dma: of: Remove restriction that #dma-cells can't be 0
Subsystems (2)
PATCH 1/2 dma: of: Remove restriction that #dma-cells can't be 0
There is no sensible reason why #dma-cells shouldn't be allowed to be 0. It is completely up to the DMA controller how many additional parameters, besides the phandle, it needs to identify a channel. E.g. for DMA controller with only one channel or for DMA controllers which don't have a restriction on which channel can be used for which peripheral it completely legitimate to not require any additional parameters.
Also fixes the following warning:
drivers/dma/of-dma.c: In function 'of_dma_controller_register':
drivers/dma/of-dma.c:67:7: warning: 'nbcells' may be used uninitialized in this function
--- drivers/dma/of-dma.c | 11 ++++------- 1 file changed, 4 insertions(+), 7 deletions(-)
drivers/dma/of-dma.c +4 -7
@@ -64,7 +64,6 @@ int of_dma_controller_register(struct device_node *np, void *data) {struct of_dma *ofdma; -int nbcells;const __be32 *prop;if (!np || !of_dma_xlate) {
@@ -77,18 +76,16 @@ int of_dma_controller_register(struct device_node *np,return -ENOMEM ; prop =of_get_property (np,"#dma-cells" ,NULL ); -if (prop) - nbcells =be32_to_cpup (prop); - -if (!prop || !nbcells) { -pr_err ("%s: #dma-cells property is missing or invalid\n" , +if (!prop) { +pr_err ("%s: #dma-cells property is missing\n" , __func__);kfree (ofdma);return -EINVAL ; } + ofdma->of_node = np; - ofdma->of_dma_nbcells = nbcells; + ofdma->of_dma_nbcells =be32_to_cpup (prop); ofdma->of_dma_xlate = of_dma_xlate; ofdma->of_dma_data = data;
Signature · 1 line
1.8.0
PATCH 2/2 dma: of: Remove check on always true condition
Both of_dma_nbcells field of the of_dma_controller and the args_count field of the dma_spec are initialized by parsing the #dma-cells attribute of their device tree node. So if the device tree nodes of a DMA controller and the dma_spec match this means that of_dma_nbcells and args_count will also match. So the second test in the of_dma_find_controller loop is redundant because given the first test yields true the second test will also yield true. So we can safely remove the test whether of_dma_nbcells matches args_count. Since this was the last user of the of_dma_nbcells field we can remove it altogether.
--- drivers/dma/of-dma.c | 14 +------------- include/linux/of_dma.h | 1 - 2 files changed, 1 insertion(+), 14 deletions(-)
drivers/dma/of-dma.c +1 -13
@@ -35,8 +35,7 @@ static struct of_dma *of_dma_find_controller(struct of_phandle_args *dma_spec)struct of_dma *ofdma;list_for_each_entry (ofdma, &of_dma_list, of_dma_controllers) -if ((ofdma->of_node == dma_spec->np) && - (ofdma->of_dma_nbcells == dma_spec->args_count)) +if (ofdma->of_node == dma_spec->np)return ofdma; pr_debug("%s: can't find DMA controller %s\n", __func__,
@@ -64,7 +63,6 @@ int of_dma_controller_register(struct device_node *np, void *data) {struct of_dma *ofdma; -const __be32 *prop;if (!np || !of_dma_xlate) {pr_err ("%s: not enough information provided\n" , __func__);
@@ -75,17 +73,7 @@ int of_dma_controller_register(struct device_node *np,if (!ofdma)return -ENOMEM ; - prop =of_get_property (np,"#dma-cells" ,NULL ); -if (!prop) { -pr_err ("%s: #dma-cells property is missing\n" , - __func__); -kfree (ofdma); -return -EINVAL ; - } - - ofdma->of_node = np; - ofdma->of_dma_nbcells =be32_to_cpup (prop); ofdma->of_dma_xlate = of_dma_xlate; ofdma->of_dma_data = data;
include/linux/of_dma.h +0 -1
@@ -21,7 +21,6 @@ struct device_node;struct of_dma {struct list_head of_dma_controllers;struct device_node *of_node; -int of_dma_nbcells;struct dma_chan *(*of_dma_xlate) (struct of_phandle_args *,struct of_dma *);void *of_dma_data;
Signature · 1 line
1.8.0
Re: [PATCH 2/2] dma: of: Remove check on always true condition
Quoted 11 lines · Lars-Peter Clausen
Both
of_dma_nbcellsfield of theof_dma_controllerand theargs_countfield of thedma_specare initialized by parsing the #dma-cells attribute of their device tree node. So if the device tree nodes of a DMA controller and thedma_specmatch this means thatof_dma_nbcellsandargs_countwill also match. So the second test in theof_dma_find_controllerloop is redundant because given the first test yields true the second test will also yield true. So we can safely remove the test whetherof_dma_nbcellsmatchesargs_count. Since this was the last user of theof_dma_nbcellsfield we can remove it altogether.Signed-off-by Lars-Peter Clausen <[email protected]>---
Re: [PATCH 2/2] dma: of: Remove check on always true condition
In reply to Lars-Peter ClausenQuoted 8 lines · Lars-Peter Clausen
Both
of_dma_nbcellsfield of theof_dma_controllerand theargs_countfield of thedma_specare initialized by parsing the #dma-cells attribute of their device tree node. So if the device tree nodes of a DMA controller and thedma_specmatch this means thatof_dma_nbcellsandargs_countwill also match. So the second test in theof_dma_find_controllerloop is redundant because given the first test yields true the second test will also yield true. So we can safely remove the test whetherof_dma_nbcellsmatchesargs_count. Since this was the last user of theof_dma_nbcellsfield we can remove it altogether.
This assumes that someone has correctly added the dma information to the dma slave binding. I could see systems where different dma controllers have different of_dma_nbcells and so someone could put the enter wrong number of cells for a dma slave binding. Its really to catch user error.
Quoted 11 lines
Signed-off-by Lars-Peter Clausen <[email protected]>--- drivers/dma/of-dma.c | 14 +------------- include/linux/of_dma.h | 1 - 2 files changed, 1 insertion(+), 14 deletions(-)drivers/dma/of-dma.c +0
-0
drivers/dma/of-dma.c static struct of_dma *of_dma_find_controller(struct of_phandle_args *dma_spec) · line 39
Quoted 7 lines
drivers/dma/of-dma.c +1
-2@@ -35,8 +35,7 @@ static struct of_dma *of_dma_find_controller(struct of_phandle_args *dma_spec)struct of_dma *ofdma;list_for_each_entry (ofdma, &of_dma_list, of_dma_controllers) -if ((ofdma->of_node == dma_spec->np) && - (ofdma->of_dma_nbcells == dma_spec->args_count)) +if (ofdma->of_node == dma_spec->np)return ofdma;
Other device-tree functions perform similar tests to this such as ...
static int of_gpiochip_find_and_xlate(struct gpio_chip *gc, void *data)
{
struct gg_data *gg_data = data;
int ret;
if ((gc->of_node != gg_data->gpiospec.np) ||
(gc->of_gpio_n_cells != gg_data->gpiospec.args_count) ||
(!gc->of_xlate))
return false;...
Cheers
Jon
Re: [PATCH 2/2] dma: of: Remove check on always true condition
Quoted 11 lines · Jon Hunter
Quoted 8 lines · Lars-Peter Clausen
Both
of_dma_nbcellsfield of theof_dma_controllerand theargs_countfield of thedma_specare initialized by parsing the #dma-cells attribute of their device tree node. So if the device tree nodes of a DMA controller and thedma_specmatch this means thatof_dma_nbcellsandargs_countwill also match. So the second test in theof_dma_find_controllerloop is redundant because given the first test yields true the second test will also yield true. So we can safely remove the test whetherof_dma_nbcellsmatchesargs_count. Since this was the last user of theof_dma_nbcellsfield we can remove it altogether.
Quoted 4 lines
This assumes that someone has correctly added the dma information to the dma slave binding. I could see systems where different dma controllers have different
of_dma_nbcellsand so someone could put the enter wrong number of cells for a dma slave binding. Its really to catch user error.
No, this assumes nothing. The condition will _always_ be true.
dma_spec->args_count is initialized by parsing the #dma-cells attribute of dma_sepc->np. of_dma->of_dma_nbcells is initialized by parsing the #dma-cells attribute of of_dma->of_node. If ofdma->of_node equals dma_spec->np then dma_spec->args_count will also equal of_dma->of_dma_nbcells.
Quoted 14 lines
Quoted 14 lines
Signed-off-by Lars-Peter Clausen <[email protected]>--- drivers/dma/of-dma.c | 14 +------------- include/linux/of_dma.h | 1 - 2 files changed, 1 insertion(+), 14 deletions(-)drivers/dma/of-dma.c +0
-0@@ -35,8 +35,7 @@ static struct of_dma *of_dma_find_controller(struct of_phandle_args *dma_spec)struct of_dma *ofdma;
Quoted 8 lines
drivers/dma/of-dma.c static struct of_dma *of_dma_find_controller(struct of_phandle_args *dma_spec) · line 39
Quoted 5 lines
drivers/dma/of-dma.c +1
-2list_for_each_entry (ofdma, &of_dma_list, of_dma_controllers) -if ((ofdma->of_node == dma_spec->np) && - (ofdma->of_dma_nbcells == dma_spec->args_count)) +if (ofdma->of_node == dma_spec->np)return ofdma;Other device-tree functions perform similar tests to this such as ...
So it's redundant there as well ;)
Quoted 14 lines
static int
of_gpiochip_find_and_xlate(struct gpio_chip *gc, void *data){ struct gg_data *gg_data = data; int ret; if ((gc->of_node != gg_data->gpiospec.np) || (gc->of_gpio_n_cells != gg_data->gpiospec.args_count) || (!gc->of_xlate)) return false;...
Cheers
Jon
Re: [PATCH 2/2] dma: of: Remove check on always true condition
Quoted 17 lines · Lars-Peter Clausen
Quoted 11 lines · Jon Hunter
Quoted 8 lines · Lars-Peter Clausen
Both
of_dma_nbcellsfield of theof_dma_controllerand theargs_countfield of thedma_specare initialized by parsing the #dma-cells attribute of their device tree node. So if the device tree nodes of a DMA controller and thedma_specmatch this means thatof_dma_nbcellsandargs_countwill also match. So the second test in theof_dma_find_controllerloop is redundant because given the first test yields true the second test will also yield true. So we can safely remove the test whetherof_dma_nbcellsmatchesargs_count. Since this was the last user of theof_dma_nbcellsfield we can remove it altogether.Quoted 4 lines
This assumes that someone has correctly added the dma information to the dma slave binding. I could see systems where different dma controllers have different
of_dma_nbcellsand so someone could put the enter wrong number of cells for a dma slave binding. Its really to catch user error.
Quoted 6 lines
No, this assumes nothing. The condition will _always_ be true.
dma_spec->args_countis initialized by parsing the #dma-cells attribute ofdma_sepc->np.of_dma->of_dma_nbcellsis initialized by parsing the #dma-cells attribute ofof_dma->of_node. Ifofdma->of_nodeequalsdma_spec->npthendma_spec->args_countwill also equalof_dma->of_dma_nbcells.
Thanks for the clarification. I should have looked more closely at of_parse_phandle_with_args().
Yes I agree it will always be true if count is less than or equal to MAX_PHANDLE_ARGS (which defaults to 8). It is very unlikely that someone would use more than 8 and I guess it does warn on this condition.
if (out_args) {
int i;
if (WARN_ON(count > MAX_PHANDLE_ARGS))
count = MAX_PHANDLE_ARGS;
out_args->np = node;
out_args->args_count = count;
for (i = 0; i < count; i++)
out_args->args[i] = be32_to_cpup(list++);
}Cheers
Jon
Quoted 8 lines · Lars-Peter Clausen
There is no sensible reason why #dma-cells shouldn't be allowed to be 0. It is completely up to the DMA controller how many additional parameters, besides the phandle, it needs to identify a channel. E.g. for DMA controller with only one channel or for DMA controllers which don't have a restriction on which channel can be used for which peripheral it completely legitimate to not require any additional parameters.
Quoted 5 lines
Also fixes the following warning:
drivers/dma/of-dma.c: In function 'of_dma_controller_register':
drivers/dma/of-dma.c:67:7: warning: 'nbcells' may be used uninitialized in this functionSigned-off-by Lars-Peter Clausen <[email protected]>
Do you have an example for this? If a dma engine has only one request line, why would you even use the dmaengine subsystem for it, rather than including the code to program it in the slave driver?
Arnd
Quoted 13 lines · Arnd Bergmann
Quoted 12 lines · Lars-Peter Clausen
There is no sensible reason why #dma-cells shouldn't be allowed to be 0. It is completely up to the DMA controller how many additional parameters, besides the phandle, it needs to identify a channel. E.g. for DMA controller with only one channel or for DMA controllers which don't have a restriction on which channel can be used for which peripheral it completely legitimate to not require any additional parameters.
Also fixes the following warning:
drivers/dma/of-dma.c: In function 'of_dma_controller_register':
drivers/dma/of-dma.c:67:7: warning: 'nbcells' may be used uninitialized in this function
Quoted 5 lines
Quoted 1 line
Signed-off-by Lars-Peter Clausen <[email protected]>Do you have an example for this? If a dma engine has only one request line, why would you even use the dmaengine subsystem for it, rather than including the code to program it in the slave driver?
Why wouldn't I use the dmaengine subsystem for a DMA controller? In my particular case different instances of the same DMA core will be used with different DMA slaves. And the DMA slaves can also have different DMA master cores, depending on the system.
- Lars
Arnd Bergmann
Acked-byQuoted 17 lines · Lars-Peter Clausen
Quoted 15 lines · Arnd Bergmann
Quoted 13 lines · Lars-Peter Clausen
There is no sensible reason why #dma-cells shouldn't be allowed to be 0. It is completely up to the DMA controller how many additional parameters, besides the phandle, it needs to identify a channel. E.g. for DMA controller with only one channel or for DMA controllers which don't have a restriction on which channel can be used for which peripheral it completely legitimate to not require any additional parameters.
Also fixes the following warning:
drivers/dma/of-dma.c: In function 'of_dma_controller_register':
drivers/dma/of-dma.c:67:7: warning: 'nbcells' may be used uninitialized in this functionSigned-off-by Lars-Peter Clausen <[email protected]>
Quoted 8 lines
Quoted 3 lines
Do you have an example for this? If a dma engine has only one request line, why would you even use the dmaengine subsystem for it, rather than including the code to program it in the slave driver?
Why wouldn't I use the dmaengine subsystem for a DMA controller? In my particular case different instances of the same DMA core will be used with different DMA slaves. And the DMA slaves can also have different DMA master cores, depending on the system.
Right, that makes sense.
Vinod Koul
AppliedQuoted 7 lines · Lars-Peter Clausen
There is no sensible reason why #dma-cells shouldn't be allowed to be 0. It is completely up to the DMA controller how many additional parameters, besides the phandle, it needs to identify a channel. E.g. for DMA controller with only one channel or for DMA controllers which don't have a restriction on which channel can be used for which peripheral it completely legitimate to not require any additional parameters.
Quoted 5 lines
Also fixes the following warning:
drivers/dma/of-dma.c: In function 'of_dma_controller_register':
drivers/dma/of-dma.c:67:7: warning: 'nbcells' may be used uninitialized in this functionSigned-off-by Lars-Peter Clausen <[email protected]>
Applied, both thanks
Sorry looks like I missed applying them earlier...
--
~Vinod
Quoted 42 lines
--- drivers/dma/of-dma.c | 11 ++++------- 1 file changed, 4 insertions(+), 7 deletions(-)drivers/dma/of-dma.c +4
-7@@ -64,7 +64,6 @@ int of_dma_controller_register(struct device_node *np, void *data) {struct of_dma *ofdma; -int nbcells;const __be32 *prop;if (!np || !of_dma_xlate) {@@ -77,18 +76,16 @@ int of_dma_controller_register(struct device_node *np,return -ENOMEM ; prop =of_get_property (np,"#dma-cells" ,NULL ); -if (prop) - nbcells =be32_to_cpup (prop); - -if (!prop || !nbcells) { -pr_err ("%s: #dma-cells property is missing or invalid\n" , +if (!prop) { +pr_err ("%s: #dma-cells property is missing\n" , __func__);kfree (ofdma);return -EINVAL ; } + ofdma->of_node = np; - ofdma->of_dma_nbcells = nbcells; + ofdma->of_dma_nbcells =be32_to_cpup (prop); ofdma->of_dma_xlate = of_dma_xlate; ofdma->of_dma_data = data;Signature · 1 line
1.8.0