ixgbe: Fix ixgbe_tx_map error path

Commit e5a43549f7a58509a91b299a51337d386697b92c (ixgbe: remove skb_dma_map/unmap calls from driver) looks to have introduced a bug in ixgbe_tx_map. If we get an error from a PCI DMA call, we loop backwards through count until it becomes -1 and return that.

The caller of ixgbe_tx_map expects 0 on error, so return that instead.

---

index 1a2ea62..af660ba 100644
@@ -5174,7 +5174,7 @@ dma_error:
 		ixgbe_unmap_and_free_tx_resource(adapter, tx_buffer_info);
 	}
 
-	return count;
+	return 0;
 }
 
 static void ixgbe_tx_queue(struct ixgbe_adapter *adapter,

Actually it looks like the bug for ixgbe is in the while (count >= 0) check. It should be just while (count > 0) insead.

It also appears there is an issue I may have introduced in igb as well. I will go back through the drivers and see about submitting some patches to address these issues.

Thanks,

Alex

-----Original Message-----
From: Anton Blanchard [mailto:[email protected]]
Sent: Tuesday, January 12, 2010 3:02 AM
To: Kirsher, Jeffrey T; Brandeburg, Jesse; Allan, Bruce W; Waskiewicz Jr, Peter P; Ronciak, John; Skidmore, Donald C; Zou, Yi; Duyck, Alexander H
Cc: [email protected]; [email protected]
Subject: [PATCH] ixgbe: Fix ixgbe_tx_map error path

Commit e5a43549f7a58509a91b299a51337d386697b92c (ixgbe: remove skb_dma_map/unmap calls from driver) looks to have introduced a bug in ixgbe_tx_map. If we get an error from a PCI DMA call, we loop backwards through count until it becomes -1 and return that.

The caller of ixgbe_tx_map expects 0 on error, so return that instead.

---

index 1a2ea62..af660ba 100644
@@ -5174,7 +5174,7 @@ dma_error:
 		ixgbe_unmap_and_free_tx_resource(adapter, tx_buffer_info);
 	}
 
-	return count;
+	return 0;
 }
 
 static void ixgbe_tx_queue(struct ixgbe_adapter *adapter,

It looks like we actually have some other patches in place that are addressing the count >= 0 issue. This patch addresses the remaining return value of -1 issues for ixgbe. I think we should probably pull it into Jeff's tree for testing, and I will generate a separate patch for the igb issues.

Thanks,

Alex