ice: add functions to get and set Tx queue context

The live migration driver will need to save and restore the Tx queue
context state from the hardware registers. This state contains both static
fields which do not change during Tx traffic as well as dynamic fields
which may change during Tx traffic.

Unlike the Rx context, the Tx queue context is accessed indirectly from
GLCOMM_QTX_CNTX_CTL and GLCOMM_QTX_CNTX_DATA registers. These registers are
shared by multiple PFs on the same PCIe card. Multiple PFs cannot safely
access the registers simultaneously, and there is no hardware semaphore or
logic to control access. To handle this, introduce the txq_ctx_lock to the
ice_adapter structure. This is similar to the ptp_gltsyn_time_lock. All PFs
on the same adapter share this structure, and use it to serialize access to
the registers to prevent error.

Add a new functions to get and set the Tx queue context through the
GLCOMM_QTX_CNTX_CTL interface. The hardware context values are stored in
the registers using the same packed format as the Admin Queue buffer.

The hardware buffer is 40 bytes wide, as it contains an additional 18 bytes
of internal state not sent with the Admin Queue buffer. For this reason, a
separate typedef and packing function must be used. We can share the same
packed fields definitions because we never need to unpack the internal
state. This is preferred, as it ensures the internal state is zero'd when
writing into HW, and avoids issues with reading by u32 registers into a
buffer of 22 bytes in length. Thanks to the typedefs, misuse of the API
with the wrong size buffer can easily be caught at compile time.

Note reading this data from hardware is essential because the current Tx
queue context may be different from the context as initially programmed by
the driver during VF initialization. When migrating a VF we must ensure the
target VF has identical context as the source VF did.

Co-developed-by: Yahui Cao <yahui.cao@intel.com>
Signed-off-by: Yahui Cao <yahui.cao@intel.com>
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Madhu Chittim <madhu.chittim@intel.com>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

1 files changed, 170 insertions, 3 deletions

diff --git a/drivers/net/ethernet/intel/ice/ice_common.c b/drivers/net/ethernet/intel/ice/ice_common.c
index ddde3487aea9..1b435e108d3c 100644
--- a/drivers/net/ethernet/intel/ice/ice_common.c
+++ b/drivers/net/ethernet/intel/ice/ice_common.c
@@ -1503,12 +1503,12 @@ static const struct packed_field_u8 ice_tlan_ctx_fields[] = {
 };
 
 /**
- * ice_pack_txq_ctx - Pack Tx queue context into a HW buffer
+ * ice_pack_txq_ctx - Pack Tx queue context into Admin Queue buffer
  * @ctx: the Tx queue context to pack
- * @buf: the HW buffer to pack into
+ * @buf: the Admin Queue HW buffer to pack into
  *
  * Pack the Tx queue context from the CPU-friendly unpacked buffer into its
- * bit-packed HW layout.
+ * bit-packed Admin Queue layout.
  */
 void ice_pack_txq_ctx(const struct ice_tlan_ctx *ctx, ice_txq_ctx_buf_t *buf)
 {
@@ -1516,6 +1516,173 @@ void ice_pack_txq_ctx(const struct ice_tlan_ctx *ctx, ice_txq_ctx_buf_t *buf)
 		    QUIRK_LITTLE_ENDIAN | QUIRK_LSW32_IS_FIRST);
 }
 
+/**
+ * ice_pack_txq_ctx_full - Pack Tx queue context into a HW buffer
+ * @ctx: the Tx queue context to pack
+ * @buf: the HW buffer to pack into
+ *
+ * Pack the Tx queue context from the CPU-friendly unpacked buffer into its
+ * bit-packed HW layout, including the internal data portion.
+ */
+static void ice_pack_txq_ctx_full(const struct ice_tlan_ctx *ctx,
+				  ice_txq_ctx_buf_full_t *buf)
+{
+	pack_fields(buf, sizeof(*buf), ctx, ice_tlan_ctx_fields,
+		    QUIRK_LITTLE_ENDIAN | QUIRK_LSW32_IS_FIRST);
+}
+
+/**
+ * ice_unpack_txq_ctx_full - Unpack Tx queue context from a HW buffer
+ * @buf: the HW buffer to unpack from
+ * @ctx: the Tx queue context to unpack
+ *
+ * Unpack the Tx queue context from the HW buffer (including the full internal
+ * state) into the CPU-friendly structure.
+ */
+static void ice_unpack_txq_ctx_full(const ice_txq_ctx_buf_full_t *buf,
+				    struct ice_tlan_ctx *ctx)
+{
+	unpack_fields(buf, sizeof(*buf), ctx, ice_tlan_ctx_fields,
+		      QUIRK_LITTLE_ENDIAN | QUIRK_LSW32_IS_FIRST);
+}
+
+/**
+ * ice_copy_txq_ctx_from_hw - Copy Tx Queue context from HW registers
+ * @hw: pointer to the hardware structure
+ * @txq_ctx: pointer to the packed Tx queue context, including internal state
+ * @txq_index: the index of the Tx queue
+ *
+ * Copy Tx Queue context from HW register space to dense structure
+ */
+static void ice_copy_txq_ctx_from_hw(struct ice_hw *hw,
+				     ice_txq_ctx_buf_full_t *txq_ctx,
+				     u32 txq_index)
+{
+	struct ice_pf *pf = container_of(hw, struct ice_pf, hw);
+	u32 *ctx = (u32 *)txq_ctx;
+	u32 txq_base, reg;
+
+	/* Get Tx queue base within card space */
+	txq_base = rd32(hw, PFLAN_TX_QALLOC(hw->pf_id));
+	txq_base = FIELD_GET(PFLAN_TX_QALLOC_FIRSTQ_M, txq_base);
+
+	reg = FIELD_PREP(GLCOMM_QTX_CNTX_CTL_CMD_M,
+			 GLCOMM_QTX_CNTX_CTL_CMD_READ) |
+	      FIELD_PREP(GLCOMM_QTX_CNTX_CTL_QUEUE_ID_M,
+			 txq_base + txq_index) |
+	      GLCOMM_QTX_CNTX_CTL_CMD_EXEC_M;
+
+	/* Prevent other PFs on the same adapter from accessing the Tx queue
+	 * context interface concurrently.
+	 */
+	spin_lock(&pf->adapter->txq_ctx_lock);
+
+	wr32(hw, GLCOMM_QTX_CNTX_CTL, reg);
+	ice_flush(hw);
+
+	/* Copy each dword separately from HW */
+	for (int i = 0; i < ICE_TXQ_CTX_FULL_SIZE_DWORDS; i++, ctx++) {
+		*ctx = rd32(hw, GLCOMM_QTX_CNTX_DATA(i));
+
+		ice_debug(hw, ICE_DBG_QCTX, "qtxdata[%d]: %08X\n", i, *ctx);
+	}
+
+	spin_unlock(&pf->adapter->txq_ctx_lock);
+}
+
+/**
+ * ice_copy_txq_ctx_to_hw - Copy Tx Queue context into HW registers
+ * @hw: pointer to the hardware structure
+ * @txq_ctx: pointer to the packed Tx queue context, including internal state
+ * @txq_index: the index of the Tx queue
+ */
+static void ice_copy_txq_ctx_to_hw(struct ice_hw *hw,
+				   const ice_txq_ctx_buf_full_t *txq_ctx,
+				   u32 txq_index)
+{
+	struct ice_pf *pf = container_of(hw, struct ice_pf, hw);
+	u32 txq_base, reg;
+
+	/* Get Tx queue base within card space */
+	txq_base = rd32(hw, PFLAN_TX_QALLOC(hw->pf_id));
+	txq_base = FIELD_GET(PFLAN_TX_QALLOC_FIRSTQ_M, txq_base);
+
+	reg = FIELD_PREP(GLCOMM_QTX_CNTX_CTL_CMD_M,
+			 GLCOMM_QTX_CNTX_CTL_CMD_WRITE_NO_DYN) |
+	      FIELD_PREP(GLCOMM_QTX_CNTX_CTL_QUEUE_ID_M,
+			 txq_base + txq_index) |
+	      GLCOMM_QTX_CNTX_CTL_CMD_EXEC_M;
+
+	/* Prevent other PFs on the same adapter from accessing the Tx queue
+	 * context interface concurrently.
+	 */
+	spin_lock(&pf->adapter->txq_ctx_lock);
+
+	/* Copy each dword separately to HW */
+	for (int i = 0; i < ICE_TXQ_CTX_FULL_SIZE_DWORDS; i++) {
+		u32 ctx = ((const u32 *)txq_ctx)[i];
+
+		wr32(hw, GLCOMM_QTX_CNTX_DATA(i), ctx);
+
+		ice_debug(hw, ICE_DBG_QCTX, "qtxdata[%d]: %08X\n", i, ctx);
+	}
+
+	wr32(hw, GLCOMM_QTX_CNTX_CTL, reg);
+	ice_flush(hw);
+
+	spin_unlock(&pf->adapter->txq_ctx_lock);
+}
+
+/**
+ * ice_read_txq_ctx - Read Tx queue context from HW
+ * @hw: pointer to the hardware structure
+ * @tlan_ctx: pointer to the Tx queue context
+ * @txq_index: the index of the Tx queue
+ *
+ * Read the Tx queue context from the HW registers, then unpack it into the
+ * ice_tlan_ctx structure for use.
+ *
+ * Returns: 0 on success, or -EINVAL on an invalid Tx queue index.
+ */
+int ice_read_txq_ctx(struct ice_hw *hw, struct ice_tlan_ctx *tlan_ctx,
+		     u32 txq_index)
+{
+	ice_txq_ctx_buf_full_t buf = {};
+
+	if (txq_index > QTX_COMM_HEAD_MAX_INDEX)
+		return -EINVAL;
+
+	ice_copy_txq_ctx_from_hw(hw, &buf, txq_index);
+	ice_unpack_txq_ctx_full(&buf, tlan_ctx);
+
+	return 0;
+}
+
+/**
+ * ice_write_txq_ctx - Write Tx queue context to HW
+ * @hw: pointer to the hardware structure
+ * @tlan_ctx: pointer to the Tx queue context
+ * @txq_index: the index of the Tx queue
+ *
+ * Pack the Tx queue context into the dense HW layout, then write it into the
+ * HW registers.
+ *
+ * Returns: 0 on success, or -EINVAL on an invalid Tx queue index.
+ */
+int ice_write_txq_ctx(struct ice_hw *hw, struct ice_tlan_ctx *tlan_ctx,
+		      u32 txq_index)
+{
+	ice_txq_ctx_buf_full_t buf = {};
+
+	if (txq_index > QTX_COMM_HEAD_MAX_INDEX)
+		return -EINVAL;
+
+	ice_pack_txq_ctx_full(tlan_ctx, &buf);
+	ice_copy_txq_ctx_to_hw(hw, &buf, txq_index);
+
+	return 0;
+}
+
 /* Sideband Queue command wrappers */
 
 /**