remoteproc carveout\n",
> +				addr, copy_sz);
> +			return -ENOMEM;

I think it would be best to maintain the same behavior between INLINE
and DEFAULT here.

> +		}
> +		memcpy(buffer, device_mem, copy_sz);

This won't work for modem on e.g. SDM845, because we need to do some
special tricks to make the memory readable, that's why we invoke
segment->dump() in the DEFAULT scenario. Doing a memcpy here instead
will result in a security violation.

Perhaps this snippet can be extracted to a separate helper function,
which would allow you to avoid the next_seg goto label below.

> +
> +		offset += copy_sz;
> +		buffer += copy_sz;
> +		bytes_left -= copy_sz;
> +	}
> +
> +	return count - bytes_left;
> +}
[..]
> diff --git a/include/linux/remoteproc.h b/include/linux/remoteproc.h
> index 0468be4..ab2b9b7 100644
> --- a/include/linux/remoteproc.h
> +++ b/include/linux/remoteproc.h
> @@ -435,6 +435,19 @@ enum rproc_crash_type {
>  };
>  
>  /**
> + * enum rproc_dump_mechanism - Coredump options for core
> + * @COREDUMP_DEFAULT:	Copy dump to separate buffer and carry on with recovery
> + * @COREDUMP_INLINE:	Read segments directly from device memory. Stall
> +			recovery until all segments are read
> + * @COREDUMP_DISABLED:	Don't perform any dump
> + */
> +enum rproc_dump_mechanism {
> +	COREDUMP_DEFAULT,
> +	COREDUMP_INLINE,
> +	COREDUMP_DISABLED,

Please prefix these with RPROC_, as "coredump" has a meaning outside
remoteproc as well.

Regards,
Bjorn