Skip to contentSkip to conversation

LoongArch: BPF: Use move_addr() for BPF_PSEUDO_FUNC

PATCH Applied Hengqi Chen 3 participants 4 messages Mar 17, 2025 – Mar 30, 2025

Subsystems (3)
lore
Share & tools
lore

Hengqi Chen

author reviewer
#0 lore raw

PATCH 1/1 LoongArch: BPF: Use move_addr() for BPF_PSEUDO_FUNC

Vincent reported that running XDP synproxy program on LoongArch results in the following error:
JIT doesn't support bpf-to-bpf calls
With dmesg:

    multi-func JIT bug 1391 != 1390

The root cause is that verifier will refill the imm with the correct addresses of bpf_calls for BPF_PSEUDO_FUNC instructions and then run the last pass of JIT. So we generate different JIT code for the same instruction in two passes (one for placeholder and one for real address). Let's use move_addr() instead.

See commit 64f50f657572 ("LoongArch, bpf: Use 4 instructions for
function address in JIT") for a similar fix.

Fixes 69c087ba6225 ("bpf: Add bpf_for_each_map_elem() helper")
Fixes bb035ef0cc91 ("LoongArch: BPF: Support mixing bpf2bpf and tailcalls")
---
 arch/loongarch/net/bpf_jit.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
arch/loongarch/net/bpf_jit.c +4 -1
index ea357a3edc09..b25b0bb43428 100644
@@ -930,7 +930,10 @@ static int build_insn(const struct bpf_insn *insn, struct jit_ctx *ctx, bool ext
 	{
 		const u64 imm64 = (u64)(insn + 1)->imm << 32 | (u32)insn->imm;
 
-		move_imm(ctx, dst, imm64, is32);
+		if (bpf_pseudo_func(insn))
+			move_addr(ctx, dst, imm64);
+		else
+			move_imm(ctx, dst, imm64, is32);
 		return 1;
 	}
 
Signature · 1 line
2.43.5
#1 lore raw
Quoted 42 lines · Hengqi Chen

Vincent reported that running XDP synproxy program on LoongArch results in the following error:
JIT doesn't support bpf-to-bpf calls
With dmesg:

    multi-func JIT bug 1391 != 1390

The root cause is that verifier will refill the imm with the correct addresses of bpf_calls for BPF_PSEUDO_FUNC instructions and then run the last pass of JIT. So we generate different JIT code for the same instruction in two passes (one for placeholder and one for real address). Let's use move_addr() instead.

See commit 64f50f657572 ("LoongArch, bpf: Use 4 instructions for
function address in JIT") for a similar fix.

Fixes 69c087ba6225 ("bpf: Add bpf_for_each_map_elem() helper")
Fixes bb035ef0cc91 ("LoongArch: BPF: Support mixing bpf2bpf and tailcalls")
---
 arch/loongarch/net/bpf_jit.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
arch/loongarch/net/bpf_jit.c +4 -2
index ea357a3edc09..b25b0bb43428 100644
@@ -930,7 +930,10 @@ static int build_insn(const struct bpf_insn *insn, struct jit_ctx *ctx, bool ext
        {
                const u64 imm64 = (u64)(insn + 1)->imm << 32 | (u32)insn->imm;
 
-               move_imm(ctx, dst, imm64, is32);
+               if (bpf_pseudo_func(insn))
+                       move_addr(ctx, dst, imm64);
+               else
+                       move_imm(ctx, dst, imm64, is32);
                return 1;
        }
 
--
Quoted 2 lines

2.43.5

Thanks Hengqi for the quick fix! tested and verified working now.

[root@fedora xdp-tools]# uname -a
Linux fedora 6.14.0-rc5 #2 SMP PREEMPT_DYNAMIC Sun Mar 16 17:16:21 PDT 2025 loongarch64 GNU/Linux

[root@fedora xdp-tools]# ./xdp-loader/xdp-loader load -vvv lo -m skb -P 80 -p /sys/fs/bpf/xdp-synproxy-tailcall -n synproxy_tailcall ./xdp-synproxy-tailcall/xdp_synproxy_tailcall.bpf.o
Current rlimit 8388608 already >= minimum 1048576
Loading 1 files on interface 'lo'.
libbpf: loading object from
./xdp-synproxy-tailcall/xdp_synproxy_tailcall.bpf.o
...
libbpf: map 'tail_call_tbl': created successfully, fd=4
libbpf: pinned map '/sys/fs/bpf/xdp-synproxy-tailcall/tail_call_tbl'
libbpf: found no pinned map to reuse at
'/sys/fs/bpf/xdp-synproxy-tailcall/values'
libbpf: map 'values': created successfully, fd=5
libbpf: pinned map '/sys/fs/bpf/xdp-synproxy-tailcall/values'
libbpf: found no pinned map to reuse at
'/sys/fs/bpf/xdp-synproxy-tailcall/allowed_ports'
libbpf: map 'allowed_ports': created successfully, fd=6
libbpf: pinned map '/sys/fs/bpf/xdp-synproxy-tailcall/allowed_ports'
libbpf: map 'xdp_synp.rodata': created successfully, fd=7
libbpf: map 'tail_call_tbl': slot [0] set to prog 'syncookie_xdp' fd=65
libxdp: Loaded XDP program synproxy_tailcall, got fd 66
libxdp: Duplicated fd 66 to 3 for prog synproxy_tailcall
libxdp: Replacing XDP fd -1 with 3 on ifindex 1

[root@fedora xdp-tools]# bpftool prog
...
55: xdp name syncookie_xdp tag 1426f5e6593da050 gpl loaded_at 2025-03-16T19:38:49-0700 uid 0
xlated 8392B jited 6412B memlock 16384B map_ids 11,10,12 btf_id 75

56: xdp name synproxy_tailcall tag 0433e599459b925f gpl loaded_at 2025-03-16T19:38:49-0700 uid 0
xlated 192B jited 328B memlock 16384B map_ids 9,12 btf_id 75

Tested-by
#2 lore raw
Quoted 84 lines · Vincent Li
Quoted 42 lines · Hengqi Chen

Vincent reported that running XDP synproxy program on LoongArch results in the following error:
JIT doesn't support bpf-to-bpf calls
With dmesg:

    multi-func JIT bug 1391 != 1390

The root cause is that verifier will refill the imm with the correct addresses of bpf_calls for BPF_PSEUDO_FUNC instructions and then run the last pass of JIT. So we generate different JIT code for the same instruction in two passes (one for placeholder and one for real address). Let's use move_addr() instead.

See commit 64f50f657572 ("LoongArch, bpf: Use 4 instructions for
function address in JIT") for a similar fix.

Fixes 69c087ba6225 ("bpf: Add bpf_for_each_map_elem() helper")
Fixes bb035ef0cc91 ("LoongArch: BPF: Support mixing bpf2bpf and tailcalls")
---
 arch/loongarch/net/bpf_jit.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
arch/loongarch/net/bpf_jit.c +4 -2
index ea357a3edc09..b25b0bb43428 100644
@@ -930,7 +930,10 @@ static int build_insn(const struct bpf_insn *insn, struct jit_ctx *ctx, bool ext
        {
                const u64 imm64 = (u64)(insn + 1)->imm << 32 | (u32)insn->imm;
 
-               move_imm(ctx, dst, imm64, is32);
+               if (bpf_pseudo_func(insn))
+                       move_addr(ctx, dst, imm64);
+               else
+                       move_imm(ctx, dst, imm64, is32);
                return 1;
        }
 
--
Quoted 2 lines

2.43.5

Thanks Hengqi for the quick fix! tested and verified working now.

[root@fedora xdp-tools]# uname -a
Linux fedora 6.14.0-rc5 #2 SMP PREEMPT_DYNAMIC Sun Mar 16 17:16:21 PDT 2025 loongarch64 GNU/Linux

[root@fedora xdp-tools]# ./xdp-loader/xdp-loader load -vvv lo -m skb -P 80 -p /sys/fs/bpf/xdp-synproxy-tailcall -n synproxy_tailcall ./xdp-synproxy-tailcall/xdp_synproxy_tailcall.bpf.o
Current rlimit 8388608 already >= minimum 1048576
Loading 1 files on interface 'lo'.
libbpf: loading object from
./xdp-synproxy-tailcall/xdp_synproxy_tailcall.bpf.o
...
libbpf: map 'tail_call_tbl': created successfully, fd=4
libbpf: pinned map '/sys/fs/bpf/xdp-synproxy-tailcall/tail_call_tbl'
libbpf: found no pinned map to reuse at
'/sys/fs/bpf/xdp-synproxy-tailcall/values'
libbpf: map 'values': created successfully, fd=5
libbpf: pinned map '/sys/fs/bpf/xdp-synproxy-tailcall/values'
libbpf: found no pinned map to reuse at
'/sys/fs/bpf/xdp-synproxy-tailcall/allowed_ports'
libbpf: map 'allowed_ports': created successfully, fd=6
libbpf: pinned map '/sys/fs/bpf/xdp-synproxy-tailcall/allowed_ports'
libbpf: map 'xdp_synp.rodata': created successfully, fd=7
libbpf: map 'tail_call_tbl': slot [0] set to prog 'syncookie_xdp' fd=65
libxdp: Loaded XDP program synproxy_tailcall, got fd 66
libxdp: Duplicated fd 66 to 3 for prog synproxy_tailcall
libxdp: Replacing XDP fd -1 with 3 on ifindex 1

[root@fedora xdp-tools]# bpftool prog
...
55: xdp name syncookie_xdp tag 1426f5e6593da050 gpl loaded_at 2025-03-16T19:38:49-0700 uid 0
xlated 8392B jited 6412B memlock 16384B map_ids 11,10,12 btf_id 75

Quoted 4 lines

56: xdp name synproxy_tailcall tag 0433e599459b925f gpl loaded_at 2025-03-16T19:38:49-0700 uid 0
xlated 192B jited 328B memlock 16384B map_ids 9,12 btf_id 75

Sorry missed the Tested-by, you can add:

maintainer Applied
#3 lore raw
In reply to Hengqi Chen

Applied, thanks.

Huacai

Quoted 45 lines · Hengqi Chen

Vincent reported that running XDP synproxy program on LoongArch results in the following error:
JIT doesn't support bpf-to-bpf calls
With dmesg:

    multi-func JIT bug 1391 != 1390

The root cause is that verifier will refill the imm with the correct addresses of bpf_calls for BPF_PSEUDO_FUNC instructions and then run the last pass of JIT. So we generate different JIT code for the same instruction in two passes (one for placeholder and one for real address). Let's use move_addr() instead.

See commit 64f50f657572 ("LoongArch, bpf: Use 4 instructions for
function address in JIT") for a similar fix.

Fixes 69c087ba6225 ("bpf: Add bpf_for_each_map_elem() helper")
Fixes bb035ef0cc91 ("LoongArch: BPF: Support mixing bpf2bpf and tailcalls")
---
 arch/loongarch/net/bpf_jit.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
arch/loongarch/net/bpf_jit.c +4 -2
index ea357a3edc09..b25b0bb43428 100644
@@ -930,7 +930,10 @@ static int build_insn(const struct bpf_insn *insn, struct jit_ctx *ctx, bool ext
        {
                const u64 imm64 = (u64)(insn + 1)->imm << 32 | (u32)insn->imm;
 
-               move_imm(ctx, dst, imm64, is32);
+               if (bpf_pseudo_func(insn))
+                       move_addr(ctx, dst, imm64);
+               else
+                       move_imm(ctx, dst, imm64, is32);
                return 1;
        }
 
--

2.43.5