ssb: Implement fast powerup delay calculation

PATCH New Gábor Stefanik 3 participants 12 messages May 1, 2010 Sonics silicon backplane driver (ssb)

lore

Share & tools

lore

Gábor Stefanik

author

2010-05-01 16:29 UTC #0 lore raw

PATCH 1/1 ssb: Implement fast powerup delay calculation

Implement fast powerup delay calculation, as described in the recently updated specs. The 4325 part is coming soon.

Signed-off-by G?bor Stefanik<[email protected]>

---

To people experiencing DMA errors: please test this patch, it touches the PMU code, one of the suspected problem areas.

I'm submitting this patch using Thunderbird 3; so I cannot be held liable if it gets damaged in transit. Blame Mozilla. :-)

drivers/ssb/driver_chipcommon.c +29 -5

Index: wireless-testing/drivers/ssb/driver_chipcommon.c
===================================================================
@@ -209,7 +209,24 @@ static void chipco_powercontrol_init(str
  	}
  }
 
-static void calc_fast_powerup_delay(struct ssb_chipcommon *cc)
+static u16 pmu_fast_powerup_delay(struct ssb_chipcommon *cc)
+{
+	struct ssb_bus *bus = cc->dev->bus;
+
+	switch (bus->chip_id) {
+	case 0x4312:
+	case 0x4322:
+	case 0x4328:
+		return 7000;
+	case 0x4325:
+		/* TODO: */
+	default:
+		break;
+	}
+	return 15000;
+}
+
+static u16 calc_fast_powerup_delay(struct ssb_chipcommon *cc)
  {
  	struct ssb_bus *bus = cc->dev->bus;
  	int minfreq;

@@ -217,26 +234,33 @@ static void calc_fast_powerup_delay(stru
  	u32 pll_on_delay;
 
  	if (bus->bustype != SSB_BUSTYPE_PCI)
-		return;
+		return 0;
+	if (cc->capabilities&  SSB_CHIPCO_CAP_PMU)
+		return pmu_fast_powerup_delay(cc);
  	if (!(cc->capabilities&  SSB_CHIPCO_CAP_PCTL))
-		return;
+		return 0;
 
  	minfreq = chipco_pctl_clockfreqlimit(cc, 0);
  	pll_on_delay = chipco_read32(cc, SSB_CHIPCO_PLLONDELAY);
  	tmp = (((pll_on_delay + 2) * 1000000) + (minfreq - 1)) / minfreq;
  	SSB_WARN_ON(tmp&  ~0xFFFF);
 
-	cc->fast_pwrup_delay = tmp;
+	return tmp;
  }
 
  void ssb_chipcommon_init(struct ssb_chipcommon *cc)
  {
+	u16 delay;
+
  	if (!cc->dev)
  		return; /* We don't have a ChipCommon */
  	ssb_pmu_init(cc);
  	chipco_powercontrol_init(cc);
  	ssb_chipco_set_clockmode(cc, SSB_CLKMODE_FAST);
-	calc_fast_powerup_delay(cc);
+	delay = calc_fast_powerup_delay(cc);
+	ssb_printk(KERN_INFO PFX "fast_pwrup_delay is %d\n", delay);
+	cc->fast_pwrup_delay = delay;
+	ssb_write16(cc->dev, SSB_MMIO_POWERUP_DELAY, delay);
  }
 
  void ssb_chipco_suspend(struct ssb_chipcommon *cc)

include/linux/ssb/ssb_regs.h +1 -0

Index: wireless-testing/include/linux/ssb/ssb_regs.h
===================================================================
@@ -26,6 +26,7 @@
  #define	SSB_EUART		(SSB_EXTIF_BASE + 0x00800000)
  #define	SSB_LED			(SSB_EXTIF_BASE + 0x00900000)
 
+#define SSB_MMIO_POWERUP_DELAY	0x06A8
 
  /* Enumeration space constants */
  #define SSB_CORE_SIZE		0x1000	/* Size of a core MMIO area */

Gábor Stefanik

author

2010-05-01 16:32 UTC #1 lore raw

PATCH 1/1 ssb: Implement fast powerup delay calculation

2010/5/1 G?bor Stefanik <[email protected]>:

Quoted 11 lines

Implement fast powerup delay calculation, as described in the recently updated specs. The 4325 part is coming soon.
Signed-off-by G?bor Stefanik<[email protected]>
---
To people experiencing DMA errors: please test this patch, it touches the PMU code, one of the suspected problem areas.
I'm submitting this patch using Thunderbird 3; so I cannot be held liable if it gets damaged in transit. Blame Mozilla. :-)

Apparently the nasty space-doubling issue seen in the Tb3 beta is still present in the final version... here is the patch as an attachment.

--G?bor
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ssb_fast_pwrup_delay.patch
Type: application/octet-stream
Size: 2488 bytes
Desc: not available
URL: <http://lists.infradead.org/pipermail/b43-dev/attachments/20100501/df8a418b/attachment.obj>

Michael Buesch

2010-05-01 17:07 UTC #2 lore raw

PATCH 1/1 ssb: Implement fast powerup delay calculation

In reply to Gábor Stefanik

include/linux/ssb/ssb_regs.h line 29

Quoted 13 lines · On Saturday 01 May 2010 18:29:06 G?bor Stefanik wrote:

drivers/ssb/driver_chipcommon.c +1 -0
+	ssb_write16(cc->dev, SSB_MMIO_POWERUP_DELAY, delay);
  }
 
  void ssb_chipco_suspend(struct ssb_chipcommon *cc)
Index: wireless-testing/include/linux/ssb/ssb_regs.h
===================================================================
include/linux/ssb/ssb_regs.h +1 -0
@@ -26,6 +26,7 @@
  #define	SSB_EUART		(SSB_EXTIF_BASE + 0x00800000)
  #define	SSB_LED			(SSB_EXTIF_BASE + 0x00900000)
 
+#define SSB_MMIO_POWERUP_DELAY	0x06A8

I think you are really confusing something here.
That register is a wireless core register and we already write it in b43.

Signature · 1 line

Greetings, Michael.

Gábor Stefanik

author

2010-05-01 17:13 UTC #3 lore raw

PATCH 1/1 ssb: Implement fast powerup delay calculation

2010/5/1 Michael Buesch <[email protected]>:

Quoted 13 lines · On Saturday 01 May 2010 18:29:06 G?bor Stefanik wrote:

Quoted 12 lines
+ ? ? ssb_write16(cc->dev, SSB_MMIO_POWERUP_DELAY, delay); ? }
? void ssb_chipco_suspend(struct ssb_chipcommon *cc)
Index: wireless-testing/include/linux/ssb/ssb_regs.h
===================================================================
include/linux/ssb/ssb_regs.h +0 -0
@@ -26,6 +26,7 @@
 ? #define ? ? SSB_EUART ? ? ? ? ? ? ? (SSB_EXTIF_BASE + 0x00800000)
 ? #define ? ? SSB_LED ? ? ? ? ? ? ? ? (SSB_EXTIF_BASE + 0x00900000)
 

Quoted 8 lines

include/linux/ssb/ssb_regs.h line 29
Quoted 1 line
include/linux/ssb/ssb_regs.h +1 -0
+#define SSB_MMIO_POWERUP_DELAY ? ? ? 0x06A8
I think you are really confusing something here.
That register is a wireless core register and we already write it in b43.
--
Greetings, Michael.

This is what I am implementing: http://bcm-v4.sipsolutions.net/802.11/Init
Here, it clearly says MMIO offset.

Signature · 1 line

Vista: [V]iruses, [I]ntruders, [S]pyware, [T]rojans and [A]dware. :-)

Gábor Stefanik

author

2010-05-01 17:19 UTC #4 lore raw

PATCH 1/1 ssb: Implement fast powerup delay calculation

2010/5/1 G?bor Stefanik <[email protected]>:

Quoted 23 lines

2010/5/1 Michael Buesch <[email protected]>:
Quoted 13 lines · On Saturday 01 May 2010 18:29:06 G?bor Stefanik wrote:
Quoted 12 lines
+ ? ? ssb_write16(cc->dev, SSB_MMIO_POWERUP_DELAY, delay); ? }
? void ssb_chipco_suspend(struct ssb_chipcommon *cc)
Index: wireless-testing/include/linux/ssb/ssb_regs.h
===================================================================
include/linux/ssb/ssb_regs.h +0 -0
@@ -26,6 +26,7 @@
 ? #define ? ? SSB_EUART ? ? ? ? ? ? ? (SSB_EXTIF_BASE + 0x00800000)
 ? #define ? ? SSB_LED ? ? ? ? ? ? ? ? (SSB_EXTIF_BASE + 0x00900000)
 
Quoted 8 lines
include/linux/ssb/ssb_regs.h line 29
Quoted 1 line
include/linux/ssb/ssb_regs.h +1 -0
+#define SSB_MMIO_POWERUP_DELAY ? ? ? 0x06A8
I think you are really confusing something here.
That register is a wireless core register and we already write it in b43.
--
Greetings, Michael.

Quoted 6 lines

This is what I am implementing: http://bcm-v4.sipsolutions.net/802.11/Init
Here, it clearly says MMIO offset.
--
Vista: [V]iruses, [I]ntruders, [S]pyware, [T]rojans and [A]dware. :-)

Sorry, now I understand. The MMIO write is unnecessary.

Signature · 1 line

Vista: [V]iruses, [I]ntruders, [S]pyware, [T]rojans and [A]dware. :-)

Michael Buesch

2010-05-01 17:19 UTC #5 lore raw

PATCH 1/1 ssb: Implement fast powerup delay calculation

In reply to Gábor Stefanik

Quoted 19 lines · On Saturday 01 May 2010 19:13:31 G?bor Stefanik wrote:

2010/5/1 Michael Buesch <[email protected]>:
Quoted 18 lines · On Saturday 01 May 2010 18:29:06 G?bor Stefanik wrote:
include/linux/ssb/ssb_regs.h line 29
Quoted 13 lines
+ ssb_write16(cc->dev, SSB_MMIO_POWERUP_DELAY, delay);
}
void ssb_chipco_suspend(struct ssb_chipcommon *cc)
Index: wireless-testing/include/linux/ssb/ssb_regs.h
===================================================================
include/linux/ssb/ssb_regs.h +1 -0
@@ -26,6 +26,7 @@
  #define     SSB_EUART               (SSB_EXTIF_BASE + 0x00800000)
  #define     SSB_LED                 (SSB_EXTIF_BASE + 0x00900000)
 
+#define SSB_MMIO_POWERUP_DELAY       0x06A8
I think you are really confusing something here.
That register is a wireless core register and we already write it in b43.

Quoted 6 lines

Quoted 3 lines
--
Greetings, Michael.
This is what I am implementing: http://bcm-v4.sipsolutions.net/802.11/Init
Here, it clearly says MMIO offset.

Yeah. Just what I said. It is an 802.11 core register. We already write it in b43/main.c/b43_chip_init(). It is _wrong_ to write to the chipcommon at that offset.

Signature · 1 line

Greetings, Michael.

Larry Finger

2010-05-01 17:42 UTC #6 lore raw

PATCH 1/1 ssb: Implement fast powerup delay calculation

Quoted 24 lines · Michael Buesch

Quoted 23 lines · On Saturday 01 May 2010 19:13:31 G?bor Stefanik wrote:
2010/5/1 Michael Buesch <[email protected]>:
Quoted 21 lines · On Saturday 01 May 2010 18:29:06 G?bor Stefanik wrote:
include/linux/ssb/ssb_regs.h line 29
Quoted 13 lines
+ ssb_write16(cc->dev, SSB_MMIO_POWERUP_DELAY, delay);
}
void ssb_chipco_suspend(struct ssb_chipcommon *cc)
Index: wireless-testing/include/linux/ssb/ssb_regs.h
===================================================================
include/linux/ssb/ssb_regs.h +1 -0
@@ -26,6 +26,7 @@
  #define     SSB_EUART               (SSB_EXTIF_BASE + 0x00800000)
  #define     SSB_LED                 (SSB_EXTIF_BASE + 0x00900000)
 
+#define SSB_MMIO_POWERUP_DELAY       0x06A8
I think you are really confusing something here.
That register is a wireless core register and we already write it in b43.
--
Greetings, Michael.

Quoted 6 lines

Quoted 2 lines
This is what I am implementing: http://bcm-v4.sipsolutions.net/802.11/Init
Here, it clearly says MMIO offset.
Yeah. Just what I said. It is an 802.11 core register. We already write it in b43/main.c/b43_chip_init(). It is _wrong_ to write to the chipcommon at that offset.

I may have gotten the MMIO offset part wrong in the specs, but the Broadcom driver definitely writes to offset 0x648 at that point.

I know that b43 writes that location; however, it is much later in the startup sequence. Whether that is important is unknown at this point.

The previous code resulted in a value of zero being written to the location in question, but the value for the 4315 is 7000.

Larry

Gábor Stefanik

author

2010-05-01 17:48 UTC #7 lore raw

PATCH 1/1 ssb: Implement fast powerup delay calculation

Quoted 35 lines · Larry Finger

Quoted 24 lines · Michael Buesch
Quoted 23 lines · On Saturday 01 May 2010 19:13:31 G?bor Stefanik wrote:
2010/5/1 Michael Buesch <[email protected]>:
Quoted 21 lines · On Saturday 01 May 2010 18:29:06 G?bor Stefanik wrote:
include/linux/ssb/ssb_regs.h line 29
Quoted 13 lines
+ ? ? ssb_write16(cc->dev, SSB_MMIO_POWERUP_DELAY, delay); ? }
? void ssb_chipco_suspend(struct ssb_chipcommon *cc)
Index: wireless-testing/include/linux/ssb/ssb_regs.h
===================================================================
include/linux/ssb/ssb_regs.h +1 -0
@@ -26,6 +26,7 @@
 ? #define ? ? SSB_EUART ? ? ? ? ? ? ? (SSB_EXTIF_BASE + 0x00800000)
 ? #define ? ? SSB_LED ? ? ? ? ? ? ? ? (SSB_EXTIF_BASE + 0x00900000)
 
+#define SSB_MMIO_POWERUP_DELAY ? ? ? 0x06A8
I think you are really confusing something here.
That register is a wireless core register and we already write it in b43.
--
Greetings, Michael.
Quoted 6 lines
Quoted 2 lines
This is what I am implementing: http://bcm-v4.sipsolutions.net/802.11/Init
Here, it clearly says MMIO offset.
Yeah. Just what I said. It is an 802.11 core register. We already write it in b43/main.c/b43_chip_init(). It is _wrong_ to write to the chipcommon at that offset.
I may have gotten the MMIO offset part wrong in the specs, but the Broadcom driver definitely writes to offset 0x648 at that point.

Quoted 8 lines

I know that b43 writes that location; however, it is much later in the startup sequence. Whether that is important is unknown at this point.
The previous code resulted in a value of zero being written to the location in question, but the value for the 4315 is 7000.
Larry

According to http://bcm-v4.sipsolutions.net/802.11/Init, b43 is right
- the write should go to the 802.11 core, not ChipCommon.

Signature · 1 line

Vista: [V]iruses, [I]ntruders, [S]pyware, [T]rojans and [A]dware. :-)

Michael Buesch

2010-05-01 18:15 UTC #8 lore raw

PATCH 1/1 ssb: Implement fast powerup delay calculation

In reply to Larry Finger

Quoted 2 lines · On Saturday 01 May 2010 19:42:27 Larry Finger wrote:

I may have gotten the MMIO offset part wrong in the specs, but the Broadcom driver definitely writes to offset 0x648 at that point.

I do not think you are right.

First: The write has to happen to the 802.11 core and not to the chipcommon core.

Second: The code that was touched by G?bor is _very_ early in the SSB init. Way ahead of the wireless core init. The specs state that the delay value has to be written after the wireless core init is nearly done.
We do _exactly_ that in b43.

The only thing that needs to be done is to add the additional switch statement. No MMIO writes have to be touched.

Signature · 1 line

Greetings, Michael.

Larry Finger

2010-05-01 18:27 UTC #9 lore raw

PATCH 1/1 ssb: Implement fast powerup delay calculation

Quoted 8 lines · Michael Buesch

Quoted 2 lines · On Saturday 01 May 2010 19:42:27 Larry Finger wrote:
I may have gotten the MMIO offset part wrong in the specs, but the Broadcom driver definitely writes to offset 0x648 at that point.
I do not think you are right.
First: The write has to happen to the 802.11 core and not to the chipcommon core.

Quoted 7 lines

Second: The code that was touched by G?bor is _very_ early in the SSB init. Way ahead of the wireless core init. The specs state that the delay value has to be written after the wireless core init is nearly done.
We do _exactly_ that in b43.
The only thing that needs to be done is to add the additional switch statement. No MMIO writes have to be touched.

OK, I changed the spec to remove the write.

Larry

Michael Buesch

2010-05-01 18:40 UTC #10 lore raw

PATCH 1/1 ssb: Implement fast powerup delay calculation

Quoted 1 line · On Saturday 01 May 2010 20:27:42 Larry Finger wrote:

OK, I changed the spec to remove the write.

Uhm, I think the spec was perfecly fine.

Signature · 1 line

Greetings, Michael.

Gábor Stefanik

author

2010-05-01 18:43 UTC #11 lore raw

PATCH 1/1 ssb: Implement fast powerup delay calculation

Quoted 8 lines · Michael Buesch

Quoted 1 line · On Saturday 01 May 2010 20:27:42 Larry Finger wrote:
OK, I changed the spec to remove the write.
Uhm, I think the spec was perfecly fine.
--
Greetings, Michael.

Yes. The specs stated "do this during 802.11 core init", the problem was the patch doing the write during ChipCommon init, much earlier.

Signature · 1 line

Vista: [V]iruses, [I]ntruders, [S]pyware, [T]rojans and [A]dware. :-)