Re: [PATCH] printk/console: Check consistent sequence number when handling race in console_unlock()

[Petr Mladek <pmladek@suse.com>]

On Tue 2021-06-29 22:59:57, John Ogness wrote:
> On 2021-06-30, kernel test robot <lkp@intel.com> wrote:
> >>> kernel/printk/printk.c:2548:6: warning: variable 'next_seq' set but not used [-Wunused-but-set-variable]
> 
> I suppose the correct fix for this warning would be to change the NOP
> macros. Currently they are:
> 
> #define prb_read_valid(rb, seq, r)      false
> #define prb_first_valid_seq(rb)         0
> 
> They should probably be something like (untested):
> 
> #define prb_read_valid(rb, seq, r)     \
> ({                                     \
>         (void)(rb);                    \
>         (void)(seq);                   \
>         (void)(r);                     \
>         false;                         \
> })

This did not work:

kernel/printk/printk.c: In function ‘console_unlock’:
kernel/printk/printk.c:2600:23: error: ‘prb’ undeclared (first use in this function)
   if (!prb_read_valid(prb, console_seq, &r))
                       ^
kernel/printk/printk.c:2230:16: note: in definition of macro ‘prb_read_valid’
         (void)(rb);                    \
                ^~
kernel/printk/printk.c:2600:23: note: each undeclared identifier is reported only once for each function it appears in
   if (!prb_read_valid(prb, console_seq, &r))
                       ^
kernel/printk/printk.c:2230:16: note: in definition of macro ‘prb_read_valid’
         (void)(rb);                    \
                ^~


Instead, it might be solved by declaring next_seq as:

	u64 __maybe_unused next_seq;

Any better idea is welcome. Well, it is not worth any big complexity.

Best Regards,
Petr