[PATCH 1/4] RISC-V: Do not allocate memblock while iterating reserved memblocks

[Geert Uytterhoeven]

Hi Atish,

On Thu, Jan 7, 2021 at 10:28 AM Atish Patra <atish.patra at wdc.com> wrote:
> Currently, resource tree allocates memory blocks while iterating on the
> list. It leads to following kernel warning because memblock allocation
> also invokes memory block reservation API.
>
> [    0.000000] ------------[ cut here ]------------
> [    0.000000] WARNING: CPU: 0 PID: 0 at kernel/resource.c:795
> __insert_resource+0x8e/0xd0
> [    0.000000] Modules linked in:
> [    0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted
> 5.10.0-00022-ge20097fb37e2-dirty #549
> [    0.000000] epc: c00125c2 ra : c001262c sp : c1c01f50
> [    0.000000]  gp : c1d456e0 tp : c1c0a980 t0 : ffffcf20
> [    0.000000]  t1 : 00000000 t2 : 00000000 s0 : c1c01f60
> [    0.000000]  s1 : ffffcf00 a0 : ffffff00 a1 : c1c0c0c4
> [    0.000000]  a2 : 80c12b15 a3 : 80402000 a4 : 80402000
> [    0.000000]  a5 : c1c0c0c4 a6 : 80c12b15 a7 : f5faf600
> [    0.000000]  s2 : c1c0c0c4 s3 : c1c0e000 s4 : c1009a80
> [    0.000000]  s5 : c1c0c000 s6 : c1d48000 s7 : c1613b4c
> [    0.000000]  s8 : 00000fff s9 : 80000200 s10: c1613b40
> [    0.000000]  s11: 00000000 t3 : c1d4a000 t4 : ffffffff
>
> This is also unnecessary as we can pre-compute the total memblocks required
> for each memory region and allocate it before the loop. It save precious
> boot time not going through memblock allocation code every time.
>
> Fixes: 00ab027a3b82 ("RISC-V: Add kernel image sections to the resource tree")
>
> Signed-off-by: Atish Patra <atish.patra at wdc.com>

Thanks for your patch!

I never saw the warning (on linux-on-litex-vexriscv), but instead I got:

    Failed to add a Kernel code resource at 40001000

after Initmem setup. Adding some debug info to init_resources() showed
that the memblock.reserved list kept on increasing, until memory
corruption happened (pointers started to look like ASCII strings), the
error message above is printed, and after which the boot continued.
Changing L1_CACHE_SHIFT in arch/riscv/include/asm/cache.h to 5 fixed that.

With this patch, the error message above is no longer printed, so
Tested-by: Geert Uytterhoeven <geert at linux-m68k.org>

Noted that the kernel still crashes later, with

    Unable to handle kernel paging request at virtual address 61636473

Again, 0x61636473 looks like ASCII, and your PATCH 3/4 for
L1_CACHE_SHIFT fixes that, too.

Gr{oetje,eeting}s,

                        Geert


--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert at linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds