Skip to contentSkip to conversation

I/OAT fixes

PATCH New Maciej Sosnowski 1 participant 5 messages Nov 7, 2008

Subsystems (4)
Networking Network drivers dmaengine Intel i/oat DMA driver
lore
Share & tools
lore

Maciej Sosnowski

author
#0 lore raw

cover I/OAT fixes

This patch series contains couple of fixes for I/OAT DMA and DCA.

Maciej Sosnowski (3):
I/OAT: fix channel resources free for not allocated channels I/OAT: fix dma_pin_iovec_pages() error handling I/OAT: fix async_tx.callback checking

Dan Williams (1):
dca: fixup initialization dependency

 drivers/dca/dca-core.c |    2 +-
 drivers/dma/ioat_dma.c |   11 +++++++++--
 drivers/dma/iovlock.c  |   17 ++++++-----------
 3 files changed, 16 insertions(+), 14 deletions(-)

--
Maciej

author
#1 lore raw

PATCH 1/4 I/OAT: fix channel resources free for not allocated channels

From: Maciej Sosnowski <[email protected]>

If the ioatdma driver is loaded but not used it does not allocate descriptors. Before it frees channel resources it should first be sure
that they have been previously allocated.

Cc <[email protected]>
Signed-off-by Maciej Sosnowski <[email protected]>
Tested-by Tom Picard <[email protected]>
Signed-off-by Dan Williams <[email protected]>

---

 drivers/dma/ioat_dma.c |    7 +++++++
 1 files changed, 7 insertions(+), 0 deletions(-)
drivers/dma/ioat_dma.c +7 -0
index b0438c4..687eef7 100644
@@ -807,6 +807,12 @@ static void ioat_dma_free_chan_resources
 	struct ioat_desc_sw *desc, *_desc;
 	int in_use_descs = 0;
 
+	/* Before freeing channel resources first check 
+	 * if they have been previously allocated for this channel.
+	 */
+	if (ioat_chan->desccount == 0)
+		return;
+
 	tasklet_disable(&ioat_chan->cleanup_task);
 	ioat_dma_memcpy_cleanup(ioat_chan);
 

@@ -869,6 +875,7 @@ static void ioat_dma_free_chan_resources
 	ioat_chan->last_completion = ioat_chan->completion_addr = 0;
 	ioat_chan->pending = 0;
 	ioat_chan->dmacount = 0;
+	ioat_chan->desccount = 0;
 	ioat_chan->watchdog_completion = 0;
 	ioat_chan->last_compl_desc_addr_hw = 0;
 	ioat_chan->watchdog_tcp_cookie =
author
#2 lore raw

PATCH 2/4 I/OAT: fix dma_pin_iovec_pages() error handling

In reply to Maciej Sosnowski

From: Maciej Sosnowski <[email protected]>

Error handling needs to be modified in dma_pin_iovec_pages(). It should return NULL instead of ERR_PTR (pinned_list is checked for NULL in tcp_recvmsg() to determine if iovec pages have been successfully pinned down). In case of error for the first iovec, local_list->nr_iovecs needs to be initialized.

Cc <[email protected]>
Signed-off-by Maciej Sosnowski <[email protected]>

---

 drivers/dma/iovlock.c |   17 ++++++-----------
 1 files changed, 6 insertions(+), 11 deletions(-)
drivers/dma/iovlock.c +6 -11
index e763d72..9f6fe46 100644
@@ -55,7 +55,6 @@ struct dma_pinned_list *dma_pin_iovec_pa
 	int nr_iovecs = 0;
 	int iovec_len_used = 0;
 	int iovec_pages_used = 0;
-	long err;
 
 	/* don't pin down non-user-based iovecs */
 	if (segment_eq(get_fs(), KERNEL_DS))

@@ -72,23 +71,21 @@ struct dma_pinned_list *dma_pin_iovec_pa
 	local_list = kmalloc(sizeof(*local_list)
 		+ (nr_iovecs * sizeof (struct dma_page_list))
 		+ (iovec_pages_used * sizeof (struct page*)), GFP_KERNEL);
-	if (!local_list) {
-		err = -ENOMEM;
+	if (!local_list)
 		goto out;
-	}
 
 	/* list of pages starts right after the page list array */
 	pages = (struct page **) &local_list->page_list[nr_iovecs];
 
+	local_list->nr_iovecs = 0;
+
 	for (i = 0; i < nr_iovecs; i++) {
 		struct dma_page_list *page_list = &local_list->page_list[i];
 
 		len -= iov[i].iov_len;
 
-		if (!access_ok(VERIFY_WRITE, iov[i].iov_base, iov[i].iov_len)) {
-			err = -EFAULT;
+		if (!access_ok(VERIFY_WRITE, iov[i].iov_base, iov[i].iov_len))
 			goto unpin;
-		}
 
 		page_list->nr_pages = num_pages_spanned(&iov[i]);
 		page_list->base_address = iov[i].iov_base;

@@ -109,10 +106,8 @@ struct dma_pinned_list *dma_pin_iovec_pa
 			NULL);
 		up_read(&current->mm->mmap_sem);
 
-		if (ret != page_list->nr_pages) {
-			err = -ENOMEM;
+		if (ret != page_list->nr_pages)
 			goto unpin;
-		}
 
 		local_list->nr_iovecs = i + 1;
 	}

@@ -122,7 +117,7 @@ struct dma_pinned_list *dma_pin_iovec_pa
 unpin:
 	dma_unpin_iovec_pages(local_list);
 out:
-	return ERR_PTR(err);
+	return NULL;
 }
 
 void dma_unpin_iovec_pages(struct dma_pinned_list *pinned_list)
author
#3 lore raw

PATCH 3/4 I/OAT: fix async_tx.callback checking

In reply to Maciej Sosnowski

From: Maciej Sosnowski <[email protected]>

async_tx.callback should be checked for the first
not the last descriptor in the chain.

Cc <[email protected]>
Signed-off-by Maciej Sosnowski <[email protected]>

---

 drivers/dma/ioat_dma.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
drivers/dma/ioat_dma.c +2 -2
index 687eef7..c5af17a 100644
@@ -525,7 +525,7 @@ static dma_cookie_t ioat1_tx_submit(stru
 	}
 
 	hw->ctl = IOAT_DMA_DESCRIPTOR_CTL_CP_STS;
-	if (new->async_tx.callback) {
+	if (first->async_tx.callback) {
 		hw->ctl |= IOAT_DMA_DESCRIPTOR_CTL_INT_GN;
 		if (first != new) {
 			/* move callback into to last desc */

@@ -617,7 +617,7 @@ static dma_cookie_t ioat2_tx_submit(stru
 	}
 
 	hw->ctl |= IOAT_DMA_DESCRIPTOR_CTL_CP_STS;
-	if (new->async_tx.callback) {
+	if (first->async_tx.callback) {
 		hw->ctl |= IOAT_DMA_DESCRIPTOR_CTL_INT_GN;
 		if (first != new) {
 			/* move callback into to last desc */
author
#4 lore raw

PATCH 4/4 dca: fixup initialization dependency

In reply to Maciej Sosnowski

From: Dan Williams <[email protected]>

Mark dca_init as a subsys_initcall since it needs to be ready to go before dependent drivers start registering themselves.

Cc <[email protected]>
Reported-and-tested-by Mark Rustad <[email protected]>
Acked-by Maciej Sosnowski <[email protected]>
Signed-off-by Dan Williams <[email protected]>

---

 drivers/dca/dca-core.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
drivers/dca/dca-core.c +1 -1
index ec249d2..d883e1b 100644
@@ -270,6 +270,6 @@ static void __exit dca_exit(void)
 	dca_sysfs_exit();
 }
 
-module_init(dca_init);
+subsys_initcall(dca_init);
 module_exit(dca_exit);