net/mlx5: Add retry mechanism to the command entry index allocation

It is possible that new command entry index allocation will temporarily
fail. The new command holds the semaphore, so it means that a free entry
should be ready soon. Add one second retry mechanism before returning an
error.

Patch "net/mlx5: Avoid possible free of command entry while timeout comp
handler" increase the possibility to bump into this temporarily failure
as it delays the entry index release for non-callback commands.

Fixes: 7c120481afcc ("mlx5: Add driver for Mellanox Connect-IB adapters")
Signed-off-by: Eran Ben Elisha <eranbe@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/cmd.c b/drivers/net/ethernet/mellanox/mlx5/core/cmd.c
index 37dae95e61d5fe8612a06671bf71db83a36ab512..2b597ac365f84672d2da86bf0c5edf985424177e 100644 (file)
--- a/drivers/net/ethernet/mellanox/mlx5/core/cmd.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/cmd.c
@@ -883,6 +883,25 @@ static bool opcode_allowed(struct mlx5_cmd *cmd, u16 opcode)
        return cmd->allowed_opcode == opcode;
 }
 
+static int cmd_alloc_index_retry(struct mlx5_cmd *cmd)
+{
+       unsigned long alloc_end = jiffies + msecs_to_jiffies(1000);
+       int idx;
+
+retry:
+       idx = cmd_alloc_index(cmd);
+       if (idx < 0 && time_before(jiffies, alloc_end)) {
+               /* Index allocation can fail on heavy load of commands. This is a temporary
+                * situation as the current command already holds the semaphore, meaning that
+                * another command completion is being handled and it is expected to release
+                * the entry index soon.
+                */
+               cpu_relax();
+               goto retry;
+       }
+       return idx;
+}
+
 static void cmd_work_handler(struct work_struct *work)
 {
        struct mlx5_cmd_work_ent *ent = container_of(work, struct mlx5_cmd_work_ent, work);
@@ -900,7 +919,7 @@ static void cmd_work_handler(struct work_struct *work)
        sem = ent->page_queue ? &cmd->pages_sem : &cmd->sem;
        down(sem);
        if (!ent->page_queue) {
-               alloc_ret = cmd_alloc_index(cmd);
+               alloc_ret = cmd_alloc_index_retry(cmd);
                if (alloc_ret < 0) {
                        mlx5_core_err_rl(dev, "failed to allocate command entry\n");
                        if (ent->callback) {