Re: [oliver.sang@xxxxxxxxx: [cpumask] b9a7ecc71f: WARNING:at_include/linux/cpumask.h:#__is_kernel_percpu_address]

From: Andrew Jones
Date: Tue Oct 11 2022 - 13:09:58 EST


On Sat, Oct 01, 2022 at 10:16:36AM -0700, Yury Norov wrote:
> On Sat, Oct 01, 2022 at 09:20:53AM -0700, Linus Torvalds wrote:
> > On Fri, Sep 30, 2022 at 6:51 PM Yury Norov <yury.norov@xxxxxxxxx> wrote:
> > >
> > > The commit b9a7ecc71fe582e ("cpumask: fix checking valid cpu range")
> > > fixes broken cpumask_check(), which for now doesn't warn user when it
> > > should. After the fix, I observed many false-positive warnings which
> > > were addressed in the following patches.
> >
> > Are all the false positives fixed?
>
> I build-tested on x86_64 and arm64. All fixed except for those
> generated by cpumask_next_wrap(). And I'm not even sure they
> are false positives.
>
> This is what I'm working on right now. Hope moving it in next
> merge window.

Hi Yury,

I just wanted to report that the warning fires when doing
'cat /proc/cpuinfo' on at least x86 and riscv. I don't think
those are false positives. I'm guessing a patch should be
something like the following diff. If you haven't already
addressed this and I'm not off in left field, then I guess
we should integrate it into your series.

Thanks,
drew


diff --git a/arch/riscv/kernel/cpu.c b/arch/riscv/kernel/cpu.c
index 4aa8cd749441..4c5dfa230d4b 100644
--- a/arch/riscv/kernel/cpu.c
+++ b/arch/riscv/kernel/cpu.c
@@ -166,9 +166,12 @@ static void print_mmu(struct seq_file *f)

static void *c_start(struct seq_file *m, loff_t *pos)
{
- *pos = cpumask_next(*pos - 1, cpu_online_mask);
- if ((*pos) < nr_cpu_ids)
- return (void *)(uintptr_t)(1 + *pos);
+ if (*pos < nr_cpu_ids) {
+ *pos = cpumask_next(*pos - 1, cpu_online_mask);
+ if ((*pos) < nr_cpu_ids)
+ return (void *)(uintptr_t)(1 + *pos);
+ }
+
return NULL;
}

diff --git a/arch/x86/kernel/cpu/proc.c b/arch/x86/kernel/cpu/proc.c
index 099b6f0d96bd..2ea614e78e28 100644
--- a/arch/x86/kernel/cpu/proc.c
+++ b/arch/x86/kernel/cpu/proc.c
@@ -153,9 +153,12 @@ static int show_cpuinfo(struct seq_file *m, void *v)

static void *c_start(struct seq_file *m, loff_t *pos)
{
- *pos = cpumask_next(*pos - 1, cpu_online_mask);
- if ((*pos) < nr_cpu_ids)
- return &cpu_data(*pos);
+ if (*pos < nr_cpu_ids) {
+ *pos = cpumask_next(*pos - 1, cpu_online_mask);
+ if ((*pos) < nr_cpu_ids)
+ return &cpu_data(*pos);
+ }
+
return NULL;
}

>
> > I suspect that to avoid any automation noise, you should just rebase
> > so that the fixes come first. Otherwise we'll end up wasting a lot of
> > time on the noise.
> >
> > This is not that different from introducing new buil;d-time warnings:
> > the things they point out need to be fixed before the warning can be
> > integrated, or it causes bisection problems.
>
> OK, I'll reorder the patches. Thanks for your help.
>