devicetree-compiler.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: David Gibson <david-xT8FGy+AXnRB3Ne2BGzF6laj5H9X9Tb+@public.gmane.org>
To: "Wang Tim(王艇艇)" <timwang-XOWFcpiLszpWk0Htik3J/w@public.gmane.org>
Cc: "devicetree-compiler-u79uwXL29TY76Z2rM5mHXA@public.gmane.org"
	<devicetree-compiler-u79uwXL29TY76Z2rM5mHXA@public.gmane.org>,
	"wtt_usst-9Onoh4P/yGk@public.gmane.org"
	<wtt_usst-9Onoh4P/yGk@public.gmane.org>
Subject: Re: [PATCH] Implement the -a option to pad dtb aligned
Date: Tue, 28 Jun 2016 15:28:56 +1000	[thread overview]
Message-ID: <20160628052855.GB4242@voom.fritz.box> (raw)
In-Reply-To: <7bbb538e64e740cdbb97e21c1ff0522b-w/Y+SVW4SlHx9BLjpE4ezaBKnGwkPULj@public.gmane.org>

[-- Attachment #1: Type: text/plain, Size: 7176 bytes --]

On Mon, Jun 27, 2016 at 08:49:37AM +0000, Wang Tim(王艇艇) wrote:
> Hi David,
> (Seems my pervious mail rejected by vger.kernel.org).
> 
> Thanks for your kindly review!
> About the usage of this patch, as I know,
> some android phone and google TV soc vendors or production vendors
> need such option to cat all dtb files to support several products.
> Such as one soc vendor support two customers products. And each of
> them use different screen panel and camera sensor. But most other peripherals
> are same as the common board.
> We can use one kernel Image and several dtb files to support both of them.

Right, I see why you need the alignment in the output.  But it's
pretty easy to introduce this as you produce the archive, without
folding it into dtc.

For example:
	$ dd if=1.dtb bs=8 conv=sync of=dtbset
	$ dd if=2.dtb bs=8 conv=sync,notrunc oflag=append of=dtbset
	$ dd if=3.dtb bs=8 conv=sync,notrunc oflag=append of=dtbset
		.
		.
		.

Will create dtbset with all the dtbs aligned to 8 byte boundaries.

> The calculate method of -a option after -p/-S is totally wrong.
> I've fix it in attached patch file, please help to review again.
> 
> But I'm not sure whether my dt_to_asm part change is right.
> Please show me more details if my patch is wrong again.
> 
> Here is the updated patch:
> [PATCH] Implement the -a option to pad dtb aligned
> 
> There is one condition that need cat the dtb files
> into one dtb.img which can support several boards
> use same SoC platform.
> 
> And the original dtb file size is not aligned to any base.
> This may cause "Synchronous Abort" when load from a unligned
> address on some SoC machine, such as ARM.
> 
> So this patch implement the -a <aligned number> option to
> pad zero at the end of dtb files and make the dtb size aligned
> to <aligned number>.
> 
> Then, the aligned dtbs can cat together and load without "Synchronous
> Abort".
> 
> Signed-off-by: Tim Wang <timwang-XOWFcpiLszpWk0Htik3J/w@public.gmane.org>
> ---
>  dtc.c      |  9 ++++++++-
>  dtc.h      |  1 +
>  flattree.c | 13 ++++++++++++-
>  3 files changed, 21 insertions(+), 2 deletions(-)
> 
> diff --git a/dtc.c b/dtc.c
> index 5fa23c4..1749d26 100644
> --- a/dtc.c
> +++ b/dtc.c
> @@ -30,6 +30,7 @@ int quiet;		/* Level of quietness */
>  int reservenum;		/* Number of memory reservation slots */
>  int minsize;		/* Minimum blob size */
>  int padsize;		/* Additional padding to blob */
> +int align_size;		/* Additional padding to blob accroding to the align size */
>  int phandle_format = PHANDLE_BOTH;	/* Use linux,phandle or phandle properties */
>  
>  static void fill_fullpaths(struct node *tree, const char *prefix)
> @@ -53,7 +54,7 @@ static void fill_fullpaths(struct node *tree, const char *prefix)
>  #define FDT_VERSION(version)	_FDT_VERSION(version)
>  #define _FDT_VERSION(version)	#version
>  static const char usage_synopsis[] = "dtc [options] <input file>";
> -static const char usage_short_opts[] = "qI:O:o:V:d:R:S:p:fb:i:H:sW:E:hv";
> +static const char usage_short_opts[] = "qI:O:o:V:d:R:S:p:a:fb:i:H:sW:E:hv";
>  static struct option const usage_long_opts[] = {
>  	{"quiet",            no_argument, NULL, 'q'},
>  	{"in-format",         a_argument, NULL, 'I'},
> @@ -64,6 +65,7 @@ static struct option const usage_long_opts[] = {
>  	{"reserve",           a_argument, NULL, 'R'},
>  	{"space",             a_argument, NULL, 'S'},
>  	{"pad",               a_argument, NULL, 'p'},
> +	{"align",             a_argument, NULL, 'a'},
>  	{"boot-cpu",          a_argument, NULL, 'b'},
>  	{"force",            no_argument, NULL, 'f'},
>  	{"include",           a_argument, NULL, 'i'},
> @@ -91,6 +93,7 @@ static const char * const usage_opts_help[] = {
>  	"\n\tMake space for <number> reserve map entries (for dtb and asm output)",
>  	"\n\tMake the blob at least <bytes> long (extra space)",
>  	"\n\tAdd padding to the blob of <bytes> long (extra space)",
> +	"\n\tMake the blob align to the <bytes> (extra space)",
>  	"\n\tSet the physical boot cpu",
>  	"\n\tTry to produce output even if the input tree has errors",
>  	"\n\tAdd a path to search for include files",
> @@ -169,6 +172,7 @@ int main(int argc, char *argv[])
>  	reservenum = 0;
>  	minsize    = 0;
>  	padsize    = 0;
> +	align_size = 0;
>  
>  	while ((opt = util_getopt_long()) != EOF) {
>  		switch (opt) {
> @@ -196,6 +200,9 @@ int main(int argc, char *argv[])
>  		case 'p':
>  			padsize = strtol(optarg, NULL, 0);
>  			break;
> +		case 'a':
> +			align_size = strtol(optarg, NULL, 0);
> +			break;
>  		case 'f':
>  			force = true;
>  			break;
> diff --git a/dtc.h b/dtc.h
> index 56212c8..b406d21 100644
> --- a/dtc.h
> +++ b/dtc.h
> @@ -53,6 +53,7 @@ extern int quiet;		/* Level of quietness */
>  extern int reservenum;		/* Number of memory reservation slots */
>  extern int minsize;		/* Minimum blob size */
>  extern int padsize;		/* Additional padding to blob */
> +extern int align_size;		/* Additional padding to blob accroding to the align size */
>  extern int phandle_format;	/* Use linux,phandle or phandle properties */
>  
>  #define PHANDLE_LEGACY	0x1
> diff --git a/flattree.c b/flattree.c
> index ec14954..29f0a54 100644
> --- a/flattree.c
> +++ b/flattree.c
> @@ -398,10 +398,12 @@ void dt_to_blob(FILE *f, struct boot_info *bi, int version)
>  	 */
>  	if (minsize > 0) {
>  		padlen = minsize - fdt32_to_cpu(fdt.totalsize);
> -		if ((padlen < 0) && (quiet < 1))
> +		if ((padlen < 0) && (quiet < 1)) {
>  			fprintf(stderr,
>  				"Warning: blob size %d >= minimum size %d\n",
>  				fdt32_to_cpu(fdt.totalsize), minsize);
> +			padlen = 0;
> +		}
>  	}
>  
>  	if (padsize > 0)
> @@ -413,6 +415,13 @@ void dt_to_blob(FILE *f, struct boot_info *bi, int version)
>  		fdt.totalsize = cpu_to_fdt32(tsize);
>  	}
>  
> +	if (align_size > 0) {
> +		int tsize = fdt32_to_cpu(fdt.totalsize);
> +		padlen += ALIGN(tsize, align_size) - tsize;
> +		tsize = ALIGN(tsize, align_size);
> +		fdt.totalsize = cpu_to_fdt32(tsize);
> +	}
> +

This is still wrong.  The previous if already increased totalsize by
padlen.  Now you've adjusted padlen and added it to totalsize again.

Please, actually test this with both -p and -a options.

>  	/*
>  	 * Assemble the blob: start with the header, add with alignment
>  	 * the reserve buffer, add the reserve map terminating zeroes,
> @@ -478,6 +487,8 @@ void dt_to_asm(FILE *f, struct boot_info *bi, int version)
>  	fprintf(f, "/* autogenerated by dtc, do not edit */\n\n");
>  
>  	emit_label(f, symprefix, "blob_start");
> +	if (align_size > 0)
> +		asm_emit_align(f, align_size);

This is correct, as long as align_size is a power of two - otherwise
it will probably cause an assembler error.

>  	emit_label(f, symprefix, "header");
>  	fprintf(f, "\t/* magic */\n");
>  	asm_emit_cell(f, FDT_MAGIC);

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

  parent reply	other threads:[~2016-06-28  5:28 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-06-24 20:32 [PATCH] Implement the -a option to pad dtb aligned Tim Wang
     [not found] ` <1466800337-1729-1-git-send-email-timwang-XOWFcpiLszpWk0Htik3J/w@public.gmane.org>
2016-06-27  5:45   ` David Gibson
     [not found]     ` <20160627054507.GO4242-RXTfZT5YzpxwFLYp8hBm2A@public.gmane.org>
2016-06-27  8:49       ` Wang Tim(王艇艇)
     [not found]         ` <7bbb538e64e740cdbb97e21c1ff0522b-w/Y+SVW4SlHx9BLjpE4ezaBKnGwkPULj@public.gmane.org>
2016-06-28  5:28           ` David Gibson [this message]
     [not found]             ` <20160628052855.GB4242-RXTfZT5YzpxwFLYp8hBm2A@public.gmane.org>
2016-07-01  2:26               ` [PATCH V2] " Wang Tim(王艇艇)

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20160628052855.GB4242@voom.fritz.box \
    --to=david-xt8fgy+axnrb3ne2bgzf6laj5h9x9tb+@public.gmane.org \
    --cc=devicetree-compiler-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \
    --cc=timwang-XOWFcpiLszpWk0Htik3J/w@public.gmane.org \
    --cc=wtt_usst-9Onoh4P/yGk@public.gmane.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).