Re: [syzbot] [erofs?] WARNING: CPU: NUM PID: NUM at mm/page_alloc.c:LINE get_page_from_freeli

[Xiang Gao]

Hi Aleksandr,

On 2023/1/5 19:14, Aleksandr Nogikh wrote:
> Hi,
>
> On Thu, Jan 5, 2023 at 11:54 AM Xiang Gao <hsiangkao@xxxxxxxxxxxxxxxxx> wrote:
>
> > I wasn't able to build the kernel with this kernel config, it shows:
> > "...
> > FATAL: modpost: vmlinux.o is truncated. sechdrs[i].sh_offset=1399394064 > sizeof(*hrd)=64
> > make[2]: *** [Module.symvers] Error 1
> > make[1]: *** [modpost] Error 2
> > make: *** [__sub-make] Error 2
> > "
>
> Could you please tell, what exact compiler/linker version did you use?

Thanks for your help.

GCC 9.2.1 on my developping server.

> > > Downloadable assets:
> > > disk image: https://storage.googleapis.com/syzbot-assets/0c8a5f06ceb3/disk-f9ff5644.raw.xz
> > > vmlinux: https://storage.googleapis.com/syzbot-assets/be222e852ae2/vmlinux-f9ff5644.xz
> > > kernel image: https://storage.googleapis.com/syzbot-assets/d9f42a53b05e/bzImage-f9ff5644.xz
> >
> > Finally I tried the original kernel image, and it printed some other
> > random bug when booting system and then reboot, like:
> >
> > [   36.991123][    T1] ==================================================================
> > [   36.991800][    T1] BUG: KASAN: slab-out-of-bounds in copy_array+0x96/0x100
> > [   36.992438][    T1] Write of size 32 at addr ffff888018c34640 by task systemd/1
> < .. >
>
> Interesting!
> I've just tried to boot it with qemu and it was fine.
>
> qemu-system-x86_64 -smp 2,sockets=2,cores=1 -m 4G -drive
> file=disk-f9ff5644.raw,format=raw -snapshot -nographic -enable-kvm
>
> So it looks like it's some difference between these VMMs that causes
> that bug to fire.

I think the problem is that the rootfs which I used has more complicated
workload than the given one.

> > May I ask it can be reproducable on the latest -rc kernel?
>
> We can ask syzbot about v6.2-rc2:
>
> #syz test git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git
> 88603b6dc419445847923fcb7fe5080067a30f98

I think I know the root cause: It seems that kvcalloc doesn't support
GFP_NOFAIL, I will use kcalloc directly instead.

Thanks,
Gao Xiang

> > Thanks,
> > Gao Xiang
>
> --
> Aleksandr