Skip to contentSkip to conversation

hw/i2c/aspeed: Fix old reg slave receive

PATCH In review Peter Delevoryas 3 participants 7 messages Aug 20, 2022 – Aug 25, 2022 Arm/Aspeed machine support

lore
Share & tools
lore
#0 lore raw

cover hw/i2c/aspeed: Fix old reg slave receive

Hey everyone,

I haven't gotten a chance to work on the Aspeed I2C controller in a little while, but I finally started looking at it again and noticed the old-register mode slave receive function (master-send-to-slave) does the wrong thing for the first byte. See the commit message for details.

I noticed this because I have a qtest for slave mode rx in old-register mode downstream [1] (I'm also working on a version of the test that can be upstreamed) that broke when I updated our QEMU branch to the 7.1 release. Previously I was using Klaus's original slave I2C patches from [2].

An example of the test running successfully with this change is pasted below, for whatever that's worth.

Thanks,
Peter

[1]: https://github.com/facebook/openbmc/blob/helium/common/recipes-devtools/qemu/qemu/0008-hw-misc-Add-byte-by-byte-i2c-network-device.patch
[2]: https://lore.kernel.org/qemu-devel/[email protected]/

# random seed: R02S5d2728d1347dc8b50533a0d85ebb1b02
# starting QEMU: exec build/qemu-system-arm -qtest unix:/tmp/qtest-711521.sock -qtest-log /dev/null -chardev socket,path=/tmp/qtest-711521.qmp,id=char0 -mon chardev=char0,mode=control -display none -machine fby35-bmc -netdev socket,id=socket0,udp=127.0.0.1:5000,localaddr=127.0.0.1:6000 -device i2c-netdev2,bus=aspeed.i2c.bus.0,address=0x32,netdev=socket0 -accel qtest
i2c_netdev2_class_init
i2c_netdev2_realize
i2c_netdev2_can_receive
1..2
# Start of arm tests
# Start of ast2600 tests
# Start of i2c tests
i2c_netdev2_handle_event: 1
../hw/misc/i2c-netdev2.c: tx [64, 00, 00]
../hw/misc/i2c-netdev2.c: tx [de]
../hw/misc/i2c-netdev2.c: tx [ad]
../hw/misc/i2c-netdev2.c: tx [be]
../hw/misc/i2c-netdev2.c: tx [ef]
i2c_netdev2_handle_event: 3
../hw/misc/i2c-netdev2.c: tx [00, 00, 00, 00]
i2c_netdev2_can_receive
../hw/misc/i2c-netdev2.c: rx [01, 00]
i2c_netdev2_can_receive
../hw/misc/i2c-netdev2.c: rx [01, 00]
i2c_netdev2_can_receive
../hw/misc/i2c-netdev2.c: rx [01, 00]
i2c_netdev2_can_receive
../hw/misc/i2c-netdev2.c: rx [01, 00]
i2c_netdev2_can_receive
../hw/misc/i2c-netdev2.c: rx [01, 00]
ok 1 /arm/ast2600/i2c/write_in_old_byte_mode
i2c_netdev2_can_receive
../hw/misc/i2c-netdev2.c: rx [20, 00, 00]
prev rx_buf: [00, 00, 00, 00, 00, 00, 00, 00, 00, 00]
next rx_buf: [20, 00, 00, 00, 00, 00, 00, 00, 00, 00]
i2c_netdev2_slave_mode_rx: rx_len=3
i2c_netdev2_slave_mode_rx: rx_len=3
i2c_netdev2_slave_mode_rx: guest OS ack rx, clearing rx_len
i2c_netdev2_can_receive
../hw/misc/i2c-netdev2.c: rx [de]
prev rx_buf: [20, 00, 00, 00, 00, 00, 00, 00, 00, 00]
next rx_buf: [de, 00, 00, 00, 00, 00, 00, 00, 00, 00]
i2c_netdev2_slave_mode_rx: rx_len=1
i2c_netdev2_slave_mode_rx: rx_len=1
i2c_netdev2_slave_mode_rx: guest OS ack rx, clearing rx_len
i2c_netdev2_can_receive
../hw/misc/i2c-netdev2.c: rx [ad]
prev rx_buf: [de, 00, 00, 00, 00, 00, 00, 00, 00, 00]
next rx_buf: [ad, 00, 00, 00, 00, 00, 00, 00, 00, 00]
i2c_netdev2_slave_mode_rx: rx_len=1
i2c_netdev2_slave_mode_rx: rx_len=1
i2c_netdev2_slave_mode_rx: guest OS ack rx, clearing rx_len
i2c_netdev2_can_receive
../hw/misc/i2c-netdev2.c: rx [be]
prev rx_buf: [ad, 00, 00, 00, 00, 00, 00, 00, 00, 00]
next rx_buf: [be, 00, 00, 00, 00, 00, 00, 00, 00, 00]
i2c_netdev2_slave_mode_rx: rx_len=1
i2c_netdev2_slave_mode_rx: rx_len=1
i2c_netdev2_slave_mode_rx: guest OS ack rx, clearing rx_len
i2c_netdev2_can_receive
../hw/misc/i2c-netdev2.c: rx [ef]
prev rx_buf: [be, 00, 00, 00, 00, 00, 00, 00, 00, 00]
next rx_buf: [ef, 00, 00, 00, 00, 00, 00, 00, 00, 00]
i2c_netdev2_slave_mode_rx: rx_len=1
i2c_netdev2_slave_mode_rx: rx_len=1
i2c_netdev2_slave_mode_rx: guest OS ack rx, clearing rx_len
i2c_netdev2_can_receive
../hw/misc/i2c-netdev2.c: rx [ef, 00, 00, 00]
prev rx_buf: [ef, 00, 00, 00, 00, 00, 00, 00, 00, 00]
next rx_buf: [ef, 00, 00, 00, 00, 00, 00, 00, 00, 00]
i2c_netdev2_slave_mode_rx: rx_len=4
ok 2 /arm/ast2600/i2c/slave_mode_rx_byte_buf
# End of i2c tests
# End of ast2600 tests
# End of arm tests
i2c_netdev2_nic_cleanup

Peter Delevoryas (1):
hw/i2c/aspeed: Fix old reg slave receive

 hw/i2c/aspeed_i2c.c         | 8 +++++---
 include/hw/i2c/aspeed_i2c.h | 1 +
 2 files changed, 6 insertions(+), 3 deletions(-)
Signature · 1 line
2.37.1
#1 lore raw

PATCH 1/1 hw/i2c/aspeed: Fix old reg slave receive

I think when Klaus ported his slave mode changes from the original patch series to the rewritten I2C module, he changed the behavior of the first byte that is received by the slave device.

What's supposed to happen is that the AspeedI2CBus's slave device's i2c_event callback should run, and if the event is "send_async", then it should populate the byte buffer with the 8-bit I2C address that is being sent to. Since we only support "send_async", the lowest bit should always be 0 (indicating that the master is requesting to send data).

This is the code Klaus had previously, for reference. [1]

    switch (event) {
    case I2C_START_SEND:
        bus->buf = bus->dev_addr << 1;

        bus->buf &= I2CD_BYTE_BUF_RX_MASK;
        bus->buf <<= I2CD_BYTE_BUF_RX_SHIFT;

        bus->intr_status |= (I2CD_INTR_SLAVE_ADDR_RX_MATCH | I2CD_INTR_RX_DONE);
        aspeed_i2c_set_state(bus, I2CD_STXD);

        break;

[1]: https://lore.kernel.org/qemu-devel/[email protected]/

Fixes a8d48f59cd021b25 ("hw/i2c/aspeed: add slave device in old register mode")
---
 hw/i2c/aspeed_i2c.c         | 8 +++++---
 include/hw/i2c/aspeed_i2c.h | 1 +
 2 files changed, 6 insertions(+), 3 deletions(-)
hw/i2c/aspeed_i2c.c +5 -3
index 42c6d69b82..c166fd20fa 100644
@@ -1131,7 +1131,9 @@ static int aspeed_i2c_bus_slave_event(I2CSlave *slave, enum i2c_event event)
     AspeedI2CBus *bus = ASPEED_I2C_BUS(qbus->parent);
     uint32_t reg_intr_sts = aspeed_i2c_bus_intr_sts_offset(bus);
     uint32_t reg_byte_buf = aspeed_i2c_bus_byte_buf_offset(bus);
-    uint32_t value;
+    uint32_t reg_dev_addr = aspeed_i2c_bus_dev_addr_offset(bus);
+    uint32_t dev_addr = SHARED_ARRAY_FIELD_EX32(bus->regs, reg_dev_addr,
+                                                SLAVE_DEV_ADDR1);
 
     if (aspeed_i2c_is_new_mode(bus->controller)) {
         return aspeed_i2c_bus_new_slave_event(bus, event);
@@ -1139,8 +1141,8 @@ static int aspeed_i2c_bus_slave_event(I2CSlave *slave, enum i2c_event event)
 
     switch (event) {
     case I2C_START_SEND_ASYNC:
-        value = SHARED_ARRAY_FIELD_EX32(bus->regs, reg_byte_buf, TX_BUF);
-        SHARED_ARRAY_FIELD_DP32(bus->regs, reg_byte_buf, RX_BUF, value << 1);
+        /* Bit[0] == 0 indicates "send". */
+        SHARED_ARRAY_FIELD_DP32(bus->regs, reg_byte_buf, RX_BUF, dev_addr << 1);
 
         ARRAY_FIELD_DP32(bus->regs, I2CD_INTR_STS, SLAVE_ADDR_RX_MATCH, 1);
         SHARED_ARRAY_FIELD_DP32(bus->regs, reg_intr_sts, RX_DONE, 1);
include/hw/i2c/aspeed_i2c.h +1 -0
index 300a89b343..adc904d6c1 100644
@@ -130,6 +130,7 @@ REG32(I2CD_CMD, 0x14) /* I2CD Command/Status */
     SHARED_FIELD(M_TX_CMD, 1, 1)
     SHARED_FIELD(M_START_CMD, 0, 1)
 REG32(I2CD_DEV_ADDR, 0x18) /* Slave Device Address */
+    SHARED_FIELD(SLAVE_DEV_ADDR1, 0, 7)
 REG32(I2CD_POOL_CTRL, 0x1C) /* Pool Buffer Control */
     SHARED_FIELD(RX_COUNT, 24, 5)
     SHARED_FIELD(RX_SIZE, 16, 5)
Signature · 1 line
2.37.1
Reviewed-by on v1 1/1
#2 lore raw
Quoted 74 lines · Peter Delevoryas

I think when Klaus ported his slave mode changes from the original patch series to the rewritten I2C module, he changed the behavior of the first byte that is received by the slave device.

What's supposed to happen is that the AspeedI2CBus's slave device's i2c_event callback should run, and if the event is "send_async", then it should populate the byte buffer with the 8-bit I2C address that is being sent to. Since we only support "send_async", the lowest bit should always be 0 (indicating that the master is requesting to send data).

This is the code Klaus had previously, for reference. [1]

    switch (event) {
    case I2C_START_SEND:
        bus->buf = bus->dev_addr << 1;

        bus->buf &= I2CD_BYTE_BUF_RX_MASK;
        bus->buf <<= I2CD_BYTE_BUF_RX_SHIFT;

        bus->intr_status |= (I2CD_INTR_SLAVE_ADDR_RX_MATCH | I2CD_INTR_RX_DONE);
        aspeed_i2c_set_state(bus, I2CD_STXD);

        break;

[1]: https://lore.kernel.org/qemu-devel/[email protected]/

Fixes a8d48f59cd021b25 ("hw/i2c/aspeed: add slave device in old register mode")
---
 hw/i2c/aspeed_i2c.c         | 8 +++++---
 include/hw/i2c/aspeed_i2c.h | 1 +
 2 files changed, 6 insertions(+), 3 deletions(-)
hw/i2c/aspeed_i2c.c +5 -3
index 42c6d69b82..c166fd20fa 100644
@@ -1131,7 +1131,9 @@ static int aspeed_i2c_bus_slave_event(I2CSlave *slave, enum i2c_event event)
     AspeedI2CBus *bus = ASPEED_I2C_BUS(qbus->parent);
     uint32_t reg_intr_sts = aspeed_i2c_bus_intr_sts_offset(bus);
     uint32_t reg_byte_buf = aspeed_i2c_bus_byte_buf_offset(bus);
-    uint32_t value;
+    uint32_t reg_dev_addr = aspeed_i2c_bus_dev_addr_offset(bus);
+    uint32_t dev_addr = SHARED_ARRAY_FIELD_EX32(bus->regs, reg_dev_addr,
+                                                SLAVE_DEV_ADDR1);
 
     if (aspeed_i2c_is_new_mode(bus->controller)) {
         return aspeed_i2c_bus_new_slave_event(bus, event);
@@ -1139,8 +1141,8 @@ static int aspeed_i2c_bus_slave_event(I2CSlave *slave, enum i2c_event event)
 
     switch (event) {
     case I2C_START_SEND_ASYNC:
-        value = SHARED_ARRAY_FIELD_EX32(bus->regs, reg_byte_buf, TX_BUF);
-        SHARED_ARRAY_FIELD_DP32(bus->regs, reg_byte_buf, RX_BUF, value << 1);
+        /* Bit[0] == 0 indicates "send". */
+        SHARED_ARRAY_FIELD_DP32(bus->regs, reg_byte_buf, RX_BUF, dev_addr << 1);
 
         ARRAY_FIELD_DP32(bus->regs, I2CD_INTR_STS, SLAVE_ADDR_RX_MATCH, 1);
         SHARED_ARRAY_FIELD_DP32(bus->regs, reg_intr_sts, RX_DONE, 1);
include/hw/i2c/aspeed_i2c.h +1 -0
index 300a89b343..adc904d6c1 100644
@@ -130,6 +130,7 @@ REG32(I2CD_CMD, 0x14) /* I2CD Command/Status */
     SHARED_FIELD(M_TX_CMD, 1, 1)
     SHARED_FIELD(M_START_CMD, 0, 1)
 REG32(I2CD_DEV_ADDR, 0x18) /* Slave Device Address */
+    SHARED_FIELD(SLAVE_DEV_ADDR1, 0, 7)
 REG32(I2CD_POOL_CTRL, 0x1C) /* Pool Buffer Control */
     SHARED_FIELD(RX_COUNT, 24, 5)
     SHARED_FIELD(RX_SIZE, 16, 5)
Signature · 1 line
2.37.1

Nice catch Peter! I'm not sure how I messed that up like that.

#3 lore raw
Quoted 78 lines · Klaus Jensen
Quoted 74 lines · Peter Delevoryas

I think when Klaus ported his slave mode changes from the original patch series to the rewritten I2C module, he changed the behavior of the first byte that is received by the slave device.

What's supposed to happen is that the AspeedI2CBus's slave device's i2c_event callback should run, and if the event is "send_async", then it should populate the byte buffer with the 8-bit I2C address that is being sent to. Since we only support "send_async", the lowest bit should always be 0 (indicating that the master is requesting to send data).

This is the code Klaus had previously, for reference. [1]

    switch (event) {
    case I2C_START_SEND:
        bus->buf = bus->dev_addr << 1;

        bus->buf &= I2CD_BYTE_BUF_RX_MASK;
        bus->buf <<= I2CD_BYTE_BUF_RX_SHIFT;

        bus->intr_status |= (I2CD_INTR_SLAVE_ADDR_RX_MATCH | I2CD_INTR_RX_DONE);
        aspeed_i2c_set_state(bus, I2CD_STXD);

        break;

[1]: https://lore.kernel.org/qemu-devel/[email protected]/

Fixes a8d48f59cd021b25 ("hw/i2c/aspeed: add slave device in old register mode")
---
 hw/i2c/aspeed_i2c.c         | 8 +++++---
 include/hw/i2c/aspeed_i2c.h | 1 +
 2 files changed, 6 insertions(+), 3 deletions(-)
hw/i2c/aspeed_i2c.c +5 -3
index 42c6d69b82..c166fd20fa 100644
@@ -1131,7 +1131,9 @@ static int aspeed_i2c_bus_slave_event(I2CSlave *slave, enum i2c_event event)
     AspeedI2CBus *bus = ASPEED_I2C_BUS(qbus->parent);
     uint32_t reg_intr_sts = aspeed_i2c_bus_intr_sts_offset(bus);
     uint32_t reg_byte_buf = aspeed_i2c_bus_byte_buf_offset(bus);
-    uint32_t value;
+    uint32_t reg_dev_addr = aspeed_i2c_bus_dev_addr_offset(bus);
+    uint32_t dev_addr = SHARED_ARRAY_FIELD_EX32(bus->regs, reg_dev_addr,
+                                                SLAVE_DEV_ADDR1);
 
     if (aspeed_i2c_is_new_mode(bus->controller)) {
         return aspeed_i2c_bus_new_slave_event(bus, event);
@@ -1139,8 +1141,8 @@ static int aspeed_i2c_bus_slave_event(I2CSlave *slave, enum i2c_event event)
 
     switch (event) {
     case I2C_START_SEND_ASYNC:
-        value = SHARED_ARRAY_FIELD_EX32(bus->regs, reg_byte_buf, TX_BUF);
-        SHARED_ARRAY_FIELD_DP32(bus->regs, reg_byte_buf, RX_BUF, value << 1);
+        /* Bit[0] == 0 indicates "send". */
+        SHARED_ARRAY_FIELD_DP32(bus->regs, reg_byte_buf, RX_BUF, dev_addr << 1);
 
         ARRAY_FIELD_DP32(bus->regs, I2CD_INTR_STS, SLAVE_ADDR_RX_MATCH, 1);
         SHARED_ARRAY_FIELD_DP32(bus->regs, reg_intr_sts, RX_DONE, 1);
include/hw/i2c/aspeed_i2c.h +1 -0
index 300a89b343..adc904d6c1 100644
@@ -130,6 +130,7 @@ REG32(I2CD_CMD, 0x14) /* I2CD Command/Status */
     SHARED_FIELD(M_TX_CMD, 1, 1)
     SHARED_FIELD(M_START_CMD, 0, 1)
 REG32(I2CD_DEV_ADDR, 0x18) /* Slave Device Address */
+    SHARED_FIELD(SLAVE_DEV_ADDR1, 0, 7)
 REG32(I2CD_POOL_CTRL, 0x1C) /* Pool Buffer Control */
     SHARED_FIELD(RX_COUNT, 24, 5)
     SHARED_FIELD(RX_SIZE, 16, 5)
Signature · 1 line
2.37.1

Nice catch Peter! I'm not sure how I messed that up like that.

Quoted 1 line

Thanks Klaus. Just realized I forgot to cc you on this, sorry about that.

#4 lore raw
Quoted 79 lines · Peter Delevoryas
Quoted 78 lines · Klaus Jensen
Quoted 74 lines · Peter Delevoryas

I think when Klaus ported his slave mode changes from the original patch series to the rewritten I2C module, he changed the behavior of the first byte that is received by the slave device.

What's supposed to happen is that the AspeedI2CBus's slave device's i2c_event callback should run, and if the event is "send_async", then it should populate the byte buffer with the 8-bit I2C address that is being sent to. Since we only support "send_async", the lowest bit should always be 0 (indicating that the master is requesting to send data).

This is the code Klaus had previously, for reference. [1]

     switch (event) {
     case I2C_START_SEND:
         bus->buf = bus->dev_addr << 1;

         bus->buf &= I2CD_BYTE_BUF_RX_MASK;
         bus->buf <<= I2CD_BYTE_BUF_RX_SHIFT;

         bus->intr_status |= (I2CD_INTR_SLAVE_ADDR_RX_MATCH | I2CD_INTR_RX_DONE);
         aspeed_i2c_set_state(bus, I2CD_STXD);

         break;

[1]: https://lore.kernel.org/qemu-devel/[email protected]/

Fixes a8d48f59cd021b25 ("hw/i2c/aspeed: add slave device in old register mode")
---
  hw/i2c/aspeed_i2c.c         | 8 +++++---
  include/hw/i2c/aspeed_i2c.h | 1 +
  2 files changed, 6 insertions(+), 3 deletions(-)
hw/i2c/aspeed_i2c.c +5 -3
index 42c6d69b82..c166fd20fa 100644
@@ -1131,7 +1131,9 @@ static int aspeed_i2c_bus_slave_event(I2CSlave *slave, enum i2c_event event)
      AspeedI2CBus *bus = ASPEED_I2C_BUS(qbus->parent);
      uint32_t reg_intr_sts = aspeed_i2c_bus_intr_sts_offset(bus);
      uint32_t reg_byte_buf = aspeed_i2c_bus_byte_buf_offset(bus);
-    uint32_t value;
+    uint32_t reg_dev_addr = aspeed_i2c_bus_dev_addr_offset(bus);
+    uint32_t dev_addr = SHARED_ARRAY_FIELD_EX32(bus->regs, reg_dev_addr,
+                                                SLAVE_DEV_ADDR1);
 
      if (aspeed_i2c_is_new_mode(bus->controller)) {
          return aspeed_i2c_bus_new_slave_event(bus, event);
@@ -1139,8 +1141,8 @@ static int aspeed_i2c_bus_slave_event(I2CSlave *slave, enum i2c_event event)
 
      switch (event) {
      case I2C_START_SEND_ASYNC:
-        value = SHARED_ARRAY_FIELD_EX32(bus->regs, reg_byte_buf, TX_BUF);
-        SHARED_ARRAY_FIELD_DP32(bus->regs, reg_byte_buf, RX_BUF, value << 1);
+        /* Bit[0] == 0 indicates "send". */
+        SHARED_ARRAY_FIELD_DP32(bus->regs, reg_byte_buf, RX_BUF, dev_addr << 1);
 
          ARRAY_FIELD_DP32(bus->regs, I2CD_INTR_STS, SLAVE_ADDR_RX_MATCH, 1);
          SHARED_ARRAY_FIELD_DP32(bus->regs, reg_intr_sts, RX_DONE, 1);
include/hw/i2c/aspeed_i2c.h +1 -0
index 300a89b343..adc904d6c1 100644
@@ -130,6 +130,7 @@ REG32(I2CD_CMD, 0x14) /* I2CD Command/Status */
      SHARED_FIELD(M_TX_CMD, 1, 1)
      SHARED_FIELD(M_START_CMD, 0, 1)
  REG32(I2CD_DEV_ADDR, 0x18) /* Slave Device Address */
+    SHARED_FIELD(SLAVE_DEV_ADDR1, 0, 7)
  REG32(I2CD_POOL_CTRL, 0x1C) /* Pool Buffer Control */
      SHARED_FIELD(RX_COUNT, 24, 5)
      SHARED_FIELD(RX_SIZE, 16, 5)
Signature · 1 line
2.37.1

Nice catch Peter! I'm not sure how I messed that up like that.

Quoted 4 lines
Quoted 1 line

Thanks Klaus. Just realized I forgot to cc you on this, sorry about that.

Do we still have time for 7.1 ?

Thanks,

C.

#5 lore raw
Quoted 77 lines · Cédric Le Goater
Quoted 76 lines · Peter Delevoryas
Quoted 75 lines · Klaus Jensen
Quoted 71 lines · Peter Delevoryas

I think when Klaus ported his slave mode changes from the original patch series to the rewritten I2C module, he changed the behavior of the first byte that is received by the slave device.

What's supposed to happen is that the AspeedI2CBus's slave device's i2c_event callback should run, and if the event is "send_async", then it should populate the byte buffer with the 8-bit I2C address that is being sent to. Since we only support "send_async", the lowest bit should always be 0 (indicating that the master is requesting to send data).

This is the code Klaus had previously, for reference. [1]

     switch (event) {
     case I2C_START_SEND:
         bus->buf = bus->dev_addr << 1;

         bus->buf &= I2CD_BYTE_BUF_RX_MASK;
         bus->buf <<= I2CD_BYTE_BUF_RX_SHIFT;

         bus->intr_status |= (I2CD_INTR_SLAVE_ADDR_RX_MATCH | I2CD_INTR_RX_DONE);
         aspeed_i2c_set_state(bus, I2CD_STXD);

         break;

[1]: https://lore.kernel.org/qemu-devel/[email protected]/

Fixes a8d48f59cd021b25 ("hw/i2c/aspeed: add slave device in old register mode")
---
  hw/i2c/aspeed_i2c.c         | 8 +++++---
  include/hw/i2c/aspeed_i2c.h | 1 +
  2 files changed, 6 insertions(+), 3 deletions(-)
hw/i2c/aspeed_i2c.c +5 -3
index 42c6d69b82..c166fd20fa 100644
@@ -1131,7 +1131,9 @@ static int aspeed_i2c_bus_slave_event(I2CSlave *slave, enum i2c_event event)
      AspeedI2CBus *bus = ASPEED_I2C_BUS(qbus->parent);
      uint32_t reg_intr_sts = aspeed_i2c_bus_intr_sts_offset(bus);
      uint32_t reg_byte_buf = aspeed_i2c_bus_byte_buf_offset(bus);
-    uint32_t value;
+    uint32_t reg_dev_addr = aspeed_i2c_bus_dev_addr_offset(bus);
+    uint32_t dev_addr = SHARED_ARRAY_FIELD_EX32(bus->regs, reg_dev_addr,
+                                                SLAVE_DEV_ADDR1);
      if (aspeed_i2c_is_new_mode(bus->controller)) {
          return aspeed_i2c_bus_new_slave_event(bus, event);
@@ -1139,8 +1141,8 @@ static int aspeed_i2c_bus_slave_event(I2CSlave *slave, enum i2c_event event)
      switch (event) {
      case I2C_START_SEND_ASYNC:
-        value = SHARED_ARRAY_FIELD_EX32(bus->regs, reg_byte_buf, TX_BUF);
-        SHARED_ARRAY_FIELD_DP32(bus->regs, reg_byte_buf, RX_BUF, value << 1);
+        /* Bit[0] == 0 indicates "send". */
+        SHARED_ARRAY_FIELD_DP32(bus->regs, reg_byte_buf, RX_BUF, dev_addr << 1);
          ARRAY_FIELD_DP32(bus->regs, I2CD_INTR_STS, SLAVE_ADDR_RX_MATCH, 1);
          SHARED_ARRAY_FIELD_DP32(bus->regs, reg_intr_sts, RX_DONE, 1);
include/hw/i2c/aspeed_i2c.h +1 -0
index 300a89b343..adc904d6c1 100644
@@ -130,6 +130,7 @@ REG32(I2CD_CMD, 0x14) /* I2CD Command/Status */
      SHARED_FIELD(M_TX_CMD, 1, 1)
      SHARED_FIELD(M_START_CMD, 0, 1)
  REG32(I2CD_DEV_ADDR, 0x18) /* Slave Device Address */
+    SHARED_FIELD(SLAVE_DEV_ADDR1, 0, 7)
  REG32(I2CD_POOL_CTRL, 0x1C) /* Pool Buffer Control */
      SHARED_FIELD(RX_COUNT, 24, 5)
      SHARED_FIELD(RX_SIZE, 16, 5)
Signature · 1 line
2.37.1

Nice catch Peter! I'm not sure how I messed that up like that.

Quoted 6 lines
Quoted 4 lines
Quoted 1 line

Thanks Klaus. Just realized I forgot to cc you on this, sorry about that.

Do we still have time for 7.1 ?

Is this question for me, or for Peter Maydell or someone else working on the release? I think they might still be accepting some patches, or deciding if rc4 is necessary: I've created this issue to bring awareness to this, since that seems like the right way to track this for the release.

https://gitlab.com/qemu-project/qemu/-/issues/1174

I don't have any special need for 7.1, since our team branches off of master and regularly pulls in updates.

Quoted 6 lines

Thanks,

C.

#6 lore raw
Quoted 83 lines · Peter Delevoryas
Quoted 82 lines · Cédric Le Goater
Quoted 80 lines · Peter Delevoryas
Quoted 75 lines · Klaus Jensen
Quoted 71 lines · Peter Delevoryas

I think when Klaus ported his slave mode changes from the original patch series to the rewritten I2C module, he changed the behavior of the first byte that is received by the slave device.

What's supposed to happen is that the AspeedI2CBus's slave device's i2c_event callback should run, and if the event is "send_async", then it should populate the byte buffer with the 8-bit I2C address that is being sent to. Since we only support "send_async", the lowest bit should always be 0 (indicating that the master is requesting to send data).

This is the code Klaus had previously, for reference. [1]

      switch (event) {
      case I2C_START_SEND:
          bus->buf = bus->dev_addr << 1;

          bus->buf &= I2CD_BYTE_BUF_RX_MASK;
          bus->buf <<= I2CD_BYTE_BUF_RX_SHIFT;

          bus->intr_status |= (I2CD_INTR_SLAVE_ADDR_RX_MATCH | I2CD_INTR_RX_DONE);
          aspeed_i2c_set_state(bus, I2CD_STXD);

          break;

[1]: https://lore.kernel.org/qemu-devel/[email protected]/

Fixes a8d48f59cd021b25 ("hw/i2c/aspeed: add slave device in old register mode")
---
   hw/i2c/aspeed_i2c.c         | 8 +++++---
   include/hw/i2c/aspeed_i2c.h | 1 +
   2 files changed, 6 insertions(+), 3 deletions(-)
hw/i2c/aspeed_i2c.c +5 -3
index 42c6d69b82..c166fd20fa 100644
@@ -1131,7 +1131,9 @@ static int aspeed_i2c_bus_slave_event(I2CSlave *slave, enum i2c_event event)
       AspeedI2CBus *bus = ASPEED_I2C_BUS(qbus->parent);
       uint32_t reg_intr_sts = aspeed_i2c_bus_intr_sts_offset(bus);
       uint32_t reg_byte_buf = aspeed_i2c_bus_byte_buf_offset(bus);
-    uint32_t value;
+    uint32_t reg_dev_addr = aspeed_i2c_bus_dev_addr_offset(bus);
+    uint32_t dev_addr = SHARED_ARRAY_FIELD_EX32(bus->regs, reg_dev_addr,
+                                                SLAVE_DEV_ADDR1);
       if (aspeed_i2c_is_new_mode(bus->controller)) {
           return aspeed_i2c_bus_new_slave_event(bus, event);
@@ -1139,8 +1141,8 @@ static int aspeed_i2c_bus_slave_event(I2CSlave *slave, enum i2c_event event)
       switch (event) {
       case I2C_START_SEND_ASYNC:
-        value = SHARED_ARRAY_FIELD_EX32(bus->regs, reg_byte_buf, TX_BUF);
-        SHARED_ARRAY_FIELD_DP32(bus->regs, reg_byte_buf, RX_BUF, value << 1);
+        /* Bit[0] == 0 indicates "send". */
+        SHARED_ARRAY_FIELD_DP32(bus->regs, reg_byte_buf, RX_BUF, dev_addr << 1);
           ARRAY_FIELD_DP32(bus->regs, I2CD_INTR_STS, SLAVE_ADDR_RX_MATCH, 1);
           SHARED_ARRAY_FIELD_DP32(bus->regs, reg_intr_sts, RX_DONE, 1);
include/hw/i2c/aspeed_i2c.h +1 -0
index 300a89b343..adc904d6c1 100644
@@ -130,6 +130,7 @@ REG32(I2CD_CMD, 0x14) /* I2CD Command/Status */
       SHARED_FIELD(M_TX_CMD, 1, 1)
       SHARED_FIELD(M_START_CMD, 0, 1)
   REG32(I2CD_DEV_ADDR, 0x18) /* Slave Device Address */
+    SHARED_FIELD(SLAVE_DEV_ADDR1, 0, 7)
   REG32(I2CD_POOL_CTRL, 0x1C) /* Pool Buffer Control */
       SHARED_FIELD(RX_COUNT, 24, 5)
       SHARED_FIELD(RX_SIZE, 16, 5)
Signature · 1 line
2.37.1

Nice catch Peter! I'm not sure how I messed that up like that.

Quoted 1 line

Thanks Klaus. Just realized I forgot to cc you on this, sorry about that.

Quoted 8 lines
Quoted 1 line

Do we still have time for 7.1 ?

Is this question for me, or for Peter Maydell or someone else working on the release? I think they might still be accepting some patches, or deciding if rc4 is necessary: I've created this issue to bring awareness to this, since that seems like the right way to track this for the release.

https://gitlab.com/qemu-project/qemu/-/issues/1174

Thanks,

Quoted 2 lines

I don't have any special need for 7.1, since our team branches off of master and regularly pulls in updates.

I think it is worth fixing it for 7.1 since we just introduced slave mode in this release. Hence the late PR. I would have preferred doing it sooner but I am out of office and this increases latency.

C.