Skip to contentSkip to conversation

staging: rtl8723au: core: endianness issues

PATCH In review David Decotigny 3 participants 5 messages Jun 8, 2015 – Jun 25, 2015 Staging

lore
Share & tools
lore
#0 lore raw

v2 cover staging: rtl8723au: core: endianness issues

The code shows a couple inconsistencies (described in commit descriptions) which would not be an issue on little-endian cpus, but could cause breakage on non-LE cpus. Note: I could not test on real hardware, these patches created based on sparse reports.

Hostory:
- resending the same patches to correct recipients, only changed commit descriptions (credits to Dan Carpenter)

############################################
# Patch Set Summary:

David Decotigny (2):
staging: rtl8723au: core: avoid bitwise arithmetic with forced
endianness
staging: rtl8723au: core: remove redundant endianness conversion

 drivers/staging/rtl8723au/core/rtw_mlme_ext.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)
Signature · 1 line
2.2.0.rc0.207.ga3a616c
#1 lore raw

PATCH v2 1/2 staging: rtl8723au: core: avoid bitwise arithmetic with forced endianness

This fixes bitwise arithmetic performed on the host on a variable previously converted to little-endian, and subsequently converted again to little-endian:
- issue_action_BA23a() called with "status" crafted in host byte order
- "status" converted to LE
- bitwise arithmetic on the (LE) "status", performed with masks and shifts in host byte order
- result converted to LE (again) and stored in device structure

Sparse warning addressed by this patch:
drivers/staging/rtl8723au/core/rtw_mlme_ext.c:3806:16: warning: incorrect type in assignment (different base types)
drivers/staging/rtl8723au/core/rtw_mlme_ext.c:3806:16: expected unsigned short [unsigned] status
drivers/staging/rtl8723au/core/rtw_mlme_ext.c:3806:16: got restricted __le16 [usertype] <noident>

Additional notes: initial cpu_to_le16 was introduced by kernel bulk
commit 5e93f3520 "staging: r8723au: Add source files for new driver
- part 1", initially from github according to commit description. On github, this traces back to another bulk commit: 2896bda04353 "Add new files in core directory", which is the 1st version of the driver.

---
 drivers/staging/rtl8723au/core/rtw_mlme_ext.c | 2 --
 1 file changed, 2 deletions(-)
drivers/staging/rtl8723au/core/rtw_mlme_ext.c +0 -2
index 196beaf..7c3b5dd 100644
@@ -3803,8 +3803,6 @@ void issue_action_BA23a(struct rtw_adapter *padapter,
 
 	pattrib->pktlen = sizeof(struct ieee80211_hdr_3addr) + 1;
 
-	status = cpu_to_le16(status);
-
 	switch (action) {
 	case WLAN_ACTION_ADDBA_REQ:
 		pattrib->pktlen += sizeof(mgmt->u.action.u.addba_req);
Signature · 1 line
2.2.0.rc0.207.ga3a616c
#2 lore raw

PATCH v2 2/2 staging: rtl8723au: core: remove redundant endianness conversion

In reply to David Decotigny

Source and destination have the same little-endian annotation: this patch removes incorrect byte-swap on non-LE cpus.

This addresses the following sparse warning:
drivers/staging/rtl8723au/core/rtw_mlme_ext.c:3911:56: warning: incorrect type in argument 1 (different base types)
drivers/staging/rtl8723au/core/rtw_mlme_ext.c:3911:56: expected unsigned short [unsigned] [usertype] val
drivers/staging/rtl8723au/core/rtw_mlme_ext.c:3911:56: got restricted __le16 [usertype] BA_timeout_value

---
 drivers/staging/rtl8723au/core/rtw_mlme_ext.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/staging/rtl8723au/core/rtw_mlme_ext.c +2 -2
index 7c3b5dd..142f214 100644
@@ -3906,8 +3906,8 @@ void issue_action_BA23a(struct rtw_adapter *padapter,
 		put_unaligned_le16(BA_para_set,
 				   &mgmt->u.action.u.addba_resp.capab);
 
-		put_unaligned_le16(pmlmeinfo->ADDBA_req.BA_timeout_value,
-				   &mgmt->u.action.u.addba_resp.timeout);
+		mgmt->u.action.u.addba_resp.timeout
+			= pmlmeinfo->ADDBA_req.BA_timeout_value;
 
 		pattrib->pktlen += 8;
 		break;
Signature · 1 line
2.2.0.rc0.207.ga3a616c
#4 lore raw
In reply to David Decotigny
Quoted 12 lines · David Decotigny

The code shows a couple inconsistencies (described in commit descriptions) which would not be an issue on little-endian cpus, but could cause breakage on non-LE cpus. Note: I could not test on real hardware, these patches created based on sparse reports.

Hostory:
- resending the same patches to correct recipients, only changed commit descriptions (credits to Dan Carpenter)

############################################
# Patch Set Summary:

Quoted 7 lines

David Decotigny (2):
staging: rtl8723au: core: avoid bitwise arithmetic with forced
endianness
staging: rtl8723au: core: remove redundant endianness conversion

 drivers/staging/rtl8723au/core/rtw_mlme_ext.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

Looks fine to me, however if you fiddle with this same value twice, wouldn't it be better to do it in one patch?

Cheers,
Jes