2024-04-04 12:49:35

by Arnd Bergmann

Subject: [PATCH] bpf: x86: avoid link error with CONFIG_SMP=n

From: Arnd Bergmann <[email protected]>

On UP systems, this_cpu_off is not defined, so the new percpu code in bpf
fails to link:

x86_64-linux-ld: vmlinux.o: in function `do_jit':
bpf_jit_comp.c:(.text+0xbab14): undefined reference to `this_cpu_off'

Use offset zero on UP builds instead.

Fixes: 7bdbf7446305 ("bpf: add special internal-only MOV instruction to resolve per-CPU addrs")
Signed-off-by: Arnd Bergmann <[email protected]>
---
I assume this is not the correct fix, or at least not the best one, so
please treat this as a bug report. It does address the link failure for
me, so I applied this to my randconfig build tree.
---
arch/x86/net/bpf_jit_comp.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/arch/x86/net/bpf_jit_comp.c b/arch/x86/net/bpf_jit_comp.c
index d6ebb9136f3c..8b8eebb89a9b 100644
--- a/arch/x86/net/bpf_jit_comp.c
+++ b/arch/x86/net/bpf_jit_comp.c
@@ -1383,7 +1383,10 @@ static int do_jit(struct bpf_prog *bpf_prog, int *addrs, u8 *image, u8 *rw_image
EMIT3(0x0F, 0x44, add_2reg(0xC0, AUX_REG, dst_reg));
break;
} else if (insn_is_mov_percpu_addr(insn)) {
- u32 off = (u32)(unsigned long)&this_cpu_off;
+ u32 off = 0;
+
+ if (IS_ENABLED(CONFIG_SMP))
+ off = (u32)(unsigned long)&this_cpu_off;

/* mov <dst>, <src> (if necessary) */
EMIT_mov(dst_reg, src_reg);
--
2.39.2

2024-04-04 14:43:05

by Daniel Borkmann

Subject: Re: [PATCH] bpf: x86: avoid link error with CONFIG_SMP=n

On 4/4/24 2:33 PM, Arnd Bergmann wrote:
> From: Arnd Bergmann <[email protected]>
>
> On UP systems, this_cpu_off is not defined, so the new percpu code in bpf
> fails to link:
>
> x86_64-linux-ld: vmlinux.o: in function `do_jit':
> bpf_jit_comp.c:(.text+0xbab14): undefined reference to `this_cpu_off'
>
> Use offset zero on UP builds instead.
>
> Fixes: 7bdbf7446305 ("bpf: add special internal-only MOV instruction to resolve per-CPU addrs")
> Signed-off-by: Arnd Bergmann <[email protected]>
> ---
> I assume this is not the correct fix, or at least not the best one, so
> please treat this as a bug report. It does address the link failure for
> me, so I applied this to my randconfig build tree.
> ---
> arch/x86/net/bpf_jit_comp.c | 5 ++++-
> 1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/arch/x86/net/bpf_jit_comp.c b/arch/x86/net/bpf_jit_comp.c
> index d6ebb9136f3c..8b8eebb89a9b 100644
> --- a/arch/x86/net/bpf_jit_comp.c
> +++ b/arch/x86/net/bpf_jit_comp.c
> @@ -1383,7 +1383,10 @@ static int do_jit(struct bpf_prog *bpf_prog, int *addrs, u8 *image, u8 *rw_image
> EMIT3(0x0F, 0x44, add_2reg(0xC0, AUX_REG, dst_reg));
> break;
> } else if (insn_is_mov_percpu_addr(insn)) {
> - u32 off = (u32)(unsigned long)&this_cpu_off;
> + u32 off = 0;
> +
> + if (IS_ENABLED(CONFIG_SMP))
> + off = (u32)(unsigned long)&this_cpu_off;
>

Thanks, there was already a fix in bpf-next tree in here:
https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next.git/commit/?id=1e9e0b85255e6eca6036b59d8a5fbca6501905ac

> /* mov <dst>, <src> (if necessary) */
> EMIT_mov(dst_reg, src_reg);
>