Re: [dpdk-dev] [PATCH v9 06/10] eal: add thread lifetime management

[Dmitry Kozlyuk <dmitry.kozliuk@gmail.com>]

2021-06-04 16:44 (UTC-0700), Narcisa Ana Maria Vasile:
[...]
> diff --git a/lib/eal/include/rte_thread.h b/lib/eal/include/rte_thread.h
> index 5c54cd9d67..1d481b9ad5 100644
> --- a/lib/eal/include/rte_thread.h
> +++ b/lib/eal/include/rte_thread.h
> @@ -208,6 +208,73 @@ __rte_experimental
>  int rte_thread_attr_set_priority(rte_thread_attr_t *thread_attr,
>  		enum rte_thread_priority priority);
>  
> +/**
> + * Create a new thread that will invoke the 'thread_func' routine.
> + *
> + * @param thread_id
> + *    A pointer that will store the id of the newly created thread.
> + *
> + * @param thread_attr
> + *    Attributes that are used at the creation of the new thread.
> + *
> + * @param thread_func
> + *    The routine that the new thread will invoke when starting execution.
> + *
> + * @param args
> + *    Arguments to be passed to the 'thread_func' routine.
> + *
> + * @return
> + *   On success, return 0.
> + *   On failure, return a positive errno-style error number.
> + */
> +__rte_experimental
> +int rte_thread_create(rte_thread_t *thread_id,
> +		      const rte_thread_attr_t *thread_attr,
> +		      void *(*thread_func)(void *), void *args);

1. Thread function prototype is used at least in 4 places,
maybe give it a name, like `rte_thread_func`?

2. We can't easily support it's `void*` return type on Windows,
because it doesn't fit in DWORD. In `rte_thread_join` below you use `int`.
All `pthread_join` usages in DPDK ignore return value, but I'd rather keep it.
Do you think it's OK to stick to `int`?

[...]
> +/**
> + * Terminates a thread.
> + *
> + * @param thread_id
> + *    The id of the thread to be cancelled.
> + *
> + * @return
> + *   On success, return 0.
> + *   On failure, return a positive errno-style error number.
> + */
> +__rte_experimental
> +int rte_thread_cancel(rte_thread_t thread_id);

What do you think of making this function internal for now?
We don't want applications to rely on this prototype.
To hide it from Doxygen, `/*` comment or #ifndef __DOXYGEN__ can be used.
It is worth noting in commit message
that it's not implemented for Windows and why.

> +
> +/**
> + * Detaches a thread.

Please explain what it means, because detaching is a pthread concept.

> + *
> + * @param thread_id
> + *    The id of the thread to be cancelled.

Not "cancelled".

> + *
> + * @return
> + *   On success, return 0.
> + *   On failure, return a positive errno-style error number.
> + */
> +__rte_experimental
> +int rte_thread_detach(rte_thread_t thread_id);
> +
> +

Redundant empty line.

>  #ifdef RTE_HAS_CPUSET
>  
>  /**
> diff --git a/lib/eal/windows/rte_thread.c b/lib/eal/windows/rte_thread.c
> index 6dc3d575c0..5afdd54e15 100644
> --- a/lib/eal/windows/rte_thread.c
> +++ b/lib/eal/windows/rte_thread.c
> @@ -14,6 +14,11 @@ struct eal_tls_key {
>  	DWORD thread_index;
>  };
>  
> +struct thread_routine_ctx {
> +	void* (*start_routine) (void*);
> +	void *routine_args;
> +};
> +
>  /* Translates the most common error codes related to threads */
>  static int
>  thread_translate_win32_error(DWORD error)
> @@ -346,6 +351,163 @@ rte_thread_attr_set_priority(rte_thread_attr_t *thread_attr,
>  	return 0;
>  }
>  
> +static DWORD
> +thread_func_wrapper(void *args)
> +{
> +	struct thread_routine_ctx *pctx = args;
> +	intptr_t func_ret = 0;
> +	struct thread_routine_ctx ctx = { NULL, NULL };
> +
> +	ctx.start_routine = pctx->start_routine;
> +	ctx.routine_args = pctx->routine_args;
> +
> +	free(pctx);
> +
> +	func_ret = (intptr_t)ctx.start_routine(ctx.routine_args);
> +	return (DWORD)func_ret;
> +}
> +
> +int
> +rte_thread_create(rte_thread_t *thread_id,
> +		  const rte_thread_attr_t *thread_attr,
> +		  void *(*thread_func)(void *), void *args)
> +{
> +	int ret = 0;
> +	DWORD tid = 0;
> +	HANDLE thread_handle = NULL;
> +	GROUP_AFFINITY thread_affinity;
> +	struct thread_routine_ctx *ctx = NULL;
> +
> +	ctx = calloc(1, sizeof(*ctx));

Why use `calloc()` for a scalar?

[...]
> +int
> +rte_thread_cancel(rte_thread_t thread_id)
> +{
> +	int ret = 0;
> +	HANDLE thread_handle = NULL;
> +
> +	thread_handle = OpenThread(THREAD_TERMINATE, FALSE, thread_id.opaque_id);
> +	if (thread_handle == NULL) {
> +		ret = thread_log_last_error("OpenThread()");
> +		goto cleanup;
> +	}
> +
> +	/*
> +	 * TODO: Behavior is different between POSIX and Windows threads.
> +	 * POSIX threads wait for a cancellation point.
> +	 * Current Windows emulation kills thread at any point.
> +	 */
> +	ret = TerminateThread(thread_handle, 0);
> +	if (ret != 0) {
> +		ret = thread_log_last_error("TerminateThread()");
> +		goto cleanup;
> +	}
> +
> +cleanup:
> +	if (thread_handle != NULL) {
> +		CloseHandle(thread_handle);
> +		thread_handle = NULL;
> +	}
> +	return ret;
> +}

As we've discussed before, such implementation should never be used.
I suggest removing it for now, otherwise if someone enables the code
calling rte_thread_cancel() for Windows, it will almost certainly lead
to deadlock bugs.

> +
> +int
> +rte_thread_detach(rte_thread_t thread_id)
> +{
> +	(void)thread_id;

RTE_SET_USED/__rte_unused

> +	return ENOTSUP;
> +}
> +

It should return success as the thread is in detached state after the call.

>  int
>  rte_thread_key_create(rte_thread_key *key,
>  		__rte_unused void (*destructor)(void *))