Skip to contentSkip to conversation

OMAP: mailbox: fixes and enhancements

PATCH Superseded Hari Kanigeri 4 participants 22 messages Nov 23, 2010 – Nov 25, 2010

Subsystems (3)

v4 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

v4 cover OMAP: mailbox: fixes and enhancements

Thanks to Benoit and Felipe Balbi for their comments.

Here is the revised set taking into account comments received.
========================

- Added more description for checkpatch warnings
- Created revision field to check for mailbox IP version to use for rx interrupt disable functionality.
- Cleaned up description in change full flag per mailbox patch.

References:
http://marc.info/?l=linux-omap&m=129012283514652&w=2
http://marc.info/?l=linux-omap&m=129016767426046&w=2
http://www.mail-archive.com/[email protected]/msg38826.html

The previous versions of this patch set can be found here
=======================

http://www.spinics.net/lists/arm-kernel/msg104544.html
http://www.spinics.net/lists/linux-omap/msg39988.html
http://www.spinics.net/lists/linux-omap/msg39931.html
http://www.mail-archive.com/[email protected]/msg37278.html

Fernando Guzman Lugo (1):
OMAP: mailbox: change full flag per mailbox queue instead of global

Hari Kanigeri (4):
OMAP: mailbox: fix rx interrupt disable in omap4
OMAP: mailbox: fix checkpatch warnings
OMAP: mailbox: send message in process context
OMAP: mailbox: add notification support for multiple readers

 arch/arm/mach-omap2/mailbox.c             |    7 ++-
 arch/arm/plat-omap/include/plat/mailbox.h |   12 ++-
 arch/arm/plat-omap/mailbox.c              |  135 +++++++++++++++++------------
 3 files changed, 96 insertions(+), 58 deletions(-)
#1 lore raw

PATCH v4 1/5 OMAP: mailbox: change full flag per mailbox queue instead of global

From: Fernando Guzman Lugo <[email protected]>

The variable rq_full flag is a global variable, so if there are multiple mailbox users there will be conflicts. Now there is a full flag per mailbox queue.

---
 arch/arm/plat-omap/include/plat/mailbox.h |    1 +
 arch/arm/plat-omap/mailbox.c              |    9 +++++++--
 2 files changed, 8 insertions(+), 2 deletions(-)
arch/arm/plat-omap/include/plat/mailbox.h +1 -0
index 9976565..13f2ef3 100644
@@ -48,6 +48,7 @@ struct omap_mbox_queue {
 	struct tasklet_struct	tasklet;
 	int	(*callback)(void *);
 	struct omap_mbox	*mbox;
+	bool full;
 };
 
 struct omap_mbox {
arch/arm/plat-omap/mailbox.c +7 -2
index d2fafb8..48e161c 100644
@@ -33,7 +33,6 @@
 
 static struct workqueue_struct *mboxd;
 static struct omap_mbox **mboxes;
-static bool rq_full;
 
 static int mbox_configured;
 static DEFINE_MUTEX(mbox_configured_lock);
@@ -148,6 +147,12 @@ static void mbox_rx_work(struct work_struct *work)
 
 		if (mq->callback)
 			mq->callback((void *)msg);
+		spin_lock_irq(&mq->lock);
+		if (mq->full) {
+			mq->full = false;
+			omap_mbox_enable_irq(mq->mbox, IRQ_RX);
+		}
+		spin_unlock_irq(&mq->lock);
 	}
 }
 
@@ -170,7 +175,7 @@ static void __mbox_rx_interrupt(struct omap_mbox *mbox)
 	while (!mbox_fifo_empty(mbox)) {
 		if (unlikely(kfifo_avail(&mq->fifo) < sizeof(msg))) {
 			omap_mbox_disable_irq(mbox, IRQ_RX);
-			rq_full = true;
+			mq->full = true;
 			goto nomem;
 		}
 
Signature · 1 line
1.7.0
#2 lore raw

PATCH v4 1/5 OMAP: mailbox: change full flag per mailbox queue instead of global

Quoted 52 lines · Hari Kanigeri

From: Fernando Guzman Lugo <[email protected]>

The variable rq_full flag is a global variable, so if there are multiple mailbox users there will be conflicts. Now there is a full flag per mailbox queue.

---
?arch/arm/plat-omap/include/plat/mailbox.h | ? ?1 + ?arch/arm/plat-omap/mailbox.c ? ? ? ? ? ? ?| ? ?9 +++++++-- ?2 files changed, 8 insertions(+), 2 deletions(-)

arch/arm/plat-omap/include/plat/mailbox.h +1 -0
index 9976565..13f2ef3 100644
@@ -48,6 +48,7 @@ struct omap_mbox_queue {
 ? ? ? ?struct tasklet_struct ? tasklet;
 ? ? ? ?int ? ? (*callback)(void *);
 ? ? ? ?struct omap_mbox ? ? ? ?*mbox;
+ ? ? ? bool full;
 ?};
 
 ?struct omap_mbox {
arch/arm/plat-omap/mailbox.c +6 -1
index d2fafb8..48e161c 100644
@@ -33,7 +33,6 @@
 
 ?static struct workqueue_struct *mboxd;
 ?static struct omap_mbox **mboxes;
-static bool rq_full;
 
 ?static int mbox_configured;
 ?static DEFINE_MUTEX(mbox_configured_lock);
@@ -148,6 +147,12 @@ static void mbox_rx_work(struct work_struct *work)
 
 ? ? ? ? ? ? ? ?if (mq->callback)
 ? ? ? ? ? ? ? ? ? ? ? ?mq->callback((void *)msg);
+ ? ? ? ? ? ? ? spin_lock_irq(&mq->lock);
+ ? ? ? ? ? ? ? if (mq->full) {
+ ? ? ? ? ? ? ? ? ? ? ? mq->full = false;
+ ? ? ? ? ? ? ? ? ? ? ? omap_mbox_enable_irq(mq->mbox, IRQ_RX);
+ ? ? ? ? ? ? ? }
+ ? ? ? ? ? ? ? spin_unlock_irq(&mq->lock);
 ? ? ? ?}
 ?}
 

arch/arm/plat-omap/mailbox.c static void __mbox_rx_interrupt(struct omap_mbox *mbox) · line 178

Quoted 5 lines
arch/arm/plat-omap/mailbox.c +1 -1
@@ -170,7 +175,7 @@ static void __mbox_rx_interrupt(struct omap_mbox *mbox)
 ? ? ? ?while (!mbox_fifo_empty(mbox)) {
 ? ? ? ? ? ? ? ?if (unlikely(kfifo_avail(&mq->fifo) < sizeof(msg))) {
 ? ? ? ? ? ? ? ? ? ? ? ?omap_mbox_disable_irq(mbox, IRQ_RX);
- ? ? ? ? ? ? ? ? ? ? ? rq_full = true;
+ ? ? ? ? ? ? ? ? ? ? ? mq->full = true;

Should this also be inside spin_lock_irq?

Quoted 6 lines
arch/arm/plat-omap/mailbox.c +0 -1
 ? ? ? ? ? ? ? ? ? ? ? ?goto nomem;
 ? ? ? ? ? ? ? ?}
 
--

1.7.0

#3 lore raw

PATCH v4 1/5 OMAP: mailbox: change full flag per mailbox queue instead of global

Quoted 52 lines · Charulatha
Quoted 51 lines · Hari Kanigeri

From: Fernando Guzman Lugo <[email protected]>

The variable rq_full flag is a global variable, so if there are multiple mailbox users there will be conflicts. Now there is a full flag per mailbox queue.

---
?arch/arm/plat-omap/include/plat/mailbox.h | ? ?1 + ?arch/arm/plat-omap/mailbox.c ? ? ? ? ? ? ?| ? ?9 +++++++-- ?2 files changed, 8 insertions(+), 2 deletions(-)

arch/arm/plat-omap/include/plat/mailbox.h +1 -0
index 9976565..13f2ef3 100644
@@ -48,6 +48,7 @@ struct omap_mbox_queue {
 ? ? ? ?struct tasklet_struct ? tasklet;
 ? ? ? ?int ? ? (*callback)(void *);
 ? ? ? ?struct omap_mbox ? ? ? ?*mbox;
+ ? ? ? bool full;
 ?};
 
 ?struct omap_mbox {
arch/arm/plat-omap/mailbox.c +6 -1
index d2fafb8..48e161c 100644
@@ -33,7 +33,6 @@
 
 ?static struct workqueue_struct *mboxd;
 ?static struct omap_mbox **mboxes;
-static bool rq_full;
 
 ?static int mbox_configured;
 ?static DEFINE_MUTEX(mbox_configured_lock);
@@ -148,6 +147,12 @@ static void mbox_rx_work(struct work_struct *work)
 
 ? ? ? ? ? ? ? ?if (mq->callback)
 ? ? ? ? ? ? ? ? ? ? ? ?mq->callback((void *)msg);
+ ? ? ? ? ? ? ? spin_lock_irq(&mq->lock);
+ ? ? ? ? ? ? ? if (mq->full) {
+ ? ? ? ? ? ? ? ? ? ? ? mq->full = false;
+ ? ? ? ? ? ? ? ? ? ? ? omap_mbox_enable_irq(mq->mbox, IRQ_RX);
+ ? ? ? ? ? ? ? }
+ ? ? ? ? ? ? ? spin_unlock_irq(&mq->lock);
 ? ? ? ?}
 ?}
 
Quoted 9 lines

arch/arm/plat-omap/mailbox.c static void __mbox_rx_interrupt(struct omap_mbox *mbox) · line 178

Quoted 6 lines
arch/arm/plat-omap/mailbox.c +1 -1
@@ -170,7 +175,7 @@ static void __mbox_rx_interrupt(struct omap_mbox *mbox)
 ? ? ? ?while (!mbox_fifo_empty(mbox)) {
 ? ? ? ? ? ? ? ?if (unlikely(kfifo_avail(&mq->fifo) < sizeof(msg))) {
 ? ? ? ? ? ? ? ? ? ? ? ?omap_mbox_disable_irq(mbox, IRQ_RX);
- ? ? ? ? ? ? ? ? ? ? ? rq_full = true;
+ ? ? ? ? ? ? ? ? ? ? ? mq->full = true;

Should this also be inside spin_lock_irq?

Not needed. this is already run from interrupt context.

Thank you,
Best regards,
Hari

#4 lore raw

PATCH v4 1/5 OMAP: mailbox: change full flag per mailbox queue instead of global

Quoted 63 lines · Hari
Quoted 52 lines · Charulatha
Quoted 51 lines · Hari Kanigeri

From: Fernando Guzman Lugo <[email protected]>

The variable rq_full flag is a global variable, so if there are multiple mailbox users there will be conflicts. Now there is a full flag per mailbox queue.

---
?arch/arm/plat-omap/include/plat/mailbox.h | ? ?1 + ?arch/arm/plat-omap/mailbox.c ? ? ? ? ? ? ?| ? ?9 +++++++-- ?2 files changed, 8 insertions(+), 2 deletions(-)

arch/arm/plat-omap/include/plat/mailbox.h +1 -0
index 9976565..13f2ef3 100644
@@ -48,6 +48,7 @@ struct omap_mbox_queue {
 ? ? ? ?struct tasklet_struct ? tasklet;
 ? ? ? ?int ? ? (*callback)(void *);
 ? ? ? ?struct omap_mbox ? ? ? ?*mbox;
+ ? ? ? bool full;
 ?};
 
 ?struct omap_mbox {
arch/arm/plat-omap/mailbox.c +6 -1
index d2fafb8..48e161c 100644
@@ -33,7 +33,6 @@
 
 ?static struct workqueue_struct *mboxd;
 ?static struct omap_mbox **mboxes;
-static bool rq_full;
 
 ?static int mbox_configured;
 ?static DEFINE_MUTEX(mbox_configured_lock);
@@ -148,6 +147,12 @@ static void mbox_rx_work(struct work_struct *work)
 
 ? ? ? ? ? ? ? ?if (mq->callback)
 ? ? ? ? ? ? ? ? ? ? ? ?mq->callback((void *)msg);
+ ? ? ? ? ? ? ? spin_lock_irq(&mq->lock);
+ ? ? ? ? ? ? ? if (mq->full) {
+ ? ? ? ? ? ? ? ? ? ? ? mq->full = false;
+ ? ? ? ? ? ? ? ? ? ? ? omap_mbox_enable_irq(mq->mbox, IRQ_RX);
+ ? ? ? ? ? ? ? }
+ ? ? ? ? ? ? ? spin_unlock_irq(&mq->lock);
 ? ? ? ?}
 ?}
 
Quoted 9 lines

arch/arm/plat-omap/mailbox.c static void __mbox_rx_interrupt(struct omap_mbox *mbox) · line 178

Quoted 6 lines
arch/arm/plat-omap/mailbox.c +1 -1
@@ -170,7 +175,7 @@ static void __mbox_rx_interrupt(struct omap_mbox *mbox)
 ? ? ? ?while (!mbox_fifo_empty(mbox)) {
 ? ? ? ? ? ? ? ?if (unlikely(kfifo_avail(&mq->fifo) < sizeof(msg))) {
 ? ? ? ? ? ? ? ? ? ? ? ?omap_mbox_disable_irq(mbox, IRQ_RX);
- ? ? ? ? ? ? ? ? ? ? ? rq_full = true;
+ ? ? ? ? ? ? ? ? ? ? ? mq->full = true;

Should this also be inside spin_lock_irq?

Quoted 1 line

Not needed. this is already run from interrupt context.

oh ok.

Quoted 5 lines

Thank you,
Best regards,
Hari

#5 lore raw

PATCH v4 2/5 OMAP: mailbox: fix rx interrupt disable in omap4

In reply to Hari Kanigeri

disablign rx interrupt on omap4 is different than its pre-decessors. The bit in OMAP4_MAILBOX_IRQENABLE_CLR should be set to disable the interrupts instead of clearing the bit.

Defined rev field in mailbox structure to differentiate the mailbox versions.

---
 arch/arm/mach-omap2/mailbox.c             |    7 ++++++-
 arch/arm/plat-omap/include/plat/mailbox.h |    4 ++++
 arch/arm/plat-omap/mailbox.c              |    4 ++++
 3 files changed, 14 insertions(+), 1 deletions(-)
arch/arm/mach-omap2/mailbox.c +6 -1
index 42dbfa4..c32058d 100644
@@ -195,7 +195,12 @@ static void omap2_mbox_disable_irq(struct omap_mbox *mbox,
 	struct omap_mbox2_priv *p = (struct omap_mbox2_priv *)mbox->priv;
 	u32 l, bit = (irq == IRQ_TX) ? p->notfull_bit : p->newmsg_bit;
 	l = mbox_read_reg(p->irqdisable);
-	l &= ~bit;
+
+	if (mbox->rev == OMAP_MBOX_IP_VERSION_2)
+		l |= bit;
+	else
+		l &= ~bit;
+
 	mbox_write_reg(l, p->irqdisable);
 }
 
arch/arm/plat-omap/include/plat/mailbox.h +4 -0
index 13f2ef3..1655c61 100644
@@ -20,6 +20,9 @@ typedef int __bitwise omap_mbox_type_t;
 #define OMAP_MBOX_TYPE1 ((__force omap_mbox_type_t) 1)
 #define OMAP_MBOX_TYPE2 ((__force omap_mbox_type_t) 2)
 
+#define OMAP_MBOX_IP_LEGACY                    0x1
+#define OMAP_MBOX_IP_VERSION_2                 0x2
+
 struct omap_mbox_ops {
 	omap_mbox_type_t	type;
 	int		(*startup)(struct omap_mbox *mbox);
@@ -58,6 +61,7 @@ struct omap_mbox {
 	struct omap_mbox_ops	*ops;
 	struct device		*dev;
 	void			*priv;
+	int			rev;
 };
 
 int omap_mbox_msg_send(struct omap_mbox *, mbox_msg_t msg);
arch/arm/plat-omap/mailbox.c +4 -0
index 48e161c..a1c6bd9 100644
@@ -358,6 +358,10 @@ int omap_mbox_register(struct device *parent, struct omap_mbox **list)
 			ret = PTR_ERR(mbox->dev);
 			goto err_out;
 		}
+		if (cpu_is_omap44xx())
+			mbox->rev = OMAP_MBOX_IP_VERSION_2;
+		else
+			mbox->rev = OMAP_MBOX_IP_LEGACY;
 	}
 	return 0;
 
Signature · 1 line
1.7.0
#6 lore raw

PATCH v4 2/5 OMAP: mailbox: fix rx interrupt disable in omap4

On Wed, Nov 24, 2010 at 02:56, Hari Kanigeri <[email protected]> wrote:

Should the subject be like this?
OMAP4: mailbox: fix rx interrupt disable

Quoted 1 line

disablign rx interrupt on omap4 is different than its pre-decessors.

Typo ->disablign

Quoted 58 lines

The bit in OMAP4_MAILBOX_IRQENABLE_CLR should be set to disable the interrupts instead of clearing the bit.

Defined rev field in mailbox structure to differentiate the mailbox versions.

---

?arch/arm/mach-omap2/mailbox.c ? ? ? ? ? ? | ? ?7 ++++++-
?arch/arm/plat-omap/include/plat/mailbox.h | ? ?4 ++++
?arch/arm/plat-omap/mailbox.c ? ? ? ? ? ? ?| ? ?4 ++++

?3 files changed, 14 insertions(+), 1 deletions(-)

arch/arm/mach-omap2/mailbox.c +6 -1
index 42dbfa4..c32058d 100644
@@ -195,7 +195,12 @@ static void omap2_mbox_disable_irq(struct omap_mbox *mbox,
 ? ? ? ?struct omap_mbox2_priv *p = (struct omap_mbox2_priv *)mbox->priv;
 ? ? ? ?u32 l, bit = (irq == IRQ_TX) ? p->notfull_bit : p->newmsg_bit;
 ? ? ? ?l = mbox_read_reg(p->irqdisable);
- ? ? ? l &= ~bit;
+
+ ? ? ? if (mbox->rev == OMAP_MBOX_IP_VERSION_2)
+ ? ? ? ? ? ? ? l |= bit;
+ ? ? ? else
+ ? ? ? ? ? ? ? l &= ~bit;
+
 ? ? ? ?mbox_write_reg(l, p->irqdisable);
 ?}
 
arch/arm/plat-omap/include/plat/mailbox.h +4 -0
index 13f2ef3..1655c61 100644
@@ -20,6 +20,9 @@ typedef int __bitwise omap_mbox_type_t;
 ?#define OMAP_MBOX_TYPE1 ((__force omap_mbox_type_t) 1)
 ?#define OMAP_MBOX_TYPE2 ((__force omap_mbox_type_t) 2)
 
+#define OMAP_MBOX_IP_LEGACY ? ? ? ? ? ? ? ? ? ?0x1
+#define OMAP_MBOX_IP_VERSION_2 ? ? ? ? ? ? ? ? 0x2
+
 ?struct omap_mbox_ops {
 ? ? ? ?omap_mbox_type_t ? ? ? ?type;
 ? ? ? ?int ? ? ? ? ? ? (*startup)(struct omap_mbox *mbox);
@@ -58,6 +61,7 @@ struct omap_mbox {
 ? ? ? ?struct omap_mbox_ops ? ?*ops;
 ? ? ? ?struct device ? ? ? ? ? *dev;
 ? ? ? ?void ? ? ? ? ? ? ? ? ? ?*priv;
+ ? ? ? int ? ? ? ? ? ? ? ? ? ? rev;
 ?};
 
 ?int omap_mbox_msg_send(struct omap_mbox *, mbox_msg_t msg);
arch/arm/plat-omap/mailbox.c +0 -0
index 48e161c..a1c6bd9 100644

arch/arm/plat-omap/mailbox.c int omap_mbox_register(struct device *parent, struct omap_mbox **list) · line 361

Quoted 4 lines
arch/arm/plat-omap/mailbox.c +1 -0
@@ -358,6 +358,10 @@ int omap_mbox_register(struct device *parent, struct omap_mbox **list)
 ? ? ? ? ? ? ? ? ? ? ? ?ret = PTR_ERR(mbox->dev);
 ? ? ? ? ? ? ? ? ? ? ? ?goto err_out;
 ? ? ? ? ? ? ? ?}
+ ? ? ? ? ? ? ? if (cpu_is_omap44xx())

Do not use cpu_is* checks in plat-omap/*

Quoted 9 lines
arch/arm/plat-omap/mailbox.c +3 -1
+ ? ? ? ? ? ? ? ? ? ? ? mbox->rev = OMAP_MBOX_IP_VERSION_2;
+ ? ? ? ? ? ? ? else
+ ? ? ? ? ? ? ? ? ? ? ? mbox->rev = OMAP_MBOX_IP_LEGACY;
 ? ? ? ?}
 ? ? ? ?return 0;
 
--

1.7.0

#7 lore raw

PATCH v4 2/5 OMAP: mailbox: fix rx interrupt disable in omap4

Quoted 11 lines · Charulatha

arch/arm/plat-omap/mailbox.c int omap_mbox_register(struct device *parent, struct omap_mbox **list) · line 361

Quoted 9 lines
arch/arm/plat-omap/mailbox.c +1 -0
index 48e161c..a1c6bd9 100644
@@ -358,6 +358,10 @@ int omap_mbox_register(struct device *parent, struct omap_mbox **list)
 ? ? ? ? ? ? ? ? ? ? ? ?ret = PTR_ERR(mbox->dev);
 ? ? ? ? ? ? ? ? ? ? ? ?goto err_out;
 ? ? ? ? ? ? ? ?}
+ ? ? ? ? ? ? ? if (cpu_is_omap44xx())

Do not use cpu_is* checks in plat-omap/*

see the previous thread.

Signature · 1 line
balbi
#8 lore raw

PATCH v4 2/5 OMAP: mailbox: fix rx interrupt disable in omap4

Quoted 15 lines · Felipe Balbi
Quoted 13 lines · Charulatha

arch/arm/plat-omap/mailbox.c int omap_mbox_register(struct device *parent, struct · line 362

Quoted 11 lines
arch/arm/plat-omap/mailbox.c +1 -0
index 48e161c..a1c6bd9 100644
@@ -358,6 +358,10 @@ int omap_mbox_register(struct device *parent, struct
 omap_mbox **list)
 ? ? ? ? ? ? ? ? ? ? ? ?ret = PTR_ERR(mbox->dev);
 ? ? ? ? ? ? ? ? ? ? ? ?goto err_out;
 ? ? ? ? ? ? ? ?}
+ ? ? ? ? ? ? ? if (cpu_is_omap44xx())

Do not use cpu_is* checks in plat-omap/*

Quoted 1 line

see the previous thread.

Referring to [1], I do not find why cpu_is* checks is used in plat-omap and why it can't be avoided.

In [1], it was suggested to create the pdata field that will be populated at init
time using the cpu_is* check. But in this version, I am finding that this is done
in plat-omap. This can be handled in mach-omap layer itself and can be passed as a pdata field which can be extracted during probe.

[1] https://patchwork.kernel.org/patch/337131/

#9 lore raw

PATCH v4 2/5 OMAP: mailbox: fix rx interrupt disable in omap4

Quoted 21 lines · Charulatha
Quoted 15 lines · Felipe Balbi
Quoted 13 lines · Charulatha

arch/arm/plat-omap/mailbox.c int omap_mbox_register(struct device *parent, struct · line 362

Quoted 11 lines
arch/arm/plat-omap/mailbox.c +1 -0
index 48e161c..a1c6bd9 100644
@@ -358,6 +358,10 @@ int omap_mbox_register(struct device *parent, struct
 omap_mbox **list)
 ? ? ? ? ? ? ? ? ? ? ? ?ret = PTR_ERR(mbox->dev);
 ? ? ? ? ? ? ? ? ? ? ? ?goto err_out;
 ? ? ? ? ? ? ? ?}
+ ? ? ? ? ? ? ? if (cpu_is_omap44xx())

Do not use cpu_is* checks in plat-omap/*

Quoted 1 line

see the previous thread.

Referring to [1], I do not find why cpu_is* checks is used in plat-omap and why it can't be avoided.

Quoted 8 lines

In [1], it was suggested to create the pdata field that will be populated at init
time using the cpu_is* check. But in this version, I am finding that this is done
in plat-omap. This can be handled in mach-omap layer itself and can be passed as a pdata field which can be extracted during probe.

[1] https://patchwork.kernel.org/patch/337131/

true, since I saw mbox->rev, I confused that with something that was coming from pdata, sorry. You're right, this is not the right way to do it.

Signature · 1 line
balbi
#10 lore raw

PATCH v4 2/5 OMAP: mailbox: fix rx interrupt disable in omap4

In reply to Varadarajan, Charulatha
Quoted 21 lines · Charulatha
Quoted 15 lines · Felipe Balbi
Quoted 13 lines · Charulatha

arch/arm/plat-omap/mailbox.c int omap_mbox_register(struct device *parent, struct · line 362

Quoted 11 lines
arch/arm/plat-omap/mailbox.c +1 -0
index 48e161c..a1c6bd9 100644
@@ -358,6 +358,10 @@ int omap_mbox_register(struct device *parent, struct
 omap_mbox **list)
 ? ? ? ? ? ? ? ? ? ? ? ?ret = PTR_ERR(mbox->dev);
 ? ? ? ? ? ? ? ? ? ? ? ?goto err_out;
 ? ? ? ? ? ? ? ?}
+ ? ? ? ? ? ? ? if (cpu_is_omap44xx())

Do not use cpu_is* checks in plat-omap/*

Quoted 1 line

see the previous thread.

Referring to [1], I do not find why cpu_is* checks is used in plat-omap and why it can't be avoided.

Quoted 9 lines

In [1], it was suggested to create the pdata field that will be populated at init
time using the cpu_is* check. But in this version, I am finding that this is done
in plat-omap. This can be handled in mach-omap layer itself and can be passed as a pdata field which can be extracted during probe.

[1] https://patchwork.kernel.org/patch/337131/

Here are these reasons why I did this way.

- The function omap_mbox_register is called only once during probe time, and I thought it was ok to call cpu check once during probe time. Since each mbox instance needs to be aware of the rev field this was the right place to add since this function iterates through the list during probe time. There are already calls to cpu_is_omap44xx in mailbox probe function.

- platform data is not present for mailbox module. I could add it for revision sake but I would prefer not to do that since this will be a throw away code once the hwmod infrastructure is ready (note: mailbox hwmod patches are under review), and amount of mailbox driver rework might be considerable.

- I could wait till the hwmod patches are ready to include this change, but don't want to put the dependency with hwmod since this is a critical fix and want to make it available in the mainline kernel.

Please let me know what you suggest.

Thank you,
Best regards,
Hari Kanigeri

#11 lore raw

PATCH v4 2/5 OMAP: mailbox: fix rx interrupt disable in omap4

Hari,

Quoted 29 lines · Hari
Quoted 28 lines · Charulatha
Quoted 15 lines · Felipe Balbi
Quoted 13 lines · Charulatha

arch/arm/plat-omap/mailbox.c int omap_mbox_register(struct device *parent, struct · line 362

Quoted 11 lines
arch/arm/plat-omap/mailbox.c +1 -0
index 48e161c..a1c6bd9 100644
@@ -358,6 +358,10 @@ int omap_mbox_register(struct device *parent, struct
 omap_mbox **list)
 ? ? ? ? ? ? ? ? ? ? ? ?ret = PTR_ERR(mbox->dev);
 ? ? ? ? ? ? ? ? ? ? ? ?goto err_out;
 ? ? ? ? ? ? ? ?}
+ ? ? ? ? ? ? ? if (cpu_is_omap44xx())

Do not use cpu_is* checks in plat-omap/*

Quoted 1 line

see the previous thread.

Referring to [1], I do not find why cpu_is* checks is used in plat-omap and why it can't be avoided.

In [1], it was suggested to create the pdata field that will be populated at init
time using the cpu_is* check. But in this version, I am finding that this is done
in plat-omap. This can be handled in mach-omap layer itself and can be passed as a pdata field which can be extracted during probe.

Quoted 8 lines
Quoted 2 lines

[1] https://patchwork.kernel.org/patch/337131/

Here are these ?reasons why I did this way.

- The function omap_mbox_register is called only once during probe time, and I thought it was ok to call cpu check once during probe time.

AFAIK it is not okay. Patches are not accepted by maintainers if cpu_is* checks are newly added in plat-omap/*
Moreover, the plat-omap is for common code between omap1 and omap2+ If any specific info is required due to unavoidable reasons, they should be managed using pdata fld/ by some other means.

Quoted 4 lines

Since each mbox instance needs to be aware of the rev field this was the right place to add since this function iterates through the list during probe time. There are already calls to cpu_is_omap44xx in mailbox probe function.

Infact , there are some more drivers like that. But some of them are getting cleaned up slowly (work under progress). But old code having cpu_is* checks does not justify adding a new cpu_is* check. As I mentioned, it is repeatedly being insisted in LO mailing list to avoid cpu_is* checks in plat-omap and to clean up the drivers to avoid these checks.

Quoted 6 lines

- platform data is not present for mailbox module. I could add it for revision sake but I would prefer not to do that since this will be a throw away code once the hwmod infrastructure is ready (note: mailbox hwmod patches are under review), and amount of mailbox driver rework might be considerable.

I would leave this to Tony/ Felipe/ Benoit to decide on this. If agreed, I don't see any issue. But there should be atleast a "TODO" mentioning that this needs to be cleaned up.

Quoted 11 lines

- I could wait till the hwmod patches are ready to include this change, but don't want to put the dependency with hwmod since this is a critical fix and want to make it available in the mainline kernel.

Please let me know what you suggest.

Thank you,
Best regards,
Hari Kanigeri

#12 lore raw

PATCH v4 2/5 OMAP: mailbox: fix rx interrupt disable in omap4

Quoted 65 lines · Charulatha

Hari,

Quoted 29 lines · Hari
Quoted 28 lines · Charulatha
Quoted 15 lines · Felipe Balbi
Quoted 13 lines · Charulatha

arch/arm/plat-omap/mailbox.c int omap_mbox_register(struct device *parent, struct · line 362

Quoted 11 lines
arch/arm/plat-omap/mailbox.c +1 -0
index 48e161c..a1c6bd9 100644
@@ -358,6 +358,10 @@ int omap_mbox_register(struct device *parent, struct
 omap_mbox **list)
 ? ? ? ? ? ? ? ? ? ? ? ?ret = PTR_ERR(mbox->dev);
 ? ? ? ? ? ? ? ? ? ? ? ?goto err_out;
 ? ? ? ? ? ? ? ?}
+ ? ? ? ? ? ? ? if (cpu_is_omap44xx())

Do not use cpu_is* checks in plat-omap/*

Quoted 1 line

see the previous thread.

Referring to [1], I do not find why cpu_is* checks is used in plat-omap and why it can't be avoided.

In [1], it was suggested to create the pdata field that will be populated at init
time using the cpu_is* check. But in this version, I am finding that this is done
in plat-omap. This can be handled in mach-omap layer itself and can be passed as a pdata field which can be extracted during probe.

Quoted 8 lines
Quoted 2 lines

[1] https://patchwork.kernel.org/patch/337131/

Here are these ?reasons why I did this way.

- The function omap_mbox_register is called only once during probe time, and I thought it was ok to call cpu check once during probe time.

AFAIK it is not okay. Patches are not accepted by maintainers if cpu_is* checks are newly added in plat-omap/*
Moreover, the plat-omap is for common code between omap1 and omap2+ If any specific info is required due to unavoidable reasons, they should be managed using pdata fld/ by some other means.

Quoted 4 lines

Since each mbox instance needs to be aware of the rev field this was the right place to add since this function iterates through the list during probe time. There are already calls to cpu_is_omap44xx in mailbox probe function.

Infact , there are some more drivers like that. But some of them are getting cleaned up slowly (work under progress). But old code having cpu_is* checks does not justify adding a new cpu_is* check. As I mentioned, it is repeatedly being insisted in LO mailing list to avoid cpu_is* checks in plat-omap and to clean up the drivers to avoid these checks.

Quoted 6 lines

- platform data is not present for mailbox module. I could add it for revision sake but I would prefer not to do that since this will be a throw away code once the hwmod infrastructure is ready (note: mailbox hwmod patches are under review), and amount of mailbox driver rework might be considerable.

Quoted 4 lines

I would leave this to Tony/ Felipe/ Benoit to decide on this. If agreed, I don't see any issue. But there should be atleast a "TODO" mentioning that this needs to be cleaned up.

fair enough. I am dropping this patch from the patch set and will send it with Omar's hwmod patches.
I will send the revised patch set excluding this patch.

Thank you,
Best regards,
Hari

#13 lore raw

PATCH v4 3/5 OMAP: mailbox: fix checkpatch warnings

In reply to Hari Kanigeri

Fix the following checkpatch warnings observed in mailbox module.

WARNING: please, no space for starting a line,
excluding comments
+ fail_alloc_rxq:$

WARNING: please, no space for starting a line,
excluding comments
+ fail_alloc_txq:$

WARNING: please, no space for starting a line,
excluding comments
+ fail_request_irq:$

WARNING: line over 80 characters
+ mbox_kfifo_size = max_t(unsigned int, mbox_kfifo_size, sizeof(mbox_msg_t));

---
 arch/arm/plat-omap/mailbox.c |    9 +++++----
 1 files changed, 5 insertions(+), 4 deletions(-)
arch/arm/plat-omap/mailbox.c +5 -4
index a1c6bd9..13698ab 100644
@@ -281,11 +281,11 @@ static int omap_mbox_startup(struct omap_mbox *mbox)
 
 	return 0;
 
- fail_alloc_rxq:
+fail_alloc_rxq:
 	mbox_queue_free(mbox->txq);
- fail_alloc_txq:
+fail_alloc_txq:
 	free_irq(mbox->irq, mbox);
- fail_request_irq:
+fail_request_irq:
 	if (mbox->ops->shutdown)
 		mbox->ops->shutdown(mbox);
 
@@ -400,7 +400,8 @@ static int __init omap_mbox_init(void)
 
 	/* kfifo size sanity check: alignment and minimal size */
 	mbox_kfifo_size = ALIGN(mbox_kfifo_size, sizeof(mbox_msg_t));
-	mbox_kfifo_size = max_t(unsigned int, mbox_kfifo_size, sizeof(mbox_msg_t));
+	mbox_kfifo_size = max_t(unsigned int, mbox_kfifo_size,
+							sizeof(mbox_msg_t));
 
 	return 0;
 }
Signature · 1 line
1.7.0
#14 lore raw

PATCH v4 4/5 OMAP: mailbox: send message in process context

In reply to Hari Kanigeri

Schedule the Tasklet to send only when mailbox fifo is full and there are pending messages in kifo, else send the message directly in the Process context. This would avoid needless scheduling of Tasklet for every message transfer

---
 arch/arm/plat-omap/mailbox.c |    9 +++++++--
 1 files changed, 7 insertions(+), 2 deletions(-)
arch/arm/plat-omap/mailbox.c +7 -2
index 13698ab..f4177df 100644
@@ -92,20 +92,25 @@ int omap_mbox_msg_send(struct omap_mbox *mbox, mbox_msg_t msg)
 	struct omap_mbox_queue *mq = mbox->txq;
 	int ret = 0, len;
 
-	spin_lock(&mq->lock);
+	spin_lock_bh(&mq->lock);
 
 	if (kfifo_avail(&mq->fifo) < sizeof(msg)) {
 		ret = -ENOMEM;
 		goto out;
 	}
 
+	if (kfifo_is_empty(&mq->fifo) && !__mbox_poll_for_space(mbox)) {
+		mbox_fifo_write(mbox, msg);
+		goto out;
+	}
+
 	len = kfifo_in(&mq->fifo, (unsigned char *)&msg, sizeof(msg));
 	WARN_ON(len != sizeof(msg));
 
 	tasklet_schedule(&mbox->txq->tasklet);
 
 out:
-	spin_unlock(&mq->lock);
+	spin_unlock_bh(&mq->lock);
 	return ret;
 }
 EXPORT_SYMBOL(omap_mbox_msg_send);
Signature · 1 line
1.7.0
#15 lore raw

PATCH v4 4/5 OMAP: mailbox: send message in process context

Quoted 2 lines · Hari Kanigeri

Schedule the Tasklet to send only when mailbox fifo is full and there are pending messages in kifo, else send the message directly in the Process

Typo -> kifo

Quoted 42 lines

context. This would avoid needless scheduling of Tasklet for every message transfer

---
?arch/arm/plat-omap/mailbox.c | ? ?9 +++++++--
?1 files changed, 7 insertions(+), 2 deletions(-)

arch/arm/plat-omap/mailbox.c +6 -1
index 13698ab..f4177df 100644
@@ -92,20 +92,25 @@ int omap_mbox_msg_send(struct omap_mbox *mbox, mbox_msg_t msg)
 ? ? ? ?struct omap_mbox_queue *mq = mbox->txq;
 ? ? ? ?int ret = 0, len;
 
- ? ? ? spin_lock(&mq->lock);
+ ? ? ? spin_lock_bh(&mq->lock);
 
 ? ? ? ?if (kfifo_avail(&mq->fifo) < sizeof(msg)) {
 ? ? ? ? ? ? ? ?ret = -ENOMEM;
 ? ? ? ? ? ? ? ?goto out;
 ? ? ? ?}
 
+ ? ? ? if (kfifo_is_empty(&mq->fifo) && !__mbox_poll_for_space(mbox)) {
+ ? ? ? ? ? ? ? mbox_fifo_write(mbox, msg);
+ ? ? ? ? ? ? ? goto out;
+ ? ? ? }
+
 ? ? ? ?len = kfifo_in(&mq->fifo, (unsigned char *)&msg, sizeof(msg));
 ? ? ? ?WARN_ON(len != sizeof(msg));
 
 ? ? ? ?tasklet_schedule(&mbox->txq->tasklet);
 

?out:
- ? ? ? spin_unlock(&mq->lock);
+ ? ? ? spin_unlock_bh(&mq->lock);
? ? ? ?return ret;
?}
?EXPORT_SYMBOL(omap_mbox_msg_send);
--
1.7.0

#16 lore raw

PATCH v4 4/5 OMAP: mailbox: send message in process context

Quoted 5 lines · Charulatha
Quoted 2 lines · Hari Kanigeri

Schedule the Tasklet to send only when mailbox fifo is full and there are pending messages in kifo, else send the message directly in the Process

Typo -> kifo

Thanks :). My bad I I think overlooked Felipe's comment on my previous patch.

Thank you,
Best regards,
Hari Kanigeri

#17 lore raw

PATCH v4 5/5 OMAP: mailbox: add notification support for multiple readers

In reply to Hari Kanigeri

In the current mailbox driver, the mailbox internal pointer for callback can be directly manipulated by the Users, so a second User can easily corrupt the first user's callback pointer. The initial effort to correct this issue can be referred here:
https://patchwork.kernel.org/patch/107520/

Along with fixing the above stated issue, this patch adds the flexibility option to register notifications from multiple readers to the events received on a mailbox instance. The discussion regarding this can be referred here.
http://www.mail-archive.com/[email protected]/msg30671.html

---
 arch/arm/plat-omap/include/plat/mailbox.h |    7 +-
 arch/arm/plat-omap/mailbox.c              |  104 ++++++++++++++++-------------
 2 files changed, 62 insertions(+), 49 deletions(-)
arch/arm/plat-omap/include/plat/mailbox.h +4 -3
index 1655c61..839c6c3 100644
@@ -49,7 +49,6 @@ struct omap_mbox_queue {
 	struct kfifo		fifo;
 	struct work_struct	work;
 	struct tasklet_struct	tasklet;
-	int	(*callback)(void *);
 	struct omap_mbox	*mbox;
 	bool full;
 };
@@ -62,13 +61,15 @@ struct omap_mbox {
 	struct device		*dev;
 	void			*priv;
 	int			rev;
+	int			use_count;
+	struct blocking_notifier_head   notifier;
 };
 
 int omap_mbox_msg_send(struct omap_mbox *, mbox_msg_t msg);
 void omap_mbox_init_seq(struct omap_mbox *);
 
-struct omap_mbox *omap_mbox_get(const char *);
-void omap_mbox_put(struct omap_mbox *);
+struct omap_mbox *omap_mbox_get(const char *, struct notifier_block *nb);
+void omap_mbox_put(struct omap_mbox *mbox, struct notifier_block *nb);
 
 int omap_mbox_register(struct device *parent, struct omap_mbox **);
 int omap_mbox_unregister(void);
arch/arm/plat-omap/mailbox.c +58 -46

164 lines not on this page · raw message

Signature · 1 line
1.7.0
#18 lore raw

PATCH v4 5/5 OMAP: mailbox: add notification support for multiple readers

Quoted 12 lines · Hari Kanigeri

In the current mailbox driver, the mailbox internal pointer for callback can be directly manipulated by the Users, so a second User can easily corrupt the first user's callback pointer. The initial effort to correct this issue can be referred here:
https://patchwork.kernel.org/patch/107520/

Along with fixing the above stated issue, this patch ?adds the flexibility option to register notifications from
multiple readers to the events received on a mailbox instance. The discussion regarding this can be referred here.
http://www.mail-archive.com/linux-omap at vger.kernel.org/msg30671.html

Quoted 7 lines

---
?arch/arm/plat-omap/include/plat/mailbox.h | ? ?7 +- ?arch/arm/plat-omap/mailbox.c ? ? ? ? ? ? ?| ?104 ++++++++++++++++------------- ?2 files changed, 62 insertions(+), 49 deletions(-)

<<snip>>

Review int omap_mbox_register(struct device *parent, struct omap_mbox **list) · line 375

Quoted 5 lines
(quoted diff) +1 -0
@@ -363,10 +372,13 @@ int omap_mbox_register(struct device *parent, struct omap_mbox **list)
 ? ? ? ? ? ? ? ? ? ? ? ?ret = PTR_ERR(mbox->dev);
 ? ? ? ? ? ? ? ? ? ? ? ?goto err_out;
 ? ? ? ? ? ? ? ?}
+

Do this change in the patch where the below code was added.

Quoted 12 lines
(quoted diff) +0 -0
 ? ? ? ? ? ? ? ?if (cpu_is_omap44xx())
 ? ? ? ? ? ? ? ? ? ? ? ?mbox->rev = OMAP_MBOX_IP_VERSION_2;

? ? ? ? ? ? ? ?else
? ? ? ? ? ? ? ? ? ? ? ?mbox->rev = OMAP_MBOX_IP_LEGACY; +
+ ? ? ? ? ? ? ? BLOCKING_INIT_NOTIFIER_HEAD(&mbox->notifier); ? ? ? ?}
? ? ? ?return 0;

--
1.7.0

#19 lore raw

PATCH v4 5/5 OMAP: mailbox: add notification support for multiple readers

Quoted 23 lines · Charulatha
Quoted 12 lines · Hari Kanigeri

In the current mailbox driver, the mailbox internal pointer for callback can be directly manipulated by the Users, so a second User can easily corrupt the first user's callback pointer. The initial effort to correct this issue can be referred here:
https://patchwork.kernel.org/patch/107520/

Along with fixing the above stated issue, this patch ?adds the flexibility option to register notifications from
multiple readers to the events received on a mailbox instance. The discussion regarding this can be referred here.
http://www.mail-archive.com/linux-omap at vger.kernel.org/msg30671.html

Quoted 7 lines

---
?arch/arm/plat-omap/include/plat/mailbox.h | ? ?7 +- ?arch/arm/plat-omap/mailbox.c ? ? ? ? ? ? ?| ?104 ++++++++++++++++------------- ?2 files changed, 62 insertions(+), 49 deletions(-)

<<snip>>

Quoted 7 lines

Review int omap_mbox_register(struct device *parent, struct omap_mbox **list) · line 375

Quoted 5 lines
(quoted diff) +1 -0
@@ -363,10 +372,13 @@ int omap_mbox_register(struct device *parent, struct omap_mbox **list)
 ? ? ? ? ? ? ? ? ? ? ? ?ret = PTR_ERR(mbox->dev);
 ? ? ? ? ? ? ? ? ? ? ? ?goto err_out;
 ? ? ? ? ? ? ? ?}
+

Do this change in the patch where the below code was added.

I am sorry I didn't get you. you mean BLOCKING_INIT_NOTIFIER_HEAD line ? yes, it is needed for this patch.

Quoted 14 lines
Quoted 13 lines
(quoted diff) +0 -0
 ? ? ? ? ? ? ? ?if (cpu_is_omap44xx())
 ? ? ? ? ? ? ? ? ? ? ? ?mbox->rev = OMAP_MBOX_IP_VERSION_2;

? ? ? ? ? ? ? ?else
? ? ? ? ? ? ? ? ? ? ? ?mbox->rev = OMAP_MBOX_IP_LEGACY; +
+ ? ? ? ? ? ? ? BLOCKING_INIT_NOTIFIER_HEAD(&mbox->notifier); ? ? ? ?}
? ? ? ?return 0;

--
1.7.0

Signature · 3 lines
Thank you,
Best regards,
Hari Kanigeri
#20 lore raw

PATCH v4 5/5 OMAP: mailbox: add notification support for multiple readers

Quoted 32 lines · Hari
Quoted 23 lines · Charulatha
Quoted 12 lines · Hari Kanigeri

In the current mailbox driver, the mailbox internal pointer for callback can be directly manipulated by the Users, so a second User can easily corrupt the first user's callback pointer. The initial effort to correct this issue can be referred here:
https://patchwork.kernel.org/patch/107520/

Along with fixing the above stated issue, this patch ?adds the flexibility option to register notifications from
multiple readers to the events received on a mailbox instance. The discussion regarding this can be referred here.
http://www.mail-archive.com/linux-omap at vger.kernel.org/msg30671.html

Quoted 7 lines

---
?arch/arm/plat-omap/include/plat/mailbox.h | ? ?7 +- ?arch/arm/plat-omap/mailbox.c ? ? ? ? ? ? ?| ?104 ++++++++++++++++------------- ?2 files changed, 62 insertions(+), 49 deletions(-)

<<snip>>

Quoted 7 lines

Review int omap_mbox_register(struct device *parent, struct omap_mbox **list) · line 375

Quoted 5 lines
(quoted diff) +1 -0
@@ -363,10 +372,13 @@ int omap_mbox_register(struct device *parent, struct omap_mbox **list)
 ? ? ? ? ? ? ? ? ? ? ? ?ret = PTR_ERR(mbox->dev);
 ? ? ? ? ? ? ? ? ? ? ? ?goto err_out;
 ? ? ? ? ? ? ? ?}
+

Do this change in the patch where the below code was added.

Quoted 2 lines

I am sorry I didn't get you. you mean BLOCKING_INIT_NOTIFIER_HEAD line ? yes, it is needed for this patch.

No. I was saying that adding a new line could be done in the patch where this piece of code (below 4 lines) was introduced. Sorry if that was not clear :-(

Quoted 22 lines
Quoted 10 lines
Quoted 10 lines
(quoted diff) +0 -0
 ? ? ? ? ? ? ? ?if (cpu_is_omap44xx())
 ? ? ? ? ? ? ? ? ? ? ? ?mbox->rev = OMAP_MBOX_IP_VERSION_2;

? ? ? ? ? ? ? ?else
? ? ? ? ? ? ? ? ? ? ? ?mbox->rev = OMAP_MBOX_IP_LEGACY; +
+ ? ? ? ? ? ? ? BLOCKING_INIT_NOTIFIER_HEAD(&mbox->notifier); ? ? ? ?}
? ? ? ?return 0;

Quoted 4 lines
Quoted 3 lines

--
1.7.0

--
Thank you,
Best regards,
Hari Kanigeri

#21 lore raw

PATCH v4 5/5 OMAP: mailbox: add notification support for multiple readers

Quoted 33 lines · Charulatha
Quoted 32 lines · Hari
Quoted 30 lines · Charulatha
Quoted 12 lines · Hari Kanigeri

In the current mailbox driver, the mailbox internal pointer for callback can be directly manipulated by the Users, so a second User can easily corrupt the first user's callback pointer. The initial effort to correct this issue can be referred here:
https://patchwork.kernel.org/patch/107520/

Along with fixing the above stated issue, this patch ?adds the flexibility option to register notifications from
multiple readers to the events received on a mailbox instance. The discussion regarding this can be referred here.
http://www.mail-archive.com/linux-omap at vger.kernel.org/msg30671.html

Quoted 7 lines

---
?arch/arm/plat-omap/include/plat/mailbox.h | ? ?7 +- ?arch/arm/plat-omap/mailbox.c ? ? ? ? ? ? ?| ?104 ++++++++++++++++------------- ?2 files changed, 62 insertions(+), 49 deletions(-)

<<snip>>

Review int omap_mbox_register(struct device *parent, struct omap_mbox **list) · line 375

Quoted 5 lines
(quoted diff) +1 -0
@@ -363,10 +372,13 @@ int omap_mbox_register(struct device *parent, struct omap_mbox **list)
 ? ? ? ? ? ? ? ? ? ? ? ?ret = PTR_ERR(mbox->dev);
 ? ? ? ? ? ? ? ? ? ? ? ?goto err_out;
 ? ? ? ? ? ? ? ?}
+

Do this change in the patch where the below code was added.

Quoted 5 lines
Quoted 2 lines

I am sorry I didn't get you. you mean BLOCKING_INIT_NOTIFIER_HEAD line ? yes, it is needed for this patch.

No. I was saying that adding a new line could be done in the patch where this piece of code (below 4 lines) was introduced. Sorry if that was not clear :-(

Got it. extra line can be removed.

Thank you,
Best regards,
Hari Kanigeri