When using internal regulatory data base kconfig option
do not try to call crda, this complely defeats the
purposes of enabling internal regulatory db.
Signed-off-by: Chaitanya T K <[email protected]>
---
net/wireless/reg.c | 29 +++++++++++++++--------------
1 file changed, 15 insertions(+), 14 deletions(-)
diff --git a/net/wireless/reg.c b/net/wireless/reg.c
index 7d20d84..1cb0f99 100644
--- a/net/wireless/reg.c
+++ b/net/wireless/reg.c
@@ -454,16 +454,16 @@ static void reg_regdb_search(struct work_struct *work)
static DECLARE_WORK(reg_regdb_work, reg_regdb_search);
-static void reg_regdb_query(const char *alpha2)
+static int reg_regdb_query(const char *alpha2)
{
struct reg_regdb_search_request *request;
if (!alpha2)
- return;
+ return -1;
request = kzalloc(sizeof(struct reg_regdb_search_request), GFP_KERNEL);
if (!request)
- return;
+ return -1;
memcpy(request->alpha2, alpha2, 2);
@@ -472,6 +472,7 @@ static void reg_regdb_query(const char *alpha2)
mutex_unlock(®_regdb_search_mutex);
schedule_work(®_regdb_work);
+ return 0;
}
/* Feel free to add any other sanity checks here */
@@ -482,7 +483,15 @@ static void reg_regdb_size_check(void)
}
#else
static inline void reg_regdb_size_check(void) {}
-static inline void reg_regdb_query(const char *alpha2) {}
+static inline int reg_regdb_query(const char *alpha2)
+{
+ if (!is_world_regdom((char *) alpha2))
+ pr_info("Calling CRDA for country: %c%c\n",
+ alpha2[0], alpha2[1]);
+ else
+ pr_info("Calling CRDA to update world regulatory domain\n");
+ return kobject_uevent(®_pdev->dev.kobj, KOBJ_CHANGE);
+}
#endif /* CONFIG_CFG80211_INTERNAL_REGDB */
/*
@@ -492,16 +501,8 @@ static inline void reg_regdb_query(const char *alpha2) {}
*/
static int call_crda(const char *alpha2)
{
- if (!is_world_regdom((char *) alpha2))
- pr_info("Calling CRDA for country: %c%c\n",
- alpha2[0], alpha2[1]);
- else
- pr_info("Calling CRDA to update world regulatory domain\n");
-
- /* query internal regulatory database (if it exists) */
- reg_regdb_query(alpha2);
-
- return kobject_uevent(®_pdev->dev.kobj, KOBJ_CHANGE);
+ /* query internal regulatory database/crda depending on config */
+ return reg_regdb_query(alpha2);
}
static enum reg_request_treatment
On Sat, Jan 18, 2014 at 12:25 AM, Johannes Berg
<[email protected]> wrote:
> On Fri, 2014-01-17 at 03:32 +0530, Chaitanya T K wrote:
>> When using internal regulatory data base kconfig option
>> do not try to call crda, this complely defeats the
>> purposes of enabling internal regulatory db.
>
> In general I don't think you can just break user visible behaviour this
> way.
Well, users are habituated the wrong way :-). Any new user seeing the
internal_regdb might just forget about CRDA and face problems, in case
the DB is different in both.
How should we go about it, then?
On Sat, 2014-01-18 at 00:22 +0530, Krishna Chaitanya wrote:
> On Fri, Jan 17, 2014 at 3:32 AM, Chaitanya T K <[email protected]> wrote:
> >
> > When using internal regulatory data base kconfig option
> > do not try to call crda, this complely defeats the
> > purposes of enabling internal regulatory db.
> >
> > Signed-off-by: Chaitanya T K <[email protected]>
> >
> > ---
> > net/wireless/reg.c | 29 +++++++++++++++--------------
> > 1 file changed, 15 insertions(+), 14 deletions(-)
> >
> > diff --git a/net/wireless/reg.c b/net/wireless/reg.c
> > index 7d20d84..1cb0f99 100644
> > --- a/net/wireless/reg.c
> > +++ b/net/wireless/reg.c
> > @@ -454,16 +454,16 @@ static void reg_regdb_search(struct work_struct *work)
> >
> > static DECLARE_WORK(reg_regdb_work, reg_regdb_search);
> >
> > -static void reg_regdb_query(const char *alpha2)
> > +static int reg_regdb_query(const char *alpha2)
> > {
> > struct reg_regdb_search_request *request;
> >
> > if (!alpha2)
> > - return;
> > + return -1;
> >
> > request = kzalloc(sizeof(struct reg_regdb_search_request), GFP_KERNEL);
> > if (!request)
> > - return;
> > + return -1;
> >
> I guess the return values must be 1 so that in case of failure
> we ignore the reg update. Will wait for sometime for other
> comments and send V2.
-1 is -EPERM anyway, which doesn't seem appropriate.
johannes
On Sat, Jan 18, 2014 at 12:33 AM, Krishna Chaitanya
<[email protected]> wrote:
> On Sat, Jan 18, 2014 at 12:25 AM, Johannes Berg
> <[email protected]> wrote:
>> On Fri, 2014-01-17 at 03:32 +0530, Chaitanya T K wrote:
>>> When using internal regulatory data base kconfig option
>>> do not try to call crda, this complely defeats the
>>> purposes of enabling internal regulatory db.
>>
>> In general I don't think you can just break user visible behaviour this
>> way.
> Well, users are habituated the wrong way :-). Any new user seeing the
> internal_regdb might just forget about CRDA and face problems, in case
> the DB is different in both.
>
> How should we go about it, then?
When user enabled internal_regdb target, if he forgets to copy the db.txt
we throw a warn_on and continue with WORLD regdom. So user has to copy
the db.txt and we will read rules from there.This all happens independent of
CRDA running/not in userspace. So i dont see any visible user impact
if at all someone decides to enabled internal_regdb even after seeing
all the warnings in place (kconfig, doc ..) :-).
One more bug in the below check:
if (call_crda(lr->alpha2))
return REG_REQ_IGNORE;
If call_crda fails (kobject_event return negative no) then also
we return REG_REQ_OK. It should be
if (call_crda(lr->alpha2) < 0)
return REG_REQ_IGNORE;
On Fri, 2014-01-17 at 03:32 +0530, Chaitanya T K wrote:
> When using internal regulatory data base kconfig option
> do not try to call crda, this complely defeats the
> purposes of enabling internal regulatory db.
In general I don't think you can just break user visible behaviour this
way.
johannes
On Fri, Jan 17, 2014 at 3:32 AM, Chaitanya T K <[email protected]> wrote:
>
> When using internal regulatory data base kconfig option
> do not try to call crda, this complely defeats the
> purposes of enabling internal regulatory db.
>
> Signed-off-by: Chaitanya T K <[email protected]>
>
> ---
> net/wireless/reg.c | 29 +++++++++++++++--------------
> 1 file changed, 15 insertions(+), 14 deletions(-)
>
> diff --git a/net/wireless/reg.c b/net/wireless/reg.c
> index 7d20d84..1cb0f99 100644
> --- a/net/wireless/reg.c
> +++ b/net/wireless/reg.c
> @@ -454,16 +454,16 @@ static void reg_regdb_search(struct work_struct *work)
>
> static DECLARE_WORK(reg_regdb_work, reg_regdb_search);
>
> -static void reg_regdb_query(const char *alpha2)
> +static int reg_regdb_query(const char *alpha2)
> {
> struct reg_regdb_search_request *request;
>
> if (!alpha2)
> - return;
> + return -1;
>
> request = kzalloc(sizeof(struct reg_regdb_search_request), GFP_KERNEL);
> if (!request)
> - return;
> + return -1;
>
I guess the return values must be 1 so that in case of failure
we ignore the reg update. Will wait for sometime for other
comments and send V2.