lib: move alignment attribute on types for MSVC

The current location used for __rte_aligned(a) for alignment of types
is not compatible with MSVC. There is only a single location accepted
by both toolchains.

The standard offers no alignment facility that compatibly interoperates
with C and C++ but it may be achieved by relocating the placement of
__rte_aligned(a) to the aforementioned location accepted by all currently
supported toolchains.

To allow alignment for both compilers, do the following:

* Expand __rte_aligned(a) to __declspec(align(a)) when building
  with MSVC.

* Move __rte_aligned from the end of {struct,union} definitions to
  be between {struct,union} and tag.

The placement between {struct,union} and the tag allows the desired
alignment to be imparted on the type regardless of the toolchain being
used for all of GCC, LLVM, MSVC compilers building both C and C++.

Note: this move has an additional benefit as Doxygen is not confused
anymore like for the rte_event_vector struct definition.

Signed-off-by: Tyler Retzlaff <[email protected]>
Acked-by: Morten Brørup <[email protected]>
Acked-by: Bruce Richardson <[email protected]>
Acked-by: Konstantin Ananyev <[email protected]>
Acked-by: Chengwen Feng <[email protected]>
Reviewed-by: Maxime Coquelin <[email protected]>
Signed-off-by: David Marchand <[email protected]>

Author: Tyler Retzlaff

devtools/checkpatches.sh

@@ -336,6 +336,21 @@ check_internal_tags() { # <patch>
 	return $res
 }
 
+check_aligned_attributes() { # <patch>
+	res=0
+
+	for token in __rte_aligned __rte_cache_aligned __rte_cache_min_aligned; do
+		if [ $(grep -E '^\+.*\<'$token'\>' "$1" | \
+				grep -vE '\<(struct|union)[[:space:]]*'$token'\>' | \
+				wc -l) != 0 ]; then
+			echo "Please use $token only for struct or union types alignment."
+			res=1
+		fi
+	done
+
+	return $res
+}
+
 check_release_notes() { # <patch>
 	rel_notes_prefix=doc/guides/rel_notes/release_
 	IFS=. read year month release < VERSION
@@ -445,6 +460,14 @@ check () { # <patch-file> <commit>
 		ret=1
 	fi
 
+	! $verbose || printf '\nChecking alignment attributes:\n'
+	report=$(check_aligned_attributes "$tmpinput")
+	if [ $? -ne 0 ] ; then
+		$headline_printed || print_headline "$subject"
+		printf '%s\n' "$report"
+		ret=1
+	fi
+
 	! $verbose || printf '\nChecking release notes updates:\n'
 	report=$(check_release_notes "$tmpinput")
 	if [ $? -ne 0 ] ; then

lib/acl/acl_run.h

@@ -55,12 +55,12 @@ struct acl_flow_data {
  * Structure to maintain running results for
  * a single packet (up to 4 tries).
  */
-struct completion {
+struct __rte_aligned(XMM_SIZE) completion {
 	uint32_t *results;                          /* running results. */
 	int32_t   priority[RTE_ACL_MAX_CATEGORIES]; /* running priorities. */
 	uint32_t  count;                            /* num of remaining tries */
 	/* true for allocated struct */
-} __rte_aligned(XMM_SIZE);
+};
 
 /*
  * One parms structure for each slot in the search engine.

lib/bpf/bpf_pkt.c

@@ -23,7 +23,7 @@
  * information about installed BPF rx/tx callback
  */
 
-struct bpf_eth_cbi {
+struct __rte_cache_aligned bpf_eth_cbi {
 	/* used by both data & control path */
 	RTE_ATOMIC(uint32_t) use;    /*usage counter */
 	const struct rte_eth_rxtx_callback *cb;  /* callback handle */
@@ -33,7 +33,7 @@ struct bpf_eth_cbi {
 	LIST_ENTRY(bpf_eth_cbi) link;
 	uint16_t port;
 	uint16_t queue;
-} __rte_cache_aligned;
+};
 
 /*
  * Odd number means that callback is used by datapath.

lib/compressdev/rte_comp.h

@@ -356,7 +356,7 @@ struct rte_comp_xform {
  * Comp operations are enqueued and dequeued in comp PMDs using the
  * rte_compressdev_enqueue_burst() / rte_compressdev_dequeue_burst() APIs
  */
-struct rte_comp_op {
+struct __rte_cache_aligned rte_comp_op {
 	enum rte_comp_op_type op_type;
 	union {
 		void *private_xform;
@@ -478,7 +478,7 @@ struct rte_comp_op {
 	 * will be set to RTE_COMP_OP_STATUS_SUCCESS after operation
 	 * is successfully processed by a PMD
 	 */
-} __rte_cache_aligned;
+};
 
 /**
  * Creates an operation pool

lib/compressdev/rte_compressdev_internal.h

@@ -69,7 +69,7 @@ typedef uint16_t (*compressdev_enqueue_pkt_burst_t)(void *qp,
 		struct rte_comp_op **ops, uint16_t nb_ops);
 
 /** The data structure associated with each comp device. */
-struct rte_compressdev {
+struct __rte_cache_aligned rte_compressdev {
 	compressdev_dequeue_pkt_burst_t dequeue_burst;
 	/**< Pointer to PMD receive function */
 	compressdev_enqueue_pkt_burst_t enqueue_burst;
@@ -87,7 +87,7 @@ struct rte_compressdev {
 	__extension__
 	uint8_t attached : 1;
 	/**< Flag indicating the device is attached */
-} __rte_cache_aligned;
+};
 
 /**
  *
@@ -96,7 +96,7 @@ struct rte_compressdev {
  * This structure is safe to place in shared memory to be common among
  * different processes in a multi-process configuration.
  */
-struct rte_compressdev_data {
+struct __rte_cache_aligned rte_compressdev_data {
 	uint8_t dev_id;
 	/**< Compress device identifier */
 	int socket_id;
@@ -115,7 +115,7 @@ struct rte_compressdev_data {
 
 	void *dev_private;
 	/**< PMD-specific private data */
-} __rte_cache_aligned;
+};
 
 #ifdef __cplusplus
 }

lib/cryptodev/cryptodev_pmd.h

@@ -61,7 +61,7 @@ struct rte_cryptodev_pmd_init_params {
  * This structure is safe to place in shared memory to be common among
  * different processes in a multi-process configuration.
  */
-struct rte_cryptodev_data {
+struct __rte_cache_aligned rte_cryptodev_data {
 	/** Device ID for this instance */
 	uint8_t dev_id;
 	/** Socket ID where memory is allocated */
@@ -82,10 +82,10 @@ struct rte_cryptodev_data {
 
 	/** PMD-specific private data */
 	void *dev_private;
-} __rte_cache_aligned;
+};
 
 /** @internal The data structure associated with each crypto device. */
-struct rte_cryptodev {
+struct __rte_cache_aligned rte_cryptodev {
 	/** Pointer to PMD dequeue function. */
 	dequeue_pkt_burst_t dequeue_burst;
 	/** Pointer to PMD enqueue function. */
@@ -117,7 +117,7 @@ struct rte_cryptodev {
 	struct rte_cryptodev_cb_rcu *enq_cbs;
 	/** User application callback for post dequeue processing */
 	struct rte_cryptodev_cb_rcu *deq_cbs;
-} __rte_cache_aligned;
+};
 
 /** Global structure used for maintaining state of allocated crypto devices */
 struct rte_cryptodev_global {

lib/cryptodev/rte_cryptodev_core.h

@@ -40,7 +40,7 @@ struct rte_cryptodev_qpdata {
 	struct rte_cryptodev_cb_rcu *deq_cb;
 };
 
-struct rte_crypto_fp_ops {
+struct __rte_cache_aligned rte_crypto_fp_ops {
 	/** PMD enqueue burst function. */
 	enqueue_pkt_burst_t enqueue_burst;
 	/** PMD dequeue burst function. */
@@ -49,7 +49,7 @@ struct rte_crypto_fp_ops {
 	struct rte_cryptodev_qpdata qp;
 	/** Reserved for future ops. */
 	uintptr_t reserved[3];
-} __rte_cache_aligned;
+};
 
 extern struct rte_crypto_fp_ops rte_crypto_fp_ops[RTE_CRYPTO_MAX_DEVS];
 

lib/dispatcher/rte_dispatcher.c

@@ -41,15 +41,15 @@ struct rte_dispatcher_finalizer {
 	void *finalize_data;
 };
 
-struct rte_dispatcher_lcore {
+struct __rte_cache_aligned rte_dispatcher_lcore {
 	uint8_t num_ports;
 	uint16_t num_handlers;
 	int32_t prio_count;
 	struct rte_dispatcher_lcore_port ports[EVD_MAX_PORTS_PER_LCORE];
 	struct rte_dispatcher_handler handlers[EVD_MAX_HANDLERS];
 	struct rte_dispatcher_stats stats;
 	RTE_CACHE_GUARD;
-} __rte_cache_aligned;
+};
 
 struct rte_dispatcher {
 	uint8_t event_dev_id;

lib/distributor/distributor_private.h

@@ -53,23 +53,23 @@
  * the next cache line to worker 0, we pad this out to three cache lines.
  * Only 64-bits of the memory is actually used though.
  */
-union rte_distributor_buffer_single {
+union __rte_cache_aligned rte_distributor_buffer_single {
 	volatile RTE_ATOMIC(int64_t) bufptr64;
 	char pad[RTE_CACHE_LINE_SIZE*3];
-} __rte_cache_aligned;
+};
 
 /*
  * Transfer up to 8 mbufs at a time to/from workers, and
  * flow matching algorithm optimized for 8 flow IDs at a time
  */
 #define RTE_DIST_BURST_SIZE 8
 
-struct rte_distributor_backlog {
+struct __rte_cache_aligned rte_distributor_backlog {
 	unsigned int start;
 	unsigned int count;
 	alignas(RTE_CACHE_LINE_SIZE) int64_t pkts[RTE_DIST_BURST_SIZE];
 	uint16_t *tags; /* will point to second cacheline of inflights */
-} __rte_cache_aligned;
+};
 
 
 struct rte_distributor_returned_pkts {

lib/dmadev/rte_dmadev_core.h

@@ -61,7 +61,7 @@ typedef uint16_t (*rte_dma_burst_capacity_t)(const void *dev_private, uint16_t v
  * The 'dev_private' field was placed in the first cache line to optimize
  * performance because the PMD mainly depends on this field.
  */
-struct rte_dma_fp_object {
+struct __rte_cache_aligned rte_dma_fp_object {
 	/** PMD-specific private data. The driver should copy
 	 * rte_dma_dev.data->dev_private to this field during initialization.
 	 */
@@ -73,7 +73,7 @@ struct rte_dma_fp_object {
 	rte_dma_completed_t        completed;
 	rte_dma_completed_status_t completed_status;
 	rte_dma_burst_capacity_t   burst_capacity;
-} __rte_cache_aligned;
+};
 
 extern struct rte_dma_fp_object *rte_dma_fp_objs;
 

lib/dmadev/rte_dmadev_pmd.h

@@ -94,7 +94,7 @@ struct rte_dma_dev_ops {
  *
  * @see struct rte_dma_dev::data
  */
-struct rte_dma_dev_data {
+struct __rte_cache_aligned rte_dma_dev_data {
 	char dev_name[RTE_DEV_NAME_MAX_LEN]; /**< Unique identifier name */
 	int16_t dev_id; /**< Device [external] identifier. */
 	int16_t numa_node; /**< Local NUMA memory ID. -1 if unknown. */
@@ -103,7 +103,7 @@ struct rte_dma_dev_data {
 	__extension__
 	uint8_t dev_started : 1; /**< Device state: STARTED(1)/STOPPED(0). */
 	uint64_t reserved[2]; /**< Reserved for future fields */
-} __rte_cache_aligned;
+};
 
 /**
  * Possible states of a DMA device.
@@ -122,7 +122,7 @@ enum rte_dma_dev_state {
  * @internal
  * The generic data structure associated with each DMA device.
  */
-struct rte_dma_dev {
+struct __rte_cache_aligned rte_dma_dev {
 	/** Device info which supplied during device initialization. */
 	struct rte_device *device;
 	struct rte_dma_dev_data *data; /**< Pointer to shared device data. */
@@ -132,7 +132,7 @@ struct rte_dma_dev {
 	const struct rte_dma_dev_ops *dev_ops;
 	enum rte_dma_dev_state state; /**< Flag indicating the device state. */
 	uint64_t reserved[2]; /**< Reserved for future fields. */
-} __rte_cache_aligned;
+};
 
 /**
  * @internal

lib/eal/arm/include/rte_vect.h

@@ -24,14 +24,14 @@ typedef int32x4_t xmm_t;
 #define	XMM_SIZE	(sizeof(xmm_t))
 #define	XMM_MASK	(XMM_SIZE - 1)
 
-typedef union rte_xmm {
+typedef union __rte_aligned(16) rte_xmm {
 	xmm_t    x;
 	uint8_t  u8[XMM_SIZE / sizeof(uint8_t)];
 	uint16_t u16[XMM_SIZE / sizeof(uint16_t)];
 	uint32_t u32[XMM_SIZE / sizeof(uint32_t)];
 	uint64_t u64[XMM_SIZE / sizeof(uint64_t)];
 	double   pd[XMM_SIZE / sizeof(double)];
-} __rte_aligned(16) rte_xmm_t;
+} rte_xmm_t;
 
 #if defined(RTE_ARCH_ARM) && defined(RTE_ARCH_32)
 /* NEON intrinsic vqtbl1q_u8() is not supported in ARMv7-A(AArch32) */

lib/eal/common/malloc_elem.h

@@ -20,7 +20,7 @@ enum elem_state {
 	ELEM_PAD  /* element is a padding-only header */
 };
 
-struct malloc_elem {
+struct __rte_cache_aligned malloc_elem {
 	struct malloc_heap *heap;
 	struct malloc_elem *volatile prev;
 	/**< points to prev elem in memseg */
@@ -48,7 +48,7 @@ struct malloc_elem {
 	size_t user_size;
 	uint64_t asan_cookie[2]; /* must be next to header_cookie */
 #endif
-} __rte_cache_aligned;
+};
 
 static const unsigned int MALLOC_ELEM_HEADER_LEN = sizeof(struct malloc_elem);
 

lib/eal/common/malloc_heap.h

@@ -21,7 +21,7 @@ struct malloc_elem;
 /**
  * Structure to hold malloc heap
  */
-struct malloc_heap {
+struct __rte_cache_aligned malloc_heap {
 	rte_spinlock_t lock;
 	LIST_HEAD(, malloc_elem) free_head[RTE_HEAP_NUM_FREELISTS];
 	struct malloc_elem *volatile first;
@@ -31,7 +31,7 @@ struct malloc_heap {
 	unsigned int socket_id;
 	size_t total_size;
 	char name[RTE_HEAP_NAME_MAX_LEN];
-} __rte_cache_aligned;
+};
 
 void *
 malloc_heap_alloc(const char *type, size_t size, int socket, unsigned int flags,

lib/eal/common/rte_random.c

@@ -13,14 +13,14 @@
 #include <rte_lcore.h>
 #include <rte_random.h>
 
-struct rte_rand_state {
+struct __rte_cache_aligned rte_rand_state {
 	uint64_t z1;
 	uint64_t z2;
 	uint64_t z3;
 	uint64_t z4;
 	uint64_t z5;
 	RTE_CACHE_GUARD;
-} __rte_cache_aligned;
+};
 
 /* One instance each for every lcore id-equipped thread, and one
  * additional instance to be shared by all others threads (i.e., all

lib/eal/common/rte_service.c

@@ -32,7 +32,7 @@
 #define RUNSTATE_RUNNING 1
 
 /* internal representation of a service */
-struct rte_service_spec_impl {
+struct __rte_cache_aligned rte_service_spec_impl {
 	/* public part of the struct */
 	struct rte_service_spec spec;
 
@@ -53,15 +53,15 @@ struct rte_service_spec_impl {
 	 * on currently.
 	 */
 	RTE_ATOMIC(uint32_t) num_mapped_cores;
-} __rte_cache_aligned;
+};
 
 struct service_stats {
 	RTE_ATOMIC(uint64_t) calls;
 	RTE_ATOMIC(uint64_t) cycles;
 };
 
 /* the internal values of a service core */
-struct core_state {
+struct __rte_cache_aligned core_state {
 	/* map of services IDs are run on this core */
 	uint64_t service_mask;
 	RTE_ATOMIC(uint8_t) runstate; /* running or stopped */
@@ -71,7 +71,7 @@ struct core_state {
 	RTE_ATOMIC(uint64_t) loops;
 	RTE_ATOMIC(uint64_t) cycles;
 	struct service_stats service_stats[RTE_SERVICE_NUM_MAX];
-} __rte_cache_aligned;
+};
 
 static uint32_t rte_service_count;
 static struct rte_service_spec_impl *rte_services;

lib/eal/include/generic/rte_atomic.h

@@ -1094,7 +1094,7 @@ static inline void rte_atomic64_clear(rte_atomic64_t *v)
 /**
  * 128-bit integer structure.
  */
-typedef struct {
+typedef struct __rte_aligned(16) {
 	union {
 		uint64_t val[2];
 #ifdef RTE_ARCH_64
@@ -1103,7 +1103,7 @@ typedef struct {
 #endif
 #endif
 	};
-} __rte_aligned(16) rte_int128_t;
+} rte_int128_t;
 
 #ifdef __DOXYGEN__