Re: [PATCH] firmware: dmi_scan: Always show system identification string

[Ard Biesheuvel]

On 13 December 2016 at 14:42, Jean Delvare <jdelvare@xxxxxxx> wrote:
> On mar., 2016-12-13 at 14:25 +0000, Ard Biesheuvel wrote:
>> On 12 December 2016 at 01:51, Kefeng Wang <wangkefeng.wang@xxxxxxxxxx> wrote:
>> >
>> >
>> > On 2016/12/9 22:55, Ard Biesheuvel wrote:
>> >> On 9 December 2016 at 06:54, Kefeng Wang <wangkefeng.wang@xxxxxxxxxx> wrote:
>> >>> Let's keep consistent when print dmi_ids_string between SMBIOS 2.x
>> >>> and SMBIOS 3.x, and always show the system identification string,
>> >>> like Vendor, Product/Board name and BIOS infos.
>> >>>
>> >>
>> >> Are you saying the printk(KERN_DEBUG) was inconsistent with the
>> >> pr_debug() before?
>> >
>> > Yes, found it when check the boot dmesg on qemu, there is no print using pr_debug
>> > with defconfig on ARM64( smbios3 present), but could shown with printk(KERN_DEBUG.
>> >
>>
>> That's interesting. I would assume pr_debug and printk(KERN_DEBUG) to
>> behave identically ...
>
> And you'd be wrong ;-)
>
> From include/linux/printk.h:
>
> 304 #if defined(CONFIG_DYNAMIC_DEBUG)
> 305 #include <linux/dynamic_debug.h>
> 306
> 307 /* dynamic_pr_debug() uses pr_fmt() internally so we don't need it here */
> 308 #define pr_debug(fmt, ...) \
> 309         dynamic_pr_debug(fmt, ##__VA_ARGS__)
> 310 #elif defined(DEBUG)
> 311 #define pr_debug(fmt, ...) \
> 312         printk(KERN_DEBUG pr_fmt(fmt), ##__VA_ARGS__)
> 313 #else
> 314 #define pr_debug(fmt, ...) \
> 315         no_printk(KERN_DEBUG pr_fmt(fmt), ##__VA_ARGS__)
> 316 #endif
>
> Which basically means that pr_debug() has no effect unless debugging is
> actually enabled, while printk(KERN_DEBUG ...) is always printed by the
> kernel (with a log level which may or may not make it to a specific
> output.)
>

Good to know. Thanks for digging that up ...