ath5k: Update PCU code
PATCH 1/1 ath5k: Update PCU code
* In set_opmode don't handle antenna settings and preserve other STA_ID1 settings (shouldn't matter because we call it during reset but it makes things cleaner) Also set properly AP/ADHOC indicator flag on CFG while setting AP/ADHOC modes and always enable key search mode.
* Properly set BSSID Mask during reset (cache it and reuse it durring set_associd)
* Update beacon_init to flush pending BMISS interrupts and handle setting of adhoc beacon ATIM policy flag for ad-hoc mode. Also set TSF to 0 to start TSF increment on AP mode. We need to handle sleep timers for AR5212 there + add support for PCF.
* Properly clean MIC key from keytable when TKIP is used (Bob is working on set_key function etc so i leave it for now).
Tested on AR5212 (Hainan) and AR5413 and works fine
Changes-Licensed-under: ISC
---
drivers/net/wireless/ath5k/ath5k.h +2 -1
@@ -1093,10 +1093,11 @@ struct ath5k_hw {u8 ah_sta_id[ETH_ALEN ]; -/* Current BSSID we are trying to assoc to / creating. +/* Current BSSID we are trying to assoc to / create. * This is passed by mac80211 on config_interface() and cached here for * use in resets */ u8 ah_bssid[ETH_ALEN ]; +u8 ah_bssid_mask[ETH_ALEN ];u32 ah_gpio[AR5K_MAX_GPIO ];int ah_gpio_npins;
drivers/net/wireless/ath5k/pcu.c +88 -20
203 lines not on this page · raw message
drivers/net/wireless/ath5k/reg.h +9 -3
@@ -1114,14 +1114,16 @@#define AR5K_PCU_MAX 0x8fff/* -* First station id register (MAC address in lower 32 bits) +* First station id register (Lower 32 bits of MAC address) */ -#define AR5K_STA_ID0 0x8000 +#define AR5K_STA_ID0 0x8000 +#define AR5K_STA_ID0_ARRD_L32 0xffffffff/* -* Second station id register (MAC address in upper 16 bits) +* Second station id register (Upper 16 bits of MAC address + PCU settings) */ #define AR5K_STA_ID1 0x8004/* Register Address */ +#define AR5K_STA_ID1_ADDR_U16 0x0000ffff/* Upper 16 bits of MAC addres */ #define AR5K_STA_ID1_AP 0x00010000/* Set AP mode */ #define AR5K_STA_ID1_ADHOC 0x00020000/* Set Ad-Hoc mode */ #define AR5K_STA_ID1_PWR_SV 0x00040000/* Power save reporting */
@@ -1811,6 +1813,10 @@#define AR5K_KEYTABLE_MAC1 (_n) AR5K_KEYTABLE_OFF(_n, 7)#define AR5K_KEYTABLE_VALID 0x00008000 +/* If key type is TKIP and MIC is enabled +* MIC key goes in offset entry + 64 */ +#define AR5K_KEYTABLE_MIC_OFFSET 64 +/* WEP 40-bit = 40-bit entered key + 24 bit IV = 64-bit * WEP 104-bit = 104-bit entered key + 24-bit IV = 128-bit * WEP 128-bit = 128-bit entered key + 24 bit IV = 152-bit
* In set_opmode don't handle antenna settings and preserve other STA_ID1 settings (shouldn't matter because we call it during reset but it makes things cleaner) Also set properly AP/ADHOC indicator flag on CFG while setting AP/ADHOC modes and always enable key search mode.
* Properly set BSSID Mask during reset (cache it and reuse it durring set_associd)
* Update beacon_init to flush pending BMISS interrupts and handle setting of adhoc beacon ATIM policy flag for ad-hoc mode. Also set TSF to 0 to start TSF increment on AP mode. We need to handle sleep timers for AR5212 there + add support for PCF.
* Properly clean MIC key from keytable when TKIP is used (Bob is working on set_key function etc so i leave it for now).
Tested on AR5212 (Hainan) and AR5413 and works fine
v2 Set PISR on AR5211+ and ISR on AR5210, got to sleep more ;-)
Changes-Licensed-under: ISC
---
drivers/net/wireless/ath5k/ath5k.h +2 -1
@@ -1093,10 +1093,11 @@ struct ath5k_hw {u8 ah_sta_id[ETH_ALEN ]; -/* Current BSSID we are trying to assoc to / creating. +/* Current BSSID we are trying to assoc to / create. * This is passed by mac80211 on config_interface() and cached here for * use in resets */ u8 ah_bssid[ETH_ALEN ]; +u8 ah_bssid_mask[ETH_ALEN ];u32 ah_gpio[AR5K_MAX_GPIO ];int ah_gpio_npins;
drivers/net/wireless/ath5k/pcu.c +91 -20
206 lines not on this page · raw message
drivers/net/wireless/ath5k/reg.h +9 -3
@@ -1114,14 +1114,16 @@#define AR5K_PCU_MAX 0x8fff/* -* First station id register (MAC address in lower 32 bits) +* First station id register (Lower 32 bits of MAC address) */ -#define AR5K_STA_ID0 0x8000 +#define AR5K_STA_ID0 0x8000 +#define AR5K_STA_ID0_ARRD_L32 0xffffffff/* -* Second station id register (MAC address in upper 16 bits) +* Second station id register (Upper 16 bits of MAC address + PCU settings) */ #define AR5K_STA_ID1 0x8004/* Register Address */ +#define AR5K_STA_ID1_ADDR_U16 0x0000ffff/* Upper 16 bits of MAC addres */ #define AR5K_STA_ID1_AP 0x00010000/* Set AP mode */ #define AR5K_STA_ID1_ADHOC 0x00020000/* Set Ad-Hoc mode */ #define AR5K_STA_ID1_PWR_SV 0x00040000/* Power save reporting */
@@ -1811,6 +1813,10 @@#define AR5K_KEYTABLE_MAC1 (_n) AR5K_KEYTABLE_OFF(_n, 7)#define AR5K_KEYTABLE_VALID 0x00008000 +/* If key type is TKIP and MIC is enabled +* MIC key goes in offset entry + 64 */ +#define AR5K_KEYTABLE_MIC_OFFSET 64 +/* WEP 40-bit = 40-bit entered key + 24 bit IV = 64-bit * WEP 104-bit = 104-bit entered key + 24-bit IV = 128-bit * WEP 128-bit = 128-bit entered key + 24 bit IV = 152-bit
Quoted 2 lines · Nick Kossifidis
* Properly clean MIC key from keytable when TKIP is used (Bob is working on
set_keyfunction etc so i leave it for now).
Heh, I was just prepping patches to send out. I had basically the same thing for _reset_key too so I'll just rebase mine on top of this.
Signature · 1 line
Bob Copeland %% www.bobcopeland.com
2008/10/29 Bob Copeland <[email protected]>:
Quoted 7 lines · Nick Kossifidis
Quoted 2 lines
* Properly clean MIC key from keytable when TKIP is used (Bob is working on
set_keyfunction etc so i leave it for now).Heh, I was just prepping patches to send out. I had basically the same thing for
_reset_keytoo so I'll just rebase mine on top of this.
ACK, sorry for the bad coordination ;-(
Signature · 3 lines
GPG ID: 0xD21DB2DB As you read this post global entropy rises. Have Fun ;-) Nick
Quoted 10 lines · Nick Kossifidis
2008/10/29 Bob Copeland <[email protected]>:
Quoted 7 lines · Nick Kossifidis
Quoted 2 lines
* Properly clean MIC key from keytable when TKIP is used (Bob is working on
set_keyfunction etc so i leave it for now).Heh, I was just prepping patches to send out. I had basically the same thing for
_reset_keytoo so I'll just rebase mine on top of this.ACK, sorry for the bad coordination ;-(
No worries, it was easy enough to drop that hunk. But I do think this:
drivers/net/wireless/ath5k/pcu.c int ath5k_hw_beaconq_finish(struct ath5k_hw *ah, unsigned long phys_addr) · line 977
Quoted 9 lines
drivers/net/wireless/ath5k/pcu.c +9
-0+/* Reset associated MIC entry if TKIP +* is enabled located at offset (entry + 64) */ +if (type ==AR5K_KEYTABLE_TYPE_TKIP ) { + entry = entry +AR5K_KEYTABLE_MIC_OFFSET ; +AR5K_ASSERT_ENTRY (entry,AR5K_KEYTABLE_SIZE ); +for (i =0 ; i <AR5K_KEYCACHE_SIZE /2 ; i++) +ath5k_hw_reg_write (ah,0 ,AR5K_KEYTABLE_OFF (entry, i)); + } +
...should move a couple of lines down or use a temporary for the micentry slot; otherwise we are setting keytype to NULL in the mic entry and not the original key slot.
The only other difference in mine was I went ahead and reset the whole line instead of just the first 4 words... since nothing goes in that half of the cache line that's probably wasted effort, and the HAL does the above so it should be fine. I just wasn't sure if power-up state clears the keytype properly.
Signature · 1 line
Bob Copeland %% www.bobcopeland.com
2008/10/29 Bob Copeland <[email protected]>:
Quoted 24 lines · Nick Kossifidis
Quoted 10 lines
2008/10/29 Bob Copeland <[email protected]>:
Quoted 7 lines · Nick Kossifidis
Quoted 2 lines
* Properly clean MIC key from keytable when TKIP is used (Bob is working on
set_keyfunction etc so i leave it for now).Heh, I was just prepping patches to send out. I had basically the same thing for
_reset_keytoo so I'll just rebase mine on top of this.ACK, sorry for the bad coordination ;-(
No worries, it was easy enough to drop that hunk. But I do think this:
drivers/net/wireless/ath5k/pcu.c int ath5k_hw_beaconq_finish(struct ath5k_hw *ah, unsigned long phys_addr) · line 977
Quoted 9 lines
drivers/net/wireless/ath5k/pcu.c +9
-0+/* Reset associated MIC entry if TKIP +* is enabled located at offset (entry + 64) */ +if (type ==AR5K_KEYTABLE_TYPE_TKIP ) { + entry = entry +AR5K_KEYTABLE_MIC_OFFSET ; +AR5K_ASSERT_ENTRY (entry,AR5K_KEYTABLE_SIZE ); +for (i =0 ; i <AR5K_KEYCACHE_SIZE /2 ; i++) +ath5k_hw_reg_write (ah,0 ,AR5K_KEYTABLE_OFF (entry, i)); + } +
Quoted 4 lines
...should move a couple of lines down or use a temporary for the micentry slot; otherwise we are setting keytype to NULL in the mic entry and not the original key slot.
Yup you are right, we should set both to NULL so a temp micentry var would do the trick.
Should i resend this one or you 'll resend yours ?
Signature · 3 lines
GPG ID: 0xD21DB2DB As you read this post global entropy rises. Have Fun ;-) Nick
Quoted 5 lines · Nick Kossifidis
2008/10/29 Bob Copeland <[email protected]>:
Yup you are right, we should set both to NULL so a temp micentry var would do the trick.Should i resend this one or you 'll resend yours ?
Well I just decided to put this in a different patch. Then I saw the 5211 check.. we already read the type without such a check and the other tkip stuff may well all blow up on such hardware. So not sure what to do about that case.
Anyway feel free to just roll this up into your original patch, or to write it differently given the above.
From: Bob Copeland <[email protected]>
Subject: [PATCH] ath5k: fix keytable type buglet in ath5k_hw_reset_key
Be sure we clear out both the mic (if applicable) and the encryption key type.
--- drivers/net/wireless/ath5k/pcu.c | 15 +++++++++++---- 1 files changed, 11 insertions(+), 4 deletions(-)
drivers/net/wireless/ath5k/pcu.c +11 -4
@@ -960,6 +960,7 @@ int ath5k_hw_beaconq_finish(struct ath5k_hw *ah, unsigned long phys_addr)int ath5k_hw_reset_key (struct ath5k_hw *ah,u16 entry) {unsigned int i, type; +u16 micentry = entry +AR5K_KEYTABLE_MIC_OFFSET ;ATH5K_TRACE (ah->ah_sc);AR5K_ASSERT_ENTRY (entry,AR5K_KEYTABLE_SIZE );
@@ -972,10 +973,10 @@ int ath5k_hw_reset_key(struct ath5k_hw *ah, u16 entry)/* Reset associated MIC entry if TKIP * is enabled located at offset (entry + 64) */ if (type ==AR5K_KEYTABLE_TYPE_TKIP ) { - entry = entry +AR5K_KEYTABLE_MIC_OFFSET ; -AR5K_ASSERT_ENTRY (entry,AR5K_KEYTABLE_SIZE ); +AR5K_ASSERT_ENTRY (micentry,AR5K_KEYTABLE_SIZE );for (i =0 ; i <AR5K_KEYCACHE_SIZE /2 ; i++) -ath5k_hw_reg_write (ah,0 ,AR5K_KEYTABLE_OFF (entry, i)); +ath5k_hw_reg_write (ah,0 , +AR5K_KEYTABLE_OFF (micentry, i)); }/*
@@ -987,10 +988,16 @@ int ath5k_hw_reset_key(struct ath5k_hw *ah, u16 entry)* Note2: Windows driver (ndiswrapper) sets this to * 0x00000714 instead of 0x00000007 */ -if (ah->ah_version >AR5K_AR5211 ) +if (ah->ah_version >AR5K_AR5211 ) {ath5k_hw_reg_write (ah,AR5K_KEYTABLE_TYPE_NULL ,AR5K_KEYTABLE_TYPE (entry)); +if (type ==AR5K_KEYTABLE_TYPE_TKIP ) { +ath5k_hw_reg_write (ah,AR5K_KEYTABLE_TYPE_NULL , +AR5K_KEYTABLE_TYPE (micentry)); + } + } +return 0 ; }
Signature · 6 lines
1.5.4.2.182.gb3092 -- Bob Copeland %% www.bobcopeland.com
2008/10/30 Bob Copeland <[email protected]>:
Quoted 60 lines · Nick Kossifidis
Quoted 5 lines
2008/10/29 Bob Copeland <[email protected]>:
Yup you are right, we should set both to NULL so a temp micentry var would do the trick.Should i resend this one or you 'll resend yours ?
Well I just decided to put this in a different patch. Then I saw the 5211 check.. we already read the type without such a check and the other tkip stuff may well all blow up on such hardware. So not sure what to do about that case.
Anyway feel free to just roll this up into your original patch, or to write it differently given the above.
From: Bob Copeland <[email protected]>
Subject: [PATCH] ath5k: fix keytable type buglet inath5k_hw_reset_keyBe sure we clear out both the mic (if applicable) and the encryption key type.
Signed-off-by Bob Copeland <[email protected]>--- drivers/net/wireless/ath5k/pcu.c | 15 +++++++++++---- 1 files changed, 11 insertions(+), 4 deletions(-)drivers/net/wireless/ath5k/pcu.c +5
-4@@ -960,6 +960,7 @@ int ath5k_hw_beaconq_finish(struct ath5k_hw *ah, unsigned long phys_addr)int ath5k_hw_reset_key (struct ath5k_hw *ah,u16 entry) {unsigned int i, type; +u16 micentry = entry +AR5K_KEYTABLE_MIC_OFFSET ;ATH5K_TRACE (ah->ah_sc);AR5K_ASSERT_ENTRY (entry,AR5K_KEYTABLE_SIZE );@@ -972,10 +973,10 @@ int ath5k_hw_reset_key(struct ath5k_hw *ah, u16 entry)/* Reset associated MIC entry if TKIP * is enabled located at offset (entry + 64) */ if (type ==AR5K_KEYTABLE_TYPE_TKIP ) { - entry = entry +AR5K_KEYTABLE_MIC_OFFSET ; -AR5K_ASSERT_ENTRY (entry,AR5K_KEYTABLE_SIZE ); +AR5K_ASSERT_ENTRY (micentry,AR5K_KEYTABLE_SIZE );for (i =0 ; i <AR5K_KEYCACHE_SIZE /2 ; i++) -ath5k_hw_reg_write (ah,0 ,AR5K_KEYTABLE_OFF (entry, i)); +ath5k_hw_reg_write (ah,0 , +AR5K_KEYTABLE_OFF (micentry, i)); }/* @@ -987,10 +988,16 @@ int ath5k_hw_reset_key(struct ath5k_hw *ah, u16 entry)* Note2: Windows driver (ndiswrapper) sets this to * 0x00000714 instead of 0x00000007 */ -if (ah->ah_version >AR5K_AR5211 ) +if (ah->ah_version >AR5K_AR5211 ) {ath5k_hw_reg_write (ah,AR5K_KEYTABLE_TYPE_NULL ,AR5K_KEYTABLE_TYPE (entry));
drivers/net/wireless/ath5k/pcu.c int ath5k_hw_reset_key(struct ath5k_hw *ah, u16 entry) · line 1003
Quoted 10 lines
drivers/net/wireless/ath5k/pcu.c +6
-0+if (type ==AR5K_KEYTABLE_TYPE_TKIP ) { +ath5k_hw_reg_write (ah,AR5K_KEYTABLE_TYPE_NULL , +AR5K_KEYTABLE_TYPE (micentry)); + } + } +return 0 ; }
Let's apply this for now and i'll take a look on what's going on for AR5211 (i don't know if TKIP/MIC is supported or not right now, i'll have to test this). Let's finish with AR5212 first (i also have a patch for QCU updates pending) since many people want AR5212 hw encryption support ;-)
Signature · 3 lines
GPG ID: 0xD21DB2DB As you read this post global entropy rises. Have Fun ;-) Nick