| 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: <F8E336E9-E400-4918-8E65-B3EFFC943D20@konsulko.com>
Date: Mon, 9 May 2016 17:34:19 +0300
From: Pantelis Antoniou <pantelis.antoniou@...sulko.com>
To: Rob Herring <robherring2@...il.com>
Cc: Frank Rowand <frowand.list@...il.com>,
Matt Porter <mporter@...sulko.com>,
Koen Kooi <koen@...inion.thruhere.net>,
Guenter Roeck <linux@...ck-us.net>,
Marek Vasut <marex@...x.de>,
"devicetree@...r.kernel.org" <devicetree@...r.kernel.org>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 3/4] of: changesets: Introduce changeset helper methods
Hi Rob,
> On May 9, 2016, at 17:27 , Rob Herring <robherring2@...il.com> wrote:
>
> On Mon, May 9, 2016 at 8:20 AM, Pantelis Antoniou
> <pantelis.antoniou@...sulko.com> wrote:
>> Changesets are very powerful, but the lack of a helper API
>> makes using them cumbersome. Introduce a simple copy based
>> API that makes things considerably easier.
>>
>> To wit, adding a property using the raw API.
>>
>> struct property *prop;
>> prop = kzalloc(sizeof(*prop)), GFP_KERNEL);
>> prop->name = kstrdup("compatible");
>> prop->value = kstrdup("foo,bar");
>> prop->length = strlen(prop->value) + 1;
>> of_changeset_add_property(ocs, np, prop);
>>
>> while using the helper API
>>
>> of_changeset_add_property_string(ocs, np, "compatible",
>> "foo,bar");
>
> Seems useful. Do we not have any users that can be converted?
>
There is a user for this, which will be shortly posted.
> [...]
>
>> +static int __of_changeset_add_update_property_copy(struct of_changeset *ocs,
>> + struct device_node *np, const char *name, const void *value,
>> + int length, bool update)
>> +{
>> + struct property *prop;
>> + char *new_name;
>> + void *new_value;
>> + int ret = -ENOMEM;
>> +
>> + prop = kzalloc(sizeof(*prop), GFP_KERNEL);
>> + if (!prop)
>> + goto out_no_prop;
>
> Just return here.
>
OK
>> +
>> + new_name = kstrdup(name, GFP_KERNEL);
>> + if (!new_name)
>> + goto out_no_name;
>> +
>> + /*
>> + * NOTE: There is no check for zero length value.
>> + * In case of a boolean property, this will allocate a value
>> + * of zero bytes. We do this to work around the use
>> + * of of_get_property() calls on boolean values.
>> + */
>> + new_value = kmemdup(value, length, GFP_KERNEL);
>> + if (!new_value)
>> + goto out_no_value;
>> +
>> + of_property_set_flag(prop, OF_DYNAMIC);
>> +
>> + prop->name = new_name;
>> + prop->value = new_value;
>> + prop->length = length;
>> +
>> + if (!update)
>> + ret = of_changeset_add_property(ocs, np, prop);
>> + else
>> + ret = of_changeset_update_property(ocs, np, prop);
>> +
>> + if (ret != 0)
>
> if (!ret)
> return 0;
>
>
>> + goto out_no_add;
>> +
>> + return 0;
>> +
>> +out_no_add:
>
> ... and remove all this.
>
Err, there’s an exit path here from kmemdup (goto err_no_value).
We’ll be leaking memory on error.
>> + kfree(prop->value);
>> +out_no_value:
>> + kfree(prop->name);
>> +out_no_name:
>> + kfree(prop);
>> +out_no_prop:
>> + return ret;
>> +}
>> +
>> +static int __of_changeset_add_update_property_string(struct of_changeset *ocs,
>> + struct device_node *np, const char *name, const char *str,
>> + bool update)
>> +{
>> + return __of_changeset_add_update_property_copy(ocs, np, name, str,
>> + strlen(str) + 1, update);
>> +}
>> +
>> +static int __of_changeset_add_update_property_stringv(struct of_changeset *ocs,
>> + struct device_node *np, const char *name,
>> + const char *fmt, va_list vargs, bool update)
>> +{
>> + char *str;
>> + int ret;
>> +
>> + str = kvasprintf(GFP_KERNEL, fmt, vargs);
>> + if (!str)
>> + return -ENOMEM;
>> +
>> + ret = __of_changeset_add_update_property_string(ocs, np, name, str, update);
>> +
>> + kfree(str);
>> +
>> + return ret;
>> +}
>> +
>> +static int __of_changeset_add_update_property_string_list(
>> + struct of_changeset *ocs, struct device_node *np, const char *name,
>> + const char **strs, int count, bool update)
>> +{
>> + int total = 0, i, ret;
>> + char *value, *s;
>> +
>> + for (i = 0; i < count; i++) {
>> + /* check if it's NULL */
>> + if (!strs[i])
>> + return -EINVAL;
>> + total += strlen(strs[i]) + 1;
>> + }
>> +
>> + value = kmalloc(total, GFP_KERNEL);
>> + if (!value)
>> + return -ENOMEM;
>> +
>> + for (i = 0, s = value; i < count; i++) {
>> + /* no need to check for NULL, check above */
>> + strcpy(s, strs[i]);
>> + s += strlen(strs[i]) + 1;
>> + }
>> +
>> + ret = __of_changeset_add_update_property_copy(ocs, np, name, value,
>> + total, update);
>> +
>> + kfree(value);
>> +
>> + return ret;
>> +}
>> +
>> +static int __of_changeset_add_update_property_u32(struct of_changeset *ocs,
>> + struct device_node *np, const char *name, u32 val, bool update)
>> +{
>> + /* in place */
>> + val = cpu_to_be32(val);
>
> Kill this function and move this to the 2 callers.
>
>> + return __of_changeset_add_update_property_copy(ocs, np, name, &val,
>> + sizeof(val), update);
>> +}
>> +
>> +static int __of_changeset_add_update_property_bool(struct of_changeset *ocs,
>> + struct device_node *np, const char *name, bool update)
>> +{
>
> I think this intermediate function could be killed off too.
>
>> + return __of_changeset_add_update_property_copy(ocs, np, name, "", 0,
>> + update);
>> +}
>
> [...]
>
>> +static struct device_node *
>> +__of_changeset_node_move_one(struct of_changeset *ocs,
>> + struct device_node *np, struct device_node *new_parent)
>> +{
>> + struct device_node *np2;
>> + const char *unitname;
>> + int err;
>> +
>> + err = of_changeset_detach_node(ocs, np);
>> + if (err)
>> + return ERR_PTR(err);
>> +
>> + unitname = strrchr(np->full_name, '/');
>> + if (!unitname)
>> + unitname = np->full_name;
>> +
>> + np2 = __of_node_dup(np, "%s/%s",
>> + new_parent->full_name, unitname);
>> + if (!np2)
>> + return ERR_PTR(-ENOMEM);
>> + np2->parent = new_parent;
>> +
>> + err = of_changeset_attach_node(ocs, np2);
>> + if (err)
>> + return ERR_PTR(err);
>> +
>> + return np2;
>> +}
>> +
>> +/**
>> + * of_changeset_node_move_to - Moves a subtree to a new place in
>> + * the tree
>> + *
>> + * @ocs: changeset pointer
>> + * @np: device node pointer to be moved
>> + * @to: device node of the new parent
>> + *
>> + * Moves a subtree to a new place in the tree.
>> + * Note that a move is a safe operation because the phandles
>> + * remain valid.
>
> I'm having a hard time imagining a usecase for this one. I would drop
> until we have a user in tree.
>
There’s a use case - it will be apparent with the next patches.
>> + *
>> + * Returns zero on success, a negative error value otherwise.
>> + */
>> +int of_changeset_node_move(struct of_changeset *ocs,
>> + struct device_node *np, struct device_node *new_parent)
>> +{
>> + struct device_node *npc, *nppc;
>> +
>> + /* move the root first */
>> + nppc = __of_changeset_node_move_one(ocs, np, new_parent);
>> + if (IS_ERR(nppc))
>> + return PTR_ERR(nppc);
>> +
>> + /* move the subtrees next */
>> + for_each_child_of_node(np, npc) {
>> + nppc = __of_changeset_node_move_one(ocs, npc, nppc);
>> + if (IS_ERR(nppc)) {
>> + of_node_put(npc);
>> + return PTR_ERR(nppc);
>> + }
>> + }
>> +
>> + return 0;
>> +}
>> diff --git a/include/linux/of.h b/include/linux/of.h
>> index 3175803..f015911 100644
>> --- a/include/linux/of.h
>> +++ b/include/linux/of.h
>> @@ -1037,6 +1037,42 @@ static inline int of_changeset_update_property(struct of_changeset *ocs,
>> {
>> return of_changeset_action(ocs, OF_RECONFIG_UPDATE_PROPERTY, np, prop);
>> }
>> +
>> +struct device_node *of_changeset_create_device_nodev(
>> + struct of_changeset *ocs, struct device_node *parent,
>> + const char *fmt, va_list vargs);
>> +__printf(3, 4) struct device_node *of_changeset_create_device_node(
>> + struct of_changeset *ocs, struct device_node *parent,
>> + const char *fmt, ...);
>> +int of_changeset_add_property_copy(struct of_changeset *ocs,
>> + struct device_node *np, const char *name,
>> + const void *value, int length);
>> +int of_changeset_add_property_string(struct of_changeset *ocs,
>> + struct device_node *np, const char *name, const char *str);
>> +__printf(4, 5) int of_changeset_add_property_stringf(struct of_changeset *ocs,
>> + struct device_node *np, const char *name, const char *fmt, ...);
>> +int of_changeset_add_property_string_list(struct of_changeset *ocs,
>> + struct device_node *np, const char *name, const char **strs, int count);
>> +int of_changeset_add_property_u32(struct of_changeset *ocs,
>> + struct device_node *np, const char *name, u32 val);
>> +int of_changeset_add_property_bool(struct of_changeset *ocs,
>> + struct device_node *np, const char *name);
>> +int of_changeset_update_property_copy(struct of_changeset *ocs,
>> + struct device_node *np, const char *name,
>> + const void *value, int length);
>> +int of_changeset_update_property_string(struct of_changeset *ocs,
>> + struct device_node *np, const char *name, const char *str);
>> +__printf(4, 5) int of_changeset_update_property_stringf(struct of_changeset *ocs,
>> + struct device_node *np, const char *name, const char *fmt, ...);
>> +int of_changeset_update_property_string_list(struct of_changeset *ocs,
>> + struct device_node *np, const char *name, const char **strs, int count);
>> +int of_changeset_update_property_u32(struct of_changeset *ocs,
>> + struct device_node *np, const char *name, u32 val);
>> +int of_changeset_update_property_bool(struct of_changeset *ocs,
>> + struct device_node *np, const char *name);
>> +int of_changeset_node_move(struct of_changeset *ocs,
>> + struct device_node *np, struct device_node *new_parent);
>
> A bunch of these can be static inline wrappers.
>
> You need EXPORT_SYMBOL_GPL on these as well.
>
I will add the exports.
> Rob
Regards
— Pantelis
Powered by blists - more mailing lists