bpf: Fix linked register tracking
Subsystems (4)
cover bpf: Fix linked register tracking
This patch fixes the linked register tracking when multiple links from the same register are created with a sync between the creation of these links. The sync corrupts the id of the register and therefore the second link is not created properly. See the patch description to understand more.
The fix is to preserve the id while doing the sync similar to the off.
Puranjay Mohan (2):
bpf: Preserve id of register in sync_linked_regs()
selftests: bpf: Add test for multiple syncs from linked register
kernel/bpf/verifier.c | 4 ++- .../bpf/progs/verifier_linked_scalars.c | 33 +++++++++++++++++++ 2 files changed, 36 insertions(+), 1 deletion(-)
base-commit: 9a403a4aea32f1801a7f29b2385ec345d4faaf78
Signature · 1 line
2.47.3
PATCH 1/2 bpf: Preserve id of register in sync_linked_regs()
sync_linked_regs() copies the id of known_reg to reg when propagating bounds of known_reg to reg using the off of known_reg, but when known_reg was linked to reg like:
known_reg = reg ; both known_reg and reg get same idknown_reg += 4 ; known_reg gets off = 4, and its id gets BPF_ADD_CONST
now when a call to sync_linked_regs() happens, let's say with the following:
if known_reg >= 10 goto pc+2
known_reg's new bounds are propagated to reg but now reg gets BPF_ADD_CONST from the copy.
This means if another link to reg is created like:
another_reg = reg ; another_reg should get the id of reg but
assign_scalar_id_before_mov() sees BPF_ADD_CONST on reg and assigns a new id to it.
As reg has a new id now, known_reg's link to reg is broken. If we find new bounds for known_reg, they will not be propagated to reg.
This can be seen in the selftest added in the next commit:
0: (85) call bpf_get_prandom_u32#7 ; R0=scalar() 1: (57) r0 &= 255 ; R0=scalar(smin=smin32=0,smax=umax=smax32=umax32=255,var_off=(0x0; 0xff)) 2: (bf) r1 = r0 ; R0=scalar(id=1,smin=smin32=0,smax=umax=smax32=umax32=255,var_off=(0x0; 0xff)) R1=scalar(id=1,smin=smin32=0,smax=umax=smax32=umax32=255,var_off=(0x0; 0xff)) 3: (07) r1 += 4 ; R1=scalar(id=1+4,smin=umin=smin32=umin32=4,smax=umax=smax32=umax32=259,var_off=(0x0; 0x1ff)) 4: (a5) if r1 < 0xa goto pc+4 ; R1=scalar(id=1+4,smin=umin=smin32=umin32=10,smax=umax=smax32=umax32=259,var_off=(0x0; 0x1ff)) 5: (bf) r2 = r0 ; R0=scalar(id=2,smin=umin=smin32=umin32=6,smax=umax=smax32=umax32=255) R2=scalar(id=2,smin=umin=smin32=umin32=6,smax=umax=smax32=umax32=255) 6: (a5) if r1 < 0xe goto pc+2 ; R1=scalar(id=1+4,smin=umin=smin32=umin32=14,smax=umax=smax32=umax32=259,var_off=(0x0; 0x1ff)) 7: (35) if r0 >= 0xa goto pc+1 ; R0=scalar(id=2,smin=umin=smin32=umin32=6,smax=umax=smax32=umax32=9,var_off=(0x0; 0xf))
8: (37) r0 /= 0
div by zero
When 4 is verified, r1's bounds are propagated to r0 but r0 also gets BPF_ADD_CONST (bug).
When 5 is verified, r0 gets a new id (2) and its link with r1 is broken.
After 6 we know r1 has bounds [14, 259] and therefore r0 should have bounds [10, 255], therefore the branch at 7 is always taken. But because r0's id was changed to 2, r1's new bounds are not propagated to r0. The verifier still thinks r0 has bounds [6, 255] before 7 and execution can reach div by zero.
Fix this by preserving id in sync_linked_regs() like off and subreg_def.
--- kernel/bpf/verifier.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
kernel/bpf/verifier.c +3 -1
@@ -16871,6 +16871,7 @@ static void sync_linked_regs(struct bpf_verifier_state *vstate, struct bpf_reg_s }else {s32 saved_subreg_def = reg->subreg_def;s32 saved_off = reg->off; +u32 saved_id = reg->id; fake_reg.type =SCALAR_VALUE ;__mark_reg_known (&fake_reg, (s32 )reg->off - (s32 )known_reg->off);
@@ -16878,10 +16879,11 @@ static void sync_linked_regs(struct bpf_verifier_state *vstate, struct bpf_reg_s/* reg = known_reg; reg += delta */ copy_register_state (reg, known_reg);/* -* Must preserve off, id and add_const flag, +* Must preserve off, id and subreg_def flag, * otherwise another sync_linked_regs() will be incorrect. */ reg->off = saved_off; + reg->id = saved_id; reg->subreg_def = saved_subreg_def;scalar32_min_max_add (reg, &fake_reg);
Signature · 1 line
2.47.3
Re: [PATCH bpf-next 1/2] bpf: Preserve id of register in sync_linked_regs()
Quoted 46 lines · Puranjay Mohan
sync_linked_regs()copies the id ofknown_regto reg when propagating bounds ofknown_regto reg using the off ofknown_reg, but whenknown_regwas linked to reg like:
known_reg= reg ; bothknown_regand reg get same idknown_reg+= 4 ;known_reggets off = 4, and its id getsBPF_ADD_CONSTnow when a call to
sync_linked_regs()happens, let's say with the following:if
known_reg>= 10 goto pc+2known_reg's new bounds are propagated to reg but now reg gets
BPF_ADD_CONSTfrom the copy.This means if another link to reg is created like:
another_reg= reg ;another_regshould get the id of reg but
assign_scalar_id_before_mov()seesBPF_ADD_CONSTon reg and assigns a new id to it.As reg has a new id now, known_reg's link to reg is broken. If we find new bounds for
known_reg, they will not be propagated to reg.This can be seen in the selftest added in the next commit:
0: (85) call bpf_get_prandom_u32#7 ; R0=scalar() 1: (57) r0 &= 255 ; R0=scalar(smin=smin32=0,smax=umax=smax32=umax32=255,var_off=(0x0; 0xff)) 2: (bf) r1 = r0 ; R0=scalar(id=1,smin=smin32=0,smax=umax=smax32=umax32=255,var_off=(0x0; 0xff)) R1=scalar(id=1,smin=smin32=0,smax=umax=smax32=umax32=255,var_off=(0x0; 0xff)) 3: (07) r1 += 4 ; R1=scalar(id=1+4,smin=umin=smin32=umin32=4,smax=umax=smax32=umax32=259,var_off=(0x0; 0x1ff)) 4: (a5) if r1 < 0xa goto pc+4 ; R1=scalar(id=1+4,smin=umin=smin32=umin32=10,smax=umax=smax32=umax32=259,var_off=(0x0; 0x1ff)) 5: (bf) r2 = r0 ; R0=scalar(id=2,smin=umin=smin32=umin32=6,smax=umax=smax32=umax32=255) R2=scalar(id=2,smin=umin=smin32=umin32=6,smax=umax=smax32=umax32=255) 6: (a5) if r1 < 0xe goto pc+2 ; R1=scalar(id=1+4,smin=umin=smin32=umin32=14,smax=umax=smax32=umax32=259,var_off=(0x0; 0x1ff)) 7: (35) if r0 >= 0xa goto pc+1 ; R0=scalar(id=2,smin=umin=smin32=umin32=6,smax=umax=smax32=umax32=9,var_off=(0x0; 0xf))8: (37) r0 /= 0
div by zeroWhen 4 is verified, r1's bounds are propagated to r0 but r0 also gets
BPF_ADD_CONST(bug).
When 5 is verified, r0 gets a new id (2) and its link with r1 is broken.After 6 we know r1 has bounds [14, 259] and therefore r0 should have bounds [10, 255], therefore the branch at 7 is always taken. But because r0's id was changed to 2, r1's new bounds are not propagated to r0. The verifier still thinks r0 has bounds [6, 255] before 7 and execution can reach div by zero.
Quoted 5 lines
Fix this by preserving id in
sync_linked_regs()like off andsubreg_def.Fixes 98d7ca374ba4 ("bpf: Track delta between "linked" registers.")Signed-off-by Puranjay Mohan <[email protected]>---
[...]
Re: [PATCH bpf-next 1/2] bpf: Preserve id of register in sync_linked_regs()
In reply to Puranjay MohanQuoted 41 lines · Puranjay Mohan
sync_linked_regs()copies the id ofknown_regto reg when propagating bounds ofknown_regto reg using the off ofknown_reg, but whenknown_regwas linked to reg like:
known_reg= reg ; bothknown_regand reg get same idknown_reg+= 4 ;known_reggets off = 4, and its id getsBPF_ADD_CONSTnow when a call to
sync_linked_regs()happens, let's say with the following:if
known_reg>= 10 goto pc+2known_reg's new bounds are propagated to reg but now reg gets
BPF_ADD_CONSTfrom the copy.This means if another link to reg is created like:
another_reg= reg ;another_regshould get the id of reg but
assign_scalar_id_before_mov()seesBPF_ADD_CONSTon reg and assigns a new id to it.As reg has a new id now, known_reg's link to reg is broken. If we find new bounds for
known_reg, they will not be propagated to reg.This can be seen in the selftest added in the next commit:
0: (85) call bpf_get_prandom_u32#7 ; R0=scalar() 1: (57) r0 &= 255 ; R0=scalar(smin=smin32=0,smax=umax=smax32=umax32=255,var_off=(0x0; 0xff)) 2: (bf) r1 = r0 ; R0=scalar(id=1,smin=smin32=0,smax=umax=smax32=umax32=255,var_off=(0x0; 0xff)) R1=scalar(id=1,smin=smin32=0,smax=umax=smax32=umax32=255,var_off=(0x0; 0xff)) 3: (07) r1 += 4 ; R1=scalar(id=1+4,smin=umin=smin32=umin32=4,smax=umax=smax32=umax32=259,var_off=(0x0; 0x1ff)) 4: (a5) if r1 < 0xa goto pc+4 ; R1=scalar(id=1+4,smin=umin=smin32=umin32=10,smax=umax=smax32=umax32=259,var_off=(0x0; 0x1ff)) 5: (bf) r2 = r0 ; R0=scalar(id=2,smin=umin=smin32=umin32=6,smax=umax=smax32=umax32=255) R2=scalar(id=2,smin=umin=smin32=umin32=6,smax=umax=smax32=umax32=255) 6: (a5) if r1 < 0xe goto pc+2 ; R1=scalar(id=1+4,smin=umin=smin32=umin32=14,smax=umax=smax32=umax32=259,var_off=(0x0; 0x1ff)) 7: (35) if r0 >= 0xa goto pc+1 ; R0=scalar(id=2,smin=umin=smin32=umin32=6,smax=umax=smax32=umax32=9,var_off=(0x0; 0xf))8: (37) r0 /= 0
div by zeroWhen 4 is verified, r1's bounds are propagated to r0 but r0 also gets
BPF_ADD_CONST(bug).
When 5 is verified, r0 gets a new id (2) and its link with r1 is broken.
Quoted 7 lines
After 6 we know r1 has bounds [14, 259] and therefore r0 should have bounds [10, 255], therefore the branch at 7 is always taken. But because r0's id was changed to 2, r1's new bounds are not propagated to r0. The verifier still thinks r0 has bounds [6, 255] before 7 and execution can reach div by zero.
Fix this by preserving id in
sync_linked_regs()like off andsubreg_def.
We should mark_reg_scratched() all the registers that got new IDs or new ranges or offsets. Basically, if anything about register state was changed, even if verified instruction doesn't work with that register directly, all affected registers (we can think about this as side effects) should be "scratched" and emitted in the verifier log.
Quoted 35 lines
Fixes 98d7ca374ba4 ("bpf: Track delta between "linked" registers.")Signed-off-by Puranjay Mohan <[email protected]>--- kernel/bpf/verifier.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)kernel/bpf/verifier.c +3
-2@@ -16871,6 +16871,7 @@ static void sync_linked_regs(struct bpf_verifier_state *vstate, struct bpf_reg_s }else {s32 saved_subreg_def = reg->subreg_def;s32 saved_off = reg->off; +u32 saved_id = reg->id; fake_reg.type =SCALAR_VALUE ;__mark_reg_known (&fake_reg, (s32 )reg->off - (s32 )known_reg->off);@@ -16878,10 +16879,11 @@ static void sync_linked_regs(struct bpf_verifier_state *vstate, struct bpf_reg_s/* reg = known_reg; reg += delta */ copy_register_state (reg, known_reg);/* -* Must preserve off, id and add_const flag, +* Must preserve off, id and subreg_def flag, * otherwise another sync_linked_regs() will be incorrect. */ reg->off = saved_off; + reg->id = saved_id; reg->subreg_def = saved_subreg_def;scalar32_min_max_add (reg, &fake_reg); --2.47.3
Re: [PATCH bpf-next 1/2] bpf: Preserve id of register in sync_linked_regs()
Quoted 49 lines · Andrii Nakryiko
Quoted 47 lines · Puranjay Mohan
sync_linked_regs()copies the id ofknown_regto reg when propagating bounds ofknown_regto reg using the off ofknown_reg, but whenknown_regwas linked to reg like:
known_reg= reg ; bothknown_regand reg get same idknown_reg+= 4 ;known_reggets off = 4, and its id getsBPF_ADD_CONSTnow when a call to
sync_linked_regs()happens, let's say with the following:if
known_reg>= 10 goto pc+2known_reg's new bounds are propagated to reg but now reg gets
BPF_ADD_CONSTfrom the copy.This means if another link to reg is created like:
another_reg= reg ;another_regshould get the id of reg but
assign_scalar_id_before_mov()seesBPF_ADD_CONSTon reg and assigns a new id to it.As reg has a new id now, known_reg's link to reg is broken. If we find new bounds for
known_reg, they will not be propagated to reg.This can be seen in the selftest added in the next commit:
0: (85) call bpf_get_prandom_u32#7 ; R0=scalar() 1: (57) r0 &= 255 ; R0=scalar(smin=smin32=0,smax=umax=smax32=umax32=255,var_off=(0x0; 0xff)) 2: (bf) r1 = r0 ; R0=scalar(id=1,smin=smin32=0,smax=umax=smax32=umax32=255,var_off=(0x0; 0xff)) R1=scalar(id=1,smin=smin32=0,smax=umax=smax32=umax32=255,var_off=(0x0; 0xff)) 3: (07) r1 += 4 ; R1=scalar(id=1+4,smin=umin=smin32=umin32=4,smax=umax=smax32=umax32=259,var_off=(0x0; 0x1ff)) 4: (a5) if r1 < 0xa goto pc+4 ; R1=scalar(id=1+4,smin=umin=smin32=umin32=10,smax=umax=smax32=umax32=259,var_off=(0x0; 0x1ff)) 5: (bf) r2 = r0 ; R0=scalar(id=2,smin=umin=smin32=umin32=6,smax=umax=smax32=umax32=255) R2=scalar(id=2,smin=umin=smin32=umin32=6,smax=umax=smax32=umax32=255) 6: (a5) if r1 < 0xe goto pc+2 ; R1=scalar(id=1+4,smin=umin=smin32=umin32=14,smax=umax=smax32=umax32=259,var_off=(0x0; 0x1ff)) 7: (35) if r0 >= 0xa goto pc+1 ; R0=scalar(id=2,smin=umin=smin32=umin32=6,smax=umax=smax32=umax32=9,var_off=(0x0; 0xf))8: (37) r0 /= 0
div by zeroWhen 4 is verified, r1's bounds are propagated to r0 but r0 also gets
BPF_ADD_CONST(bug).
When 5 is verified, r0 gets a new id (2) and its link with r1 is broken.After 6 we know r1 has bounds [14, 259] and therefore r0 should have bounds [10, 255], therefore the branch at 7 is always taken. But because r0's id was changed to 2, r1's new bounds are not propagated to r0. The verifier still thinks r0 has bounds [6, 255] before 7 and execution can reach div by zero.
Quoted 8 lines
Quoted 1 line
Fix this by preserving id in
sync_linked_regs()like off andsubreg_def.We should
mark_reg_scratched()all the registers that got new IDs or new ranges or offsets. Basically, if anything about register state was changed, even if verified instruction doesn't work with that register directly, all affected registers (we can think about this as side effects) should be "scratched" and emitted in the verifier log.
I agree, will send a separate patch for this.
Re: [PATCH bpf-next 1/2] bpf: Preserve id of register in sync_linked_regs()
In reply to Puranjay MohanQuoted 68 lines · Puranjay Mohan
sync_linked_regs()copies the id ofknown_regto reg when propagating bounds ofknown_regto reg using the off ofknown_reg, but whenknown_regwas linked to reg like:
known_reg= reg ; bothknown_regand reg get same idknown_reg+= 4 ;known_reggets off = 4, and its id getsBPF_ADD_CONSTnow when a call to
sync_linked_regs()happens, let's say with the following:if
known_reg>= 10 goto pc+2known_reg's new bounds are propagated to reg but now reg gets
BPF_ADD_CONSTfrom the copy.This means if another link to reg is created like:
another_reg= reg ;another_regshould get the id of reg but
assign_scalar_id_before_mov()seesBPF_ADD_CONSTon reg and assigns a new id to it.As reg has a new id now, known_reg's link to reg is broken. If we find new bounds for
known_reg, they will not be propagated to reg.This can be seen in the selftest added in the next commit:
0: (85) call bpf_get_prandom_u32#7 ; R0=scalar() 1: (57) r0 &= 255 ; R0=scalar(smin=smin32=0,smax=umax=smax32=umax32=255,var_off=(0x0; 0xff)) 2: (bf) r1 = r0 ; R0=scalar(id=1,smin=smin32=0,smax=umax=smax32=umax32=255,var_off=(0x0; 0xff)) R1=scalar(id=1,smin=smin32=0,smax=umax=smax32=umax32=255,var_off=(0x0; 0xff)) 3: (07) r1 += 4 ; R1=scalar(id=1+4,smin=umin=smin32=umin32=4,smax=umax=smax32=umax32=259,var_off=(0x0; 0x1ff)) 4: (a5) if r1 < 0xa goto pc+4 ; R1=scalar(id=1+4,smin=umin=smin32=umin32=10,smax=umax=smax32=umax32=259,var_off=(0x0; 0x1ff)) 5: (bf) r2 = r0 ; R0=scalar(id=2,smin=umin=smin32=umin32=6,smax=umax=smax32=umax32=255) R2=scalar(id=2,smin=umin=smin32=umin32=6,smax=umax=smax32=umax32=255) 6: (a5) if r1 < 0xe goto pc+2 ; R1=scalar(id=1+4,smin=umin=smin32=umin32=14,smax=umax=smax32=umax32=259,var_off=(0x0; 0x1ff)) 7: (35) if r0 >= 0xa goto pc+1 ; R0=scalar(id=2,smin=umin=smin32=umin32=6,smax=umax=smax32=umax32=9,var_off=(0x0; 0xf))8: (37) r0 /= 0
div by zeroWhen 4 is verified, r1's bounds are propagated to r0 but r0 also gets
BPF_ADD_CONST(bug).
When 5 is verified, r0 gets a new id (2) and its link with r1 is broken.After 6 we know r1 has bounds [14, 259] and therefore r0 should have bounds [10, 255], therefore the branch at 7 is always taken. But because r0's id was changed to 2, r1's new bounds are not propagated to r0. The verifier still thinks r0 has bounds [6, 255] before 7 and execution can reach div by zero.
Fix this by preserving id in
sync_linked_regs()like off andsubreg_def.Fixes 98d7ca374ba4 ("bpf: Track delta between "linked" registers.")Signed-off-by Puranjay Mohan <[email protected]>--- kernel/bpf/verifier.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)kernel/bpf/verifier.c +1
-0@@ -16871,6 +16871,7 @@ static void sync_linked_regs(struct bpf_verifier_state *vstate, struct bpf_reg_s }else {s32 saved_subreg_def = reg->subreg_def;s32 saved_off = reg->off; +u32 saved_id = reg->id; fake_reg.type =SCALAR_VALUE ;__mark_reg_known (&fake_reg, (s32 )reg->off - (s32 )known_reg->off);
kernel/bpf/verifier.c static void sync_linked_regs(struct bpf_verifier_state *vstate, struct bpf_reg_s · line 16886
Quoted 9 lines
kernel/bpf/verifier.c +2
-1@@ -16878,10 +16879,11 @@ static void sync_linked_regs(struct bpf_verifier_state *vstate, struct bpf_reg_s/* reg = known_reg; reg += delta */ copy_register_state (reg, known_reg);/* -* Must preserve off, id and add_const flag, +* Must preserve off, id and subreg_def flag, * otherwise another sync_linked_regs() will be incorrect. */ reg->off = saved_off; + reg->id = saved_id;
What is the veristat difference for meta/scx ?
I don't trust CI at the moment:
026-01-16T01:26:34.7163235Z Failed to open 'c-scheds_lib_sdt_task.bpf.o': -2
2026-01-16T01:26:34.7240161Z Failed to open 'c-scheds_scheds_c_scx_sdt.bpf.o': -2 2026-01-16T01:26:34.7298286Z Failed to open 'rust-scheds_scx_arena-9355999175dda454_out_arena.bpf.o': -2 2026-01-16T01:26:34.7353438Z Failed to open 'rust-scheds_scx_arena-9355999175dda454_out_atq.bpf.o': -2 2026-01-16T01:26:34.7414081Z Failed to open
'rust-scheds_scx_arena-9355999175dda454_out_bitmap.bpf.o': -2
...
2026-01-16T01:27:52.3087135Z # No changes in verification performance
something fishy.
Re: [PATCH bpf-next 1/2] bpf: Preserve id of register in sync_linked_regs()
Quoted 92 lines · Alexei Starovoitov
Quoted 68 lines · Puranjay Mohan
sync_linked_regs()copies the id ofknown_regto reg when propagating bounds ofknown_regto reg using the off ofknown_reg, but whenknown_regwas linked to reg like:
known_reg= reg ; bothknown_regand reg get same idknown_reg+= 4 ;known_reggets off = 4, and its id getsBPF_ADD_CONSTnow when a call to
sync_linked_regs()happens, let's say with the following:if
known_reg>= 10 goto pc+2known_reg's new bounds are propagated to reg but now reg gets
BPF_ADD_CONSTfrom the copy.This means if another link to reg is created like:
another_reg= reg ;another_regshould get the id of reg but
assign_scalar_id_before_mov()seesBPF_ADD_CONSTon reg and assigns a new id to it.As reg has a new id now, known_reg's link to reg is broken. If we find new bounds for
known_reg, they will not be propagated to reg.This can be seen in the selftest added in the next commit:
0: (85) call bpf_get_prandom_u32#7 ; R0=scalar() 1: (57) r0 &= 255 ; R0=scalar(smin=smin32=0,smax=umax=smax32=umax32=255,var_off=(0x0; 0xff)) 2: (bf) r1 = r0 ; R0=scalar(id=1,smin=smin32=0,smax=umax=smax32=umax32=255,var_off=(0x0; 0xff)) R1=scalar(id=1,smin=smin32=0,smax=umax=smax32=umax32=255,var_off=(0x0; 0xff)) 3: (07) r1 += 4 ; R1=scalar(id=1+4,smin=umin=smin32=umin32=4,smax=umax=smax32=umax32=259,var_off=(0x0; 0x1ff)) 4: (a5) if r1 < 0xa goto pc+4 ; R1=scalar(id=1+4,smin=umin=smin32=umin32=10,smax=umax=smax32=umax32=259,var_off=(0x0; 0x1ff)) 5: (bf) r2 = r0 ; R0=scalar(id=2,smin=umin=smin32=umin32=6,smax=umax=smax32=umax32=255) R2=scalar(id=2,smin=umin=smin32=umin32=6,smax=umax=smax32=umax32=255) 6: (a5) if r1 < 0xe goto pc+2 ; R1=scalar(id=1+4,smin=umin=smin32=umin32=14,smax=umax=smax32=umax32=259,var_off=(0x0; 0x1ff)) 7: (35) if r0 >= 0xa goto pc+1 ; R0=scalar(id=2,smin=umin=smin32=umin32=6,smax=umax=smax32=umax32=9,var_off=(0x0; 0xf))8: (37) r0 /= 0
div by zeroWhen 4 is verified, r1's bounds are propagated to r0 but r0 also gets
BPF_ADD_CONST(bug).
When 5 is verified, r0 gets a new id (2) and its link with r1 is broken.After 6 we know r1 has bounds [14, 259] and therefore r0 should have bounds [10, 255], therefore the branch at 7 is always taken. But because r0's id was changed to 2, r1's new bounds are not propagated to r0. The verifier still thinks r0 has bounds [6, 255] before 7 and execution can reach div by zero.
Fix this by preserving id in
sync_linked_regs()like off andsubreg_def.Fixes 98d7ca374ba4 ("bpf: Track delta between "linked" registers.")Signed-off-by Puranjay Mohan <[email protected]>--- kernel/bpf/verifier.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)kernel/bpf/verifier.c +1
-0@@ -16871,6 +16871,7 @@ static void sync_linked_regs(struct bpf_verifier_state *vstate, struct bpf_reg_s }else {s32 saved_subreg_def = reg->subreg_def;s32 saved_off = reg->off; +u32 saved_id = reg->id; fake_reg.type =SCALAR_VALUE ;__mark_reg_known (&fake_reg, (s32 )reg->off - (s32 )known_reg->off);kernel/bpf/verifier.c static void sync_linked_regs(struct bpf_verifier_state *vstate, struct bpf_reg_s · line 16886
Quoted 9 lines
kernel/bpf/verifier.c +2
-1@@ -16878,10 +16879,11 @@ static void sync_linked_regs(struct bpf_verifier_state *vstate, struct bpf_reg_s/* reg = known_reg; reg += delta */ copy_register_state (reg, known_reg);/* -* Must preserve off, id and add_const flag, +* Must preserve off, id and subreg_def flag, * otherwise another sync_linked_regs() will be incorrect. */ reg->off = saved_off; + reg->id = saved_id;What is the veristat difference for meta/scx ?
I don't trust CI at the moment:
026-01-16T01:26:34.7163235Z Failed to open 'c-scheds_lib_sdt_task.bpf.o': -2
2026-01-16T01:26:34.7240161Z Failed to open 'c-scheds_scheds_c_scx_sdt.bpf.o': -2 2026-01-16T01:26:34.7298286Z Failed to open 'rust-scheds_scx_arena-9355999175dda454_out_arena.bpf.o': -2 2026-01-16T01:26:34.7353438Z Failed to open 'rust-scheds_scx_arena-9355999175dda454_out_atq.bpf.o': -2 2026-01-16T01:26:34.7414081Z Failed to open
Quoted 6 lines
'rust-scheds_scx_arena-9355999175dda454_out_bitmap.bpf.o': -2
...2026-01-16T01:27:52.3087135Z # No changes in verification performancesomething fishy.
I don't see any change in veristat output:
../../../veristat/src/veristat -C -e file,prog,states,insns -f "insns_pct>0" fb_before.csv fb_after.csv
File Program States (A) States (B) States (DIFF) Insns (A) Insns
(B) Insns (DIFF)
---- ------- ---------- ---------- ------------- --------- --------- ------------
../../../veristat/src/veristat -C -e file,prog,states,insns -f "insns_pct>0" scx_before.csv scx_after.csv
File Program States (A) States (B) States (DIFF) Insns (A) Insns
(B) Insns (DIFF)
---- ------- ---------- ---------- ------------- --------- --------- ------------
PATCH 2/2 selftests: bpf: Add test for multiple syncs from linked register
In reply to Puranjay MohanBefore the last commit, sync_linked_regs() corrupted the register whose bounds are being updated by copying known_reg's id to it. The ids are the same in value but known_reg has the BPF_ADD_CONST flag which is wrongly copied to reg.
This later causes issues when creating new links to this reg. assign_scalar_id_before_mov() sees this BPF_ADD_CONST and gives a new id to this register and breaks the old links. This is exposed by the added selftest.
--- .../bpf/progs/verifier_linked_scalars.c | 33 +++++++++++++++++++ 1 file changed, 33 insertions(+)
tools/testing/selftests/bpf/progs/verifier_linked_scalars.c +33 -0
@@ -31,4 +31,37 @@ l1: \ " ::: __clobber_all); } +/* +* Test that sync_linked_regs() preserves register IDs. +* +* The sync_linked_regs() function copies bounds from known_reg to linked +* registers. When doing so, it must preserve each register's original id +* to allow subsequent syncs from the same source to work correctly. +* +*/ +SEC("socket") +__success +__naked void sync_linked_regs_preserves_id (void ) +{ + asmvolatile (" \ +call %[bpf_get_prandom_u32]; \ +r0 &= 0xff; /* r0 in [0, 255] */ \ +r1 = r0; /* r0, r1 linked with id 1 */ \ +r1 += 4; /* r1 has id=1 and off=4 in [4, 259] */ \ +if r1 < 10 goto l0_%=; \ +/* r1 in [10, 259], r0 synced to [6, 255] */ \ +r2 = r0; /* r2 has id=1 and in [6, 255] */ \ +if r1 < 14 goto l0_%=; \ +/* r1 in [14, 259], r0 synced to [10, 255] */ \ +if r0 >= 10 goto l0_%=; \ +/* Never executed */ \ +r0 /= 0; \ +l0_%=: \ +r0 = 0; \ +exit; \ +" : + : __imm(bpf_get_prandom_u32) + : __clobber_all); +} + char _license[] SEC("license") = "GPL";
Signature · 1 line
2.47.3
Re: [PATCH bpf-next 2/2] selftests: bpf: Add test for multiple syncs from linked register
Quoted 5 lines · Puranjay Mohan
Before the last commit,
sync_linked_regs()corrupted the register whose bounds are being updated by copying known_reg's id to it. The ids are the same in value butknown_reghas theBPF_ADD_CONSTflag which is wrongly copied to reg.
Quoted 7 lines
This later causes issues when creating new links to this reg.
assign_scalar_id_before_mov()sees thisBPF_ADD_CONSTand gives a new id to this register and breaks the old links. This is exposed by the added selftest.Signed-off-by Puranjay Mohan <[email protected]>---
Confirm the test behavior on master vs master+patch #1.
[...]
This series was applied to bpf/bpf-next.git (master) by Alexei Starovoitov <[email protected]>:
Hello:
This series was applied to bpf/bpf-next.git (master) by Alexei Starovoitov <[email protected]>:
Quoted 9 lines · On Thu, 15 Jan 2026 07:11:39 -0800 you wrote:
This patch fixes the linked register tracking when multiple links from the same register are created with a sync between the creation of these links. The sync corrupts the id of the register and therefore the second link is not created properly. See the patch description to understand more.
The fix is to preserve the id while doing the sync similar to the off.
[...]
Here is the summary with links:
- [bpf-next,1/2] bpf: Preserve id of register in sync_linked_regs()
https://git.kernel.org/bpf/bpf-next/c/af9e89d8dd39 - [bpf-next,2/2] selftests: bpf: Add test for multiple syncs from linked register https://git.kernel.org/bpf/bpf-next/c/086c99fbe450
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html