On 14/08/2024 18:38, Jonathan Cameron wrote:
> On Mon, 12 Aug 2024 15:33:55 +0200
> Krzysztof Kozlowski <krzysztof.kozlowski@xxxxxxxxxx> wrote:
>
>> Obtain the device node reference with scoped/cleanup.h to reduce error
>> handling and make the code a bit simpler.
>>
>> Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@xxxxxxxxxx>
>
> Hi,
>
> Comments inline.
>> ---
>> drivers/memory/atmel-ebi.c | 29 ++++++++++-------------------
>> 1 file changed, 10 insertions(+), 19 deletions(-)
>>
>> diff --git a/drivers/memory/atmel-ebi.c b/drivers/memory/atmel-ebi.c
>> index e8bb5f37f5cb..fcbfc2655d8d 100644
>> --- a/drivers/memory/atmel-ebi.c
>> +++ b/drivers/memory/atmel-ebi.c
>> @@ -6,6 +6,7 @@
>> * Copyright (C) 2013 Jean-Jacques Hiblot <jjhiblot@xxxxxxxxxxxxxxx>
>> */
>>
>> +#include <linux/cleanup.h>
>> #include <linux/clk.h>
>> #include <linux/io.h>
>> #include <linux/mfd/syscon.h>
>> @@ -517,7 +518,7 @@ static int atmel_ebi_dev_disable(struct atmel_ebi *ebi, struct device_node *np)
>> static int atmel_ebi_probe(struct platform_device *pdev)
>> {
>> struct device *dev = &pdev->dev;
>> - struct device_node *child, *np = dev->of_node, *smc_np;
>> + struct device_node *child, *np = dev->of_node;
>> struct atmel_ebi *ebi;
>> int ret, reg_cells;
>> struct clk *clk;
>> @@ -541,30 +542,24 @@ static int atmel_ebi_probe(struct platform_device *pdev)
>>
>> ebi->clk = clk;
>>
>> - smc_np = of_parse_phandle(dev->of_node, "atmel,smc", 0);
>> + struct device_node *smc_np __free(device_node) = of_parse_phandle(dev->of_node,
>> + "atmel,smc", 0);
> Trivial:
> I'd line break this as
>> + struct device_node *smc_np __free(device_node) =
> of_parse_phandle(dev->of_node, "atmel,smc", 0);
Yeah, I have troubles with this constructor+destructor syntaxes. They
are way past 80 and 100 column, so maybe indeed should be wrapped at '='.
>
>>
>> ebi->smc.regmap = syscon_node_to_regmap(smc_np);
>> - if (IS_ERR(ebi->smc.regmap)) {
>> - ret = PTR_ERR(ebi->smc.regmap);
>> - goto put_node;
>> - }
>> + if (IS_ERR(ebi->smc.regmap))
>> + return PTR_ERR(ebi->smc.regmap);
>>
>> ebi->smc.layout = atmel_hsmc_get_reg_layout(smc_np);
>> - if (IS_ERR(ebi->smc.layout)) {
>> - ret = PTR_ERR(ebi->smc.layout);
>> - goto put_node;
>> - }
>> + if (IS_ERR(ebi->smc.layout))
>> + return PTR_ERR(ebi->smc.layout);
>>
>> ebi->smc.clk = of_clk_get(smc_np, 0);
>> if (IS_ERR(ebi->smc.clk)) {
>> - if (PTR_ERR(ebi->smc.clk) != -ENOENT) {
>> - ret = PTR_ERR(ebi->smc.clk);
>> - goto put_node;
>> - }
>> + if (PTR_ERR(ebi->smc.clk) != -ENOENT)
>> + return PTR_ERR(ebi->smc.clk);
>>
>> ebi->smc.clk = NULL;
>> }
>> - of_node_put(smc_np);
>
> The large change in scope is a bit inelegant as it now hangs on to
> the smc_np much longer than before.
>
> Maybe it's worth pulling out the modified code as a
> atem_eb_probe_smc(struct device_node *smc_np, struct atmel_ebi_smc *smc )
>
> or something like with a struct_group to define the atmel_ebi_smc
>
> That would keep the tight scope for the data and generally simplify it
> a bit.
Are you speaking about any particular code optimization/performance
concerns or readability? Because scope in the latter, I believe, is not
the problem here. The entire point of __free() is that you do not care
about scope of variable or destructor. You know it will be freed, sooner
or later. If it happens later - no problem, anyway we don't have to
"think" about this variable or cleaning up because of __free().
Best regards,
Krzysztof
