| Message ID | 1406427279.29010.139.camel@deadeye.wl.decadent.org.uk |
|---|---|
| State | Accepted, archived |
| Delegated to: | David Miller |
| Headers | show |
On 27/07/14 03:14, Ben Hutchings wrote: > __iowrite64_copy() isn't quite the same as efx_memcpy_64(), but > it looks close enough: > > - The length is in units of qwords not bytes > - It never byte-swaps, but that doesn't make a difference now as PIO > is only enabled for x86_64 > - It doesn't include any memory barriers, but that's OK as there is a > barrier just before pushing the doorbell > - mlx4_en uses it for the same purpose > > Compile-tested only. > > Signed-off-by: Ben Hutchings <ben@decadent.org.uk> Acked-by: Edward Cree <ecree@solarflare.com> On my test system, __iowrite64_copy compiles to use "mov %rax,(%rdi)" which we believe is safe (all writes will be naturally-aligned 64-bit accesses, as required by the hardware). -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
From: Ben Hutchings <ben@decadent.org.uk> Date: Sun, 27 Jul 2014 03:14:39 +0100 > __iowrite64_copy() isn't quite the same as efx_memcpy_64(), but > it looks close enough: > > - The length is in units of qwords not bytes > - It never byte-swaps, but that doesn't make a difference now as PIO > is only enabled for x86_64 > - It doesn't include any memory barriers, but that's OK as there is a > barrier just before pushing the doorbell > - mlx4_en uses it for the same purpose > > Compile-tested only. > > Signed-off-by: Ben Hutchings <ben@decadent.org.uk> Applied, thanks. -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/net/ethernet/sfc/tx.c b/drivers/net/ethernet/sfc/tx.c index ede8dcc..e03350e 100644 --- a/drivers/net/ethernet/sfc/tx.c +++ b/drivers/net/ethernet/sfc/tx.c @@ -189,18 +189,6 @@ struct efx_short_copy_buffer { u8 buf[L1_CACHE_BYTES]; }; -/* Copy in explicit 64-bit writes. */ -static void efx_memcpy_64(void __iomem *dest, void *src, size_t len) -{ - u64 *src64 = src; - u64 __iomem *dest64 = dest; - size_t l64 = len / 8; - size_t i; - - for (i = 0; i < l64; i++) - writeq(src64[i], &dest64[i]); -} - /* Copy to PIO, respecting that writes to PIO buffers must be dword aligned. * Advances piobuf pointer. Leaves additional data in the copy buffer. */ @@ -210,7 +198,7 @@ static void efx_memcpy_toio_aligned(struct efx_nic *efx, u8 __iomem **piobuf, { int block_len = len & ~(sizeof(copy_buf->buf) - 1); - efx_memcpy_64(*piobuf, data, block_len); + __iowrite64_copy(*piobuf, data, block_len >> 3); *piobuf += block_len; len -= block_len; @@ -242,7 +230,8 @@ static void efx_memcpy_toio_aligned_cb(struct efx_nic *efx, u8 __iomem **piobuf, if (copy_buf->used < sizeof(copy_buf->buf)) return; - efx_memcpy_64(*piobuf, copy_buf->buf, sizeof(copy_buf->buf)); + __iowrite64_copy(*piobuf, copy_buf->buf, + sizeof(copy_buf->buf) >> 3); *piobuf += sizeof(copy_buf->buf); data += copy_to_buf; len -= copy_to_buf; @@ -257,7 +246,8 @@ static void efx_flush_copy_buffer(struct efx_nic *efx, u8 __iomem *piobuf, { /* if there's anything in it, write the whole buffer, including junk */ if (copy_buf->used) - efx_memcpy_64(piobuf, copy_buf->buf, sizeof(copy_buf->buf)); + __iowrite64_copy(piobuf, copy_buf->buf, + sizeof(copy_buf->buf) >> 3); } /* Traverse skb structure and copy fragments in to PIO buffer. @@ -316,8 +306,8 @@ efx_enqueue_skb_pio(struct efx_tx_queue *tx_queue, struct sk_buff *skb) */ BUILD_BUG_ON(L1_CACHE_BYTES > SKB_DATA_ALIGN(sizeof(struct skb_shared_info))); - efx_memcpy_64(tx_queue->piobuf, skb->data, - ALIGN(skb->len, L1_CACHE_BYTES)); + __iowrite64_copy(tx_queue->piobuf, skb->data, + ALIGN(skb->len, L1_CACHE_BYTES) >> 3); } EFX_POPULATE_QWORD_5(buffer->option,
__iowrite64_copy() isn't quite the same as efx_memcpy_64(), but it looks close enough: - The length is in units of qwords not bytes - It never byte-swaps, but that doesn't make a difference now as PIO is only enabled for x86_64 - It doesn't include any memory barriers, but that's OK as there is a barrier just before pushing the doorbell - mlx4_en uses it for the same purpose Compile-tested only. Signed-off-by: Ben Hutchings <ben@decadent.org.uk> --- drivers/net/ethernet/sfc/tx.c | 24 +++++++----------------- 1 file changed, 7 insertions(+), 17 deletions(-)