[linux-2.6-block.git] / fs / orangefs / orangefs-bufmap.c

/*
 * (C) 2001 Clemson University and The University of Chicago
 *
 * See COPYING in top-level directory.
 */
#include "protocol.h"
#include "orangefs-kernel.h"
#include "orangefs-bufmap.h"

DECLARE_WAIT_QUEUE_HEAD(orangefs_bufmap_init_waitq);

/* used to describe mapped buffers */
struct orangefs_bufmap_desc {
	void *uaddr;			/* user space address pointer */
	struct page **page_array;	/* array of mapped pages */
	int array_count;		/* size of above arrays */
	struct list_head list_link;
};

static struct orangefs_bufmap {
	atomic_t refcnt;

	int desc_size;
	int desc_shift;
	int desc_count;
	int total_size;
	int page_count;

	struct page **page_array;
	struct orangefs_bufmap_desc *desc_array;

	/* array to track usage of buffer descriptors */
	int *buffer_index_array;
	spinlock_t buffer_index_lock;

	/* array to track usage of buffer descriptors for readdir */
	int readdir_index_array[ORANGEFS_READDIR_DEFAULT_DESC_COUNT];
	spinlock_t readdir_index_lock;
} *__orangefs_bufmap;

static DEFINE_SPINLOCK(orangefs_bufmap_lock);

static void
orangefs_bufmap_unmap(struct orangefs_bufmap *bufmap)
{
	int i;

	for (i = 0; i < bufmap->page_count; i++)
		page_cache_release(bufmap->page_array[i]);
}

static void
orangefs_bufmap_free(struct orangefs_bufmap *bufmap)
{
	kfree(bufmap->page_array);
	kfree(bufmap->desc_array);
	kfree(bufmap->buffer_index_array);
	kfree(bufmap);
}

static struct orangefs_bufmap *orangefs_bufmap_ref(void)
{
	struct orangefs_bufmap *bufmap = NULL;

	spin_lock(&orangefs_bufmap_lock);
	if (__orangefs_bufmap) {
		bufmap = __orangefs_bufmap;
		atomic_inc(&bufmap->refcnt);
	}
	spin_unlock(&orangefs_bufmap_lock);
	return bufmap;
}

static void orangefs_bufmap_unref(struct orangefs_bufmap *bufmap)
{
	if (atomic_dec_and_lock(&bufmap->refcnt, &orangefs_bufmap_lock)) {
		__orangefs_bufmap = NULL;
		spin_unlock(&orangefs_bufmap_lock);

		orangefs_bufmap_unmap(bufmap);
		orangefs_bufmap_free(bufmap);
	}
}

inline int orangefs_bufmap_size_query(void)
{
	struct orangefs_bufmap *bufmap = orangefs_bufmap_ref();
	int size = bufmap ? bufmap->desc_size : 0;

	orangefs_bufmap_unref(bufmap);
	return size;
}

inline int orangefs_bufmap_shift_query(void)
{
	struct orangefs_bufmap *bufmap = orangefs_bufmap_ref();
	int shift = bufmap ? bufmap->desc_shift : 0;

	orangefs_bufmap_unref(bufmap);
	return shift;
}

static DECLARE_WAIT_QUEUE_HEAD(bufmap_waitq);
static DECLARE_WAIT_QUEUE_HEAD(readdir_waitq);

/*
 * get_bufmap_init
 *
 * If bufmap_init is 1, then the shared memory system, including the
 * buffer_index_array, is available.  Otherwise, it is not.
 *
 * returns the value of bufmap_init
 */
int get_bufmap_init(void)
{
	return __orangefs_bufmap ? 1 : 0;
}


static struct orangefs_bufmap *
orangefs_bufmap_alloc(struct ORANGEFS_dev_map_desc *user_desc)
{
	struct orangefs_bufmap *bufmap;

	bufmap = kzalloc(sizeof(*bufmap), GFP_KERNEL);
	if (!bufmap)
		goto out;

	atomic_set(&bufmap->refcnt, 1);
	bufmap->total_size = user_desc->total_size;
	bufmap->desc_count = user_desc->count;
	bufmap->desc_size = user_desc->size;
	bufmap->desc_shift = ilog2(bufmap->desc_size);

	spin_lock_init(&bufmap->buffer_index_lock);
	bufmap->buffer_index_array =
		kcalloc(bufmap->desc_count, sizeof(int), GFP_KERNEL);
	if (!bufmap->buffer_index_array) {
		gossip_err("orangefs: could not allocate %d buffer indices\n",
				bufmap->desc_count);
		goto out_free_bufmap;
	}
	spin_lock_init(&bufmap->readdir_index_lock);

	bufmap->desc_array =
		kcalloc(bufmap->desc_count, sizeof(struct orangefs_bufmap_desc),
			GFP_KERNEL);
	if (!bufmap->desc_array) {
		gossip_err("orangefs: could not allocate %d descriptors\n",
				bufmap->desc_count);
		goto out_free_index_array;
	}

	bufmap->page_count = bufmap->total_size / PAGE_SIZE;

	/* allocate storage to track our page mappings */
	bufmap->page_array =
		kcalloc(bufmap->page_count, sizeof(struct page *), GFP_KERNEL);
	if (!bufmap->page_array)
		goto out_free_desc_array;

	return bufmap;

out_free_desc_array:
	kfree(bufmap->desc_array);
out_free_index_array:
	kfree(bufmap->buffer_index_array);
out_free_bufmap:
	kfree(bufmap);
out:
	return NULL;
}

static int
orangefs_bufmap_map(struct orangefs_bufmap *bufmap,
		struct ORANGEFS_dev_map_desc *user_desc)
{
	int pages_per_desc = bufmap->desc_size / PAGE_SIZE;
	int offset = 0, ret, i;

	/* map the pages */
	ret = get_user_pages_fast((unsigned long)user_desc->ptr,
			     bufmap->page_count, 1, bufmap->page_array);

	if (ret < 0)
		return ret;

	if (ret != bufmap->page_count) {
		gossip_err("orangefs error: asked for %d pages, only got %d.\n",
				bufmap->page_count, ret);

		for (i = 0; i < ret; i++) {
			SetPageError(bufmap->page_array[i]);
			page_cache_release(bufmap->page_array[i]);
		}
		return -ENOMEM;
	}

	/*
	 * ideally we want to get kernel space pointers for each page, but
	 * we can't kmap that many pages at once if highmem is being used.
	 * so instead, we just kmap/kunmap the page address each time the
	 * kaddr is needed.
	 */
	for (i = 0; i < bufmap->page_count; i++)
		flush_dcache_page(bufmap->page_array[i]);

	/* build a list of available descriptors */
	for (offset = 0, i = 0; i < bufmap->desc_count; i++) {
		bufmap->desc_array[i].page_array = &bufmap->page_array[offset];
		bufmap->desc_array[i].array_count = pages_per_desc;
		bufmap->desc_array[i].uaddr =
		    (user_desc->ptr + (i * pages_per_desc * PAGE_SIZE));
		offset += pages_per_desc;
	}

	return 0;
}

/*
 * orangefs_bufmap_initialize()
 *
 * initializes the mapped buffer interface
 *
 * returns 0 on success, -errno on failure
 */
int orangefs_bufmap_initialize(struct ORANGEFS_dev_map_desc *user_desc)
{
	struct orangefs_bufmap *bufmap;
	int ret = -EINVAL;

	gossip_debug(GOSSIP_BUFMAP_DEBUG,
		     "orangefs_bufmap_initialize: called (ptr ("
		     "%p) sz (%d) cnt(%d).\n",
		     user_desc->ptr,
		     user_desc->size,
		     user_desc->count);

	/*
	 * sanity check alignment and size of buffer that caller wants to
	 * work with
	 */
	if (PAGE_ALIGN((unsigned long)user_desc->ptr) !=
	    (unsigned long)user_desc->ptr) {
		gossip_err("orangefs error: memory alignment (front). %p\n",
			   user_desc->ptr);
		goto out;
	}

	if (PAGE_ALIGN(((unsigned long)user_desc->ptr + user_desc->total_size))
	    != (unsigned long)(user_desc->ptr + user_desc->total_size)) {
		gossip_err("orangefs error: memory alignment (back).(%p + %d)\n",
			   user_desc->ptr,
			   user_desc->total_size);
		goto out;
	}

	if (user_desc->total_size != (user_desc->size * user_desc->count)) {
		gossip_err("orangefs error: user provided an oddly sized buffer: (%d, %d, %d)\n",
			   user_desc->total_size,
			   user_desc->size,
			   user_desc->count);
		goto out;
	}

	if ((user_desc->size % PAGE_SIZE) != 0) {
		gossip_err("orangefs error: bufmap size not page size divisible (%d).\n",
			   user_desc->size);
		goto out;
	}

	ret = -ENOMEM;
	bufmap = orangefs_bufmap_alloc(user_desc);
	if (!bufmap)
		goto out;

	ret = orangefs_bufmap_map(bufmap, user_desc);
	if (ret)
		goto out_free_bufmap;


	spin_lock(&orangefs_bufmap_lock);
	if (__orangefs_bufmap) {
		spin_unlock(&orangefs_bufmap_lock);
		gossip_err("orangefs: error: bufmap already initialized.\n");
		ret = -EALREADY;
		goto out_unmap_bufmap;
	}
	__orangefs_bufmap = bufmap;
	spin_unlock(&orangefs_bufmap_lock);

	/*
	 * If there are operations in orangefs_bufmap_init_waitq, wake them up.
	 * This scenario occurs when the client-core is restarted and I/O
	 * requests in the in-progress or waiting tables are restarted.  I/O
	 * requests cannot be restarted until the shared memory system is
	 * completely re-initialized, so we put the I/O requests in this
	 * waitq until initialization has completed.  NOTE:  the I/O requests
	 * are also on a timer, so they don't wait forever just in case the
	 * client-core doesn't come back up.
	 */
	wake_up_interruptible(&orangefs_bufmap_init_waitq);

	gossip_debug(GOSSIP_BUFMAP_DEBUG,
		     "orangefs_bufmap_initialize: exiting normally\n");
	return 0;

out_unmap_bufmap:
	orangefs_bufmap_unmap(bufmap);
out_free_bufmap:
	orangefs_bufmap_free(bufmap);
out:
	return ret;
}

/*
 * orangefs_bufmap_finalize()
 *
 * shuts down the mapped buffer interface and releases any resources
 * associated with it
 *
 * no return value
 */
void orangefs_bufmap_finalize(void)
{
	gossip_debug(GOSSIP_BUFMAP_DEBUG, "orangefs_bufmap_finalize: called\n");
	BUG_ON(!__orangefs_bufmap);
	orangefs_bufmap_unref(__orangefs_bufmap);
	gossip_debug(GOSSIP_BUFMAP_DEBUG,
		     "orangefs_bufmap_finalize: exiting normally\n");
}

struct slot_args {
	int slot_count;
	int *slot_array;
	spinlock_t *slot_lock;
	wait_queue_head_t *slot_wq;
};

static int wait_for_a_slot(struct slot_args *slargs, int *buffer_index)
{
	int ret = -1;
	int i = 0;
	DEFINE_WAIT(wait_entry);

	while (1) {
		/*
		 * check for available desc, slot_lock is the appropriate
		 * index_lock
		 */
		spin_lock(slargs->slot_lock);
		prepare_to_wait_exclusive(slargs->slot_wq,
					  &wait_entry,
					  TASK_INTERRUPTIBLE);
		for (i = 0; i < slargs->slot_count; i++)
			if (slargs->slot_array[i] == 0) {
				slargs->slot_array[i] = 1;
				*buffer_index = i;
				ret = 0;
				break;
			}
		spin_unlock(slargs->slot_lock);

		/* if we acquired a buffer, then break out of while */
		if (ret == 0)
			break;

		if (!signal_pending(current)) {
			int timeout =
			    MSECS_TO_JIFFIES(1000 * slot_timeout_secs);
			gossip_debug(GOSSIP_BUFMAP_DEBUG,
				     "[BUFMAP]: waiting %d "
				     "seconds for a slot\n",
				     slot_timeout_secs);
			if (!schedule_timeout(timeout)) {
				gossip_debug(GOSSIP_BUFMAP_DEBUG,
					     "*** wait_for_a_slot timed out\n");
				ret = -ETIMEDOUT;
				break;
			}
			gossip_debug(GOSSIP_BUFMAP_DEBUG,
			  "[BUFMAP]: woken up by a slot becoming available.\n");
			continue;
		}

		gossip_debug(GOSSIP_BUFMAP_DEBUG, "orangefs: %s interrupted.\n",
			     __func__);
		ret = -EINTR;
		break;
	}

	spin_lock(slargs->slot_lock);
	finish_wait(slargs->slot_wq, &wait_entry);
	spin_unlock(slargs->slot_lock);
	return ret;
}

static void put_back_slot(struct slot_args *slargs, int buffer_index)
{
	/* slot_lock is the appropriate index_lock */
	spin_lock(slargs->slot_lock);
	if (buffer_index < 0 || buffer_index >= slargs->slot_count) {
		spin_unlock(slargs->slot_lock);
		return;
	}

	/* put the desc back on the queue */
	slargs->slot_array[buffer_index] = 0;
	spin_unlock(slargs->slot_lock);

	/* wake up anyone who may be sleeping on the queue */
	wake_up_interruptible(slargs->slot_wq);
}

/*
 * orangefs_bufmap_get()
 *
 * gets a free mapped buffer descriptor, will sleep until one becomes
 * available if necessary
 *
 * returns 0 on success, -errno on failure
 */
int orangefs_bufmap_get(struct orangefs_bufmap **mapp, int *buffer_index)
{
	struct orangefs_bufmap *bufmap = orangefs_bufmap_ref();
	struct slot_args slargs;
	int ret;

	if (!bufmap) {
		gossip_err("orangefs: please confirm that pvfs2-client daemon is running.\n");
		return -EIO;
	}

	slargs.slot_count = bufmap->desc_count;
	slargs.slot_array = bufmap->buffer_index_array;
	slargs.slot_lock = &bufmap->buffer_index_lock;
	slargs.slot_wq = &bufmap_waitq;
	ret = wait_for_a_slot(&slargs, buffer_index);
	if (ret)
		orangefs_bufmap_unref(bufmap);
	*mapp = bufmap;
	return ret;
}

/*
 * orangefs_bufmap_put()
 *
 * returns a mapped buffer descriptor to the collection
 *
 * no return value
 */
void orangefs_bufmap_put(struct orangefs_bufmap *bufmap, int buffer_index)
{
	struct slot_args slargs;

	slargs.slot_count = bufmap->desc_count;
	slargs.slot_array = bufmap->buffer_index_array;
	slargs.slot_lock = &bufmap->buffer_index_lock;
	slargs.slot_wq = &bufmap_waitq;
	put_back_slot(&slargs, buffer_index);
	orangefs_bufmap_unref(bufmap);
}

/*
 * readdir_index_get()
 *
 * gets a free descriptor, will sleep until one becomes
 * available if necessary.
 * Although the readdir buffers are not mapped into kernel space
 * we could do that at a later point of time. Regardless, these
 * indices are used by the client-core.
 *
 * returns 0 on success, -errno on failure
 */
int readdir_index_get(struct orangefs_bufmap **mapp, int *buffer_index)
{
	struct orangefs_bufmap *bufmap = orangefs_bufmap_ref();
	struct slot_args slargs;
	int ret;

	if (!bufmap) {
		gossip_err("orangefs: please confirm that pvfs2-client daemon is running.\n");
		return -EIO;
	}

	slargs.slot_count = ORANGEFS_READDIR_DEFAULT_DESC_COUNT;
	slargs.slot_array = bufmap->readdir_index_array;
	slargs.slot_lock = &bufmap->readdir_index_lock;
	slargs.slot_wq = &readdir_waitq;
	ret = wait_for_a_slot(&slargs, buffer_index);
	if (ret)
		orangefs_bufmap_unref(bufmap);
	*mapp = bufmap;
	return ret;
}

void readdir_index_put(struct orangefs_bufmap *bufmap, int buffer_index)
{
	struct slot_args slargs;

	slargs.slot_count = ORANGEFS_READDIR_DEFAULT_DESC_COUNT;
	slargs.slot_array = bufmap->readdir_index_array;
	slargs.slot_lock = &bufmap->readdir_index_lock;
	slargs.slot_wq = &readdir_waitq;
	put_back_slot(&slargs, buffer_index);
	orangefs_bufmap_unref(bufmap);
}

/*
 * we've been handed an iovec, we need to copy it to 
 * the shared memory descriptor at "buffer_index".
 */
int orangefs_bufmap_copy_from_iovec(struct orangefs_bufmap *bufmap,
				struct iov_iter *iter,
				int buffer_index,
				size_t size)
{
	struct orangefs_bufmap_desc *to = &bufmap->desc_array[buffer_index];
	int i;

	gossip_debug(GOSSIP_BUFMAP_DEBUG,
		     "%s: buffer_index:%d: size:%zu:\n",
		     __func__, buffer_index, size);


	for (i = 0; size; i++) {
		struct page *page = to->page_array[i];
		size_t n = size;
		if (n > PAGE_SIZE)
			n = PAGE_SIZE;
		n = copy_page_from_iter(page, 0, n, iter);
		if (!n)
			return -EFAULT;
		size -= n;
	}
	return 0;

}

/*
 * we've been handed an iovec, we need to fill it from
 * the shared memory descriptor at "buffer_index".
 */
int orangefs_bufmap_copy_to_iovec(struct orangefs_bufmap *bufmap,
				    struct iov_iter *iter,
				    int buffer_index,
				    size_t size)
{
	struct orangefs_bufmap_desc *from = &bufmap->desc_array[buffer_index];
	int i;

	gossip_debug(GOSSIP_BUFMAP_DEBUG,
		     "%s: buffer_index:%d: size:%zu:\n",
		     __func__, buffer_index, size);


	for (i = 0; size; i++) {
		struct page *page = from->page_array[i];
		size_t n = size;
		if (n > PAGE_SIZE)
			n = PAGE_SIZE;
		n = copy_page_to_iter(page, 0, n, iter);
		if (!n)
			return -EFAULT;
		size -= n;
	}
	return 0;
}
Commit	Line	Data
274dcf55 MM	1	/*
	2	* (C) 2001 Clemson University and The University of Chicago
	3	*
	4	* See COPYING in top-level directory.
	5	*/
	6	#include "protocol.h"
575e9461 MM	7	#include "orangefs-kernel.h"
575e9461 MM	8	#include "orangefs-bufmap.h"
274dcf55	9
8bb8aefd	10	DECLARE_WAIT_QUEUE_HEAD(orangefs_bufmap_init_waitq);
274dcf55	11
bf89f584 MB	12	/* used to describe mapped buffers */
	13	struct orangefs_bufmap_desc {
	14	void uaddr; / user space address pointer */
	15	struct page *page_array; / array of mapped pages */
	16	int array_count; /* size of above arrays */
	17	struct list_head list_link;
	18	};
	19
8bb8aefd	20	static struct orangefs_bufmap {
274dcf55 MM	21	atomic_t refcnt;
	22
	23	int desc_size;
	24	int desc_shift;
	25	int desc_count;
	26	int total_size;
	27	int page_count;
	28
	29	struct page **page_array;
8bb8aefd	30	struct orangefs_bufmap_desc *desc_array;
274dcf55 MM	31
	32	/* array to track usage of buffer descriptors */
	33	int *buffer_index_array;
	34	spinlock_t buffer_index_lock;
	35
	36	/* array to track usage of buffer descriptors for readdir */
8bb8aefd	37	int readdir_index_array[ORANGEFS_READDIR_DEFAULT_DESC_COUNT];
274dcf55	38	spinlock_t readdir_index_lock;
8bb8aefd	39	} *__orangefs_bufmap;
274dcf55	40
8bb8aefd	41	static DEFINE_SPINLOCK(orangefs_bufmap_lock);
274dcf55 MM	42
274dcf55 MM	43	static void
8bb8aefd	44	orangefs_bufmap_unmap(struct orangefs_bufmap *bufmap)
274dcf55 MM	45	{
	46	int i;
	47
	48	for (i = 0; i < bufmap->page_count; i++)
	49	page_cache_release(bufmap->page_array[i]);
	50	}
	51
	52	static void
8bb8aefd	53	orangefs_bufmap_free(struct orangefs_bufmap *bufmap)
274dcf55 MM	54	{
	55	kfree(bufmap->page_array);
	56	kfree(bufmap->desc_array);
	57	kfree(bufmap->buffer_index_array);
	58	kfree(bufmap);
	59	}
	60
bf89f584	61	static struct orangefs_bufmap *orangefs_bufmap_ref(void)
274dcf55	62	{
8bb8aefd	63	struct orangefs_bufmap *bufmap = NULL;
274dcf55	64
8bb8aefd YL	65	spin_lock(&orangefs_bufmap_lock);
	66	if (__orangefs_bufmap) {
	67	bufmap = __orangefs_bufmap;
274dcf55 MM	68	atomic_inc(&bufmap->refcnt);
274dcf55 MM	69	}
8bb8aefd	70	spin_unlock(&orangefs_bufmap_lock);
274dcf55 MM	71	return bufmap;
	72	}
	73
bf89f584	74	static void orangefs_bufmap_unref(struct orangefs_bufmap *bufmap)
274dcf55	75	{
8bb8aefd YL	76	if (atomic_dec_and_lock(&bufmap->refcnt, &orangefs_bufmap_lock)) {
	77	__orangefs_bufmap = NULL;
	78	spin_unlock(&orangefs_bufmap_lock);
274dcf55	79
8bb8aefd YL	80	orangefs_bufmap_unmap(bufmap);
8bb8aefd YL	81	orangefs_bufmap_free(bufmap);
274dcf55 MM	82	}
	83	}
	84
8bb8aefd	85	inline int orangefs_bufmap_size_query(void)
274dcf55	86	{
8bb8aefd	87	struct orangefs_bufmap *bufmap = orangefs_bufmap_ref();
274dcf55 MM	88	int size = bufmap ? bufmap->desc_size : 0;
274dcf55 MM	89
8bb8aefd	90	orangefs_bufmap_unref(bufmap);
274dcf55 MM	91	return size;
	92	}
	93
8bb8aefd	94	inline int orangefs_bufmap_shift_query(void)
274dcf55	95	{
8bb8aefd	96	struct orangefs_bufmap *bufmap = orangefs_bufmap_ref();
274dcf55 MM	97	int shift = bufmap ? bufmap->desc_shift : 0;
274dcf55 MM	98
8bb8aefd	99	orangefs_bufmap_unref(bufmap);
274dcf55 MM	100	return shift;
	101	}
	102
	103	static DECLARE_WAIT_QUEUE_HEAD(bufmap_waitq);
	104	static DECLARE_WAIT_QUEUE_HEAD(readdir_waitq);
	105
	106	/*
	107	* get_bufmap_init
	108	*
	109	* If bufmap_init is 1, then the shared memory system, including the
	110	* buffer_index_array, is available. Otherwise, it is not.
	111	*
	112	* returns the value of bufmap_init
	113	*/
	114	int get_bufmap_init(void)
	115	{
8bb8aefd	116	return __orangefs_bufmap ? 1 : 0;
274dcf55 MM	117	}
	118
	119
8bb8aefd YL	120	static struct orangefs_bufmap *
8bb8aefd YL	121	orangefs_bufmap_alloc(struct ORANGEFS_dev_map_desc *user_desc)
274dcf55	122	{
8bb8aefd	123	struct orangefs_bufmap *bufmap;
274dcf55 MM	124
	125	bufmap = kzalloc(sizeof(*bufmap), GFP_KERNEL);
	126	if (!bufmap)
	127	goto out;
	128
	129	atomic_set(&bufmap->refcnt, 1);
	130	bufmap->total_size = user_desc->total_size;
	131	bufmap->desc_count = user_desc->count;
	132	bufmap->desc_size = user_desc->size;
	133	bufmap->desc_shift = ilog2(bufmap->desc_size);
	134
	135	spin_lock_init(&bufmap->buffer_index_lock);
	136	bufmap->buffer_index_array =
	137	kcalloc(bufmap->desc_count, sizeof(int), GFP_KERNEL);
	138	if (!bufmap->buffer_index_array) {
8bb8aefd	139	gossip_err("orangefs: could not allocate %d buffer indices\n",
274dcf55 MM	140	bufmap->desc_count);
	141	goto out_free_bufmap;
	142	}
	143	spin_lock_init(&bufmap->readdir_index_lock);
	144
	145	bufmap->desc_array =
8bb8aefd	146	kcalloc(bufmap->desc_count, sizeof(struct orangefs_bufmap_desc),
274dcf55 MM	147	GFP_KERNEL);
274dcf55 MM	148	if (!bufmap->desc_array) {
8bb8aefd	149	gossip_err("orangefs: could not allocate %d descriptors\n",
274dcf55 MM	150	bufmap->desc_count);
	151	goto out_free_index_array;
	152	}
	153
	154	bufmap->page_count = bufmap->total_size / PAGE_SIZE;
	155
	156	/* allocate storage to track our page mappings */
	157	bufmap->page_array =
	158	kcalloc(bufmap->page_count, sizeof(struct page *), GFP_KERNEL);
	159	if (!bufmap->page_array)
	160	goto out_free_desc_array;
	161
	162	return bufmap;
	163
	164	out_free_desc_array:
	165	kfree(bufmap->desc_array);
	166	out_free_index_array:
	167	kfree(bufmap->buffer_index_array);
	168	out_free_bufmap:
	169	kfree(bufmap);
	170	out:
	171	return NULL;
	172	}
	173
	174	static int
8bb8aefd YL	175	orangefs_bufmap_map(struct orangefs_bufmap *bufmap,
8bb8aefd YL	176	struct ORANGEFS_dev_map_desc *user_desc)
274dcf55 MM	177	{
	178	int pages_per_desc = bufmap->desc_size / PAGE_SIZE;
	179	int offset = 0, ret, i;
	180
	181	/* map the pages */
16742f2d AV	182	ret = get_user_pages_fast((unsigned long)user_desc->ptr,
16742f2d AV	183	bufmap->page_count, 1, bufmap->page_array);
274dcf55 MM	184
	185	if (ret < 0)
	186	return ret;
	187
	188	if (ret != bufmap->page_count) {
8bb8aefd	189	gossip_err("orangefs error: asked for %d pages, only got %d.\n",
274dcf55 MM	190	bufmap->page_count, ret);
	191
	192	for (i = 0; i < ret; i++) {
	193	SetPageError(bufmap->page_array[i]);
	194	page_cache_release(bufmap->page_array[i]);
	195	}
	196	return -ENOMEM;
	197	}
	198
	199	/*
	200	* ideally we want to get kernel space pointers for each page, but
	201	* we can't kmap that many pages at once if highmem is being used.
	202	* so instead, we just kmap/kunmap the page address each time the
	203	* kaddr is needed.
	204	*/
	205	for (i = 0; i < bufmap->page_count; i++)
	206	flush_dcache_page(bufmap->page_array[i]);
	207
	208	/* build a list of available descriptors */
	209	for (offset = 0, i = 0; i < bufmap->desc_count; i++) {
	210	bufmap->desc_array[i].page_array = &bufmap->page_array[offset];
	211	bufmap->desc_array[i].array_count = pages_per_desc;
	212	bufmap->desc_array[i].uaddr =
	213	(user_desc->ptr + (i * pages_per_desc * PAGE_SIZE));
	214	offset += pages_per_desc;
	215	}
	216
	217	return 0;
	218	}
	219
	220	/*
8bb8aefd	221	* orangefs_bufmap_initialize()
274dcf55 MM	222	*
	223	* initializes the mapped buffer interface
	224	*
	225	* returns 0 on success, -errno on failure
	226	*/
8bb8aefd	227	int orangefs_bufmap_initialize(struct ORANGEFS_dev_map_desc *user_desc)
274dcf55	228	{
8bb8aefd	229	struct orangefs_bufmap *bufmap;
274dcf55 MM	230	int ret = -EINVAL;
	231
	232	gossip_debug(GOSSIP_BUFMAP_DEBUG,
8bb8aefd	233	"orangefs_bufmap_initialize: called (ptr ("
274dcf55 MM	234	"%p) sz (%d) cnt(%d).\n",
	235	user_desc->ptr,
	236	user_desc->size,
	237	user_desc->count);
	238
	239	/*
	240	* sanity check alignment and size of buffer that caller wants to
	241	* work with
	242	*/
	243	if (PAGE_ALIGN((unsigned long)user_desc->ptr) !=
	244	(unsigned long)user_desc->ptr) {
8bb8aefd	245	gossip_err("orangefs error: memory alignment (front). %p\n",
274dcf55 MM	246	user_desc->ptr);
	247	goto out;
	248	}
	249
	250	if (PAGE_ALIGN(((unsigned long)user_desc->ptr + user_desc->total_size))
	251	!= (unsigned long)(user_desc->ptr + user_desc->total_size)) {
8bb8aefd	252	gossip_err("orangefs error: memory alignment (back).(%p + %d)\n",
274dcf55 MM	253	user_desc->ptr,
	254	user_desc->total_size);
	255	goto out;
	256	}
	257
	258	if (user_desc->total_size != (user_desc->size * user_desc->count)) {
8bb8aefd	259	gossip_err("orangefs error: user provided an oddly sized buffer: (%d, %d, %d)\n",
274dcf55 MM	260	user_desc->total_size,
	261	user_desc->size,
	262	user_desc->count);
	263	goto out;
	264	}
	265
	266	if ((user_desc->size % PAGE_SIZE) != 0) {
8bb8aefd	267	gossip_err("orangefs error: bufmap size not page size divisible (%d).\n",
274dcf55 MM	268	user_desc->size);
	269	goto out;
	270	}
	271
	272	ret = -ENOMEM;
8bb8aefd	273	bufmap = orangefs_bufmap_alloc(user_desc);
274dcf55 MM	274	if (!bufmap)
	275	goto out;
	276
8bb8aefd	277	ret = orangefs_bufmap_map(bufmap, user_desc);
274dcf55 MM	278	if (ret)
	279	goto out_free_bufmap;
	280
	281
8bb8aefd YL	282	spin_lock(&orangefs_bufmap_lock);
	283	if (__orangefs_bufmap) {
	284	spin_unlock(&orangefs_bufmap_lock);
	285	gossip_err("orangefs: error: bufmap already initialized.\n");
274dcf55 MM	286	ret = -EALREADY;
	287	goto out_unmap_bufmap;
	288	}
8bb8aefd YL	289	__orangefs_bufmap = bufmap;
8bb8aefd YL	290	spin_unlock(&orangefs_bufmap_lock);
274dcf55 MM	291
274dcf55 MM	292	/*
8bb8aefd	293	* If there are operations in orangefs_bufmap_init_waitq, wake them up.
274dcf55 MM	294	* This scenario occurs when the client-core is restarted and I/O
	295	* requests in the in-progress or waiting tables are restarted. I/O
	296	* requests cannot be restarted until the shared memory system is
	297	* completely re-initialized, so we put the I/O requests in this
	298	* waitq until initialization has completed. NOTE: the I/O requests
	299	* are also on a timer, so they don't wait forever just in case the
	300	* client-core doesn't come back up.
	301	*/
8bb8aefd	302	wake_up_interruptible(&orangefs_bufmap_init_waitq);
274dcf55 MM	303
274dcf55 MM	304	gossip_debug(GOSSIP_BUFMAP_DEBUG,
8bb8aefd	305	"orangefs_bufmap_initialize: exiting normally\n");
274dcf55 MM	306	return 0;
	307
	308	out_unmap_bufmap:
8bb8aefd	309	orangefs_bufmap_unmap(bufmap);
274dcf55	310	out_free_bufmap:
8bb8aefd	311	orangefs_bufmap_free(bufmap);
274dcf55 MM	312	out:
	313	return ret;
	314	}
	315
	316	/*
8bb8aefd	317	* orangefs_bufmap_finalize()
274dcf55 MM	318	*
	319	* shuts down the mapped buffer interface and releases any resources
	320	* associated with it
	321	*
	322	* no return value
	323	*/
8bb8aefd	324	void orangefs_bufmap_finalize(void)
274dcf55	325	{
8bb8aefd YL	326	gossip_debug(GOSSIP_BUFMAP_DEBUG, "orangefs_bufmap_finalize: called\n");
	327	BUG_ON(!__orangefs_bufmap);
	328	orangefs_bufmap_unref(__orangefs_bufmap);
274dcf55	329	gossip_debug(GOSSIP_BUFMAP_DEBUG,
8bb8aefd	330	"orangefs_bufmap_finalize: exiting normally\n");
274dcf55 MM	331	}
	332
	333	struct slot_args {
	334	int slot_count;
	335	int *slot_array;
	336	spinlock_t *slot_lock;
	337	wait_queue_head_t *slot_wq;
	338	};
	339
	340	static int wait_for_a_slot(struct slot_args slargs, int buffer_index)
	341	{
	342	int ret = -1;
	343	int i = 0;
ce6c414e	344	DEFINE_WAIT(wait_entry);
274dcf55 MM	345
274dcf55 MM	346	while (1) {
274dcf55 MM	347	/*
	348	* check for available desc, slot_lock is the appropriate
	349	* index_lock
	350	*/
	351	spin_lock(slargs->slot_lock);
ce6c414e MM	352	prepare_to_wait_exclusive(slargs->slot_wq,
	353	&wait_entry,
	354	TASK_INTERRUPTIBLE);
274dcf55 MM	355	for (i = 0; i < slargs->slot_count; i++)
	356	if (slargs->slot_array[i] == 0) {
	357	slargs->slot_array[i] = 1;
	358	*buffer_index = i;
	359	ret = 0;
	360	break;
	361	}
	362	spin_unlock(slargs->slot_lock);
	363
	364	/* if we acquired a buffer, then break out of while */
	365	if (ret == 0)
	366	break;
	367
	368	if (!signal_pending(current)) {
	369	int timeout =
	370	MSECS_TO_JIFFIES(1000 * slot_timeout_secs);
	371	gossip_debug(GOSSIP_BUFMAP_DEBUG,
	372	"[BUFMAP]: waiting %d "
	373	"seconds for a slot\n",
	374	slot_timeout_secs);
	375	if (!schedule_timeout(timeout)) {
	376	gossip_debug(GOSSIP_BUFMAP_DEBUG,
	377	"*** wait_for_a_slot timed out\n");
	378	ret = -ETIMEDOUT;
	379	break;
	380	}
	381	gossip_debug(GOSSIP_BUFMAP_DEBUG,
	382	"[BUFMAP]: woken up by a slot becoming available.\n");
	383	continue;
	384	}
	385
8bb8aefd	386	gossip_debug(GOSSIP_BUFMAP_DEBUG, "orangefs: %s interrupted.\n",
274dcf55 MM	387	__func__);
	388	ret = -EINTR;
	389	break;
	390	}
	391
ce6c414e MM	392	spin_lock(slargs->slot_lock);
	393	finish_wait(slargs->slot_wq, &wait_entry);
	394	spin_unlock(slargs->slot_lock);
274dcf55 MM	395	return ret;
	396	}
	397
	398	static void put_back_slot(struct slot_args *slargs, int buffer_index)
	399	{
	400	/* slot_lock is the appropriate index_lock */
	401	spin_lock(slargs->slot_lock);
	402	if (buffer_index < 0 \|\| buffer_index >= slargs->slot_count) {
	403	spin_unlock(slargs->slot_lock);
	404	return;
	405	}
	406
	407	/* put the desc back on the queue */
	408	slargs->slot_array[buffer_index] = 0;
	409	spin_unlock(slargs->slot_lock);
	410
	411	/* wake up anyone who may be sleeping on the queue */
	412	wake_up_interruptible(slargs->slot_wq);
	413	}
	414
	415	/*
8bb8aefd	416	* orangefs_bufmap_get()
274dcf55 MM	417	*
	418	* gets a free mapped buffer descriptor, will sleep until one becomes
	419	* available if necessary
	420	*
	421	* returns 0 on success, -errno on failure
	422	*/
8bb8aefd	423	int orangefs_bufmap_get(struct orangefs_bufmap *mapp, int buffer_index)
274dcf55	424	{
8bb8aefd	425	struct orangefs_bufmap *bufmap = orangefs_bufmap_ref();
274dcf55 MM	426	struct slot_args slargs;
	427	int ret;
	428
	429	if (!bufmap) {
8bb8aefd	430	gossip_err("orangefs: please confirm that pvfs2-client daemon is running.\n");
274dcf55 MM	431	return -EIO;
	432	}
	433
	434	slargs.slot_count = bufmap->desc_count;
	435	slargs.slot_array = bufmap->buffer_index_array;
	436	slargs.slot_lock = &bufmap->buffer_index_lock;
	437	slargs.slot_wq = &bufmap_waitq;
	438	ret = wait_for_a_slot(&slargs, buffer_index);
	439	if (ret)
8bb8aefd	440	orangefs_bufmap_unref(bufmap);
274dcf55 MM	441	*mapp = bufmap;
	442	return ret;
	443	}
	444
	445	/*
8bb8aefd	446	* orangefs_bufmap_put()
274dcf55 MM	447	*
	448	* returns a mapped buffer descriptor to the collection
	449	*
	450	* no return value
	451	*/
8bb8aefd	452	void orangefs_bufmap_put(struct orangefs_bufmap *bufmap, int buffer_index)
274dcf55 MM	453	{
	454	struct slot_args slargs;
	455
	456	slargs.slot_count = bufmap->desc_count;
	457	slargs.slot_array = bufmap->buffer_index_array;
	458	slargs.slot_lock = &bufmap->buffer_index_lock;
	459	slargs.slot_wq = &bufmap_waitq;
	460	put_back_slot(&slargs, buffer_index);
8bb8aefd	461	orangefs_bufmap_unref(bufmap);
274dcf55 MM	462	}
	463
	464	/*
	465	* readdir_index_get()
	466	*
	467	* gets a free descriptor, will sleep until one becomes
	468	* available if necessary.
	469	* Although the readdir buffers are not mapped into kernel space
	470	* we could do that at a later point of time. Regardless, these
	471	* indices are used by the client-core.
	472	*
	473	* returns 0 on success, -errno on failure
	474	*/
8bb8aefd	475	int readdir_index_get(struct orangefs_bufmap *mapp, int buffer_index)
274dcf55	476	{
8bb8aefd	477	struct orangefs_bufmap *bufmap = orangefs_bufmap_ref();
274dcf55 MM	478	struct slot_args slargs;
	479	int ret;
	480
	481	if (!bufmap) {
8bb8aefd	482	gossip_err("orangefs: please confirm that pvfs2-client daemon is running.\n");
274dcf55 MM	483	return -EIO;
	484	}
	485
8bb8aefd	486	slargs.slot_count = ORANGEFS_READDIR_DEFAULT_DESC_COUNT;
274dcf55 MM	487	slargs.slot_array = bufmap->readdir_index_array;
	488	slargs.slot_lock = &bufmap->readdir_index_lock;
	489	slargs.slot_wq = &readdir_waitq;
	490	ret = wait_for_a_slot(&slargs, buffer_index);
	491	if (ret)
8bb8aefd	492	orangefs_bufmap_unref(bufmap);
274dcf55 MM	493	*mapp = bufmap;
	494	return ret;
	495	}
	496
8bb8aefd	497	void readdir_index_put(struct orangefs_bufmap *bufmap, int buffer_index)
274dcf55 MM	498	{
	499	struct slot_args slargs;
	500
8bb8aefd	501	slargs.slot_count = ORANGEFS_READDIR_DEFAULT_DESC_COUNT;
274dcf55 MM	502	slargs.slot_array = bufmap->readdir_index_array;
	503	slargs.slot_lock = &bufmap->readdir_index_lock;
	504	slargs.slot_wq = &readdir_waitq;
	505	put_back_slot(&slargs, buffer_index);
8bb8aefd	506	orangefs_bufmap_unref(bufmap);
274dcf55 MM	507	}
274dcf55 MM	508
b5e376ea MM	509	/*
	510	* we've been handed an iovec, we need to copy it to
	511	* the shared memory descriptor at "buffer_index".
	512	*/
8bb8aefd	513	int orangefs_bufmap_copy_from_iovec(struct orangefs_bufmap *bufmap,
54804949 MM	514	struct iov_iter *iter,
	515	int buffer_index,
	516	size_t size)
274dcf55	517	{
8bb8aefd	518	struct orangefs_bufmap_desc *to = &bufmap->desc_array[buffer_index];
4d1c4404	519	int i;
274dcf55 MM	520
274dcf55 MM	521	gossip_debug(GOSSIP_BUFMAP_DEBUG,
34204fde	522	"%s: buffer_index:%d: size:%zu:\n",
4d1c4404	523	__func__, buffer_index, size);
274dcf55	524
274dcf55	525
4d1c4404	526	for (i = 0; size; i++) {
34204fde AV	527	struct page *page = to->page_array[i];
	528	size_t n = size;
	529	if (n > PAGE_SIZE)
	530	n = PAGE_SIZE;
	531	n = copy_page_from_iter(page, 0, n, iter);
	532	if (!n)
	533	return -EFAULT;
	534	size -= n;
274dcf55	535	}
34204fde	536	return 0;
274dcf55	537
274dcf55 MM	538	}
	539
	540	/*
b5e376ea MM	541	* we've been handed an iovec, we need to fill it from
b5e376ea MM	542	* the shared memory descriptor at "buffer_index".
274dcf55	543	*/
8bb8aefd	544	int orangefs_bufmap_copy_to_iovec(struct orangefs_bufmap *bufmap,
4d1c4404	545	struct iov_iter *iter,
5c278228 AV	546	int buffer_index,
5c278228 AV	547	size_t size)
274dcf55	548	{
8bb8aefd	549	struct orangefs_bufmap_desc *from = &bufmap->desc_array[buffer_index];
4d1c4404	550	int i;
274dcf55 MM	551
274dcf55 MM	552	gossip_debug(GOSSIP_BUFMAP_DEBUG,
5c278228 AV	553	"%s: buffer_index:%d: size:%zu:\n",
5c278228 AV	554	__func__, buffer_index, size);
274dcf55	555
274dcf55	556
5c278228 AV	557	for (i = 0; size; i++) {
	558	struct page *page = from->page_array[i];
	559	size_t n = size;
	560	if (n > PAGE_SIZE)
	561	n = PAGE_SIZE;
	562	n = copy_page_to_iter(page, 0, n, iter);
	563	if (!n)
	564	return -EFAULT;
	565	size -= n;
274dcf55	566	}
5c278228	567	return 0;
274dcf55	568	}