Currently dma_request_channel() set DMA_PRIVATE capability but never
clear it. So if a public channel was once grabbed by
dma_request_channel(), the device stay PRIVATE forever. Add
privatecnt member to dma_device to correctly revert it.
Signed-off-by: Atsushi Nemoto <[email protected]>
---
drivers/dma/dmaengine.c | 8 ++++++++
include/linux/dmaengine.h | 9 +++++++++
2 files changed, 17 insertions(+), 0 deletions(-)
diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c
index 280a9d2..14be8fa 100644
--- a/drivers/dma/dmaengine.c
+++ b/drivers/dma/dmaengine.c
@@ -507,6 +507,7 @@ struct dma_chan *__dma_request_channel(dma_cap_mask_t *mask, dma_filter_fn fn, v
* published in the general-purpose allocator
*/
dma_cap_set(DMA_PRIVATE, device->cap_mask);
+ device->privatecnt++;
err = dma_chan_get(chan);
if (err == -ENODEV) {
@@ -518,6 +519,8 @@ struct dma_chan *__dma_request_channel(dma_cap_mask_t *mask, dma_filter_fn fn, v
dma_chan_name(chan), err);
else
break;
+ if (--device->privatecnt == 0)
+ dma_cap_clear(DMA_PRIVATE, device->cap_mask);
chan->private = NULL;
chan = NULL;
}
@@ -537,6 +540,9 @@ void dma_release_channel(struct dma_chan *chan)
WARN_ONCE(chan->client_count != 1,
"chan reference count %d != 1\n", chan->client_count);
dma_chan_put(chan);
+ /* drop PRIVATE cap enabled by __dma_request_channel() */
+ if (--chan->device->privatecnt == 0)
+ dma_cap_clear(DMA_PRIVATE, chan->device->cap_mask);
chan->private = NULL;
mutex_unlock(&dma_list_mutex);
}
@@ -701,6 +707,8 @@ int dma_async_device_register(struct dma_device *device)
}
}
list_add_tail_rcu(&device->global_node, &dma_device_list);
+ if (dma_has_cap(DMA_PRIVATE, device->cap_mask))
+ chan->device->privatecnt++; /* Always private */
dma_channel_rebalance();
mutex_unlock(&dma_list_mutex);
diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h
index f041384..f834fa1 100644
--- a/include/linux/dmaengine.h
+++ b/include/linux/dmaengine.h
@@ -209,6 +209,7 @@ struct dma_async_tx_descriptor {
/**
* struct dma_device - info on the entity supplying DMA services
* @chancnt: how many DMA channels are supported
+ * @privatecnt: how many DMA channels are requested by dma_request_channel
* @channels: the list of struct dma_chan
* @global_node: list_head for global dma_device_list
* @cap_mask: one or more dma_capability flags
@@ -232,6 +233,7 @@ struct dma_async_tx_descriptor {
struct dma_device {
unsigned int chancnt;
+ unsigned int privatecnt;
struct list_head channels;
struct list_head global_node;
dma_cap_mask_t cap_mask;
@@ -342,6 +344,13 @@ __dma_cap_set(enum dma_transaction_type tx_type, dma_cap_mask_t *dstp)
set_bit(tx_type, dstp->bits);
}
+#define dma_cap_clear(tx, mask) __dma_cap_clear((tx), &(mask))
+static inline void
+__dma_cap_clear(enum dma_transaction_type tx_type, dma_cap_mask_t *dstp)
+{
+ clear_bit(tx_type, dstp->bits);
+}
+
#define dma_cap_zero(mask) __dma_cap_zero(&(mask))
static inline void __dma_cap_zero(dma_cap_mask_t *dstp)
{
--
1.5.6.3
Atsushi Nemoto wrote:
> Currently dma_request_channel() set DMA_PRIVATE capability but never
> clear it. So if a public channel was once grabbed by
> dma_request_channel(), the device stay PRIVATE forever. Add
> privatecnt member to dma_device to correctly revert it.
>
> Signed-off-by: Atsushi Nemoto <[email protected]>
> ---
I think it's a good idea and the patch looks fine to me.
Acked-by: Maciej Sosnowski <[email protected]>
On Fri, Mar 6, 2009 at 4:07 AM, Atsushi Nemoto <[email protected]> wrote:
> Currently dma_request_channel() set DMA_PRIVATE capability but never
> clear it. ?So if a public channel was once grabbed by
> dma_request_channel(), the device stay PRIVATE forever. ?Add
> privatecnt member to dma_device to correctly revert it.
>
> Signed-off-by: Atsushi Nemoto <[email protected]>
> ---
This looks good to me. Please rebase it on top of the 'next' branch
of the async_tx tree[1] and I will queue it for -next.
Thanks,
Dan
http://git.kernel.org/?p=linux/kernel/git/djbw/async_tx.git;a=shortlog;h=next
On Mon, 16 Mar 2009 13:39:49 -0700, Dan Williams <[email protected]> wrote:
> > Currently dma_request_channel() set DMA_PRIVATE capability but never
> > clear it. ?So if a public channel was once grabbed by
> > dma_request_channel(), the device stay PRIVATE forever. ?Add
> > privatecnt member to dma_device to correctly revert it.
> >
> > Signed-off-by: Atsushi Nemoto <[email protected]>
> > ---
>
> This looks good to me. Please rebase it on top of the 'next' branch
> of the async_tx tree[1] and I will queue it for -next.
>
> Thanks,
> Dan
>
> http://git.kernel.org/?p=linux/kernel/git/djbw/async_tx.git;a=shortlog;h=next
Sorry, I have missed this mail. I will rebase it, but while the
'next' branch of async_tx tree lacks commit 287d85 ("atmel-mci: fix
initialization of dma slave data"), rebasing will cause another
conflict with mainline. What should I do?
---
Atsushi Nemoto
On Wed, Mar 25, 2009 at 2:42 AM, Atsushi Nemoto <[email protected]> wrote:
> On Mon, 16 Mar 2009 13:39:49 -0700, Dan Williams <[email protected]> wrote:
>> > Currently dma_request_channel() set DMA_PRIVATE capability but never
>> > clear it. ?So if a public channel was once grabbed by
>> > dma_request_channel(), the device stay PRIVATE forever. ?Add
>> > privatecnt member to dma_device to correctly revert it.
>> >
>> > Signed-off-by: Atsushi Nemoto <[email protected]>
>> > ---
>>
>> This looks good to me. ?Please rebase it on top of the 'next' branch
>> of the async_tx tree[1] and I will queue it for -next.
>>
>> Thanks,
>> Dan
>>
>> http://git.kernel.org/?p=linux/kernel/git/djbw/async_tx.git;a=shortlog;h=next
>
> Sorry, I have missed this mail. ?I will rebase it, but while the
> 'next' branch of async_tx tree lacks commit 287d85 ("atmel-mci: fix
> initialization of dma slave data"), rebasing will cause another
> conflict with mainline. ?What should I do?
I have updated the 'next' branch to include this commit.
Thanks,
Dan
On Wed, 25 Mar 2009 09:27:27 -0700, Dan Williams <[email protected]> wrote:
> >> > Currently dma_request_channel() set DMA_PRIVATE capability but never
> >> > clear it. ?So if a public channel was once grabbed by
> >> > dma_request_channel(), the device stay PRIVATE forever. ?Add
> >> > privatecnt member to dma_device to correctly revert it.
> >> >
> >> > Signed-off-by: Atsushi Nemoto <[email protected]>
> >>
> >> This looks good to me. ?Please rebase it on top of the 'next' branch
> >> of the async_tx tree[1] and I will queue it for -next.
> >>
> > Sorry, I have missed this mail. ?I will rebase it, but while the
> > 'next' branch of async_tx tree lacks commit 287d85 ("atmel-mci: fix
> > initialization of dma slave data"), rebasing will cause another
> > conflict with mainline. ?What should I do?
>
> I have updated the 'next' branch to include this commit.
Thanks. And now the patch can be applied using git-am as is.
Should I rebase and send again?
---
Atsushi Nemoto
On Thu, Mar 26, 2009 at 4:44 AM, Atsushi Nemoto <[email protected]> wrote:
> Thanks. ?And now the patch can be applied using git-am as is.
> Should I rebase and send again?
No need, it's now applied.
Thanks,
Dan