* [net-next 0/3][pull request] Intel Wired LAN ver Updates 2019-07-03
@ 2019-07-04 2:12 Jeff Kirsher
2019-07-04 2:12 ` [net-next 1/3] ice: Initialize and register platform device to provide RDMA Jeff Kirsher
` (3 more replies)
0 siblings, 4 replies; 15+ messages in thread
From: Jeff Kirsher @ 2019-07-04 2:12 UTC (permalink / raw)
To: davem, dledford, jgg
Cc: Jeff Kirsher, netdev, linux-rdma, nhorman, sassmann,
mustafa.ismail, shiraz.saleem, david.m.ertman
This series contains updates to i40e an ice drivers only and is required
for a series of changes being submitted to the RDMA maintainer/tree.
Vice Versa, the Intel RDMA driver patches could not be applied to
net-next due to dependencies to the changes currently in the for-next
branch of the rdma git tree. The Intel RDMA driver patches are
available in the following git repository:
git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/rdma for-next
(reference only)
Dave initializes the ice driver to support RDMA by creating and
registering a platform device for the RDMA driver to register to on a
virtual bus by utilizing the platform bus to provide this access.
Followed up by operations for the peer device/driver to communicate with
each other to request resources and manage event notification.
Shiraz does the similar work in the i40e driver, while retaining the
old global register/unregister calls exported for i40iw until that
driver is removed from the kernel.
The following are changes since commit a51df9f8da43e8bf9e508143630849b7d696e053:
gve: fix -ENOMEM null check on a page allocation
and are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next-queue 100GbE
Shiraz Saleem (1):
i40e: Register RDMA client devices to the virtual platform bus
Tony Nguyen (2):
ice: Initialize and register platform device to provide RDMA
ice: Implement peer communications
drivers/net/ethernet/intel/i40e/i40e_client.c | 116 +-
drivers/net/ethernet/intel/i40e/i40e_client.h | 8 +
drivers/net/ethernet/intel/ice/Makefile | 1 +
drivers/net/ethernet/intel/ice/ice.h | 15 +
.../net/ethernet/intel/ice/ice_adminq_cmd.h | 33 +
drivers/net/ethernet/intel/ice/ice_common.c | 194 +++
drivers/net/ethernet/intel/ice/ice_common.h | 9 +
drivers/net/ethernet/intel/ice/ice_dcb_lib.c | 64 +
drivers/net/ethernet/intel/ice/ice_dcb_lib.h | 2 +
.../net/ethernet/intel/ice/ice_hw_autogen.h | 1 +
drivers/net/ethernet/intel/ice/ice_idc.c | 1354 +++++++++++++++++
drivers/net/ethernet/intel/ice/ice_idc.h | 377 +++++
drivers/net/ethernet/intel/ice/ice_idc_int.h | 82 +
drivers/net/ethernet/intel/ice/ice_lib.c | 35 +-
drivers/net/ethernet/intel/ice/ice_lib.h | 2 +
drivers/net/ethernet/intel/ice/ice_main.c | 94 +-
drivers/net/ethernet/intel/ice/ice_sched.c | 69 +-
drivers/net/ethernet/intel/ice/ice_switch.c | 27 +
drivers/net/ethernet/intel/ice/ice_switch.h | 4 +
drivers/net/ethernet/intel/ice/ice_type.h | 4 +
.../net/ethernet/intel/ice/ice_virtchnl_pf.c | 25 -
21 files changed, 2462 insertions(+), 54 deletions(-)
create mode 100644 drivers/net/ethernet/intel/ice/ice_idc.c
create mode 100644 drivers/net/ethernet/intel/ice/ice_idc.h
create mode 100644 drivers/net/ethernet/intel/ice/ice_idc_int.h
--
2.21.0
^ permalink raw reply [flat|nested] 15+ messages in thread
* [net-next 1/3] ice: Initialize and register platform device to provide RDMA
2019-07-04 2:12 [net-next 0/3][pull request] Intel Wired LAN ver Updates 2019-07-03 Jeff Kirsher
@ 2019-07-04 2:12 ` Jeff Kirsher
2019-07-04 12:16 ` Jason Gunthorpe
2019-07-04 2:12 ` [net-next 2/3] ice: Implement peer communications Jeff Kirsher
` (2 subsequent siblings)
3 siblings, 1 reply; 15+ messages in thread
From: Jeff Kirsher @ 2019-07-04 2:12 UTC (permalink / raw)
To: davem, dledford, jgg
Cc: Tony Nguyen, netdev, linux-rdma, nhorman, sassmann, poswald,
mustafa.ismail, shiraz.saleem, Dave Ertman, Andrew Bowers,
Jeff Kirsher
From: Tony Nguyen <anthony.l.nguyen@intel.com>
The RDMA block does not advertise on the PCI bus or any other bus.
Thus the ice driver needs to provide access to the RDMA hardware block
via a virtual bus; utilize the platform bus to provide this access.
This patch initializes the driver to support RDMA as well as creates
and registers a platform device for the RDMA driver to register to. At
this point the driver is fully initialized to register a platform
driver, however, can not yet register as the ops have not been
implemented.
We refer to the interaction of this platform device as Inter-Driver
Communication (IDC); where the platform device is referred to as the peer
device and the platform driver is referred to as the peer driver.
Signed-off-by: Dave Ertman <david.m.ertman@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
drivers/net/ethernet/intel/ice/Makefile | 1 +
drivers/net/ethernet/intel/ice/ice.h | 13 +
.../net/ethernet/intel/ice/ice_adminq_cmd.h | 1 +
drivers/net/ethernet/intel/ice/ice_common.c | 5 +
drivers/net/ethernet/intel/ice/ice_dcb_lib.c | 31 ++
drivers/net/ethernet/intel/ice/ice_dcb_lib.h | 2 +
.../net/ethernet/intel/ice/ice_hw_autogen.h | 1 +
drivers/net/ethernet/intel/ice/ice_idc.c | 427 ++++++++++++++++++
drivers/net/ethernet/intel/ice/ice_idc.h | 360 +++++++++++++++
drivers/net/ethernet/intel/ice/ice_idc_int.h | 67 +++
drivers/net/ethernet/intel/ice/ice_main.c | 42 ++
drivers/net/ethernet/intel/ice/ice_type.h | 1 +
12 files changed, 951 insertions(+)
create mode 100644 drivers/net/ethernet/intel/ice/ice_idc.c
create mode 100644 drivers/net/ethernet/intel/ice/ice_idc.h
create mode 100644 drivers/net/ethernet/intel/ice/ice_idc_int.h
diff --git a/drivers/net/ethernet/intel/ice/Makefile b/drivers/net/ethernet/intel/ice/Makefile
index 2d140ba83781..1500f7724b6f 100644
--- a/drivers/net/ethernet/intel/ice/Makefile
+++ b/drivers/net/ethernet/intel/ice/Makefile
@@ -15,6 +15,7 @@ ice-y := ice_main.o \
ice_sched.o \
ice_lib.o \
ice_txrx.o \
+ ice_idc.o \
ice_ethtool.o
ice-$(CONFIG_PCI_IOV) += ice_virtchnl_pf.o ice_sriov.o
ice-$(CONFIG_DCB) += ice_dcb.o ice_dcb_lib.o
diff --git a/drivers/net/ethernet/intel/ice/ice.h b/drivers/net/ethernet/intel/ice/ice.h
index 9ee6b55553c0..6ee2774699f0 100644
--- a/drivers/net/ethernet/intel/ice/ice.h
+++ b/drivers/net/ethernet/intel/ice/ice.h
@@ -38,6 +38,7 @@
#include "ice_switch.h"
#include "ice_common.h"
#include "ice_sched.h"
+#include "ice_idc_int.h"
#include "ice_virtchnl_pf.h"
#include "ice_sriov.h"
@@ -84,6 +85,7 @@ extern const char ice_drv_ver[];
#define ICE_MAX_SMALL_RSS_QS 8
#define ICE_RES_VALID_BIT 0x8000
#define ICE_RES_MISC_VEC_ID (ICE_RES_VALID_BIT - 1)
+#define ICE_RES_RDMA_VEC_ID (ICE_RES_MISC_VEC_ID - 1)
#define ICE_INVAL_Q_INDEX 0xffff
#define ICE_INVAL_VFID 256
#define ICE_MAX_VF_COUNT 256
@@ -331,6 +333,7 @@ struct ice_q_vector {
enum ice_pf_flags {
ICE_FLAG_MSIX_ENA,
ICE_FLAG_FLTR_SYNC,
+ ICE_FLAG_IWARP_ENA,
ICE_FLAG_RSS_ENA,
ICE_FLAG_SRIOV_ENA,
ICE_FLAG_SRIOV_CAPABLE,
@@ -373,6 +376,9 @@ struct ice_pf {
struct mutex avail_q_mutex; /* protects access to avail_[rx|tx]qs */
struct mutex sw_mutex; /* lock for protecting VSI alloc flow */
u32 msg_enable;
+ /* Total number of MSIX vectors reserved for base driver */
+ u32 num_rdma_msix;
+ u32 rdma_base_vector;
u32 hw_csum_rx_error;
u32 oicr_idx; /* Other interrupt cause MSIX vector index */
u32 num_avail_sw_msix; /* remaining MSIX SW vectors left unclaimed */
@@ -399,6 +405,7 @@ struct ice_pf {
unsigned long tx_timeout_last_recovery;
u32 tx_timeout_recovery_level;
char int_name[ICE_INT_NAME_STR_LEN];
+ struct ice_peer_dev_int **peers;
u32 sw_int_count;
};
@@ -406,6 +413,8 @@ struct ice_netdev_priv {
struct ice_vsi *vsi;
};
+extern struct ida ice_peer_index_ida;
+
/**
* ice_irq_dynamic_ena - Enable default interrupt generation settings
* @hw: pointer to HW struct
@@ -463,6 +472,10 @@ int ice_set_rss(struct ice_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size);
int ice_get_rss(struct ice_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size);
void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size);
void ice_print_link_msg(struct ice_vsi *vsi, bool isup);
+int ice_init_peer_devices(struct ice_pf *pf);
+int
+ice_for_each_peer(struct ice_pf *pf, void *data,
+ int (*fn)(struct ice_peer_dev_int *, void *));
#ifdef CONFIG_DCB
int ice_pf_ena_all_vsi(struct ice_pf *pf, bool locked);
void ice_pf_dis_all_vsi(struct ice_pf *pf, bool locked);
diff --git a/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h b/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h
index 765e3c2ed045..1e1bd5f0c2a2 100644
--- a/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h
+++ b/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h
@@ -96,6 +96,7 @@ struct ice_aqc_list_caps_elem {
#define ICE_AQC_CAPS_TXQS 0x0042
#define ICE_AQC_CAPS_MSIX 0x0043
#define ICE_AQC_CAPS_MAX_MTU 0x0047
+#define ICE_AQC_CAPS_IWARP 0x0051
u8 major_ver;
u8 minor_ver;
diff --git a/drivers/net/ethernet/intel/ice/ice_common.c b/drivers/net/ethernet/intel/ice/ice_common.c
index 2e0731c1e1a3..2788b2ede440 100644
--- a/drivers/net/ethernet/intel/ice/ice_common.c
+++ b/drivers/net/ethernet/intel/ice/ice_common.c
@@ -1613,6 +1613,11 @@ ice_parse_caps(struct ice_hw *hw, void *buf, u32 cap_count,
"%s: MSIX first vector index = %d\n", prefix,
caps->msix_vector_first_id);
break;
+ case ICE_AQC_CAPS_IWARP:
+ caps->iwarp = (number == 1);
+ ice_debug(hw, ICE_DBG_INIT,
+ "%s: iWARP = %d\n", prefix, caps->iwarp);
+ break;
case ICE_AQC_CAPS_MAX_MTU:
caps->max_mtu = number;
ice_debug(hw, ICE_DBG_INIT, "%s: max MTU = %d\n",
diff --git a/drivers/net/ethernet/intel/ice/ice_dcb_lib.c b/drivers/net/ethernet/intel/ice/ice_dcb_lib.c
index fe88b127ca42..c6466e16eef1 100644
--- a/drivers/net/ethernet/intel/ice/ice_dcb_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_dcb_lib.c
@@ -501,6 +501,37 @@ ice_tx_prepare_vlan_flags_dcb(struct ice_ring *tx_ring,
return 0;
}
+/**
+ * ice_setup_dcb_qos_info - Setup DCB QoS information
+ * @pf: ptr to ice_pf
+ * @qos_info: QoS param instance
+ */
+void ice_setup_dcb_qos_info(struct ice_pf *pf, struct ice_qos_params *qos_info)
+{
+ struct ice_dcbx_cfg *dcbx_cfg;
+ u32 up2tc;
+ int i;
+
+ dcbx_cfg = &pf->hw.port_info->local_dcbx_cfg;
+ up2tc = rd32(&pf->hw, PRTDCB_TUP2TC);
+ qos_info->num_apps = dcbx_cfg->numapps;
+
+ qos_info->num_tc = ice_dcb_get_num_tc(dcbx_cfg);
+
+ for (i = 0; i < ICE_IDC_MAX_USER_PRIORITY; i++)
+ qos_info->up2tc[i] = (up2tc >> (i * 3)) & 0x7;
+
+ for (i = 0; i < IEEE_8021QAZ_MAX_TCS; i++)
+ qos_info->tc_info[i].rel_bw =
+ dcbx_cfg->etscfg.tcbwtable[i];
+
+ for (i = 0; i < qos_info->num_apps; i++) {
+ qos_info->apps[i].priority = dcbx_cfg->app[i].priority;
+ qos_info->apps[i].prot_id = dcbx_cfg->app[i].prot_id;
+ qos_info->apps[i].selector = dcbx_cfg->app[i].selector;
+ }
+}
+
/**
* ice_dcb_need_recfg - Check if DCB needs reconfig
* @pf: board private structure
diff --git a/drivers/net/ethernet/intel/ice/ice_dcb_lib.h b/drivers/net/ethernet/intel/ice/ice_dcb_lib.h
index 819081053ff5..bce3899afb3a 100644
--- a/drivers/net/ethernet/intel/ice/ice_dcb_lib.h
+++ b/drivers/net/ethernet/intel/ice/ice_dcb_lib.h
@@ -19,6 +19,7 @@ void ice_update_dcb_stats(struct ice_pf *pf);
int
ice_tx_prepare_vlan_flags_dcb(struct ice_ring *tx_ring,
struct ice_tx_buf *first);
+void ice_setup_dcb_qos_info(struct ice_pf *pf, struct ice_qos_params *qos_info);
void
ice_dcb_process_lldp_set_mib_change(struct ice_pf *pf,
struct ice_rq_event_info *event);
@@ -56,6 +57,7 @@ ice_tx_prepare_vlan_flags_dcb(struct ice_ring __always_unused *tx_ring,
#define ice_update_dcb_stats(pf) do {} while (0)
#define ice_vsi_cfg_dcb_rings(vsi) do {} while (0)
+#define ice_setup_dcb_qos_info(pf, qos_info) do {} while (0)
#define ice_dcb_process_lldp_set_mib_change(pf, event) do {} while (0)
#define ice_set_cgd_num(tlan_ctx, ring) do {} while (0)
#endif /* CONFIG_DCB */
diff --git a/drivers/net/ethernet/intel/ice/ice_hw_autogen.h b/drivers/net/ethernet/intel/ice/ice_hw_autogen.h
index 6c5ce05742b1..0fa393f909dd 100644
--- a/drivers/net/ethernet/intel/ice/ice_hw_autogen.h
+++ b/drivers/net/ethernet/intel/ice/ice_hw_autogen.h
@@ -55,6 +55,7 @@
#define PRTDCB_GENS 0x00083020
#define PRTDCB_GENS_DCBX_STATUS_S 0
#define PRTDCB_GENS_DCBX_STATUS_M ICE_M(0x7, 0)
+#define PRTDCB_TUP2TC 0x001D26C0
#define GLFLXP_RXDID_FLAGS(_i, _j) (0x0045D000 + ((_i) * 4 + (_j) * 256))
#define GLFLXP_RXDID_FLAGS_FLEXIFLAG_4N_S 0
#define GLFLXP_RXDID_FLAGS_FLEXIFLAG_4N_M ICE_M(0x3F, 0)
diff --git a/drivers/net/ethernet/intel/ice/ice_idc.c b/drivers/net/ethernet/intel/ice/ice_idc.c
new file mode 100644
index 000000000000..0bf66e9fa159
--- /dev/null
+++ b/drivers/net/ethernet/intel/ice/ice_idc.c
@@ -0,0 +1,427 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright (c) 2019, Intel Corporation. */
+
+/* Inter-Driver Communication */
+#include "ice.h"
+#include "ice_lib.h"
+#include "ice_dcb_lib.h"
+
+DEFINE_IDA(ice_peer_index_ida);
+
+static const struct peer_dev_id peer_dev_ids[] = ASSIGN_PEER_INFO;
+
+/**
+ * ice_peer_state_change - manage state machine for peer
+ * @peer_dev: pointer to peer's configuration
+ * @new_state: the state requested to transition into
+ * @locked: boolean to determine if call made with mutex held
+ *
+ * This function handles all state transitions for peer devices.
+ * The state machine is as follows:
+ *
+ * +<-----------------------+<-----------------------------+
+ * |<-------+<----------+ +
+ * \/ + + +
+ * INIT --------------> PROBED --> OPENING CLOSED --> REMOVED
+ * + +
+ * OPENED --> CLOSING
+ * + +
+ * PREP_RST +
+ * + +
+ * PREPPED +
+ * +---------->+
+ */
+static void
+ice_peer_state_change(struct ice_peer_dev_int *peer_dev, long new_state,
+ bool locked)
+{
+ if (!locked)
+ mutex_lock(&peer_dev->peer_dev_state_mutex);
+
+ switch (new_state) {
+ case ICE_PEER_DEV_STATE_INIT:
+ if (test_and_clear_bit(ICE_PEER_DEV_STATE_REMOVED,
+ peer_dev->state)) {
+ set_bit(ICE_PEER_DEV_STATE_INIT, peer_dev->state);
+ dev_info(&peer_dev->peer_dev.platform_dev.dev,
+ "state transition from _REMOVED to _INIT\n");
+ } else {
+ set_bit(ICE_PEER_DEV_STATE_INIT, peer_dev->state);
+ dev_info(&peer_dev->peer_dev.platform_dev.dev,
+ "state set to _INIT\n");
+ }
+ break;
+ case ICE_PEER_DEV_STATE_PROBED:
+ if (test_and_clear_bit(ICE_PEER_DEV_STATE_INIT,
+ peer_dev->state)) {
+ set_bit(ICE_PEER_DEV_STATE_PROBED, peer_dev->state);
+ dev_info(&peer_dev->peer_dev.platform_dev.dev,
+ "state transition from _INIT to _PROBED\n");
+ } else if (test_and_clear_bit(ICE_PEER_DEV_STATE_REMOVED,
+ peer_dev->state)) {
+ set_bit(ICE_PEER_DEV_STATE_PROBED, peer_dev->state);
+ dev_info(&peer_dev->peer_dev.platform_dev.dev,
+ "state transition from _REMOVED to _PROBED\n");
+ } else if (test_and_clear_bit(ICE_PEER_DEV_STATE_OPENING,
+ peer_dev->state)) {
+ set_bit(ICE_PEER_DEV_STATE_PROBED, peer_dev->state);
+ dev_info(&peer_dev->peer_dev.platform_dev.dev,
+ "state transition from _OPENING to _PROBED\n");
+ }
+ break;
+ case ICE_PEER_DEV_STATE_OPENING:
+ if (test_and_clear_bit(ICE_PEER_DEV_STATE_PROBED,
+ peer_dev->state)) {
+ set_bit(ICE_PEER_DEV_STATE_OPENING, peer_dev->state);
+ dev_info(&peer_dev->peer_dev.platform_dev.dev,
+ "state transition from _PROBED to _OPENING\n");
+ } else if (test_and_clear_bit(ICE_PEER_DEV_STATE_CLOSED,
+ peer_dev->state)) {
+ set_bit(ICE_PEER_DEV_STATE_OPENING, peer_dev->state);
+ dev_info(&peer_dev->peer_dev.platform_dev.dev,
+ "state transition from _CLOSED to _OPENING\n");
+ }
+ break;
+ case ICE_PEER_DEV_STATE_OPENED:
+ if (test_and_clear_bit(ICE_PEER_DEV_STATE_OPENING,
+ peer_dev->state)) {
+ set_bit(ICE_PEER_DEV_STATE_OPENED, peer_dev->state);
+ dev_info(&peer_dev->peer_dev.platform_dev.dev,
+ "state transition from _OPENING to _OPENED\n");
+ }
+ break;
+ case ICE_PEER_DEV_STATE_PREP_RST:
+ if (test_and_clear_bit(ICE_PEER_DEV_STATE_OPENED,
+ peer_dev->state)) {
+ set_bit(ICE_PEER_DEV_STATE_PREP_RST, peer_dev->state);
+ dev_info(&peer_dev->peer_dev.platform_dev.dev,
+ "state transition from _OPENED to _PREP_RST\n");
+ }
+ break;
+ case ICE_PEER_DEV_STATE_PREPPED:
+ if (test_and_clear_bit(ICE_PEER_DEV_STATE_PREP_RST,
+ peer_dev->state)) {
+ set_bit(ICE_PEER_DEV_STATE_PREPPED, peer_dev->state);
+ dev_info(&peer_dev->peer_dev.platform_dev.dev,
+ "state transition _PREP_RST to _PREPPED\n");
+ }
+ break;
+ case ICE_PEER_DEV_STATE_CLOSING:
+ if (test_and_clear_bit(ICE_PEER_DEV_STATE_OPENED,
+ peer_dev->state)) {
+ set_bit(ICE_PEER_DEV_STATE_CLOSING, peer_dev->state);
+ dev_info(&peer_dev->peer_dev.platform_dev.dev,
+ "state transition from _OPENED to _CLOSING\n");
+ }
+ if (test_and_clear_bit(ICE_PEER_DEV_STATE_PREPPED,
+ peer_dev->state)) {
+ set_bit(ICE_PEER_DEV_STATE_CLOSING, peer_dev->state);
+ dev_info(&peer_dev->peer_dev.platform_dev.dev,
+ "state transition _PREPPED to _CLOSING\n");
+ }
+ /* NOTE - up to peer to handle this situation correctly */
+ if (test_and_clear_bit(ICE_PEER_DEV_STATE_PREP_RST,
+ peer_dev->state)) {
+ set_bit(ICE_PEER_DEV_STATE_CLOSING, peer_dev->state);
+ dev_warn(&peer_dev->peer_dev.platform_dev.dev,
+ "WARN: Peer state PREP_RST to _CLOSING\n");
+ }
+ break;
+ case ICE_PEER_DEV_STATE_CLOSED:
+ if (test_and_clear_bit(ICE_PEER_DEV_STATE_CLOSING,
+ peer_dev->state)) {
+ set_bit(ICE_PEER_DEV_STATE_CLOSED, peer_dev->state);
+ dev_info(&peer_dev->peer_dev.platform_dev.dev,
+ "state transition from _CLOSING to _CLOSED\n");
+ }
+ break;
+ case ICE_PEER_DEV_STATE_REMOVED:
+ if (test_and_clear_bit(ICE_PEER_DEV_STATE_OPENED,
+ peer_dev->state) ||
+ test_and_clear_bit(ICE_PEER_DEV_STATE_CLOSED,
+ peer_dev->state)) {
+ set_bit(ICE_PEER_DEV_STATE_REMOVED, peer_dev->state);
+ dev_info(&peer_dev->peer_dev.platform_dev.dev,
+ "state from _OPENED/_CLOSED to _REMOVED\n");
+ /* Clear registration for events when peer removed */
+ bitmap_zero(peer_dev->events, ICE_PEER_DEV_STATE_NBITS);
+ }
+ break;
+ default:
+ break;
+ }
+
+ if (!locked)
+ mutex_unlock(&peer_dev->peer_dev_state_mutex);
+}
+
+/**
+ * ice_for_each_peer - iterate across and call function for each peer dev
+ * @pf: pointer to private board struct
+ * @data: data to pass to function on each call
+ * @fn: pointer to function to call for each peer
+ *
+ * This function is to be used similarly to ice_for_each_peer
+ */
+int
+ice_for_each_peer(struct ice_pf *pf, void *data,
+ int (*fn)(struct ice_peer_dev_int *, void *))
+{
+ int i;
+
+ for (i = 0; i < ARRAY_SIZE(peer_dev_ids); i++) {
+ struct ice_peer_dev_int *peer_dev_int;
+ int ret;
+
+ peer_dev_int = pf->peers[i];
+ if (peer_dev_int) {
+ ret = fn(peer_dev_int, data);
+ if (ret)
+ return ret;
+ }
+ }
+
+ return 0;
+}
+
+/**
+ * ice_unreg_peer_device - unregister specified device
+ * @peer_dev_int: ptr to peer device internal
+ * @data: ptr to opaque data
+ *
+ * This function invokes device unregistration, removes ID associated with
+ * the specified device.
+ */
+int
+ice_unreg_peer_device(struct ice_peer_dev_int *peer_dev_int,
+ void __always_unused *data)
+{
+ struct ice_pf *pf;
+
+ if (!peer_dev_int)
+ return 0;
+
+ platform_device_unregister(&peer_dev_int->peer_dev.platform_dev);
+
+ pf = pci_get_drvdata(peer_dev_int->peer_dev.pdev);
+ if (!pf)
+ return 0;
+
+ if (peer_dev_int->ice_peer_wq) {
+ if (peer_dev_int->peer_prep_task.func)
+ cancel_work_sync(&peer_dev_int->peer_prep_task);
+
+ if (peer_dev_int->peer_close_task.func)
+ cancel_work_sync(&peer_dev_int->peer_close_task);
+ destroy_workqueue(peer_dev_int->ice_peer_wq);
+ }
+
+ /* Cleanup the allocated ID for this peer device */
+ ida_simple_remove(&ice_peer_index_ida, peer_dev_int->peer_dev.index);
+
+ devm_kfree(&pf->pdev->dev, peer_dev_int);
+
+ return 0;
+}
+
+/**
+ * ice_unroll_peer - destroy peers and peer_wq in case of error
+ * @peer_dev_int: ptr to peer device internal struct
+ * @data: ptr to opaque data
+ *
+ * This function releases resources in the event of a failure in creating
+ * peer devices or their individual work_queues. Meant to be called from
+ * a ice_for_each_peer invocation
+ */
+int
+ice_unroll_peer(struct ice_peer_dev_int *peer_dev_int,
+ void __always_unused *data)
+{
+ struct ice_pf *pf;
+
+ pf = pci_get_drvdata(peer_dev_int->peer_dev.pdev);
+
+ if (peer_dev_int->ice_peer_wq)
+ destroy_workqueue(peer_dev_int->ice_peer_wq);
+ devm_kfree(&pf->pdev->dev, peer_dev_int);
+
+ return 0;
+}
+
+/**
+ * ice_reserve_peer_qvector - Reserve vector resources for peer drivers
+ * @pf: board private structure to initialize
+ */
+static int ice_reserve_peer_qvector(struct ice_pf *pf)
+{
+ if (test_bit(ICE_FLAG_IWARP_ENA, pf->flags)) {
+ int index;
+
+ index = ice_get_res(pf, pf->irq_tracker, pf->num_rdma_msix,
+ ICE_RES_RDMA_VEC_ID);
+ if (index < 0)
+ return index;
+ pf->num_avail_sw_msix -= pf->num_rdma_msix;
+ pf->rdma_base_vector = index;
+ }
+ return 0;
+}
+
+/**
+ * ice_peer_device_release - release function for platform peer device
+ * @dev: pointer to device structure
+ *
+ * This callback function is accessed by platform bus infrastructure when
+ * all references on the peer platform device we registered are removed
+ * (e.g. the peer priver has been removed). We will use this opportunity
+ * to set the peer's state to init to be ready if the peer driver re-loads.
+ */
+static void ice_peer_device_release(struct device *dev)
+{
+ struct ice_peer_dev_int *peer_dev_int;
+ struct ice_peer_dev *peer_dev;
+
+ peer_dev = dev_to_ice_peer(dev);
+ peer_dev_int = container_of(peer_dev, struct ice_peer_dev_int,
+ peer_dev);
+
+ ice_peer_state_change(peer_dev_int, ICE_PEER_DEV_STATE_INIT, false);
+}
+
+/**
+ * ice_init_peer_devices - initializes peer devices
+ * @pf: ptr to ice_pf
+ *
+ * This function initializes peer devices and associates them with specified
+ * pci_dev as their parent.
+ */
+int ice_init_peer_devices(struct ice_pf *pf)
+{
+ struct pci_dev *pdev = pf->pdev;
+ struct msix_entry *entry = NULL;
+ struct ice_port_info *port_info;
+ struct ice_vsi *vsi;
+ int status = 0;
+ int i;
+
+ /* Reserve vector resources */
+ status = ice_reserve_peer_qvector(pf);
+ if (status < 0) {
+ dev_err(&pdev->dev,
+ "failed to reserve vectors for peer drivers\n");
+ return status;
+ }
+ for (i = 0; i < ARRAY_SIZE(peer_dev_ids); i++) {
+ struct ice_peer_dev_int *peer_dev_int;
+ struct platform_device *platform_dev;
+ struct ice_qos_params *qos_info;
+ struct ice_peer_dev *peer_dev;
+ int j;
+
+ /* don't create an RDMA platform_device if NIC does not
+ * support RDMA functionality
+ */
+ if (peer_dev_ids[i].id == ICE_PEER_RDMA_DEV &&
+ !test_bit(ICE_FLAG_IWARP_ENA, pf->flags)) {
+ dev_warn(&pf->pdev->dev,
+ "RDMA not possible with this device config\n");
+ continue;
+ }
+
+ peer_dev_int = devm_kzalloc(&pdev->dev, sizeof(*peer_dev_int),
+ GFP_KERNEL);
+ if (!peer_dev_int)
+ return -ENOMEM;
+ pf->peers[i] = peer_dev_int;
+
+ mutex_init(&peer_dev_int->peer_dev_state_mutex);
+
+ peer_dev = &peer_dev_int->peer_dev;
+ peer_dev->peer_ops = NULL;
+ peer_dev->hw_addr = (u8 __iomem *)pf->hw.hw_addr;
+ peer_dev->ver.major = ICE_PEER_MAJOR_VER;
+ peer_dev->ver.minor = ICE_PEER_MINOR_VER;
+ peer_dev->ver.support = ICE_IDC_FEATURES;
+ peer_dev->peer_dev_id = peer_dev_ids[i].id;
+ port_info = pf->hw.port_info;
+ vsi = pf->vsi[0];
+ peer_dev->pf_vsi_num = vsi->vsi_num;
+ peer_dev->netdev = vsi->netdev;
+ peer_dev->initial_mtu = vsi->netdev->mtu;
+ ether_addr_copy(peer_dev->lan_addr, port_info->mac.lan_addr);
+
+ platform_dev = &peer_dev->platform_dev;
+ platform_dev->name = peer_dev_ids[i].name;
+ platform_dev->id = PLATFORM_DEVID_AUTO;
+ platform_dev->id_auto = true;
+ platform_dev->dev.release = ice_peer_device_release;
+ platform_dev->dev.parent = &pdev->dev;
+ peer_dev_int->ice_peer_wq =
+ alloc_ordered_workqueue("ice_peer_wq_%d", WQ_UNBOUND,
+ i);
+ if (!peer_dev_int->ice_peer_wq)
+ return -ENOMEM;
+
+ /* Assign a unique index and hence name for peer device */
+ status = ida_simple_get(&ice_peer_index_ida, 0, 0, GFP_KERNEL);
+ if (status < 0) {
+ dev_err(&pdev->dev,
+ "failed to get unique index for device\n");
+ devm_kfree(&pdev->dev, peer_dev);
+ return status;
+ }
+ peer_dev->index = status;
+ peer_dev->pdev = pdev;
+ peer_dev->ari_ena = pci_ari_enabled(pdev->bus);
+ peer_dev->bus_num = PCI_BUS_NUM(pdev->devfn);
+ if (!peer_dev->ari_ena) {
+ peer_dev->dev_num = PCI_SLOT(pdev->devfn);
+ peer_dev->fn_num = PCI_FUNC(pdev->devfn);
+ } else {
+ peer_dev->dev_num = 0;
+ peer_dev->fn_num = pdev->devfn & 0xff;
+ }
+
+ qos_info = &peer_dev->initial_qos_info;
+
+ /* setup qos_info fields with defaults */
+ qos_info->num_apps = 0;
+ qos_info->num_tc = 1;
+
+ for (j = 0; j < ICE_IDC_MAX_USER_PRIORITY; j++)
+ qos_info->up2tc[j] = 0;
+
+ qos_info->tc_info[0].rel_bw = 100;
+ for (j = 1; j < IEEE_8021QAZ_MAX_TCS; j++)
+ qos_info->tc_info[j].rel_bw = 0;
+
+ /* for DCB, override the qos_info defaults. */
+ ice_setup_dcb_qos_info(pf, qos_info);
+
+ /* make sure peer specific resources such as msix_count and
+ * msix_entries are initialized
+ */
+ switch (peer_dev_ids[i].id) {
+ case ICE_PEER_RDMA_DEV:
+ if (test_bit(ICE_FLAG_IWARP_ENA, pf->flags)) {
+ peer_dev->msix_count = pf->num_rdma_msix;
+ entry = &pf->msix_entries[pf->rdma_base_vector];
+ }
+ break;
+ default:
+ break;
+ }
+
+ peer_dev->msix_entries = entry;
+ ice_peer_state_change(peer_dev_int, ICE_PEER_DEV_STATE_INIT,
+ false);
+ status = platform_device_register(platform_dev);
+ if (status) {
+ pr_err("Failed to register peer device %s\n",
+ platform_dev->name);
+ return status;
+ }
+ }
+
+ return status;
+}
diff --git a/drivers/net/ethernet/intel/ice/ice_idc.h b/drivers/net/ethernet/intel/ice/ice_idc.h
new file mode 100644
index 000000000000..d29e946fe2f1
--- /dev/null
+++ b/drivers/net/ethernet/intel/ice/ice_idc.h
@@ -0,0 +1,360 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/* Copyright (c) 2019, Intel Corporation. */
+
+#ifndef _ICE_IDC_H_
+#define _ICE_IDC_H_
+
+#include <linux/kernel.h>
+#include <linux/device.h>
+#include <linux/if_ether.h>
+#include <linux/netdevice.h>
+#include <linux/dcbnl.h>
+#include <linux/platform_device.h>
+
+/* This major and minor version represent IDC API version information.
+ * During peer driver registration, peer driver specifies major and minor
+ * version information (via. peer_driver:ver_info). It gets checked against
+ * following defines and if mismatch, then peer driver registration
+ * fails and appropriate message gets logged.
+ */
+#define ICE_PEER_MAJOR_VER 6
+#define ICE_PEER_MINOR_VER 1
+
+enum ice_peer_features {
+ ICE_PEER_FEATURE_ADK_SUPPORT,
+ ICE_PEER_FEATURE_PTP_SUPPORT,
+ ICE_PEER_FEATURE_SRIOV_SUPPORT,
+ ICE_PEER_FEATURE_PCIIOV_SUPPORT,
+ ICE_PEER_FEATURE_NBITS
+};
+
+#define ICE_SRIOV_SUP BIT(ICE_PEER_FEATURE_SRIOV_SUPPORT)
+
+#ifdef CONFIG_PCI_IOV
+#define ICE_PCIIOV_SUP BIT(ICE_PEER_FEATURE_PCIIOV_SUPPORT)
+#else
+#define ICE_PCIIOV_SUP 0
+#endif /* CONFIG_PCI_IOV */
+
+#define ICE_IDC_FEATURES (ICE_SRIOV_SUP | ICE_PCIIOV_SUP)
+
+enum ice_event_type {
+ ICE_EVENT_LINK_CHANGE = 0x0,
+ ICE_EVENT_MTU_CHANGE,
+ ICE_EVENT_TC_CHANGE,
+ ICE_EVENT_API_CHANGE,
+ ICE_EVENT_MBX_CHANGE,
+ ICE_EVENT_NBITS /* must be last */
+};
+
+enum ice_res_type {
+ ICE_INVAL_RES = 0x0,
+ ICE_VSI,
+ ICE_VEB,
+ ICE_EVENT_Q,
+ ICE_EGRESS_CMPL_Q,
+ ICE_CMPL_EVENT_Q,
+ ICE_ASYNC_EVENT_Q,
+ ICE_DOORBELL_Q,
+ ICE_RDMA_QSETS_TXSCHED,
+};
+
+enum ice_peer_reset_type {
+ ICE_PEER_PFR = 0,
+ ICE_PEER_CORER,
+ ICE_PEER_CORER_SW_CORE,
+ ICE_PEER_CORER_SW_FULL,
+ ICE_PEER_GLOBR,
+};
+
+/* reason notified to peer driver as part of event handling */
+enum ice_close_reason {
+ ICE_REASON_INVAL = 0x0,
+ ICE_REASON_HW_UNRESPONSIVE,
+ ICE_REASON_INTERFACE_DOWN, /* Administrative down */
+ ICE_REASON_PEER_DRV_UNREG, /* peer driver getting unregistered */
+ ICE_REASON_PEER_DEV_UNINIT,
+ ICE_REASON_GLOBR_REQ,
+ ICE_REASON_CORER_REQ,
+ ICE_REASON_EMPR_REQ,
+ ICE_REASON_PFR_REQ,
+ ICE_REASON_HW_RESET_PENDING,
+ ICE_REASON_PARAM_CHANGE,
+};
+
+enum ice_rdma_filter {
+ ICE_RDMA_FILTER_INVAL = 0x0,
+ ICE_RDMA_FILTER_IWARP,
+ ICE_RDMA_FILTER_ROCEV2,
+ ICE_RDMA_FILTER_BOTH,
+};
+
+/* This information is needed to handle peer driver registration,
+ * instead of adding more params to peer_drv_registration function,
+ * let's get it thru' peer_drv object.
+ */
+struct ice_ver_info {
+ u16 major;
+ u16 minor;
+ u64 support;
+};
+
+/* Struct to hold per DCB APP info */
+struct ice_dcb_app_info {
+ u8 priority;
+ u8 selector;
+ u16 prot_id;
+};
+
+struct ice_peer_dev;
+struct ice_peer_dev_int;
+
+#define ICE_IDC_MAX_USER_PRIORITY 8
+#define ICE_IDC_MAX_APPS 8
+
+/* Struct to hold per RDMA Qset info */
+struct ice_rdma_qset_params {
+ u32 teid; /* qset TEID */
+ u16 qs_handle; /* RDMA driver provides this */
+ u16 vsi_id; /* VSI index */
+ u8 tc; /* TC branch the QSet should belong to */
+ u8 reserved[3];
+};
+
+struct ice_res_base {
+ /* Union for future provision e.g. other res_type */
+ union {
+ struct ice_rdma_qset_params qsets;
+ } res;
+};
+
+struct ice_res {
+ /* Type of resource. Filled by peer driver */
+ enum ice_res_type res_type;
+ /* Count requested by peer driver */
+ u16 cnt_req;
+
+ /* Number of resources allocated. Filled in by callee.
+ * Based on this value, caller to fill up "resources"
+ */
+ u16 res_allocated;
+
+ /* Unique handle to resources allocated. Zero if call fails.
+ * Allocated by callee and for now used by caller for internal
+ * tracking purpose.
+ */
+ u32 res_handle;
+
+ /* Peer driver has to allocate sufficient memory, to accommodate
+ * cnt_requested before calling this function.
+ * Memory has to be zero initialized. It is input/output param.
+ * As a result of alloc_res API, this structures will be populated.
+ */
+ struct ice_res_base res[1];
+};
+
+struct ice_qos_info {
+ u64 tc_ctx;
+ u8 rel_bw;
+ u8 prio_type;
+ u8 egress_virt_up;
+ u8 ingress_virt_up;
+};
+
+/* Struct to hold QoS info */
+struct ice_qos_params {
+ struct ice_qos_info tc_info[IEEE_8021QAZ_MAX_TCS];
+ u8 up2tc[ICE_IDC_MAX_USER_PRIORITY];
+ u8 vsi_relative_bw;
+ u8 vsi_priority_type;
+ u32 num_apps;
+ struct ice_dcb_app_info apps[ICE_IDC_MAX_APPS];
+ u8 num_tc;
+};
+
+union ice_event_info {
+ /* ICE_EVENT_LINK_CHANGE */
+ struct {
+ struct net_device *lwr_nd;
+ u16 vsi_num; /* HW index of VSI corresponding to lwr ndev */
+ u8 new_link_state;
+ u8 lport;
+ } link_info;
+ /* ICE_EVENT_MTU_CHANGE */
+ u16 mtu;
+ /* ICE_EVENT_TC_CHANGE */
+ struct ice_qos_params port_qos;
+ /* ICE_EVENT_API_CHANGE */
+ u8 api_rdy;
+ /* ICE_EVENT_MBX_CHANGE */
+ u8 mbx_rdy;
+};
+
+/* ice_event elements are to be passed back and forth between the ice driver
+ * and the peer drivers. They are to be used to both register/unregister
+ * for event reporting and to report an event (events can be either ice
+ * generated or peer generated).
+ *
+ * For (un)registering for events, the structure needs to be populated with:
+ * reporter - pointer to the ice_peer_dev struct of the peer (un)registering
+ * type - bitmap with bits set for event types to (un)register for
+ *
+ * For reporting events, the structure needs to be populated with:
+ * reporter - pointer to peer that generated the event (NULL for ice)
+ * type - bitmap with single bit set for this event type
+ * info - union containing data relevant to this event type
+ */
+struct ice_event {
+ struct ice_peer_dev *reporter;
+ DECLARE_BITMAP(type, ICE_EVENT_NBITS);
+ union ice_event_info info;
+};
+
+/* Following APIs are implemented by ICE driver and invoked by peer drivers */
+struct ice_ops {
+ /* APIs to allocate resources such as VEB, VSI, Doorbell queues,
+ * completion queues, Tx/Rx queues, etc...
+ */
+ int (*alloc_res)(struct ice_peer_dev *peer_dev,
+ struct ice_res *res,
+ int partial_acceptable);
+ int (*free_res)(struct ice_peer_dev *peer_dev,
+ struct ice_res *res);
+
+ int (*is_vsi_ready)(struct ice_peer_dev *peer_dev);
+ int (*peer_register)(struct ice_peer_dev *peer_dev);
+ int (*peer_unregister)(struct ice_peer_dev *peer_dev);
+ int (*request_reset)(struct ice_peer_dev *dev,
+ enum ice_peer_reset_type reset_type);
+
+ void (*notify_state_change)(struct ice_peer_dev *dev,
+ struct ice_event *event);
+
+ /* Notification APIs */
+ void (*reg_for_notification)(struct ice_peer_dev *dev,
+ struct ice_event *event);
+ void (*unreg_for_notification)(struct ice_peer_dev *dev,
+ struct ice_event *event);
+ int (*update_vsi_filter)(struct ice_peer_dev *peer_dev,
+ enum ice_rdma_filter filter, bool enable);
+ int (*vc_send)(struct ice_peer_dev *peer_dev, u32 vf_id, u8 *msg,
+ u16 len);
+};
+
+/* Following APIs are implemented by peer drivers and invoked by ICE driver */
+struct ice_peer_ops {
+ void (*event_handler)(struct ice_peer_dev *peer_dev,
+ struct ice_event *event);
+
+ /* Why we have 'open' and when it is expected to be called:
+ * 1. symmetric set of API w.r.t close
+ * 2. To be invoked form driver initialization path
+ * - call peer_driver:open once ice driver is fully initialized
+ * 3. To be invoked upon RESET complete
+ *
+ * Calls to open are performed from ice_finish_init_peer_device
+ * which is invoked from the service task. This helps keep devices
+ * from having their open called until the ice driver is ready and
+ * has scheduled its service task.
+ */
+ int (*open)(struct ice_peer_dev *peer_dev);
+
+ /* Peer's close function is to be called when the peer needs to be
+ * quiesced. This can be for a variety of reasons (enumerated in the
+ * ice_close_reason enum struct). A call to close will only be
+ * followed by a call to either remove or open. No IDC calls from the
+ * peer should be accepted until it is re-opened.
+ *
+ * The *reason* parameter is the reason for the call to close. This
+ * can be for any reason enumerated in the ice_close_reason struct.
+ * It's primary reason is for the peer's bookkeeping and in case the
+ * peer want to perform any different tasks dictated by the reason.
+ */
+ void (*close)(struct ice_peer_dev *peer_dev,
+ enum ice_close_reason reason);
+
+ int (*vc_receive)(struct ice_peer_dev *peer_dev, u32 vf_id, u8 *msg,
+ u16 len);
+ /* tell RDMA peer to prepare for TC change in a blocking call
+ * that will directly precede the change event
+ */
+ void (*prep_tc_change)(struct ice_peer_dev *peer_dev);
+};
+
+#define ICE_PEER_RDMA_NAME "ice_rdma"
+#define ICE_PEER_RDMA_DEV 0x00000010
+#define ICE_MAX_NUM_PEERS 4
+
+/* The const struct that instantiates peer_dev_id needs to be initialized
+ * in the .c with the macro ASSIGN_PEER_INFO.
+ * For example:
+ * static const struct peer_dev_id peer_dev_ids[] = ASSIGN_PEER_INFO;
+ */
+struct peer_dev_id {
+ char *name;
+ int id;
+};
+
+#define ASSIGN_PEER_INFO \
+{ \
+ { ICE_PEER_RDMA_NAME, ICE_PEER_RDMA_DEV }, \
+}
+
+/* structure representing peer device */
+struct ice_peer_dev {
+ struct platform_device platform_dev;
+ struct ice_ver_info ver;
+ struct pci_dev *pdev; /* PCI device of corresponding to main function */
+ /* KVA / Linear address corresponding to BAR0 of underlying
+ * pci_device.
+ */
+ u8 __iomem *hw_addr;
+ int peer_dev_id;
+
+ unsigned int index;
+
+ u8 ftype; /* PF(false) or VF (true) */
+
+ /* Data VSI created by driver */
+ u16 pf_vsi_num;
+
+ u8 lan_addr[ETH_ALEN]; /* default MAC address of main netdev */
+ u16 initial_mtu; /* Initial MTU of main netdev */
+ struct ice_qos_params initial_qos_info;
+ struct net_device *netdev;
+ /* PCI info */
+ u8 ari_ena;
+ u16 bus_num;
+ u16 dev_num;
+ u16 fn_num;
+
+ /* Based on peer driver type, this shall point to corresponding MSIx
+ * entries in pf->msix_entries (which were allocated as part of driver
+ * initialization) e.g. for RDMA driver, msix_entries reserved will be
+ * num_online_cpus + 1.
+ */
+ u16 msix_count; /* How many vectors are reserved for this device */
+ struct msix_entry *msix_entries;
+
+ /* Following struct contains function pointers to be initialized
+ * by ICE driver and called by peer driver
+ */
+ const struct ice_ops *ops;
+
+ /* Following struct contains function pointers to be initialized
+ * by peer driver and called by ICE driver
+ */
+ const struct ice_peer_ops *peer_ops;
+
+ /* Pointer to peer_drv struct to be populated by peer driver */
+ struct ice_peer_drv *peer_drv;
+};
+
+static inline struct ice_peer_dev *dev_to_ice_peer(struct device *_dev)
+{
+ struct platform_device *platform_dev;
+
+ platform_dev = container_of(_dev, struct platform_device, dev);
+ return container_of(platform_dev, struct ice_peer_dev, platform_dev);
+}
+#endif /* _ICE_IDC_H_*/
diff --git a/drivers/net/ethernet/intel/ice/ice_idc_int.h b/drivers/net/ethernet/intel/ice/ice_idc_int.h
new file mode 100644
index 000000000000..f8ff8cac7a26
--- /dev/null
+++ b/drivers/net/ethernet/intel/ice/ice_idc_int.h
@@ -0,0 +1,67 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/* Copyright (c) 2019, Intel Corporation. */
+
+#ifndef _ICE_IDC_INT_H_
+#define _ICE_IDC_INT_H_
+
+#include "ice_idc.h"
+
+int ice_unroll_peer(struct ice_peer_dev_int *peer_dev_int, void *data);
+int ice_unreg_peer_device(struct ice_peer_dev_int *peer_dev_int, void *data);
+
+enum ice_peer_dev_state {
+ ICE_PEER_DEV_STATE_INIT,
+ ICE_PEER_DEV_STATE_PROBED,
+ ICE_PEER_DEV_STATE_OPENING,
+ ICE_PEER_DEV_STATE_OPENED,
+ ICE_PEER_DEV_STATE_PREP_RST,
+ ICE_PEER_DEV_STATE_PREPPED,
+ ICE_PEER_DEV_STATE_CLOSING,
+ ICE_PEER_DEV_STATE_CLOSED,
+ ICE_PEER_DEV_STATE_REMOVED,
+ ICE_PEER_DEV_STATE_API_RDY,
+ ICE_PEER_DEV_STATE_NBITS, /* must be last */
+};
+
+enum ice_peer_drv_state {
+ ICE_PEER_DRV_STATE_MBX_RDY,
+ ICE_PEER_DRV_STATE_NBITS, /* must be last */
+};
+
+struct ice_peer_drv_int {
+ struct ice_peer_drv *peer_drv;
+ struct ice_peer_dev *peer_dev;
+
+ /* States associated with peer driver */
+ DECLARE_BITMAP(state, ICE_PEER_DRV_STATE_NBITS);
+
+ /* if this peer_dev is the originator of an event, these are the
+ * most recent events of each type
+ */
+ struct ice_event current_events[ICE_EVENT_NBITS];
+};
+
+struct ice_peer_dev_int {
+ struct ice_peer_dev peer_dev; /* public structure */
+ struct ice_peer_drv_int *peer_drv_int; /* driver private structure */
+
+ /* if this peer_dev is the originator of an event, these are the
+ * most recent events of each type
+ */
+ struct ice_event current_events[ICE_EVENT_NBITS];
+ /* Events a peer has registered to be notified about */
+ DECLARE_BITMAP(events, ICE_EVENT_NBITS);
+
+ /* States associated with peer device */
+ DECLARE_BITMAP(state, ICE_PEER_DEV_STATE_NBITS);
+ struct mutex peer_dev_state_mutex; /* peer_dev state mutex */
+
+ /* per peer workqueue */
+ struct workqueue_struct *ice_peer_wq;
+
+ struct work_struct peer_prep_task;
+ struct work_struct peer_close_task;
+
+ enum ice_close_reason rst_type;
+};
+#endif /* !_ICE_IDC_INT_H_ */
diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c
index 28ec0d57941d..08bce0d67634 100644
--- a/drivers/net/ethernet/intel/ice/ice_main.c
+++ b/drivers/net/ethernet/intel/ice/ice_main.c
@@ -2092,6 +2092,8 @@ static void ice_init_pf(struct ice_pf *pf)
{
bitmap_zero(pf->flags, ICE_PF_FLAGS_NBITS);
set_bit(ICE_FLAG_MSIX_ENA, pf->flags);
+ if (pf->hw.func_caps.common_cap.iwarp)
+ set_bit(ICE_FLAG_IWARP_ENA, pf->flags);
#ifdef CONFIG_PCI_IOV
if (pf->hw.func_caps.common_cap.sr_iov_1_1) {
struct ice_hw *hw = &pf->hw;
@@ -2145,6 +2147,17 @@ static int ice_ena_msix_range(struct ice_pf *pf)
v_budget += pf->num_lan_msix;
v_left -= pf->num_lan_msix;
+ /* reserve vectors for RDMA peer driver */
+ if (test_bit(ICE_FLAG_IWARP_ENA, pf->flags)) {
+ /* RDMA peer driver needs one extra to handle misc causes */
+ needed = min_t(int, num_online_cpus(), v_left) + 1;
+ if (v_left < needed)
+ goto no_vecs_left_err;
+ pf->num_rdma_msix = needed;
+ v_budget += needed;
+ v_left -= needed;
+ }
+
pf->msix_entries = devm_kcalloc(&pf->pdev->dev, v_budget,
sizeof(*pf->msix_entries), GFP_KERNEL);
@@ -2171,6 +2184,8 @@ static int ice_ena_msix_range(struct ice_pf *pf)
"not enough vectors. requested = %d, obtained = %d\n",
v_budget, v_actual);
if (v_actual >= (pf->num_lan_msix + 1)) {
+ clear_bit(ICE_FLAG_IWARP_ENA, pf->flags);
+ pf->num_rdma_msix = 0;
pf->num_avail_sw_msix = v_actual -
(pf->num_lan_msix + 1);
} else if (v_actual >= 2) {
@@ -2189,6 +2204,11 @@ static int ice_ena_msix_range(struct ice_pf *pf)
devm_kfree(&pf->pdev->dev, pf->msix_entries);
goto exit_err;
+no_vecs_left_err:
+ dev_err(&pf->pdev->dev,
+ "not enough vectors. requested = %d, available = %d\n",
+ needed, v_left);
+ err = -ERANGE;
exit_err:
pf->num_lan_msix = 0;
clear_bit(ICE_FLAG_MSIX_ENA, pf->flags);
@@ -2351,6 +2371,12 @@ ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent)
ice_init_pf(pf);
+ pf->peers = devm_kcalloc(dev, ICE_MAX_NUM_PEERS, sizeof(*pf->peers),
+ GFP_KERNEL);
+ if (!pf->peers) {
+ err = -ENOMEM;
+ goto err_init_peer_unroll;
+ }
err = ice_init_pf_dcb(pf, false);
if (err) {
clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
@@ -2432,10 +2458,20 @@ ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent)
goto err_alloc_sw_unroll;
}
+ err = ice_init_peer_devices(pf);
+ if (err) {
+ dev_err(dev, "Failed to initialize peer devices: 0x%x\n", err);
+ err = -EIO;
+ goto err_init_peer_unroll;
+ }
+
ice_verify_cacheline_size(pf);
return 0;
+ /* Unwind non-managed device resources, etc. if something failed */
+err_init_peer_unroll:
+ ice_for_each_peer(pf, NULL, ice_unroll_peer);
err_alloc_sw_unroll:
set_bit(__ICE_SERVICE_DIS, pf->state);
set_bit(__ICE_DOWN, pf->state);
@@ -2477,6 +2513,7 @@ static void ice_remove(struct pci_dev *pdev)
if (test_bit(ICE_FLAG_SRIOV_ENA, pf->flags))
ice_free_vfs(pf);
ice_vsi_release_all(pf);
+ ice_for_each_peer(pf, NULL, ice_unreg_peer_device);
ice_free_irq_msix_misc(pf);
ice_for_each_vsi(pf, i) {
if (!pf->vsi[i])
@@ -2676,6 +2713,7 @@ static int __init ice_module_init(void)
if (status) {
pr_err("failed to register PCI driver, err %d\n", status);
destroy_workqueue(ice_wq);
+ ida_destroy(&ice_peer_index_ida);
}
return status;
@@ -2692,6 +2730,10 @@ static void __exit ice_module_exit(void)
{
pci_unregister_driver(&ice_driver);
destroy_workqueue(ice_wq);
+ /* release all cached layer within ida tree, associated with
+ * ice_peer_index_ida object
+ */
+ ida_destroy(&ice_peer_index_ida);
pr_info("module unloaded\n");
}
module_exit(ice_module_exit);
diff --git a/drivers/net/ethernet/intel/ice/ice_type.h b/drivers/net/ethernet/intel/ice/ice_type.h
index 24bbef8bbe69..99822b4e8621 100644
--- a/drivers/net/ethernet/intel/ice/ice_type.h
+++ b/drivers/net/ethernet/intel/ice/ice_type.h
@@ -161,6 +161,7 @@ struct ice_hw_common_caps {
u8 rss_table_entry_width; /* RSS Entry width in bits */
u8 dcb;
+ u8 iwarp;
};
/* Function specific capabilities */
--
2.21.0
^ permalink raw reply related [flat|nested] 15+ messages in thread
* [net-next 2/3] ice: Implement peer communications
2019-07-04 2:12 [net-next 0/3][pull request] Intel Wired LAN ver Updates 2019-07-03 Jeff Kirsher
2019-07-04 2:12 ` [net-next 1/3] ice: Initialize and register platform device to provide RDMA Jeff Kirsher
@ 2019-07-04 2:12 ` Jeff Kirsher
2019-07-04 2:12 ` [net-next 3/3] i40e: Register RDMA client devices to the virtual platform bus Jeff Kirsher
2019-07-04 12:15 ` [net-next 0/3][pull request] Intel Wired LAN ver Updates 2019-07-03 Jason Gunthorpe
3 siblings, 0 replies; 15+ messages in thread
From: Jeff Kirsher @ 2019-07-04 2:12 UTC (permalink / raw)
To: davem, dledford, jgg
Cc: Tony Nguyen, netdev, linux-rdma, nhorman, sassmann,
mustafa.ismail, shiraz.saleem, Dave Ertman, Andrew Bowers,
Jeff Kirsher
From: Tony Nguyen <anthony.l.nguyen@intel.com>
Set and implement operations for the peer device and peer driver to
communicate with each other, via ice_ops and ice_peer_ops, to request
resources and manage event notification.
Signed-off-by: Dave Ertman <david.m.ertman@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
drivers/net/ethernet/intel/ice/ice.h | 2 +
.../net/ethernet/intel/ice/ice_adminq_cmd.h | 32 +
drivers/net/ethernet/intel/ice/ice_common.c | 189 ++++
drivers/net/ethernet/intel/ice/ice_common.h | 9 +
drivers/net/ethernet/intel/ice/ice_dcb_lib.c | 33 +
drivers/net/ethernet/intel/ice/ice_idc.c | 927 ++++++++++++++++++
drivers/net/ethernet/intel/ice/ice_idc.h | 17 +
drivers/net/ethernet/intel/ice/ice_idc_int.h | 15 +
drivers/net/ethernet/intel/ice/ice_lib.c | 35 +-
drivers/net/ethernet/intel/ice/ice_lib.h | 2 +
drivers/net/ethernet/intel/ice/ice_main.c | 52 +-
drivers/net/ethernet/intel/ice/ice_sched.c | 69 +-
drivers/net/ethernet/intel/ice/ice_switch.c | 27 +
drivers/net/ethernet/intel/ice/ice_switch.h | 4 +
drivers/net/ethernet/intel/ice/ice_type.h | 3 +
.../net/ethernet/intel/ice/ice_virtchnl_pf.c | 25 -
16 files changed, 1409 insertions(+), 32 deletions(-)
diff --git a/drivers/net/ethernet/intel/ice/ice.h b/drivers/net/ethernet/intel/ice/ice.h
index 6ee2774699f0..4855d3425043 100644
--- a/drivers/net/ethernet/intel/ice/ice.h
+++ b/drivers/net/ethernet/intel/ice/ice.h
@@ -302,6 +302,7 @@ struct ice_vsi {
u16 num_rxq; /* Used Rx queues */
u16 num_rx_desc;
u16 num_tx_desc;
+ u16 qset_handle[ICE_MAX_TRAFFIC_CLASS];
struct ice_tc_cfg tc_cfg;
} ____cacheline_internodealigned_in_smp;
@@ -471,6 +472,7 @@ struct ice_vsi *ice_lb_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi);
int ice_set_rss(struct ice_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size);
int ice_get_rss(struct ice_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size);
void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size);
+int ice_schedule_reset(struct ice_pf *pf, enum ice_reset_req reset);
void ice_print_link_msg(struct ice_vsi *vsi, bool isup);
int ice_init_peer_devices(struct ice_pf *pf);
int
diff --git a/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h b/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h
index 1e1bd5f0c2a2..8e4f20b696ae 100644
--- a/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h
+++ b/drivers/net/ethernet/intel/ice/ice_adminq_cmd.h
@@ -1441,6 +1441,36 @@ struct ice_aqc_dis_txq {
struct ice_aqc_dis_txq_item qgrps[1];
};
+/* Add Tx RDMA Queue Set (indirect 0x0C33) */
+struct ice_aqc_add_rdma_qset {
+ u8 num_qset_grps;
+ u8 reserved[7];
+ __le32 addr_high;
+ __le32 addr_low;
+};
+
+/* This is the descriptor of each qset entry for the Add Tx RDMA Queue Set
+ * command (0x0C33). Only used within struct ice_aqc_add_rdma_qset.
+ */
+struct ice_aqc_add_tx_rdma_qset_entry {
+ __le16 tx_qset_id;
+ u8 rsvd[2];
+ __le32 qset_teid;
+ struct ice_aqc_txsched_elem info;
+};
+
+/* The format of the command buffer for Add Tx RDMA Queue Set(0x0C33)
+ * is an array of the following structs. Please note that the length of
+ * each struct ice_aqc_add_rdma_qset is variable due to the variable
+ * number of queues in each group!
+ */
+struct ice_aqc_add_rdma_qset_data {
+ __le32 parent_teid;
+ __le16 num_qsets;
+ u8 rsvd[2];
+ struct ice_aqc_add_tx_rdma_qset_entry rdma_qsets[1];
+};
+
/* Configure Firmware Logging Command (indirect 0xFF09)
* Logging Information Read Response (indirect 0xFF10)
* Note: The 0xFF10 command has no input parameters.
@@ -1576,6 +1606,7 @@ struct ice_aq_desc {
struct ice_aqc_get_set_rss_key get_set_rss_key;
struct ice_aqc_add_txqs add_txqs;
struct ice_aqc_dis_txqs dis_txqs;
+ struct ice_aqc_add_rdma_qset add_rdma_qset;
struct ice_aqc_add_get_update_free_vsi vsi_cmd;
struct ice_aqc_add_update_free_vsi_resp add_update_free_vsi_res;
struct ice_aqc_fw_logging fw_logging;
@@ -1697,6 +1728,7 @@ enum ice_adminq_opc {
/* Tx queue handling commands/events */
ice_aqc_opc_add_txqs = 0x0C30,
ice_aqc_opc_dis_txqs = 0x0C31,
+ ice_aqc_opc_add_rdma_qset = 0x0C33,
/* debug commands */
ice_aqc_opc_fw_logging = 0xFF09,
diff --git a/drivers/net/ethernet/intel/ice/ice_common.c b/drivers/net/ethernet/intel/ice/ice_common.c
index 2788b2ede440..6cc8cfd97f88 100644
--- a/drivers/net/ethernet/intel/ice/ice_common.c
+++ b/drivers/net/ethernet/intel/ice/ice_common.c
@@ -2711,6 +2711,59 @@ ice_aq_dis_lan_txq(struct ice_hw *hw, u8 num_qgrps,
return status;
}
+/**
+ * ice_aq_add_rdma_qsets
+ * @hw: pointer to the hardware structure
+ * @num_qset_grps: Number of RDMA Qset groups
+ * @qset_list: list of qset groups to be added
+ * @buf_size: size of buffer for indirect command
+ * @cd: pointer to command details structure or NULL
+ *
+ * Add Tx RDMA Qsets (0x0C33)
+ */
+static enum ice_status
+ice_aq_add_rdma_qsets(struct ice_hw *hw, u8 num_qset_grps,
+ struct ice_aqc_add_rdma_qset_data *qset_list,
+ u16 buf_size, struct ice_sq_cd *cd)
+{
+ struct ice_aqc_add_rdma_qset_data *list;
+ u16 i, sum_header_size, sum_q_size = 0;
+ struct ice_aqc_add_rdma_qset *cmd;
+ struct ice_aq_desc desc;
+
+ cmd = &desc.params.add_rdma_qset;
+
+ ice_fill_dflt_direct_cmd_desc(&desc, ice_aqc_opc_add_rdma_qset);
+
+ if (!qset_list)
+ return ICE_ERR_PARAM;
+
+ if (num_qset_grps > ICE_LAN_TXQ_MAX_QGRPS)
+ return ICE_ERR_PARAM;
+
+ sum_header_size = num_qset_grps *
+ (sizeof(*qset_list) - sizeof(*qset_list->rdma_qsets));
+
+ list = qset_list;
+ for (i = 0; i < num_qset_grps; i++) {
+ struct ice_aqc_add_tx_rdma_qset_entry *qset = list->rdma_qsets;
+ u16 num_qsets = le16_to_cpu(list->num_qsets);
+
+ sum_q_size += num_qsets * sizeof(*qset);
+ list = (struct ice_aqc_add_rdma_qset_data *)
+ (qset + num_qsets);
+ }
+
+ if (buf_size != (sum_header_size + sum_q_size))
+ return ICE_ERR_PARAM;
+
+ desc.flags |= cpu_to_le16(ICE_AQ_FLAG_RD);
+
+ cmd->num_qset_grps = num_qset_grps;
+
+ return ice_aq_send_cmd(hw, &desc, qset_list, buf_size, cd);
+}
+
/* End of FW Admin Queue command wrappers */
/**
@@ -3179,6 +3232,142 @@ ice_cfg_vsi_lan(struct ice_port_info *pi, u16 vsi_handle, u8 tc_bitmap,
ICE_SCHED_NODE_OWNER_LAN);
}
+/**
+ * ice_cfg_vsi_rdma - configure the VSI RDMA queues
+ * @pi: port information structure
+ * @vsi_handle: software VSI handle
+ * @tc_bitmap: TC bitmap
+ * @max_rdmaqs: max RDMA queues array per TC
+ *
+ * This function adds/updates the VSI RDMA queues per TC.
+ */
+enum ice_status
+ice_cfg_vsi_rdma(struct ice_port_info *pi, u16 vsi_handle, u8 tc_bitmap,
+ u16 *max_rdmaqs)
+{
+ return ice_cfg_vsi_qs(pi, vsi_handle, tc_bitmap, max_rdmaqs,
+ ICE_SCHED_NODE_OWNER_RDMA);
+}
+
+/**
+ * ice_ena_vsi_rdma_qset
+ * @pi: port information structure
+ * @vsi_handle: software VSI handle
+ * @tc: TC number
+ * @rdma_qset: pointer to RDMA qset
+ * @num_qsets: number of RDMA qsets
+ * @qset_teid: pointer to qset node teids
+ *
+ * This function adds RDMA qset
+ */
+enum ice_status
+ice_ena_vsi_rdma_qset(struct ice_port_info *pi, u16 vsi_handle, u8 tc,
+ u16 *rdma_qset, u16 num_qsets, u32 *qset_teid)
+{
+ struct ice_aqc_txsched_elem_data node = { 0 };
+ struct ice_aqc_add_rdma_qset_data *buf;
+ struct ice_sched_node *parent;
+ enum ice_status status;
+ struct ice_hw *hw;
+ u16 buf_size;
+ u8 i;
+
+ if (!pi || pi->port_state != ICE_SCHED_PORT_STATE_READY)
+ return ICE_ERR_CFG;
+ hw = pi->hw;
+
+ if (!ice_is_vsi_valid(hw, vsi_handle))
+ return ICE_ERR_PARAM;
+
+ buf_size = sizeof(*buf) + sizeof(*buf->rdma_qsets) * (num_qsets - 1);
+ buf = devm_kzalloc(ice_hw_to_dev(hw), buf_size, GFP_KERNEL);
+ if (!buf)
+ return ICE_ERR_NO_MEMORY;
+ mutex_lock(&pi->sched_lock);
+
+ parent = ice_sched_get_free_qparent(pi, vsi_handle, tc,
+ ICE_SCHED_NODE_OWNER_RDMA);
+ if (!parent) {
+ status = ICE_ERR_PARAM;
+ goto rdma_error_exit;
+ }
+ buf->parent_teid = parent->info.node_teid;
+ node.parent_teid = parent->info.node_teid;
+
+ buf->num_qsets = cpu_to_le16(num_qsets);
+ for (i = 0; i < num_qsets; i++) {
+ buf->rdma_qsets[i].tx_qset_id = cpu_to_le16(rdma_qset[i]);
+ buf->rdma_qsets[i].info.valid_sections =
+ ICE_AQC_ELEM_VALID_GENERIC;
+ }
+ status = ice_aq_add_rdma_qsets(hw, 1, buf, buf_size, NULL);
+ if (status) {
+ ice_debug(hw, ICE_DBG_RDMA, "add RDMA qset failed\n");
+ goto rdma_error_exit;
+ }
+ node.data.elem_type = ICE_AQC_ELEM_TYPE_LEAF;
+ for (i = 0; i < num_qsets; i++) {
+ node.node_teid = buf->rdma_qsets[i].qset_teid;
+ status = ice_sched_add_node(pi, hw->num_tx_sched_layers - 1,
+ &node);
+ if (status)
+ break;
+ qset_teid[i] = le32_to_cpu(node.node_teid);
+ }
+rdma_error_exit:
+ mutex_unlock(&pi->sched_lock);
+ devm_kfree(ice_hw_to_dev(hw), buf);
+ return status;
+}
+
+/**
+ * ice_dis_vsi_rdma_qset - free RDMA resources
+ * @pi: port_info struct
+ * @count: number of RDMA qsets to free
+ * @qset_teid: TEID of qset node
+ * @q_id: list of queue IDs being disabled
+ */
+enum ice_status
+ice_dis_vsi_rdma_qset(struct ice_port_info *pi, u16 count, u32 *qset_teid,
+ u16 *q_id)
+{
+ struct ice_aqc_dis_txq_item qg_list;
+ enum ice_status status = 0;
+ u16 qg_size;
+ int i;
+
+ if (!pi || pi->port_state != ICE_SCHED_PORT_STATE_READY)
+ return ICE_ERR_CFG;
+
+ qg_size = sizeof(qg_list);
+
+ mutex_lock(&pi->sched_lock);
+
+ for (i = 0; i < count; i++) {
+ struct ice_sched_node *node;
+
+ node = ice_sched_find_node_by_teid(pi->root, qset_teid[i]);
+ if (!node)
+ continue;
+
+ qg_list.parent_teid = node->info.parent_teid;
+ qg_list.num_qs = 1;
+ qg_list.q_id[0] =
+ cpu_to_le16(q_id[i] |
+ ICE_AQC_Q_DIS_BUF_ELEM_TYPE_RDMA_QSET);
+
+ status = ice_aq_dis_lan_txq(pi->hw, 1, &qg_list, qg_size,
+ ICE_NO_RESET, 0, NULL);
+ if (status)
+ break;
+
+ ice_free_sched_node(pi, node);
+ }
+
+ mutex_unlock(&pi->sched_lock);
+ return status;
+}
+
/**
* ice_replay_pre_init - replay pre initialization
* @hw: pointer to the HW struct
diff --git a/drivers/net/ethernet/intel/ice/ice_common.h b/drivers/net/ethernet/intel/ice/ice_common.h
index d1f8353fe6bb..f56c05119378 100644
--- a/drivers/net/ethernet/intel/ice/ice_common.h
+++ b/drivers/net/ethernet/intel/ice/ice_common.h
@@ -108,6 +108,15 @@ ice_aq_set_port_id_led(struct ice_port_info *pi, bool is_orig_mode,
struct ice_sq_cd *cd);
enum ice_status
+ice_cfg_vsi_rdma(struct ice_port_info *pi, u16 vsi_handle, u8 tc_bitmap,
+ u16 *max_rdmaqs);
+enum ice_status
+ice_ena_vsi_rdma_qset(struct ice_port_info *pi, u16 vsi_handle, u8 tc,
+ u16 *rdma_qset, u16 num_qsets, u32 *qset_teid);
+enum ice_status
+ice_dis_vsi_rdma_qset(struct ice_port_info *pi, u16 count, u32 *qset_teid,
+ u16 *q_id);
+enum ice_status
ice_dis_vsi_txq(struct ice_port_info *pi, u16 vsi_handle, u8 tc, u8 num_queues,
u16 *q_handle, u16 *q_ids, u32 *q_teids,
enum ice_disq_rst_src rst_src, u16 vmvf_num,
diff --git a/drivers/net/ethernet/intel/ice/ice_dcb_lib.c b/drivers/net/ethernet/intel/ice/ice_dcb_lib.c
index c6466e16eef1..ded039a1e3e7 100644
--- a/drivers/net/ethernet/intel/ice/ice_dcb_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_dcb_lib.c
@@ -106,6 +106,7 @@ void ice_vsi_cfg_dcb_rings(struct ice_vsi *vsi)
static void ice_pf_dcb_recfg(struct ice_pf *pf)
{
struct ice_dcbx_cfg *dcbcfg = &pf->hw.port_info->local_dcbx_cfg;
+ struct ice_event *event;
u8 tc_map = 0;
int v, ret;
@@ -129,6 +130,35 @@ static void ice_pf_dcb_recfg(struct ice_pf *pf)
ice_vsi_map_rings_to_vectors(pf->vsi[v]);
}
+ event = devm_kzalloc(&pf->pdev->dev, sizeof(*event), GFP_KERNEL);
+ if (!event)
+ return;
+
+ set_bit(ICE_EVENT_TC_CHANGE, event->type);
+ event->reporter = NULL;
+ ice_setup_dcb_qos_info(pf, &event->info.port_qos);
+ ice_for_each_peer(pf, event, ice_peer_check_for_reg);
+ devm_kfree(&pf->pdev->dev, event);
+}
+
+/**
+ * ice_peer_prep_tc_change - Pre-notify RDMA Peer in blocking call of TC change
+ * @peer_dev_int: ptr to peer device internal struct
+ * @data: ptr to opaque data
+ */
+static int
+ice_peer_prep_tc_change(struct ice_peer_dev_int *peer_dev_int,
+ void __always_unused *data)
+{
+ struct ice_peer_dev *peer_dev = &peer_dev_int->peer_dev;
+
+ if (!ice_validate_peer_dev(peer_dev))
+ return 0;
+
+ if (peer_dev->peer_ops && peer_dev->peer_ops->prep_tc_change)
+ peer_dev->peer_ops->prep_tc_change(peer_dev);
+
+ return 0;
}
/**
@@ -160,6 +190,9 @@ int ice_pf_dcb_cfg(struct ice_pf *pf, struct ice_dcbx_cfg *new_cfg, bool locked)
return ret;
}
+ /* Notify capable peers about impending change to TCs */
+ ice_for_each_peer(pf, NULL, ice_peer_prep_tc_change);
+
/* Store old config in case FW config fails */
old_cfg = devm_kzalloc(&pf->pdev->dev, sizeof(*old_cfg), GFP_KERNEL);
memcpy(old_cfg, curr_cfg, sizeof(*old_cfg));
diff --git a/drivers/net/ethernet/intel/ice/ice_idc.c b/drivers/net/ethernet/intel/ice/ice_idc.c
index 0bf66e9fa159..550c3839b581 100644
--- a/drivers/net/ethernet/intel/ice/ice_idc.c
+++ b/drivers/net/ethernet/intel/ice/ice_idc.c
@@ -10,6 +10,44 @@ DEFINE_IDA(ice_peer_index_ida);
static const struct peer_dev_id peer_dev_ids[] = ASSIGN_PEER_INFO;
+/**
+ * ice_validate_peer_dev - validate peer device state
+ * @peer_dev: ptr to peer device
+ *
+ * This helper function checks if PF is in a minimal state and if the peer
+ * device is valid. This should be called before engaging in peer operations.
+ *
+ * Returns true if the peer device is valid, false otherwise.
+ */
+#ifndef CONFIG_DCB
+static
+#endif
+bool ice_validate_peer_dev(struct ice_peer_dev *peer_dev)
+{
+ struct ice_peer_dev_int *peer_dev_int;
+ struct ice_pf *pf;
+
+ if (!peer_dev || !peer_dev->pdev)
+ return false;
+
+ if (!peer_dev->peer_ops)
+ return false;
+
+ pf = pci_get_drvdata(peer_dev->pdev);
+ if (!pf)
+ return false;
+
+ peer_dev_int = peer_to_ice_dev_int(peer_dev);
+ if (!peer_dev_int)
+ return false;
+
+ if (test_bit(ICE_PEER_DEV_STATE_REMOVED, peer_dev_int->state) ||
+ test_bit(ICE_PEER_DEV_STATE_INIT, peer_dev_int->state))
+ return false;
+
+ return true;
+}
+
/**
* ice_peer_state_change - manage state machine for peer
* @peer_dev: pointer to peer's configuration
@@ -155,6 +193,155 @@ ice_peer_state_change(struct ice_peer_dev_int *peer_dev, long new_state,
mutex_unlock(&peer_dev->peer_dev_state_mutex);
}
+/**
+ * ice_peer_close - close a peer device
+ * @peer_dev_int: device to close
+ * @data: pointer to opaque data
+ *
+ * This function will also set the state bit for the peer to CLOSED. This
+ * function is meant to be called from a ice_for_each_peer().
+ */
+int ice_peer_close(struct ice_peer_dev_int *peer_dev_int, void *data)
+{
+ enum ice_close_reason reason = *(enum ice_close_reason *)(data);
+ struct ice_peer_dev *peer_dev = &peer_dev_int->peer_dev;
+ struct ice_pf *pf;
+ int i;
+
+ /* return 0 so ice_for_each_peer will continue closing other peers */
+ if (!ice_validate_peer_dev(peer_dev))
+ return 0;
+ pf = pci_get_drvdata(peer_dev->pdev);
+
+ if (test_bit(__ICE_DOWN, pf->state) ||
+ test_bit(__ICE_SUSPENDED, pf->state) ||
+ test_bit(__ICE_NEEDS_RESTART, pf->state))
+ return 0;
+
+ mutex_lock(&peer_dev_int->peer_dev_state_mutex);
+
+ /* no peer driver, already closed, closing or opening nothing to do */
+ if (test_bit(ICE_PEER_DEV_STATE_CLOSED, peer_dev_int->state) ||
+ test_bit(ICE_PEER_DEV_STATE_CLOSING, peer_dev_int->state) ||
+ test_bit(ICE_PEER_DEV_STATE_OPENING, peer_dev_int->state) ||
+ test_bit(ICE_PEER_DEV_STATE_REMOVED, peer_dev_int->state))
+ goto peer_close_out;
+
+ /* Set the peer state to CLOSING */
+ ice_peer_state_change(peer_dev_int, ICE_PEER_DEV_STATE_CLOSING, true);
+
+ for (i = 0; i < ICE_EVENT_NBITS; i++)
+ bitmap_zero(peer_dev_int->current_events[i].type,
+ ICE_EVENT_NBITS);
+
+ if (peer_dev->peer_ops && peer_dev->peer_ops->close)
+ peer_dev->peer_ops->close(peer_dev, reason);
+
+ /* Set the peer state to CLOSED */
+ ice_peer_state_change(peer_dev_int, ICE_PEER_DEV_STATE_CLOSED, true);
+
+peer_close_out:
+ mutex_unlock(&peer_dev_int->peer_dev_state_mutex);
+
+ return 0;
+}
+
+/**
+ * ice_close_peer_for_reset - queue work to close peer for reset
+ * @peer_dev_int: pointer peer dev internal struct
+ * @data: pointer to opaque data used for reset type
+ */
+int ice_close_peer_for_reset(struct ice_peer_dev_int *peer_dev_int, void *data)
+{
+ struct ice_peer_dev *peer_dev = &peer_dev_int->peer_dev;
+ enum ice_reset_req reset;
+
+ if (!ice_validate_peer_dev(peer_dev))
+ return 0;
+
+ reset = *(enum ice_reset_req *)data;
+
+ switch (reset) {
+ case ICE_RESET_GLOBR:
+ peer_dev_int->rst_type = ICE_REASON_GLOBR_REQ;
+ break;
+ case ICE_RESET_CORER:
+ peer_dev_int->rst_type = ICE_REASON_CORER_REQ;
+ break;
+ case ICE_RESET_PFR:
+ peer_dev_int->rst_type = ICE_REASON_PFR_REQ;
+ break;
+ default:
+ /* reset type is invalid */
+ return 1;
+ }
+ queue_work(peer_dev_int->ice_peer_wq, &peer_dev_int->peer_close_task);
+ return 0;
+}
+
+/**
+ * ice_check_peer_drv_for_events - check peer_drv for events to report
+ * @peer_dev: peer device to report to
+ */
+static void ice_check_peer_drv_for_events(struct ice_peer_dev *peer_dev)
+{
+ const struct ice_peer_ops *p_ops = peer_dev->peer_ops;
+ struct ice_peer_dev_int *peer_dev_int;
+ struct ice_peer_drv_int *peer_drv_int;
+ int i;
+
+ peer_dev_int = peer_to_ice_dev_int(peer_dev);
+ peer_drv_int = peer_dev_int->peer_drv_int;
+
+ for_each_set_bit(i, peer_dev_int->events, ICE_EVENT_NBITS) {
+ struct ice_event *curr = &peer_drv_int->current_events[i];
+
+ if (!bitmap_empty(curr->type, ICE_EVENT_NBITS) &&
+ p_ops->event_handler)
+ p_ops->event_handler(peer_dev, curr);
+ }
+}
+
+/**
+ * ice_check_peer_for_events - check peer_devs for events new peer reg'd for
+ * @src_peer_int: peer to check for events
+ * @data: ptr to opaque data, to be used for the peer struct that opened
+ *
+ * This function is to be called when a peer device is opened.
+ *
+ * Since a new peer opening would have missed any events that would
+ * have happened before its opening, we need to walk the peers and see
+ * if any of them have events that the new peer cares about
+ *
+ * This function is meant to be called by a device_for_each_child.
+ */
+static int
+ice_check_peer_for_events(struct ice_peer_dev_int *src_peer_int, void *data)
+{
+ struct ice_peer_dev *new_peer = (struct ice_peer_dev *)data;
+ struct ice_peer_dev *src_peer = &src_peer_int->peer_dev;
+ const struct ice_peer_ops *p_ops = new_peer->peer_ops;
+ struct ice_peer_dev_int *new_peer_int;
+ int i;
+
+ if (!ice_validate_peer_dev(new_peer) ||
+ !ice_validate_peer_dev(src_peer))
+ return 0;
+
+ new_peer_int = peer_to_ice_dev_int(new_peer);
+
+ for_each_set_bit(i, new_peer_int->events, ICE_EVENT_NBITS) {
+ struct ice_event *curr = &src_peer_int->current_events[i];
+
+ if (!bitmap_empty(curr->type, ICE_EVENT_NBITS) &&
+ new_peer->index != src_peer->index &&
+ p_ops->event_handler)
+ p_ops->event_handler(new_peer, curr);
+ }
+
+ return 0;
+}
+
/**
* ice_for_each_peer - iterate across and call function for each peer dev
* @pf: pointer to private board struct
@@ -184,6 +371,88 @@ ice_for_each_peer(struct ice_pf *pf, void *data,
return 0;
}
+/**
+ * ice_finish_init_peer_device - complete peer device initialization
+ * @peer_dev_int: ptr to peer device internal struct
+ * @data: ptr to opaque data
+ *
+ * This function completes remaining initialization of peer_devices
+ */
+int
+ice_finish_init_peer_device(struct ice_peer_dev_int *peer_dev_int,
+ void __always_unused *data)
+{
+ struct ice_peer_dev *peer_dev = &peer_dev_int->peer_dev;
+ struct ice_peer_drv *peer_drv;
+ struct ice_pf *pf;
+ int ret = 0;
+
+ /* peer_dev will not always be populated at the time of this check */
+ if (!ice_validate_peer_dev(peer_dev))
+ return ret;
+
+ peer_drv = peer_dev->peer_drv;
+ pf = pci_get_drvdata(peer_dev->pdev);
+ /* There will be several assessments of the peer_dev's state in this
+ * chunk of logic. We need to hold the peer_dev_int's state mutex
+ * for the entire part so that the flow progresses without another
+ * context changing things mid-flow
+ */
+ mutex_lock(&peer_dev_int->peer_dev_state_mutex);
+
+ if (!peer_dev->peer_ops) {
+ dev_err(&pf->pdev->dev,
+ "peer_ops not defined on peer dev\n");
+ goto init_unlock;
+ }
+
+ if (!peer_dev->peer_ops->open) {
+ dev_err(&pf->pdev->dev,
+ "peer_ops:open not defined on peer dev\n");
+ goto init_unlock;
+ }
+
+ if (!peer_dev->peer_ops->close) {
+ dev_err(&pf->pdev->dev,
+ "peer_ops:close not defined on peer dev\n");
+ goto init_unlock;
+ }
+
+ /* Peer driver expected to set driver_id during registration */
+ if (!peer_drv->driver_id) {
+ dev_err(&pf->pdev->dev,
+ "Peer driver did not set driver_id\n");
+ goto init_unlock;
+ }
+
+ if ((test_bit(ICE_PEER_DEV_STATE_CLOSED, peer_dev_int->state) ||
+ test_bit(ICE_PEER_DEV_STATE_PROBED, peer_dev_int->state)) &&
+ ice_pf_state_is_nominal(pf)) {
+ ice_peer_state_change(peer_dev_int, ICE_PEER_DEV_STATE_OPENING,
+ true);
+ ret = peer_dev->peer_ops->open(peer_dev);
+ if (ret) {
+ dev_err(&pf->pdev->dev,
+ "Peer %d failed to open\n",
+ peer_dev->peer_dev_id);
+ ice_peer_state_change(peer_dev_int,
+ ICE_PEER_DEV_STATE_PROBED, true);
+ goto init_unlock;
+ }
+
+ ice_peer_state_change(peer_dev_int, ICE_PEER_DEV_STATE_OPENED,
+ true);
+ ret = ice_for_each_peer(pf, peer_dev,
+ ice_check_peer_for_events);
+ ice_check_peer_drv_for_events(peer_dev);
+ }
+
+init_unlock:
+ mutex_unlock(&peer_dev_int->peer_dev_state_mutex);
+
+ return ret;
+}
+
/**
* ice_unreg_peer_device - unregister specified device
* @peer_dev_int: ptr to peer device internal
@@ -196,6 +465,7 @@ int
ice_unreg_peer_device(struct ice_peer_dev_int *peer_dev_int,
void __always_unused *data)
{
+ struct ice_peer_drv_int *peer_drv_int;
struct ice_pf *pf;
if (!peer_dev_int)
@@ -207,6 +477,8 @@ ice_unreg_peer_device(struct ice_peer_dev_int *peer_dev_int,
if (!pf)
return 0;
+ peer_drv_int = peer_dev_int->peer_drv_int;
+
if (peer_dev_int->ice_peer_wq) {
if (peer_dev_int->peer_prep_task.func)
cancel_work_sync(&peer_dev_int->peer_prep_task);
@@ -219,6 +491,8 @@ ice_unreg_peer_device(struct ice_peer_dev_int *peer_dev_int,
/* Cleanup the allocated ID for this peer device */
ida_simple_remove(&ice_peer_index_ida, peer_dev_int->peer_dev.index);
+ devm_kfree(&pf->pdev->dev, peer_drv_int);
+
devm_kfree(&pf->pdev->dev, peer_dev_int);
return 0;
@@ -248,6 +522,607 @@ ice_unroll_peer(struct ice_peer_dev_int *peer_dev_int,
return 0;
}
+/**
+ * ice_find_vsi - Find the VSI from VSI ID
+ * @pf: The PF pointer to search in
+ * @vsi_num: The VSI ID to search for
+ */
+static struct ice_vsi *ice_find_vsi(struct ice_pf *pf, u16 vsi_num)
+{
+ int i;
+
+ ice_for_each_vsi(pf, i)
+ if (pf->vsi[i] && pf->vsi[i]->vsi_num == vsi_num)
+ return pf->vsi[i];
+ return NULL;
+}
+
+/**
+ * ice_peer_alloc_rdma_qsets - Allocate Leaf Nodes for RDMA Qset
+ * @peer_dev: peer that is requesting the Leaf Nodes
+ * @res: Resources to be allocated
+ * @partial_acceptable: If partial allocation is acceptable to the peer
+ *
+ * This function allocates Leaf Nodes for given RDMA Qset resources
+ * for the peer device.
+ */
+static int
+ice_peer_alloc_rdma_qsets(struct ice_peer_dev *peer_dev, struct ice_res *res,
+ int __maybe_unused partial_acceptable)
+{
+ u16 max_rdmaqs[ICE_MAX_TRAFFIC_CLASS];
+ enum ice_status status;
+ struct ice_vsi *vsi;
+ struct ice_pf *pf;
+ int i, ret = 0;
+ u32 *qset_teid;
+ u16 *qs_handle;
+
+ if (!ice_validate_peer_dev(peer_dev) || !res)
+ return -EINVAL;
+
+ pf = pci_get_drvdata(peer_dev->pdev);
+
+ if (res->cnt_req > ICE_MAX_TXQ_PER_TXQG)
+ return -EINVAL;
+
+ qset_teid = devm_kcalloc(&pf->pdev->dev, res->cnt_req,
+ sizeof(*qset_teid), GFP_KERNEL);
+ if (!qset_teid)
+ return -ENOMEM;
+
+ qs_handle = devm_kcalloc(&pf->pdev->dev, res->cnt_req,
+ sizeof(*qs_handle), GFP_KERNEL);
+ if (!qs_handle) {
+ devm_kfree(&pf->pdev->dev, qset_teid);
+ return -ENOMEM;
+ }
+
+ ice_for_each_traffic_class(i)
+ max_rdmaqs[i] = 0;
+
+ for (i = 0; i < res->cnt_req; i++) {
+ struct ice_rdma_qset_params *qset;
+
+ qset = &res->res[i].res.qsets;
+ if (qset->vsi_id != peer_dev->pf_vsi_num) {
+ dev_err(&pf->pdev->dev,
+ "RDMA QSet invalid VSI requested\n");
+ ret = -EINVAL;
+ goto out;
+ }
+ max_rdmaqs[qset->tc]++;
+ qs_handle[i] = qset->qs_handle;
+ }
+
+ vsi = ice_find_vsi(pf, peer_dev->pf_vsi_num);
+ if (!vsi) {
+ dev_err(&pf->pdev->dev, "RDMA QSet invalid VSI\n");
+ ret = -EINVAL;
+ goto out;
+ }
+
+ status = ice_cfg_vsi_rdma(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
+ max_rdmaqs);
+ if (status) {
+ dev_err(&pf->pdev->dev, "Failed VSI RDMA qset config\n");
+ ret = -EINVAL;
+ goto out;
+ }
+
+ for (i = 0; i < res->cnt_req; i++) {
+ struct ice_rdma_qset_params *qset;
+
+ qset = &res->res[i].res.qsets;
+ status = ice_ena_vsi_rdma_qset(vsi->port_info, vsi->idx,
+ qset->tc, &qs_handle[i], 1,
+ &qset_teid[i]);
+ if (status) {
+ dev_err(&pf->pdev->dev,
+ "Failed VSI RDMA qset enable\n");
+ ret = -EINVAL;
+ goto out;
+ }
+ vsi->qset_handle[qset->tc] = qset->qs_handle;
+ qset->teid = qset_teid[i];
+ }
+
+out:
+ devm_kfree(&pf->pdev->dev, qset_teid);
+ devm_kfree(&pf->pdev->dev, qs_handle);
+ return ret;
+}
+
+/**
+ * ice_peer_free_rdma_qsets - Free leaf nodes for RDMA Qset
+ * @peer_dev: peer that requested qsets to be freed
+ * @res: Resource to be freed
+ */
+static int
+ice_peer_free_rdma_qsets(struct ice_peer_dev *peer_dev, struct ice_res *res)
+{
+ enum ice_status status;
+ int count, i, ret = 0;
+ struct ice_vsi *vsi;
+ struct ice_pf *pf;
+ u16 vsi_id;
+ u32 *teid;
+ u16 *q_id;
+
+ if (!ice_validate_peer_dev(peer_dev) || !res)
+ return -EINVAL;
+
+ pf = pci_get_drvdata(peer_dev->pdev);
+
+ count = res->res_allocated;
+ if (count > ICE_MAX_TXQ_PER_TXQG)
+ return -EINVAL;
+
+ teid = devm_kcalloc(&pf->pdev->dev, count, sizeof(*teid), GFP_KERNEL);
+ if (!teid)
+ return -ENOMEM;
+
+ q_id = devm_kcalloc(&pf->pdev->dev, count, sizeof(*q_id), GFP_KERNEL);
+ if (!q_id) {
+ devm_kfree(&pf->pdev->dev, teid);
+ return -ENOMEM;
+ }
+
+ vsi_id = res->res[0].res.qsets.vsi_id;
+ vsi = ice_find_vsi(pf, vsi_id);
+ if (!vsi) {
+ dev_err(&pf->pdev->dev, "RDMA Invalid VSI\n");
+ ret = -EINVAL;
+ goto rdma_free_out;
+ }
+
+ for (i = 0; i < count; i++) {
+ struct ice_rdma_qset_params *qset;
+
+ qset = &res->res[i].res.qsets;
+ if (qset->vsi_id != vsi_id) {
+ dev_err(&pf->pdev->dev, "RDMA Invalid VSI ID\n");
+ ret = -EINVAL;
+ goto rdma_free_out;
+ }
+ q_id[i] = qset->qs_handle;
+ teid[i] = qset->teid;
+
+ vsi->qset_handle[qset->tc] = 0;
+ }
+
+ status = ice_dis_vsi_rdma_qset(vsi->port_info, count, teid, q_id);
+ if (status)
+ ret = -EINVAL;
+
+rdma_free_out:
+ devm_kfree(&pf->pdev->dev, teid);
+ devm_kfree(&pf->pdev->dev, q_id);
+
+ return ret;
+}
+
+/**
+ * ice_peer_alloc_res - Allocate requested resources for peer device
+ * @peer_dev: peer that is requesting resources
+ * @res: Resources to be allocated
+ * @partial_acceptable: If partial allocation is acceptable to the peer
+ *
+ * This function allocates requested resources for the peer device.
+ */
+static int
+ice_peer_alloc_res(struct ice_peer_dev *peer_dev, struct ice_res *res,
+ int partial_acceptable)
+{
+ struct ice_pf *pf;
+ int ret;
+
+ if (!ice_validate_peer_dev(peer_dev) || !res)
+ return -EINVAL;
+
+ pf = pci_get_drvdata(peer_dev->pdev);
+ if (!ice_pf_state_is_nominal(pf))
+ return -EBUSY;
+
+ switch (res->res_type) {
+ case ICE_RDMA_QSETS_TXSCHED:
+ ret = ice_peer_alloc_rdma_qsets(peer_dev, res,
+ partial_acceptable);
+ break;
+ default:
+ ret = -EINVAL;
+ break;
+ }
+
+ return ret;
+}
+
+/**
+ * ice_peer_free_res - Free given resources
+ * @peer_dev: peer that is requesting freeing of resources
+ * @res: Resources to be freed
+ *
+ * Free/Release resources allocated to given peer device.
+ */
+static int
+ice_peer_free_res(struct ice_peer_dev *peer_dev, struct ice_res *res)
+{
+ int ret;
+
+ if (!ice_validate_peer_dev(peer_dev) || !res)
+ return -EINVAL;
+
+ switch (res->res_type) {
+ case ICE_RDMA_QSETS_TXSCHED:
+ ret = ice_peer_free_rdma_qsets(peer_dev, res);
+ break;
+ default:
+ ret = -EINVAL;
+ break;
+ }
+
+ return ret;
+}
+
+/**
+ * ice_peer_reg_for_notif - register a peer to receive specific notifications
+ * @peer_dev: peer that is registering for event notifications
+ * @events: mask of event types peer is registering for
+ */
+static void
+ice_peer_reg_for_notif(struct ice_peer_dev *peer_dev, struct ice_event *events)
+{
+ struct ice_peer_dev_int *peer_dev_int;
+ struct ice_pf *pf;
+
+ if (!ice_validate_peer_dev(peer_dev) || !events)
+ return;
+
+ peer_dev_int = peer_to_ice_dev_int(peer_dev);
+ pf = pci_get_drvdata(peer_dev->pdev);
+
+ bitmap_or(peer_dev_int->events, peer_dev_int->events, events->type,
+ ICE_EVENT_NBITS);
+
+ /* Check to see if any events happened previous to peer registering */
+ ice_for_each_peer(pf, peer_dev, ice_check_peer_for_events);
+ ice_check_peer_drv_for_events(peer_dev);
+}
+
+/**
+ * ice_peer_unreg_for_notif - unreg a peer from receiving certain notifications
+ * @peer_dev: peer that is unregistering from event notifications
+ * @events: mask of event types peer is unregistering for
+ */
+static void
+ice_peer_unreg_for_notif(struct ice_peer_dev *peer_dev,
+ struct ice_event *events)
+{
+ struct ice_peer_dev_int *peer_dev_int;
+
+ if (!ice_validate_peer_dev(peer_dev) || !events)
+ return;
+
+ peer_dev_int = peer_to_ice_dev_int(peer_dev);
+
+ bitmap_andnot(peer_dev_int->events, peer_dev_int->events, events->type,
+ ICE_EVENT_NBITS);
+}
+
+/**
+ * ice_peer_check_for_reg - check to see if any peers are reg'd for event
+ * @peer_dev_int: ptr to peer device internal struct
+ * @data: ptr to opaque data, to be used for ice_event to report
+ *
+ * This function is to be called by device_for_each_child to handle an
+ * event reported by a peer or the ice driver.
+ */
+int ice_peer_check_for_reg(struct ice_peer_dev_int *peer_dev_int, void *data)
+{
+ struct ice_event *event = (struct ice_event *)data;
+ DECLARE_BITMAP(comp_events, ICE_EVENT_NBITS);
+ struct ice_peer_dev *peer_dev;
+ bool check = true;
+
+ peer_dev = &peer_dev_int->peer_dev;
+
+ if (!ice_validate_peer_dev(peer_dev) || !data)
+ /* If invalid dev, in this case return 0 instead of error
+ * because caller ignores this return value
+ */
+ return 0;
+
+ if (event->reporter)
+ check = event->reporter->index != peer_dev->index;
+
+ if (bitmap_and(comp_events, event->type, peer_dev_int->events,
+ ICE_EVENT_NBITS) &&
+ (test_bit(ICE_PEER_DEV_STATE_OPENED, peer_dev_int->state) ||
+ test_bit(ICE_PEER_DEV_STATE_PREP_RST, peer_dev_int->state) ||
+ test_bit(ICE_PEER_DEV_STATE_PREPPED, peer_dev_int->state)) &&
+ check &&
+ peer_dev->peer_ops->event_handler)
+ peer_dev->peer_ops->event_handler(peer_dev, event);
+
+ return 0;
+}
+
+/**
+ * ice_peer_report_state_change - accept report of a peer state change
+ * @peer_dev: peer that is sending notification about state change
+ * @event: ice_event holding info on what the state change is
+ *
+ * We also need to parse the list of peers to see if anyone is registered
+ * for notifications about this state change event, and if so, notify them.
+ */
+static void
+ice_peer_report_state_change(struct ice_peer_dev *peer_dev,
+ struct ice_event *event)
+{
+ struct ice_peer_dev_int *peer_dev_int;
+ struct ice_peer_drv_int *peer_drv_int;
+ int e_type, drv_event = 0;
+ struct ice_pf *pf;
+
+ if (!ice_validate_peer_dev(peer_dev) || !event)
+ return;
+
+ pf = pci_get_drvdata(peer_dev->pdev);
+ peer_dev_int = peer_to_ice_dev_int(peer_dev);
+ peer_drv_int = peer_dev_int->peer_drv_int;
+
+ e_type = find_first_bit(event->type, ICE_EVENT_NBITS);
+ if (!e_type)
+ return;
+
+ switch (e_type) {
+ /* Check for peer_drv events */
+ case ICE_EVENT_MBX_CHANGE:
+ drv_event = 1;
+ if (event->info.mbx_rdy)
+ set_bit(ICE_PEER_DRV_STATE_MBX_RDY,
+ peer_drv_int->state);
+ else
+ clear_bit(ICE_PEER_DRV_STATE_MBX_RDY,
+ peer_drv_int->state);
+ break;
+
+ /* Check for peer_dev events */
+ case ICE_EVENT_API_CHANGE:
+ if (event->info.api_rdy)
+ set_bit(ICE_PEER_DEV_STATE_API_RDY,
+ peer_dev_int->state);
+ else
+ clear_bit(ICE_PEER_DEV_STATE_API_RDY,
+ peer_dev_int->state);
+ break;
+
+ default:
+ return;
+ }
+
+ /* store the event and state to notify any new peers opening */
+ if (drv_event)
+ memcpy(&peer_drv_int->current_events[e_type], event,
+ sizeof(*event));
+ else
+ memcpy(&peer_dev_int->current_events[e_type], event,
+ sizeof(*event));
+
+ ice_for_each_peer(pf, event, ice_peer_check_for_reg);
+}
+
+/**
+ * ice_peer_unregister - request to unregister peer
+ * @peer_dev: peer device
+ *
+ * This function triggers close/remove on peer_dev allowing peer
+ * to unregister.
+ */
+static int ice_peer_unregister(struct ice_peer_dev *peer_dev)
+{
+ enum ice_close_reason reason = ICE_REASON_PEER_DEV_UNINIT;
+ struct ice_peer_dev_int *peer_dev_int;
+ struct ice_pf *pf;
+ int ret;
+
+ if (!ice_validate_peer_dev(peer_dev))
+ return -EINVAL;
+
+ pf = pci_get_drvdata(peer_dev->pdev);
+ if (ice_is_reset_in_progress(pf->state))
+ return -EBUSY;
+
+ peer_dev_int = peer_to_ice_dev_int(peer_dev);
+
+ ret = ice_peer_close(peer_dev_int, &reason);
+ if (ret)
+ return ret;
+
+ peer_dev->peer_ops = NULL;
+
+ ice_peer_state_change(peer_dev_int, ICE_PEER_DEV_STATE_REMOVED, false);
+
+ return ret;
+}
+
+/**
+ * ice_peer_register - Called by peer to open communication with LAN
+ * @peer_dev: ptr to peer device
+ *
+ * registering peer is expected to populate the ice_peerdrv->name field
+ * before calling this function.
+ */
+static int ice_peer_register(struct ice_peer_dev *peer_dev)
+{
+ struct ice_peer_drv_int *peer_drv_int;
+ struct ice_peer_dev_int *peer_dev_int;
+ struct ice_peer_drv *peer_drv;
+
+ if (!peer_dev) {
+ pr_err("Failed to reg peer dev: peer_dev ptr NULL\n");
+ return -EINVAL;
+ }
+
+ if (!peer_dev->pdev) {
+ pr_err("Failed to reg peer dev: peer dev pdev NULL\n");
+ return -EINVAL;
+ }
+
+ if (!peer_dev->peer_ops || !peer_dev->ops) {
+ pr_err("Failed to reg peer dev: peer dev peer_ops/ops NULL\n");
+ return -EINVAL;
+ }
+
+ peer_drv = peer_dev->peer_drv;
+ if (!peer_drv) {
+ pr_err("Failed to reg peer dev: peer drv NULL\n");
+ return -EINVAL;
+ }
+
+ if (peer_drv->ver.major != ICE_PEER_MAJOR_VER ||
+ peer_drv->ver.minor != ICE_PEER_MINOR_VER) {
+ pr_err("failed to register due to version mismatch:\n");
+ pr_err("expected major ver %d, caller specified major ver %d\n",
+ ICE_PEER_MAJOR_VER, peer_drv->ver.major);
+ pr_err("expected minor ver %d, caller specified minor ver %d\n",
+ ICE_PEER_MINOR_VER, peer_drv->ver.minor);
+ return -EINVAL;
+ }
+
+ peer_dev_int = peer_to_ice_dev_int(peer_dev);
+ peer_drv_int = peer_dev_int->peer_drv_int;
+ if (!peer_drv_int) {
+ pr_err("Failed to match peer_drv_int to peer_dev\n");
+ return -EINVAL;
+ }
+
+ peer_drv_int->peer_drv = peer_drv;
+
+ ice_peer_state_change(peer_dev_int, ICE_PEER_DEV_STATE_PROBED, false);
+
+ return 0;
+}
+
+/**
+ * ice_peer_request_reset - accept request from peer to perform a reset
+ * @peer_dev: peer device that is request a reset
+ * @reset_type: type of reset the peer is requesting
+ */
+static int
+ice_peer_request_reset(struct ice_peer_dev *peer_dev,
+ enum ice_peer_reset_type reset_type)
+{
+ enum ice_reset_req reset;
+ struct ice_pf *pf;
+
+ if (!ice_validate_peer_dev(peer_dev))
+ return -EINVAL;
+
+ pf = pci_get_drvdata(peer_dev->pdev);
+
+ switch (reset_type) {
+ case ICE_PEER_PFR:
+ reset = ICE_RESET_PFR;
+ break;
+ case ICE_PEER_CORER:
+ reset = ICE_RESET_CORER;
+ break;
+ case ICE_PEER_GLOBR:
+ reset = ICE_RESET_GLOBR;
+ break;
+ default:
+ dev_err(&pf->pdev->dev, "incorrect reset request from peer\n");
+ return -EINVAL;
+ }
+
+ return ice_schedule_reset(pf, reset);
+}
+
+/**
+ * ice_peer_is_vsi_ready - query if VSI in nominal state
+ * @peer_dev: pointer to ice_peer_dev struct
+ */
+static int ice_peer_is_vsi_ready(struct ice_peer_dev *peer_dev)
+{
+ DECLARE_BITMAP(check_bits, __ICE_STATE_NBITS) = { 0 };
+ struct ice_netdev_priv *np;
+ struct ice_vsi *vsi;
+
+ /* If the peer_dev or associated values are not valid, then return
+ * 0 as there is no ready port associated with the values passed in
+ * as parameters.
+ */
+
+ if (!ice_validate_peer_dev(peer_dev))
+ return 0;
+
+ if (!peer_dev->netdev)
+ return 0;
+
+ np = netdev_priv(peer_dev->netdev);
+ vsi = np->vsi;
+ if (!vsi)
+ return 0;
+
+ bitmap_set(check_bits, 0, __ICE_STATE_NOMINAL_CHECK_BITS);
+ if (bitmap_intersects(vsi->state, check_bits, __ICE_STATE_NBITS))
+ return 0;
+
+ return 1;
+}
+
+/**
+ * ice_peer_update_vsi_filter - update filters for RDMA VSI
+ * @peer_dev: pointer to RDMA peer device
+ * @filter: selection of filters to enable or disable
+ * @enable: bool whether to enable or disable filters
+ */
+static int
+ice_peer_update_vsi_filter(struct ice_peer_dev *peer_dev,
+ enum ice_rdma_filter __maybe_unused filter,
+ bool enable)
+{
+ struct ice_pf *pf;
+ int ret, v;
+ u16 idx;
+
+ if (!ice_validate_peer_dev(peer_dev))
+ return -EINVAL;
+
+ pf = pci_get_drvdata(peer_dev->pdev);
+
+ ice_for_each_vsi(pf, v)
+ if (peer_dev->pf_vsi_num == pf->vsi[v]->vsi_num) {
+ idx = pf->vsi[v]->idx;
+ break;
+ }
+ if (v >= pf->num_alloc_vsi)
+ return -EINVAL;
+
+ ret = ice_cfg_iwarp_fltr(&pf->hw, idx, enable);
+
+ if (ret)
+ dev_err(&pf->pdev->dev, "Failed to %sable iWARP filtering\n",
+ enable ? "en" : "dis");
+
+ return ret;
+}
+
+/* Initialize the ice_ops struct, which is used in 'ice_init_peer_devices' */
+static const struct ice_ops ops = {
+ .alloc_res = ice_peer_alloc_res,
+ .free_res = ice_peer_free_res,
+ .is_vsi_ready = ice_peer_is_vsi_ready,
+ .reg_for_notification = ice_peer_reg_for_notif,
+ .unreg_for_notification = ice_peer_unreg_for_notif,
+ .notify_state_change = ice_peer_report_state_change,
+ .request_reset = ice_peer_request_reset,
+ .peer_register = ice_peer_register,
+ .peer_unregister = ice_peer_unregister,
+ .update_vsi_filter = ice_peer_update_vsi_filter,
+};
+
/**
* ice_reserve_peer_qvector - Reserve vector resources for peer drivers
* @pf: board private structure to initialize
@@ -267,6 +1142,41 @@ static int ice_reserve_peer_qvector(struct ice_pf *pf)
return 0;
}
+/**
+ * ice_peer_close_task - call peer's close asynchronously
+ * @work: pointer to work_struct contained by the peer_dev_int struct
+ *
+ * This method (asynchronous) of calling a peer's close function is
+ * meant to be used in the reset path.
+ */
+static void ice_peer_close_task(struct work_struct *work)
+{
+ struct ice_peer_dev_int *peer_dev_int;
+ struct ice_peer_dev *peer_dev;
+
+ peer_dev_int = container_of(work, struct ice_peer_dev_int,
+ peer_close_task);
+
+ peer_dev = &peer_dev_int->peer_dev;
+ if (!peer_dev || !peer_dev->peer_ops)
+ return;
+
+ /* If this peer_dev is going to close, we do not want any state changes
+ * to happen until after we successfully finish or abort the close.
+ * Grab the peer_dev_state_mutex to protect this flow
+ */
+ mutex_lock(&peer_dev_int->peer_dev_state_mutex);
+
+ ice_peer_state_change(peer_dev_int, ICE_PEER_DEV_STATE_CLOSING, true);
+
+ if (peer_dev->peer_ops->close)
+ peer_dev->peer_ops->close(peer_dev, peer_dev_int->rst_type);
+
+ ice_peer_state_change(peer_dev_int, ICE_PEER_DEV_STATE_CLOSED, true);
+
+ mutex_unlock(&peer_dev_int->peer_dev_state_mutex);
+}
+
/**
* ice_peer_device_release - release function for platform peer device
* @dev: pointer to device structure
@@ -313,6 +1223,7 @@ int ice_init_peer_devices(struct ice_pf *pf)
}
for (i = 0; i < ARRAY_SIZE(peer_dev_ids); i++) {
struct ice_peer_dev_int *peer_dev_int;
+ struct ice_peer_drv_int *peer_drv_int;
struct platform_device *platform_dev;
struct ice_qos_params *qos_info;
struct ice_peer_dev *peer_dev;
@@ -334,6 +1245,19 @@ int ice_init_peer_devices(struct ice_pf *pf)
return -ENOMEM;
pf->peers[i] = peer_dev_int;
+ peer_drv_int = devm_kzalloc(&pdev->dev, sizeof(*peer_drv_int),
+ GFP_KERNEL);
+ if (!peer_drv_int)
+ return -ENOMEM;
+
+ peer_drv_int->peer_dev = &peer_dev_int->peer_dev;
+ peer_dev_int->peer_drv_int = peer_drv_int;
+
+ /* Initialize driver values */
+ for (j = 0; j < ICE_EVENT_NBITS; j++)
+ bitmap_zero(peer_drv_int->current_events[j].type,
+ ICE_EVENT_NBITS);
+
mutex_init(&peer_dev_int->peer_dev_state_mutex);
peer_dev = &peer_dev_int->peer_dev;
@@ -361,6 +1285,7 @@ int ice_init_peer_devices(struct ice_pf *pf)
i);
if (!peer_dev_int->ice_peer_wq)
return -ENOMEM;
+ INIT_WORK(&peer_dev_int->peer_close_task, ice_peer_close_task);
/* Assign a unique index and hence name for peer device */
status = ida_simple_get(&ice_peer_index_ida, 0, 0, GFP_KERNEL);
@@ -397,6 +1322,8 @@ int ice_init_peer_devices(struct ice_pf *pf)
/* for DCB, override the qos_info defaults. */
ice_setup_dcb_qos_info(pf, qos_info);
+ /* Initialize ice_ops */
+ peer_dev->ops = &ops;
/* make sure peer specific resources such as msix_count and
* msix_entries are initialized
diff --git a/drivers/net/ethernet/intel/ice/ice_idc.h b/drivers/net/ethernet/intel/ice/ice_idc.h
index d29e946fe2f1..455c1192ca4f 100644
--- a/drivers/net/ethernet/intel/ice/ice_idc.h
+++ b/drivers/net/ethernet/intel/ice/ice_idc.h
@@ -357,4 +357,21 @@ static inline struct ice_peer_dev *dev_to_ice_peer(struct device *_dev)
platform_dev = container_of(_dev, struct platform_device, dev);
return container_of(platform_dev, struct ice_peer_dev, platform_dev);
}
+
+/* structure representing peer driver
+ * Peer driver to initialize those function ptrs and
+ * it will be invoked by ICE as part of driver_registration
+ * via bus infrastructure
+ */
+struct ice_peer_drv {
+ u16 driver_id;
+#define ICE_PEER_LAN_DRIVER 0
+#define ICE_PEER_RDMA_DRIVER 4
+#define ICE_PEER_ADK_DRIVER 5
+
+ struct ice_ver_info ver;
+ const char *name;
+
+};
+
#endif /* _ICE_IDC_H_*/
diff --git a/drivers/net/ethernet/intel/ice/ice_idc_int.h b/drivers/net/ethernet/intel/ice/ice_idc_int.h
index f8ff8cac7a26..e67955a36342 100644
--- a/drivers/net/ethernet/intel/ice/ice_idc_int.h
+++ b/drivers/net/ethernet/intel/ice/ice_idc_int.h
@@ -6,8 +6,16 @@
#include "ice_idc.h"
+int ice_close_peer_for_reset(struct ice_peer_dev_int *peer_dev_int, void *data);
int ice_unroll_peer(struct ice_peer_dev_int *peer_dev_int, void *data);
int ice_unreg_peer_device(struct ice_peer_dev_int *peer_dev_int, void *data);
+int ice_peer_close(struct ice_peer_dev_int *peer_dev_int, void *data);
+int ice_peer_check_for_reg(struct ice_peer_dev_int *peer_dev_int, void *data);
+int
+ice_finish_init_peer_device(struct ice_peer_dev_int *peer_dev_int, void *data);
+#ifdef CONFIG_DCB
+bool ice_validate_peer_dev(struct ice_peer_dev *peer);
+#endif /* CONFIG_DCB */
enum ice_peer_dev_state {
ICE_PEER_DEV_STATE_INIT,
@@ -64,4 +72,11 @@ struct ice_peer_dev_int {
enum ice_close_reason rst_type;
};
+
+static inline struct
+ice_peer_dev_int *peer_to_ice_dev_int(struct ice_peer_dev *peer_dev)
+{
+ return container_of(peer_dev, struct ice_peer_dev_int, peer_dev);
+}
+
#endif /* !_ICE_IDC_INT_H_ */
diff --git a/drivers/net/ethernet/intel/ice/ice_lib.c b/drivers/net/ethernet/intel/ice/ice_lib.c
index a19f5920733b..65a3f814b87d 100644
--- a/drivers/net/ethernet/intel/ice/ice_lib.c
+++ b/drivers/net/ethernet/intel/ice/ice_lib.c
@@ -1464,6 +1464,31 @@ int ice_add_mac_to_list(struct ice_vsi *vsi, struct list_head *add_list,
return 0;
}
+/**
+ * ice_pf_state_is_nominal - checks the PF for nominal state
+ * @pf: pointer to PF to check
+ *
+ * Check the PF's state for a collection of bits that would indicate
+ * the PF is in a state that would inhibit normal operation for
+ * driver functionality.
+ *
+ * Returns true if PF is in a nominal state.
+ * Returns false otherwise
+ */
+bool ice_pf_state_is_nominal(struct ice_pf *pf)
+{
+ DECLARE_BITMAP(check_bits, __ICE_STATE_NBITS) = { 0 };
+
+ if (!pf)
+ return false;
+
+ bitmap_set(check_bits, 0, __ICE_STATE_NOMINAL_CHECK_BITS);
+ if (bitmap_intersects(pf->state, check_bits, __ICE_STATE_NBITS))
+ return false;
+
+ return true;
+}
+
/**
* ice_update_eth_stats - Update VSI-specific ethernet statistics counters
* @vsi: the VSI to be updated
@@ -2684,9 +2709,17 @@ void ice_vsi_free_rx_rings(struct ice_vsi *vsi)
*/
void ice_vsi_close(struct ice_vsi *vsi)
{
+ enum ice_close_reason reason = ICE_REASON_INTERFACE_DOWN;
+ struct device *dev = &vsi->back->pdev->dev;
+ int ret = 0;
+
+ if (vsi->type == ICE_VSI_PF)
+ ret = ice_for_each_peer(vsi->back, &reason, ice_peer_close);
+
+ if (ret)
+ dev_dbg(dev, "Peer device did not implement close function\n");
if (!test_and_set_bit(__ICE_DOWN, vsi->state))
ice_down(vsi);
-
ice_vsi_free_irq(vsi);
ice_vsi_free_tx_rings(vsi);
ice_vsi_free_rx_rings(vsi);
diff --git a/drivers/net/ethernet/intel/ice/ice_lib.h b/drivers/net/ethernet/intel/ice/ice_lib.h
index 6e43ef03bfc3..5353a71fcca0 100644
--- a/drivers/net/ethernet/intel/ice/ice_lib.h
+++ b/drivers/net/ethernet/intel/ice/ice_lib.h
@@ -11,6 +11,8 @@ int ice_add_mac_to_list(struct ice_vsi *vsi, struct list_head *add_list,
void ice_free_fltr_list(struct device *dev, struct list_head *h);
+bool ice_pf_state_is_nominal(struct ice_pf *pf);
+
void ice_update_eth_stats(struct ice_vsi *vsi);
int ice_vsi_cfg_rxqs(struct ice_vsi *vsi);
diff --git a/drivers/net/ethernet/intel/ice/ice_main.c b/drivers/net/ethernet/intel/ice/ice_main.c
index 08bce0d67634..62b45c918ab5 100644
--- a/drivers/net/ethernet/intel/ice/ice_main.c
+++ b/drivers/net/ethernet/intel/ice/ice_main.c
@@ -576,6 +576,7 @@ static void ice_reset_subtask(struct ice_pf *pf)
/* return if no valid reset type requested */
if (reset_type == ICE_RESET_INVAL)
return;
+ ice_for_each_peer(pf, &reset_type, ice_close_peer_for_reset);
ice_prepare_for_reset(pf);
/* make sure we are ready to rebuild */
@@ -1336,6 +1337,8 @@ static void ice_service_task(struct work_struct *work)
return;
}
+ /* Invoke remaining initialization of peer devices */
+ ice_for_each_peer(pf, NULL, ice_finish_init_peer_device);
ice_check_for_hang_subtask(pf);
ice_sync_fltr_subtask(pf);
ice_handle_mdd_event(pf);
@@ -1375,6 +1378,42 @@ static void ice_set_ctrlq_len(struct ice_hw *hw)
hw->mailboxq.sq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
}
+/**
+ * ice_schedule_reset - schedule a reset
+ * @pf: board private structure
+ * @reset: reset being requested
+ */
+int ice_schedule_reset(struct ice_pf *pf, enum ice_reset_req reset)
+{
+ /* bail out if earlier reset has failed */
+ if (test_bit(__ICE_RESET_FAILED, pf->state)) {
+ dev_dbg(&pf->pdev->dev, "earlier reset has failed\n");
+ return -EIO;
+ }
+ /* bail if reset/recovery already in progress */
+ if (ice_is_reset_in_progress(pf->state)) {
+ dev_dbg(&pf->pdev->dev, "Reset already in progress\n");
+ return -EBUSY;
+ }
+
+ switch (reset) {
+ case ICE_RESET_PFR:
+ set_bit(__ICE_PFR_REQ, pf->state);
+ break;
+ case ICE_RESET_CORER:
+ set_bit(__ICE_CORER_REQ, pf->state);
+ break;
+ case ICE_RESET_GLOBR:
+ set_bit(__ICE_GLOBR_REQ, pf->state);
+ break;
+ default:
+ return -EINVAL;
+ }
+
+ ice_service_task_schedule(pf);
+ return 0;
+}
+
/**
* ice_irq_affinity_notify - Callback for affinity changes
* @notify: context as to what irq was changed
@@ -2496,6 +2535,7 @@ ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent)
static void ice_remove(struct pci_dev *pdev)
{
struct ice_pf *pf = pci_get_drvdata(pdev);
+ enum ice_close_reason reason;
int i;
if (!pf)
@@ -2507,8 +2547,10 @@ static void ice_remove(struct pci_dev *pdev)
msleep(100);
}
- set_bit(__ICE_DOWN, pf->state);
ice_service_task_stop(pf);
+ reason = ICE_REASON_INTERFACE_DOWN;
+ ice_for_each_peer(pf, &reason, ice_peer_close);
+ set_bit(__ICE_DOWN, pf->state);
if (test_bit(ICE_FLAG_SRIOV_ENA, pf->flags))
ice_free_vfs(pf);
@@ -3946,6 +3988,7 @@ static int ice_change_mtu(struct net_device *netdev, int new_mtu)
struct ice_netdev_priv *np = netdev_priv(netdev);
struct ice_vsi *vsi = np->vsi;
struct ice_pf *pf = vsi->back;
+ struct ice_event *event;
u8 count = 0;
if (new_mtu == netdev->mtu) {
@@ -3997,6 +4040,13 @@ static int ice_change_mtu(struct net_device *netdev, int new_mtu)
}
}
+ event = devm_kzalloc(&pf->pdev->dev, sizeof(*event), GFP_KERNEL);
+ set_bit(ICE_EVENT_MTU_CHANGE, event->type);
+ event->reporter = NULL;
+ event->info.mtu = new_mtu;
+ ice_for_each_peer(pf, event, ice_peer_check_for_reg);
+ devm_kfree(&pf->pdev->dev, event);
+
netdev_info(netdev, "changed MTU to %d\n", new_mtu);
return 0;
}
diff --git a/drivers/net/ethernet/intel/ice/ice_sched.c b/drivers/net/ethernet/intel/ice/ice_sched.c
index 2a232504379d..1bc5b8e61eee 100644
--- a/drivers/net/ethernet/intel/ice/ice_sched.c
+++ b/drivers/net/ethernet/intel/ice/ice_sched.c
@@ -576,6 +576,50 @@ ice_alloc_lan_q_ctx(struct ice_hw *hw, u16 vsi_handle, u8 tc, u16 new_numqs)
return 0;
}
+/**
+ * ice_alloc_rdma_q_ctx - allocate RDMA queue contexts for the given VSI and TC
+ * @hw: pointer to the HW struct
+ * @vsi_handle: VSI handle
+ * @tc: TC number
+ * @new_numqs: number of queues
+ */
+static enum ice_status
+ice_alloc_rdma_q_ctx(struct ice_hw *hw, u16 vsi_handle, u8 tc, u16 new_numqs)
+{
+ struct ice_vsi_ctx *vsi_ctx;
+ struct ice_q_ctx *q_ctx;
+
+ vsi_ctx = ice_get_vsi_ctx(hw, vsi_handle);
+ if (!vsi_ctx)
+ return ICE_ERR_PARAM;
+ /* allocate RDMA queue contexts */
+ if (!vsi_ctx->rdma_q_ctx[tc]) {
+ vsi_ctx->rdma_q_ctx[tc] = devm_kcalloc(ice_hw_to_dev(hw),
+ new_numqs,
+ sizeof(*q_ctx),
+ GFP_KERNEL);
+ if (!vsi_ctx->rdma_q_ctx[tc])
+ return ICE_ERR_NO_MEMORY;
+ vsi_ctx->num_rdma_q_entries[tc] = new_numqs;
+ return 0;
+ }
+ /* num queues are increased, update the queue contexts */
+ if (new_numqs > vsi_ctx->num_rdma_q_entries[tc]) {
+ u16 prev_num = vsi_ctx->num_rdma_q_entries[tc];
+
+ q_ctx = devm_kcalloc(ice_hw_to_dev(hw), new_numqs,
+ sizeof(*q_ctx), GFP_KERNEL);
+ if (!q_ctx)
+ return ICE_ERR_NO_MEMORY;
+ memcpy(q_ctx, vsi_ctx->rdma_q_ctx[tc],
+ prev_num * sizeof(*q_ctx));
+ devm_kfree(ice_hw_to_dev(hw), vsi_ctx->rdma_q_ctx[tc]);
+ vsi_ctx->rdma_q_ctx[tc] = q_ctx;
+ vsi_ctx->num_rdma_q_entries[tc] = new_numqs;
+ }
+ return 0;
+}
+
/**
* ice_sched_clear_agg - clears the aggregator related information
* @hw: pointer to the hardware structure
@@ -1447,13 +1491,22 @@ ice_sched_update_vsi_child_nodes(struct ice_port_info *pi, u16 vsi_handle,
if (!vsi_ctx)
return ICE_ERR_PARAM;
- prev_numqs = vsi_ctx->sched.max_lanq[tc];
+ if (owner == ICE_SCHED_NODE_OWNER_LAN)
+ prev_numqs = vsi_ctx->sched.max_lanq[tc];
+ else
+ prev_numqs = vsi_ctx->sched.max_rdmaq[tc];
/* num queues are not changed or less than the previous number */
if (new_numqs <= prev_numqs)
return status;
- status = ice_alloc_lan_q_ctx(hw, vsi_handle, tc, new_numqs);
- if (status)
- return status;
+ if (owner == ICE_SCHED_NODE_OWNER_LAN) {
+ status = ice_alloc_lan_q_ctx(hw, vsi_handle, tc, new_numqs);
+ if (status)
+ return status;
+ } else {
+ status = ice_alloc_rdma_q_ctx(hw, vsi_handle, tc, new_numqs);
+ if (status)
+ return status;
+ }
if (new_numqs)
ice_sched_calc_vsi_child_nodes(hw, new_numqs, new_num_nodes);
@@ -1468,7 +1521,10 @@ ice_sched_update_vsi_child_nodes(struct ice_port_info *pi, u16 vsi_handle,
new_num_nodes, owner);
if (status)
return status;
- vsi_ctx->sched.max_lanq[tc] = new_numqs;
+ if (owner == ICE_SCHED_NODE_OWNER_LAN)
+ vsi_ctx->sched.max_lanq[tc] = new_numqs;
+ else
+ vsi_ctx->sched.max_rdmaq[tc] = new_numqs;
return 0;
}
@@ -1534,6 +1590,7 @@ ice_sched_cfg_vsi(struct ice_port_info *pi, u16 vsi_handle, u8 tc, u16 maxqs,
* recreate the child nodes all the time in these cases.
*/
vsi_ctx->sched.max_lanq[tc] = 0;
+ vsi_ctx->sched.max_rdmaq[tc] = 0;
}
/* update the VSI child nodes */
@@ -1665,6 +1722,8 @@ ice_sched_rm_vsi_cfg(struct ice_port_info *pi, u16 vsi_handle, u8 owner)
}
if (owner == ICE_SCHED_NODE_OWNER_LAN)
vsi_ctx->sched.max_lanq[i] = 0;
+ else
+ vsi_ctx->sched.max_rdmaq[i] = 0;
}
status = 0;
diff --git a/drivers/net/ethernet/intel/ice/ice_switch.c b/drivers/net/ethernet/intel/ice/ice_switch.c
index 8271fd651725..92c44973af51 100644
--- a/drivers/net/ethernet/intel/ice/ice_switch.c
+++ b/drivers/net/ethernet/intel/ice/ice_switch.c
@@ -346,6 +346,10 @@ static void ice_clear_vsi_q_ctx(struct ice_hw *hw, u16 vsi_handle)
devm_kfree(ice_hw_to_dev(hw), vsi->lan_q_ctx[i]);
vsi->lan_q_ctx[i] = NULL;
}
+ if (vsi->rdma_q_ctx[i]) {
+ devm_kfree(ice_hw_to_dev(hw), vsi->rdma_q_ctx[i]);
+ vsi->rdma_q_ctx[i] = NULL;
+ }
}
}
@@ -467,6 +471,29 @@ ice_update_vsi(struct ice_hw *hw, u16 vsi_handle, struct ice_vsi_ctx *vsi_ctx,
return ice_aq_update_vsi(hw, vsi_ctx, cd);
}
+/**
+ * ice_cfg_iwarp_fltr - enable/disable iWARP filtering on VSI
+ * @hw: pointer to HW struct
+ * @vsi_handle: VSI SW index
+ * @enable: boolean for enable/disable
+ */
+enum ice_status
+ice_cfg_iwarp_fltr(struct ice_hw *hw, u16 vsi_handle, bool enable)
+{
+ struct ice_vsi_ctx *ctx;
+
+ ctx = ice_get_vsi_ctx(hw, vsi_handle);
+ if (!ctx)
+ return ICE_ERR_DOES_NOT_EXIST;
+
+ if (enable)
+ ctx->info.q_opt_flags |= ICE_AQ_VSI_Q_OPT_PE_FLTR_EN;
+ else
+ ctx->info.q_opt_flags &= ~ICE_AQ_VSI_Q_OPT_PE_FLTR_EN;
+
+ return ice_update_vsi(hw, vsi_handle, ctx, NULL);
+}
+
/**
* ice_aq_alloc_free_vsi_list
* @hw: pointer to the HW struct
diff --git a/drivers/net/ethernet/intel/ice/ice_switch.h b/drivers/net/ethernet/intel/ice/ice_switch.h
index cb123fbe30be..a81a9dd509d7 100644
--- a/drivers/net/ethernet/intel/ice/ice_switch.h
+++ b/drivers/net/ethernet/intel/ice/ice_switch.h
@@ -31,6 +31,8 @@ struct ice_vsi_ctx {
u8 vf_num;
u16 num_lan_q_entries[ICE_MAX_TRAFFIC_CLASS];
struct ice_q_ctx *lan_q_ctx[ICE_MAX_TRAFFIC_CLASS];
+ u16 num_rdma_q_entries[ICE_MAX_TRAFFIC_CLASS];
+ struct ice_q_ctx *rdma_q_ctx[ICE_MAX_TRAFFIC_CLASS];
};
enum ice_sw_fwd_act_type {
@@ -225,6 +227,8 @@ void ice_remove_vsi_fltr(struct ice_hw *hw, u16 vsi_handle);
enum ice_status
ice_add_vlan(struct ice_hw *hw, struct list_head *m_list);
enum ice_status ice_remove_vlan(struct ice_hw *hw, struct list_head *v_list);
+enum ice_status
+ice_cfg_iwarp_fltr(struct ice_hw *hw, u16 vsi_handle, bool enable);
/* Promisc/defport setup for VSIs */
enum ice_status
diff --git a/drivers/net/ethernet/intel/ice/ice_type.h b/drivers/net/ethernet/intel/ice/ice_type.h
index 99822b4e8621..737181f006ee 100644
--- a/drivers/net/ethernet/intel/ice/ice_type.h
+++ b/drivers/net/ethernet/intel/ice/ice_type.h
@@ -31,6 +31,7 @@ static inline bool ice_is_tc_ena(u8 bitmap, u8 tc)
#define ICE_DBG_LAN BIT_ULL(8)
#define ICE_DBG_SW BIT_ULL(13)
#define ICE_DBG_SCHED BIT_ULL(14)
+#define ICE_DBG_RDMA BIT_ULL(15)
#define ICE_DBG_RES BIT_ULL(17)
#define ICE_DBG_AQ_MSG BIT_ULL(24)
#define ICE_DBG_AQ_CMD BIT_ULL(27)
@@ -242,6 +243,7 @@ struct ice_sched_node {
u8 tc_num;
u8 owner;
#define ICE_SCHED_NODE_OWNER_LAN 0
+#define ICE_SCHED_NODE_OWNER_RDMA 2
};
/* Access Macros for Tx Sched Elements data */
@@ -267,6 +269,7 @@ struct ice_sched_vsi_info {
struct ice_sched_node *ag_node[ICE_MAX_TRAFFIC_CLASS];
struct list_head list_entry;
u16 max_lanq[ICE_MAX_TRAFFIC_CLASS];
+ u16 max_rdmaq[ICE_MAX_TRAFFIC_CLASS];
};
/* driver defines the policy */
diff --git a/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c b/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c
index 5d24b539648f..8feb0ddad729 100644
--- a/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c
+++ b/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c
@@ -1310,31 +1310,6 @@ static int ice_alloc_vfs(struct ice_pf *pf, u16 num_alloc_vfs)
return ret;
}
-/**
- * ice_pf_state_is_nominal - checks the PF for nominal state
- * @pf: pointer to PF to check
- *
- * Check the PF's state for a collection of bits that would indicate
- * the PF is in a state that would inhibit normal operation for
- * driver functionality.
- *
- * Returns true if PF is in a nominal state.
- * Returns false otherwise
- */
-static bool ice_pf_state_is_nominal(struct ice_pf *pf)
-{
- DECLARE_BITMAP(check_bits, __ICE_STATE_NBITS) = { 0 };
-
- if (!pf)
- return false;
-
- bitmap_set(check_bits, 0, __ICE_STATE_NOMINAL_CHECK_BITS);
- if (bitmap_intersects(pf->state, check_bits, __ICE_STATE_NBITS))
- return false;
-
- return true;
-}
-
/**
* ice_pci_sriov_ena - Enable or change number of VFs
* @pf: pointer to the PF structure
--
2.21.0
^ permalink raw reply related [flat|nested] 15+ messages in thread
* [net-next 3/3] i40e: Register RDMA client devices to the virtual platform bus
2019-07-04 2:12 [net-next 0/3][pull request] Intel Wired LAN ver Updates 2019-07-03 Jeff Kirsher
2019-07-04 2:12 ` [net-next 1/3] ice: Initialize and register platform device to provide RDMA Jeff Kirsher
2019-07-04 2:12 ` [net-next 2/3] ice: Implement peer communications Jeff Kirsher
@ 2019-07-04 2:12 ` Jeff Kirsher
2019-07-04 12:15 ` [net-next 0/3][pull request] Intel Wired LAN ver Updates 2019-07-03 Jason Gunthorpe
3 siblings, 0 replies; 15+ messages in thread
From: Jeff Kirsher @ 2019-07-04 2:12 UTC (permalink / raw)
To: davem, dledford, jgg
Cc: Shiraz Saleem, netdev, linux-rdma, nhorman, sassmann,
mustafa.ismail, david.m.ertman, Andrew Bowers, Jeff Kirsher
From: Shiraz Saleem <shiraz.saleem@intel.com>
Register RDMA client devices to the virtual platform bus. The
unified RDMA driver 'irdma' registers as a platform driver and
will bind to these devices. This model is inspired by the discussion
in [1]. It realizes a single RDMA driver capable of working with
multiple LAN drivers over multi-generation Intel HW supporting RDMA.
There is also no load ordering dependencies between i40e and irdma.
Summary of changes:
* Register a platform client device per PF and unregister the platform
device when PF goes away.
* Add 2 new client ops.
* i40e_client_device_register() which is called during RDMA
probe() per PF. Validate client drv OPs and schedule service
task to call open()
* i40e_client_device_unregister() called during RDMA remove()
per PF. Call client close() and release_qvlist.
* The global register/unregister calls exported for i40iw are retained
until i40iw is removed from the kernel.
[1] https://patchwork.kernel.org/patch/10815567/
Signed-off-by: Shiraz Saleem <shiraz.saleem@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
drivers/net/ethernet/intel/i40e/i40e_client.c | 116 ++++++++++++++----
drivers/net/ethernet/intel/i40e/i40e_client.h | 8 ++
2 files changed, 102 insertions(+), 22 deletions(-)
diff --git a/drivers/net/ethernet/intel/i40e/i40e_client.c b/drivers/net/ethernet/intel/i40e/i40e_client.c
index e81530ca08d0..133950d190a6 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_client.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_client.c
@@ -30,11 +30,17 @@ static int i40e_client_update_vsi_ctxt(struct i40e_info *ldev,
bool is_vf, u32 vf_id,
u32 flag, u32 valid_flag);
+static int i40e_client_device_register(struct i40e_info *ldev);
+
+static void i40e_client_device_unregister(struct i40e_info *ldev);
+
static struct i40e_ops i40e_lan_ops = {
.virtchnl_send = i40e_client_virtchnl_send,
.setup_qvlist = i40e_client_setup_qvlist,
.request_reset = i40e_client_request_reset,
.update_vsi_ctxt = i40e_client_update_vsi_ctxt,
+ .client_device_register = i40e_client_device_register,
+ .client_device_unregister = i40e_client_device_unregister,
};
/**
@@ -71,6 +77,10 @@ int i40e_client_get_params(struct i40e_vsi *vsi, struct i40e_params *params)
return 0;
}
+static void i40e_client_device_release(struct device *dev)
+{
+}
+
/**
* i40e_notify_client_of_vf_msg - call the client vf message callback
* @vsi: the VSI with the message
@@ -287,9 +297,7 @@ static void i40e_client_add_instance(struct i40e_pf *pf)
struct i40e_client_instance *cdev = NULL;
struct netdev_hw_addr *mac = NULL;
struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
-
- if (!registered_client || pf->cinst)
- return;
+ struct platform_device *platform_dev;
cdev = kzalloc(sizeof(*cdev), GFP_KERNEL);
if (!cdev)
@@ -308,6 +316,12 @@ static void i40e_client_add_instance(struct i40e_pf *pf)
cdev->lan_info.fw_maj_ver = pf->hw.aq.fw_maj_ver;
cdev->lan_info.fw_min_ver = pf->hw.aq.fw_min_ver;
cdev->lan_info.fw_build = pf->hw.aq.fw_build;
+ platform_dev = &cdev->lan_info.platform_dev;
+ platform_dev->name = "i40e_rdma";
+ platform_dev->id = PLATFORM_DEVID_AUTO;
+ platform_dev->id_auto = true;
+ platform_dev->dev.release = i40e_client_device_release;
+ platform_dev->dev.parent = &pf->pdev->dev;
set_bit(__I40E_CLIENT_INSTANCE_NONE, &cdev->state);
if (i40e_client_get_params(vsi, &cdev->lan_info.params)) {
@@ -323,10 +337,12 @@ static void i40e_client_add_instance(struct i40e_pf *pf)
else
dev_err(&pf->pdev->dev, "MAC address list is empty!\n");
- cdev->client = registered_client;
+ cdev->client = NULL;
pf->cinst = cdev;
- i40e_client_update_msix_info(pf);
+ cdev->lan_info.msix_count = pf->num_iwarp_msix;
+ cdev->lan_info.msix_entries = &pf->msix_entries[pf->iwarp_base_vector];
+ platform_device_register(platform_dev);
}
/**
@@ -347,7 +363,7 @@ void i40e_client_del_instance(struct i40e_pf *pf)
**/
void i40e_client_subtask(struct i40e_pf *pf)
{
- struct i40e_client *client = registered_client;
+ struct i40e_client *client;
struct i40e_client_instance *cdev;
struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
int ret = 0;
@@ -361,9 +377,11 @@ void i40e_client_subtask(struct i40e_pf *pf)
test_bit(__I40E_CONFIG_BUSY, pf->state))
return;
- if (!client || !cdev)
+ if (!cdev || !cdev->client)
return;
+ client = cdev->client;
+
/* Here we handle client opens. If the client is down, and
* the netdev is registered, then open the client.
*/
@@ -424,16 +442,7 @@ int i40e_lan_add_device(struct i40e_pf *pf)
pf->hw.pf_id, pf->hw.bus.bus_id,
pf->hw.bus.device, pf->hw.bus.func);
- /* If a client has already been registered, we need to add an instance
- * of it to our new LAN device.
- */
- if (registered_client)
- i40e_client_add_instance(pf);
-
- /* Since in some cases register may have happened before a device gets
- * added, we can schedule a subtask to go initiate the clients if
- * they can be launched at probe time.
- */
+ i40e_client_add_instance(pf);
set_bit(__I40E_CLIENT_SERVICE_REQUESTED, pf->state);
i40e_service_event_schedule(pf);
@@ -453,6 +462,8 @@ int i40e_lan_del_device(struct i40e_pf *pf)
struct i40e_device *ldev, *tmp;
int ret = -ENODEV;
+ platform_device_unregister(&pf->cinst->lan_info.platform_dev);
+
/* First, remove any client instance. */
i40e_client_del_instance(pf);
@@ -505,10 +516,7 @@ static void i40e_client_release(struct i40e_client *client)
"Client %s instance for PF id %d closed\n",
client->name, pf->hw.pf_id);
}
- /* delete the client instance */
- i40e_client_del_instance(pf);
- dev_info(&pf->pdev->dev, "Deleted client instance of Client %s\n",
- client->name);
+ cdev->client = NULL;
clear_bit(__I40E_SERVICE_SCHED, pf->state);
}
mutex_unlock(&i40e_device_mutex);
@@ -527,7 +535,7 @@ static void i40e_client_prepare(struct i40e_client *client)
mutex_lock(&i40e_device_mutex);
list_for_each_entry(ldev, &i40e_devices, list) {
pf = ldev->pf;
- i40e_client_add_instance(pf);
+ pf->cinst->client = registered_client;
/* Start the client subtask */
set_bit(__I40E_CLIENT_SERVICE_REQUESTED, pf->state);
i40e_service_event_schedule(pf);
@@ -733,6 +741,70 @@ static int i40e_client_update_vsi_ctxt(struct i40e_info *ldev,
return err;
}
+static int i40e_client_device_register(struct i40e_info *ldev)
+{
+ struct i40e_client *client;
+ struct i40e_pf *pf;
+
+ if (!ldev) {
+ pr_err("Failed to reg client dev: ldev ptr NULL\n");
+ return -EINVAL;
+ }
+
+ client = ldev->client;
+ pf = ldev->pf;
+ if (!client) {
+ pr_err("Failed to reg client dev: client ptr NULL\n");
+ return -EINVAL;
+ }
+
+ if (!ldev->ops || !client->ops) {
+ pr_err("Failed to reg client dev: client dev peer_ops/ops NULL\n");
+ return -EINVAL;
+ }
+
+ if (client->version.major != I40E_CLIENT_VERSION_MAJOR ||
+ client->version.minor != I40E_CLIENT_VERSION_MINOR) {
+ pr_err("i40e: Failed to register client %s due to mismatched client interface version\n",
+ client->name);
+ pr_err("Client is using version: %02d.%02d.%02d while LAN driver supports %s\n",
+ client->version.major, client->version.minor,
+ client->version.build,
+ i40e_client_interface_version_str);
+ return -EINVAL;
+ }
+
+ pf->cinst->client = ldev->client;
+ set_bit(__I40E_CLIENT_SERVICE_REQUESTED, pf->state);
+ i40e_service_event_schedule(pf);
+
+ return 0;
+}
+
+static void i40e_client_device_unregister(struct i40e_info *ldev)
+{
+ struct i40e_pf *pf = ldev->pf;
+ struct i40e_client_instance *cdev = pf->cinst;
+
+ while (test_and_set_bit(__I40E_SERVICE_SCHED, pf->state))
+ usleep_range(500, 1000);
+
+ if (!cdev || !cdev->client || !cdev->client->ops ||
+ !cdev->client->ops->close) {
+ dev_err(&pf->pdev->dev, "Cannot close client device\n");
+ return;
+ }
+ cdev->client->ops->close(&cdev->lan_info, cdev->client, false);
+ clear_bit(__I40E_CLIENT_INSTANCE_OPENED, &cdev->state);
+ i40e_client_release_qvlist(&cdev->lan_info);
+ pf->cinst->client = NULL;
+ clear_bit(__I40E_SERVICE_SCHED, pf->state);
+}
+
+/* Retain legacy global registration/unregistration calls till i40iw is
+ * deprecated from the kernel. The irdma unified driver does not use these
+ * exported symbols.
+ */
/**
* i40e_register_client - Register a i40e client driver with the L2 driver
* @client: pointer to the i40e_client struct
diff --git a/drivers/net/ethernet/intel/i40e/i40e_client.h b/drivers/net/ethernet/intel/i40e/i40e_client.h
index 72994baf4941..0cfdb35e93d6 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_client.h
+++ b/drivers/net/ethernet/intel/i40e/i40e_client.h
@@ -4,6 +4,8 @@
#ifndef _I40E_CLIENT_H_
#define _I40E_CLIENT_H_
+#include <linux/platform_device.h>
+
#define I40E_CLIENT_STR_LENGTH 10
/* Client interface version should be updated anytime there is a change in the
@@ -80,6 +82,7 @@ struct i40e_params {
/* Structure to hold Lan device info for a client device */
struct i40e_info {
+ struct platform_device platform_dev;
struct i40e_client_version version;
u8 lanmac[6];
struct net_device *netdev;
@@ -97,6 +100,7 @@ struct i40e_info {
struct i40e_qvlist_info *qvlist_info;
struct i40e_params params;
struct i40e_ops *ops;
+ struct i40e_client *client;
u16 msix_count; /* number of msix vectors*/
/* Array down below will be dynamically allocated based on msix_count */
@@ -132,6 +136,10 @@ struct i40e_ops {
struct i40e_client *client,
bool is_vf, u32 vf_id,
u32 flag, u32 valid_flag);
+
+ int (*client_device_register)(struct i40e_info *ldev);
+
+ void (*client_device_unregister)(struct i40e_info *ldev);
};
struct i40e_client_ops {
--
2.21.0
^ permalink raw reply related [flat|nested] 15+ messages in thread
* Re: [net-next 0/3][pull request] Intel Wired LAN ver Updates 2019-07-03
2019-07-04 2:12 [net-next 0/3][pull request] Intel Wired LAN ver Updates 2019-07-03 Jeff Kirsher
` (2 preceding siblings ...)
2019-07-04 2:12 ` [net-next 3/3] i40e: Register RDMA client devices to the virtual platform bus Jeff Kirsher
@ 2019-07-04 12:15 ` Jason Gunthorpe
3 siblings, 0 replies; 15+ messages in thread
From: Jason Gunthorpe @ 2019-07-04 12:15 UTC (permalink / raw)
To: Jeff Kirsher
Cc: davem, dledford, netdev, linux-rdma, nhorman, sassmann,
mustafa.ismail, shiraz.saleem, david.m.ertman
On Wed, Jul 03, 2019 at 07:12:49PM -0700, Jeff Kirsher wrote:
> This series contains updates to i40e an ice drivers only and is required
> for a series of changes being submitted to the RDMA maintainer/tree.
> Vice Versa, the Intel RDMA driver patches could not be applied to
> net-next due to dependencies to the changes currently in the for-next
> branch of the rdma git tree.
RDMA driver code is not to be applied to the net-next tree. We've
learned this causes too many work flow problems.
You must co-ordinate your driver with a shared git tree as Mellanox is
doing, or wait for alternating kernel releases.
I'm not sure what to do with this PR, it is far too late in the cycle
to submit a new driver so most likely net should not go ahead with
this prep work until this new driver model scheme is properly
reviewed.
> The following are changes since commit a51df9f8da43e8bf9e508143630849b7d696e053:
> gve: fix -ENOMEM null check on a page allocation
> and are available in the git repository at:
> git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next-queue 100GbE
And if you expect anything to be shared with rdma you need to produce
pull requests that are based on some sensible -rc tag.
Jason
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [net-next 1/3] ice: Initialize and register platform device to provide RDMA
2019-07-04 2:12 ` [net-next 1/3] ice: Initialize and register platform device to provide RDMA Jeff Kirsher
@ 2019-07-04 12:16 ` Jason Gunthorpe
2019-07-04 12:29 ` Greg KH
0 siblings, 1 reply; 15+ messages in thread
From: Jason Gunthorpe @ 2019-07-04 12:16 UTC (permalink / raw)
To: Jeff Kirsher, Greg KH
Cc: davem, dledford, Tony Nguyen, netdev, linux-rdma, nhorman,
sassmann, poswald, mustafa.ismail, shiraz.saleem, Dave Ertman,
Andrew Bowers
On Wed, Jul 03, 2019 at 07:12:50PM -0700, Jeff Kirsher wrote:
> From: Tony Nguyen <anthony.l.nguyen@intel.com>
>
> The RDMA block does not advertise on the PCI bus or any other bus.
> Thus the ice driver needs to provide access to the RDMA hardware block
> via a virtual bus; utilize the platform bus to provide this access.
>
> This patch initializes the driver to support RDMA as well as creates
> and registers a platform device for the RDMA driver to register to. At
> this point the driver is fully initialized to register a platform
> driver, however, can not yet register as the ops have not been
> implemented.
I think you need Greg's ack on all this driver stuff - particularly
that a platform_device is OK.
Jason
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [net-next 1/3] ice: Initialize and register platform device to provide RDMA
2019-07-04 12:16 ` Jason Gunthorpe
@ 2019-07-04 12:29 ` Greg KH
2019-07-04 12:37 ` Jason Gunthorpe
0 siblings, 1 reply; 15+ messages in thread
From: Greg KH @ 2019-07-04 12:29 UTC (permalink / raw)
To: Jason Gunthorpe
Cc: Jeff Kirsher, davem, dledford, Tony Nguyen, netdev, linux-rdma,
nhorman, sassmann, poswald, mustafa.ismail, shiraz.saleem,
Dave Ertman, Andrew Bowers
On Thu, Jul 04, 2019 at 12:16:41PM +0000, Jason Gunthorpe wrote:
> On Wed, Jul 03, 2019 at 07:12:50PM -0700, Jeff Kirsher wrote:
> > From: Tony Nguyen <anthony.l.nguyen@intel.com>
> >
> > The RDMA block does not advertise on the PCI bus or any other bus.
> > Thus the ice driver needs to provide access to the RDMA hardware block
> > via a virtual bus; utilize the platform bus to provide this access.
> >
> > This patch initializes the driver to support RDMA as well as creates
> > and registers a platform device for the RDMA driver to register to. At
> > this point the driver is fully initialized to register a platform
> > driver, however, can not yet register as the ops have not been
> > implemented.
>
> I think you need Greg's ack on all this driver stuff - particularly
> that a platform_device is OK.
A platform_device is almost NEVER ok.
Don't abuse it, make a real device on a real bus. If you don't have a
real bus and just need to create a device to hang other things off of,
then use the virtual one, that's what it is there for.
thanks,
greg k-h
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [net-next 1/3] ice: Initialize and register platform device to provide RDMA
2019-07-04 12:29 ` Greg KH
@ 2019-07-04 12:37 ` Jason Gunthorpe
2019-07-04 12:42 ` Greg KH
0 siblings, 1 reply; 15+ messages in thread
From: Jason Gunthorpe @ 2019-07-04 12:37 UTC (permalink / raw)
To: Greg KH
Cc: Jeff Kirsher, davem, dledford, Tony Nguyen, netdev, linux-rdma,
nhorman, sassmann, poswald, mustafa.ismail, shiraz.saleem,
Dave Ertman, Andrew Bowers
On Thu, Jul 04, 2019 at 02:29:50PM +0200, Greg KH wrote:
> On Thu, Jul 04, 2019 at 12:16:41PM +0000, Jason Gunthorpe wrote:
> > On Wed, Jul 03, 2019 at 07:12:50PM -0700, Jeff Kirsher wrote:
> > > From: Tony Nguyen <anthony.l.nguyen@intel.com>
> > >
> > > The RDMA block does not advertise on the PCI bus or any other bus.
> > > Thus the ice driver needs to provide access to the RDMA hardware block
> > > via a virtual bus; utilize the platform bus to provide this access.
> > >
> > > This patch initializes the driver to support RDMA as well as creates
> > > and registers a platform device for the RDMA driver to register to. At
> > > this point the driver is fully initialized to register a platform
> > > driver, however, can not yet register as the ops have not been
> > > implemented.
> >
> > I think you need Greg's ack on all this driver stuff - particularly
> > that a platform_device is OK.
>
> A platform_device is almost NEVER ok.
>
> Don't abuse it, make a real device on a real bus. If you don't have a
> real bus and just need to create a device to hang other things off of,
> then use the virtual one, that's what it is there for.
Ideally I'd like to see all the RDMA drivers that connect to ethernet
drivers use some similar scheme.
Should it be some generic virtual bus?
This is for a PCI device that plugs into multiple subsystems in the
kernel, ie it has net driver functionality, rdma functionality, some
even have SCSI functionality
Jason
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [net-next 1/3] ice: Initialize and register platform device to provide RDMA
2019-07-04 12:37 ` Jason Gunthorpe
@ 2019-07-04 12:42 ` Greg KH
2019-07-04 12:48 ` Jason Gunthorpe
0 siblings, 1 reply; 15+ messages in thread
From: Greg KH @ 2019-07-04 12:42 UTC (permalink / raw)
To: Jason Gunthorpe
Cc: Jeff Kirsher, davem, dledford, Tony Nguyen, netdev, linux-rdma,
nhorman, sassmann, poswald, mustafa.ismail, shiraz.saleem,
Dave Ertman, Andrew Bowers
On Thu, Jul 04, 2019 at 12:37:33PM +0000, Jason Gunthorpe wrote:
> On Thu, Jul 04, 2019 at 02:29:50PM +0200, Greg KH wrote:
> > On Thu, Jul 04, 2019 at 12:16:41PM +0000, Jason Gunthorpe wrote:
> > > On Wed, Jul 03, 2019 at 07:12:50PM -0700, Jeff Kirsher wrote:
> > > > From: Tony Nguyen <anthony.l.nguyen@intel.com>
> > > >
> > > > The RDMA block does not advertise on the PCI bus or any other bus.
> > > > Thus the ice driver needs to provide access to the RDMA hardware block
> > > > via a virtual bus; utilize the platform bus to provide this access.
> > > >
> > > > This patch initializes the driver to support RDMA as well as creates
> > > > and registers a platform device for the RDMA driver to register to. At
> > > > this point the driver is fully initialized to register a platform
> > > > driver, however, can not yet register as the ops have not been
> > > > implemented.
> > >
> > > I think you need Greg's ack on all this driver stuff - particularly
> > > that a platform_device is OK.
> >
> > A platform_device is almost NEVER ok.
> >
> > Don't abuse it, make a real device on a real bus. If you don't have a
> > real bus and just need to create a device to hang other things off of,
> > then use the virtual one, that's what it is there for.
>
> Ideally I'd like to see all the RDMA drivers that connect to ethernet
> drivers use some similar scheme.
Why? They should be attached to a "real" device, why make any up?
> Should it be some generic virtual bus?
There is a generic virtual bus today.
> This is for a PCI device that plugs into multiple subsystems in the
> kernel, ie it has net driver functionality, rdma functionality, some
> even have SCSI functionality
Sounds like a MFD device, why aren't you using that functionality
instead?
thanks,
greg k-h
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [net-next 1/3] ice: Initialize and register platform device to provide RDMA
2019-07-04 12:42 ` Greg KH
@ 2019-07-04 12:48 ` Jason Gunthorpe
2019-07-04 13:46 ` Greg KH
0 siblings, 1 reply; 15+ messages in thread
From: Jason Gunthorpe @ 2019-07-04 12:48 UTC (permalink / raw)
To: Greg KH
Cc: Jeff Kirsher, davem, dledford, Tony Nguyen, netdev, linux-rdma,
nhorman, sassmann, poswald, mustafa.ismail, shiraz.saleem,
Dave Ertman, Andrew Bowers
On Thu, Jul 04, 2019 at 02:42:47PM +0200, Greg KH wrote:
> On Thu, Jul 04, 2019 at 12:37:33PM +0000, Jason Gunthorpe wrote:
> > On Thu, Jul 04, 2019 at 02:29:50PM +0200, Greg KH wrote:
> > > On Thu, Jul 04, 2019 at 12:16:41PM +0000, Jason Gunthorpe wrote:
> > > > On Wed, Jul 03, 2019 at 07:12:50PM -0700, Jeff Kirsher wrote:
> > > > > From: Tony Nguyen <anthony.l.nguyen@intel.com>
> > > > >
> > > > > The RDMA block does not advertise on the PCI bus or any other bus.
> > > > > Thus the ice driver needs to provide access to the RDMA hardware block
> > > > > via a virtual bus; utilize the platform bus to provide this access.
> > > > >
> > > > > This patch initializes the driver to support RDMA as well as creates
> > > > > and registers a platform device for the RDMA driver to register to. At
> > > > > this point the driver is fully initialized to register a platform
> > > > > driver, however, can not yet register as the ops have not been
> > > > > implemented.
> > > >
> > > > I think you need Greg's ack on all this driver stuff - particularly
> > > > that a platform_device is OK.
> > >
> > > A platform_device is almost NEVER ok.
> > >
> > > Don't abuse it, make a real device on a real bus. If you don't have a
> > > real bus and just need to create a device to hang other things off of,
> > > then use the virtual one, that's what it is there for.
> >
> > Ideally I'd like to see all the RDMA drivers that connect to ethernet
> > drivers use some similar scheme.
>
> Why? They should be attached to a "real" device, why make any up?
? A "real" device, like struct pci_device, can only bind to one
driver. How can we bind it concurrently to net, rdma, scsi, etc?
> > This is for a PCI device that plugs into multiple subsystems in the
> > kernel, ie it has net driver functionality, rdma functionality, some
> > even have SCSI functionality
>
> Sounds like a MFD device, why aren't you using that functionality
> instead?
This was also my advice, but in another email Jeff says:
MFD architecture was also considered, and we selected the simpler
platform model. Supporting a MFD architecture would require an
additional MFD core driver, individual platform netdev, RDMA function
drivers, and stripping a large portion of the netdev drivers into
MFD core. The sub-devices registered by MFD core for function
drivers are indeed platform devices.
Thanks,
Jason
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [net-next 1/3] ice: Initialize and register platform device to provide RDMA
2019-07-04 12:48 ` Jason Gunthorpe
@ 2019-07-04 13:46 ` Greg KH
2019-07-04 13:53 ` Jason Gunthorpe
2019-07-05 16:33 ` Saleem, Shiraz
0 siblings, 2 replies; 15+ messages in thread
From: Greg KH @ 2019-07-04 13:46 UTC (permalink / raw)
To: Jason Gunthorpe
Cc: Jeff Kirsher, davem, dledford, Tony Nguyen, netdev, linux-rdma,
nhorman, sassmann, poswald, mustafa.ismail, shiraz.saleem,
Dave Ertman, Andrew Bowers
On Thu, Jul 04, 2019 at 12:48:29PM +0000, Jason Gunthorpe wrote:
> On Thu, Jul 04, 2019 at 02:42:47PM +0200, Greg KH wrote:
> > On Thu, Jul 04, 2019 at 12:37:33PM +0000, Jason Gunthorpe wrote:
> > > On Thu, Jul 04, 2019 at 02:29:50PM +0200, Greg KH wrote:
> > > > On Thu, Jul 04, 2019 at 12:16:41PM +0000, Jason Gunthorpe wrote:
> > > > > On Wed, Jul 03, 2019 at 07:12:50PM -0700, Jeff Kirsher wrote:
> > > > > > From: Tony Nguyen <anthony.l.nguyen@intel.com>
> > > > > >
> > > > > > The RDMA block does not advertise on the PCI bus or any other bus.
> > > > > > Thus the ice driver needs to provide access to the RDMA hardware block
> > > > > > via a virtual bus; utilize the platform bus to provide this access.
> > > > > >
> > > > > > This patch initializes the driver to support RDMA as well as creates
> > > > > > and registers a platform device for the RDMA driver to register to. At
> > > > > > this point the driver is fully initialized to register a platform
> > > > > > driver, however, can not yet register as the ops have not been
> > > > > > implemented.
> > > > >
> > > > > I think you need Greg's ack on all this driver stuff - particularly
> > > > > that a platform_device is OK.
> > > >
> > > > A platform_device is almost NEVER ok.
> > > >
> > > > Don't abuse it, make a real device on a real bus. If you don't have a
> > > > real bus and just need to create a device to hang other things off of,
> > > > then use the virtual one, that's what it is there for.
> > >
> > > Ideally I'd like to see all the RDMA drivers that connect to ethernet
> > > drivers use some similar scheme.
> >
> > Why? They should be attached to a "real" device, why make any up?
>
> ? A "real" device, like struct pci_device, can only bind to one
> driver. How can we bind it concurrently to net, rdma, scsi, etc?
MFD was designed for this very problem.
> > > This is for a PCI device that plugs into multiple subsystems in the
> > > kernel, ie it has net driver functionality, rdma functionality, some
> > > even have SCSI functionality
> >
> > Sounds like a MFD device, why aren't you using that functionality
> > instead?
>
> This was also my advice, but in another email Jeff says:
>
> MFD architecture was also considered, and we selected the simpler
> platform model. Supporting a MFD architecture would require an
> additional MFD core driver, individual platform netdev, RDMA function
> drivers, and stripping a large portion of the netdev drivers into
> MFD core. The sub-devices registered by MFD core for function
> drivers are indeed platform devices.
So, "mfd is too hard, let's abuse a platform device" is ok?
People have been wanting to do MFD drivers for PCI devices for a long
time, it's about time someone actually did the work for it, I bet it
will not be all that complex if tiny embedded drivers can do it :)
thanks,
greg k-h
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [net-next 1/3] ice: Initialize and register platform device to provide RDMA
2019-07-04 13:46 ` Greg KH
@ 2019-07-04 13:53 ` Jason Gunthorpe
2019-07-05 16:33 ` Saleem, Shiraz
1 sibling, 0 replies; 15+ messages in thread
From: Jason Gunthorpe @ 2019-07-04 13:53 UTC (permalink / raw)
To: Greg KH
Cc: Jeff Kirsher, davem, dledford, Tony Nguyen, netdev, linux-rdma,
nhorman, sassmann, poswald, mustafa.ismail, shiraz.saleem,
Dave Ertman, Andrew Bowers
On Thu, Jul 04, 2019 at 03:46:12PM +0200, Greg KH wrote:
> On Thu, Jul 04, 2019 at 12:48:29PM +0000, Jason Gunthorpe wrote:
> > On Thu, Jul 04, 2019 at 02:42:47PM +0200, Greg KH wrote:
> > > On Thu, Jul 04, 2019 at 12:37:33PM +0000, Jason Gunthorpe wrote:
> > > > On Thu, Jul 04, 2019 at 02:29:50PM +0200, Greg KH wrote:
> > > > > On Thu, Jul 04, 2019 at 12:16:41PM +0000, Jason Gunthorpe wrote:
> > > > > > On Wed, Jul 03, 2019 at 07:12:50PM -0700, Jeff Kirsher wrote:
> > > > > > > From: Tony Nguyen <anthony.l.nguyen@intel.com>
> > > > > > >
> > > > > > > The RDMA block does not advertise on the PCI bus or any other bus.
> > > > > > > Thus the ice driver needs to provide access to the RDMA hardware block
> > > > > > > via a virtual bus; utilize the platform bus to provide this access.
> > > > > > >
> > > > > > > This patch initializes the driver to support RDMA as well as creates
> > > > > > > and registers a platform device for the RDMA driver to register to. At
> > > > > > > this point the driver is fully initialized to register a platform
> > > > > > > driver, however, can not yet register as the ops have not been
> > > > > > > implemented.
> > > > > >
> > > > > > I think you need Greg's ack on all this driver stuff - particularly
> > > > > > that a platform_device is OK.
> > > > >
> > > > > A platform_device is almost NEVER ok.
> > > > >
> > > > > Don't abuse it, make a real device on a real bus. If you don't have a
> > > > > real bus and just need to create a device to hang other things off of,
> > > > > then use the virtual one, that's what it is there for.
> > > >
> > > > Ideally I'd like to see all the RDMA drivers that connect to ethernet
> > > > drivers use some similar scheme.
> > >
> > > Why? They should be attached to a "real" device, why make any up?
> >
> > ? A "real" device, like struct pci_device, can only bind to one
> > driver. How can we bind it concurrently to net, rdma, scsi, etc?
>
> MFD was designed for this very problem.
>
> > > > This is for a PCI device that plugs into multiple subsystems in the
> > > > kernel, ie it has net driver functionality, rdma functionality, some
> > > > even have SCSI functionality
> > >
> > > Sounds like a MFD device, why aren't you using that functionality
> > > instead?
> >
> > This was also my advice, but in another email Jeff says:
> >
> > MFD architecture was also considered, and we selected the simpler
> > platform model. Supporting a MFD architecture would require an
> > additional MFD core driver, individual platform netdev, RDMA function
> > drivers, and stripping a large portion of the netdev drivers into
> > MFD core. The sub-devices registered by MFD core for function
> > drivers are indeed platform devices.
>
> So, "mfd is too hard, let's abuse a platform device" is ok?
>
> People have been wanting to do MFD drivers for PCI devices for a long
> time, it's about time someone actually did the work for it, I bet it
> will not be all that complex if tiny embedded drivers can do it :)
Okay, sounds like a NAK to me. I'll drop these patches from the RDMA
patchworks and Jeff can work through the MFD stuff first.
Jason
^ permalink raw reply [flat|nested] 15+ messages in thread
* RE: [net-next 1/3] ice: Initialize and register platform device to provide RDMA
2019-07-04 13:46 ` Greg KH
2019-07-04 13:53 ` Jason Gunthorpe
@ 2019-07-05 16:33 ` Saleem, Shiraz
2019-07-06 8:25 ` Greg KH
1 sibling, 1 reply; 15+ messages in thread
From: Saleem, Shiraz @ 2019-07-05 16:33 UTC (permalink / raw)
To: Greg KH, Jason Gunthorpe
Cc: Kirsher, Jeffrey T, davem, dledford, Nguyen, Anthony L, netdev,
linux-rdma, nhorman, sassmann, poswald, Ismail, Mustafa, Ertman,
David M, Bowers, AndrewX
> Subject: Re: [net-next 1/3] ice: Initialize and register platform device to provide
> RDMA
>
> On Thu, Jul 04, 2019 at 12:48:29PM +0000, Jason Gunthorpe wrote:
> > On Thu, Jul 04, 2019 at 02:42:47PM +0200, Greg KH wrote:
> > > On Thu, Jul 04, 2019 at 12:37:33PM +0000, Jason Gunthorpe wrote:
> > > > On Thu, Jul 04, 2019 at 02:29:50PM +0200, Greg KH wrote:
> > > > > On Thu, Jul 04, 2019 at 12:16:41PM +0000, Jason Gunthorpe wrote:
> > > > > > On Wed, Jul 03, 2019 at 07:12:50PM -0700, Jeff Kirsher wrote:
> > > > > > > From: Tony Nguyen <anthony.l.nguyen@intel.com>
> > > > > > >
> > > > > > > The RDMA block does not advertise on the PCI bus or any other bus.
> > > > > > > Thus the ice driver needs to provide access to the RDMA
> > > > > > > hardware block via a virtual bus; utilize the platform bus to provide this
> access.
> > > > > > >
> > > > > > > This patch initializes the driver to support RDMA as well as
> > > > > > > creates and registers a platform device for the RDMA driver
> > > > > > > to register to. At this point the driver is fully
> > > > > > > initialized to register a platform driver, however, can not
> > > > > > > yet register as the ops have not been implemented.
> > > > > >
> > > > > > I think you need Greg's ack on all this driver stuff -
> > > > > > particularly that a platform_device is OK.
> > > > >
> > > > > A platform_device is almost NEVER ok.
> > > > >
> > > > > Don't abuse it, make a real device on a real bus. If you don't
> > > > > have a real bus and just need to create a device to hang other
> > > > > things off of, then use the virtual one, that's what it is there for.
> > > >
> > > > Ideally I'd like to see all the RDMA drivers that connect to
> > > > ethernet drivers use some similar scheme.
> > >
> > > Why? They should be attached to a "real" device, why make any up?
> >
> > ? A "real" device, like struct pci_device, can only bind to one
> > driver. How can we bind it concurrently to net, rdma, scsi, etc?
>
> MFD was designed for this very problem.
>
> > > > This is for a PCI device that plugs into multiple subsystems in
> > > > the kernel, ie it has net driver functionality, rdma
> > > > functionality, some even have SCSI functionality
> > >
> > > Sounds like a MFD device, why aren't you using that functionality
> > > instead?
> >
> > This was also my advice, but in another email Jeff says:
> >
> > MFD architecture was also considered, and we selected the simpler
> > platform model. Supporting a MFD architecture would require an
> > additional MFD core driver, individual platform netdev, RDMA function
> > drivers, and stripping a large portion of the netdev drivers into
> > MFD core. The sub-devices registered by MFD core for function
> > drivers are indeed platform devices.
>
> So, "mfd is too hard, let's abuse a platform device" is ok?
>
> People have been wanting to do MFD drivers for PCI devices for a long time, it's
> about time someone actually did the work for it, I bet it will not be all that complex
> if tiny embedded drivers can do it :)
>
Hi Greg - Thanks for your feedback!
We currently have 2 PCI function netdev drivers in the kernel (i40e & ice) that support devices (x722 & e810)
which are RDMA capable. Our objective is to add a single unified RDMA driver
(as this a subsystem specific requirement) which needs to access HW resources from the
netdev PF drivers. Attaching platform devices from the netdev drivers to the platform bus
and having a single RDMA platform driver bind to them and access these resources seemed
like a simple approach to realize our objective. But seems like attaching platform devices is
wrong. I would like to understand why.
Are platform sub devices only to be added from an MFD core driver? I am also wondering if MFD arch.
would allow for realizing a single RDMA driver and whether we need an MFD core driver for
each device, x722 & e810 or whether it can be a single driver.
Shiraz
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [net-next 1/3] ice: Initialize and register platform device to provide RDMA
2019-07-05 16:33 ` Saleem, Shiraz
@ 2019-07-06 8:25 ` Greg KH
2019-07-06 16:03 ` Saleem, Shiraz
0 siblings, 1 reply; 15+ messages in thread
From: Greg KH @ 2019-07-06 8:25 UTC (permalink / raw)
To: Saleem, Shiraz
Cc: Jason Gunthorpe, Kirsher, Jeffrey T, davem, dledford, Nguyen,
Anthony L, netdev, linux-rdma, nhorman, sassmann, poswald,
Ismail, Mustafa, Ertman, David M, Bowers, AndrewX
On Fri, Jul 05, 2019 at 04:33:07PM +0000, Saleem, Shiraz wrote:
> > Subject: Re: [net-next 1/3] ice: Initialize and register platform device to provide
> > RDMA
> >
> > On Thu, Jul 04, 2019 at 12:48:29PM +0000, Jason Gunthorpe wrote:
> > > On Thu, Jul 04, 2019 at 02:42:47PM +0200, Greg KH wrote:
> > > > On Thu, Jul 04, 2019 at 12:37:33PM +0000, Jason Gunthorpe wrote:
> > > > > On Thu, Jul 04, 2019 at 02:29:50PM +0200, Greg KH wrote:
> > > > > > On Thu, Jul 04, 2019 at 12:16:41PM +0000, Jason Gunthorpe wrote:
> > > > > > > On Wed, Jul 03, 2019 at 07:12:50PM -0700, Jeff Kirsher wrote:
> > > > > > > > From: Tony Nguyen <anthony.l.nguyen@intel.com>
> > > > > > > >
> > > > > > > > The RDMA block does not advertise on the PCI bus or any other bus.
> > > > > > > > Thus the ice driver needs to provide access to the RDMA
> > > > > > > > hardware block via a virtual bus; utilize the platform bus to provide this
> > access.
> > > > > > > >
> > > > > > > > This patch initializes the driver to support RDMA as well as
> > > > > > > > creates and registers a platform device for the RDMA driver
> > > > > > > > to register to. At this point the driver is fully
> > > > > > > > initialized to register a platform driver, however, can not
> > > > > > > > yet register as the ops have not been implemented.
> > > > > > >
> > > > > > > I think you need Greg's ack on all this driver stuff -
> > > > > > > particularly that a platform_device is OK.
> > > > > >
> > > > > > A platform_device is almost NEVER ok.
> > > > > >
> > > > > > Don't abuse it, make a real device on a real bus. If you don't
> > > > > > have a real bus and just need to create a device to hang other
> > > > > > things off of, then use the virtual one, that's what it is there for.
> > > > >
> > > > > Ideally I'd like to see all the RDMA drivers that connect to
> > > > > ethernet drivers use some similar scheme.
> > > >
> > > > Why? They should be attached to a "real" device, why make any up?
> > >
> > > ? A "real" device, like struct pci_device, can only bind to one
> > > driver. How can we bind it concurrently to net, rdma, scsi, etc?
> >
> > MFD was designed for this very problem.
> >
> > > > > This is for a PCI device that plugs into multiple subsystems in
> > > > > the kernel, ie it has net driver functionality, rdma
> > > > > functionality, some even have SCSI functionality
> > > >
> > > > Sounds like a MFD device, why aren't you using that functionality
> > > > instead?
> > >
> > > This was also my advice, but in another email Jeff says:
> > >
> > > MFD architecture was also considered, and we selected the simpler
> > > platform model. Supporting a MFD architecture would require an
> > > additional MFD core driver, individual platform netdev, RDMA function
> > > drivers, and stripping a large portion of the netdev drivers into
> > > MFD core. The sub-devices registered by MFD core for function
> > > drivers are indeed platform devices.
> >
> > So, "mfd is too hard, let's abuse a platform device" is ok?
> >
> > People have been wanting to do MFD drivers for PCI devices for a long time, it's
> > about time someone actually did the work for it, I bet it will not be all that complex
> > if tiny embedded drivers can do it :)
> >
> Hi Greg - Thanks for your feedback!
>
> We currently have 2 PCI function netdev drivers in the kernel (i40e & ice) that support devices (x722 & e810)
> which are RDMA capable. Our objective is to add a single unified RDMA driver
> (as this a subsystem specific requirement) which needs to access HW resources from the
> netdev PF drivers. Attaching platform devices from the netdev drivers to the platform bus
> and having a single RDMA platform driver bind to them and access these resources seemed
> like a simple approach to realize our objective. But seems like attaching platform devices is
> wrong. I would like to understand why.
Because that is NOT what a platform device is for.
It was originally created for those types of devices that live on the
"platform" bus, i.e. things that are hard-wired and you just "know" are
in your system because they are located at specific locations. We used
to generate them from board files, and then when we got DT, we create
them from the resources that DT says where the locations of the devices
are.
They are NOT to be abused and used whenever someone wants to put them
somewhere in the "middle" of the device tree because they feel like they
are easier to use instead of creating a real bus and drivers.
Yes, they do get abused, and I need to sweep the tree again and fix up
all of the places where this has crept back in. But now that I know you
are thinking of doing this, I'll keep saying to NOT do it for your use
case either :)
> Are platform sub devices only to be added from an MFD core driver? I
> am also wondering if MFD arch. would allow for realizing a single
> RDMA driver and whether we need an MFD core driver for each device,
> x722 & e810 or whether it can be a single driver.
I do not know the details of how MFD works, please ask those developers
for specifics. If MFD doesn't work, then create a tiny virtual bus and
make sub-devices out of that. If you need a "generic" way to do this
for PCI devices, then please create that as you are not the only one
that keeps wanting this, as for some reason PCI hardware vendors don't
like dividing up their devices in ways that would have made it much
simpler to create individual devices (probably saves some gates and
firmware complexity on the device).
thanks,
greg k-h
^ permalink raw reply [flat|nested] 15+ messages in thread
* RE: [net-next 1/3] ice: Initialize and register platform device to provide RDMA
2019-07-06 8:25 ` Greg KH
@ 2019-07-06 16:03 ` Saleem, Shiraz
0 siblings, 0 replies; 15+ messages in thread
From: Saleem, Shiraz @ 2019-07-06 16:03 UTC (permalink / raw)
To: Greg KH
Cc: Jason Gunthorpe, Kirsher, Jeffrey T, davem, dledford, Nguyen,
Anthony L, netdev, linux-rdma, nhorman, sassmann, poswald,
Ismail, Mustafa, Ertman, David M, Bowers, AndrewX
> Subject: Re: [net-next 1/3] ice: Initialize and register platform device to provide
> RDMA
>
> On Fri, Jul 05, 2019 at 04:33:07PM +0000, Saleem, Shiraz wrote:
> > > Subject: Re: [net-next 1/3] ice: Initialize and register platform
> > > device to provide RDMA
> > >
> > > On Thu, Jul 04, 2019 at 12:48:29PM +0000, Jason Gunthorpe wrote:
> > > > On Thu, Jul 04, 2019 at 02:42:47PM +0200, Greg KH wrote:
> > > > > On Thu, Jul 04, 2019 at 12:37:33PM +0000, Jason Gunthorpe wrote:
> > > > > > On Thu, Jul 04, 2019 at 02:29:50PM +0200, Greg KH wrote:
> > > > > > > On Thu, Jul 04, 2019 at 12:16:41PM +0000, Jason Gunthorpe wrote:
> > > > > > > > On Wed, Jul 03, 2019 at 07:12:50PM -0700, Jeff Kirsher wrote:
> > > > > > > > > From: Tony Nguyen <anthony.l.nguyen@intel.com>
> > > > > > > > >
> > > > > > > > > The RDMA block does not advertise on the PCI bus or any other
> bus.
> > > > > > > > > Thus the ice driver needs to provide access to the RDMA
> > > > > > > > > hardware block via a virtual bus; utilize the platform
> > > > > > > > > bus to provide this
> > > access.
> > > > > > > > >
> > > > > > > > > This patch initializes the driver to support RDMA as
> > > > > > > > > well as creates and registers a platform device for the
> > > > > > > > > RDMA driver to register to. At this point the driver is
> > > > > > > > > fully initialized to register a platform driver,
> > > > > > > > > however, can not yet register as the ops have not been
> implemented.
> > > > > > > >
> > > > > > > > I think you need Greg's ack on all this driver stuff -
> > > > > > > > particularly that a platform_device is OK.
> > > > > > >
> > > > > > > A platform_device is almost NEVER ok.
> > > > > > >
> > > > > > > Don't abuse it, make a real device on a real bus. If you
> > > > > > > don't have a real bus and just need to create a device to
> > > > > > > hang other things off of, then use the virtual one, that's what it is there
> for.
> > > > > >
> > > > > > Ideally I'd like to see all the RDMA drivers that connect to
> > > > > > ethernet drivers use some similar scheme.
> > > > >
> > > > > Why? They should be attached to a "real" device, why make any up?
> > > >
> > > > ? A "real" device, like struct pci_device, can only bind to one
> > > > driver. How can we bind it concurrently to net, rdma, scsi, etc?
> > >
> > > MFD was designed for this very problem.
> > >
> > > > > > This is for a PCI device that plugs into multiple subsystems
> > > > > > in the kernel, ie it has net driver functionality, rdma
> > > > > > functionality, some even have SCSI functionality
> > > > >
> > > > > Sounds like a MFD device, why aren't you using that
> > > > > functionality instead?
> > > >
> > > > This was also my advice, but in another email Jeff says:
> > > >
> > > > MFD architecture was also considered, and we selected the simpler
> > > > platform model. Supporting a MFD architecture would require an
> > > > additional MFD core driver, individual platform netdev, RDMA function
> > > > drivers, and stripping a large portion of the netdev drivers into
> > > > MFD core. The sub-devices registered by MFD core for function
> > > > drivers are indeed platform devices.
> > >
> > > So, "mfd is too hard, let's abuse a platform device" is ok?
> > >
> > > People have been wanting to do MFD drivers for PCI devices for a
> > > long time, it's about time someone actually did the work for it, I
> > > bet it will not be all that complex if tiny embedded drivers can do
> > > it :)
> > >
> > Hi Greg - Thanks for your feedback!
> >
> > We currently have 2 PCI function netdev drivers in the kernel (i40e &
> > ice) that support devices (x722 & e810) which are RDMA capable. Our
> > objective is to add a single unified RDMA driver (as this a subsystem
> > specific requirement) which needs to access HW resources from the
> > netdev PF drivers. Attaching platform devices from the netdev drivers
> > to the platform bus and having a single RDMA platform driver bind to
> > them and access these resources seemed like a simple approach to realize our
> objective. But seems like attaching platform devices is wrong. I would like to
> understand why.
>
> Because that is NOT what a platform device is for.
>
> It was originally created for those types of devices that live on the "platform" bus,
> i.e. things that are hard-wired and you just "know" are in your system because they
> are located at specific locations. We used to generate them from board files, and
> then when we got DT, we create them from the resources that DT says where the
> locations of the devices are.
>
> They are NOT to be abused and used whenever someone wants to put them
> somewhere in the "middle" of the device tree because they feel like they are easier
> to use instead of creating a real bus and drivers.
>
> Yes, they do get abused, and I need to sweep the tree again and fix up all of the
> places where this has crept back in. But now that I know you are thinking of doing
> this, I'll keep saying to NOT do it for your use case either :)
Thanks Greg for the explanation.
And yes, we went by some example usages in the tree.
>
> > Are platform sub devices only to be added from an MFD core driver? I
> > am also wondering if MFD arch. would allow for realizing a single
> > RDMA driver and whether we need an MFD core driver for each device,
> > x722 & e810 or whether it can be a single driver.
>
> I do not know the details of how MFD works, please ask those developers for
> specifics. If MFD doesn't work, then create a tiny virtual bus and make sub-
> devices out of that. If you need a "generic" way to do this for PCI devices, then
> please create that as you are not the only one that keeps wanting this, as for some
> reason PCI hardware vendors don't like dividing up their devices in ways that
> would have made it much simpler to create individual devices (probably saves
> some gates and firmware complexity on the device).
>
Thank you for laying out options. We will review internally and get back.
Shiraz
^ permalink raw reply [flat|nested] 15+ messages in thread
end of thread, other threads:[~2019-07-06 16:03 UTC | newest]
Thread overview: 15+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-07-04 2:12 [net-next 0/3][pull request] Intel Wired LAN ver Updates 2019-07-03 Jeff Kirsher
2019-07-04 2:12 ` [net-next 1/3] ice: Initialize and register platform device to provide RDMA Jeff Kirsher
2019-07-04 12:16 ` Jason Gunthorpe
2019-07-04 12:29 ` Greg KH
2019-07-04 12:37 ` Jason Gunthorpe
2019-07-04 12:42 ` Greg KH
2019-07-04 12:48 ` Jason Gunthorpe
2019-07-04 13:46 ` Greg KH
2019-07-04 13:53 ` Jason Gunthorpe
2019-07-05 16:33 ` Saleem, Shiraz
2019-07-06 8:25 ` Greg KH
2019-07-06 16:03 ` Saleem, Shiraz
2019-07-04 2:12 ` [net-next 2/3] ice: Implement peer communications Jeff Kirsher
2019-07-04 2:12 ` [net-next 3/3] i40e: Register RDMA client devices to the virtual platform bus Jeff Kirsher
2019-07-04 12:15 ` [net-next 0/3][pull request] Intel Wired LAN ver Updates 2019-07-03 Jason Gunthorpe
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).