ath5k: AHB port. Fix reset and interrupts for AHB type of devices.
A message that is not in the archive (sent off-list or not yet received): 5575430.811289486702581.JavaMail.wlan@CHBU500181
PATCH 9/9 ath5k: AHB port. Fix reset and interrupts for AHB type of devices.
On WiSoc we cannot access mac register before it is resetted. Otherwise it will crash hardware.
--- drivers/net/wireless/ath/ath5k/base.c | 7 ++- drivers/net/wireless/ath/ath5k/reset.c | 113 ++++++++++++++++++++++++------- 2 files changed, 93 insertions(+), 27 deletions(-)
drivers/net/wireless/ath/ath5k/base.c +6 -1
@@ -2175,7 +2175,8 @@ ath5k_intr(int irq, void *dev_id)unsigned int counter =1000 ;if (unlikely (test_bit (ATH_STAT_INVALID , sc->status) || - !ath5k_hw_is_intr_pending (ah))) + ((ath5k_get_bus_type (ah) !=ATH_AHB ) && + !ath5k_hw_is_intr_pending (ah))))return IRQ_NONE ; do {
@@ -2241,6 +2242,10 @@ ath5k_intr(int irq, void *dev_id)tasklet_schedule (&sc->rf_kill.toggleq); } + +if (ath5k_get_bus_type (ah) ==ATH_AHB ) +break ; + }while (ath5k_hw_is_intr_pending (ah) && --counter >0 );if (unlikely (!counter))
drivers/net/wireless/ath/ath5k/reset.c +87 -26
178 lines not on this page · raw message
Signature · 1 line
1.7.1
Quoted 16 lines · Wojciech Dubowik
On WiSoc we cannot access mac register before it is resetted. Otherwise it will crash hardware.
Signed-off-by Wojciech Dubowik <[email protected]>---
drivers/net/wireless/ath/ath5k/base.c | 7 ++-drivers/net/wireless/ath/ath5k/reset.c| 113 ++++++++++++++++++++++++-------2 files changed, 93 insertions(+), 27 deletions(-)drivers/net/wireless/ath/ath5k/base.c +0
-0@@ -2175,7 +2175,8 @@ ath5k_intr(int irq, void *dev_id) unsigned int counter = 1000;
Quoted 6 lines
if (unlikely(test_bit(ATH_STAT_INVALID,
sc->status) ||
- !ath5k_hw_is_intr_pending(ah)))
+ ((ath5k_get_bus_type(ah)!=ATH_AHB) &&
+ !ath5k_hw_is_intr_pending(ah)))) returnIRQ_NONE;
Hrm, we really don't want to sprinkle these special cases all around the driver. Perhaps it'd be better to make sure ATH_STAT_INVALID isn't set until the soc is ready. And, in other cases, using capability bits or something? Bus type seems a bit crude.
drivers/net/wireless/ath/ath5k/base.c ath5k_intr(int irq, void *dev_id) · line 2249
Quoted 9 lines
drivers/net/wireless/ath/ath5k/base.c +4
-0@@ -2241,6 +2242,10 @@ ath5k_intr(int irq, void *dev_id) tasklet_schedule(&sc->rf_kill.toggleq); } + + if(ath5k_get_bus_type(ah) == ATH_AHB) + break; + } while (ath5k_hw_is_intr_pending(ah) && --counter > 0);
Why?
drivers/net/wireless/ath/ath5k/base.c ath5k_intr(int irq, void *dev_id)
Quoted 4 lines
drivers/net/wireless/ath/ath5k/base.c +1
-1static int ath5k_hw_nic_reset(struct ath5k_hw *ah, u32 val) { - int ret; + int ret = 0;
In general, I dislike assigning ret until the last possible moment, because leaving it unassigned sometimes catches errors like:
int ret;
+ if (new_boolean_test)
+ goto out;
ret = xyz();
out:
return ret;
In this particular case it doesn't matter that much, but the early assignment shouldn't really be needed.
Quoted 40 lines
/*
* Reset configuration register (for hw byte-swap). Note that thisdrivers/net/wireless/ath/ath5k/base.c +10
-10@@ -334,6 +377,9 @@ int ath5k_hw_on_hold(struct ath5k_hw *ah) u32 bus_flags; int ret; + if(ath5k_get_bus_type(ah) == ATH_AHB) + return 0; + /* Make sure device is awake */ ret = ath5k_hw_set_power(ah, AR5K_PM_AWAKE, true, 0); if (ret) {@@ -390,22 +436,30 @@ int ath5k_hw_nic_wakeup(struct ath5k_hw *ah, int flags, bool initial) mode = 0; clock = 0; - /* Wakeup the device */ - ret = ath5k_hw_set_power(ah, AR5K_PM_AWAKE, true, 0); - if (ret) { - ATH5K_ERR(ah->ah_sc, "failed to wakeup the MAC Chip\n"); - return ret; + if(ath5k_get_bus_type(ah) == ATH_AHB && !initial){ + /* Wakeup the device */ + ret = ath5k_hw_set_power(ah, AR5K_PM_AWAKE, true, 0); + if (ret) { + ATH5K_ERR(ah->ah_sc, "failed to wakeup the MAC Chip\n"); + return ret; + } } /** Put chipset on warm reset... * - * Note: putting PCI core on warm reset on PCI-E cards - * results card to hang and always return 0xffff... so - * we ingore that flag for PCI-E cards. On PCI cards - * this flag gets cleared after 64 PCI clocks.*/ - bus_flags = (pdev && pdev->is_pcie) ? 0 : AR5K_RESET_CTL_PCI;
drivers/net/wireless/ath/ath5k/base.c int ath5k_hw_nic_wakeup(struct ath5k_hw *ah, int flags, bool initial) · line 461
Quoted 10 lines
drivers/net/wireless/ath/ath5k/base.c +10
-0+ if(ath5k_get_bus_type(ah) == ATH_AHB){ + /* Reset MAC on WiSoc devices */ + bus_flags = (initial) ? AR5K_RESET_CTL_MAC : 0; + } else { + /* Note: putting PCI core on warm reset on PCI-E cards + * results card to hang and always return 0xffff... so + * we ingore that flag for PCI-E cards. On PCI cards + * this flag gets cleared after 64 PCI clocks. + */ + bus_flags = (pdev && pdev->is_pcie) ? 0 : AR5K_RESET_CTL_PCI;
Maybe ath5k_get_bus_flags or something here.
+ }
drivers/net/wireless/ath/ath5k/base.c static void ath5k_hw_set_sleep_clock(struct ath5k_hw *ah, bool enable) · line 595
Quoted 10 lines
drivers/net/wireless/ath/ath5k/base.c +0
-0if (ah->ah_version == AR5K_AR5210) {ret = ath5k_hw_nic_reset(ah,
AR5K_RESET_CTL_PCU|drivers/net/wireless/ath/ath5k/base.c +3
-0@@ -536,6 +590,9 @@ static void ath5k_hw_set_sleep_clock(struct ath5k_hw *ah, bool enable) struct ath5k_eeprom_info *ee = &ah->ah_capabilities.cap_eeprom; u32 scal, spending, usec32; + if(ath5k_get_bus_type(ah) == ATH_AHB) + enable = false; +
Do you need this hunk? We never enable the sleep clock for AP mode.
Signature · 1 line
Bob Copeland %% www.bobcopeland.com
Quoted 32 lines · Bob Copeland
Quoted 14 lines · Wojciech Dubowik
On WiSoc we cannot access mac register before it is resetted. Otherwise it will crash hardware.
Signed-off-by Wojciech Dubowik <[email protected]>--- drivers/net/wireless/ath/ath5k/base.c | 7 ++- drivers/net/wireless/ath/ath5k/reset.c | 113 ++++++++++++++++++++++++------- 2 files changed, 93 insertions(+), 27 deletions(-)drivers/net/wireless/ath/ath5k/base.c +0
-0drivers/net/wireless/ath/ath5k/base.c ath5k_intr(int irq, void *dev_id) · line 2181
Quoted 8 lines
drivers/net/wireless/ath/ath5k/base.c +2
-1@@ -2175,7 +2175,8 @@ ath5k_intr(int irq, void *dev_id)unsigned int counter =1000 ;if (unlikely (test_bit (ATH_STAT_INVALID , sc->status) || - !ath5k_hw_is_intr_pending (ah))) + ((ath5k_get_bus_type (ah) !=ATH_AHB ) && + !ath5k_hw_is_intr_pending (ah))))return IRQ_NONE ;Hrm, we really don't want to sprinkle these special cases all around the driver. Perhaps it'd be better to make sure
ATH_STAT_INVALIDisn't set until the soc is ready. And, in other cases, using capability bits or something? Bus type seems a bit crude.Quoted 3 lines
drivers/net/wireless/ath/ath5k/base.c +0
-0@@ -2241,6 +2242,10 @@ ath5k_intr(int irq, void *dev_id)tasklet_schedule (&sc->rf_kill.toggleq);
Quoted 8 lines
drivers/net/wireless/ath/ath5k/base.c ath5k_intr(int irq, void *dev_id) · line 2249
Quoted 6 lines
drivers/net/wireless/ath/ath5k/base.c +4
-0} + +if (ath5k_get_bus_type (ah) ==ATH_AHB ) +break ; + }while (ath5k_hw_is_intr_pending (ah) && --counter >0 );Why?
I think I had this in one of my old patches, because occasionally the device would get duplicate interrupt events (causing the kernel IRQ code to eventually disable the interrupt completely).
Quoted 10 lines
drivers/net/wireless/ath/ath5k/base.c static void ath5k_hw_set_sleep_clock(struct ath5k_hw *ah, bool enable) · line 595
Quoted 7 lines
drivers/net/wireless/ath/ath5k/base.c +3
-0@@ -536,6 +590,9 @@ static void ath5k_hw_set_sleep_clock(struct ath5k_hw *ah, bool enable)struct ath5k_eeprom_info *ee = &ah->ah_capabilities.cap_eeprom;u32 scal, spending, usec32; +if (ath5k_get_bus_type (ah) ==ATH_AHB ) + enable = false; +Do you need this hunk? We never enable the sleep clock for AP mode.
What does this have to do with AP mode? Some people use these devices in station mode.
If I remember correctly, these devices have problems with sleep clock, maybe because parts of the WMAC clock overlap with the CPU clock (they're inside the same chip after all).
- Felix
Quoted 7 lines · Felix Fietkau
Quoted 2 lines
Do you need this hunk? We never enable the sleep clock for AP mode.
What does this have to do with AP mode? Some people use these devices in station mode.
If I remember correctly, these devices have problems with sleep clock, maybe because parts of the WMAC clock overlap with the CPU clock (they're inside the same chip after all).
Well I mentioned it because I didn't know how old the patchset was originally, and we had problems with the TSF and sleep clock in general in AP mode, so I didn't know if this line was working around the same issue.
Signature · 1 line
Bob Copeland %% www.bobcopeland.com