| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-Id: <20140518092730.109AAC40B8A@trevor.secretlab.ca>
Date: Sun, 18 May 2014 10:27:30 +0100
From: Grant Likely <grant.likely@...aro.org>
To: frowand.list@...il.com
Cc: linux-kernel@...r.kernel.org, devicetree@...r.kernel.org,
David Daney <david.daney@...ium.com>,
Pantelis Antoniou <pantelis.antoniou@...sulko.com>
Subject: Re: [PATCH 2/3] of: Make of_find_node_by_path() handle /aliases
On Fri, 16 May 2014 11:54:44 +0100, Grant Likely <grant.likely@...aro.org> wrote:
> On Thu, 15 May 2014 19:51:17 -0700, Frank Rowand <frowand.list@...il.com> wrote:
> > On 5/13/2014 7:58 AM, Grant Likely wrote:
> > > Make of_find_node_by_path() handle aliases as prefixes. To make this
> > > work the name search is refactored to search by path component instead
> > > of by full string. This should be a more efficient search, and it makes
> > > it possible to start a search at a subnode of a tree.
> > >
> > > Signed-off-by: David Daney <david.daney@...ium.com>
> > > Signed-off-by: Pantelis Antoniou <pantelis.antoniou@...sulko.com>
> > > [grant.likely: Rework to not require allocating at runtime]
> > > Acked-by: Rob Herring <robh@...nel.org>
> > > Signed-off-by: Grant Likely <grant.likely@...aro.org>
> > > ---
> > > drivers/of/base.c | 60 +++++++++++++++++++++++++++++++++++++++++++++++++++----
> > > 1 file changed, 56 insertions(+), 4 deletions(-)
> > >
> > > diff --git a/drivers/of/base.c b/drivers/of/base.c
> > > index 6e240698353b..60089b9a3014 100644
> > > --- a/drivers/of/base.c
> > > +++ b/drivers/of/base.c
> > > @@ -771,9 +771,38 @@ struct device_node *of_get_child_by_name(const struct device_node *node,
> > > }
> > > EXPORT_SYMBOL(of_get_child_by_name);
> > >
> > > +static struct device_node *__of_find_node_by_path(struct device_node *parent,
> > > + const char *path)
> > > +{
> > > + struct device_node *child;
> > > + int len = strchrnul(path, '/') - path;
> > > +
> > > + if (!len)
> > > + return parent;
> >
> > (!len) is true if the the final character of the path passed into of_find_node_by_path()
> > was "/". Strictly speaking, ->full_name will never end with "/", so the return value
> > should be NULL, indicating that the match fails.
>
> Ah, good catch. I should add a test case for that.
In my testing this looks okay. The while loop that calls into
__of_find_node_by_path() looks like this:
while (np && *path == '/') {
path++; /* Increment past '/' delimiter */
np = __of_find_node_by_path(np, path);
path = strchrnul(path, '/');
}
If the path ends with a '/', then the loop will go around one more time.
The pointer will be incremented to point at the null character and len
will be null because strchrnul() will point at the last item.
I've added a couple of test cases to make sure it works correctly:
diff --git a/drivers/of/selftest.c b/drivers/of/selftest.c
index a9d00e8c17ea..10900b18fc06 100644
--- a/drivers/of/selftest.c
+++ b/drivers/of/selftest.c
@@ -40,6 +40,12 @@ static void __init of_selftest_find_node_by_name(void)
"find /testcase-data failed\n");
of_node_put(np);
+ /* Test if trailing '/' works */
+ np = of_find_node_by_path("/testcase-data/");
+ selftest(np && !strcmp("/testcase-data", np->full_name),
+ "find /testcase-data/ failed\n");
+ of_node_put(np);
+
np = of_find_node_by_path("/testcase-data/phandle-tests/consumer-a");
selftest(np && !strcmp("/testcase-data/phandle-tests/consumer-a", np->full_name),
"find /testcase-data/phandle-tests/consumer-a failed\n");
@@ -50,6 +56,12 @@ static void __init of_selftest_find_node_by_name(void)
"find testcase-alias failed\n");
of_node_put(np);
+ /* Test if trailing '/' works on aliases */
+ np = of_find_node_by_path("testcase-alias/");
+ selftest(np && !strcmp("/testcase-data", np->full_name),
+ "find testcase-alias/ failed\n");
+ of_node_put(np);
+
np = of_find_node_by_path("testcase-alias/phandle-tests/consumer-a");
selftest(np && !strcmp("/testcase-data/phandle-tests/consumer-a", np->full_name),
"find testcase-alias/phandle-tests/consumer-a failed\n");
g.
>
> >
> > > +
> > > + for_each_child_of_node(parent, child) {
> > > + const char *name = strrchr(child->full_name, '/');
> > > + if (WARN(!name, "malformed device_node %s\n", child->full_name))
> > > + continue;
> > > + name++;
> >
> > Why go to the effort of finding the final component of child->full_name instead
> > of just using child->name?
>
> Because child->name and the final component of child->full_name is not
> the same thing. child->name has the unit address stripped off. It is
> part of how OpenFirmware expects to be used. New drivers are never
> supposed to be matching by node name, but I don't currently have a good
> way to go through any driver depending on the OpenFirmware behaviour.
>
> Ideally I'd like to get rid of the difference.
>
> >
> > > + if (strncmp(path, name, len) == 0 && (strlen(name) == len))
> > > + return child;
> > > + }
> > > + return NULL;
> > > +}
> > > +
> > > /**
> > > * of_find_node_by_path - Find a node matching a full OF path
> > > * @path: The full path to match
> > > + * @path: Either the full path to match, or if the path does not
> >
> > Delete the old @path description.
>
> Oops, thanks.
>
> >
> >
> > > + * start with '/', the name of a property of the /aliases
> > > + * node (an alias). In the case of an alias, the node
> > > + * matching the alias' value will be returned.
> > > + *
> > > + * Valid paths:
> > > + * /foo/bar Full path
> > > + * foo Valid alias
> > > + * foo/bar Valid alias + relative path
> > > *
> > > * Returns a node pointer with refcount incremented, use
> > > * of_node_put() on it when done.
> > > @@ -781,13 +810,36 @@ EXPORT_SYMBOL(of_get_child_by_name);
> > > struct device_node *of_find_node_by_path(const char *path)
> > > {
> > > struct device_node *np = of_allnodes;
> > > + struct property *pp;
> > > unsigned long flags;
> > >
> > > + /* The path could begin with an alias */
> > > + if (*path != '/') {
> > > + char *p = strchrnul(path, '/');
> > > + int len = p - path;
> > > +
> > > + /* of_aliases must not be NULL */
> > > + if (!of_aliases)
> > > + return NULL;
> > > +
> > > + np = NULL;
> > > + for_each_property_of_node(of_aliases, pp) {
> > > + if (strlen(pp->name) == len && !strncmp(pp->name, path, len)) {
> > > + np = of_find_node_by_path(pp->value);
> > > + break;
> > > + }
> > > + }
> > > + if (!np)
> > > + return NULL;
> > > + path = p;
> > > + }
> > > +
> > > + /* Step down the tree matching path components */
> > > raw_spin_lock_irqsave(&devtree_lock, flags);
> > > - for (; np; np = np->allnext) {
> > > - if (np->full_name && (of_node_cmp(np->full_name, path) == 0)
> > > - && of_node_get(np))
> > > - break;
> > > + while (np && *path == '/') {
> > > + path++; /* Increment past '/' delimiter */
> > > + np = __of_find_node_by_path(np, path);
> > > + path = strchrnul(path, '/');
> > > }
> > > raw_spin_unlock_irqrestore(&devtree_lock, flags);
> > > return np;
> > >
> >
>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists