percpu: km: ensure it is used with NOMMU (either UP or SMP)
authorVladimir Murzin <vladimir.murzin@arm.com>
Tue, 30 Nov 2021 17:29:54 +0000 (17:29 +0000)
committerDennis Zhou <dennis@kernel.org>
Mon, 6 Dec 2021 17:45:09 +0000 (12:45 -0500)
Currently, NOMMU pull km allocator via !SMP dependency because most of
them are UP, yet for SMP+NOMMU vm allocator gets pulled which:

* may lead to broken build [1]
* ...or not working runtime due to [2]

It looks like SMP+NOMMU case was overlooked in bbddff054587 ("percpu:
use percpu allocator on UP too") so restore that.

[1]
For ARM SMP+NOMMU (R-class cores)

arm-none-linux-gnueabihf-ld: mm/percpu.o: in function `pcpu_post_unmap_tlb_flush':
mm/percpu-vm.c:188: undefined reference to `flush_tlb_kernel_range'

[2]
static inline
int vmap_pages_range_noflush(unsigned long addr, unsigned long end,
                pgprot_t prot, struct page **pages, unsigned int page_shift)
{
       return -EINVAL;
}

Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
Tested-by: Rob Landley <rob@landley.net>
Tested-by: Rich Felker <dalias@libc.org>
[Dennis: use depends instead of default for condition]
Signed-off-by: Dennis Zhou <dennis@kernel.org>
mm/Kconfig

index 28edafc820adfc9b434cbc76fbb774992a45ff5b..356f4f2c779e5fda55abcd8bc5b652d85e5af054 100644 (file)
@@ -428,7 +428,7 @@ config THP_SWAP
 # UP and nommu archs use km based percpu allocator
 #
 config NEED_PER_CPU_KM
-       depends on !SMP
+       depends on !SMP || !MMU
        bool
        default y