| Message ID | 1375423458-6868-1-git-send-email-rizhao@nvidia.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 08/02/2013 12:04 AM, Richard Zhao wrote: > DMA client device driver usually needs to know at probe time whether > dma controller has been registered to deffer probe. So add a help > function of_dma_check_controller. > > DMA request channel functions can also used to check it, but they > are usually called at open() time. > diff --git a/drivers/dma/of-dma.c b/drivers/dma/of-dma.c > index e1c4d3b..b6828c1 100644 > +int of_dma_check_controller(struct device *dev, const char *name) > + for (i = 0; i < count; i++) { > + if (of_dma_match_channel(np, name, i, &dma_spec)) > + continue; > + > + mutex_lock(&of_dma_lock); > + ofdma = of_dma_find_controller(&dma_spec); > + mutex_unlock(&of_dma_lock); > + of_node_put(dma_spec.np); Do we need to add the following here: if (ofdma) break To ensure that as soon as a successful match is found, the loop exits? Otherwise, if there are multiple providers for that name, and the first N are registered but the last isn't, this function will still return failure. > + if (ofdma) > + return 0; > + else > + return -ENODEV; That probably should be -EPROBE_DEFER? Although, what about differentiating between "entry not found by of_dma_match_channel" and "controller not yet probed"?
On Sat, Aug 03, 2013 at 03:59:59AM +0800, Stephen Warren wrote: > On 08/02/2013 12:04 AM, Richard Zhao wrote: > > DMA client device driver usually needs to know at probe time whether > > dma controller has been registered to deffer probe. So add a help > > function of_dma_check_controller. > > > > DMA request channel functions can also used to check it, but they > > are usually called at open() time. > > > diff --git a/drivers/dma/of-dma.c b/drivers/dma/of-dma.c > > index e1c4d3b..b6828c1 100644 > > > +int of_dma_check_controller(struct device *dev, const char *name) > > > > + for (i = 0; i < count; i++) { > > + if (of_dma_match_channel(np, name, i, &dma_spec)) > > + continue; > > + > > + mutex_lock(&of_dma_lock); > > + ofdma = of_dma_find_controller(&dma_spec); > > + mutex_unlock(&of_dma_lock); > > + of_node_put(dma_spec.np); > > Do we need to add the following here: > > if (ofdma) > break > > To ensure that as soon as a successful match is found, the loop exits? > Otherwise, if there are multiple providers for that name, and the first > N are registered but the last isn't, this function will still return > failure. Right. Thanks. > > > + if (ofdma) > > + return 0; > > + else > > + return -ENODEV; > > That probably should be -EPROBE_DEFER? > > Although, what about differentiating between "entry not found by > of_dma_match_channel" and "controller not yet probed"? I thought it might be called in non-probe functions. If no people against it, I'll change it to EPROBE_DEFER. Thanks Richard
diff --git a/drivers/dma/of-dma.c b/drivers/dma/of-dma.c index e1c4d3b..b6828c1 100644 --- a/drivers/dma/of-dma.c +++ b/drivers/dma/of-dma.c @@ -188,6 +188,44 @@ struct dma_chan *of_dma_request_slave_channel(struct device_node *np, } /** + * of_dma_check_controller - Check whether dma controller registered + * @dev: pointer to client device structure + * @name: slave channel name + */ +int of_dma_check_controller(struct device *dev, const char *name) +{ + struct device_node *np = dev->of_node; + struct of_phandle_args dma_spec; + struct of_dma *ofdma = NULL; + int count, i; + + if (!np || !name) + return -EINVAL; + + count = of_property_count_strings(np, "dma-names"); + if (count < 0) { + dev_err(dev, "dma-names property missing or empty\n"); + return -EINVAL; + } + + for (i = 0; i < count; i++) { + if (of_dma_match_channel(np, name, i, &dma_spec)) + continue; + + mutex_lock(&of_dma_lock); + ofdma = of_dma_find_controller(&dma_spec); + mutex_unlock(&of_dma_lock); + of_node_put(dma_spec.np); + } + + if (ofdma) + return 0; + else + return -ENODEV; +} +EXPORT_SYMBOL_GPL(of_dma_check_controller); + +/** * of_dma_simple_xlate - Simple DMA engine translation function * @dma_spec: pointer to DMA specifier as found in the device tree * @of_dma: pointer to DMA controller data diff --git a/include/linux/of_dma.h b/include/linux/of_dma.h index ae36298..bc7195f 100644 --- a/include/linux/of_dma.h +++ b/include/linux/of_dma.h @@ -39,6 +39,7 @@ extern int of_dma_controller_register(struct device_node *np, extern void of_dma_controller_free(struct device_node *np); extern struct dma_chan *of_dma_request_slave_channel(struct device_node *np, const char *name); +extern int of_dma_check_controller(struct device *dev, const char *name); extern struct dma_chan *of_dma_simple_xlate(struct of_phandle_args *dma_spec, struct of_dma *ofdma); #else @@ -60,6 +61,11 @@ static inline struct dma_chan *of_dma_request_slave_channel(struct device_node * return NULL; } +static inline int of_dma_check_controller(struct device *dev, const char *name) +{ + return 0; +} + static inline struct dma_chan *of_dma_simple_xlate(struct of_phandle_args *dma_spec, struct of_dma *ofdma) {
DMA client device driver usually needs to know at probe time whether dma controller has been registered to deffer probe. So add a help function of_dma_check_controller. DMA request channel functions can also used to check it, but they are usually called at open() time. Signed-off-by: Richard Zhao <rizhao@nvidia.com> --- drivers/dma/of-dma.c | 38 ++++++++++++++++++++++++++++++++++++++ include/linux/of_dma.h | 6 ++++++ 2 files changed, 44 insertions(+)