提交 7be85676 编写于 作者: S Sebastian Sanchez 提交者: Doug Ledford

IB/hfi1: Don't remove RB entry when not needed.

An RB tree is used for the SDMA pinning cache. Cache
entries are extracted and reinserted from the tree
in case the address range for it changes. However,
if the address range for the entry doesn't change,
deleting the entry from the RB tree is not necessary.

This affects performance since the tree needs to be
rebalanced for each insertion, and this happens in
the hot path. Optimize RB search by not removing
entries when it's not needed.
Reviewed-by: NMike Marciniszyn <mike.marciniszyn@intel.com>
Reviewed-by: NMitko Haralanov <mitko.haralanov@intel.com>
Signed-off-by: NSebastian Sanchez <sebastian.sanchez@intel.com>
Signed-off-by: NDennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: NDoug Ledford <dledford@redhat.com>
上级 14fe13fc
......@@ -217,21 +217,27 @@ static struct mmu_rb_node *__mmu_rb_search(struct mmu_rb_handler *handler,
return node;
}
struct mmu_rb_node *hfi1_mmu_rb_extract(struct mmu_rb_handler *handler,
unsigned long addr, unsigned long len)
bool hfi1_mmu_rb_remove_unless_exact(struct mmu_rb_handler *handler,
unsigned long addr, unsigned long len,
struct mmu_rb_node **rb_node)
{
struct mmu_rb_node *node;
unsigned long flags;
bool ret = false;
spin_lock_irqsave(&handler->lock, flags);
node = __mmu_rb_search(handler, addr, len);
if (node) {
if (node->addr == addr && node->len == len)
goto unlock;
__mmu_int_rb_remove(node, &handler->root);
list_del(&node->list); /* remove from LRU list */
ret = true;
}
unlock:
spin_unlock_irqrestore(&handler->lock, flags);
return node;
*rb_node = node;
return ret;
}
void hfi1_mmu_rb_evict(struct mmu_rb_handler *handler, void *evict_arg)
......
......@@ -81,7 +81,8 @@ int hfi1_mmu_rb_insert(struct mmu_rb_handler *handler,
void hfi1_mmu_rb_evict(struct mmu_rb_handler *handler, void *evict_arg);
void hfi1_mmu_rb_remove(struct mmu_rb_handler *handler,
struct mmu_rb_node *mnode);
struct mmu_rb_node *hfi1_mmu_rb_extract(struct mmu_rb_handler *handler,
unsigned long addr, unsigned long len);
bool hfi1_mmu_rb_remove_unless_exact(struct mmu_rb_handler *handler,
unsigned long addr, unsigned long len,
struct mmu_rb_node **rb_node);
#endif /* _HFI1_MMU_RB_H */
......@@ -1165,14 +1165,23 @@ static int pin_vector_pages(struct user_sdma_request *req,
struct hfi1_user_sdma_pkt_q *pq = req->pq;
struct sdma_mmu_node *node = NULL;
struct mmu_rb_node *rb_node;
rb_node = hfi1_mmu_rb_extract(pq->handler,
(unsigned long)iovec->iov.iov_base,
iovec->iov.iov_len);
if (rb_node)
bool extracted;
extracted =
hfi1_mmu_rb_remove_unless_exact(pq->handler,
(unsigned long)
iovec->iov.iov_base,
iovec->iov.iov_len, &rb_node);
if (rb_node) {
node = container_of(rb_node, struct sdma_mmu_node, rb);
else
rb_node = NULL;
if (!extracted) {
atomic_inc(&node->refcount);
iovec->pages = node->pages;
iovec->npages = node->npages;
iovec->node = node;
return 0;
}
}
if (!node) {
node = kzalloc(sizeof(*node), GFP_KERNEL);
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册