(we've fixed it in 2.6.23-rc3, let's get the same fix into -mm)
The layout for struct ioat_desc_sw is non-optimal and causes an extra
cache hit for every descriptor processed. By tightening up the struct
layout and removing one item, we pull in the fields that get used in
the speedpath and get a little better performance.
Before:
-------
struct ioat_desc_sw {
struct ioat_dma_descriptor * hw; /* 0 8 */
struct list_head node; /* 8 16 */
int tx_cnt; /* 24 4 */
/* XXX 4 bytes hole, try to pack */
dma_addr_t src; /* 32 8 */
__u32 src_len; /* 40 4 */
/* XXX 4 bytes hole, try to pack */
dma_addr_t dst; /* 48 8 */
__u32 dst_len; /* 56 4 */
/* XXX 4 bytes hole, try to pack */
/* --- cacheline 1 boundary (64 bytes) --- */
struct dma_async_tx_descriptor async_tx; /* 64 144 */
/* --- cacheline 3 boundary (192 bytes) was 16 bytes ago --- */
/* size: 208, cachelines: 4 */
/* sum members: 196, holes: 3, sum holes: 12 */
/* last cacheline: 16 bytes */
}; /* definitions: 1 */
After:
------
struct ioat_desc_sw {
struct ioat_dma_descriptor * hw; /* 0 8 */
struct list_head node; /* 8 16 */
int tx_cnt; /* 24 4 */
__u32 len; /* 28 4 */
dma_addr_t src; /* 32 8 */
dma_addr_t dst; /* 40 8 */
struct dma_async_tx_descriptor async_tx; /* 48 144 */
/* --- cacheline 3 boundary (192 bytes) --- */
/* size: 192, cachelines: 3 */
}; /* definitions: 1 */
Signed-off-by: Shannon Nelson <[email protected]>
---
drivers/dma/ioat_dma.c | 9 ++++-----
drivers/dma/ioatdma.h | 3 +--
2 files changed, 5 insertions(+), 7 deletions(-)
diff --git a/drivers/dma/ioat_dma.c b/drivers/dma/ioat_dma.c
index 55f4179..c4c7343 100644
--- a/drivers/dma/ioat_dma.c
+++ b/drivers/dma/ioat_dma.c
@@ -432,9 +432,8 @@ static struct dma_async_tx_descriptor *ioat_dma_prep_memcpy(
new->async_tx.ack = 0; /* client is in control of this ack */
new->async_tx.cookie = -EBUSY;
- pci_unmap_len_set(new, src_len, orig_len);
- pci_unmap_len_set(new, dst_len, orig_len);
- spin_unlock_bh(&ioat_chan->desc_lock);
+ pci_unmap_len_set(new, len, orig_len);
+ spin_unlock_bh(&ioat_chan->desc_lock);:w
return new ? &new->async_tx : NULL;
}
@@ -518,11 +517,11 @@ static void ioat_dma_memcpy_cleanup(struct ioat_dma_chan *ioat_chan)
*/
pci_unmap_page(ioat_chan->device->pdev,
pci_unmap_addr(desc, dst),
- pci_unmap_len(desc, dst_len),
+ pci_unmap_len(desc, len),
PCI_DMA_FROMDEVICE);
pci_unmap_page(ioat_chan->device->pdev,
pci_unmap_addr(desc, src),
- pci_unmap_len(desc, src_len),
+ pci_unmap_len(desc, len),
PCI_DMA_TODEVICE);
}
diff --git a/drivers/dma/ioatdma.h b/drivers/dma/ioatdma.h
index bab2a72..d59c90b 100644
--- a/drivers/dma/ioatdma.h
+++ b/drivers/dma/ioatdma.h
@@ -123,10 +123,9 @@ struct ioat_desc_sw {
struct ioat_dma_descriptor *hw;
struct list_head node;
int tx_cnt;
+ DECLARE_PCI_UNMAP_LEN(len)
DECLARE_PCI_UNMAP_ADDR(src)
- DECLARE_PCI_UNMAP_LEN(src_len)
DECLARE_PCI_UNMAP_ADDR(dst)
- DECLARE_PCI_UNMAP_LEN(dst_len)
struct dma_async_tx_descriptor async_tx;
};