Skip to contentSkip to conversation

hw/ssi: imx_spi: Fix various bugs in the imx_spi model

PATCH Superseded Bin Meng 3 participants 18 messages Jan 10, 2021 – Jan 12, 2021 Arm i.MX and Layerscape

v4 of 9 Superseded Latest is v9 →

lore
Share & tools
lore

Bin Meng

author
#0 lore raw

v4 cover hw/ssi: imx_spi: Fix various bugs in the imx_spi model

From: Bin Meng <[email protected]>

This series fixes a bunch of bugs in current implementation of the imx spi controller, including the following issues:

- chip select signal was not lower down when spi controller is disabled
- remove imx_spi_update_irq() in imx_spi_reset()
- transfer incorrect data when the burst length is larger than 32 bit
- spi controller tx and rx fifo endianness is incorrect

Tested with upstream U-Boot v2020.10 (polling mode) and VxWorks 7 (interrupt mode).

Changes in v4:
- adujst the patch 2,3 order
- rename imx_spi_soft_reset() to imx_spi_hard_reset() to avoid confusion
- s/normal/common/ in the commit message
- log the burst length value in the log message

Changes in v3:
- new patch: remove imx_spi_update_irq() in imx_spi_reset()
- Move the chip selects disable out of imx_spi_reset()
- new patch: log unimplemented burst length
- Simplify the tx fifo endianness handling

Changes in v2:
- Fix the "Fixes" tag in the commit message
- Use ternary operator as Philippe suggested

Bin Meng (5):
hw/ssi: imx_spi: Use a macro for number of chip selects supported hw/ssi: imx_spi: Remove imx_spi_update_irq() in imx_spi_reset() hw/ssi: imx_spi: Log unimplemented burst length hw/ssi: imx_spi: Correct the burst length > 32 bit transfer logic hw/ssi: imx_spi: Correct tx and rx fifo endianness

Xuzhou Cheng (1):
hw/ssi: imx_spi: Disable chip selects when controller is disabled

 include/hw/ssi/imx_spi.h |  5 ++++-
 hw/ssi/imx_spi.c         | 42 ++++++++++++++++++++++++++++------------
 2 files changed, 34 insertions(+), 13 deletions(-)
Signature · 1 line
2.25.1
author
#1 lore raw

PATCH v4 1/6 hw/ssi: imx_spi: Use a macro for number of chip selects supported

From: Bin Meng <[email protected]>

Avoid using a magic number (4) everywhere for the number of chip selects supported.

---

(no changes since v1)

 include/hw/ssi/imx_spi.h | 5 ++++-
 hw/ssi/imx_spi.c         | 4 ++--
 2 files changed, 6 insertions(+), 3 deletions(-)
include/hw/ssi/imx_spi.h +4 -1
index b82b17f364..eeaf49bbac 100644
@@ -77,6 +77,9 @@
 
 #define EXTRACT(value, name) extract32(value, name##_SHIFT, name##_LENGTH)
 
+/* number of chip selects supported */
+#define ECSPI_NUM_CS 4
+
 #define TYPE_IMX_SPI "imx.spi"
 OBJECT_DECLARE_SIMPLE_TYPE(IMXSPIState, IMX_SPI)
 
@@ -89,7 +92,7 @@ struct IMXSPIState {
 
     qemu_irq irq;
 
-    qemu_irq cs_lines[4];
+    qemu_irq cs_lines[ECSPI_NUM_CS];
 
     SSIBus *bus;
 
hw/ssi/imx_spi.c +2 -2
index d8885ae454..e605049a21 100644
@@ -361,7 +361,7 @@ static void imx_spi_write(void *opaque, hwaddr offset, uint64_t value,
 
             /* We are in master mode */
 
-            for (i = 0; i < 4; i++) {
+            for (i = 0; i < ECSPI_NUM_CS; i++) {
                 qemu_set_irq(s->cs_lines[i],
                              i == imx_spi_selected_channel(s) ? 0 : 1);
             }
@@ -424,7 +424,7 @@ static void imx_spi_realize(DeviceState *dev, Error **errp)
     sysbus_init_mmio(SYS_BUS_DEVICE(dev), &s->iomem);
     sysbus_init_irq(SYS_BUS_DEVICE(dev), &s->irq);
 
-    for (i = 0; i < 4; ++i) {
+    for (i = 0; i < ECSPI_NUM_CS; ++i) {
         sysbus_init_irq(SYS_BUS_DEVICE(dev), &s->cs_lines[i]);
     }
 
Signature · 1 line
2.25.1
author
#2 lore raw

PATCH v4 2/6 hw/ssi: imx_spi: Remove imx_spi_update_irq() in imx_spi_reset()

In reply to Bin Meng

From: Bin Meng <[email protected]>

Usually the approach is that the device on the other end of the line is going to reset its state anyway, so there's no need to actively signal an irq line change during the reset hook.

Move imx_spi_update_irq() out of imx_spi_reset(), to a new function imx_spi_hard_reset() that is called when the controller is disabled.

---

Changes in v4:
- adujst the patch 2,3 order
- rename imx_spi_soft_reset() to imx_spi_hard_reset() to avoid confusion

Changes in v3:
- new patch: remove imx_spi_update_irq() in imx_spi_reset()

 hw/ssi/imx_spi.c | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)
hw/ssi/imx_spi.c +10 -4
index e605049a21..2c4c5ec1b8 100644
@@ -241,11 +241,16 @@ static void imx_spi_reset(DeviceState *dev)
     imx_spi_rxfifo_reset(s);
     imx_spi_txfifo_reset(s);
 
-    imx_spi_update_irq(s);
-
     s->burst_length = 0;
 }
 
+static void imx_spi_hard_reset(IMXSPIState *s)
+{
+    imx_spi_reset(DEVICE(s));
+
+    imx_spi_update_irq(s);
+}
+
 static uint64_t imx_spi_read(void *opaque, hwaddr offset, unsigned size)
 {
     uint32_t value = 0;
@@ -351,8 +356,9 @@ static void imx_spi_write(void *opaque, hwaddr offset, uint64_t value,
         s->regs[ECSPI_CONREG] = value;
 
         if (!imx_spi_is_enabled(s)) {
-            /* device is disabled, so this is a reset */
-            imx_spi_reset(DEVICE(s));
+            /* device is disabled, so this is a hard reset */
+            imx_spi_hard_reset(s);
+
             return;
         }
 
Signature · 1 line
2.25.1
#3 lore raw

Re: [PATCH v4 2/6] hw/ssi: imx_spi: Remove imx_spi_update_irq() in imx_spi_reset()

Quoted 48 lines · Bin Meng

From: Bin Meng <[email protected]>

Usually the approach is that the device on the other end of the line is going to reset its state anyway, so there's no need to actively signal an irq line change during the reset hook.

Move imx_spi_update_irq() out of imx_spi_reset(), to a new function imx_spi_hard_reset() that is called when the controller is disabled.

---

Changes in v4:
- adujst the patch 2,3 order
- rename imx_spi_soft_reset() to imx_spi_hard_reset() to avoid confusion

Changes in v3:
- new patch: remove imx_spi_update_irq() in imx_spi_reset()

 hw/ssi/imx_spi.c | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)
hw/ssi/imx_spi.c +7 -2
index e605049a21..2c4c5ec1b8 100644
@@ -241,11 +241,16 @@ static void imx_spi_reset(DeviceState *dev)
     imx_spi_rxfifo_reset(s);
     imx_spi_txfifo_reset(s);
 
-    imx_spi_update_irq(s);
-
     s->burst_length = 0;
 }
 
+static void imx_spi_hard_reset(IMXSPIState *s)
+{
+    imx_spi_reset(DEVICE(s));
+
+    imx_spi_update_irq(s);
+}
+
 static uint64_t imx_spi_read(void *opaque, hwaddr offset, unsigned size)
 {
     uint32_t value = 0;
@@ -351,8 +356,9 @@ static void imx_spi_write(void *opaque, hwaddr offset, uint64_t value,
         s->regs[ECSPI_CONREG] = value;

hw/ssi/imx_spi.c static void imx_spi_write(void *opaque, hwaddr offset, uint64_t value, · line 364

Quoted 10 lines
hw/ssi/imx_spi.c +3 -2
 
         if (!imx_spi_is_enabled(s)) {
-            /* device is disabled, so this is a reset */
-            imx_spi_reset(DEVICE(s));
+            /* device is disabled, so this is a hard reset */
+            imx_spi_hard_reset(s);
+
             return;
         }
 

Almost good :)

DeviceReset handler is a hard reset, so you need:

  dc->reset = imx_spi_hard_reset;

Thus you also need this function prototype:

void imx_spi_hard_reset(DeviceState *dev)

Eventually renaming imx_spi_reset() -> imx_spi_soft_reset() will make things even easier.

Regards,

Phil.

author
#4 lore raw

Re: [PATCH v4 2/6] hw/ssi: imx_spi: Remove imx_spi_update_irq() in imx_spi_reset()

Hi Philippe,

Quoted 65 lines · Philippe Mathieu-Daudé
Quoted 48 lines · Bin Meng

From: Bin Meng <[email protected]>

Usually the approach is that the device on the other end of the line is going to reset its state anyway, so there's no need to actively signal an irq line change during the reset hook.

Move imx_spi_update_irq() out of imx_spi_reset(), to a new function imx_spi_hard_reset() that is called when the controller is disabled.

---

Changes in v4:
- adujst the patch 2,3 order
- rename imx_spi_soft_reset() to imx_spi_hard_reset() to avoid confusion

Changes in v3:
- new patch: remove imx_spi_update_irq() in imx_spi_reset()

 hw/ssi/imx_spi.c | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)
hw/ssi/imx_spi.c +7 -2
index e605049a21..2c4c5ec1b8 100644
@@ -241,11 +241,16 @@ static void imx_spi_reset(DeviceState *dev)
     imx_spi_rxfifo_reset(s);
     imx_spi_txfifo_reset(s);
 
-    imx_spi_update_irq(s);
-
     s->burst_length = 0;
 }
 
+static void imx_spi_hard_reset(IMXSPIState *s)
+{
+    imx_spi_reset(DEVICE(s));
+
+    imx_spi_update_irq(s);
+}
+
 static uint64_t imx_spi_read(void *opaque, hwaddr offset, unsigned size)
 {
     uint32_t value = 0;
@@ -351,8 +356,9 @@ static void imx_spi_write(void *opaque, hwaddr offset, uint64_t value,
         s->regs[ECSPI_CONREG] = value;

hw/ssi/imx_spi.c static void imx_spi_write(void *opaque, hwaddr offset, uint64_t value, · line 364

Quoted 10 lines
hw/ssi/imx_spi.c +3 -2
 
         if (!imx_spi_is_enabled(s)) {
-            /* device is disabled, so this is a reset */
-            imx_spi_reset(DEVICE(s));
+            /* device is disabled, so this is a hard reset */
+            imx_spi_hard_reset(s);
+
             return;
         }
 

Almost good :)

DeviceReset handler is a hard reset, so you need:

Quoted 8 lines
  dc->reset = imx_spi_hard_reset;

Thus you also need this function prototype:

void imx_spi_hard_reset(DeviceState *dev)

Eventually renaming imx_spi_reset() -> imx_spi_soft_reset() will make things even easier.

Now I am confused. The v3 patch did the above already, but you mentioned that usually the hard reset includes the soft reset. But in the hard reset the imx_spi_update_irq() does not need to be called.

Regards,
Bin

#5 lore raw

Re: [PATCH v4 2/6] hw/ssi: imx_spi: Remove imx_spi_update_irq() in imx_spi_reset()

In reply to Bin Meng
Quoted 48 lines · Bin Meng

From: Bin Meng <[email protected]>

Usually the approach is that the device on the other end of the line is going to reset its state anyway, so there's no need to actively signal an irq line change during the reset hook.

Move imx_spi_update_irq() out of imx_spi_reset(), to a new function imx_spi_hard_reset() that is called when the controller is disabled.

---

Changes in v4:
- adujst the patch 2,3 order
- rename imx_spi_soft_reset() to imx_spi_hard_reset() to avoid confusion

Changes in v3:
- new patch: remove imx_spi_update_irq() in imx_spi_reset()

 hw/ssi/imx_spi.c | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)
hw/ssi/imx_spi.c +7 -2
index e605049a21..2c4c5ec1b8 100644
@@ -241,11 +241,16 @@ static void imx_spi_reset(DeviceState *dev)
     imx_spi_rxfifo_reset(s);
     imx_spi_txfifo_reset(s);
 
-    imx_spi_update_irq(s);
-
     s->burst_length = 0;
 }
 
+static void imx_spi_hard_reset(IMXSPIState *s)
+{
+    imx_spi_reset(DEVICE(s));
+
+    imx_spi_update_irq(s);
+}
+
 static uint64_t imx_spi_read(void *opaque, hwaddr offset, unsigned size)
 {
     uint32_t value = 0;

hw/ssi/imx_spi.c static void imx_spi_write(void *opaque, hwaddr offset, uint64_t value, · line 363

Quoted 10 lines
hw/ssi/imx_spi.c +3 -2
@@ -351,8 +356,9 @@ static void imx_spi_write(void *opaque, hwaddr offset, uint64_t value,
         s->regs[ECSPI_CONREG] = value;
 
         if (!imx_spi_is_enabled(s)) {
-            /* device is disabled, so this is a reset */
-            imx_spi_reset(DEVICE(s));
+            /* device is disabled, so this is a hard reset */
+            imx_spi_hard_reset(s);
+
             return;
         }

The function of the code is correct, but you seem to have the function naming backwards here. Generally:
* soft reset == the reset triggered by the register write
* hard reset == power-on reset == the dc->reset function

I think this is what Philippe was trying to say.

thanks
-- PMM

author
#6 lore raw

Re: [PATCH v4 2/6] hw/ssi: imx_spi: Remove imx_spi_update_irq() in imx_spi_reset()

Quoted 66 lines · Peter Maydell
Quoted 48 lines · Bin Meng

From: Bin Meng <[email protected]>

Usually the approach is that the device on the other end of the line is going to reset its state anyway, so there's no need to actively signal an irq line change during the reset hook.

Move imx_spi_update_irq() out of imx_spi_reset(), to a new function imx_spi_hard_reset() that is called when the controller is disabled.

---

Changes in v4:
- adujst the patch 2,3 order
- rename imx_spi_soft_reset() to imx_spi_hard_reset() to avoid confusion

Changes in v3:
- new patch: remove imx_spi_update_irq() in imx_spi_reset()

 hw/ssi/imx_spi.c | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)
hw/ssi/imx_spi.c +7 -2
index e605049a21..2c4c5ec1b8 100644
@@ -241,11 +241,16 @@ static void imx_spi_reset(DeviceState *dev)
     imx_spi_rxfifo_reset(s);
     imx_spi_txfifo_reset(s);
 
-    imx_spi_update_irq(s);
-
     s->burst_length = 0;
 }
 
+static void imx_spi_hard_reset(IMXSPIState *s)
+{
+    imx_spi_reset(DEVICE(s));
+
+    imx_spi_update_irq(s);
+}
+
 static uint64_t imx_spi_read(void *opaque, hwaddr offset, unsigned size)
 {
     uint32_t value = 0;

hw/ssi/imx_spi.c static void imx_spi_write(void *opaque, hwaddr offset, uint64_t value, · line 363

Quoted 10 lines
hw/ssi/imx_spi.c +3 -2
@@ -351,8 +356,9 @@ static void imx_spi_write(void *opaque, hwaddr offset, uint64_t value,
         s->regs[ECSPI_CONREG] = value;
 
         if (!imx_spi_is_enabled(s)) {
-            /* device is disabled, so this is a reset */
-            imx_spi_reset(DEVICE(s));
+            /* device is disabled, so this is a hard reset */
+            imx_spi_hard_reset(s);
+
             return;
         }

The function of the code is correct, but you seem to have the function naming backwards here. Generally:
* soft reset == the reset triggered by the register write
* hard reset == power-on reset == the dc->reset function

Quoted 1 line

I think this is what Philippe was trying to say.

Philippe said: "Hmm usually hard reset include soft reset."

Since we are moving imx_spi_update_irq() out of imx_spi_reset() to a new function called imx_spi_soft_reset() (what I did in v3), that confused him (and I felt the same thing), so I renamed imx_spi_soft_reset() to imx_spi_hard_reset() in v4..

Regards,
Bin

#7 lore raw

Re: [PATCH v4 2/6] hw/ssi: imx_spi: Remove imx_spi_update_irq() in imx_spi_reset()

Quoted 68 lines · Bin Meng
Quoted 66 lines · Peter Maydell
Quoted 48 lines · Bin Meng

From: Bin Meng <[email protected]>

Usually the approach is that the device on the other end of the line is going to reset its state anyway, so there's no need to actively signal an irq line change during the reset hook.

Move imx_spi_update_irq() out of imx_spi_reset(), to a new function imx_spi_hard_reset() that is called when the controller is disabled.

---

Changes in v4:
- adujst the patch 2,3 order
- rename imx_spi_soft_reset() to imx_spi_hard_reset() to avoid confusion

Changes in v3:
- new patch: remove imx_spi_update_irq() in imx_spi_reset()

 hw/ssi/imx_spi.c | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)
hw/ssi/imx_spi.c +7 -2
index e605049a21..2c4c5ec1b8 100644
@@ -241,11 +241,16 @@ static void imx_spi_reset(DeviceState *dev)
     imx_spi_rxfifo_reset(s);
     imx_spi_txfifo_reset(s);
 
-    imx_spi_update_irq(s);
-
     s->burst_length = 0;
 }
 
+static void imx_spi_hard_reset(IMXSPIState *s)
+{
+    imx_spi_reset(DEVICE(s));
+
+    imx_spi_update_irq(s);
+}
+
 static uint64_t imx_spi_read(void *opaque, hwaddr offset, unsigned size)
 {
     uint32_t value = 0;

hw/ssi/imx_spi.c static void imx_spi_write(void *opaque, hwaddr offset, uint64_t value, · line 363

Quoted 10 lines
hw/ssi/imx_spi.c +3 -2
@@ -351,8 +356,9 @@ static void imx_spi_write(void *opaque, hwaddr offset, uint64_t value,
         s->regs[ECSPI_CONREG] = value;
 
         if (!imx_spi_is_enabled(s)) {
-            /* device is disabled, so this is a reset */
-            imx_spi_reset(DEVICE(s));
+            /* device is disabled, so this is a hard reset */
+            imx_spi_hard_reset(s);
+
             return;
         }

The function of the code is correct, but you seem to have the function naming backwards here. Generally:
* soft reset == the reset triggered by the register write
* hard reset == power-on reset == the dc->reset function

Quoted 3 lines
Quoted 1 line

I think this is what Philippe was trying to say.

Philippe said: "Hmm usually hard reset include soft reset."

True in hardware, but for QEMU there are some things we don't want to do in what we would call a hard or power-on reset.

Quoted 4 lines

Since we are moving imx_spi_update_irq() out of imx_spi_reset() to a new function called imx_spi_soft_reset() (what I did in v3), that confused him (and I felt the same thing), so I renamed imx_spi_soft_reset() to imx_spi_hard_reset() in v4..

thanks
-- PMM

author
#8 lore raw

Re: [PATCH v4 2/6] hw/ssi: imx_spi: Remove imx_spi_update_irq() in imx_spi_reset()

Quoted 70 lines · Peter Maydell
Quoted 68 lines · Bin Meng
Quoted 66 lines · Peter Maydell
Quoted 48 lines · Bin Meng

From: Bin Meng <[email protected]>

Usually the approach is that the device on the other end of the line is going to reset its state anyway, so there's no need to actively signal an irq line change during the reset hook.

Move imx_spi_update_irq() out of imx_spi_reset(), to a new function imx_spi_hard_reset() that is called when the controller is disabled.

---

Changes in v4:
- adujst the patch 2,3 order
- rename imx_spi_soft_reset() to imx_spi_hard_reset() to avoid confusion

Changes in v3:
- new patch: remove imx_spi_update_irq() in imx_spi_reset()

 hw/ssi/imx_spi.c | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)
hw/ssi/imx_spi.c +7 -2
index e605049a21..2c4c5ec1b8 100644
@@ -241,11 +241,16 @@ static void imx_spi_reset(DeviceState *dev)
     imx_spi_rxfifo_reset(s);
     imx_spi_txfifo_reset(s);
 
-    imx_spi_update_irq(s);
-
     s->burst_length = 0;
 }
 
+static void imx_spi_hard_reset(IMXSPIState *s)
+{
+    imx_spi_reset(DEVICE(s));
+
+    imx_spi_update_irq(s);
+}
+
 static uint64_t imx_spi_read(void *opaque, hwaddr offset, unsigned size)
 {
     uint32_t value = 0;

hw/ssi/imx_spi.c static void imx_spi_write(void *opaque, hwaddr offset, uint64_t value, · line 363

Quoted 10 lines
hw/ssi/imx_spi.c +3 -2
@@ -351,8 +356,9 @@ static void imx_spi_write(void *opaque, hwaddr offset, uint64_t value,
         s->regs[ECSPI_CONREG] = value;
 
         if (!imx_spi_is_enabled(s)) {
-            /* device is disabled, so this is a reset */
-            imx_spi_reset(DEVICE(s));
+            /* device is disabled, so this is a hard reset */
+            imx_spi_hard_reset(s);
+
             return;
         }

The function of the code is correct, but you seem to have the function naming backwards here. Generally:
* soft reset == the reset triggered by the register write
* hard reset == power-on reset == the dc->reset function

Quoted 7 lines
Quoted 3 lines
Quoted 1 line

I think this is what Philippe was trying to say.

Philippe said: "Hmm usually hard reset include soft reset."

True in hardware, but for QEMU there are some things we don't want to do in what we would call a hard or power-on reset.

OK, will revert to use imx_spi_soft_reset().

Quoted 4 lines
Quoted 4 lines

Since we are moving imx_spi_update_irq() out of imx_spi_reset() to a new function called imx_spi_soft_reset() (what I did in v3), that confused him (and I felt the same thing), so I renamed imx_spi_soft_reset() to imx_spi_hard_reset() in v4..

Regards,
Bin

#9 lore raw

Re: [PATCH v4 2/6] hw/ssi: imx_spi: Remove imx_spi_update_irq() in imx_spi_reset()

Hi Ben,

Quoted 71 lines · Bin Meng
Quoted 70 lines · Peter Maydell
Quoted 68 lines · Bin Meng
Quoted 66 lines · Peter Maydell
Quoted 48 lines · Bin Meng

From: Bin Meng <[email protected]>

Usually the approach is that the device on the other end of the line is going to reset its state anyway, so there's no need to actively signal an irq line change during the reset hook.

Move imx_spi_update_irq() out of imx_spi_reset(), to a new function imx_spi_hard_reset() that is called when the controller is disabled.

---

Changes in v4:
- adujst the patch 2,3 order
- rename imx_spi_soft_reset() to imx_spi_hard_reset() to avoid confusion

Changes in v3:
- new patch: remove imx_spi_update_irq() in imx_spi_reset()

 hw/ssi/imx_spi.c | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)
hw/ssi/imx_spi.c +7 -2
index e605049a21..2c4c5ec1b8 100644
@@ -241,11 +241,16 @@ static void imx_spi_reset(DeviceState *dev)
     imx_spi_rxfifo_reset(s);
     imx_spi_txfifo_reset(s);
 
-    imx_spi_update_irq(s);
-
     s->burst_length = 0;
 }
 
+static void imx_spi_hard_reset(IMXSPIState *s)
+{
+    imx_spi_reset(DEVICE(s));
+
+    imx_spi_update_irq(s);
+}
+
 static uint64_t imx_spi_read(void *opaque, hwaddr offset, unsigned size)
 {
     uint32_t value = 0;

hw/ssi/imx_spi.c static void imx_spi_write(void *opaque, hwaddr offset, uint64_t value, · line 363

Quoted 10 lines
hw/ssi/imx_spi.c +3 -2
@@ -351,8 +356,9 @@ static void imx_spi_write(void *opaque, hwaddr offset, uint64_t value,
         s->regs[ECSPI_CONREG] = value;
 
         if (!imx_spi_is_enabled(s)) {
-            /* device is disabled, so this is a reset */
-            imx_spi_reset(DEVICE(s));
+            /* device is disabled, so this is a hard reset */
+            imx_spi_hard_reset(s);
+
             return;
         }

The function of the code is correct, but you seem to have the function naming backwards here. Generally:
* soft reset == the reset triggered by the register write
* hard reset == power-on reset == the dc->reset function

Quoted 6 lines
Quoted 6 lines
Quoted 3 lines
Quoted 1 line

I think this is what Philippe was trying to say.

Philippe said: "Hmm usually hard reset include soft reset."

True in hardware, but for QEMU there are some things we don't want to do in what we would call a hard or power-on reset.

Sorry for the confusion. I guess you understood me well, but I was wrong. Anyhow I'll try to sort this discussion out with my English teacher so the next time such confusion doesn't happen again.

Thanks,

Phil.

Quoted 10 lines

OK, will revert to use imx_spi_soft_reset().

Quoted 4 lines
Quoted 4 lines

Since we are moving imx_spi_update_irq() out of imx_spi_reset() to a new function called imx_spi_soft_reset() (what I did in v3), that confused him (and I felt the same thing), so I renamed imx_spi_soft_reset() to imx_spi_hard_reset() in v4..

Regards,
Bin

author
#10 lore raw

PATCH v4 3/6 hw/ssi: imx_spi: Disable chip selects when controller is disabled

In reply to Bin Meng

From: Xuzhou Cheng <[email protected]>

When a write to ECSPI_CONREG register to disable the SPI controller, imx_spi_reset() is called to reset the controller, but chip select lines should have been disabled, otherwise the state machine of any devices (e.g.: SPI flashes) connected to the SPI master is stuck to its last state and responds incorrectly to any follow-up commands.

Fixes c906a3a01582 ("i.MX: Add the Freescale SPI Controller")

---

(no changes since v3)

Changes in v3:
- Move the chip selects disable out of imx_spi_reset()

Changes in v2:
- Fix the "Fixes" tag in the commit message

 hw/ssi/imx_spi.c | 6 ++++++
 1 file changed, 6 insertions(+)
hw/ssi/imx_spi.c +6 -0
index 2c4c5ec1b8..168ea95440 100644
@@ -246,9 +246,15 @@ static void imx_spi_reset(DeviceState *dev)
 
 static void imx_spi_hard_reset(IMXSPIState *s)
 {
+    int i;
+
     imx_spi_reset(DEVICE(s));
 
     imx_spi_update_irq(s);
+
+    for (i = 0; i < ECSPI_NUM_CS; i++) {
+        qemu_set_irq(s->cs_lines[i], 1);
+    }
 }
 
 static uint64_t imx_spi_read(void *opaque, hwaddr offset, unsigned size)
Signature · 1 line
2.25.1
Reviewed-by on v4 3/6
#11 lore raw

Re: [PATCH v4 3/6] hw/ssi: imx_spi: Disable chip selects when controller is disabled

Quoted 16 lines · Bin Meng

From: Xuzhou Cheng <[email protected]>

When a write to ECSPI_CONREG register to disable the SPI controller, imx_spi_reset() is called to reset the controller, but chip select lines should have been disabled, otherwise the state machine of any devices (e.g.: SPI flashes) connected to the SPI master is stuck to its last state and responds incorrectly to any follow-up commands.

Fixes c906a3a01582 ("i.MX: Add the Freescale SPI Controller")

---

(no changes since v3)

Quoted 8 lines

Changes in v3:
- Move the chip selects disable out of imx_spi_reset()

Changes in v2:
- Fix the "Fixes" tag in the commit message

 hw/ssi/imx_spi.c | 6 ++++++
 1 file changed, 6 insertions(+)
author
#12 lore raw

PATCH v4 4/6 hw/ssi: imx_spi: Log unimplemented burst length

In reply to Bin Meng

From: Bin Meng <[email protected]>

Current implementation of the imx spi controller expects the burst length to be multiple of 8, which is the most common use case.

In case the burst length is not what we expect, log it to give user a chance to notice it.

---

Changes in v4:
- s/normal/common/ in the commit message
- log the burst length value in the log message

Changes in v3:
- new patch: log unimplemented burst length

 hw/ssi/imx_spi.c | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)
hw/ssi/imx_spi.c +10 -1
index 168ea95440..7f81b329a4 100644
@@ -128,7 +128,16 @@ static uint8_t imx_spi_selected_channel(IMXSPIState *s)
 
 static uint32_t imx_spi_burst_length(IMXSPIState *s)
 {
-    return EXTRACT(s->regs[ECSPI_CONREG], ECSPI_CONREG_BURST_LENGTH) + 1;
+    uint32_t burst;
+
+    burst = EXTRACT(s->regs[ECSPI_CONREG], ECSPI_CONREG_BURST_LENGTH) + 1;
+    if (burst % 8) {
+        qemu_log_mask(LOG_UNIMP,
+                      "[%s]%s: burst length (%d) not multiple of 8!\n",
+                      TYPE_IMX_SPI, __func__, burst);
+    }
+
+    return burst;
 }
 
 static bool imx_spi_is_enabled(IMXSPIState *s)
Signature · 1 line
2.25.1
author
#13 lore raw

PATCH v4 5/6 hw/ssi: imx_spi: Correct the burst length > 32 bit transfer logic

In reply to Bin Meng

From: Bin Meng <[email protected]>

For the ECSPIx_CONREG register BURST_LENGTH field, the manual says:

0x020 A SPI burst contains the 1 LSB in first word and all 32 bits in second word. 0x021 A SPI burst contains the 2 LSB in first word and all 32 bits in second word.

Current logic uses either s->burst_length or 32, whichever smaller, to determine how many bits it should read from the tx fifo each time. For example, for a 48 bit burst length, current logic transfers the first 32 bit from the first word in the tx fifo, followed by a 16 bit from the second word in the tx fifo, which is wrong. The correct logic should be: transfer the first 16 bit from the first word in the tx fifo, followed by a 32 bit from the second word in the tx fifo.

With this change, SPI flash can be successfully probed by U-Boot on imx6 sabrelite board.

=> sf probe
SF: Detected sst25vf016b with page size 256 Bytes, erase size 4 KiB, total 2 MiB

Fixes c906a3a01582 ("i.MX: Add the Freescale SPI Controller")

---

(no changes since v2)

Changes in v2:
- Use ternary operator as Philippe suggested

 hw/ssi/imx_spi.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
hw/ssi/imx_spi.c +1 -1
index 7f81b329a4..47c8a0f572 100644
@@ -187,7 +187,7 @@ static void imx_spi_flush_txfifo(IMXSPIState *s)
 
         DPRINTF("data tx:0x%08x\n", tx);
 
-        tx_burst = MIN(s->burst_length, 32);
+        tx_burst = (s->burst_length % 32) ? : 32;
 
         rx = 0;
 
Signature · 1 line
2.25.1
author
#14 lore raw

PATCH v4 6/6 hw/ssi: imx_spi: Correct tx and rx fifo endianness

In reply to Bin Meng

From: Bin Meng <[email protected]>

The endianness of data exchange between tx and rx fifo is incorrect. Earlier bytes are supposed to show up on MSB and later bytes on LSB,
ie: in big endian. The manual does not explicitly say this, but the U-Boot and Linux driver codes have a swap on the data transferred to tx fifo and from rx fifo.

With this change, U-Boot read from / write to SPI flash tests pass.

=> sf test 1ff000 1000
SPI flash test:
0 erase: 0 ticks, 4096000 KiB/s 32768.000 Mbps
1 check: 3 ticks, 1333 KiB/s 10.664 Mbps
2 write: 235 ticks, 17 KiB/s 0.136 Mbps
3 read: 2 ticks, 2000 KiB/s 16.000 Mbps
Test passed
0 erase: 0 ticks, 4096000 KiB/s 32768.000 Mbps
1 check: 3 ticks, 1333 KiB/s 10.664 Mbps
2 write: 235 ticks, 17 KiB/s 0.136 Mbps
3 read: 2 ticks, 2000 KiB/s 16.000 Mbps

Fixes c906a3a01582 ("i.MX: Add the Freescale SPI Controller")

---

(no changes since v3)

Changes in v3:
- Simplify the tx fifo endianness handling

 hw/ssi/imx_spi.c | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)
hw/ssi/imx_spi.c +2 -5
index 47c8a0f572..b5124a6426 100644
@@ -171,7 +171,6 @@ static void imx_spi_flush_txfifo(IMXSPIState *s)
 
     while (!fifo32_is_empty(&s->tx_fifo)) {
         int tx_burst = 0;
-        int index = 0;
 
         if (s->burst_length <= 0) {
             s->burst_length = imx_spi_burst_length(s);
@@ -192,7 +191,7 @@ static void imx_spi_flush_txfifo(IMXSPIState *s)
         rx = 0;
 
         while (tx_burst > 0) {
-            uint8_t byte = tx & 0xff;
+            uint8_t byte = tx >> (tx_burst - 8);
 
             DPRINTF("writing 0x%02x\n", (uint32_t)byte);
 
@@ -201,13 +200,11 @@ static void imx_spi_flush_txfifo(IMXSPIState *s)
 
             DPRINTF("0x%02x read\n", (uint32_t)byte);
 
-            tx = tx >> 8;
-            rx |= (byte << (index * 8));
+            rx = (rx << 8) | byte;
 
             /* Remove 8 bits from the actual burst */
             tx_burst -= 8;
             s->burst_length -= 8;
-            index++;
         }
 
         DPRINTF("data rx:0x%08x\n", rx);
Signature · 1 line
2.25.1
#15 lore raw

Re: [PATCH v4 6/6] hw/ssi: imx_spi: Correct tx and rx fifo endianness

Quoted 60 lines · Bin Meng

From: Bin Meng <[email protected]>

The endianness of data exchange between tx and rx fifo is incorrect. Earlier bytes are supposed to show up on MSB and later bytes on LSB,
ie: in big endian. The manual does not explicitly say this, but the U-Boot and Linux driver codes have a swap on the data transferred to tx fifo and from rx fifo.

With this change, U-Boot read from / write to SPI flash tests pass.

=> sf test 1ff000 1000
SPI flash test:
0 erase: 0 ticks, 4096000 KiB/s 32768.000 Mbps
1 check: 3 ticks, 1333 KiB/s 10.664 Mbps
2 write: 235 ticks, 17 KiB/s 0.136 Mbps
3 read: 2 ticks, 2000 KiB/s 16.000 Mbps
Test passed
0 erase: 0 ticks, 4096000 KiB/s 32768.000 Mbps
1 check: 3 ticks, 1333 KiB/s 10.664 Mbps
2 write: 235 ticks, 17 KiB/s 0.136 Mbps
3 read: 2 ticks, 2000 KiB/s 16.000 Mbps

Fixes c906a3a01582 ("i.MX: Add the Freescale SPI Controller")

---

(no changes since v3)

Changes in v3:
- Simplify the tx fifo endianness handling

 hw/ssi/imx_spi.c | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)
hw/ssi/imx_spi.c +1 -2
index 47c8a0f572..b5124a6426 100644
@@ -171,7 +171,6 @@ static void imx_spi_flush_txfifo(IMXSPIState *s)
 
     while (!fifo32_is_empty(&s->tx_fifo)) {
         int tx_burst = 0;
-        int index = 0;
 
         if (s->burst_length <= 0) {
             s->burst_length = imx_spi_burst_length(s);
@@ -192,7 +191,7 @@ static void imx_spi_flush_txfifo(IMXSPIState *s)
         rx = 0;
 
         while (tx_burst > 0) {
-            uint8_t byte = tx & 0xff;
+            uint8_t byte = tx >> (tx_burst - 8);
 
             DPRINTF("writing 0x%02x\n", (uint32_t)byte);
 
@@ -201,13 +200,11 @@ static void imx_spi_flush_txfifo(IMXSPIState *s)
 
             DPRINTF("0x%02x read\n", (uint32_t)byte);

hw/ssi/imx_spi.c static void imx_spi_flush_txfifo(IMXSPIState *s) · line 208

Quoted 10 lines
hw/ssi/imx_spi.c +1 -3
 
-            tx = tx >> 8;
-            rx |= (byte << (index * 8));
+            rx = (rx << 8) | byte;
 
             /* Remove 8 bits from the actual burst */
             tx_burst -= 8;
             s->burst_length -= 8;
-            index++;
         }

This version of the loop definitely looks a lot neater. However, looking at the code I don't think there's anything that forces the guest to set a burst length that's a multiple of 8, so you need to handle that somehow. Otherwise on the last time through the loop (tx_burst - 8) can be negative, which is undefined behaviour when you try to shift by it.

I think just rounding tx_burst up to a multiple of 8 before the start of the loop would do the right thing ?

thanks
-- PMM

author
#16 lore raw

Re: [PATCH v4 6/6] hw/ssi: imx_spi: Correct tx and rx fifo endianness

Quoted 67 lines · Peter Maydell
Quoted 65 lines · Bin Meng

From: Bin Meng <[email protected]>

The endianness of data exchange between tx and rx fifo is incorrect. Earlier bytes are supposed to show up on MSB and later bytes on LSB,
ie: in big endian. The manual does not explicitly say this, but the U-Boot and Linux driver codes have a swap on the data transferred to tx fifo and from rx fifo.

With this change, U-Boot read from / write to SPI flash tests pass.

=> sf test 1ff000 1000
SPI flash test:
0 erase: 0 ticks, 4096000 KiB/s 32768.000 Mbps
1 check: 3 ticks, 1333 KiB/s 10.664 Mbps
2 write: 235 ticks, 17 KiB/s 0.136 Mbps
3 read: 2 ticks, 2000 KiB/s 16.000 Mbps
Test passed
0 erase: 0 ticks, 4096000 KiB/s 32768.000 Mbps
1 check: 3 ticks, 1333 KiB/s 10.664 Mbps
2 write: 235 ticks, 17 KiB/s 0.136 Mbps
3 read: 2 ticks, 2000 KiB/s 16.000 Mbps

Fixes c906a3a01582 ("i.MX: Add the Freescale SPI Controller")

---

(no changes since v3)

Changes in v3:
- Simplify the tx fifo endianness handling

 hw/ssi/imx_spi.c | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)
hw/ssi/imx_spi.c +2 -4
index 47c8a0f572..b5124a6426 100644
@@ -171,7 +171,6 @@ static void imx_spi_flush_txfifo(IMXSPIState *s)
 
     while (!fifo32_is_empty(&s->tx_fifo)) {
         int tx_burst = 0;
-        int index = 0;
 
         if (s->burst_length <= 0) {
             s->burst_length = imx_spi_burst_length(s);
@@ -192,7 +191,7 @@ static void imx_spi_flush_txfifo(IMXSPIState *s)
         rx = 0;
 
         while (tx_burst > 0) {
-            uint8_t byte = tx & 0xff;
+            uint8_t byte = tx >> (tx_burst - 8);
 
             DPRINTF("writing 0x%02x\n", (uint32_t)byte);
 
@@ -201,13 +200,11 @@ static void imx_spi_flush_txfifo(IMXSPIState *s)
 
             DPRINTF("0x%02x read\n", (uint32_t)byte);
 
-            tx = tx >> 8;
-            rx |= (byte << (index * 8));
+            rx = (rx << 8) | byte;
 
Quoted 12 lines

hw/ssi/imx_spi.c static void imx_spi_flush_txfifo(IMXSPIState *s) · line 208

Quoted 5 lines
hw/ssi/imx_spi.c +0 -1
             /* Remove 8 bits from the actual burst */
             tx_burst -= 8;
             s->burst_length -= 8;
-            index++;
         }

This version of the loop definitely looks a lot neater. However, looking at the code I don't think there's anything that forces the guest to set a burst length that's a multiple of 8, so you need to handle that somehow. Otherwise on the last time through the loop (tx_burst - 8) can be negative, which is undefined behaviour when you try to shift by it.

Yes, that's why I added a patch to log the unimplemented behavior to notify the user.

Quoted 2 lines

I think just rounding tx_burst up to a multiple of 8 before the start of the loop would do the right thing ?

Probably. Given all flash transfers are normally multiple of 8-bits I am not sure what the real hardware behavior is when it is not multiple of 8, but I will try to add something in the next version.

Regards,
Bin

#17 lore raw

Re: [PATCH v4 6/6] hw/ssi: imx_spi: Correct tx and rx fifo endianness

Quoted 11 lines · Bin Meng
Quoted 10 lines · Peter Maydell
Quoted 8 lines · Bin Meng

From: Bin Meng <[email protected]>

The endianness of data exchange between tx and rx fifo is incorrect. Earlier bytes are supposed to show up on MSB and later bytes on LSB,
ie: in big endian. The manual does not explicitly say this, but the U-Boot and Linux driver codes have a swap on the data transferred to tx fifo and from rx fifo.

[...]

Quoted 11 lines
Quoted 7 lines

This version of the loop definitely looks a lot neater. However, looking at the code I don't think there's anything that forces the guest to set a burst length that's a multiple of 8, so you need to handle that somehow. Otherwise on the last time through the loop (tx_burst - 8) can be negative, which is undefined behaviour when you try to shift by it.

Yes, that's why I added a patch to log the unimplemented behavior to notify the user.

Quoted 6 lines
Quoted 2 lines

I think just rounding tx_burst up to a multiple of 8 before the start of the loop would do the right thing ?

Probably. Given all flash transfers are normally multiple of 8-bits I am not sure what the real hardware behavior is when it is not multiple of 8, but I will try to add something in the next version.

FWIW not multiple of 8 use is not that uncommon, see:
https://guruce.com/blogpost/freescale-imx53-and-imx6-ecspi-silicon-bug