Revision 7b5967389f5a8dfb9d32843830f5e2717e20995d authored by Thomas Falcon on 08 December 2016, 22:40:03 UTC, committed by David S. Miller on 10 December 2016, 03:47:22 UTC
This patch is based on an earlier one submitted
by Jon Maxwell with the following commit message:

"We recently encountered a bug where a few customers using ibmveth on the
same LPAR hit an issue where a TCP session hung when large receive was
enabled. Closer analysis revealed that the session was stuck because the
one side was advertising a zero window repeatedly.

We narrowed this down to the fact the ibmveth driver did not set gso_size
which is translated by TCP into the MSS later up the stack. The MSS is
used to calculate the TCP window size and as that was abnormally large,
it was calculating a zero window, even although the sockets receive buffer
was completely empty."

We rely on the Virtual I/O Server partition in a pseries
environment to provide the MSS through the TCP header checksum
field. The stipulation is that users should not disable checksum
offloading if rx packet aggregation is enabled through VIOS.

Some firmware offerings provide the MSS in the RX buffer.
This is signalled by a bit in the RX queue descriptor.

Reviewed-by: Brian King <brking@linux.vnet.ibm.com>
Reviewed-by: Pradeep Satyanarayana <pradeeps@linux.vnet.ibm.com>
Reviewed-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Reviewed-by: Jonathan Maxwell <jmaxwell37@gmail.com>
Reviewed-by: David Dai <zdai@us.ibm.com>
Signed-off-by: Thomas Falcon <tlfalcon@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
1 parent 1472d59
Raw File
list_sort.c

#define pr_fmt(fmt) "list_sort_test: " fmt

#include <linux/kernel.h>
#include <linux/bug.h>
#include <linux/compiler.h>
#include <linux/export.h>
#include <linux/string.h>
#include <linux/list_sort.h>
#include <linux/list.h>

#define MAX_LIST_LENGTH_BITS 20

/*
 * Returns a list organized in an intermediate format suited
 * to chaining of merge() calls: null-terminated, no reserved or
 * sentinel head node, "prev" links not maintained.
 */
static struct list_head *merge(void *priv,
				int (*cmp)(void *priv, struct list_head *a,
					struct list_head *b),
				struct list_head *a, struct list_head *b)
{
	struct list_head head, *tail = &head;

	while (a && b) {
		/* if equal, take 'a' -- important for sort stability */
		if ((*cmp)(priv, a, b) <= 0) {
			tail->next = a;
			a = a->next;
		} else {
			tail->next = b;
			b = b->next;
		}
		tail = tail->next;
	}
	tail->next = a?:b;
	return head.next;
}

/*
 * Combine final list merge with restoration of standard doubly-linked
 * list structure.  This approach duplicates code from merge(), but
 * runs faster than the tidier alternatives of either a separate final
 * prev-link restoration pass, or maintaining the prev links
 * throughout.
 */
static void merge_and_restore_back_links(void *priv,
				int (*cmp)(void *priv, struct list_head *a,
					struct list_head *b),
				struct list_head *head,
				struct list_head *a, struct list_head *b)
{
	struct list_head *tail = head;
	u8 count = 0;

	while (a && b) {
		/* if equal, take 'a' -- important for sort stability */
		if ((*cmp)(priv, a, b) <= 0) {
			tail->next = a;
			a->prev = tail;
			a = a->next;
		} else {
			tail->next = b;
			b->prev = tail;
			b = b->next;
		}
		tail = tail->next;
	}
	tail->next = a ? : b;

	do {
		/*
		 * In worst cases this loop may run many iterations.
		 * Continue callbacks to the client even though no
		 * element comparison is needed, so the client's cmp()
		 * routine can invoke cond_resched() periodically.
		 */
		if (unlikely(!(++count)))
			(*cmp)(priv, tail->next, tail->next);

		tail->next->prev = tail;
		tail = tail->next;
	} while (tail->next);

	tail->next = head;
	head->prev = tail;
}

/**
 * list_sort - sort a list
 * @priv: private data, opaque to list_sort(), passed to @cmp
 * @head: the list to sort
 * @cmp: the elements comparison function
 *
 * This function implements "merge sort", which has O(nlog(n))
 * complexity.
 *
 * The comparison function @cmp must return a negative value if @a
 * should sort before @b, and a positive value if @a should sort after
 * @b. If @a and @b are equivalent, and their original relative
 * ordering is to be preserved, @cmp must return 0.
 */
void list_sort(void *priv, struct list_head *head,
		int (*cmp)(void *priv, struct list_head *a,
			struct list_head *b))
{
	struct list_head *part[MAX_LIST_LENGTH_BITS+1]; /* sorted partial lists
						-- last slot is a sentinel */
	int lev;  /* index into part[] */
	int max_lev = 0;
	struct list_head *list;

	if (list_empty(head))
		return;

	memset(part, 0, sizeof(part));

	head->prev->next = NULL;
	list = head->next;

	while (list) {
		struct list_head *cur = list;
		list = list->next;
		cur->next = NULL;

		for (lev = 0; part[lev]; lev++) {
			cur = merge(priv, cmp, part[lev], cur);
			part[lev] = NULL;
		}
		if (lev > max_lev) {
			if (unlikely(lev >= ARRAY_SIZE(part)-1)) {
				printk_once(KERN_DEBUG "list too long for efficiency\n");
				lev--;
			}
			max_lev = lev;
		}
		part[lev] = cur;
	}

	for (lev = 0; lev < max_lev; lev++)
		if (part[lev])
			list = merge(priv, cmp, part[lev], list);

	merge_and_restore_back_links(priv, cmp, head, part[max_lev], list);
}
EXPORT_SYMBOL(list_sort);

#ifdef CONFIG_TEST_LIST_SORT

#include <linux/slab.h>
#include <linux/random.h>

/*
 * The pattern of set bits in the list length determines which cases
 * are hit in list_sort().
 */
#define TEST_LIST_LEN (512+128+2) /* not including head */

#define TEST_POISON1 0xDEADBEEF
#define TEST_POISON2 0xA324354C

struct debug_el {
	unsigned int poison1;
	struct list_head list;
	unsigned int poison2;
	int value;
	unsigned serial;
};

/* Array, containing pointers to all elements in the test list */
static struct debug_el **elts __initdata;

static int __init check(struct debug_el *ela, struct debug_el *elb)
{
	if (ela->serial >= TEST_LIST_LEN) {
		pr_err("error: incorrect serial %d\n", ela->serial);
		return -EINVAL;
	}
	if (elb->serial >= TEST_LIST_LEN) {
		pr_err("error: incorrect serial %d\n", elb->serial);
		return -EINVAL;
	}
	if (elts[ela->serial] != ela || elts[elb->serial] != elb) {
		pr_err("error: phantom element\n");
		return -EINVAL;
	}
	if (ela->poison1 != TEST_POISON1 || ela->poison2 != TEST_POISON2) {
		pr_err("error: bad poison: %#x/%#x\n",
			ela->poison1, ela->poison2);
		return -EINVAL;
	}
	if (elb->poison1 != TEST_POISON1 || elb->poison2 != TEST_POISON2) {
		pr_err("error: bad poison: %#x/%#x\n",
			elb->poison1, elb->poison2);
		return -EINVAL;
	}
	return 0;
}

static int __init cmp(void *priv, struct list_head *a, struct list_head *b)
{
	struct debug_el *ela, *elb;

	ela = container_of(a, struct debug_el, list);
	elb = container_of(b, struct debug_el, list);

	check(ela, elb);
	return ela->value - elb->value;
}

static int __init list_sort_test(void)
{
	int i, count = 1, err = -ENOMEM;
	struct debug_el *el;
	struct list_head *cur;
	LIST_HEAD(head);

	pr_debug("start testing list_sort()\n");

	elts = kcalloc(TEST_LIST_LEN, sizeof(*elts), GFP_KERNEL);
	if (!elts) {
		pr_err("error: cannot allocate memory\n");
		return err;
	}

	for (i = 0; i < TEST_LIST_LEN; i++) {
		el = kmalloc(sizeof(*el), GFP_KERNEL);
		if (!el) {
			pr_err("error: cannot allocate memory\n");
			goto exit;
		}
		 /* force some equivalencies */
		el->value = prandom_u32() % (TEST_LIST_LEN / 3);
		el->serial = i;
		el->poison1 = TEST_POISON1;
		el->poison2 = TEST_POISON2;
		elts[i] = el;
		list_add_tail(&el->list, &head);
	}

	list_sort(NULL, &head, cmp);

	err = -EINVAL;
	for (cur = head.next; cur->next != &head; cur = cur->next) {
		struct debug_el *el1;
		int cmp_result;

		if (cur->next->prev != cur) {
			pr_err("error: list is corrupted\n");
			goto exit;
		}

		cmp_result = cmp(NULL, cur, cur->next);
		if (cmp_result > 0) {
			pr_err("error: list is not sorted\n");
			goto exit;
		}

		el = container_of(cur, struct debug_el, list);
		el1 = container_of(cur->next, struct debug_el, list);
		if (cmp_result == 0 && el->serial >= el1->serial) {
			pr_err("error: order of equivalent elements not "
				"preserved\n");
			goto exit;
		}

		if (check(el, el1)) {
			pr_err("error: element check failed\n");
			goto exit;
		}
		count++;
	}
	if (head.prev != cur) {
		pr_err("error: list is corrupted\n");
		goto exit;
	}


	if (count != TEST_LIST_LEN) {
		pr_err("error: bad list length %d", count);
		goto exit;
	}

	err = 0;
exit:
	for (i = 0; i < TEST_LIST_LEN; i++)
		kfree(elts[i]);
	kfree(elts);
	return err;
}
late_initcall(list_sort_test);
#endif /* CONFIG_TEST_LIST_SORT */
back to top