From 2327a753b94e0556494d5499931f352a6ab1fc03 Mon Sep 17 00:00:00 2001 From: Hans Petter Selasky Date: Wed, 7 Mar 2018 14:35:28 +0000 Subject: [PATCH] Fix potential deadlock in command mode change in mlx5core. Call command completion handler in case of timeout when working in interrupts mode. Avoid flushing the commands workqueue after acquiring the semaphores to prevent a potential deadlock. linux commit commit 9cba4ebcf374c3772f6eb61f2d065294b2451b49 MFC after: 1 week Sponsored by: Mellanox Technologies --- sys/dev/mlx5/mlx5_core/mlx5_cmd.c | 77 ++++++++++++++----------------- 1 file changed, 34 insertions(+), 43 deletions(-) diff --git a/sys/dev/mlx5/mlx5_core/mlx5_cmd.c b/sys/dev/mlx5/mlx5_core/mlx5_cmd.c index 02d35664fd4..7988bebcb89 100644 --- a/sys/dev/mlx5/mlx5_core/mlx5_cmd.c +++ b/sys/dev/mlx5/mlx5_core/mlx5_cmd.c @@ -735,11 +735,6 @@ static void cmd_work_handler(struct work_struct *work) struct semaphore *sem; sem = ent->page_queue ? &cmd->pages_sem : &cmd->sem; - if (cmd->moving_to_polling) { - mlx5_core_warn(dev, "not expecting command execution, ignoring...\n"); - return; - } - down(sem); if (alloc_ent(ent) < 0) { @@ -826,12 +821,12 @@ static int wait_func(struct mlx5_core_dev *dev, struct mlx5_cmd_work_ent *ent) if (cmd->mode == CMD_MODE_POLLING) { wait_for_completion(&ent->done); err = ent->ret; - } else { - if (!wait_for_completion_timeout(&ent->done, timeout)) - err = -ETIMEDOUT; - else - err = 0; - } + } else if (!wait_for_completion_timeout(&ent->done, timeout)) { + ent->ret = -ETIMEDOUT; + mlx5_cmd_comp_handler(dev, 1UL << ent->idx); + } + + err = ent->ret; if (err == -ETIMEDOUT) { mlx5_core_warn(dev, "%s(0x%x) timeout. Will cause a leak of a command resource\n", @@ -881,26 +876,27 @@ static int mlx5_cmd_invoke(struct mlx5_core_dev *dev, struct mlx5_cmd_msg *in, goto out_free; } - if (!callback) { - err = wait_func(dev, ent); - if (err == -ETIMEDOUT) - goto out; + if (callback) + goto out; - ds = ent->ts2 - ent->ts1; - op = be16_to_cpu(((struct mlx5_inbox_hdr *)in->first.data)->opcode); - if (op < ARRAY_SIZE(cmd->stats)) { - stats = &cmd->stats[op]; - spin_lock_irq(&stats->lock); - stats->sum += ds; - ++stats->n; - spin_unlock_irq(&stats->lock); - } - mlx5_core_dbg_mask(dev, 1 << MLX5_CMD_TIME, - "fw exec time for %s is %lld nsec\n", - mlx5_command_str(op), (long long)ds); - *status = ent->status; - free_cmd(ent); - } + err = wait_func(dev, ent); + if (err == -ETIMEDOUT) + goto out; + + ds = ent->ts2 - ent->ts1; + op = be16_to_cpu(((struct mlx5_inbox_hdr *)in->first.data)->opcode); + if (op < ARRAY_SIZE(cmd->stats)) { + stats = &cmd->stats[op]; + spin_lock_irq(&stats->lock); + stats->sum += ds; + ++stats->n; + spin_unlock_irq(&stats->lock); + } + mlx5_core_dbg_mask(dev, 1 << MLX5_CMD_TIME, + "fw exec time for %s is %lld nsec\n", + mlx5_command_str(op), (long long)ds); + *status = ent->status; + free_cmd(ent); return err; @@ -1017,7 +1013,7 @@ static void clean_debug_files(struct mlx5_core_dev *dev) } -void mlx5_cmd_use_events(struct mlx5_core_dev *dev) +static void mlx5_cmd_change_mod(struct mlx5_core_dev *dev, int mode) { struct mlx5_cmd *cmd = &dev->cmd; int i; @@ -1026,26 +1022,21 @@ void mlx5_cmd_use_events(struct mlx5_core_dev *dev) down(&cmd->sem); down(&cmd->pages_sem); - - flush_workqueue(cmd->wq); - - cmd->mode = CMD_MODE_EVENTS; + cmd->mode = mode; up(&cmd->pages_sem); for (i = 0; i < cmd->max_reg_cmds; i++) up(&cmd->sem); } +void mlx5_cmd_use_events(struct mlx5_core_dev *dev) +{ + mlx5_cmd_change_mod(dev, CMD_MODE_EVENTS); +} + void mlx5_cmd_use_polling(struct mlx5_core_dev *dev) { - struct mlx5_cmd *cmd = &dev->cmd; - - synchronize_irq(dev->priv.eq_table.pages_eq.irqn); - flush_workqueue(dev->priv.pg_wq); - cmd->moving_to_polling = 1; - flush_workqueue(cmd->wq); - cmd->mode = CMD_MODE_POLLING; - cmd->moving_to_polling = 0; + mlx5_cmd_change_mod(dev, CMD_MODE_POLLING); } static void free_msg(struct mlx5_core_dev *dev, struct mlx5_cmd_msg *msg)