BPF verifier retval logic fixes
Subsystems (4)
Andrii Nakryiko
maintainercover BPF verifier retval logic fixes
This patch set fixes BPF verifier logic around validating and enforcing return values for BPF programs that have specific range of expected return values. Both sync and async callbacks have similar logic and are fixes as well. A few tests are added that would fail without the fixes in this patch set.
Also, while at it, we extend retval checking logic to use not just tnum value, but also umin/umax range, avoiding future potential issues if expected range cannot be represented precisely by tnum (e.g., [0, 2] is not representable by tnum and is treated as [0, 3]).
There is a little bit of refactoring to unify async callback and program exit logic to avoid duplication of checks as much as possible.
Andrii Nakryiko (10):
bpf: rearrange bpf_func_state fields to save a bit of memory
bpf: provide correct register name for exception callback retval check
bpf: enforce precision of R0 on callback return
bpf: enforce exact retval range on subprog/callback exit selftests/bpf: add selftest validating callback result is enforced
bpf: enforce precise retval range on program exit
bpf: unify async callback and program retval checks
bpf: enforce precision of R0 on program/async callback return selftests/bpf: validate async callback return value check correctness selftests/bpf: adjust global_func15 test to validate prog exit
precision
include/linux/bpf_verifier.h | 9 +- kernel/bpf/verifier.c | 133 +++++++++++------- .../selftests/bpf/progs/exceptions_assert.c | 2 +- .../selftests/bpf/progs/exceptions_fail.c | 2 +- .../selftests/bpf/progs/test_global_func15.c | 31 +++- .../selftests/bpf/progs/timer_failure.c | 31 ++-- .../selftests/bpf/progs/user_ringbuf_fail.c | 2 +- .../bpf/progs/verifier_cgroup_inv_retcode.c | 8 +- .../bpf/progs/verifier_netfilter_retcode.c | 2 +- .../bpf/progs/verifier_subprog_precision.c | 45 ++++++ 10 files changed, 198 insertions(+), 67 deletions(-)
Signature · 1 line
2.34.1
Andrii Nakryiko
maintainerPATCH 1/10 bpf: rearrange bpf_func_state fields to save a bit of memory
It's a trivial rearrangement saving 8 bytes. We have 4 bytes of padding at the end which can be filled with another field without increasing struct bpf_func_state.
copy_func_state() logic remains correct without any further changes.
BEFORE
======
struct bpf_func_state {
struct bpf_reg_state regs[11]; /* 0 1320 */
/* --- cacheline 20 boundary (1280 bytes) was 40 bytes ago --- */
int callsite; /* 1320 4 */
u32 frameno; /* 1324 4 */
u32 subprogno; /* 1328 4 */
u32 async_entry_cnt; /* 1332 4 */
bool in_callback_fn; /* 1336 1 */
/* XXX 7 bytes hole, try to pack */
/* --- cacheline 21 boundary (1344 bytes) --- */
struct tnum callback_ret_range; /* 1344 16 */
bool in_async_callback_fn; /* 1360 1 */
bool in_exception_callback_fn; /* 1361 1 */
/* XXX 2 bytes hole, try to pack */
int acquired_refs; /* 1364 4 */
struct bpf_reference_state * refs; /* 1368 8 */
int allocated_stack; /* 1376 4 */
/* XXX 4 bytes hole, try to pack */
struct bpf_stack_state * stack; /* 1384 8 */
/* size: 1392, cachelines: 22, members: 13 */ /* sum members: 1379, holes: 3, sum holes: 13 */ /* last cacheline: 48 bytes */
};
AFTER
=====
struct bpf_func_state {
struct bpf_reg_state regs[11]; /* 0 1320 */
/* --- cacheline 20 boundary (1280 bytes) was 40 bytes ago --- */
int callsite; /* 1320 4 */
u32 frameno; /* 1324 4 */
u32 subprogno; /* 1328 4 */
u32 async_entry_cnt; /* 1332 4 */
struct tnum callback_ret_range; /* 1336 16 */
/* --- cacheline 21 boundary (1344 bytes) was 8 bytes ago --- */
bool in_callback_fn; /* 1352 1 */
bool in_async_callback_fn; /* 1353 1 */
bool in_exception_callback_fn; /* 1354 1 */
/* XXX 1 byte hole, try to pack */
int acquired_refs; /* 1356 4 */
struct bpf_reference_state * refs; /* 1360 8 */
struct bpf_stack_state * stack; /* 1368 8 */
int allocated_stack; /* 1376 4 */
/* size: 1384, cachelines: 22, members: 13 */ /* sum members: 1379, holes: 1, sum holes: 1 */ /* padding: 4 */
/* last cacheline: 40 bytes */
};
--- include/linux/bpf_verifier.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
include/linux/bpf_verifier.h +2 -2
@@ -297,16 +297,16 @@ struct bpf_func_state {* void foo(void) { bpf_timer_set_callback(,foo); } */ u32 async_entry_cnt; -bool in_callback_fn;struct tnum callback_ret_range; +bool in_callback_fn;bool in_async_callback_fn;bool in_exception_callback_fn;/* The following fields should be last. See copy_func_state() */ int acquired_refs;struct bpf_reference_state *refs; -int allocated_stack;struct bpf_stack_state *stack; +int allocated_stack; };struct bpf_idx_pair {
Signature · 1 line
2.34.1
Re: [PATCH bpf-next 01/10] bpf: rearrange bpf_func_state fields to save a bit of memory
Quoted 4 lines · Andrii Nakryiko
It's a trivial rearrangement saving 8 bytes. We have 4 bytes of padding at the end which can be filled with another field without increasing struct
bpf_func_state.
[...]
Quoted 3 lines
Signed-off-by Andrii Nakryiko <[email protected]>---
[...]
Andrii Nakryiko
maintainerPATCH 2/10 bpf: provide correct register name for exception callback retval check
In reply to Andrii Nakryikobpf_throw() is checking R1, so let's report R1 in the log.
--- kernel/bpf/verifier.c | 12 ++++++------ .../testing/selftests/bpf/progs/exceptions_assert.c | 2 +- tools/testing/selftests/bpf/progs/exceptions_fail.c | 2 +- 3 files changed, 8 insertions(+), 8 deletions(-)
kernel/bpf/verifier.c +6 -6
@@ -11663,7 +11663,7 @@ static int fetch_kfunc_meta(struct bpf_verifier_env *env,return 0 ; } -static int check_return_code (struct bpf_verifier_env *env,int regno); +static int check_return_code (struct bpf_verifier_env *env,int regno,const char *reg_name);static int check_kfunc_call (struct bpf_verifier_env *env,struct bpf_insn *insn,int *insn_idx_p)
@@ -11799,7 +11799,7 @@ static int check_kfunc_call(struct bpf_verifier_env *env, struct bpf_insn *insn,* to bpf_throw becomes the return value of the program. */ if (!env->exception_callback_subprog) { - err =check_return_code (env,BPF_REG_1 ); + err =check_return_code (env,BPF_REG_1 ,"R1" );if (err <0 )return err; }
@@ -14818,7 +14818,7 @@ static int check_ld_abs(struct bpf_verifier_env *env, struct bpf_insn *insn)return 0 ; } -static int check_return_code (struct bpf_verifier_env *env,int regno) +static int check_return_code (struct bpf_verifier_env *env,int regno,const char *reg_name) {struct tnum enforce_attach_type_range = tnum_unknown;const struct bpf_prog *prog = env->prog;
@@ -14872,7 +14872,7 @@ static int check_return_code(struct bpf_verifier_env *env, int regno) }if (!tnum_in (const_0, reg->var_off)) { -verbose_invalid_scalar (env, reg, &const_0,"async callback" ,"R0" ); +verbose_invalid_scalar (env, reg, &const_0,"async callback" , reg_name);return -EINVAL ; }return 0 ;
@@ -14972,7 +14972,7 @@ static int check_return_code(struct bpf_verifier_env *env, int regno) }if (!tnum_in (range, reg->var_off)) { -verbose_invalid_scalar (env, reg, &range,"program exit" ,"R0" ); +verbose_invalid_scalar (env, reg, &range,"program exit" , reg_name);if (prog->expected_attach_type ==BPF_LSM_CGROUP && prog_type ==BPF_PROG_TYPE_LSM && !prog->aux->attach_func_proto->type)
@@ -17220,7 +17220,7 @@ static int do_check(struct bpf_verifier_env *env)continue ; } - err =check_return_code (env,BPF_REG_0 ); + err =check_return_code (env,BPF_REG_0 ,"R0" );if (err)return err; process_bpf_exit:
tools/testing/selftests/bpf/progs/exceptions_assert.c +1 -1
@@ -125,7 +125,7 @@ int check_assert_generic(struct __sk_buff *ctx) } SEC("?fentry/bpf_check") -__failure __msg("At program exit the register R0 has value (0x40; 0x0)") +__failure __msg("At program exit the register R1 has value (0x40; 0x0)") int check_assert_with_return(void *ctx) {bpf_assert_with (!ctx,64 );
tools/testing/selftests/bpf/progs/exceptions_fail.c +1 -1
@@ -306,7 +306,7 @@ int reject_set_exception_cb_bad_ret1(void *ctx) } SEC("?fentry/bpf_check") -__failure __msg("At program exit the register R0 has value (0x40; 0x0) should") +__failure __msg("At program exit the register R1 has value (0x40; 0x0) should") int reject_set_exception_cb_bad_ret2(void *ctx) {bpf_throw (64 );
Signature · 1 line
2.34.1
Re: [PATCH bpf-next 02/10] bpf: provide correct register name for exception callback retval check
Quoted 4 lines · Andrii Nakryiko
bpf_throw()is checking R1, so let's report R1 in the log.Signed-off-by Andrii Nakryiko <[email protected]>---
[...]
Re: [PATCH bpf-next 02/10] bpf: provide correct register name for exception callback retval check
In reply to Andrii Nakryikotools/testing/selftests/bpf/progs/exceptions_assert.c int check_assert_generic(struct __sk_buff *ctx) · line 128
Quoted 3 lines · Andrii Nakryiko
SEC("?fentry/bpf_check")tools/testing/selftests/bpf/progs/exceptions_assert.c +1
-1-__failure __msg("At program exit the register R0 has value (0x40; 0x0)") +__failure __msg("At program exit the register R1 has value (0x40; 0x0)")
While at it...
should we change tnum_strn() to print it as normal constant when mask==0 ?
Re: [PATCH bpf-next 02/10] bpf: provide correct register name for exception callback retval check
Quoted 8 lines · Alexei Starovoitov
tools/testing/selftests/bpf/progs/exceptions_assert.c int check_assert_generic(struct __sk_buff *ctx) · line 128
Quoted 3 lines · Andrii Nakryiko
SEC("?fentry/bpf_check")tools/testing/selftests/bpf/progs/exceptions_assert.c +1
-1-__failure __msg("At program exit the register R0 has value (0x40; 0x0)") +__failure __msg("At program exit the register R1 has value (0x40; 0x0)")While at it...
should we changetnum_strn()to print it as normal constant when mask==0 ?
sure, I can add that as another patch, because I suspect even more tests would have to be adjusted
Andrii Nakryiko
maintainerPATCH 3/10 bpf: enforce precision of R0 on callback return
In reply to Andrii NakryikoGiven verifier checks actual value, r0 has to be precise, so we need to propagate precision properly. r0 also has to be marked as read, otherwise subsequent state comparisons will ignore such register as unimportant and precision won't really help here.
--- kernel/bpf/verifier.c | 7 +++++++ 1 file changed, 7 insertions(+)
kernel/bpf/verifier.c +7 -0
@@ -9493,6 +9493,13 @@ static int prepare_func_exit(struct bpf_verifier_env *env, int *insn_idx)verbose (env,"R0 not a scalar value\n" );return -EACCES ; } + +/* we are going to rely on register's precise value */ + err =mark_reg_read (env, r0, r0->parent,REG_LIVE_READ64 ); + err = err ?:mark_chain_precision (env,BPF_REG_0 ); +if (err) +return err; +if (!tnum_in (range, r0->var_off)) {verbose_invalid_scalar (env, r0, &range,"callback return" ,"R0" );return -EINVAL ;
Signature · 1 line
2.34.1
Re: [PATCH bpf-next 03/10] bpf: enforce precision of R0 on callback return
Quoted 8 lines · Andrii Nakryiko
Given verifier checks actual value, r0 has to be precise, so we need to propagate precision properly. r0 also has to be marked as read, otherwise subsequent state comparisons will ignore such register as unimportant and precision won't really help here.
Fixes 69c087ba6225 ("bpf: Add bpf_for_each_map_elem() helper")Signed-off-by Andrii Nakryiko <[email protected]>---
[...]
Andrii Nakryiko
maintainerPATCH 4/10 bpf: enforce exact retval range on subprog/callback exit
In reply to Andrii NakryikoInstead of relying on potentially imprecise tnum representation of expected return value range for callbacks and subprogs, validate that both tnum and umin/umax range satisfy exact expected range of return values.
E.g., if callback would need to return [0, 2] range, tnum can't represent this precisely and instead will allow [0, 3] range. By additionally checking umin/umax range, we can make sure that subprog/callback indeed returns only valid [0, 2] range.
--- include/linux/bpf_verifier.h | 7 +++++- kernel/bpf/verifier.c | 45 +++++++++++++++++++++++++++--------- 2 files changed, 40 insertions(+), 12 deletions(-)
include/linux/bpf_verifier.h +6 -1
@@ -275,6 +275,11 @@ struct bpf_reference_state {int callback_ref; }; +struct bpf_retval_range { +u32 minval; +u32 maxval; +}; +/* state of the program: * type of all registers and stack info */
@@ -297,7 +302,7 @@ struct bpf_func_state {* void foo(void) { bpf_timer_set_callback(,foo); } */ u32 async_entry_cnt; -struct tnum callback_ret_range; +struct bpf_retval_range callback_ret_range;bool in_callback_fn;bool in_async_callback_fn;bool in_exception_callback_fn;
kernel/bpf/verifier.c +34 -11
111 lines not on this page · raw message
Signature · 1 line
2.34.1
Re: [PATCH bpf-next 04/10] bpf: enforce exact retval range on subprog/callback exit
Quoted 5 lines · Andrii Nakryiko
Instead of relying on potentially imprecise tnum representation of expected return value range for callbacks and subprogs, validate that both tnum and umin/umax range satisfy exact expected range of return values.
Quoted 7 lines
E.g., if callback would need to return [0, 2] range, tnum can't represent this precisely and instead will allow [0, 3] range. By additionally checking umin/umax range, we can make sure that subprog/callback indeed returns only valid [0, 2] range.
Signed-off-by Andrii Nakryiko <[email protected]>---
(but please see a question below)
[...]
kernel/bpf/verifier.c static bool in_rbtree_lock_required_cb(struct bpf_verifier_env *env) · line 9485
Quoted 10 lines
kernel/bpf/verifier.c +6
-0@@ -9464,6 +9477,16 @@ static bool in_rbtree_lock_required_cb(struct bpf_verifier_env *env)return is_rbtree_lock_required_kfunc (kfunc_btf_id); } +static bool retval_range_within (struct bpf_retval_range range,const struct bpf_reg_state *reg) +{ +struct tnum trange =retval_range_as_tnum (range); + +if (!tnum_in (trange, reg->var_off)) +return false;
Q: When is it necessary to do this check?
I tried commenting it and test_{verifier,progs} still pass. Are there situations when umin/umax change is not sufficient?
Quoted 4 lines
kernel/bpf/verifier.c +4
-0+ +return range.minval <= reg->umin_value && reg->umax_value <= range.maxval; +} +
[...]
Re: [PATCH bpf-next 04/10] bpf: enforce exact retval range on subprog/callback exit
Quoted 24 lines · Eduard Zingerman
Quoted 5 lines · Andrii Nakryiko
Instead of relying on potentially imprecise tnum representation of expected return value range for callbacks and subprogs, validate that both tnum and umin/umax range satisfy exact expected range of return values.
Quoted 7 lines
E.g., if callback would need to return [0, 2] range, tnum can't represent this precisely and instead will allow [0, 3] range. By additionally checking umin/umax range, we can make sure that subprog/callback indeed returns only valid [0, 2] range.
Signed-off-by Andrii Nakryiko <[email protected]>---
Acked-by Eduard Zingerman <[email protected]>(but please see a question below)
[...]
Quoted 4 lines
kernel/bpf/verifier.c +0
-0@@ -9464,6 +9477,16 @@ static bool in_rbtree_lock_required_cb(struct bpf_verifier_env *env)return is_rbtree_lock_required_kfunc (kfunc_btf_id); }
Quoted 10 lines
kernel/bpf/verifier.c static bool in_rbtree_lock_required_cb(struct bpf_verifier_env *env) · line 9485
Quoted 6 lines
kernel/bpf/verifier.c +6
-0+static bool retval_range_within (struct bpf_retval_range range,const struct bpf_reg_state *reg) +{ +struct tnum trange =retval_range_as_tnum (range); + +if (!tnum_in (trange, reg->var_off)) +return false;Q: When is it necessary to do this check?
I tried commenting it and test_{verifier,progs} still pass. Are there situations when umin/umax change is not sufficient?
I believe not. But we still check tnum in check_cond_jmp_op, for example, so I decided to keep it to not have to argue and prove why it's ok to ditch tnum.
Generally speaking, I think tnum is useful in only one use case:
checking (un)aligned memory accesses. This is the only representation that can make sure we have lower 2-3 bits as zero to prove that memory access is 4- or 8-byte aligned.
Other than this, I think ranges are more precise and easier to work with.
But I'm not ready to go on the quest to eliminate tnum usage everywhere :)
Quoted 9 lines
Quoted 5 lines
kernel/bpf/verifier.c +4
-0+ +return range.minval <= reg->umin_value && reg->umax_value <= range.maxval; +} +[...]
Re: [PATCH bpf-next 04/10] bpf: enforce exact retval range on subprog/callback exit
Quoted 32 lines · Andrii Nakryiko
Quoted 20 lines · Eduard Zingerman
Quoted 5 lines · Andrii Nakryiko
Instead of relying on potentially imprecise tnum representation of expected return value range for callbacks and subprogs, validate that both tnum and umin/umax range satisfy exact expected range of return values.
Quoted 7 lines
E.g., if callback would need to return [0, 2] range, tnum can't represent this precisely and instead will allow [0, 3] range. By additionally checking umin/umax range, we can make sure that subprog/callback indeed returns only valid [0, 2] range.
Signed-off-by Andrii Nakryiko <[email protected]>---
Acked-by Eduard Zingerman <[email protected]>(but please see a question below)
[...]
Quoted 10 lines
kernel/bpf/verifier.c static bool in_rbtree_lock_required_cb(struct bpf_verifier_env *env) · line 9485
Quoted 6 lines
kernel/bpf/verifier.c +6
-0+static bool retval_range_within (struct bpf_retval_range range,const struct bpf_reg_state *reg) +{ +struct tnum trange =retval_range_as_tnum (range); + +if (!tnum_in (trange, reg->var_off)) +return false;Q: When is it necessary to do this check?
I tried commenting it and test_{verifier,progs} still pass. Are there situations when umin/umax change is not sufficient?
Quoted 3 lines
I believe not. But we still check tnum in
check_cond_jmp_op, for example, so I decided to keep it to not have to argue and prove why it's ok to ditch tnum.
Semi-related proof[1] from awhile back :)
Quoted 6 lines
Generally speaking, I think tnum is useful in only one use case:
checking (un)aligned memory accesses. This is the only representation that can make sure we have lower 2-3 bits as zero to prove that memory access is 4- or 8-byte aligned.Other than this, I think ranges are more precise and easier to work with.
Agree with the above.
I'd vote for ditching tnum for the retval check here. With umin/umax check in place there really isn't a need for an additional tnum check at all. Keeping it probably does more harm (in the form of confusion) than good.
1: https://lore.kernel.org/bpf/[email protected]/
Quoted 8 lines
But I'm not ready to go on the quest to eliminate tnum usage everywhere :)
Quoted 6 lines
Quoted 4 lines
kernel/bpf/verifier.c +4
-0+ +return range.minval <= reg->umin_value && reg->umax_value <= range.maxval; +} +[...]
Re: [PATCH bpf-next 04/10] bpf: enforce exact retval range on subprog/callback exit
Quoted 45 lines · Shung-Hsi Yu
Quoted 32 lines · Andrii Nakryiko
Quoted 20 lines · Eduard Zingerman
Quoted 5 lines · Andrii Nakryiko
Instead of relying on potentially imprecise tnum representation of expected return value range for callbacks and subprogs, validate that both tnum and umin/umax range satisfy exact expected range of return values.
Quoted 7 lines
E.g., if callback would need to return [0, 2] range, tnum can't represent this precisely and instead will allow [0, 3] range. By additionally checking umin/umax range, we can make sure that subprog/callback indeed returns only valid [0, 2] range.
Signed-off-by Andrii Nakryiko <[email protected]>---
Acked-by Eduard Zingerman <[email protected]>(but please see a question below)
[...]
Quoted 10 lines
kernel/bpf/verifier.c static bool in_rbtree_lock_required_cb(struct bpf_verifier_env *env) · line 9485
Quoted 6 lines
kernel/bpf/verifier.c +6
-0+static bool retval_range_within (struct bpf_retval_range range,const struct bpf_reg_state *reg) +{ +struct tnum trange =retval_range_as_tnum (range); + +if (!tnum_in (trange, reg->var_off)) +return false;Q: When is it necessary to do this check?
I tried commenting it and test_{verifier,progs} still pass. Are there situations when umin/umax change is not sufficient?Quoted 3 lines
I believe not. But we still check tnum in
check_cond_jmp_op, for example, so I decided to keep it to not have to argue and prove why it's ok to ditch tnum.Semi-related proof[1] from awhile back :)
Quoted 5 lines
Generally speaking, I think tnum is useful in only one use case:
checking (un)aligned memory accesses. This is the only representation that can make sure we have lower 2-3 bits as zero to prove that memory access is 4- or 8-byte aligned.
Quoted 8 lines
Quoted 1 line
Other than this, I think ranges are more precise and easier to work with.
Agree with the above.
I'd vote for ditching tnum for the retval check here. With umin/umax check in place there really isn't a need for an additional tnum check at all. Keeping it probably does more harm (in the form of confusion) than good.
Ok, I think it's as trivial as removing two lines of code. So I'll add it as the last patch to the series and will let BPF maintainers decide if they want this change or not.
Quoted 11 lines
1: https://lore.kernel.org/bpf/[email protected]/
Quoted 8 lines
But I'm not ready to go on the quest to eliminate tnum usage everywhere :)
Quoted 6 lines
Quoted 4 lines
kernel/bpf/verifier.c +4
-0+ +return range.minval <= reg->umin_value && reg->umax_value <= range.maxval; +} +[...]
Andrii Nakryiko
maintainerPATCH 5/10 selftests/bpf: add selftest validating callback result is enforced
In reply to Andrii NakryikoBPF verifier expects callback subprogs to return values from specified range (typically [0, 1]). This requires that r0 at exit is both precise (because we rely on specific value range) and is marked as read (otherwise state comparison will ignore such register as unimportant).
Add a simple test that validates that all these conditions are enforced.
--- .../bpf/progs/verifier_subprog_precision.c | 45 +++++++++++++++++++ 1 file changed, 45 insertions(+)
tools/testing/selftests/bpf/progs/verifier_subprog_precision.c +45 -0
@@ -117,6 +117,51 @@ __naked int global_subprog_result_precise(void) ); } +__naked __noinline __used +static unsigned long loop_callback_bad() +{ +/* bpf_loop() callback that can return values outside of [0, 1] range */ + asmvolatile ( +"call %[bpf_get_prandom_u32];" +"if r0 > 1000 goto 1f;" +"r0 = 0;" +"1:" +"goto +0;" /* checkpoint */ +/* bpf_loop() expects [0, 1] values, so branch above skipping +* r0 = 0; should lead to a failure, but if exit instruction +* doesn't enforce r0's precision, this callback will be +* successfully verified +*/ +"exit;" + : + : __imm(bpf_get_prandom_u32) + : __clobber_common + ); +} + +SEC("?raw_tp") +__failure __log_level(2) +__flag(BPF_F_TEST_STATE_FREQ) +__msg("from 10 to 12: frame1: R0=scalar(umin=1001) R10=fp0 cb") +__msg("At callback return the register R0 has unknown scalar value should have been in (0x0; 0x1)") +__naked int callback_precise_return_fail(void) +{ + asmvolatile ( +"r1 = 1;" /* nr_loops */ +"r2 = %[loop_callback_bad];" /* callback_fn */ +"r3 = 0;" /* callback_ctx */ +"r4 = 0;" /* flags */ +"call %[bpf_loop];" + +"r0 = 0;" +"exit;" + : + : __imm_ptr(loop_callback_bad), +__imm (bpf_loop) + : __clobber_common + ); +} + SEC("?raw_tp") __success __log_level(2)__msg ("14: (0f) r1 += r6" )
Signature · 1 line
2.34.1
Re: [PATCH bpf-next 05/10] selftests/bpf: add selftest validating callback result is enforced
Quoted 9 lines · Andrii Nakryiko
BPF verifier expects callback subprogs to return values from specified range (typically [0, 1]). This requires that r0 at exit is both precise (because we rely on specific value range) and is marked as read (otherwise state comparison will ignore such register as unimportant).
Add a simple test that validates that all these conditions are enforced.
Signed-off-by Andrii Nakryiko <[email protected]>---
[...]
tools/testing/selftests/bpf/progs/verifier_subprog_precision.c __naked int global_subprog_result_precise(void) · line 144
Quoted 3 lines
tools/testing/selftests/bpf/progs/verifier_subprog_precision.c +3
-0+SEC("?raw_tp") +__failure __log_level(2) +__flag (BPF_F_TEST_STATE_FREQ )
Nit: although it is redundant, but maybe also check precision log to
check that r0 is indeed marked precise?
Quoted 3 lines
tools/testing/selftests/bpf/progs/verifier_subprog_precision.c +3
-0+__msg("from 10 to 12: frame1: R0=scalar(umin=1001) R10=fp0 cb") +__msg("At callback return the register R0 has unknown scalar value should have been in (0x0; 0x1)") +__naked int callback_precise_return_fail (void )
[...]
Re: [PATCH bpf-next 05/10] selftests/bpf: add selftest validating callback result is enforced
Quoted 14 lines · Eduard Zingerman
Quoted 9 lines · Andrii Nakryiko
BPF verifier expects callback subprogs to return values from specified range (typically [0, 1]). This requires that r0 at exit is both precise (because we rely on specific value range) and is marked as read (otherwise state comparison will ignore such register as unimportant).
Add a simple test that validates that all these conditions are enforced.
Signed-off-by Andrii Nakryiko <[email protected]>---
Acked-by Eduard Zingerman <[email protected]>
Quoted 9 lines
[...]
tools/testing/selftests/bpf/progs/verifier_subprog_precision.c __naked int global_subprog_result_precise(void) · line 144
Quoted 3 lines
tools/testing/selftests/bpf/progs/verifier_subprog_precision.c +3
-0+SEC("?raw_tp") +__failure __log_level(2) +__flag (BPF_F_TEST_STATE_FREQ )Nit: although it is redundant, but maybe also check precision log to
check that r0 is indeed marked precise?
sure, I'll add another expected msg, no problem
Quoted 7 lines
Quoted 3 lines
tools/testing/selftests/bpf/progs/verifier_subprog_precision.c +3
-0+__msg("from 10 to 12: frame1: R0=scalar(umin=1001) R10=fp0 cb") +__msg("At callback return the register R0 has unknown scalar value should have been in (0x0; 0x1)") +__naked int callback_precise_return_fail (void )[...]
Andrii Nakryiko
maintainerPATCH 6/10 bpf: enforce precise retval range on program exit
In reply to Andrii NakryikoSimilarly to subprog/callback logic, enforce return value of BPF program using more precise umin/umax range, in addition to tnum-based check.
We need to adjust a bunch of tests due to a changed format of an error message.
--- kernel/bpf/verifier.c | 55 +++++++++++-------- .../selftests/bpf/progs/exceptions_assert.c | 2 +- .../selftests/bpf/progs/exceptions_fail.c | 2 +- .../selftests/bpf/progs/test_global_func15.c | 2 +- .../selftests/bpf/progs/timer_failure.c | 2 +- .../selftests/bpf/progs/user_ringbuf_fail.c | 2 +- .../bpf/progs/verifier_cgroup_inv_retcode.c | 8 +-- .../bpf/progs/verifier_netfilter_retcode.c | 2 +- .../bpf/progs/verifier_subprog_precision.c | 2 +- 9 files changed, 43 insertions(+), 34 deletions(-)
kernel/bpf/verifier.c +32 -23
@@ -352,20 +352,29 @@ __printf(2, 3) static void verbose(void *private_data, const char *fmt, ...)static void verbose_invalid_scalar (struct bpf_verifier_env *env,struct bpf_reg_state *reg, -struct tnum *range,const char *ctx, +struct bpf_retval_range range,const char *ctx,const char *reg_name) {char tn_buf[48 ]; +bool unknown = true; -verbose (env,"At %s the register %s " , ctx, reg_name); +verbose (env,"At %s the register %s has" , ctx, reg_name); +if (reg->umin_value >0 ) { +verbose (env," umin=%llu" , reg->umin_value); + unknown = false; + } +if (reg->umax_value <U64_MAX ) { +verbose (env," umax=%llu" , reg->umax_value); + unknown = false; + }if (!tnum_is_unknown (reg->var_off)) {tnum_strn (tn_buf,sizeof (tn_buf), reg->var_off); -verbose (env,"has value %s" , tn_buf); - }else { -verbose (env,"has unknown scalar value" ); +verbose (env," var_off=%s" , tn_buf); + unknown = false; } -tnum_strn (tn_buf,sizeof (tn_buf), *range); -verbose (env," should have been in %s\n" , tn_buf); +if (unknown) +verbose (env," unknown scalar value" ); +verbose (env," should have been in [%u, %u]\n" , range.minval, range.maxval); }static bool type_may_be_null (u32 type)
@@ -9522,9 +9531,8 @@ static int prepare_func_exit(struct bpf_verifier_env *env, int *insn_idx)/* enforce R0 return value range */ if (!retval_range_within (callee->callback_ret_range, r0)) { -struct tnum range =retval_range_as_tnum (callee->callback_ret_range); - -verbose_invalid_scalar (env, r0, &range,"callback return" ,"R0" ); +verbose_invalid_scalar (env, r0, callee->callback_ret_range, +"callback return" ,"R0" );return -EINVAL ; } }else {
@@ -14853,7 +14861,8 @@ static int check_return_code(struct bpf_verifier_env *env, int regno, const charstruct tnum enforce_attach_type_range = tnum_unknown;const struct bpf_prog *prog = env->prog;struct bpf_reg_state *reg; -struct tnum range =tnum_range (0 ,1 ), const_0 =tnum_const (0 ); +struct bpf_retval_range range =retval_range (0 ,1 ); +struct bpf_retval_range const_0 =retval_range (0 ,0 );enum bpf_prog_type prog_type =resolve_prog_type (env->prog);int err;struct bpf_func_state *frame = env->cur_state->frame[0 ];
@@ -14901,8 +14910,8 @@ static int check_return_code(struct bpf_verifier_env *env, int regno, const charreturn -EINVAL ; } -if (!tnum_in (const_0, reg->var_off)) { -verbose_invalid_scalar (env, reg, &const_0,"async callback" , reg_name); +if (!retval_range_within (const_0, reg)) { +verbose_invalid_scalar (env, reg, const_0,"async callback" , reg_name);return -EINVAL ; }return 0 ;
@@ -14928,14 +14937,14 @@ static int check_return_code(struct bpf_verifier_env *env, int regno, const char env->prog->expected_attach_type == BPF_CGROUP_INET4_GETSOCKNAME || env->prog->expected_attach_type == BPF_CGROUP_INET6_GETSOCKNAME || env->prog->expected_attach_type == BPF_CGROUP_UNIX_GETSOCKNAME) - range =tnum_range (1 ,1 ); + range =retval_range (1 ,1 );if (env->prog->expected_attach_type ==BPF_CGROUP_INET4_BIND || env->prog->expected_attach_type ==BPF_CGROUP_INET6_BIND ) - range =tnum_range (0 ,3 ); + range =retval_range (0 ,3 );break ;case BPF_PROG_TYPE_CGROUP_SKB :if (env->prog->expected_attach_type ==BPF_CGROUP_INET_EGRESS ) { - range =tnum_range (0 ,3 ); + range =retval_range (0 ,3 ); enforce_attach_type_range =tnum_range (2 ,3 ); }break ;
@@ -14948,13 +14957,13 @@ static int check_return_code(struct bpf_verifier_env *env, int regno, const charcase BPF_PROG_TYPE_RAW_TRACEPOINT :if (!env->prog->aux->attach_btf_id)return 0 ; - range =tnum_const (0 ); + range =retval_range (0 ,0 );break ;case BPF_PROG_TYPE_TRACING :switch (env->prog->expected_attach_type) {case BPF_TRACE_FENTRY :case BPF_TRACE_FEXIT : - range =tnum_const (0 ); + range =retval_range (0 ,0 );break ;case BPF_TRACE_RAW_TP :case BPF_MODIFY_RETURN :
@@ -14966,7 +14975,7 @@ static int check_return_code(struct bpf_verifier_env *env, int regno, const char }break ;case BPF_PROG_TYPE_SK_LOOKUP : - range =tnum_range (SK_DROP ,SK_PASS ); + range =retval_range (SK_DROP ,SK_PASS );break ;case BPF_PROG_TYPE_LSM :
@@ -14980,12 +14989,12 @@ static int check_return_code(struct bpf_verifier_env *env, int regno, const char/* Make sure programs that attach to void * hooks don't try to modify return value. */ - range =tnum_range (1 ,1 ); + range =retval_range (1 ,1 ); }break ;case BPF_PROG_TYPE_NETFILTER : - range =tnum_range (NF_DROP ,NF_ACCEPT ); + range =retval_range (NF_DROP ,NF_ACCEPT );break ;case BPF_PROG_TYPE_EXT :/* freplace program can return anything as its return value
@@ -15001,8 +15010,8 @@ static int check_return_code(struct bpf_verifier_env *env, int regno, const charreturn -EINVAL ; } -if (!tnum_in (range, reg->var_off)) { -verbose_invalid_scalar (env, reg, &range,"program exit" , reg_name); +if (!retval_range_within (range, reg)) { +verbose_invalid_scalar (env, reg, range,"program exit" , reg_name);if (prog->expected_attach_type ==BPF_LSM_CGROUP && prog_type ==BPF_PROG_TYPE_LSM && !prog->aux->attach_func_proto->type)
tools/testing/selftests/bpf/progs/exceptions_assert.c +1 -1
@@ -125,7 +125,7 @@ int check_assert_generic(struct __sk_buff *ctx) } SEC("?fentry/bpf_check") -__failure __msg("At program exit the register R1 has value (0x40; 0x0)") +__failure __msg("At program exit the register R1 has umin=64 umax=64 var_off=(0x40; 0x0)") int check_assert_with_return(void *ctx) {bpf_assert_with (!ctx,64 );
tools/testing/selftests/bpf/progs/exceptions_fail.c +1 -1
@@ -306,7 +306,7 @@ int reject_set_exception_cb_bad_ret1(void *ctx) } SEC("?fentry/bpf_check") -__failure __msg("At program exit the register R1 has value (0x40; 0x0) should") +__failure __msg("At program exit the register R1 has umin=64 umax=64 var_off=(0x40; 0x0) should") int reject_set_exception_cb_bad_ret2(void *ctx) {bpf_throw (64 );
tools/testing/selftests/bpf/progs/test_global_func15.c +1 -1
@@ -13,7 +13,7 @@ __noinline int foo(unsigned int *v) } SEC("cgroup_skb/ingress") -__failure __msg("At program exit the register R0 has value") +__failure __msg("At program exit the register R0 has ") int global_func15(struct __sk_buff *skb) {unsigned int v =1 ;
tools/testing/selftests/bpf/progs/timer_failure.c +1 -1
@@ -30,7 +30,7 @@ static int timer_cb_ret1(void *map, int *key, struct bpf_timer *timer) } SEC("fentry/bpf_fentry_test1") -__failure __msg("should have been in (0x0; 0x0)") +__failure __msg("should have been in [0, 0]") int BPF_PROG2(test_ret_1, int, a) {int key =0 ;
tools/testing/selftests/bpf/progs/user_ringbuf_fail.c +1 -1
@@ -184,7 +184,7 @@ invalid_drain_callback_return(struct bpf_dynptr *dynptr, void *context)* not be able to write to that pointer. */ SEC("?raw_tp") -__failure __msg("At callback return the register R0 has value") +__failure __msg("At callback return the register R0 has ")int user_ringbuf_callback_invalid_return (void *ctx) {bpf_user_ringbuf_drain (&user_ringbuf, invalid_drain_callback_return,NULL ,0 );
tools/testing/selftests/bpf/progs/verifier_cgroup_inv_retcode.c +4 -4
@@ -7,7 +7,7 @@ SEC("cgroup/sock") __description("bpf_exit with invalid return code. test1") -__failure __msg("R0 has value (0x0; 0xffffffff)") +__failure __msg("var_off=(0x0; 0xffffffff) should have been in [0, 1]") __naked void with_invalid_return_code_test1(void) { asm volatile (" \
@@ -30,7 +30,7 @@ __naked void with_invalid_return_code_test2(void) SEC("cgroup/sock") __description("bpf_exit with invalid return code. test3") -__failure __msg("R0 has value (0x0; 0x3)") +__failure __msg("var_off=(0x0; 0x3) should have been in [0, 1]") __naked void with_invalid_return_code_test3(void) { asm volatile (" \
@@ -53,7 +53,7 @@ __naked void with_invalid_return_code_test4(void) SEC("cgroup/sock") __description("bpf_exit with invalid return code. test5") -__failure __msg("R0 has value (0x2; 0x0)") +__failure __msg("var_off=(0x2; 0x0) should have been in [0, 1]") __naked void with_invalid_return_code_test5(void) { asm volatile (" \
@@ -75,7 +75,7 @@ __naked void with_invalid_return_code_test6(void) SEC("cgroup/sock") __description("bpf_exit with invalid return code. test7") -__failure __msg("R0 has unknown scalar value") +__failure __msg("R0 has unknown scalar value should have been in [0, 1]") __naked void with_invalid_return_code_test7(void) { asm volatile (" \
tools/testing/selftests/bpf/progs/verifier_netfilter_retcode.c +1 -1
@@ -39,7 +39,7 @@ __naked void with_valid_return_code_test3(void) SEC("netfilter") __description("bpf_exit with invalid return code. test4") -__failure __msg("R0 has value (0x2; 0x0)") +__failure __msg("R0 has umin=2 umax=2 var_off=(0x2; 0x0) should have been in [0, 1]") __naked void with_invalid_return_code_test4(void) { asm volatile (" \
tools/testing/selftests/bpf/progs/verifier_subprog_precision.c +1 -1
@@ -143,7 +143,7 @@ SEC("?raw_tp") __failure __log_level(2) __flag(BPF_F_TEST_STATE_FREQ) __msg("from 10 to 12: frame1: R0=scalar(umin=1001) R10=fp0 cb") -__msg("At callback return the register R0 has unknown scalar value should have been in (0x0; 0x1)") +__msg("At callback return the register R0 has umin=1001 should have been in [0, 1]")__naked int callback_precise_return_fail (void ) { asm volatile (
Signature · 1 line
2.34.1
Re: [PATCH bpf-next 06/10] bpf: enforce precise retval range on program exit
Quoted 8 lines · Andrii Nakryiko
Similarly to subprog/callback logic, enforce return value of BPF program using more precise umin/umax range, in addition to tnum-based check.
We need to adjust a bunch of tests due to a changed format of an error message.
Signed-off-by Andrii Nakryiko <[email protected]>---
[...]
Andrii Nakryiko
maintainerPATCH 7/10 bpf: unify async callback and program retval checks
In reply to Andrii NakryikoUse common logic to verify program return values and async callback return values. This allows to avoid duplication of any extra steps necessary, like precision marking, which will be added in the next patch.
--- kernel/bpf/verifier.c | 30 ++++++++++++------------------ 1 file changed, 12 insertions(+), 18 deletions(-)
kernel/bpf/verifier.c +12 -18
@@ -358,7 +358,7 @@ static void verbose_invalid_scalar(struct bpf_verifier_env *env,char tn_buf[48 ];bool unknown = true; -verbose (env,"At %s the register %s has" , ctx, reg_name); +verbose (env,"%s the register %s has" , ctx, reg_name);if (reg->umin_value >0 ) {verbose (env," umin=%llu" , reg->umin_value); unknown = false;
@@ -9532,7 +9532,7 @@ static int prepare_func_exit(struct bpf_verifier_env *env, int *insn_idx)/* enforce R0 return value range */ if (!retval_range_within (callee->callback_ret_range, r0)) {verbose_invalid_scalar (env, r0, callee->callback_ret_range, -"callback return" ,"R0" ); +"At callback return" ,"R0" );return -EINVAL ; } }else {
@@ -14858,11 +14858,11 @@ static int check_ld_abs(struct bpf_verifier_env *env, struct bpf_insn *insn)static int check_return_code (struct bpf_verifier_env *env,int regno,const char *reg_name) { +const char *exit_ctx ="At program exit" ;struct tnum enforce_attach_type_range = tnum_unknown;const struct bpf_prog *prog = env->prog;struct bpf_reg_state *reg;struct bpf_retval_range range =retval_range (0 ,1 ); -struct bpf_retval_range const_0 =retval_range (0 ,0 );enum bpf_prog_type prog_type =resolve_prog_type (env->prog);int err;struct bpf_func_state *frame = env->cur_state->frame[0 ];
@@ -14904,17 +14904,9 @@ static int check_return_code(struct bpf_verifier_env *env, int regno, const charif (frame->in_async_callback_fn) {/* enforce return zero from async callbacks like timer */ -if (reg->type !=SCALAR_VALUE ) { -verbose (env,"In async callback the register R%d is not a known value (%s)\n" , - regno,reg_type_str (env, reg->type)); -return -EINVAL ; - } - -if (!retval_range_within (const_0, reg)) { -verbose_invalid_scalar (env, reg, const_0,"async callback" , reg_name); -return -EINVAL ; - } -return 0 ; + exit_ctx ="At async callback return" ; + range =retval_range (0 ,0 ); + goto enforce_retval; }if (is_subprog && !frame->in_exception_callback_fn) {
@@ -15004,15 +14996,17 @@ static int check_return_code(struct bpf_verifier_env *env, int regno, const charreturn 0 ; } +enforce_retval:if (reg->type !=SCALAR_VALUE ) { -verbose (env,"At program exit the register R%d is not a known value (%s)\n" , - regno,reg_type_str (env, reg->type)); +verbose (env,"%s the register R%d is not a known value (%s)\n" , + exit_ctx, regno,reg_type_str (env, reg->type));return -EINVAL ; }if (!retval_range_within (range, reg)) { -verbose_invalid_scalar (env, reg, range,"program exit" , reg_name); -if (prog->expected_attach_type ==BPF_LSM_CGROUP && +verbose_invalid_scalar (env, reg, range, exit_ctx, reg_name); +if (!is_subprog && + prog->expected_attach_type ==BPF_LSM_CGROUP && prog_type ==BPF_PROG_TYPE_LSM && !prog->aux->attach_func_proto->type)verbose (env,"Note, BPF_LSM_CGROUP that attach to void LSM hooks can't modify return value!\n" );
Signature · 1 line
2.34.1
Re: [PATCH bpf-next 07/10] bpf: unify async callback and program retval checks
Quoted 7 lines · Andrii Nakryiko
Use common logic to verify program return values and async callback return values. This allows to avoid duplication of any extra steps necessary, like precision marking, which will be added in the next patch.
Signed-off-by Andrii Nakryiko <[email protected]>---
[...]
Andrii Nakryiko
maintainerPATCH 8/10 bpf: enforce precision of R0 on program/async callback return
In reply to Andrii NakryikoGiven we enforce a valid range for program and async callback return value, we must mark R0 as precise to avoid incorrect state pruning.
--- kernel/bpf/verifier.c | 4 ++++ 1 file changed, 4 insertions(+)
kernel/bpf/verifier.c +4 -0
@@ -15003,6 +15003,10 @@ static int check_return_code(struct bpf_verifier_env *env, int regno, const charreturn -EINVAL ; } + err =mark_chain_precision (env, regno); +if (err) +return err; +if (!retval_range_within (range, reg)) {verbose_invalid_scalar (env, reg, range, exit_ctx, reg_name); if (!is_subprog &&
Signature · 1 line
2.34.1
Re: [PATCH bpf-next 08/10] bpf: enforce precision of R0 on program/async callback return
Quoted 5 lines · Andrii Nakryiko
Given we enforce a valid range for program and async callback return value, we must mark R0 as precise to avoid incorrect state pruning.
Signed-off-by Andrii Nakryiko <[email protected]>---
[...]
Andrii Nakryiko
maintainerPATCH 9/10 selftests/bpf: validate async callback return value check correctness
In reply to Andrii NakryikoAdjust timer/timer_ret_1 test to validate more carefully verifier logic of enforcing async callback return value. This test will pass only if return result is marked precise and read.
--- .../selftests/bpf/progs/timer_failure.c | 29 ++++++++++++++----- 1 file changed, 22 insertions(+), 7 deletions(-)
tools/testing/selftests/bpf/progs/timer_failure.c +22 -7
@@ -21,17 +21,32 @@ struct { __type(value, struct elem); } timer_map SEC(".maps"); -static int timer_cb_ret1 (void *map,int *key,struct bpf_timer *timer) +__naked __noinline __used +static unsigned long timer_cb_ret_bad() { -if (bpf_get_smp_processor_id () %2 ) -return 1 ; -else -return 0 ; + asmvolatile ( +"call %[bpf_get_prandom_u32];" +"if r0 > 1000 goto 1f;" +"r0 = 0;" +"1:" +"goto +0;" /* checkpoint */ +/* async callback is expected to return 0, so branch above +* skipping r0 = 0; should lead to a failure, but if exit +* instruction doesn't enforce r0's precision, this callback +* will be successfully verified +*/ +"exit;" + : + : __imm(bpf_get_prandom_u32) + : __clobber_common + ); } SEC("fentry/bpf_fentry_test1") __failure __msg("should have been in [0, 0]") -int BPF_PROG2(test_ret_1, int, a) +__log_level(2) +__flag (BPF_F_TEST_STATE_FREQ) +int BPF_PROG2(test_bad_ret, int, a) {int key =0 ;struct bpf_timer *timer;
@@ -39,7 +54,7 @@ int BPF_PROG2(test_ret_1, int, a) timer =bpf_map_lookup_elem (&timer_map, &key);if (timer) {bpf_timer_init (timer, &timer_map,CLOCK_BOOTTIME ); -bpf_timer_set_callback (timer, timer_cb_ret1); +bpf_timer_set_callback (timer, timer_cb_ret_bad);bpf_timer_start (timer,1000 ,0 ); }
Signature · 1 line
2.34.1
Re: [PATCH bpf-next 09/10] selftests/bpf: validate async callback return value check correctness
Quoted 6 lines · Andrii Nakryiko
Adjust timer/timer_ret_1 test to validate more carefully verifier logic of enforcing async callback return value. This test will pass only if return result is marked precise and read.
Signed-off-by Andrii Nakryiko <[email protected]>---
[...]
Andrii Nakryiko
maintainerPATCH 10/10 selftests/bpf: adjust global_func15 test to validate prog exit precision
In reply to Andrii NakryikoAdd one more subtest to global_func15 selftest to validate that verifier properly marks r0 as precise and avoids erroneous state pruning of the branch that has return value outside of expected [0, 1] value.
--- .../selftests/bpf/progs/test_global_func15.c | 29 +++++++++++++++++++ 1 file changed, 29 insertions(+)
tools/testing/selftests/bpf/progs/test_global_func15.c +29 -0
@@ -22,3 +22,32 @@ int global_func15(struct __sk_buff *skb)return v; } + +SEC("cgroup_skb/ingress") +__failure __msg("At program exit the register R0 has ") +__log_level(2) __flag(BPF_F_TEST_STATE_FREQ) +__naked int global_func15_tricky_pruning(void) +{ + asmvolatile ( +"r0 = 1;" +"*(u64 *)(r10 -8) = r0;" +"call %[bpf_get_prandom_u32];" +"if r0 > 1000 goto 1f;" +"r0 = 1;" +"1:" +"goto +0;" /* checkpoint */ +/* cgroup_skb/ingress program is expected to return [0, 1] +* values, so branch above makes sure that in a fallthrough +* case we have a valid 1 stored in R0 register, but in +* a branch case we assign some random value to R0. So if +* there is something wrong with precision tracking for R0 at +* program exit, we might erronenously prune branch case, +* because R0 in fallthrough case is imprecise (and thus any +* value is valid from POV of verifier is_state_equal() logic) +*/ +"exit;" + : + : __imm(bpf_get_prandom_u32) + : __clobber_common + ); +}
Signature · 1 line
2.34.1
Re: [PATCH bpf-next 10/10] selftests/bpf: adjust global_func15 test to validate prog exit precision
Quoted 6 lines · Andrii Nakryiko
Quoted 6 lines
Add one more subtest to
global_func15selftest to validate that verifier properly marks r0 as precise and avoids erroneous state pruning of the branch that has return value outside of expected [0, 1] value.Signed-off-by Andrii Nakryiko <[email protected]>---
[...]