[PATCH] When dumping from file system show results in sane order

[PATCH] When dumping from file system show results in sane order

[Matthew McClintock]

Currently, when we run the following

$> dtc -I fs -O dts /proc/device-tree

We get the output in a "reverse order", this patch will reverse that
output order and should resemble your initial device more closely

Code copied over from kexec-tools

Signed-off-by: Matthew McClintock <msm-KZfg59tc24xl57MIdRCFDg@public.gmane.org>
---
 fstree.c |   29 ++++++++++++++++++++++-------
 1 files changed, 22 insertions(+), 7 deletions(-)

diff --git a/fstree.c b/fstree.c
index f377453..fd2658d 100644
--- a/fstree.c
+++ b/fstree.c
@@ -23,21 +23,35 @@
 #include <dirent.h>
 #include <sys/stat.h>
 
+static int comparefunc(const void *dentry1, const void *dentry2)
+{
+	const char *str1 = (*(struct dirent * const *)dentry1)->d_name;
+	const char *str2 = (*(struct dirent * const *)dentry2)->d_name;
+
+	if (strchr(str1, '@') && strchr(str2, '@') &&
+		(strlen(str1) > strlen(str2)))
+		return 1;
+
+	return strcmp(str1, str2);
+}
+
 static struct node *read_fstree(const char *dirname)
 {
-	DIR *d;
-	struct dirent *de;
 	struct stat st;
 	struct node *tree;
+	struct dirent **namelist;
+	int numlist, i;
 
-	d = opendir(dirname);
-	if (!d)
-		die("Couldn't opendir() \"%s\": %s\n", dirname, strerror(errno));
+	numlist = scandir(dirname, &namelist, 0, comparefunc);
+	if (numlist <= 0)
+		die("unrecoverable error: could not scan \"%s\": %s\n",
+                    dirname, strerror(errno));
 
 	tree = build_node(NULL, NULL);
 
-	while ((de = readdir(d)) != NULL) {
+	for (i = 0; i < numlist; i++) {
 		char *tmpnam;
+		struct dirent *de = namelist[i];
 
 		if (streq(de->d_name, ".")
 		    || streq(de->d_name, ".."))
@@ -72,10 +86,11 @@ static struct node *read_fstree(const char *dirname)
 			add_child(tree, newchild);
 		}
 
+		free(de);
 		free(tmpnam);
 	}
+	free(namelist);
 
-	closedir(d);
 	return tree;
 }
 
-- 
1.6.0.6

Re: [PATCH] When dumping from file system show results in sane order

[Grant Likely]

On Tue, Oct 05, 2010 at 02:07:21PM -0500, Matthew McClintock wrote:
> Currently, when we run the following
> 
> $> dtc -I fs -O dts /proc/device-tree
> 
> We get the output in a "reverse order", this patch will reverse that
> output order and should resemble your initial device more closely
> 
> Code copied over from kexec-tools
> 
> Signed-off-by: Matthew McClintock <msm-KZfg59tc24xl57MIdRCFDg@public.gmane.org>

Hi Matthew.  David just posted a patch implementing a generic search
option when outputting a tree in dts format which would be applicable
to more use cases than just sucking up the fstree.  Would that do the
job for you?

g.

> ---
>  fstree.c |   29 ++++++++++++++++++++++-------
>  1 files changed, 22 insertions(+), 7 deletions(-)
> 
> diff --git a/fstree.c b/fstree.c
> index f377453..fd2658d 100644
> --- a/fstree.c
> +++ b/fstree.c
> @@ -23,21 +23,35 @@
>  #include <dirent.h>
>  #include <sys/stat.h>
>  
> +static int comparefunc(const void *dentry1, const void *dentry2)
> +{
> +	const char *str1 = (*(struct dirent * const *)dentry1)->d_name;
> +	const char *str2 = (*(struct dirent * const *)dentry2)->d_name;
> +
> +	if (strchr(str1, '@') && strchr(str2, '@') &&
> +		(strlen(str1) > strlen(str2)))
> +		return 1;
> +
> +	return strcmp(str1, str2);
> +}
> +
>  static struct node *read_fstree(const char *dirname)
>  {
> -	DIR *d;
> -	struct dirent *de;
>  	struct stat st;
>  	struct node *tree;
> +	struct dirent **namelist;
> +	int numlist, i;
>  
> -	d = opendir(dirname);
> -	if (!d)
> -		die("Couldn't opendir() \"%s\": %s\n", dirname, strerror(errno));
> +	numlist = scandir(dirname, &namelist, 0, comparefunc);
> +	if (numlist <= 0)
> +		die("unrecoverable error: could not scan \"%s\": %s\n",
> +                    dirname, strerror(errno));
>  
>  	tree = build_node(NULL, NULL);
>  
> -	while ((de = readdir(d)) != NULL) {
> +	for (i = 0; i < numlist; i++) {
>  		char *tmpnam;
> +		struct dirent *de = namelist[i];
>  
>  		if (streq(de->d_name, ".")
>  		    || streq(de->d_name, ".."))
> @@ -72,10 +86,11 @@ static struct node *read_fstree(const char *dirname)
>  			add_child(tree, newchild);
>  		}
>  
> +		free(de);
>  		free(tmpnam);
>  	}
> +	free(namelist);
>  
> -	closedir(d);
>  	return tree;
>  }
>  
> -- 
> 1.6.0.6
> 
> 
> _______________________________________________
> devicetree-discuss mailing list
> devicetree-discuss-uLR06cmDAlY/bJ5BZ2RsiQ@public.gmane.org
> https://lists.ozlabs.org/listinfo/devicetree-discuss

Re: [PATCH] When dumping from file system show results in sane order

[Matthew McClintock]

On Oct 11, 2010, at 12:54 AM, Grant Likely wrote:

> On Tue, Oct 05, 2010 at 02:07:21PM -0500, Matthew McClintock wrote:
>> Currently, when we run the following
>> 
>> $> dtc -I fs -O dts /proc/device-tree
>> 
>> We get the output in a "reverse order", this patch will reverse that
>> output order and should resemble your initial device more closely
>> 
>> Code copied over from kexec-tools
>> 
>> Signed-off-by: Matthew McClintock <msm-KZfg59tc24xl57MIdRCFDg@public.gmane.org>
> 
> Hi Matthew.  David just posted a patch implementing a generic search
> option when outputting a tree in dts format which would be applicable
> to more use cases than just sucking up the fstree.  Would that do the
> job for you?
> 
> g.

I just noticed that, I suspect it will work. I will test it out and get back to everyone.

-M

Re: [PATCH] When dumping from file system show results in sane order

[David Gibson]

On Tue, Oct 05, 2010 at 02:07:21PM -0500, Matthew McClintock wrote:
> Currently, when we run the following
> 
> $> dtc -I fs -O dts /proc/device-tree
> 
> We get the output in a "reverse order", this patch will reverse that
> output order and should resemble your initial device more closely

Reverse order from what?  Directory entries are unordered for most
intents and purposes - there's no guarantee that readdir() will return
entries in a consistent order, and behaviour in practice will vary
from one filesystem to another.

Your patch doesn't reverse the order - it sorts the items.  If that's
what you actually want, then my general dtdiff and sorting patch will
serve you better, as Grant suggests.

-- 
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

Re: [PATCH] When dumping from file system show results in sane order

[Matthew McClintock]

On Tue, 2010-10-12 at 10:55 +1100, David Gibson wrote:
> On Tue, Oct 05, 2010 at 02:07:21PM -0500, Matthew McClintock wrote:
> > Currently, when we run the following
> > 
> > $> dtc -I fs -O dts /proc/device-tree
> > 
> > We get the output in a "reverse order", this patch will reverse that
> > output order and should resemble your initial device more closely
> 
> Reverse order from what?  Directory entries are unordered for most
> intents and purposes - there's no guarantee that readdir() will return
> entries in a consistent order, and behaviour in practice will vary
> from one filesystem to another.

Ok, maybe I should word it as now ordered. =)

> 
> Your patch doesn't reverse the order - it sorts the items.  If that's
> what you actually want, then my general dtdiff and sorting patch will
> serve you better, as Grant suggests.
> 

Yep, I concur your version is better and I would like to see that
accepted for my own reasons.

Cheers,
Matthew