On Fri, 2008-11-14 at 23:17 -0700, Andrew Morton wrote:

On Fri, 14 Nov 2008 14:34:58 -0700 Dan Williams [off-list ref] wrote:

quoted

 static int __init dmatest_init(void)
 {
-	dma_cap_set(DMA_MEMCPY, dmatest_client.cap_mask);
-	dma_async_client_register(&dmatest_client);
-	dma_async_client_chan_request(&dmatest_client);
+	dma_cap_mask_t mask;
+	struct dma_chan *chan;
+
+	dma_cap_zero(mask);
+	dma_cap_set(DMA_MEMCPY, mask);
+	for (;;) {
+		chan = dma_request_channel(mask, filter, NULL);
+		if (chan) {
+			if (dmatest_add_channel(chan) == 0)
+				continue;

hrm.  I trust this loop can't lock up the box.

Only if the system has infinite channels, but we should break if
dma_add_channel fails.

diff --git a/drivers/dma/dmatest.c b/drivers/dma/dmatest.c
index e0f9139..c77d47c 100644
--- a/drivers/dma/dmatest.c
+++ b/drivers/dma/dmatest.c

@@ -375,23 +375,27 @@ static int __init dmatest_init(void)
 {
 	dma_cap_mask_t mask;
 	struct dma_chan *chan;
+	int err = 0;
 
 	dma_cap_zero(mask);
 	dma_cap_set(DMA_MEMCPY, mask);
 	for (;;) {
 		chan = dma_request_channel(mask, filter, NULL);
 		if (chan) {
-			if (dmatest_add_channel(chan) == 0)
+			err = dmatest_add_channel(chan);
+			if (err == 0)
 				continue;
-			else
+			else {
 				dma_release_channel(chan);
+				break; /* add_channel failed, punt */
+			}
 		} else
 			break; /* no more channels available */
 		if (max_channels && nr_channels >= max_channels)
 			break; /* we have all we need */
 	}
 
-	return 0;
+	return err;
 }
 /* when compiled-in wait for drivers to load first */
 late_initcall(dmatest_init);

---

Ok, so I now have:
# git shortlog `stg id akpm-review-spell-fixes.patch`^..
Dan Williams (6):
      dmaengine review: spelling/grammar fixes
      dmaengine review: dma_wait_for_async_tx clarification
      dmaengine review: fix up kernel doc for dma_list_mutex usage
      dmaengine review: propagate alloc_chan_resources error code
      dmaengine review: return meaningful error code from initcall
      dmaengine review: dmatest stop on dmatest_add_channel failure

May I add your Reviewed-by for these and the original 13.  I could fold
these into the originals but perhaps it is better to document fixups
when they are the result of review [1].

Thanks,
Dan

[1] http://lwn.net/Articles/306690/