This patch is to add PCI error handler function support for mlx5. Created the functions for error_detected and slot_rest, plus will send a port down event to users when the driver error_detected function is invoked. This is to prevent a hang seeing in mcast_remove_one at the time ib_unregister_device is called for the ib_sa module. It will fail hardware commands while the driver is handling a PCI error. It will reduce the hardware commands timeout to 10 msecs so it does not hang waiting for an interrupt of the completion of the hardware command.
Signed-off-by: Carol Soto <[email protected]> --- drivers/infiniband/hw/mlx5/main.c | 32 +++++++++++++++++++++++++- drivers/net/ethernet/mellanox/mlx5/core/cmd.c | 7 +++++ include/linux/mlx5/driver.h | 4 +-- 3 files changed, 40 insertions(+), 3 deletions(-) Index: b/drivers/infiniband/hw/mlx5/main.c =================================================================== --- a/drivers/infiniband/hw/mlx5/main.c +++ b/drivers/infiniband/hw/mlx5/main.c @@ -1508,11 +1508,41 @@ static DEFINE_PCI_DEVICE_TABLE(mlx5_ib_p MODULE_DEVICE_TABLE(pci, mlx5_ib_pci_table); +static pci_ers_result_t mlx5_pci_err_detected(struct pci_dev *pdev, + pci_channel_state_t state) +{ + struct mlx5_ib_dev *dev = mlx5_pci2ibdev(pdev); + struct mlx5_core_dev *mdev = &dev->mdev; + u8 port; + + /* To avoid the mcast hang with ipoib up */ + for (port = 1; port <= dev->mdev.caps.num_ports; port++) + mlx5_ib_event(mdev, MLX5_DEV_EVENT_PORT_DOWN, &port); + + remove_one(pdev); + + return state == pci_channel_io_perm_failure ? + PCI_ERS_RESULT_DISCONNECT : PCI_ERS_RESULT_NEED_RESET; +} + +static pci_ers_result_t mlx5_pci_slot_reset(struct pci_dev *pdev) +{ + int ret = init_one(pdev, 0); + + return ret ? PCI_ERS_RESULT_DISCONNECT : PCI_ERS_RESULT_RECOVERED; +} + +static const struct pci_error_handlers mlx5_err_handler = { + .error_detected = mlx5_pci_err_detected, + .slot_reset = mlx5_pci_slot_reset, +}; + static struct pci_driver mlx5_ib_driver = { .name = DRIVER_NAME, .id_table = mlx5_ib_pci_table, .probe = init_one, - .remove = remove_one + .remove = remove_one, + .err_handler = &mlx5_err_handler, }; static int __init mlx5_ib_init(void) Index: b/drivers/net/ethernet/mellanox/mlx5/core/cmd.c =================================================================== --- a/drivers/net/ethernet/mellanox/mlx5/core/cmd.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/cmd.c @@ -646,6 +646,13 @@ static int mlx5_cmd_invoke(struct mlx5_c if (callback && page_queue) return -EINVAL; + if (pci_channel_offline(dev->pdev)) { + /* Device is going through error recovery + * and cannot accept commands. + */ + return -EIO; + } + ent = alloc_cmd(cmd, in, out, uout, uout_size, callback, context, page_queue); if (IS_ERR(ent)) Index: b/include/linux/mlx5/driver.h =================================================================== --- a/include/linux/mlx5/driver.h +++ b/include/linux/mlx5/driver.h @@ -51,10 +51,10 @@ enum { }; enum { - /* one minute for the sake of bringup. Generally, commands must always + /* 10 msecs for the sake of bringup. Generally, commands must always * complete and we may need to increase this timeout value */ - MLX5_CMD_TIMEOUT_MSEC = 7200 * 1000, + MLX5_CMD_TIMEOUT_MSEC = 10 * 1000, MLX5_CMD_WQ_MAX_NAME = 32, }; -- -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
