Introduce new options on devlink reload API to enable the user to select
the reload action required and constrains limits on these actions that he
may want to ensure. Complete support for reload actions in mlx5.
The following reload actions are supported:
driver_reinit: driver entities re-initialization, applying devlink-param
and devlink-resource values.
fw_activate: firmware activate.
The uAPI is backward compatible, if the reload action option is omitted
from the reload command, the driver reinit action will be used.
Note that when required to do firmware activation some drivers may need
to reload the driver. On the other hand some drivers may need to reset
the firmware to reinitialize the driver entities. Therefore, the devlink
reload command returns the actions which were actually performed.
By default reload actions are not limited and driver implementation may
include reset or downtime as needed to perform the actions.
However, if reload limit is selected, the driver should perform only if
it can do it while keeping the limit constrains.
Reload limit added:
no_reset: No reset allowed, no down time allowed, no link flap and no
configuration is lost.
Each driver which supports devlink reload command should expose the
reload actions and limits supported.
Add reload stats to hold the history per reload action per limit.
For example, the number of times fw_activate has been done on this
device since the driver module was added or if the firmware activation
was done with or without reset.
Change log: Preceding to this version were 5 RFC versions. The changes
applied according to comments mainly from Jakub and Jiri on RFC API
patches are listed in each patch.
Patch 1 changes devlink_reload_supported() param type to enable using
it before allocating devlink.
Patch 2-3 add the new API reload action and reload limit options to
devlink reload.
Patch 4-5 add reload stats and remote reload stats. These stats are
exposed through devlink dev get.
Patches 6-11 add support on mlx5 for devlink reload action fw_activate
and handle the firmware reset events.
Patches 12-13 add devlink enable remote dev reset parameter and use it
in mlx5.
Patches 14-15 mlx5 add devlink reload limit no_reset support for
fw_activate reload action.
Patch 16 adds documentation file devlink-reload.rst
Moshe Shemesh (16):
devlink: Change devlink_reload_supported() param type
devlink: Add reload action option to devlink reload command
devlink: Add devlink reload limit option
devlink: Add reload stats
devlink: Add remote reload stats
net/mlx5: Add functions to set/query MFRL register
net/mlx5: Set cap for pci sync for fw update event
net/mlx5: Handle sync reset request event
net/mlx5: Handle sync reset now event
net/mlx5: Handle sync reset abort event
net/mlx5: Add support for devlink reload action fw activate
devlink: Add enable_remote_dev_reset generic parameter
net/mlx5: Add devlink param enable_remote_dev_reset support
net/mlx5: Add support for fw live patch event
net/mlx5: Add support for devlink reload limit no reset
devlink: Add Documentation/networking/devlink/devlink-reload.rst
.../networking/devlink/devlink-params.rst | 6 +
.../networking/devlink/devlink-reload.rst | 81 +++
Documentation/networking/devlink/index.rst | 1 +
drivers/net/ethernet/mellanox/mlx4/main.c | 9 +-
.../net/ethernet/mellanox/mlx5/core/Makefile | 2 +-
.../net/ethernet/mellanox/mlx5/core/devlink.c | 114 ++++-
.../mellanox/mlx5/core/diag/fw_tracer.c | 52 ++
.../mellanox/mlx5/core/diag/fw_tracer.h | 1 +
.../ethernet/mellanox/mlx5/core/fw_reset.c | 463 ++++++++++++++++++
.../ethernet/mellanox/mlx5/core/fw_reset.h | 21 +
.../net/ethernet/mellanox/mlx5/core/health.c | 35 +-
.../net/ethernet/mellanox/mlx5/core/main.c | 16 +
.../ethernet/mellanox/mlx5/core/mlx5_core.h | 2 +
drivers/net/ethernet/mellanox/mlxsw/core.c | 13 +-
drivers/net/netdevsim/dev.c | 8 +-
include/linux/mlx5/device.h | 1 +
include/linux/mlx5/driver.h | 2 +
include/net/devlink.h | 21 +-
include/uapi/linux/devlink.h | 42 ++
net/core/devlink.c | 318 +++++++++++-
20 files changed, 1164 insertions(+), 44 deletions(-)
create mode 100644 Documentation/networking/devlink/devlink-reload.rst
create mode 100644 drivers/net/ethernet/mellanox/mlx5/core/fw_reset.c
create mode 100644 drivers/net/ethernet/mellanox/mlx5/core/fw_reset.h
--
2.18.2
The enable_remote_dev_reset devlink param flags that the host admin
allows resets by other hosts. In case it is cleared mlx5 host PF driver
will send NACK on pci sync for firmware update reset request and the
command will fail.
By default enable_remote_dev_reset parameter is true, so pci sync for
firmware update reset is enabled.
Signed-off-by: Moshe Shemesh <[email protected]>
Reviewed-by: Saeed Mahameed <[email protected]>
---
RFCv1 -> RFCv2:
- Have MLX5_FW_RESET_FLAGS_NACK_RESET_REQUEST instead of
MLX5_HEALTH_RESET_FLAGS_NACK_RESET_REQUEST
---
.../net/ethernet/mellanox/mlx5/core/devlink.c | 21 ++++++++++++++
.../ethernet/mellanox/mlx5/core/fw_reset.c | 29 +++++++++++++++++++
.../ethernet/mellanox/mlx5/core/fw_reset.h | 2 ++
3 files changed, 52 insertions(+)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/devlink.c b/drivers/net/ethernet/mellanox/mlx5/core/devlink.c
index 13153454cc05..7b304227ad57 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/devlink.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/devlink.c
@@ -278,6 +278,24 @@ static int mlx5_devlink_large_group_num_validate(struct devlink *devlink, u32 id
}
#endif
+static int mlx5_devlink_enable_remote_dev_reset_set(struct devlink *devlink, u32 id,
+ struct devlink_param_gset_ctx *ctx)
+{
+ struct mlx5_core_dev *dev = devlink_priv(devlink);
+
+ mlx5_fw_reset_enable_remote_dev_reset_set(dev, ctx->val.vbool);
+ return 0;
+}
+
+static int mlx5_devlink_enable_remote_dev_reset_get(struct devlink *devlink, u32 id,
+ struct devlink_param_gset_ctx *ctx)
+{
+ struct mlx5_core_dev *dev = devlink_priv(devlink);
+
+ ctx->val.vbool = mlx5_fw_reset_enable_remote_dev_reset_get(dev);
+ return 0;
+}
+
static const struct devlink_param mlx5_devlink_params[] = {
DEVLINK_PARAM_DRIVER(MLX5_DEVLINK_PARAM_ID_FLOW_STEERING_MODE,
"flow_steering_mode", DEVLINK_PARAM_TYPE_STRING,
@@ -293,6 +311,9 @@ static const struct devlink_param mlx5_devlink_params[] = {
NULL, NULL,
mlx5_devlink_large_group_num_validate),
#endif
+ DEVLINK_PARAM_GENERIC(ENABLE_REMOTE_DEV_RESET, BIT(DEVLINK_PARAM_CMODE_RUNTIME),
+ mlx5_devlink_enable_remote_dev_reset_get,
+ mlx5_devlink_enable_remote_dev_reset_set, NULL),
};
static void mlx5_devlink_set_params_init_values(struct devlink *devlink)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fw_reset.c b/drivers/net/ethernet/mellanox/mlx5/core/fw_reset.c
index f5ffb6fc55c3..b2aaff8d4fcd 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fw_reset.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fw_reset.c
@@ -5,6 +5,7 @@
enum {
MLX5_FW_RESET_FLAGS_RESET_REQUESTED,
+ MLX5_FW_RESET_FLAGS_NACK_RESET_REQUEST,
MLX5_FW_RESET_FLAGS_PENDING_COMP
};
@@ -22,6 +23,23 @@ struct mlx5_fw_reset {
int ret;
};
+void mlx5_fw_reset_enable_remote_dev_reset_set(struct mlx5_core_dev *dev, bool enable)
+{
+ struct mlx5_fw_reset *fw_reset = dev->priv.fw_reset;
+
+ if (enable)
+ clear_bit(MLX5_FW_RESET_FLAGS_NACK_RESET_REQUEST, &fw_reset->reset_flags);
+ else
+ set_bit(MLX5_FW_RESET_FLAGS_NACK_RESET_REQUEST, &fw_reset->reset_flags);
+}
+
+bool mlx5_fw_reset_enable_remote_dev_reset_get(struct mlx5_core_dev *dev)
+{
+ struct mlx5_fw_reset *fw_reset = dev->priv.fw_reset;
+
+ return !test_bit(MLX5_FW_RESET_FLAGS_NACK_RESET_REQUEST, &fw_reset->reset_flags);
+}
+
static int mlx5_reg_mfrl_set(struct mlx5_core_dev *dev, u8 reset_level,
u8 reset_type_sel, u8 sync_resp, bool sync_start)
{
@@ -160,6 +178,11 @@ static int mlx5_fw_reset_set_reset_sync_ack(struct mlx5_core_dev *dev)
return mlx5_reg_mfrl_set(dev, MLX5_MFRL_REG_RESET_LEVEL3, 0, 1, false);
}
+static int mlx5_fw_reset_set_reset_sync_nack(struct mlx5_core_dev *dev)
+{
+ return mlx5_reg_mfrl_set(dev, MLX5_MFRL_REG_RESET_LEVEL3, 0, 2, false);
+}
+
static void mlx5_sync_reset_set_reset_requested(struct mlx5_core_dev *dev)
{
struct mlx5_fw_reset *fw_reset = dev->priv.fw_reset;
@@ -176,6 +199,12 @@ static void mlx5_sync_reset_request_event(struct work_struct *work)
struct mlx5_core_dev *dev = fw_reset->dev;
int err;
+ if (test_bit(MLX5_FW_RESET_FLAGS_NACK_RESET_REQUEST, &fw_reset->reset_flags)) {
+ err = mlx5_fw_reset_set_reset_sync_nack(dev);
+ mlx5_core_warn(dev, "PCI Sync FW Update Reset Nack %s",
+ err ? "Failed" : "Sent");
+ return;
+ }
mlx5_sync_reset_set_reset_requested(dev);
err = mlx5_fw_reset_set_reset_sync_ack(dev);
if (err)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fw_reset.h b/drivers/net/ethernet/mellanox/mlx5/core/fw_reset.h
index e7937447ce1d..7761ee5fc7d0 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fw_reset.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fw_reset.h
@@ -6,6 +6,8 @@
#include "mlx5_core.h"
+void mlx5_fw_reset_enable_remote_dev_reset_set(struct mlx5_core_dev *dev, bool enable);
+bool mlx5_fw_reset_enable_remote_dev_reset_get(struct mlx5_core_dev *dev);
int mlx5_fw_reset_query(struct mlx5_core_dev *dev, u8 *reset_level, u8 *reset_type);
int mlx5_fw_reset_set_reset_sync(struct mlx5_core_dev *dev, u8 reset_type_sel);
int mlx5_fw_reset_set_live_patch(struct mlx5_core_dev *dev);
--
2.18.2
On sync_reset_now event the driver does reload and PCI link toggle to
activate firmware upgrade reset. When the firmware sends this event it
syncs the event on all PFs, so all PFs will do PCI link toggle at once.
To do PCI link toggle, the driver ensures that no other device ID under
the same bridge by checking that all the PF functions under the same PCI
bridge have same device ID. If no other device it uses PCI bridge link
control to turn link down and up.
Signed-off-by: Moshe Shemesh <[email protected]>
Reviewed-by: Saeed Mahameed <[email protected]>
---
.../ethernet/mellanox/mlx5/core/fw_reset.c | 119 ++++++++++++++++++
1 file changed, 119 insertions(+)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fw_reset.c b/drivers/net/ethernet/mellanox/mlx5/core/fw_reset.c
index cd1b4e1b56ba..26caf65e9f5f 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fw_reset.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fw_reset.c
@@ -13,6 +13,7 @@ struct mlx5_fw_reset {
struct workqueue_struct *wq;
struct work_struct reset_request_work;
struct work_struct reset_reload_work;
+ struct work_struct reset_now_work;
unsigned long reset_flags;
struct timer_list timer;
};
@@ -156,6 +157,120 @@ static void mlx5_sync_reset_request_event(struct work_struct *work)
mlx5_core_warn(dev, "PCI Sync FW Update Reset Ack. Device reset is expected.\n");
}
+#define MLX5_PCI_LINK_UP_TIMEOUT 2000
+
+static int mlx5_pci_link_toggle(struct mlx5_core_dev *dev)
+{
+ struct pci_bus *bridge_bus = dev->pdev->bus;
+ struct pci_dev *bridge = bridge_bus->self;
+ u16 reg16, dev_id, sdev_id;
+ unsigned long timeout;
+ struct pci_dev *sdev;
+ int cap, err;
+ u32 reg32;
+
+ /* Check that all functions under the pci bridge are PFs of
+ * this device otherwise fail this function.
+ */
+ err = pci_read_config_word(dev->pdev, PCI_DEVICE_ID, &dev_id);
+ if (err)
+ return err;
+ list_for_each_entry(sdev, &bridge_bus->devices, bus_list) {
+ err = pci_read_config_word(sdev, PCI_DEVICE_ID, &sdev_id);
+ if (err)
+ return err;
+ if (sdev_id != dev_id)
+ return -EPERM;
+ }
+
+ cap = pci_find_capability(bridge, PCI_CAP_ID_EXP);
+ if (!cap)
+ return -EOPNOTSUPP;
+
+ list_for_each_entry(sdev, &bridge_bus->devices, bus_list) {
+ pci_save_state(sdev);
+ pci_cfg_access_lock(sdev);
+ }
+ /* PCI link toggle */
+ err = pci_read_config_word(bridge, cap + PCI_EXP_LNKCTL, ®16);
+ if (err)
+ return err;
+ reg16 |= PCI_EXP_LNKCTL_LD;
+ err = pci_write_config_word(bridge, cap + PCI_EXP_LNKCTL, reg16);
+ if (err)
+ return err;
+ msleep(500);
+ reg16 &= ~PCI_EXP_LNKCTL_LD;
+ err = pci_write_config_word(bridge, cap + PCI_EXP_LNKCTL, reg16);
+ if (err)
+ return err;
+
+ /* Check link */
+ err = pci_read_config_dword(bridge, cap + PCI_EXP_LNKCAP, ®32);
+ if (err)
+ return err;
+ if (!(reg32 & PCI_EXP_LNKCAP_DLLLARC)) {
+ mlx5_core_warn(dev, "No PCI link reporting capability (0x%08x)\n", reg32);
+ msleep(1000);
+ goto restore;
+ }
+
+ timeout = jiffies + msecs_to_jiffies(MLX5_PCI_LINK_UP_TIMEOUT);
+ do {
+ err = pci_read_config_word(bridge, cap + PCI_EXP_LNKSTA, ®16);
+ if (err)
+ return err;
+ if (reg16 & PCI_EXP_LNKSTA_DLLLA)
+ break;
+ msleep(20);
+ } while (!time_after(jiffies, timeout));
+
+ if (reg16 & PCI_EXP_LNKSTA_DLLLA) {
+ mlx5_core_info(dev, "PCI Link up\n");
+ } else {
+ mlx5_core_err(dev, "PCI link not ready (0x%04x) after %d ms\n",
+ reg16, MLX5_PCI_LINK_UP_TIMEOUT);
+ err = -ETIMEDOUT;
+ }
+
+restore:
+ list_for_each_entry(sdev, &bridge_bus->devices, bus_list) {
+ pci_cfg_access_unlock(sdev);
+ pci_restore_state(sdev);
+ }
+
+ return err;
+}
+
+static void mlx5_sync_reset_now_event(struct work_struct *work)
+{
+ struct mlx5_fw_reset *fw_reset = container_of(work, struct mlx5_fw_reset,
+ reset_now_work);
+ struct mlx5_core_dev *dev = fw_reset->dev;
+ int err;
+
+ mlx5_sync_reset_clear_reset_requested(dev, false);
+
+ mlx5_core_warn(dev, "Sync Reset now. Device is going to reset.\n");
+
+ err = mlx5_cmd_fast_teardown_hca(dev);
+ if (err) {
+ mlx5_core_warn(dev, "Fast teardown failed, no reset done, err %d\n", err);
+ goto done;
+ }
+
+ err = mlx5_pci_link_toggle(dev);
+ if (err) {
+ mlx5_core_warn(dev, "mlx5_pci_link_toggle failed, no reset done, err %d\n", err);
+ goto done;
+ }
+
+ mlx5_enter_error_state(dev, true);
+ mlx5_unload_one(dev, false);
+done:
+ mlx5_load_one(dev, false);
+}
+
static void mlx5_sync_reset_events_handle(struct mlx5_fw_reset *fw_reset, struct mlx5_eqe *eqe)
{
struct mlx5_eqe_sync_fw_update *sync_fw_update_eqe;
@@ -167,6 +282,9 @@ static void mlx5_sync_reset_events_handle(struct mlx5_fw_reset *fw_reset, struct
case MLX5_SYNC_RST_STATE_RESET_REQUEST:
queue_work(fw_reset->wq, &fw_reset->reset_request_work);
break;
+ case MLX5_SYNC_RST_STATE_RESET_NOW:
+ queue_work(fw_reset->wq, &fw_reset->reset_now_work);
+ break;
}
}
@@ -216,6 +334,7 @@ int mlx5_fw_reset_init(struct mlx5_core_dev *dev)
INIT_WORK(&fw_reset->reset_request_work, mlx5_sync_reset_request_event);
INIT_WORK(&fw_reset->reset_reload_work, mlx5_sync_reset_reload_work);
+ INIT_WORK(&fw_reset->reset_now_work, mlx5_sync_reset_now_event);
return 0;
}
--
2.18.2
Add support for devlink reload action fw_activate. To activate firmware
image the mlx5 driver resets the firmware and reloads it from flash. If
a new image was stored on flash it will be loaded. Once this reload
command is executed the driver initiates fw sync reset flow, where the
firmware synchronizes all PFs on coming reset and driver reload.
Signed-off-by: Moshe Shemesh <[email protected]>
Reviewed-by: Saeed Mahameed <[email protected]>
---
RFCv4 -> RFCv5:
- Add remote actions stats
- If devlink reload is not supported, show only remote_stats
RFCv3 -> RFCv4:
- Renamed DEVLINK_ATTR_RELOAD_ACTION_CNT to
DEVLINK_ATTR_RELOAD_ACTION_STAT
- Add stats per action per limit level
RFCv2 -> RFCv3:
- Add reload actions counters instead of supported reload actions
(reload actions counters are only for supported action so no need for
both)
RFCv1 -> RFCv2:
- Removed DEVLINK_ATTR_RELOAD_DEFAULT_LEVEL
- Removed DEVLINK_ATTR_RELOAD_LEVELS_INFO
- Have actions instead of levels
---
.../net/ethernet/mellanox/mlx5/core/devlink.c | 59 +++++++++++++++++--
.../ethernet/mellanox/mlx5/core/fw_reset.c | 59 +++++++++++++++++--
.../ethernet/mellanox/mlx5/core/fw_reset.h | 1 +
3 files changed, 108 insertions(+), 11 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/devlink.c b/drivers/net/ethernet/mellanox/mlx5/core/devlink.c
index 47f3dbc80c7c..13153454cc05 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/devlink.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/devlink.c
@@ -4,6 +4,7 @@
#include <devlink.h>
#include "mlx5_core.h"
+#include "fw_reset.h"
#include "fs_core.h"
#include "eswitch.h"
@@ -84,6 +85,32 @@ mlx5_devlink_info_get(struct devlink *devlink, struct devlink_info_req *req,
return 0;
}
+static int mlx5_devlink_reload_fw_activate(struct devlink *devlink, struct netlink_ext_ack *extack)
+{
+ struct mlx5_core_dev *dev = devlink_priv(devlink);
+ u8 reset_level, reset_type, net_port_alive;
+ int err;
+
+ err = mlx5_fw_reset_query(dev, &reset_level, &reset_type);
+ if (err)
+ return err;
+ if (!(reset_level & MLX5_MFRL_REG_RESET_LEVEL3)) {
+ NL_SET_ERR_MSG_MOD(extack, "FW activate requires reboot");
+ return -EINVAL;
+ }
+
+ net_port_alive = !!(reset_type & MLX5_MFRL_REG_RESET_TYPE_NET_PORT_ALIVE);
+ err = mlx5_fw_reset_set_reset_sync(dev, net_port_alive);
+ if (err)
+ goto out;
+
+ err = mlx5_fw_reset_wait_reset_done(dev);
+out:
+ if (err)
+ NL_SET_ERR_MSG_MOD(extack, "FW activate command failed");
+ return err;
+}
+
static int mlx5_devlink_reload_down(struct devlink *devlink, bool netns_change,
enum devlink_reload_action action,
enum devlink_reload_limit limit,
@@ -91,8 +118,17 @@ static int mlx5_devlink_reload_down(struct devlink *devlink, bool netns_change,
{
struct mlx5_core_dev *dev = devlink_priv(devlink);
- mlx5_unload_one(dev, false);
- return 0;
+ switch (action) {
+ case DEVLINK_RELOAD_ACTION_DRIVER_REINIT:
+ mlx5_unload_one(dev, false);
+ return 0;
+ case DEVLINK_RELOAD_ACTION_FW_ACTIVATE:
+ return mlx5_devlink_reload_fw_activate(devlink, extack);
+ default:
+ /* Unsupported action should not get to this function */
+ WARN_ON(1);
+ return -EOPNOTSUPP;
+ }
}
static int mlx5_devlink_reload_up(struct devlink *devlink, enum devlink_reload_action action,
@@ -101,8 +137,20 @@ static int mlx5_devlink_reload_up(struct devlink *devlink, enum devlink_reload_a
{
struct mlx5_core_dev *dev = devlink_priv(devlink);
- *actions_performed = BIT(DEVLINK_RELOAD_ACTION_DRIVER_REINIT);
- return mlx5_load_one(dev, false);
+ *actions_performed = BIT(action);
+ switch (action) {
+ case DEVLINK_RELOAD_ACTION_DRIVER_REINIT:
+ case DEVLINK_RELOAD_ACTION_FW_ACTIVATE:
+ /* On fw_activate action, also driver is reloaded and reinit performed */
+ *actions_performed |= BIT(DEVLINK_RELOAD_ACTION_DRIVER_REINIT);
+ return mlx5_load_one(dev, false);
+ default:
+ /* Unsupported action should not get to this function */
+ WARN_ON(1);
+ return -EOPNOTSUPP;
+ }
+
+ return 0;
}
static const struct devlink_ops mlx5_devlink_ops = {
@@ -118,7 +166,8 @@ static const struct devlink_ops mlx5_devlink_ops = {
#endif
.flash_update = mlx5_devlink_flash_update,
.info_get = mlx5_devlink_info_get,
- .reload_actions = BIT(DEVLINK_RELOAD_ACTION_DRIVER_REINIT),
+ .reload_actions = BIT(DEVLINK_RELOAD_ACTION_DRIVER_REINIT) |
+ BIT(DEVLINK_RELOAD_ACTION_FW_ACTIVATE),
.reload_down = mlx5_devlink_reload_down,
.reload_up = mlx5_devlink_reload_up,
};
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fw_reset.c b/drivers/net/ethernet/mellanox/mlx5/core/fw_reset.c
index 56ae72e016f1..f5ffb6fc55c3 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fw_reset.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fw_reset.c
@@ -5,6 +5,7 @@
enum {
MLX5_FW_RESET_FLAGS_RESET_REQUESTED,
+ MLX5_FW_RESET_FLAGS_PENDING_COMP
};
struct mlx5_fw_reset {
@@ -17,6 +18,8 @@ struct mlx5_fw_reset {
struct work_struct reset_abort_work;
unsigned long reset_flags;
struct timer_list timer;
+ struct completion done;
+ int ret;
};
static int mlx5_reg_mfrl_set(struct mlx5_core_dev *dev, u8 reset_level,
@@ -58,7 +61,14 @@ int mlx5_fw_reset_query(struct mlx5_core_dev *dev, u8 *reset_level, u8 *reset_ty
int mlx5_fw_reset_set_reset_sync(struct mlx5_core_dev *dev, u8 reset_type_sel)
{
- return mlx5_reg_mfrl_set(dev, MLX5_MFRL_REG_RESET_LEVEL3, reset_type_sel, 0, true);
+ struct mlx5_fw_reset *fw_reset = dev->priv.fw_reset;
+ int err;
+
+ set_bit(MLX5_FW_RESET_FLAGS_PENDING_COMP, &fw_reset->reset_flags);
+ err = mlx5_reg_mfrl_set(dev, MLX5_MFRL_REG_RESET_LEVEL3, reset_type_sel, 0, true);
+ if (err)
+ clear_bit(MLX5_FW_RESET_FLAGS_PENDING_COMP, &fw_reset->reset_flags);
+ return err;
}
int mlx5_fw_reset_set_live_patch(struct mlx5_core_dev *dev)
@@ -66,19 +76,35 @@ int mlx5_fw_reset_set_live_patch(struct mlx5_core_dev *dev)
return mlx5_reg_mfrl_set(dev, MLX5_MFRL_REG_RESET_LEVEL0, 0, 0, false);
}
+static void mlx5_fw_reset_complete_reload(struct mlx5_core_dev *dev)
+{
+ struct mlx5_fw_reset *fw_reset = dev->priv.fw_reset;
+
+ /* if this is the driver that initiated the fw reset, devlink completed the reload */
+ if (test_bit(MLX5_FW_RESET_FLAGS_PENDING_COMP, &fw_reset->reset_flags)) {
+ complete(&fw_reset->done);
+ } else {
+ mlx5_load_one(dev, false);
+ devlink_remote_reload_actions_performed(priv_to_devlink(dev), 0,
+ BIT(DEVLINK_RELOAD_ACTION_DRIVER_REINIT) |
+ BIT(DEVLINK_RELOAD_ACTION_FW_ACTIVATE));
+ }
+}
+
static void mlx5_sync_reset_reload_work(struct work_struct *work)
{
struct mlx5_fw_reset *fw_reset = container_of(work, struct mlx5_fw_reset,
reset_reload_work);
struct mlx5_core_dev *dev = fw_reset->dev;
+ int err;
mlx5_enter_error_state(dev, true);
mlx5_unload_one(dev, false);
- if (mlx5_health_wait_pci_up(dev)) {
+ err = mlx5_health_wait_pci_up(dev);
+ if (err)
mlx5_core_err(dev, "reset reload flow aborted, PCI reads still not working\n");
- return;
- }
- mlx5_load_one(dev, false);
+ fw_reset->ret = err;
+ mlx5_fw_reset_complete_reload(dev);
}
static void mlx5_stop_sync_reset_poll(struct mlx5_core_dev *dev)
@@ -269,7 +295,8 @@ static void mlx5_sync_reset_now_event(struct work_struct *work)
mlx5_enter_error_state(dev, true);
mlx5_unload_one(dev, false);
done:
- mlx5_load_one(dev, false);
+ fw_reset->ret = err;
+ mlx5_fw_reset_complete_reload(dev);
}
static void mlx5_sync_reset_abort_event(struct work_struct *work)
@@ -318,6 +345,25 @@ static int fw_reset_event_notifier(struct notifier_block *nb, unsigned long acti
return NOTIFY_OK;
}
+#define MLX5_FW_RESET_TIMEOUT_MSEC 5000
+int mlx5_fw_reset_wait_reset_done(struct mlx5_core_dev *dev)
+{
+ unsigned long timeout = msecs_to_jiffies(MLX5_FW_RESET_TIMEOUT_MSEC);
+ struct mlx5_fw_reset *fw_reset = dev->priv.fw_reset;
+ int err;
+
+ if (!wait_for_completion_timeout(&fw_reset->done, timeout)) {
+ mlx5_core_warn(dev, "FW sync reset timeout after %d seconds\n",
+ MLX5_FW_RESET_TIMEOUT_MSEC / 1000);
+ err = -ETIMEDOUT;
+ goto out;
+ }
+ err = fw_reset->ret;
+out:
+ clear_bit(MLX5_FW_RESET_FLAGS_PENDING_COMP, &fw_reset->reset_flags);
+ return err;
+}
+
void mlx5_fw_reset_events_start(struct mlx5_core_dev *dev)
{
struct mlx5_fw_reset *fw_reset = dev->priv.fw_reset;
@@ -351,6 +397,7 @@ int mlx5_fw_reset_init(struct mlx5_core_dev *dev)
INIT_WORK(&fw_reset->reset_now_work, mlx5_sync_reset_now_event);
INIT_WORK(&fw_reset->reset_abort_work, mlx5_sync_reset_abort_event);
+ init_completion(&fw_reset->done);
return 0;
}
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fw_reset.h b/drivers/net/ethernet/mellanox/mlx5/core/fw_reset.h
index a231f7848a8f..e7937447ce1d 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fw_reset.h
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fw_reset.h
@@ -10,6 +10,7 @@ int mlx5_fw_reset_query(struct mlx5_core_dev *dev, u8 *reset_level, u8 *reset_ty
int mlx5_fw_reset_set_reset_sync(struct mlx5_core_dev *dev, u8 reset_type_sel);
int mlx5_fw_reset_set_live_patch(struct mlx5_core_dev *dev);
+int mlx5_fw_reset_wait_reset_done(struct mlx5_core_dev *dev);
void mlx5_fw_reset_events_start(struct mlx5_core_dev *dev);
void mlx5_fw_reset_events_stop(struct mlx5_core_dev *dev);
int mlx5_fw_reset_init(struct mlx5_core_dev *dev);
--
2.18.2
If firmware sends sync_reset_abort to driver the driver should clear the
reset requested mode as reset is not expected any more.
Signed-off-by: Moshe Shemesh <[email protected]>
Reviewed-by: Saeed Mahameed <[email protected]>
---
.../net/ethernet/mellanox/mlx5/core/fw_reset.c | 15 +++++++++++++++
1 file changed, 15 insertions(+)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fw_reset.c b/drivers/net/ethernet/mellanox/mlx5/core/fw_reset.c
index 26caf65e9f5f..56ae72e016f1 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/fw_reset.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/fw_reset.c
@@ -14,6 +14,7 @@ struct mlx5_fw_reset {
struct work_struct reset_request_work;
struct work_struct reset_reload_work;
struct work_struct reset_now_work;
+ struct work_struct reset_abort_work;
unsigned long reset_flags;
struct timer_list timer;
};
@@ -271,6 +272,16 @@ static void mlx5_sync_reset_now_event(struct work_struct *work)
mlx5_load_one(dev, false);
}
+static void mlx5_sync_reset_abort_event(struct work_struct *work)
+{
+ struct mlx5_fw_reset *fw_reset = container_of(work, struct mlx5_fw_reset,
+ reset_abort_work);
+ struct mlx5_core_dev *dev = fw_reset->dev;
+
+ mlx5_sync_reset_clear_reset_requested(dev, true);
+ mlx5_core_warn(dev, "PCI Sync FW Update Reset Aborted.\n");
+}
+
static void mlx5_sync_reset_events_handle(struct mlx5_fw_reset *fw_reset, struct mlx5_eqe *eqe)
{
struct mlx5_eqe_sync_fw_update *sync_fw_update_eqe;
@@ -285,6 +296,9 @@ static void mlx5_sync_reset_events_handle(struct mlx5_fw_reset *fw_reset, struct
case MLX5_SYNC_RST_STATE_RESET_NOW:
queue_work(fw_reset->wq, &fw_reset->reset_now_work);
break;
+ case MLX5_SYNC_RST_STATE_RESET_ABORT:
+ queue_work(fw_reset->wq, &fw_reset->reset_abort_work);
+ break;
}
}
@@ -335,6 +349,7 @@ int mlx5_fw_reset_init(struct mlx5_core_dev *dev)
INIT_WORK(&fw_reset->reset_request_work, mlx5_sync_reset_request_event);
INIT_WORK(&fw_reset->reset_reload_work, mlx5_sync_reset_reload_work);
INIT_WORK(&fw_reset->reset_now_work, mlx5_sync_reset_now_event);
+ INIT_WORK(&fw_reset->reset_abort_work, mlx5_sync_reset_abort_event);
return 0;
}
--
2.18.2
Add devlink reload action to allow the user to request a specific reload
action. The action parameter is optional, if not specified then devlink
driver re-init action is used (backward compatible).
Note that when required to do firmware activation some drivers may need
to reload the driver. On the other hand some drivers may need to reset
the firmware to reinitialize the driver entities. Therefore, the devlink
reload command returns the actions which were actually performed.
Reload actions supported are:
driver_reinit: driver entities re-initialization, applying devlink-param
and devlink-resource values.
fw_activate: firmware activate.
command examples:
$devlink dev reload pci/0000:82:00.0 action driver_reinit
reload_actions_performed:
driver_reinit
$devlink dev reload pci/0000:82:00.0 action fw_activate
reload_actions_performed:
driver_reinit fw_activate
Signed-off-by: Moshe Shemesh <[email protected]>
---
RFCv5 -> v1:
- Rename supported_reload_actions to reload_actions.
- Rename devlink_nl_reload_actions_performed_fill() to
devlink_nl_reload_actions_performed_snd() and add genlmsg_reply() to it
- Actions_performed sent to user space as a mask
- Driver can initialize actions_performed before done as devlink ignores
in case of failure
- Use nla_poilcy range validation and remove the range check in
devlink_nl_cmd_reload
RFCv4 -> RFCv5:
- Always pass actions_performed to unload_up() instead of checking in
each driver
- Verify returned actions_performed includes the requested action
- Changed devlink_reload_actions_verify(devlink) to get ops
- Changed devlink_reload_actions_verify() to return bool and rename to
devlink_reload_actions_valid()
- Only generate the reply if request uses new attributes
RFCv3 -> RFCv4:
- Removed fw_activate_no_reset as an action (next patch adds limit
levels instead).
- Renamed actions_done to actions_performed
RFCv2 -> RFCv3:
- Replace fw_live_patch action by fw_activate_no_reset
- Devlink reload returns the actions done over netlink reply
RFCv1 -> RFCv2:
- Instead of reload levels driver,fw_reset,fw_live_patch have reload
actions driver_reinit,fw_activate,fw_live_patch
- Remove driver default level, the action driver_reinit is the default
action for all drivers
---
drivers/net/ethernet/mellanox/mlx4/main.c | 7 +-
.../net/ethernet/mellanox/mlx5/core/devlink.c | 7 +-
drivers/net/ethernet/mellanox/mlxsw/core.c | 11 +-
drivers/net/netdevsim/dev.c | 8 +-
include/net/devlink.h | 7 +-
include/uapi/linux/devlink.h | 18 ++++
net/core/devlink.c | 101 ++++++++++++++++--
7 files changed, 139 insertions(+), 20 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx4/main.c b/drivers/net/ethernet/mellanox/mlx4/main.c
index 70cf24ba71e4..a44d8b733db3 100644
--- a/drivers/net/ethernet/mellanox/mlx4/main.c
+++ b/drivers/net/ethernet/mellanox/mlx4/main.c
@@ -3946,6 +3946,7 @@ static int mlx4_restart_one_up(struct pci_dev *pdev, bool reload,
struct devlink *devlink);
static int mlx4_devlink_reload_down(struct devlink *devlink, bool netns_change,
+ enum devlink_reload_action action,
struct netlink_ext_ack *extack)
{
struct mlx4_priv *priv = devlink_priv(devlink);
@@ -3962,14 +3963,15 @@ static int mlx4_devlink_reload_down(struct devlink *devlink, bool netns_change,
return 0;
}
-static int mlx4_devlink_reload_up(struct devlink *devlink,
- struct netlink_ext_ack *extack)
+static int mlx4_devlink_reload_up(struct devlink *devlink, enum devlink_reload_action action,
+ struct netlink_ext_ack *extack, unsigned long *actions_performed)
{
struct mlx4_priv *priv = devlink_priv(devlink);
struct mlx4_dev *dev = &priv->dev;
struct mlx4_dev_persistent *persist = dev->persist;
int err;
+ *actions_performed = BIT(DEVLINK_RELOAD_ACTION_DRIVER_REINIT);
err = mlx4_restart_one_up(persist->pdev, true, devlink);
if (err)
mlx4_err(persist->dev, "mlx4_restart_one_up failed, ret=%d\n",
@@ -3980,6 +3982,7 @@ static int mlx4_devlink_reload_up(struct devlink *devlink,
static const struct devlink_ops mlx4_devlink_ops = {
.port_type_set = mlx4_devlink_port_type_set,
+ .reload_actions = BIT(DEVLINK_RELOAD_ACTION_DRIVER_REINIT),
.reload_down = mlx4_devlink_reload_down,
.reload_up = mlx4_devlink_reload_up,
};
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/devlink.c b/drivers/net/ethernet/mellanox/mlx5/core/devlink.c
index 9b14e3f805a2..35aceab79a50 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/devlink.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/devlink.c
@@ -85,6 +85,7 @@ mlx5_devlink_info_get(struct devlink *devlink, struct devlink_info_req *req,
}
static int mlx5_devlink_reload_down(struct devlink *devlink, bool netns_change,
+ enum devlink_reload_action action,
struct netlink_ext_ack *extack)
{
struct mlx5_core_dev *dev = devlink_priv(devlink);
@@ -93,11 +94,12 @@ static int mlx5_devlink_reload_down(struct devlink *devlink, bool netns_change,
return 0;
}
-static int mlx5_devlink_reload_up(struct devlink *devlink,
- struct netlink_ext_ack *extack)
+static int mlx5_devlink_reload_up(struct devlink *devlink, enum devlink_reload_action action,
+ struct netlink_ext_ack *extack, unsigned long *actions_performed)
{
struct mlx5_core_dev *dev = devlink_priv(devlink);
+ *actions_performed = BIT(DEVLINK_RELOAD_ACTION_DRIVER_REINIT);
return mlx5_load_one(dev, false);
}
@@ -114,6 +116,7 @@ static const struct devlink_ops mlx5_devlink_ops = {
#endif
.flash_update = mlx5_devlink_flash_update,
.info_get = mlx5_devlink_info_get,
+ .reload_actions = BIT(DEVLINK_RELOAD_ACTION_DRIVER_REINIT),
.reload_down = mlx5_devlink_reload_down,
.reload_up = mlx5_devlink_reload_up,
};
diff --git a/drivers/net/ethernet/mellanox/mlxsw/core.c b/drivers/net/ethernet/mellanox/mlxsw/core.c
index a21afa56e3f7..35b800dba6f5 100644
--- a/drivers/net/ethernet/mellanox/mlxsw/core.c
+++ b/drivers/net/ethernet/mellanox/mlxsw/core.c
@@ -1414,7 +1414,7 @@ mlxsw_devlink_info_get(struct devlink *devlink, struct devlink_info_req *req,
static int
mlxsw_devlink_core_bus_device_reload_down(struct devlink *devlink,
- bool netns_change,
+ bool netns_change, enum devlink_reload_action action,
struct netlink_ext_ack *extack)
{
struct mlxsw_core *mlxsw_core = devlink_priv(devlink);
@@ -1427,11 +1427,14 @@ mlxsw_devlink_core_bus_device_reload_down(struct devlink *devlink,
}
static int
-mlxsw_devlink_core_bus_device_reload_up(struct devlink *devlink,
- struct netlink_ext_ack *extack)
+mlxsw_devlink_core_bus_device_reload_up(struct devlink *devlink, enum devlink_reload_action action,
+ struct netlink_ext_ack *extack,
+ unsigned long *actions_performed)
{
struct mlxsw_core *mlxsw_core = devlink_priv(devlink);
+ *actions_performed = BIT(DEVLINK_RELOAD_ACTION_DRIVER_REINIT) |
+ BIT(DEVLINK_RELOAD_ACTION_FW_ACTIVATE);
return mlxsw_core_bus_device_register(mlxsw_core->bus_info,
mlxsw_core->bus,
mlxsw_core->bus_priv, true,
@@ -1564,6 +1567,8 @@ mlxsw_devlink_trap_policer_counter_get(struct devlink *devlink,
}
static const struct devlink_ops mlxsw_devlink_ops = {
+ .reload_actions = BIT(DEVLINK_RELOAD_ACTION_DRIVER_REINIT) |
+ BIT(DEVLINK_RELOAD_ACTION_FW_ACTIVATE),
.reload_down = mlxsw_devlink_core_bus_device_reload_down,
.reload_up = mlxsw_devlink_core_bus_device_reload_up,
.port_type_set = mlxsw_devlink_port_type_set,
diff --git a/drivers/net/netdevsim/dev.c b/drivers/net/netdevsim/dev.c
index 56213ba151f6..e1973d752148 100644
--- a/drivers/net/netdevsim/dev.c
+++ b/drivers/net/netdevsim/dev.c
@@ -701,7 +701,7 @@ static int nsim_dev_reload_create(struct nsim_dev *nsim_dev,
static void nsim_dev_reload_destroy(struct nsim_dev *nsim_dev);
static int nsim_dev_reload_down(struct devlink *devlink, bool netns_change,
- struct netlink_ext_ack *extack)
+ enum devlink_reload_action action, struct netlink_ext_ack *extack)
{
struct nsim_dev *nsim_dev = devlink_priv(devlink);
@@ -717,8 +717,8 @@ static int nsim_dev_reload_down(struct devlink *devlink, bool netns_change,
return 0;
}
-static int nsim_dev_reload_up(struct devlink *devlink,
- struct netlink_ext_ack *extack)
+static int nsim_dev_reload_up(struct devlink *devlink, enum devlink_reload_action action,
+ struct netlink_ext_ack *extack, unsigned long *actions_performed)
{
struct nsim_dev *nsim_dev = devlink_priv(devlink);
@@ -730,6 +730,7 @@ static int nsim_dev_reload_up(struct devlink *devlink,
return -EINVAL;
}
+ *actions_performed = BIT(DEVLINK_RELOAD_ACTION_DRIVER_REINIT);
return nsim_dev_reload_create(nsim_dev, extack);
}
@@ -886,6 +887,7 @@ nsim_dev_devlink_trap_policer_counter_get(struct devlink *devlink,
static const struct devlink_ops nsim_dev_devlink_ops = {
.supported_flash_update_params = DEVLINK_SUPPORT_FLASH_UPDATE_COMPONENT |
DEVLINK_SUPPORT_FLASH_UPDATE_OVERWRITE_MASK,
+ .reload_actions = BIT(DEVLINK_RELOAD_ACTION_DRIVER_REINIT),
.reload_down = nsim_dev_reload_down,
.reload_up = nsim_dev_reload_up,
.info_get = nsim_dev_info_get,
diff --git a/include/net/devlink.h b/include/net/devlink.h
index 1c286e9a3590..ddba63bce7ad 100644
--- a/include/net/devlink.h
+++ b/include/net/devlink.h
@@ -1077,10 +1077,11 @@ struct devlink_ops {
* implemementation.
*/
u32 supported_flash_update_params;
+ unsigned long reload_actions;
int (*reload_down)(struct devlink *devlink, bool netns_change,
- struct netlink_ext_ack *extack);
- int (*reload_up)(struct devlink *devlink,
- struct netlink_ext_ack *extack);
+ enum devlink_reload_action action, struct netlink_ext_ack *extack);
+ int (*reload_up)(struct devlink *devlink, enum devlink_reload_action action,
+ struct netlink_ext_ack *extack, unsigned long *actions_performed);
int (*port_type_set)(struct devlink_port *devlink_port,
enum devlink_port_type port_type);
int (*port_split)(struct devlink *devlink, unsigned int port_index,
diff --git a/include/uapi/linux/devlink.h b/include/uapi/linux/devlink.h
index ba467dc07852..f1782eaf0be4 100644
--- a/include/uapi/linux/devlink.h
+++ b/include/uapi/linux/devlink.h
@@ -298,6 +298,21 @@ enum {
DEVLINK_ATTR_TRAP_METADATA_TYPE_FA_COOKIE,
};
+/**
+ * enum devlink_reload_action - Reload action.
+ * @DEVLINK_RELOAD_ACTION_DRIVER_REINIT: Driver entities re-instantiation.
+ * @DEVLINK_RELOAD_ACTION_FW_ACTIVATE: FW activate.
+ */
+enum devlink_reload_action {
+ DEVLINK_RELOAD_ACTION_UNSPEC,
+ DEVLINK_RELOAD_ACTION_DRIVER_REINIT,
+ DEVLINK_RELOAD_ACTION_FW_ACTIVATE,
+
+ /* Add new reload actions above */
+ __DEVLINK_RELOAD_ACTION_MAX,
+ DEVLINK_RELOAD_ACTION_MAX = __DEVLINK_RELOAD_ACTION_MAX - 1
+};
+
enum devlink_attr {
/* don't change the order or add anything between, this is ABI! */
DEVLINK_ATTR_UNSPEC,
@@ -490,6 +505,9 @@ enum devlink_attr {
DEVLINK_ATTR_FLASH_UPDATE_STATUS_TIMEOUT, /* u64 */
DEVLINK_ATTR_FLASH_UPDATE_OVERWRITE_MASK, /* bitfield32 */
+ DEVLINK_ATTR_RELOAD_ACTION, /* u8 */
+ DEVLINK_ATTR_RELOAD_ACTIONS_PERFORMED, /* u64 */
+
/* add new attributes above here, update the policy in devlink.c */
__DEVLINK_ATTR_MAX,
diff --git a/net/core/devlink.c b/net/core/devlink.c
index 722a9431ff60..6f39dc76370e 100644
--- a/net/core/devlink.c
+++ b/net/core/devlink.c
@@ -462,6 +462,12 @@ static int devlink_nl_put_handle(struct sk_buff *msg, struct devlink *devlink)
return 0;
}
+static bool
+devlink_reload_action_is_supported(struct devlink *devlink, enum devlink_reload_action action)
+{
+ return test_bit(action, &devlink->ops->reload_actions);
+}
+
static int devlink_nl_fill(struct sk_buff *msg, struct devlink *devlink,
enum devlink_command cmd, u32 portid,
u32 seq, int flags)
@@ -2967,28 +2973,68 @@ bool devlink_is_reload_failed(const struct devlink *devlink)
EXPORT_SYMBOL_GPL(devlink_is_reload_failed);
static int devlink_reload(struct devlink *devlink, struct net *dest_net,
- struct netlink_ext_ack *extack)
+ enum devlink_reload_action action, struct netlink_ext_ack *extack,
+ unsigned long *actions_performed)
{
int err;
if (!devlink->reload_enabled)
return -EOPNOTSUPP;
- err = devlink->ops->reload_down(devlink, !!dest_net, extack);
+ err = devlink->ops->reload_down(devlink, !!dest_net, action, extack);
if (err)
return err;
if (dest_net && !net_eq(dest_net, devlink_net(devlink)))
devlink_reload_netns_change(devlink, dest_net);
- err = devlink->ops->reload_up(devlink, extack);
+ err = devlink->ops->reload_up(devlink, action, extack, actions_performed);
devlink_reload_failed_set(devlink, !!err);
- return err;
+ if (err)
+ return err;
+
+ WARN_ON(!test_bit(action, actions_performed));
+ return 0;
+}
+
+static int
+devlink_nl_reload_actions_performed_snd(struct devlink *devlink,
+ unsigned long actions_performed,
+ enum devlink_command cmd, struct genl_info *info)
+{
+ struct sk_buff *msg;
+ void *hdr;
+
+ msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
+ if (!msg)
+ return -ENOMEM;
+
+ hdr = genlmsg_put(msg, info->snd_portid, info->snd_seq, &devlink_nl_family, 0, cmd);
+ if (!hdr)
+ goto free_msg;
+
+ if (devlink_nl_put_handle(msg, devlink))
+ goto nla_put_failure;
+
+ if (nla_put_u64_64bit(msg, DEVLINK_ATTR_RELOAD_ACTIONS_PERFORMED, actions_performed,
+ DEVLINK_ATTR_PAD))
+ goto nla_put_failure;
+ genlmsg_end(msg, hdr);
+
+ return genlmsg_reply(msg, info);
+
+nla_put_failure:
+ genlmsg_cancel(msg, hdr);
+free_msg:
+ nlmsg_free(msg);
+ return -EMSGSIZE;
}
static int devlink_nl_cmd_reload(struct sk_buff *skb, struct genl_info *info)
{
struct devlink *devlink = info->user_ptr[0];
+ enum devlink_reload_action action;
+ unsigned long actions_performed;
struct net *dest_net = NULL;
int err;
@@ -3009,12 +3055,30 @@ static int devlink_nl_cmd_reload(struct sk_buff *skb, struct genl_info *info)
return PTR_ERR(dest_net);
}
- err = devlink_reload(devlink, dest_net, info->extack);
+ if (info->attrs[DEVLINK_ATTR_RELOAD_ACTION])
+ action = nla_get_u8(info->attrs[DEVLINK_ATTR_RELOAD_ACTION]);
+ else
+ action = DEVLINK_RELOAD_ACTION_DRIVER_REINIT;
+
+ if (!devlink_reload_action_is_supported(devlink, action)) {
+ NL_SET_ERR_MSG_MOD(info->extack,
+ "Requested reload action is not supported by the driver");
+ return -EOPNOTSUPP;
+ }
+
+ err = devlink_reload(devlink, dest_net, action, info->extack, &actions_performed);
if (dest_net)
put_net(dest_net);
- return err;
+ if (err)
+ return err;
+ /* For backward compatibility generate reply only if attributes used by user */
+ if (!info->attrs[DEVLINK_ATTR_RELOAD_ACTION])
+ return 0;
+
+ return devlink_nl_reload_actions_performed_snd(devlink, actions_performed,
+ DEVLINK_CMD_RELOAD, info);
}
static int devlink_nl_flash_update_fill(struct sk_buff *msg,
@@ -7119,6 +7183,8 @@ static const struct nla_policy devlink_nl_policy[DEVLINK_ATTR_MAX + 1] = {
[DEVLINK_ATTR_TRAP_POLICER_RATE] = { .type = NLA_U64 },
[DEVLINK_ATTR_TRAP_POLICER_BURST] = { .type = NLA_U64 },
[DEVLINK_ATTR_PORT_FUNCTION] = { .type = NLA_NESTED },
+ [DEVLINK_ATTR_RELOAD_ACTION] = NLA_POLICY_RANGE(NLA_U8, DEVLINK_RELOAD_ACTION_DRIVER_REINIT,
+ DEVLINK_RELOAD_ACTION_MAX),
};
static const struct genl_ops devlink_nl_ops[] = {
@@ -7452,6 +7518,21 @@ static struct genl_family devlink_nl_family __ro_after_init = {
.n_mcgrps = ARRAY_SIZE(devlink_nl_mcgrps),
};
+static bool devlink_reload_actions_valid(const struct devlink_ops *ops)
+{
+ if (!devlink_reload_supported(ops)) {
+ if (WARN_ON(ops->reload_actions))
+ return false;
+ return true;
+ }
+
+ if (WARN_ON(!ops->reload_actions ||
+ ops->reload_actions & BIT(DEVLINK_RELOAD_ACTION_UNSPEC) ||
+ ops->reload_actions >= BIT(__DEVLINK_RELOAD_ACTION_MAX)))
+ return false;
+ return true;
+}
+
/**
* devlink_alloc - Allocate new devlink instance resources
*
@@ -7468,6 +7549,9 @@ struct devlink *devlink_alloc(const struct devlink_ops *ops, size_t priv_size)
if (WARN_ON(!ops))
return NULL;
+ if (!devlink_reload_actions_valid(ops))
+ return NULL;
+
devlink = kzalloc(sizeof(*devlink) + priv_size, GFP_KERNEL);
if (!devlink)
return NULL;
@@ -9724,6 +9808,7 @@ int devlink_compat_switch_id_get(struct net_device *dev,
static void __net_exit devlink_pernet_pre_exit(struct net *net)
{
+ unsigned long actions_performed;
struct devlink *devlink;
int err;
@@ -9735,7 +9820,9 @@ static void __net_exit devlink_pernet_pre_exit(struct net *net)
if (net_eq(devlink_net(devlink), net)) {
if (WARN_ON(!devlink_reload_supported(devlink->ops)))
continue;
- err = devlink_reload(devlink, &init_net, NULL);
+ err = devlink_reload(devlink, &init_net,
+ DEVLINK_RELOAD_ACTION_DRIVER_REINIT,
+ NULL, &actions_performed);
if (err && err != -EOPNOTSUPP)
pr_warn("Failed to reload devlink instance into init_net\n");
}
--
2.18.2
Add devlink reload rst documentation file.
Update index file to include it.
Signed-off-by: Moshe Shemesh <[email protected]>
---
RFCv5 -> v1:
- Rename reload_action_limit_level to reload_limit
RFCv4 -> RFCv5:
- Rephrase namespace chnage section
- Rephrase note on actions performed
RFCv3 -> RFCv4:
- Remove reload action fw_activate_no_reset
- Add reload actions limit levels and document the no_reset limit level
constrains
RFCv2 -> RFCv3:
- Devlink reload returns the actions done
- Replace fw_live_patch action by fw_activate_no_reset
- Explain fw_activate meaning
RFCv1 -> RFCv2:
- Instead of reload levels driver,fw_reset,fw_live_patch have reload
actions driver_reinit,fw_activate,fw_live_patch
---
.../networking/devlink/devlink-reload.rst | 81 +++++++++++++++++++
Documentation/networking/devlink/index.rst | 1 +
2 files changed, 82 insertions(+)
create mode 100644 Documentation/networking/devlink/devlink-reload.rst
diff --git a/Documentation/networking/devlink/devlink-reload.rst b/Documentation/networking/devlink/devlink-reload.rst
new file mode 100644
index 000000000000..5abc5c2c75fd
--- /dev/null
+++ b/Documentation/networking/devlink/devlink-reload.rst
@@ -0,0 +1,81 @@
+.. SPDX-License-Identifier: GPL-2.0
+
+==============
+Devlink Reload
+==============
+
+``devlink-reload`` provides mechanism to either reinit driver entities,
+applying ``devlink-params`` and ``devlink-resources`` new values or firmware
+activation depends on reload action selected.
+
+Reload actions
+==============
+
+User may select a reload action.
+By default ``driver_reinit`` action is selected.
+
+.. list-table:: Possible reload actions
+ :widths: 5 90
+
+ * - Name
+ - Description
+ * - ``driver-reinit``
+ - Devlink driver entities re-initialization, including applying
+ new values to devlink entities which are used during driver
+ load such as ``devlink-params`` in configuration mode
+ ``driverinit`` or ``devlink-resources``
+ * - ``fw_activate``
+ - Firmware activate. Activates new firmware if such image is stored and
+ pending activation. If no limitation specified this action may involve
+ firmware reset. If no new image pending this action will reload current
+ firmware image.
+
+Note that even though user asks for a specific action, the driver
+implementation might require to perform another action alongside with
+it. For example, some driver do not support driver reinitialization
+being performed without fw activation. Therefore, the devlink reload
+command returns the list of actions which were actrually performed.
+
+Reload limits
+=============
+
+By default reload actions are not limited and driver implementation may
+include reset or downtime as needed to perform the actions.
+
+However, some drivers support action limits, which limit the action
+implementation to specific constrains.
+
+.. list-table:: Possible reload limits
+ :widths: 5 90
+
+ * - Name
+ - Description
+ * - ``no_reset``
+ - No reset allowed, no down time allowed, no link flap and no
+ configuration is lost.
+
+Change namespace
+================
+
+The netns option allow user to be able to move devlink instances into
+namespaces during devlink reload operation.
+By default all devlink instances are created in init_net and stay there.
+
+example usage
+-------------
+
+.. code:: shell
+
+ $ devlink dev reload help
+ $ devlink dev reload DEV [ netns { PID | NAME | ID } ] [ action { driver_reinit | fw_activate } ] [ limit no_reset ]
+
+ # Run reload command for devlink driver entities re-initialization:
+ $ devlink dev reload pci/0000:82:00.0 action driver_reinit
+ reload_actions_performed:
+ driver_reinit
+
+ # Run reload command to activate firmware:
+ # Note that mlx5 driver reloads the driver while activating firmware
+ $ devlink dev reload pci/0000:82:00.0 action fw_activate
+ reload_actions_performed:
+ driver_reinit fw_activate
diff --git a/Documentation/networking/devlink/index.rst b/Documentation/networking/devlink/index.rst
index 7684ae5c4a4a..d82874760ae2 100644
--- a/Documentation/networking/devlink/index.rst
+++ b/Documentation/networking/devlink/index.rst
@@ -20,6 +20,7 @@ general.
devlink-params
devlink-region
devlink-resource
+ devlink-reload
devlink-trap
Driver-specific documentation
--
2.18.2
The enable_remote_dev_reset devlink param flags that the host admin
allows device resets that can be initiated by other hosts. This
parameter is useful for setups where a device is shared by different
hosts, such as multi-host setup. Once the user set this parameter to
false, the driver should NACK any attempt to reset the device while the
driver is loaded.
Signed-off-by: Moshe Shemesh <[email protected]>
Reviewed-by: Jiri Pirko <[email protected]>
---
Documentation/networking/devlink/devlink-params.rst | 6 ++++++
include/net/devlink.h | 4 ++++
net/core/devlink.c | 5 +++++
3 files changed, 15 insertions(+)
diff --git a/Documentation/networking/devlink/devlink-params.rst b/Documentation/networking/devlink/devlink-params.rst
index d075fd090b3d..54c9f107c4b0 100644
--- a/Documentation/networking/devlink/devlink-params.rst
+++ b/Documentation/networking/devlink/devlink-params.rst
@@ -108,3 +108,9 @@ own name.
* - ``region_snapshot_enable``
- Boolean
- Enable capture of ``devlink-region`` snapshots.
+ * - ``enable_remote_dev_reset``
+ - Boolean
+ - Enable device reset by remote host. When cleared, the device driver
+ will NACK any attempt of other host to reset the device. This parameter
+ is useful for setups where a device is shared by different hosts, such
+ as multi-host setup.
diff --git a/include/net/devlink.h b/include/net/devlink.h
index a4ccb83bbd2c..0a1c88805627 100644
--- a/include/net/devlink.h
+++ b/include/net/devlink.h
@@ -464,6 +464,7 @@ enum devlink_param_generic_id {
DEVLINK_PARAM_GENERIC_ID_FW_LOAD_POLICY,
DEVLINK_PARAM_GENERIC_ID_RESET_DEV_ON_DRV_PROBE,
DEVLINK_PARAM_GENERIC_ID_ENABLE_ROCE,
+ DEVLINK_PARAM_GENERIC_ID_ENABLE_REMOTE_DEV_RESET,
/* add new param generic ids above here*/
__DEVLINK_PARAM_GENERIC_ID_MAX,
@@ -501,6 +502,9 @@ enum devlink_param_generic_id {
#define DEVLINK_PARAM_GENERIC_ENABLE_ROCE_NAME "enable_roce"
#define DEVLINK_PARAM_GENERIC_ENABLE_ROCE_TYPE DEVLINK_PARAM_TYPE_BOOL
+#define DEVLINK_PARAM_GENERIC_ENABLE_REMOTE_DEV_RESET_NAME "enable_remote_dev_reset"
+#define DEVLINK_PARAM_GENERIC_ENABLE_REMOTE_DEV_RESET_TYPE DEVLINK_PARAM_TYPE_BOOL
+
#define DEVLINK_PARAM_GENERIC(_id, _cmodes, _get, _set, _validate) \
{ \
.id = DEVLINK_PARAM_GENERIC_ID_##_id, \
diff --git a/net/core/devlink.c b/net/core/devlink.c
index 3b6bd3b4d346..4f074c81a5e4 100644
--- a/net/core/devlink.c
+++ b/net/core/devlink.c
@@ -3484,6 +3484,11 @@ static const struct devlink_param devlink_param_generic[] = {
.name = DEVLINK_PARAM_GENERIC_ENABLE_ROCE_NAME,
.type = DEVLINK_PARAM_GENERIC_ENABLE_ROCE_TYPE,
},
+ {
+ .id = DEVLINK_PARAM_GENERIC_ID_ENABLE_REMOTE_DEV_RESET,
+ .name = DEVLINK_PARAM_GENERIC_ENABLE_REMOTE_DEV_RESET_NAME,
+ .type = DEVLINK_PARAM_GENERIC_ENABLE_REMOTE_DEV_RESET_TYPE,
+ },
};
static int devlink_param_generic_verify(const struct devlink_param *param)
--
2.18.2
Set capability to notify the firmware that this host driver is capable
of handling pci sync for firmware update events.
Signed-off-by: Moshe Shemesh <[email protected]>
Reviewed-by: Saeed Mahameed <[email protected]>
---
drivers/net/ethernet/mellanox/mlx5/core/main.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/main.c b/drivers/net/ethernet/mellanox/mlx5/core/main.c
index ce43e3feccd9..871d28b09f8a 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/main.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/main.c
@@ -548,6 +548,9 @@ static int handle_hca_cap(struct mlx5_core_dev *dev, void *set_ctx)
if (MLX5_CAP_GEN_MAX(dev, dct))
MLX5_SET(cmd_hca_cap, set_hca_cap, dct, 1);
+ if (MLX5_CAP_GEN_MAX(dev, pci_sync_for_fw_update_event))
+ MLX5_SET(cmd_hca_cap, set_hca_cap, pci_sync_for_fw_update_event, 1);
+
if (MLX5_CAP_GEN_MAX(dev, num_vhca_ports))
MLX5_SET(cmd_hca_cap,
set_hca_cap,
--
2.18.2
Add support for devlink reload action fw_activate with reload limit
no_reset which does firmware live patching, updating the firmware image
without reset, no downtime and no configuration lose. The driver checks
if the firmware is capable of handling the pending firmware changes as a
live patch. If it is then it triggers firmware live patching flow.
Signed-off-by: Moshe Shemesh <[email protected]>
Reviewed-by: Saeed Mahameed <[email protected]>
---
RFCv5 -> v1:
- Renamed reload_action_limit_level to reload_limit
RFCv3 -> RFCv4:
- Have action fw_activate with limit level no_reset instead of action
fw_activate_no_reset
RFCv2 -> RFCv3:
- Replace fw_live_patch action by fw_activate_no_reset
RFCv1 -> RFCv2:
- Have fw_live_patch action instead of level
---
.../net/ethernet/mellanox/mlx5/core/devlink.c | 29 +++++++++++++++++++
1 file changed, 29 insertions(+)
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/devlink.c b/drivers/net/ethernet/mellanox/mlx5/core/devlink.c
index 7b304227ad57..db59933a7fe1 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/devlink.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/devlink.c
@@ -111,6 +111,29 @@ static int mlx5_devlink_reload_fw_activate(struct devlink *devlink, struct netli
return err;
}
+static int mlx5_devlink_trigger_fw_live_patch(struct devlink *devlink,
+ struct netlink_ext_ack *extack)
+{
+ struct mlx5_core_dev *dev = devlink_priv(devlink);
+ u8 reset_level;
+ int err;
+
+ err = mlx5_fw_reset_query(dev, &reset_level, NULL);
+ if (err)
+ return err;
+ if (!(reset_level & MLX5_MFRL_REG_RESET_LEVEL0)) {
+ NL_SET_ERR_MSG_MOD(extack,
+ "FW upgrade to the stored FW can't be done by FW live patching");
+ return -EINVAL;
+ }
+
+ err = mlx5_fw_reset_set_live_patch(dev);
+ if (err)
+ return err;
+
+ return 0;
+}
+
static int mlx5_devlink_reload_down(struct devlink *devlink, bool netns_change,
enum devlink_reload_action action,
enum devlink_reload_limit limit,
@@ -123,6 +146,8 @@ static int mlx5_devlink_reload_down(struct devlink *devlink, bool netns_change,
mlx5_unload_one(dev, false);
return 0;
case DEVLINK_RELOAD_ACTION_FW_ACTIVATE:
+ if (limit == DEVLINK_RELOAD_LIMIT_NO_RESET)
+ return mlx5_devlink_trigger_fw_live_patch(devlink, extack);
return mlx5_devlink_reload_fw_activate(devlink, extack);
default:
/* Unsupported action should not get to this function */
@@ -140,7 +165,10 @@ static int mlx5_devlink_reload_up(struct devlink *devlink, enum devlink_reload_a
*actions_performed = BIT(action);
switch (action) {
case DEVLINK_RELOAD_ACTION_DRIVER_REINIT:
+ return mlx5_load_one(dev, false);
case DEVLINK_RELOAD_ACTION_FW_ACTIVATE:
+ if (limit == DEVLINK_RELOAD_LIMIT_NO_RESET)
+ break;
/* On fw_activate action, also driver is reloaded and reinit performed */
*actions_performed |= BIT(DEVLINK_RELOAD_ACTION_DRIVER_REINIT);
return mlx5_load_one(dev, false);
@@ -168,6 +196,7 @@ static const struct devlink_ops mlx5_devlink_ops = {
.info_get = mlx5_devlink_info_get,
.reload_actions = BIT(DEVLINK_RELOAD_ACTION_DRIVER_REINIT) |
BIT(DEVLINK_RELOAD_ACTION_FW_ACTIVATE),
+ .reload_limits = BIT(DEVLINK_RELOAD_LIMIT_NO_RESET),
.reload_down = mlx5_devlink_reload_down,
.reload_up = mlx5_devlink_reload_up,
};
--
2.18.2
On Thu, 1 Oct 2020 16:59:05 +0300 Moshe Shemesh wrote:
> Add devlink reload action to allow the user to request a specific reload
> action. The action parameter is optional, if not specified then devlink
> driver re-init action is used (backward compatible).
> Note that when required to do firmware activation some drivers may need
> to reload the driver. On the other hand some drivers may need to reset
> the firmware to reinitialize the driver entities. Therefore, the devlink
> reload command returns the actions which were actually performed.
> Reload actions supported are:
> driver_reinit: driver entities re-initialization, applying devlink-param
> and devlink-resource values.
> fw_activate: firmware activate.
>
> command examples:
> $devlink dev reload pci/0000:82:00.0 action driver_reinit
> reload_actions_performed:
> driver_reinit
>
> $devlink dev reload pci/0000:82:00.0 action fw_activate
> reload_actions_performed:
> driver_reinit fw_activate
>
> Signed-off-by: Moshe Shemesh <[email protected]>
Reviewed-by: Jakub Kicinski <[email protected]>
On Thu, 1 Oct 2020 16:59:18 +0300 Moshe Shemesh wrote:
> + err = mlx5_fw_reset_set_live_patch(dev);
> + if (err)
> + return err;
> +
> + return 0;
nit return mlx...
On 10/2/2020 12:52 AM, Jakub Kicinski wrote:
> External email: Use caution opening links or attachments
>
>
> On Thu, 1 Oct 2020 16:59:18 +0300 Moshe Shemesh wrote:
>> + err = mlx5_fw_reset_set_live_patch(dev);
>> + if (err)
>> + return err;
>> +
>> + return 0;
> nit return mlx...
Right, will fix. Thanks.
Thu, Oct 01, 2020 at 03:59:05PM CEST, [email protected] wrote:
[...]
>diff --git a/include/net/devlink.h b/include/net/devlink.h
>index 1c286e9a3590..ddba63bce7ad 100644
>--- a/include/net/devlink.h
>+++ b/include/net/devlink.h
>@@ -1077,10 +1077,11 @@ struct devlink_ops {
> * implemementation.
> */
> u32 supported_flash_update_params;
>+ unsigned long reload_actions;
> int (*reload_down)(struct devlink *devlink, bool netns_change,
>- struct netlink_ext_ack *extack);
>- int (*reload_up)(struct devlink *devlink,
>- struct netlink_ext_ack *extack);
>+ enum devlink_reload_action action, struct netlink_ext_ack *extack);
>+ int (*reload_up)(struct devlink *devlink, enum devlink_reload_action action,
>+ struct netlink_ext_ack *extack, unsigned long *actions_performed);
Nit. Could you please push extack to be the last arg here? It is common
to have extack as the last arg + action and actions_performed are going
to be side by side.
Otherwise the patch looks fine.
Reviewed-by: Jiri Pirko <[email protected]>
Thu, Oct 01, 2020 at 03:59:05PM CEST, [email protected] wrote:
[...]
>+static int
>+devlink_nl_reload_actions_performed_snd(struct devlink *devlink,
>+ unsigned long actions_performed,
>+ enum devlink_command cmd, struct genl_info *info)
>+{
>+ struct sk_buff *msg;
>+ void *hdr;
>+
>+ msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
>+ if (!msg)
>+ return -ENOMEM;
>+
>+ hdr = genlmsg_put(msg, info->snd_portid, info->snd_seq, &devlink_nl_family, 0, cmd);
>+ if (!hdr)
>+ goto free_msg;
>+
>+ if (devlink_nl_put_handle(msg, devlink))
>+ goto nla_put_failure;
>+
>+ if (nla_put_u64_64bit(msg, DEVLINK_ATTR_RELOAD_ACTIONS_PERFORMED, actions_performed,
This should be NLA_BITFIELD, I believe. We use it for other bitfields
too.
>+ DEVLINK_ATTR_PAD))
[...]
Thu, Oct 01, 2020 at 03:59:19PM CEST, [email protected] wrote:
>Add devlink reload rst documentation file.
>Update index file to include it.
>
>Signed-off-by: Moshe Shemesh <[email protected]>
>---
>RFCv5 -> v1:
>- Rename reload_action_limit_level to reload_limit
>RFCv4 -> RFCv5:
>- Rephrase namespace chnage section
>- Rephrase note on actions performed
>RFCv3 -> RFCv4:
>- Remove reload action fw_activate_no_reset
>- Add reload actions limit levels and document the no_reset limit level
> constrains
>RFCv2 -> RFCv3:
>- Devlink reload returns the actions done
>- Replace fw_live_patch action by fw_activate_no_reset
>- Explain fw_activate meaning
>RFCv1 -> RFCv2:
>- Instead of reload levels driver,fw_reset,fw_live_patch have reload
> actions driver_reinit,fw_activate,fw_live_patch
>---
> .../networking/devlink/devlink-reload.rst | 81 +++++++++++++++++++
> Documentation/networking/devlink/index.rst | 1 +
> 2 files changed, 82 insertions(+)
> create mode 100644 Documentation/networking/devlink/devlink-reload.rst
>
>diff --git a/Documentation/networking/devlink/devlink-reload.rst b/Documentation/networking/devlink/devlink-reload.rst
>new file mode 100644
>index 000000000000..5abc5c2c75fd
>--- /dev/null
>+++ b/Documentation/networking/devlink/devlink-reload.rst
>@@ -0,0 +1,81 @@
>+.. SPDX-License-Identifier: GPL-2.0
>+
>+==============
>+Devlink Reload
No reason for capital "R".
>+==============
>+
>+``devlink-reload`` provides mechanism to either reinit driver entities,
>+applying ``devlink-params`` and ``devlink-resources`` new values or firmware
>+activation depends on reload action selected.
Could you perhaps split the sentense? It is hard to read.
>+
>+Reload actions
>+==============
>+
>+User may select a reload action.
>+By default ``driver_reinit`` action is selected.
>+
>+.. list-table:: Possible reload actions
>+ :widths: 5 90
>+
>+ * - Name
>+ - Description
>+ * - ``driver-reinit``
>+ - Devlink driver entities re-initialization, including applying
>+ new values to devlink entities which are used during driver
>+ load such as ``devlink-params`` in configuration mode
>+ ``driverinit`` or ``devlink-resources``
>+ * - ``fw_activate``
>+ - Firmware activate. Activates new firmware if such image is stored and
>+ pending activation. If no limitation specified this action may involve
>+ firmware reset. If no new image pending this action will reload current
>+ firmware image.
>+
>+Note that even though user asks for a specific action, the driver
>+implementation might require to perform another action alongside with
>+it. For example, some driver do not support driver reinitialization
>+being performed without fw activation. Therefore, the devlink reload
>+command returns the list of actions which were actrually performed.
>+
>+Reload limits
>+=============
>+
>+By default reload actions are not limited and driver implementation may
>+include reset or downtime as needed to perform the actions.
>+
>+However, some drivers support action limits, which limit the action
>+implementation to specific constrains.
>+
>+.. list-table:: Possible reload limits
>+ :widths: 5 90
>+
>+ * - Name
>+ - Description
>+ * - ``no_reset``
>+ - No reset allowed, no down time allowed, no link flap and no
>+ configuration is lost.
>+
>+Change namespace
>+================
>+
>+The netns option allow user to be able to move devlink instances into
"allows"
>+namespaces during devlink reload operation.
>+By default all devlink instances are created in init_net and stay there.
>+
>+example usage
>+-------------
>+
>+.. code:: shell
>+
>+ $ devlink dev reload help
>+ $ devlink dev reload DEV [ netns { PID | NAME | ID } ] [ action { driver_reinit | fw_activate } ] [ limit no_reset ]
>+
>+ # Run reload command for devlink driver entities re-initialization:
>+ $ devlink dev reload pci/0000:82:00.0 action driver_reinit
>+ reload_actions_performed:
>+ driver_reinit
>+
>+ # Run reload command to activate firmware:
>+ # Note that mlx5 driver reloads the driver while activating firmware
>+ $ devlink dev reload pci/0000:82:00.0 action fw_activate
>+ reload_actions_performed:
>+ driver_reinit fw_activate
>diff --git a/Documentation/networking/devlink/index.rst b/Documentation/networking/devlink/index.rst
>index 7684ae5c4a4a..d82874760ae2 100644
>--- a/Documentation/networking/devlink/index.rst
>+++ b/Documentation/networking/devlink/index.rst
>@@ -20,6 +20,7 @@ general.
> devlink-params
> devlink-region
> devlink-resource
>+ devlink-reload
> devlink-trap
>
> Driver-specific documentation
>--
>2.18.2
>
On 10/2/2020 6:19 PM, Jiri Pirko wrote:
> Thu, Oct 01, 2020 at 03:59:05PM CEST, [email protected] wrote:
>
> [...]
>
>
>> diff --git a/include/net/devlink.h b/include/net/devlink.h
>> index 1c286e9a3590..ddba63bce7ad 100644
>> --- a/include/net/devlink.h
>> +++ b/include/net/devlink.h
>> @@ -1077,10 +1077,11 @@ struct devlink_ops {
>> * implemementation.
>> */
>> u32 supported_flash_update_params;
>> + unsigned long reload_actions;
>> int (*reload_down)(struct devlink *devlink, bool netns_change,
>> - struct netlink_ext_ack *extack);
>> - int (*reload_up)(struct devlink *devlink,
>> - struct netlink_ext_ack *extack);
>> + enum devlink_reload_action action, struct netlink_ext_ack *extack);
>> + int (*reload_up)(struct devlink *devlink, enum devlink_reload_action action,
>> + struct netlink_ext_ack *extack, unsigned long *actions_performed);
> Nit. Could you please push extack to be the last arg here? It is common
> to have extack as the last arg + action and actions_performed are going
> to be side by side.
Sure.
> Otherwise the patch looks fine.
> Reviewed-by: Jiri Pirko <[email protected]>
On 10/3/2020 10:52 AM, Jiri Pirko wrote:
> Thu, Oct 01, 2020 at 03:59:05PM CEST, [email protected] wrote:
>
> [...]
>
>> +static int
>> +devlink_nl_reload_actions_performed_snd(struct devlink *devlink,
>> + unsigned long actions_performed,
>> + enum devlink_command cmd, struct genl_info *info)
>> +{
>> + struct sk_buff *msg;
>> + void *hdr;
>> +
>> + msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
>> + if (!msg)
>> + return -ENOMEM;
>> +
>> + hdr = genlmsg_put(msg, info->snd_portid, info->snd_seq, &devlink_nl_family, 0, cmd);
>> + if (!hdr)
>> + goto free_msg;
>> +
>> + if (devlink_nl_put_handle(msg, devlink))
>> + goto nla_put_failure;
>> +
>> + if (nla_put_u64_64bit(msg, DEVLINK_ATTR_RELOAD_ACTIONS_PERFORMED, actions_performed,
> This should be NLA_BITFIELD, I believe. We use it for other bitfields
> too.
>
OK, I see it now, NLA_BITFIELD32.
>> + DEVLINK_ATTR_PAD))
> [...]
On 10/3/2020 12:14 PM, Jiri Pirko wrote:
> Thu, Oct 01, 2020 at 03:59:19PM CEST, [email protected] wrote:
>> Add devlink reload rst documentation file.
>> Update index file to include it.
>>
>> Signed-off-by: Moshe Shemesh <[email protected]>
>> ---
>> RFCv5 -> v1:
>> - Rename reload_action_limit_level to reload_limit
>> RFCv4 -> RFCv5:
>> - Rephrase namespace chnage section
>> - Rephrase note on actions performed
>> RFCv3 -> RFCv4:
>> - Remove reload action fw_activate_no_reset
>> - Add reload actions limit levels and document the no_reset limit level
>> constrains
>> RFCv2 -> RFCv3:
>> - Devlink reload returns the actions done
>> - Replace fw_live_patch action by fw_activate_no_reset
>> - Explain fw_activate meaning
>> RFCv1 -> RFCv2:
>> - Instead of reload levels driver,fw_reset,fw_live_patch have reload
>> actions driver_reinit,fw_activate,fw_live_patch
>> ---
>> .../networking/devlink/devlink-reload.rst | 81 +++++++++++++++++++
>> Documentation/networking/devlink/index.rst | 1 +
>> 2 files changed, 82 insertions(+)
>> create mode 100644 Documentation/networking/devlink/devlink-reload.rst
>>
>> diff --git a/Documentation/networking/devlink/devlink-reload.rst b/Documentation/networking/devlink/devlink-reload.rst
>> new file mode 100644
>> index 000000000000..5abc5c2c75fd
>> --- /dev/null
>> +++ b/Documentation/networking/devlink/devlink-reload.rst
>> @@ -0,0 +1,81 @@
>> +.. SPDX-License-Identifier: GPL-2.0
>> +
>> +==============
>> +Devlink Reload
> No reason for capital "R".
It looks as the convention here for rst headers, as in
devlink-region.rst, devlink-trap.rst, devlink-resource.rst
>
>> +==============
>> +
>> +``devlink-reload`` provides mechanism to either reinit driver entities,
>> +applying ``devlink-params`` and ``devlink-resources`` new values or firmware
>> +activation depends on reload action selected.
> Could you perhaps split the sentense? It is hard to read.
Yes.
>
>> +
>> +Reload actions
>> +==============
>> +
>> +User may select a reload action.
>> +By default ``driver_reinit`` action is selected.
>> +
>> +.. list-table:: Possible reload actions
>> + :widths: 5 90
>> +
>> + * - Name
>> + - Description
>> + * - ``driver-reinit``
>> + - Devlink driver entities re-initialization, including applying
>> + new values to devlink entities which are used during driver
>> + load such as ``devlink-params`` in configuration mode
>> + ``driverinit`` or ``devlink-resources``
>> + * - ``fw_activate``
>> + - Firmware activate. Activates new firmware if such image is stored and
>> + pending activation. If no limitation specified this action may involve
>> + firmware reset. If no new image pending this action will reload current
>> + firmware image.
>> +
>> +Note that even though user asks for a specific action, the driver
>> +implementation might require to perform another action alongside with
>> +it. For example, some driver do not support driver reinitialization
>> +being performed without fw activation. Therefore, the devlink reload
>> +command returns the list of actions which were actrually performed.
>> +
>> +Reload limits
>> +=============
>> +
>> +By default reload actions are not limited and driver implementation may
>> +include reset or downtime as needed to perform the actions.
>> +
>> +However, some drivers support action limits, which limit the action
>> +implementation to specific constrains.
>> +
>> +.. list-table:: Possible reload limits
>> + :widths: 5 90
>> +
>> + * - Name
>> + - Description
>> + * - ``no_reset``
>> + - No reset allowed, no down time allowed, no link flap and no
>> + configuration is lost.
>> +
>> +Change namespace
>> +================
>> +
>> +The netns option allow user to be able to move devlink instances into
> "allows"
Ack.
>
>> +namespaces during devlink reload operation.
>> +By default all devlink instances are created in init_net and stay there.
>> +
>> +example usage
>> +-------------
>> +
>> +.. code:: shell
>> +
>> + $ devlink dev reload help
>> + $ devlink dev reload DEV [ netns { PID | NAME | ID } ] [ action { driver_reinit | fw_activate } ] [ limit no_reset ]
>> +
>> + # Run reload command for devlink driver entities re-initialization:
>> + $ devlink dev reload pci/0000:82:00.0 action driver_reinit
>> + reload_actions_performed:
>> + driver_reinit
>> +
>> + # Run reload command to activate firmware:
>> + # Note that mlx5 driver reloads the driver while activating firmware
>> + $ devlink dev reload pci/0000:82:00.0 action fw_activate
>> + reload_actions_performed:
>> + driver_reinit fw_activate
>> diff --git a/Documentation/networking/devlink/index.rst b/Documentation/networking/devlink/index.rst
>> index 7684ae5c4a4a..d82874760ae2 100644
>> --- a/Documentation/networking/devlink/index.rst
>> +++ b/Documentation/networking/devlink/index.rst
>> @@ -20,6 +20,7 @@ general.
>> devlink-params
>> devlink-region
>> devlink-resource
>> + devlink-reload
>> devlink-trap
>>
>> Driver-specific documentation
>> --
>> 2.18.2
>>
On 10/1/2020 6:59 AM, Moshe Shemesh wrote:
> Add devlink reload action to allow the user to request a specific reload
> action. The action parameter is optional, if not specified then devlink
> driver re-init action is used (backward compatible).
> Note that when required to do firmware activation some drivers may need
> to reload the driver. On the other hand some drivers may need to reset
> the firmware to reinitialize the driver entities. Therefore, the devlink
> reload command returns the actions which were actually performed.
> Reload actions supported are:
> driver_reinit: driver entities re-initialization, applying devlink-param
> and devlink-resource values.
> fw_activate: firmware activate.
>
> command examples:
> $devlink dev reload pci/0000:82:00.0 action driver_reinit
> reload_actions_performed:
> driver_reinit
>
> $devlink dev reload pci/0000:82:00.0 action fw_activate
> reload_actions_performed:
> driver_reinit fw_activate
>
> Signed-off-by: Moshe Shemesh <[email protected]>
Looks straight forward.
Reviewed-by: Jacob Keller <[email protected]>
> ---
> RFCv5 -> v1:
> - Rename supported_reload_actions to reload_actions.
> - Rename devlink_nl_reload_actions_performed_fill() to
> devlink_nl_reload_actions_performed_snd() and add genlmsg_reply() to it
> - Actions_performed sent to user space as a mask
> - Driver can initialize actions_performed before done as devlink ignores
> in case of failure
> - Use nla_poilcy range validation and remove the range check in
> devlink_nl_cmd_reload
> RFCv4 -> RFCv5:
> - Always pass actions_performed to unload_up() instead of checking in
> each driver
> - Verify returned actions_performed includes the requested action
> - Changed devlink_reload_actions_verify(devlink) to get ops
> - Changed devlink_reload_actions_verify() to return bool and rename to
> devlink_reload_actions_valid()
> - Only generate the reply if request uses new attributes
> RFCv3 -> RFCv4:
> - Removed fw_activate_no_reset as an action (next patch adds limit
> levels instead).
> - Renamed actions_done to actions_performed
> RFCv2 -> RFCv3:
> - Replace fw_live_patch action by fw_activate_no_reset
> - Devlink reload returns the actions done over netlink reply
> RFCv1 -> RFCv2:
> - Instead of reload levels driver,fw_reset,fw_live_patch have reload
> actions driver_reinit,fw_activate,fw_live_patch
> - Remove driver default level, the action driver_reinit is the default
> action for all drivers
> ---
> drivers/net/ethernet/mellanox/mlx4/main.c | 7 +-
> .../net/ethernet/mellanox/mlx5/core/devlink.c | 7 +-
> drivers/net/ethernet/mellanox/mlxsw/core.c | 11 +-
> drivers/net/netdevsim/dev.c | 8 +-
> include/net/devlink.h | 7 +-
> include/uapi/linux/devlink.h | 18 ++++
> net/core/devlink.c | 101 ++++++++++++++++--
> 7 files changed, 139 insertions(+), 20 deletions(-)
>
> diff --git a/drivers/net/ethernet/mellanox/mlx4/main.c b/drivers/net/ethernet/mellanox/mlx4/main.c
> index 70cf24ba71e4..a44d8b733db3 100644
> --- a/drivers/net/ethernet/mellanox/mlx4/main.c
> +++ b/drivers/net/ethernet/mellanox/mlx4/main.c
> @@ -3946,6 +3946,7 @@ static int mlx4_restart_one_up(struct pci_dev *pdev, bool reload,
> struct devlink *devlink);
>
> static int mlx4_devlink_reload_down(struct devlink *devlink, bool netns_change,
> + enum devlink_reload_action action,
> struct netlink_ext_ack *extack)
> {
I might have opted to convert reload_down to take a structure of
parameters, given that we're about to add 2 parameters. Not really that
significant either way.