]> git.dujemihanovic.xyz Git - u-boot.git/commitdiff
mips: octeon: Add cvmx-helper-ilk.c
authorAaron Williams <awilliams@marvell.com>
Thu, 7 Apr 2022 07:11:15 +0000 (09:11 +0200)
committerStefan Roese <sr@denx.de>
Wed, 4 May 2022 01:38:21 +0000 (03:38 +0200)
Import cvmx-helper-ilk.c from 2013 U-Boot. It will be used by the later
added drivers to support networking on the MIPS Octeon II / III
platforms.

Signed-off-by: Aaron Williams <awilliams@marvell.com>
Signed-off-by: Stefan Roese <sr@denx.de>
arch/mips/mach-octeon/cvmx-helper-ilk.c [new file with mode: 0644]

diff --git a/arch/mips/mach-octeon/cvmx-helper-ilk.c b/arch/mips/mach-octeon/cvmx-helper-ilk.c
new file mode 100644 (file)
index 0000000..9e882f1
--- /dev/null
@@ -0,0 +1,902 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2018-2022 Marvell International Ltd.
+ *
+ * Functions for ILK initialization, configuration,
+ * and monitoring.
+ */
+
+#include <time.h>
+#include <log.h>
+#include <linux/delay.h>
+
+#include <mach/cvmx-regs.h>
+#include <mach/cvmx-csr.h>
+#include <mach/cvmx-bootmem.h>
+#include <mach/octeon-model.h>
+#include <mach/cvmx-fuse.h>
+#include <mach/octeon-feature.h>
+#include <mach/cvmx-qlm.h>
+#include <mach/octeon_qlm.h>
+#include <mach/cvmx-pcie.h>
+#include <mach/cvmx-coremask.h>
+
+#include <mach/cvmx-agl-defs.h>
+#include <mach/cvmx-bgxx-defs.h>
+#include <mach/cvmx-ciu-defs.h>
+#include <mach/cvmx-gmxx-defs.h>
+#include <mach/cvmx-ilk-defs.h>
+#include <mach/cvmx-ipd-defs.h>
+#include <mach/cvmx-pcsx-defs.h>
+#include <mach/cvmx-pcsxx-defs.h>
+#include <mach/cvmx-pki-defs.h>
+#include <mach/cvmx-pko-defs.h>
+#include <mach/cvmx-xcv-defs.h>
+
+#include <mach/cvmx-hwpko.h>
+#include <mach/cvmx-ilk.h>
+#include <mach/cvmx-pki.h>
+
+#include <mach/cvmx-helper.h>
+#include <mach/cvmx-helper-board.h>
+#include <mach/cvmx-helper-cfg.h>
+
+int __cvmx_helper_ilk_enumerate(int xiface)
+{
+       struct cvmx_xiface xi = cvmx_helper_xiface_to_node_interface(xiface);
+
+       xi.interface -= CVMX_ILK_GBL_BASE();
+       return cvmx_ilk_chans[xi.node][xi.interface];
+}
+
+/**
+ * @INTERNAL
+ * Initialize all tx calendar entries to the xoff state.
+ * Initialize all rx calendar entries to the xon state. The rx calendar entries
+ * must be in the xon state to allow new pko pipe assignments. If a calendar
+ * entry is assigned a different pko pipe while in the xoff state, the old pko
+ * pipe will stay in the xoff state even when no longer used by ilk.
+ *
+ * @param intf Interface whose calendar are to be initialized.
+ */
+static void __cvmx_ilk_clear_cal_cn78xx(int intf)
+{
+       cvmx_ilk_txx_cal_entryx_t tx_entry;
+       cvmx_ilk_rxx_cal_entryx_t rx_entry;
+       int i;
+       int node = (intf >> 4) & 0xf;
+       int interface = (intf & 0xf);
+
+       /* Initialize all tx calendar entries to off */
+       tx_entry.u64 = 0;
+       tx_entry.s.ctl = XOFF;
+       for (i = 0; i < CVMX_ILK_MAX_CAL; i++) {
+               csr_wr_node(node, CVMX_ILK_TXX_CAL_ENTRYX(i, interface),
+                           tx_entry.u64);
+       }
+
+       /* Initialize all rx calendar entries to on */
+       rx_entry.u64 = 0;
+       rx_entry.s.ctl = XOFF;
+       for (i = 0; i < CVMX_ILK_MAX_CAL; i++) {
+               csr_wr_node(node, CVMX_ILK_RXX_CAL_ENTRYX(i, interface),
+                           rx_entry.u64);
+       }
+}
+
+/**
+ * @INTERNAL
+ * Initialize all tx calendar entries to the xoff state.
+ * Initialize all rx calendar entries to the xon state. The rx calendar entries
+ * must be in the xon state to allow new pko pipe assignments. If a calendar
+ * entry is assigned a different pko pipe while in the xoff state, the old pko
+ * pipe will stay in the xoff state even when no longer used by ilk.
+ *
+ * @param interface whose calendar are to be initialized.
+ */
+static void __cvmx_ilk_clear_cal_cn68xx(int interface)
+{
+       cvmx_ilk_txx_idx_cal_t tx_idx;
+       cvmx_ilk_txx_mem_cal0_t tx_cal0;
+       cvmx_ilk_txx_mem_cal1_t tx_cal1;
+       cvmx_ilk_rxx_idx_cal_t rx_idx;
+       cvmx_ilk_rxx_mem_cal0_t rx_cal0;
+       cvmx_ilk_rxx_mem_cal1_t rx_cal1;
+       int i;
+
+       /*
+        * First we initialize the tx calendar starting from entry 0,
+        * incrementing the entry with every write.
+        */
+       tx_idx.u64 = 0;
+       tx_idx.s.inc = 1;
+       csr_wr(CVMX_ILK_TXX_IDX_CAL(interface), tx_idx.u64);
+
+       /* Set state to xoff for all entries */
+       tx_cal0.u64 = 0;
+       tx_cal0.s.entry_ctl0 = XOFF;
+       tx_cal0.s.entry_ctl1 = XOFF;
+       tx_cal0.s.entry_ctl2 = XOFF;
+       tx_cal0.s.entry_ctl3 = XOFF;
+
+       tx_cal1.u64 = 0;
+       tx_cal1.s.entry_ctl4 = XOFF;
+       tx_cal1.s.entry_ctl5 = XOFF;
+       tx_cal1.s.entry_ctl6 = XOFF;
+       tx_cal1.s.entry_ctl7 = XOFF;
+
+       /* Write all 288 entries */
+       for (i = 0; i < CVMX_ILK_MAX_CAL_IDX; i++) {
+               csr_wr(CVMX_ILK_TXX_MEM_CAL0(interface), tx_cal0.u64);
+               csr_wr(CVMX_ILK_TXX_MEM_CAL1(interface), tx_cal1.u64);
+       }
+
+       /*
+        * Next we initialize the rx calendar starting from entry 0,
+        * incrementing the entry with every write.
+        */
+       rx_idx.u64 = 0;
+       rx_idx.s.inc = 1;
+       csr_wr(CVMX_ILK_RXX_IDX_CAL(interface), rx_idx.u64);
+
+       /* Set state to xon for all entries */
+       rx_cal0.u64 = 0;
+       rx_cal0.s.entry_ctl0 = XON;
+       rx_cal0.s.entry_ctl1 = XON;
+       rx_cal0.s.entry_ctl2 = XON;
+       rx_cal0.s.entry_ctl3 = XON;
+
+       rx_cal1.u64 = 0;
+       rx_cal1.s.entry_ctl4 = XON;
+       rx_cal1.s.entry_ctl5 = XON;
+       rx_cal1.s.entry_ctl6 = XON;
+       rx_cal1.s.entry_ctl7 = XON;
+
+       /* Write all 288 entries */
+       for (i = 0; i < CVMX_ILK_MAX_CAL_IDX; i++) {
+               csr_wr(CVMX_ILK_RXX_MEM_CAL0(interface), rx_cal0.u64);
+               csr_wr(CVMX_ILK_RXX_MEM_CAL1(interface), rx_cal1.u64);
+       }
+}
+
+/**
+ * @INTERNAL
+ * Initialize all calendar entries.
+ *
+ * @param interface whose calendar is to be initialized.
+ */
+void __cvmx_ilk_clear_cal(int interface)
+{
+       if (OCTEON_IS_MODEL(OCTEON_CN68XX))
+               __cvmx_ilk_clear_cal_cn68xx(interface);
+       else if (OCTEON_IS_MODEL(OCTEON_CN78XX))
+               __cvmx_ilk_clear_cal_cn78xx(interface);
+}
+
+void __cvmx_ilk_write_tx_cal_entry_cn68xx(int interface, int channel,
+                                         unsigned char bpid)
+{
+       cvmx_ilk_txx_idx_cal_t tx_idx;
+       cvmx_ilk_txx_mem_cal0_t tx_cal0;
+       cvmx_ilk_txx_mem_cal1_t tx_cal1;
+       int entry;
+       int window;
+       int window_entry;
+
+       /*
+        * The calendar has 288 entries. Each calendar entry represents a
+        * channel's flow control state or the link flow control state.
+        * Starting with the first entry, every sixteenth entry is used for the
+        * link flow control state. The other 15 entries are used for the
+        * channels flow control state:
+        * entry 0   ----> link flow control state
+        * entry 1   ----> channel 0 flow control state
+        * entry 2   ----> channel 1 flow control state
+        * ...
+        * entry 15  ----> channel 14 flow control state
+        * entry 16  ----> link flow control state
+        * entry 17  ----> channel 15 flow control state
+        *
+        * Also, the calendar is accessed via windows into it. Each window maps
+        * to 8 entries.
+        */
+       entry = 1 + channel + (channel / 15);
+       window = entry / 8;
+       window_entry = entry % 8;
+
+       /* Indicate the window we need to access */
+       tx_idx.u64 = 0;
+       tx_idx.s.index = window;
+       csr_wr(CVMX_ILK_TXX_IDX_CAL(interface), tx_idx.u64);
+
+       /* Get the window's current value */
+       tx_cal0.u64 = csr_rd(CVMX_ILK_TXX_MEM_CAL0(interface));
+       tx_cal1.u64 = csr_rd(CVMX_ILK_TXX_MEM_CAL1(interface));
+
+       /* Force every sixteenth entry as link flow control state */
+       if ((window & 1) == 0)
+               tx_cal0.s.entry_ctl0 = LINK;
+
+       /* Update the entry */
+       switch (window_entry) {
+       case 0:
+               tx_cal0.s.entry_ctl0 = 0;
+               tx_cal0.s.bpid0 = bpid;
+               break;
+       case 1:
+               tx_cal0.s.entry_ctl1 = 0;
+               tx_cal0.s.bpid1 = bpid;
+               break;
+       case 2:
+               tx_cal0.s.entry_ctl2 = 0;
+               tx_cal0.s.bpid2 = bpid;
+               break;
+       case 3:
+               tx_cal0.s.entry_ctl3 = 0;
+               tx_cal0.s.bpid3 = bpid;
+               break;
+       case 4:
+               tx_cal1.s.entry_ctl4 = 0;
+               tx_cal1.s.bpid4 = bpid;
+               break;
+       case 5:
+               tx_cal1.s.entry_ctl5 = 0;
+               tx_cal1.s.bpid5 = bpid;
+               break;
+       case 6:
+               tx_cal1.s.entry_ctl6 = 0;
+               tx_cal1.s.bpid6 = bpid;
+               break;
+       case 7:
+               tx_cal1.s.entry_ctl7 = 0;
+               tx_cal1.s.bpid7 = bpid;
+               break;
+       }
+
+       /* Write the window new value */
+       csr_wr(CVMX_ILK_TXX_MEM_CAL0(interface), tx_cal0.u64);
+       csr_wr(CVMX_ILK_TXX_MEM_CAL1(interface), tx_cal1.u64);
+}
+
+void __cvmx_ilk_write_tx_cal_entry_cn78xx(int intf, int channel,
+                                         unsigned char bpid)
+{
+       cvmx_ilk_txx_cal_entryx_t tx_cal;
+       int calender_16_block = channel / 15;
+       int calender_16_index = channel % 15 + 1;
+       int index = calender_16_block * 16 + calender_16_index;
+       int node = (intf >> 4) & 0xf;
+       int interface = intf & 0xf;
+
+       /* Program the link status on first channel */
+       if (calender_16_index == 1) {
+               tx_cal.u64 = 0;
+               tx_cal.s.ctl = 1;
+               csr_wr_node(node, CVMX_ILK_TXX_CAL_ENTRYX(index - 1, interface),
+                           tx_cal.u64);
+       }
+       tx_cal.u64 = 0;
+       tx_cal.s.ctl = 0;
+       tx_cal.s.channel = channel;
+       csr_wr_node(node, CVMX_ILK_TXX_CAL_ENTRYX(index, interface),
+                   tx_cal.u64);
+}
+
+/**
+ * @INTERNAL
+ * Setup the channel's tx calendar entry.
+ *
+ * @param interface channel belongs to
+ * @param channel whose calendar entry is to be updated
+ * @param bpid assigned to the channel
+ */
+void __cvmx_ilk_write_tx_cal_entry(int interface, int channel,
+                                  unsigned char bpid)
+{
+       if (OCTEON_IS_MODEL(OCTEON_CN68XX))
+               __cvmx_ilk_write_tx_cal_entry_cn68xx(interface, channel, bpid);
+       else
+               __cvmx_ilk_write_tx_cal_entry_cn78xx(interface, channel, bpid);
+}
+
+void __cvmx_ilk_write_rx_cal_entry_cn78xx(int intf, int channel,
+                                         unsigned char bpid)
+{
+       cvmx_ilk_rxx_cal_entryx_t rx_cal;
+       int calender_16_block = channel / 15;
+       int calender_16_index = channel % 15 + 1;
+       int index = calender_16_block * 16 + calender_16_index;
+       int node = (intf >> 4) & 0xf;
+       int interface = intf & 0xf;
+
+       /* Program the link status on first channel */
+       if (calender_16_index == 1) {
+               rx_cal.u64 = 0;
+               rx_cal.s.ctl = 1;
+               csr_wr_node(node, CVMX_ILK_RXX_CAL_ENTRYX(index - 1, interface),
+                           rx_cal.u64);
+       }
+       rx_cal.u64 = 0;
+       rx_cal.s.ctl = 0;
+       rx_cal.s.channel = channel;
+       csr_wr_node(node, CVMX_ILK_RXX_CAL_ENTRYX(index, interface),
+                   rx_cal.u64);
+}
+
+void __cvmx_ilk_write_rx_cal_entry_cn68xx(int interface, int channel,
+                                         unsigned char pipe)
+{
+       cvmx_ilk_rxx_idx_cal_t rx_idx;
+       cvmx_ilk_rxx_mem_cal0_t rx_cal0;
+       cvmx_ilk_rxx_mem_cal1_t rx_cal1;
+       int entry;
+       int window;
+       int window_entry;
+
+       /*
+        * The calendar has 288 entries. Each calendar entry represents a
+        * channel's flow control state or the link flow control state.
+        * Starting with the first entry, every sixteenth entry is used for the
+        * link flow control state. The other 15 entries are used for the
+        * channels flow control state:
+        * entry 0   ----> link flow control state
+        * entry 1   ----> channel 0 flow control state
+        * entry 2   ----> channel 1 flow control state
+        * ...
+        * entry 15  ----> channel 14 flow control state
+        * entry 16  ----> link flow control state
+        * entry 17  ----> channel 15 flow control state
+        *
+        * Also, the calendar is accessed via windows into it. Each window maps
+        * to 8 entries.
+        */
+       entry = 1 + channel + (channel / 15);
+       window = entry / 8;
+       window_entry = entry % 8;
+
+       /* Indicate the window we need to access */
+       rx_idx.u64 = 0;
+       rx_idx.s.index = window;
+       csr_wr(CVMX_ILK_RXX_IDX_CAL(interface), rx_idx.u64);
+
+       /* Get the window's current value */
+       rx_cal0.u64 = csr_rd(CVMX_ILK_RXX_MEM_CAL0(interface));
+       rx_cal1.u64 = csr_rd(CVMX_ILK_RXX_MEM_CAL1(interface));
+
+       /* Force every sixteenth entry as link flow control state */
+       if ((window & 1) == 0)
+               rx_cal0.s.entry_ctl0 = LINK;
+
+       /* Update the entry */
+       switch (window_entry) {
+       case 0:
+               rx_cal0.s.entry_ctl0 = 0;
+               rx_cal0.s.port_pipe0 = pipe;
+               break;
+       case 1:
+               rx_cal0.s.entry_ctl1 = 0;
+               rx_cal0.s.port_pipe1 = pipe;
+               break;
+       case 2:
+               rx_cal0.s.entry_ctl2 = 0;
+               rx_cal0.s.port_pipe2 = pipe;
+               break;
+       case 3:
+               rx_cal0.s.entry_ctl3 = 0;
+               rx_cal0.s.port_pipe3 = pipe;
+               break;
+       case 4:
+               rx_cal1.s.entry_ctl4 = 0;
+               rx_cal1.s.port_pipe4 = pipe;
+               break;
+       case 5:
+               rx_cal1.s.entry_ctl5 = 0;
+               rx_cal1.s.port_pipe5 = pipe;
+               break;
+       case 6:
+               rx_cal1.s.entry_ctl6 = 0;
+               rx_cal1.s.port_pipe6 = pipe;
+               break;
+       case 7:
+               rx_cal1.s.entry_ctl7 = 0;
+               rx_cal1.s.port_pipe7 = pipe;
+               break;
+       }
+
+       /* Write the window new value */
+       csr_wr(CVMX_ILK_RXX_MEM_CAL0(interface), rx_cal0.u64);
+       csr_wr(CVMX_ILK_RXX_MEM_CAL1(interface), rx_cal1.u64);
+}
+
+/**
+ * @INTERNAL
+ * Setup the channel's rx calendar entry.
+ *
+ * @param interface channel belongs to
+ * @param channel whose calendar entry is to be updated
+ * @param pipe PKO assigned to the channel
+ */
+void __cvmx_ilk_write_rx_cal_entry(int interface, int channel,
+                                  unsigned char pipe)
+{
+       if (OCTEON_IS_MODEL(OCTEON_CN68XX))
+               __cvmx_ilk_write_rx_cal_entry_cn68xx(interface, channel, pipe);
+       else
+               __cvmx_ilk_write_rx_cal_entry_cn78xx(interface, channel, pipe);
+}
+
+/**
+ * @INTERNAL
+ * Probe a ILK interface and determine the number of ports
+ * connected to it. The ILK interface should still be down
+ * after this call.
+ *
+ * @param xiface Interface to probe
+ *
+ * @return Number of ports on the interface. Zero to disable.
+ */
+int __cvmx_helper_ilk_probe(int xiface)
+{
+       int res = 0;
+       int interface;
+       struct cvmx_xiface xi = cvmx_helper_xiface_to_node_interface(xiface);
+
+       if (!octeon_has_feature(OCTEON_FEATURE_ILK))
+               return res;
+
+       interface = xi.interface - CVMX_ILK_GBL_BASE();
+       if (interface >= CVMX_NUM_ILK_INTF)
+               return 0;
+
+       /* the configuration should be done only once */
+       if (cvmx_ilk_get_intf_ena(xiface))
+               return cvmx_ilk_chans[xi.node][interface];
+
+       /* configure lanes and enable the link */
+       res = cvmx_ilk_start_interface(((xi.node << 4) | interface),
+                                      cvmx_ilk_lane_mask[xi.node][interface]);
+       if (res < 0)
+               return 0;
+
+       res = __cvmx_helper_ilk_enumerate(xiface);
+
+       return res;
+}
+
+static int __cvmx_helper_ilk_init_port_cn68xx(int xiface)
+{
+       int i, j, res = -1;
+       static int pipe_base = 0, pknd_base;
+       static cvmx_ilk_pipe_chan_t *pch = NULL, *tmp;
+       static cvmx_ilk_chan_pknd_t *chpknd = NULL, *tmp1;
+       static cvmx_ilk_cal_entry_t *calent = NULL, *tmp2;
+       int enable_rx_cal = 1;
+       int interface;
+       struct cvmx_xiface xi = cvmx_helper_xiface_to_node_interface(xiface);
+       int intf;
+       int num_chans;
+
+       interface = xi.interface - CVMX_ILK_GBL_BASE();
+       intf = (xi.node << 4) | interface;
+       if (interface >= CVMX_NUM_ILK_INTF)
+               return 0;
+
+       num_chans = cvmx_ilk_chans[0][interface];
+
+       /* set up channel to pkind mapping */
+       if (pknd_base == 0)
+               pknd_base = cvmx_helper_get_pknd(xiface, 0);
+
+       /* set up the group of pipes available to ilk */
+       if (pipe_base == 0)
+               pipe_base =
+                       __cvmx_pko_get_pipe(interface + CVMX_ILK_GBL_BASE(), 0);
+
+       if (pipe_base == -1) {
+               pipe_base = 0;
+               return 0;
+       }
+
+       res = cvmx_ilk_set_pipe(xiface, pipe_base,
+                               cvmx_ilk_chans[0][interface]);
+       if (res < 0)
+               return 0;
+
+       /* set up pipe to channel mapping */
+       i = pipe_base;
+       if (!pch) {
+               pch = (cvmx_ilk_pipe_chan_t *)cvmx_bootmem_alloc(
+                       num_chans * sizeof(cvmx_ilk_pipe_chan_t),
+                       sizeof(cvmx_ilk_pipe_chan_t));
+               if (!pch)
+                       return 0;
+       }
+
+       memset(pch, 0, num_chans * sizeof(cvmx_ilk_pipe_chan_t));
+       tmp = pch;
+       for (j = 0; j < num_chans; j++) {
+               tmp->pipe = i++;
+               tmp->chan = j;
+               tmp++;
+       }
+       res = cvmx_ilk_tx_set_channel(interface, pch,
+                                     cvmx_ilk_chans[0][interface]);
+       if (res < 0) {
+               res = 0;
+               goto err_free_pch;
+       }
+       pipe_base += cvmx_ilk_chans[0][interface];
+       i = pknd_base;
+       if (!chpknd) {
+               chpknd = (cvmx_ilk_chan_pknd_t *)cvmx_bootmem_alloc(
+                       CVMX_ILK_MAX_PKNDS * sizeof(cvmx_ilk_chan_pknd_t),
+                       sizeof(cvmx_ilk_chan_pknd_t));
+               if (!chpknd) {
+                       pipe_base -= cvmx_ilk_chans[xi.node][interface];
+                       res = 0;
+                       goto err_free_pch;
+               }
+       }
+
+       memset(chpknd, 0, CVMX_ILK_MAX_PKNDS * sizeof(cvmx_ilk_chan_pknd_t));
+       tmp1 = chpknd;
+       for (j = 0; j < cvmx_ilk_chans[xi.node][interface]; j++) {
+               tmp1->chan = j;
+               tmp1->pknd = i++;
+               tmp1++;
+       }
+
+       res = cvmx_ilk_rx_set_pknd(xiface, chpknd,
+                                  cvmx_ilk_chans[xi.node][interface]);
+       if (res < 0) {
+               pipe_base -= cvmx_ilk_chans[xi.node][interface];
+               res = 0;
+               goto err_free_chpknd;
+       }
+       pknd_base += cvmx_ilk_chans[xi.node][interface];
+
+       /* Set up tx calendar */
+       if (!calent) {
+               calent = (cvmx_ilk_cal_entry_t *)cvmx_bootmem_alloc(
+                       CVMX_ILK_MAX_PIPES * sizeof(cvmx_ilk_cal_entry_t),
+                       sizeof(cvmx_ilk_cal_entry_t));
+               if (!calent) {
+                       pipe_base -= cvmx_ilk_chans[xi.node][interface];
+                       pknd_base -= cvmx_ilk_chans[xi.node][interface];
+                       res = 0;
+                       goto err_free_chpknd;
+               }
+       }
+
+       memset(calent, 0, CVMX_ILK_MAX_PIPES * sizeof(cvmx_ilk_cal_entry_t));
+       tmp1 = chpknd;
+       tmp2 = calent;
+       for (j = 0; j < cvmx_ilk_chans[xi.node][interface]; j++) {
+               tmp2->pipe_bpid = tmp1->pknd;
+               tmp2->ent_ctrl = PIPE_BPID;
+               tmp1++;
+               tmp2++;
+       }
+       res = cvmx_ilk_cal_setup_tx(intf, cvmx_ilk_chans[xi.node][interface],
+                                   calent, 1);
+       if (res < 0) {
+               pipe_base -= cvmx_ilk_chans[xi.node][interface];
+               pknd_base -= cvmx_ilk_chans[xi.node][interface];
+               res = 0;
+               goto err_free_calent;
+       }
+
+       /* set up rx calendar. allocated memory can be reused.
+        * this is because max pkind is always less than max pipe
+        */
+       memset(calent, 0, CVMX_ILK_MAX_PIPES * sizeof(cvmx_ilk_cal_entry_t));
+       tmp = pch;
+       tmp2 = calent;
+       for (j = 0; j < cvmx_ilk_chans[0][interface]; j++) {
+               tmp2->pipe_bpid = tmp->pipe;
+               tmp2->ent_ctrl = PIPE_BPID;
+               tmp++;
+               tmp2++;
+       }
+       if (cvmx_ilk_use_la_mode(interface, 0))
+               enable_rx_cal = cvmx_ilk_la_mode_enable_rx_calendar(interface);
+       else
+               enable_rx_cal = 1;
+
+       res = cvmx_ilk_cal_setup_rx(intf, cvmx_ilk_chans[xi.node][interface],
+                                   calent, CVMX_ILK_RX_FIFO_WM, enable_rx_cal);
+       if (res < 0) {
+               pipe_base -= cvmx_ilk_chans[xi.node][interface];
+               pknd_base -= cvmx_ilk_chans[xi.node][interface];
+               res = 0;
+               goto err_free_calent;
+       }
+       goto out;
+
+err_free_calent:
+       /* no free() for cvmx_bootmem_alloc() */
+
+err_free_chpknd:
+       /* no free() for cvmx_bootmem_alloc() */
+
+err_free_pch:
+       /* no free() for cvmx_bootmem_alloc() */
+out:
+       return res;
+}
+
+static int __cvmx_helper_ilk_init_port_cn78xx(int xiface)
+{
+       struct cvmx_xiface xi = cvmx_helper_xiface_to_node_interface(xiface);
+       int interface;
+       int intf;
+
+       interface = xi.interface - CVMX_ILK_GBL_BASE();
+       intf = (xi.node << 4) | interface;
+       if (interface >= CVMX_NUM_ILK_INTF)
+               return 0;
+
+       if (OCTEON_IS_MODEL(OCTEON_CN78XX)) {
+               struct cvmx_pki_style_config style_cfg;
+               int num_channels = cvmx_ilk_chans[xi.node][interface];
+               int index, i;
+
+               for (i = 0; i < num_channels; i++) {
+                       int pknd;
+
+                       index = (i % 8);
+
+                       /* Set jabber to allow max sized packets */
+                       if (i == 0)
+                               csr_wr_node(xi.node,
+                                           CVMX_ILK_RXX_JABBER(interface),
+                                           0xfff8);
+
+                       /* Setup PKND */
+                       pknd = cvmx_helper_get_pknd(xiface, index);
+                       csr_wr_node(xi.node, CVMX_ILK_RXX_CHAX(i, interface),
+                                   pknd);
+                       cvmx_pki_read_style_config(
+                               0, pknd, CVMX_PKI_CLUSTER_ALL, &style_cfg);
+                       style_cfg.parm_cfg.qpg_port_sh = 0;
+                       /* 256 channels */
+                       style_cfg.parm_cfg.qpg_port_msb = 8;
+                       cvmx_pki_write_style_config(
+                               0, pknd, CVMX_PKI_CLUSTER_ALL, &style_cfg);
+               }
+
+               cvmx_ilk_cal_setup_tx(intf, num_channels, NULL, 1);
+               cvmx_ilk_cal_setup_rx(intf, num_channels, NULL,
+                                     CVMX_ILK_RX_FIFO_WM, 1);
+       }
+       return 0;
+}
+
+static int __cvmx_helper_ilk_init_port(int xiface)
+{
+       if (OCTEON_IS_MODEL(OCTEON_CN68XX))
+               return __cvmx_helper_ilk_init_port_cn68xx(xiface);
+       else
+               return __cvmx_helper_ilk_init_port_cn78xx(xiface);
+}
+
+/**
+ * @INTERNAL
+ * Bringup and enable ILK interface. After this call packet
+ * I/O should be fully functional. This is called with IPD
+ * enabled but PKO disabled.
+ *
+ * @param xiface Interface to bring up
+ *
+ * @return Zero on success, negative on failure
+ */
+int __cvmx_helper_ilk_enable(int xiface)
+{
+       if (__cvmx_helper_ilk_init_port(xiface) < 0)
+               return -1;
+
+       return cvmx_ilk_enable(xiface);
+}
+
+/**
+ * @INTERNAL
+ * Return the link state of an IPD/PKO port as returned by ILK link status.
+ *
+ * @param ipd_port IPD/PKO port to query
+ *
+ * @return Link state
+ */
+cvmx_helper_link_info_t __cvmx_helper_ilk_link_get(int ipd_port)
+{
+       cvmx_helper_link_info_t result;
+       int xiface = cvmx_helper_get_interface_num(ipd_port);
+       struct cvmx_xiface xi = cvmx_helper_xiface_to_node_interface(xiface);
+       int interface;
+       int retry_count = 0;
+       cvmx_ilk_rxx_cfg1_t ilk_rxx_cfg1;
+       cvmx_ilk_rxx_int_t ilk_rxx_int;
+       int lane_mask = 0;
+       int i;
+       int node = xi.node;
+
+       result.u64 = 0;
+       interface = xi.interface - CVMX_ILK_GBL_BASE();
+
+retry:
+       retry_count++;
+       if (retry_count > 200)
+               goto fail;
+
+       /* Read RX config and status bits */
+       ilk_rxx_cfg1.u64 = csr_rd_node(node, CVMX_ILK_RXX_CFG1(interface));
+       ilk_rxx_int.u64 = csr_rd_node(node, CVMX_ILK_RXX_INT(interface));
+
+       if (ilk_rxx_cfg1.s.rx_bdry_lock_ena == 0) {
+               /* (GSER-21957) GSER RX Equalization may make >= 5gbaud non-KR
+                * channel better
+                */
+               if (OCTEON_IS_MODEL(OCTEON_CN78XX)) {
+                       int qlm, lane_mask;
+
+                       for (qlm = 4; qlm < 8; qlm++) {
+                               lane_mask = 1 << (qlm - 4) * 4;
+                               if (lane_mask &
+                                   cvmx_ilk_lane_mask[node][interface]) {
+                                       if (__cvmx_qlm_rx_equalization(
+                                                   node, qlm, -1))
+                                               goto retry;
+                               }
+                       }
+               }
+
+               /* Clear the boundary lock status bit */
+               ilk_rxx_int.u64 = 0;
+               ilk_rxx_int.s.word_sync_done = 1;
+               csr_wr_node(node, CVMX_ILK_RXX_INT(interface), ilk_rxx_int.u64);
+
+               /* We need to start looking for word boundary lock */
+               ilk_rxx_cfg1.s.rx_bdry_lock_ena =
+                       cvmx_ilk_lane_mask[node][interface];
+               ilk_rxx_cfg1.s.rx_align_ena = 0;
+               csr_wr_node(node, CVMX_ILK_RXX_CFG1(interface),
+                           ilk_rxx_cfg1.u64);
+               //debug("ILK%d: Looking for word boundary lock\n", interface);
+               udelay(50);
+               goto retry;
+       }
+
+       if (ilk_rxx_cfg1.s.rx_align_ena == 0) {
+               if (ilk_rxx_int.s.word_sync_done) {
+                       /* Clear the lane align status bits */
+                       ilk_rxx_int.u64 = 0;
+                       ilk_rxx_int.s.lane_align_fail = 1;
+                       ilk_rxx_int.s.lane_align_done = 1;
+                       csr_wr_node(node, CVMX_ILK_RXX_INT(interface),
+                                   ilk_rxx_int.u64);
+
+                       ilk_rxx_cfg1.s.rx_align_ena = 1;
+                       csr_wr_node(node, CVMX_ILK_RXX_CFG1(interface),
+                                   ilk_rxx_cfg1.u64);
+                       //printf("ILK%d: Looking for lane alignment\n", interface);
+               }
+               udelay(50);
+               goto retry;
+       }
+
+       if (ilk_rxx_int.s.lane_align_fail) {
+               ilk_rxx_cfg1.s.rx_bdry_lock_ena = 0;
+               ilk_rxx_cfg1.s.rx_align_ena = 0;
+               csr_wr_node(node, CVMX_ILK_RXX_CFG1(interface),
+                           ilk_rxx_cfg1.u64);
+               //debug("ILK%d: Lane alignment failed\n", interface);
+               goto fail;
+       }
+
+       lane_mask = ilk_rxx_cfg1.s.rx_bdry_lock_ena;
+
+       if (ilk_rxx_cfg1.s.pkt_ena == 0 && ilk_rxx_int.s.lane_align_done) {
+               cvmx_ilk_txx_cfg1_t ilk_txx_cfg1;
+
+               ilk_txx_cfg1.u64 =
+                       csr_rd_node(node, CVMX_ILK_TXX_CFG1(interface));
+               ilk_rxx_cfg1.u64 =
+                       csr_rd_node(node, CVMX_ILK_RXX_CFG1(interface));
+               ilk_rxx_cfg1.s.pkt_ena = ilk_txx_cfg1.s.pkt_ena;
+               csr_wr_node(node, CVMX_ILK_RXX_CFG1(interface),
+                           ilk_rxx_cfg1.u64);
+
+               if (OCTEON_IS_MODEL(OCTEON_CN68XX)) {
+                       /*
+                        * Enable rxf_ctl_perr, rxf_lnk0_perr, rxf_lnk1_perr,
+                        * pop_empty, push_full.
+                        */
+                       csr_wr(CVMX_ILK_GBL_INT_EN, 0x1f);
+                       /* Enable bad_pipe, bad_seq, txf_err */
+                       csr_wr(CVMX_ILK_TXX_INT_EN(interface), 0x7);
+
+                       /*
+                        * Enable crc24_err, lane_bad_word,
+                        * pkt_drop_{rid,rxf,sop}
+                        */
+                       csr_wr(CVMX_ILK_RXX_INT_EN(interface), 0x1e2);
+               }
+               /* Need to enable ILK interrupts for 78xx */
+
+               for (i = 0; i < CVMX_ILK_MAX_LANES(); i++) {
+                       if ((1 << i) & lane_mask) {
+                               /* clear pending interrupts, before enabling. */
+                               csr_wr_node(node, CVMX_ILK_RX_LNEX_INT(i),
+                                           0x1ff);
+                               /* Enable bad_64b67b, bdry_sync_loss, crc32_err,
+                                * dskew_fifo_ovfl, scrm_sync_loss,
+                                * serdes_lock_loss, stat_msg, ukwn_cntl_word
+                                */
+                               if (OCTEON_IS_MODEL(OCTEON_CN68XX))
+                                       csr_wr(CVMX_ILK_RX_LNEX_INT_EN(i),
+                                              0x1ff);
+                       }
+               }
+
+               //debug("ILK%d: Lane alignment complete\n", interface);
+       }
+
+       /* Enable error interrupts, now link is up */
+       cvmx_error_enable_group(CVMX_ERROR_GROUP_ILK,
+                               node | (interface << 2) | (lane_mask << 4));
+
+       result.s.link_up = 1;
+       result.s.full_duplex = 1;
+       if (OCTEON_IS_MODEL(OCTEON_CN78XX)) {
+               int qlm = cvmx_qlm_lmac(xiface, 0);
+
+               result.s.speed = cvmx_qlm_get_gbaud_mhz(qlm) * 64 / 67;
+       } else {
+               result.s.speed =
+                       cvmx_qlm_get_gbaud_mhz(1 + interface) * 64 / 67;
+       }
+       result.s.speed *= cvmx_pop(lane_mask);
+
+       return result;
+
+fail:
+       if (ilk_rxx_cfg1.s.pkt_ena) {
+               /* Disable the interface */
+               ilk_rxx_cfg1.s.pkt_ena = 0;
+               csr_wr_node(node, CVMX_ILK_RXX_CFG1(interface),
+                           ilk_rxx_cfg1.u64);
+
+               /* Disable error interrupts */
+               for (i = 0; i < CVMX_ILK_MAX_LANES(); i++) {
+                       /* Disable bad_64b67b, bdry_sync_loss, crc32_err,
+                        * dskew_fifo_ovfl, scrm_sync_loss, serdes_lock_loss,
+                        * stat_msg, ukwn_cntl_word
+                        */
+                       if ((1 << i) & lane_mask) {
+                               csr_wr_node(node, CVMX_ILK_RX_LNEX_INT(i),
+                                           0x1ff);
+                               if (OCTEON_IS_MODEL(OCTEON_CN68XX))
+                                       csr_wr(CVMX_ILK_RX_LNEX_INT_EN(i),
+                                              ~0x1ff);
+                       }
+               }
+               /* Disable error interrupts */
+               cvmx_error_enable_group(CVMX_ERROR_GROUP_ILK, 0);
+       }
+
+       return result;
+}
+
+/**
+ * @INTERNAL
+ * Set the link state of an IPD/PKO port.
+ *
+ * @param ipd_port  IPD/PKO port to configure
+ * @param link_info The new link state
+ *
+ * @return Zero on success, negative on failure
+ */
+int __cvmx_helper_ilk_link_set(int ipd_port, cvmx_helper_link_info_t link_info)
+{
+       /* Do nothing */
+       return 0;
+}