diff mbox series

iw: strip NLA_FLAGS from printed bands

Message ID 20231221222842.1310957-1-matthias.may@westermo.com
State New
Headers show
Series iw: strip NLA_FLAGS from printed bands | expand

Commit Message

Matthias May Dec. 21, 2023, 10:29 p.m. UTC 
nl_band->nla_type might have NLA_F_NESTED (0x8000) set,
causing 'Bands' to be displayed with an according offset
when used directly.

Use the nla_type() macro instead to strip flags for
printing.

Signed-off-by: Zefir Kurtisi <zefir.kurtisi@westermo.com>
Signed-off-by: Matthias May <matthias.may@westermo.com>
---
 info.c | 2 +-
 phy.c  | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

Comments

Johannes Berg Dec. 31, 2023, 9:59 p.m. UTC | #1 
On Thu, 2023-12-21 at 22:29 +0000, Matthias May wrote:
> nl_band->nla_type might have NLA_F_NESTED (0x8000) set,

I'm curious - how does that happen?

Not that the change looks wrong per se, but ... I don't think this can
happen?

johannes
Matthias May Jan. 5, 2024, 8:35 p.m. UTC | #2 
On 31/12/2023 22:59, Johannes Berg wrote:
> On Thu, 2023-12-21 at 22:29 +0000, Matthias May wrote:
>> nl_band->nla_type might have NLA_F_NESTED (0x8000) set,
> I'm curious - how does that happen?
>
> Not that the change looks wrong per se, but ... I don't think this can
> happen?
>
> johannes
>
Hi Johannes

We saw that happen on one of our platforms where we are forced to use an 
out-of-tree driver by QCA (SPF12.2).
If you think this is something that can not happen with upstream 
drivers, feel free to ignore this patch.

BR
Matthias
Johannes Berg Jan. 5, 2024, 8:36 p.m. UTC | #3 
On Fri, 2024-01-05 at 21:35 +0100, Matthias May wrote:
> We saw that happen on one of our platforms where we are forced to use an 
> out-of-tree driver by QCA (SPF12.2).
> If you think this is something that can not happen with upstream 
> drivers, feel free to ignore this patch.
> 

It should be independent of drivers, so they must have modified cfg80211
to explicitly use the nla_put_nested() instead of
nla_put_nested_noflag()? It's just ... weird.

johannes
Jeff Johnson Jan. 6, 2024, 12:56 a.m. UTC | #4 
On 1/5/2024 12:36 PM, Johannes Berg wrote:
> On Fri, 2024-01-05 at 21:35 +0100, Matthias May wrote:
>> We saw that happen on one of our platforms where we are forced to use an 
>> out-of-tree driver by QCA (SPF12.2).
>> If you think this is something that can not happen with upstream 
>> drivers, feel free to ignore this patch.
>>
> 
> It should be independent of drivers, so they must have modified cfg80211
> to explicitly use the nla_put_nested() instead of
> nla_put_nested_noflag()? It's just ... weird.

Indeed. Matthias do you have a QCA support contact? You should open a
ticket with them to address this issue.

/jeff
diff mbox series

Patch

diff --git a/info.c b/info.c
index 317e7a3..5334436 100644
--- a/info.c
+++ b/info.c
@@ -354,7 +354,7 @@  static int print_phy_handler(struct nl_msg *msg, void *arg)
 	if (tb_msg[NL80211_ATTR_WIPHY_BANDS]) {
 		nla_for_each_nested(nl_band, tb_msg[NL80211_ATTR_WIPHY_BANDS], rem_band) {
 			if (last_band != nl_band->nla_type) {
-				printf("\tBand %d:\n", nl_band->nla_type + 1);
+				printf("\tBand %d:\n", nla_type(nl_band) + 1);
 				band_had_freq = false;
 			}
 			last_band = nl_band->nla_type;
diff --git a/phy.c b/phy.c
index 4722125..ebd7289 100644
--- a/phy.c
+++ b/phy.c
@@ -52,7 +52,7 @@  static int print_channels_handler(struct nl_msg *msg, void *arg)
 	if (tb_msg[NL80211_ATTR_WIPHY_BANDS]) {
 		nla_for_each_nested(nl_band, tb_msg[NL80211_ATTR_WIPHY_BANDS], rem_band) {
 			if (ctx->last_band != nl_band->nla_type) {
-				printf("Band %d:\n", nl_band->nla_type + 1);
+				printf("Band %d:\n", nla_type(nl_band) + 1);
 				ctx->width_40 = false;
 				ctx->width_80 = false;
 				ctx->width_160 = false;