bpf: make DEBUG_INFO_BTF_MODULES selectable independently
Subsystems (2)
PATCH 1/1 bpf: make DEBUG_INFO_BTF_MODULES selectable independently
We're having an issue where we have a recent clang that seems to generate kind_flag for enums (aka, adding signed/unsigned). Trying to install a module on a kernel that doesn't have commit 6089fb325cf7 ("bpf: Add btf enum64 support") returns the following:
[ 3.176954] BPF:Invalid btf_info kind_flag
The enum that it complains about doesn't seem to have anything special except having a sign:
[1721] ENUM 'perf_event_state' encoding=SIGNED size=4 vlen=6
'PERF_EVENT_STATE_DEAD' val=-4
'PERF_EVENT_STATE_EXIT' val=-3
'PERF_EVENT_STATE_ERROR' val=-2
'PERF_EVENT_STATE_OFF' val=-1
'PERF_EVENT_STATE_INACTIVE' val=0
'PERF_EVENT_STATE_ACTIVE' val=1We are not currently using CONFIG_DEBUG_INFO_BTF_MODULES and don't plan to use module BTF, so it's preferable to be able to explicits disable it in the kernel config. Unfortunately, because that kconfig option doesn't have a name, it's not possible to flip it independently from CONFIG_DEBUG_INFO_BTF. Let's add a name to make sure module BTF is user-controllable.
(Not sure, but maybe the right fix is to also have a stable patch
to relax that "Invalid btf_info kind_flag" check?)
--- lib/Kconfig.debug | 1 + 1 file changed, 1 insertion(+)
lib/Kconfig.debug +1 -0
@@ -326,6 +326,7 @@ config PAHOLE_HAS_SPLIT_BTF def_bool $(success, test `$(PAHOLE) --version | sed -E 's/v([0-9]+)\.([0-9]+)/\1\2/'` -ge "119") config DEBUG_INFO_BTF_MODULES + bool "Generate BTF module typeinfo" def_bool y depends on DEBUG_INFO_BTF && MODULES && PAHOLE_HAS_SPLIT_BTF help
Signature · 1 line
2.38.0.rc1.362.ged0d419d3c-goog
Quoted 24 lines · Stanislav Fomichev
We're having an issue where we have a recent clang that seems to generate
kind_flagfor enums (aka, adding signed/unsigned). Trying to install a module on a kernel that doesn't have commit6089fb325cf7("bpf: Add btf enum64 support") returns the following:[ 3.176954] BPF:Invalid btf_info kind_flagThe enum that it complains about doesn't seem to have anything special except having a sign:
[1721] ENUM '
perf_event_state' encoding=SIGNED size=4 vlen=6'PERF_EVENT_STATE_DEAD' val=-4 'PERF_EVENT_STATE_EXIT' val=-3 'PERF_EVENT_STATE_ERROR' val=-2 'PERF_EVENT_STATE_OFF' val=-1 'PERF_EVENT_STATE_INACTIVE' val=0 'PERF_EVENT_STATE_ACTIVE' val=1We are not currently using
CONFIG_DEBUG_INFO_BTF_MODULESand don't plan to use module BTF, so it's preferable to be able to explicits disable it in the kernel config. Unfortunately, because that kconfig option doesn't have a name, it's not possible to flip it independently fromCONFIG_DEBUG_INFO_BTF. Let's add a name to make sure module BTF is user-controllable.
[..]
Quoted 2 lines
(Not sure, but maybe the right fix is to also have a stable patch
to relax that "Invalidbtf_infokind_flag" check?)
Answering to myself, looks like we do need the following for non-enum64-compatible older/stable kernels:
kernel/bpf/btf.c +0 -5
@@ -3301,11 +3301,6 @@ static s32 btf_enum_check_meta(struct btf_verifier_env *env,return -EINVAL ; } -if (btf_type_kflag (t)) { -btf_verifier_log_type (env, t,"Invalid btf_info kind_flag" ); -return -EINVAL ; - } -if (t->size >8 || !is_power_of_2 (t->size)) {btf_verifier_log_type (env, t,"Unexpected size" );return -EINVAL ;
Anything I'm missing? Feels like any pre-6089fb325cf7 ("bpf: Add btf enum64 support") kernel will have an issue with a recent clang that puts sign into kflag?
Quoted 21 lines
Fixes 5f9ae91f7c0d ("kbuild: Build kernel module BTFs if BTF is enabled and pahole supports it")Signed-off-by Stanislav Fomichev <[email protected]>--- lib/Kconfig.debug | 1 + 1 file changed, 1 insertion(+)lib/Kconfig.debug +1
-2@@ -326,6 +326,7 @@ config PAHOLE_HAS_SPLIT_BTF def_bool $(success, test `$(PAHOLE) --version | sed -E 's/v([0-9]+)\.([0-9]+)/\1\2/'` -ge "119") config DEBUG_INFO_BTF_MODULES + bool "Generate BTF module typeinfo" def_bool y depends on DEBUG_INFO_BTF && MODULES && PAHOLE_HAS_SPLIT_BTF help --2.38.0.rc1.362.ged0d419d3c-goog
Quoted 53 lines · On Tue, Oct 4, 2022 at 5:33 PM <[email protected]> wrote:
Quoted 24 lines · Stanislav Fomichev
We're having an issue where we have a recent clang that seems to generate
kind_flagfor enums (aka, adding signed/unsigned). Trying to install a module on a kernel that doesn't have commit6089fb325cf7("bpf: Add btf enum64 support") returns the following:[ 3.176954] BPF:Invalid btf_info kind_flagThe enum that it complains about doesn't seem to have anything special except having a sign:
[1721] ENUM '
perf_event_state' encoding=SIGNED size=4 vlen=6'PERF_EVENT_STATE_DEAD' val=-4 'PERF_EVENT_STATE_EXIT' val=-3 'PERF_EVENT_STATE_ERROR' val=-2 'PERF_EVENT_STATE_OFF' val=-1 'PERF_EVENT_STATE_INACTIVE' val=0 'PERF_EVENT_STATE_ACTIVE' val=1We are not currently using
CONFIG_DEBUG_INFO_BTF_MODULESand don't plan to use module BTF, so it's preferable to be able to explicits disable it in the kernel config. Unfortunately, because that kconfig option doesn't have a name, it's not possible to flip it independently fromCONFIG_DEBUG_INFO_BTF. Let's add a name to make sure module BTF is user-controllable.[..]
Quoted 2 lines
(Not sure, but maybe the right fix is to also have a stable patch
to relax that "Invalidbtf_infokind_flag" check?)Answering to myself, looks like we do need the following for non-enum64-compatible older/stable kernels:
kernel/bpf/btf.c +0
-5@@ -3301,11 +3301,6 @@ static s32 btf_enum_check_meta(struct btf_verifier_env *env, return -EINVAL ; } -if (btf_type_kflag (t)) { -btf_verifier_log_type (env, t,"Invalid btf_info kind_flag" ); -return -EINVAL ; - } -if (t->size >8 || !is_power_of_2 (t->size)) {btf_verifier_log_type (env, t,"Unexpected size" );return -EINVAL;
Quoted 25 lines
Anything I'm missing? Feels like any pre-6089fb325cf7 ("bpf: Add btf enum64 support") kernel will have an issue with a recent clang that puts sign into kflag?
Quoted 20 lines
Fixes 5f9ae91f7c0d ("kbuild: Build kernel module BTFs if BTF is enabledand pahole supports it")
Signed-off-by Stanislav Fomichev <[email protected]>--- lib/Kconfig.debug | 1 + 1 file changed, 1 insertion(+)lib/Kconfig.debug +1
-1@@ -326,6 +326,7 @@ config PAHOLE_HAS_SPLIT_BTF def_bool $(success, test `$(PAHOLE) --version | sed -E 's/v([0-9]+)\.([0-9]+)/\1\2/'` -ge "119") config DEBUG_INFO_BTF_MODULES + bool "Generate BTF module typeinfo" def_bool y depends on DEBUG_INFO_BTF && MODULES && PAHOLE_HAS_SPLIT_BTF help
Not quite following.
Are you saying instead of backporting enum64 support to older kernels you'd prefer to add this patch to upstream and backport this smaller patch?
Quoted 81 lines · Alexei Starovoitov
Quoted 53 lines · On Tue, Oct 4, 2022 at 5:33 PM <[email protected]> wrote:
Quoted 24 lines · Stanislav Fomichev
We're having an issue where we have a recent clang that seems to generate
kind_flagfor enums (aka, adding signed/unsigned). Trying to install a module on a kernel that doesn't have commit6089fb325cf7("bpf: Add btf enum64 support") returns the following:[ 3.176954] BPF:Invalid btf_info kind_flagThe enum that it complains about doesn't seem to have anything special except having a sign:
[1721] ENUM '
perf_event_state' encoding=SIGNED size=4 vlen=6'PERF_EVENT_STATE_DEAD' val=-4 'PERF_EVENT_STATE_EXIT' val=-3 'PERF_EVENT_STATE_ERROR' val=-2 'PERF_EVENT_STATE_OFF' val=-1 'PERF_EVENT_STATE_INACTIVE' val=0 'PERF_EVENT_STATE_ACTIVE' val=1We are not currently using
CONFIG_DEBUG_INFO_BTF_MODULESand don't plan to use module BTF, so it's preferable to be able to explicits disable it in the kernel config. Unfortunately, because that kconfig option doesn't have a name, it's not possible to flip it independently fromCONFIG_DEBUG_INFO_BTF. Let's add a name to make sure module BTF is user-controllable.[..]
Quoted 2 lines
(Not sure, but maybe the right fix is to also have a stable patch
to relax that "Invalidbtf_infokind_flag" check?)Answering to myself, looks like we do need the following for non-enum64-compatible older/stable kernels:
kernel/bpf/btf.c +0
-5@@ -3301,11 +3301,6 @@ static s32 btf_enum_check_meta(struct btf_verifier_env *env, return -EINVAL ; } -if (btf_type_kflag (t)) { -btf_verifier_log_type (env, t,"Invalid btf_info kind_flag" ); -return -EINVAL ; - } -if (t->size >8 || !is_power_of_2 (t->size)) {btf_verifier_log_type (env, t,"Unexpected size" );return -EINVAL;Quoted 25 lines
Anything I'm missing? Feels like any pre-6089fb325cf7 ("bpf: Add btf enum64 support") kernel will have an issue with a recent clang that puts sign into kflag?
Quoted 20 lines
Fixes 5f9ae91f7c0d ("kbuild: Build kernel module BTFs if BTF is enabledand pahole supports it")
Signed-off-by Stanislav Fomichev <[email protected]>--- lib/Kconfig.debug | 1 + 1 file changed, 1 insertion(+)lib/Kconfig.debug +1
-1@@ -326,6 +326,7 @@ config PAHOLE_HAS_SPLIT_BTF def_bool $(success, test `$(PAHOLE) --version | sed -E 's/v([0-9]+)\.([0-9]+)/\1\2/'` -ge "119") config DEBUG_INFO_BTF_MODULES + bool "Generate BTF module typeinfo" def_bool y depends on DEBUG_INFO_BTF && MODULES && PAHOLE_HAS_SPLIT_BTF help
Quoted 4 lines
Not quite following.
Are you saying instead of backporting enum64 support to older kernels you'd prefer to add this patch to upstream and backport this smaller patch?
Yeah, sorry, it took me a while to build the context, I might still be missing something.
So far it seems that disabling DEBUG_INFO_BTF_MODULES is not enough. It looks like as long as the older kernels still have that btf_type_kflag enum check, compiling them with a fairly recent clang won't work?
Or do we expect those users to use pahole's --skip_encoding_btf_enum64 which seems to fallback to the old way?
I guess I'm still trying to understand whether we care about old/stable kernels + recent llvm combination.
Quoted 92 lines · Stanislav Fomichev
Quoted 81 lines · Alexei Starovoitov
Quoted 53 lines · On Tue, Oct 4, 2022 at 5:33 PM <[email protected]> wrote:
Quoted 24 lines · Stanislav Fomichev
We're having an issue where we have a recent clang that seems to generate
kind_flagfor enums (aka, adding signed/unsigned). Trying to install a module on a kernel that doesn't have commit6089fb325cf7("bpf: Add btf enum64 support") returns the following:[ 3.176954] BPF:Invalid btf_info kind_flagThe enum that it complains about doesn't seem to have anything special except having a sign:
[1721] ENUM '
perf_event_state' encoding=SIGNED size=4 vlen=6'PERF_EVENT_STATE_DEAD' val=-4 'PERF_EVENT_STATE_EXIT' val=-3 'PERF_EVENT_STATE_ERROR' val=-2 'PERF_EVENT_STATE_OFF' val=-1 'PERF_EVENT_STATE_INACTIVE' val=0 'PERF_EVENT_STATE_ACTIVE' val=1We are not currently using
CONFIG_DEBUG_INFO_BTF_MODULESand don't plan to use module BTF, so it's preferable to be able to explicits disable it in the kernel config. Unfortunately, because that kconfig option doesn't have a name, it's not possible to flip it independently fromCONFIG_DEBUG_INFO_BTF. Let's add a name to make sure module BTF is user-controllable.[..]
Quoted 2 lines
(Not sure, but maybe the right fix is to also have a stable patch
to relax that "Invalidbtf_infokind_flag" check?)Answering to myself, looks like we do need the following for non-enum64-compatible older/stable kernels:
kernel/bpf/btf.c +0
-5@@ -3301,11 +3301,6 @@ static s32 btf_enum_check_meta(struct btf_verifier_env *env, return -EINVAL ; } -if (btf_type_kflag (t)) { -btf_verifier_log_type (env, t,"Invalid btf_info kind_flag" ); -return -EINVAL ; - } -if (t->size >8 || !is_power_of_2 (t->size)) {btf_verifier_log_type (env, t,"Unexpected size" );return -EINVAL;Quoted 25 lines
Anything I'm missing? Feels like any pre-6089fb325cf7 ("bpf: Add btf enum64 support") kernel will have an issue with a recent clang that puts sign into kflag?
Quoted 20 lines
Fixes 5f9ae91f7c0d ("kbuild: Build kernel module BTFs if BTF is enabledand pahole supports it")
Signed-off-by Stanislav Fomichev <[email protected]>--- lib/Kconfig.debug | 1 + 1 file changed, 1 insertion(+)lib/Kconfig.debug +1
-1@@ -326,6 +326,7 @@ config PAHOLE_HAS_SPLIT_BTF def_bool $(success, test `$(PAHOLE) --version | sed -E 's/v([0-9]+)\.([0-9]+)/\1\2/'` -ge "119") config DEBUG_INFO_BTF_MODULES + bool "Generate BTF module typeinfo" def_bool y depends on DEBUG_INFO_BTF && MODULES && PAHOLE_HAS_SPLIT_BTF helpQuoted 4 lines
Not quite following.
Are you saying instead of backporting enum64 support to older kernels you'd prefer to add this patch to upstream and backport this smaller patch?Yeah, sorry, it took me a while to build the context, I might still be missing something.
Quoted 6 lines
So far it seems that disabling
DEBUG_INFO_BTF_MODULESis not enough. It looks like as long as the older kernels still have thatbtf_type_kflagenum check, compiling them with a fairly recent clang won't work?
Or do we expect those users to use pahole's --skip_encoding_btf_enum64 which seems to fallback to the old way?
Clang doesn't generate kernel or kernel module's BTF, so it has nothing to do with this. It's pahole that needs to be told to not generate kflag bit for enum on older kernels. --skip_encoding_btf_enum64 is not exactly that, seems like we need another flag to disable the sign bit for enum?
cc'ed Arnaldo as well.
Quoted 3 lines
I guess I'm still trying to understand whether we care about old/stable kernels + recent llvm combination.
I think it's similar to --skip_encoding_btf_enum64, so I'd say yes?
Quoted 106 lines · Andrii Nakryiko
Quoted 92 lines · Stanislav Fomichev
Quoted 81 lines · Alexei Starovoitov
Quoted 53 lines · On Tue, Oct 4, 2022 at 5:33 PM <[email protected]> wrote:
Quoted 24 lines · Stanislav Fomichev
We're having an issue where we have a recent clang that seems to generate
kind_flagfor enums (aka, adding signed/unsigned). Trying to install a module on a kernel that doesn't have commit6089fb325cf7("bpf: Add btf enum64 support") returns the following:[ 3.176954] BPF:Invalid btf_info kind_flagThe enum that it complains about doesn't seem to have anything special except having a sign:
[1721] ENUM '
perf_event_state' encoding=SIGNED size=4 vlen=6'PERF_EVENT_STATE_DEAD' val=-4 'PERF_EVENT_STATE_EXIT' val=-3 'PERF_EVENT_STATE_ERROR' val=-2 'PERF_EVENT_STATE_OFF' val=-1 'PERF_EVENT_STATE_INACTIVE' val=0 'PERF_EVENT_STATE_ACTIVE' val=1We are not currently using
CONFIG_DEBUG_INFO_BTF_MODULESand don't plan to use module BTF, so it's preferable to be able to explicits disable it in the kernel config. Unfortunately, because that kconfig option doesn't have a name, it's not possible to flip it independently fromCONFIG_DEBUG_INFO_BTF. Let's add a name to make sure module BTF is user-controllable.[..]
Quoted 2 lines
(Not sure, but maybe the right fix is to also have a stable patch
to relax that "Invalidbtf_infokind_flag" check?)Answering to myself, looks like we do need the following for non-enum64-compatible older/stable kernels:
kernel/bpf/btf.c +0
-5@@ -3301,11 +3301,6 @@ static s32 btf_enum_check_meta(struct btf_verifier_env *env, return -EINVAL ; } -if (btf_type_kflag (t)) { -btf_verifier_log_type (env, t,"Invalid btf_info kind_flag" ); -return -EINVAL ; - } -if (t->size >8 || !is_power_of_2 (t->size)) {btf_verifier_log_type (env, t,"Unexpected size" );return -EINVAL;Quoted 25 lines
Anything I'm missing? Feels like any pre-6089fb325cf7 ("bpf: Add btf enum64 support") kernel will have an issue with a recent clang that puts sign into kflag?
Quoted 20 lines
Fixes 5f9ae91f7c0d ("kbuild: Build kernel module BTFs if BTF is enabledand pahole supports it")
Signed-off-by Stanislav Fomichev <[email protected]>--- lib/Kconfig.debug | 1 + 1 file changed, 1 insertion(+)lib/Kconfig.debug +1
-1@@ -326,6 +326,7 @@ config PAHOLE_HAS_SPLIT_BTF def_bool $(success, test `$(PAHOLE) --version | sed -E 's/v([0-9]+)\.([0-9]+)/\1\2/'` -ge "119") config DEBUG_INFO_BTF_MODULES + bool "Generate BTF module typeinfo" def_bool y depends on DEBUG_INFO_BTF && MODULES && PAHOLE_HAS_SPLIT_BTF helpQuoted 4 lines
Not quite following.
Are you saying instead of backporting enum64 support to older kernels you'd prefer to add this patch to upstream and backport this smaller patch?Yeah, sorry, it took me a while to build the context, I might still be missing something.
Quoted 6 lines
So far it seems that disabling
DEBUG_INFO_BTF_MODULESis not enough. It looks like as long as the older kernels still have thatbtf_type_kflagenum check, compiling them with a fairly recent clang won't work?
Or do we expect those users to use pahole's --skip_encoding_btf_enum64 which seems to fallback to the old way?Clang doesn't generate kernel or kernel module's BTF, so it has nothing to do with this. It's pahole that needs to be told to not generate kflag bit for enum on older kernels. --skip_encoding_btf_enum64 is not exactly that, seems like we need another flag to disable the sign bit for enum?
Quoted 7 lines
cc'ed Arnaldo as well.
Quoted 3 lines
I guess I'm still trying to understand whether we care about old/stable kernels + recent llvm combination.
I think it's similar to --skip_encoding_btf_enum64, so I'd say yes?
hi,
this seems like the issue we fixed recently where BTF with enum64 won't pass stable kernel verifier.. we did fix for stable 5.19 and 5.15:
https://lore.kernel.org/bpf/[email protected]/
https://lore.kernel.org/bpf/[email protected]/
I did not do 5.10 fix as explained in here:
https://lore.kernel.org/bpf/YyeYUEHyR%2FnHM8NT@krava/
jirka
Quoted 116 lines · Jiri Olsa
Quoted 106 lines · Andrii Nakryiko
Quoted 92 lines · Stanislav Fomichev
Quoted 81 lines · Alexei Starovoitov
Quoted 53 lines · On Tue, Oct 4, 2022 at 5:33 PM <[email protected]> wrote:
Quoted 24 lines · Stanislav Fomichev
We're having an issue where we have a recent clang that seems to generate
kind_flagfor enums (aka, adding signed/unsigned). Trying to install a module on a kernel that doesn't have commit6089fb325cf7("bpf: Add btf enum64 support") returns the following:[ 3.176954] BPF:Invalid btf_info kind_flagThe enum that it complains about doesn't seem to have anything special except having a sign:
[1721] ENUM '
perf_event_state' encoding=SIGNED size=4 vlen=6'PERF_EVENT_STATE_DEAD' val=-4 'PERF_EVENT_STATE_EXIT' val=-3 'PERF_EVENT_STATE_ERROR' val=-2 'PERF_EVENT_STATE_OFF' val=-1 'PERF_EVENT_STATE_INACTIVE' val=0 'PERF_EVENT_STATE_ACTIVE' val=1We are not currently using
CONFIG_DEBUG_INFO_BTF_MODULESand don't plan to use module BTF, so it's preferable to be able to explicits disable it in the kernel config. Unfortunately, because that kconfig option doesn't have a name, it's not possible to flip it independently fromCONFIG_DEBUG_INFO_BTF. Let's add a name to make sure module BTF is user-controllable.[..]
Quoted 2 lines
(Not sure, but maybe the right fix is to also have a stable patch
to relax that "Invalidbtf_infokind_flag" check?)Answering to myself, looks like we do need the following for non-enum64-compatible older/stable kernels:
kernel/bpf/btf.c +0
-5@@ -3301,11 +3301,6 @@ static s32 btf_enum_check_meta(struct btf_verifier_env *env, return -EINVAL ; } -if (btf_type_kflag (t)) { -btf_verifier_log_type (env, t,"Invalid btf_info kind_flag" ); -return -EINVAL ; - } -if (t->size >8 || !is_power_of_2 (t->size)) {btf_verifier_log_type (env, t,"Unexpected size" );return -EINVAL;Quoted 25 lines
Anything I'm missing? Feels like any pre-6089fb325cf7 ("bpf: Add btf enum64 support") kernel will have an issue with a recent clang that puts sign into kflag?
Quoted 20 lines
Fixes 5f9ae91f7c0d ("kbuild: Build kernel module BTFs if BTF is enabledand pahole supports it")
Signed-off-by Stanislav Fomichev <[email protected]>--- lib/Kconfig.debug | 1 + 1 file changed, 1 insertion(+)lib/Kconfig.debug +1
-1@@ -326,6 +326,7 @@ config PAHOLE_HAS_SPLIT_BTF def_bool $(success, test `$(PAHOLE) --version | sed -E 's/v([0-9]+)\.([0-9]+)/\1\2/'` -ge "119") config DEBUG_INFO_BTF_MODULES + bool "Generate BTF module typeinfo" def_bool y depends on DEBUG_INFO_BTF && MODULES && PAHOLE_HAS_SPLIT_BTF helpQuoted 4 lines
Not quite following.
Are you saying instead of backporting enum64 support to older kernels you'd prefer to add this patch to upstream and backport this smaller patch?Yeah, sorry, it took me a while to build the context, I might still be missing something.
Quoted 6 lines
So far it seems that disabling
DEBUG_INFO_BTF_MODULESis not enough. It looks like as long as the older kernels still have thatbtf_type_kflagenum check, compiling them with a fairly recent clang won't work?
Or do we expect those users to use pahole's --skip_encoding_btf_enum64 which seems to fallback to the old way?Clang doesn't generate kernel or kernel module's BTF, so it has nothing to do with this. It's pahole that needs to be told to not generate kflag bit for enum on older kernels. --skip_encoding_btf_enum64 is not exactly that, seems like we need another flag to disable the sign bit for enum?
Quoted 7 lines
cc'ed Arnaldo as well.
Quoted 3 lines
I guess I'm still trying to understand whether we care about old/stable kernels + recent llvm combination.
I think it's similar to --skip_encoding_btf_enum64, so I'd say yes?
Quoted 8 lines
hi,
this seems like the issue we fixed recently where BTF with enum64 won't pass stable kernel verifier.. we did fix for stable 5.19 and 5.15:
https://lore.kernel.org/bpf/[email protected]/
https://lore.kernel.org/bpf/[email protected]/I did not do 5.10 fix as explained in here:
https://lore.kernel.org/bpf/YyeYUEHyR%2FnHM8NT@krava/
Oh, great, thanks for the pointer! That's what I was looking for :-)
Quoted 1 line
jirka