On 07/11/2018 14:54:17+0000, [email protected] wrote: > Hi Alexandre, > > On 06.11.2018 23:09, Alexandre Belloni wrote: > > Hi Claudiu, > > > > On 05/11/2018 11:14:26+0000, [email protected] wrote: > >> static int __init at91_poweroff_probe(struct platform_device *pdev) > >> @@ -154,16 +160,22 @@ static int __init at91_poweroff_probe(struct > >> platform_device *pdev) > >> u32 ddr_type; > >> int ret; > >> > >> + at91_shdwc = devm_kzalloc(&pdev->dev, sizeof(*at91_shdwc), GFP_KERNEL); > >> + if (!at91_shdwc) > >> + return -ENOMEM; > >> + > > > > Is there any real benefit that will offset the time lost for that > > allocation at boot time? > > No, I haven't run benchmarks on this. I only wanted to have them grouped in > one structure. Please let me know if you have some tests in mind. >
Well, it is probably not much but small things adds up. Havinf it as a global structure is probably good enough. > > > > I understand you are then testing at91_shdwc to know whether the driver > > already probed once. But, the driver will never probe twice as there is > > only one shutdown controller on the SoC and anyway, If it was to probe > > twice, it will still work as expected. > > I had in mind the scenario where the driver would be compiled as module. I > know insmod already does this checking. I'm ok to remove this checking. I > will do it in next version. With this I will also remove devm_kzalloc() of > at91_shdwc. > Thanks, -- Alexandre Belloni, Bootlin Embedded Linux and Kernel engineering https://bootlin.com
