Re: spinlock.c:306:9: error: implicit declaration of function '__raw_write_lock_nested'

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Wed, Nov 24, 2021 at 11:01 AM Rob Landley <rob@xxxxxxxxxxx> wrote:
> On 11/23/21 7:19 AM, Arnd Bergmann wrote:
> >
> > These happen with any compiler version, someone needs to write the correct
> > entry code for clone3 and hook up futex_waitv().
>
> I did a naieve "add them both to the .tbl" patch and the result booted to a
> shell prompt, but that doesn't mean much. What arch-specific entry code does
> clone3 need here? The SYSCALL_DEFINE2(clone3) in kernel/fork.c seems reasonably
> straightforward? (Unlike the #ifdef stack around the previous clone...)

I forget the exact issue, but I can see that 4 out of the 13
architectures that set
__ARCH_WANT_SYS_CLONE3 provide a custom version: arc, m68k,
mips and parisc. Have a look at what those do to see if you need the same
changes.

> >> include/linux/sh_intc.h:100:63: warning: division 'sizeof (void *) / sizeof (void)' does not compute the number of array elements
> >
> > These are old bugs, they show up in any kernel version with gcc-8 or higher.
>
> I looked at trying to fix that but it seems to be a compiler bug. Gcc is warning
> about an ? : else case that's dead code eliminated. It's already GOT a test
> protecting it from being evaluated...

I wouldn't call it a bug in the compiler, as there is no definite
correct ordering
between dead-code-elimination and warning generation. IIRC I fixed a bunch
of these on other architectures, and those did turn out to be actual code issues
that would go unnoticed otherwise.

> >> fs/mpage.c:336:1: warning: the frame size of 1092 bytes is larger than
> >
> > I see these going back to gcc-6, it looks like this is caused by
> > CONFIG_PAGE_SIZE_64KB.
>
> In which case the stack size is going to be 64k as well?

No, the stack is still 4KB or 8KB, depending on CONFIG_4KSTACKS, it gets
allocated using

        stack = kmem_cache_alloc_node(thread_stack_cache, THREADINFO_GFP, node);

from a THREAD_SIZE-sized naturally-aligned kmem cache in this case.
Using 1KB of stack space is definitely a red flag that something is going
wrong. This could be a bug in kernel code, in the compiler, or in the
combination of the two.

        Arnd



[Index of Archives]     [Linux Kernel]     [Linux USB Development]     [Yosemite News]     [Linux SCSI]

  Powered by Linux