btf_dump fixes for s390
Subsystems (4)
v2 cover btf_dump fixes for s390
v1: https://lore.kernel.org/bpf/[email protected]/
v1 -> v2:
- Remove redundant local variables, use t->size directly instead.
- Add btf__align_of() patch.
Pending questions:
- Can things like defined(__i386__) break cross-compilation?
- Why exactly do we need both cpu_number and cpu_profile_flip? If we do, is there a suitable replacement for cpu_number in common code?
---
Hi,
This series along with [1] and [2] fixes all the failures in the btf_dump testsuite currently present on s390, in particular:
* [1] fixes intermittent build bug causing "failed to encode tag ..."
* error messages.
* [2] fixes missing VAR entries on s390.
* Patch 1 disables Intel-specific code in a testcase.
* Patch 2 fixes an endianness-related bug.
* Patch 3 fixes an alignment-related bug.
* Patch 4 improves overly pessimistic alignment handling.
[1] https://lore.kernel.org/bpf/[email protected]/
[2] https://lore.kernel.org/bpf/[email protected]/
Best regards,
Ilya
Ilya Leoshkevich (4):
selftests/bpf: Use cpu_number only on arches that have it
libbpf: Fix dumping big-endian bitfields
libbpf: Fix dumping non-aligned __int128
libbpf: Fix ptr_is_aligned() usages
tools/lib/bpf/btf_dump.c | 34 +++++++++---------- .../selftests/bpf/prog_tests/btf_dump.c | 2 ++ 2 files changed, 19 insertions(+), 17 deletions(-)
Signature · 1 line
2.31.1
PATCH v2 1/4 selftests/bpf: Use cpu_number only on arches that have it
cpu_number exists only on Intel and aarch64, so skip the test involing it on other arches. An alternative would be to replace it with an exported non-ifdefed primitive-typed percpu variable from the common code, but there appears to be none.
--- tools/testing/selftests/bpf/prog_tests/btf_dump.c | 2 ++ 1 file changed, 2 insertions(+)
tools/testing/selftests/bpf/prog_tests/btf_dump.c +2 -0
@@ -778,8 +778,10 @@ static void test_btf_dump_struct_data(struct btf *btf, struct btf_dump *d,static void test_btf_dump_var_data (struct btf *btf,struct btf_dump *d,char *str) { +#if defined(__i386__) || defined(__x86_64__) || defined(__aarch64__)TEST_BTF_DUMP_VAR (btf, d,NULL , str,"cpu_number" , int,BTF_F_COMPACT ,"int cpu_number = (int)100" ,100 ); +#endif TEST_BTF_DUMP_VAR (btf, d,NULL , str,"cpu_profile_flip" , int,BTF_F_COMPACT ,"static int cpu_profile_flip = (int)2" ,2 ); }
Signature · 1 line
2.31.1
PATCH v2 2/4 libbpf: Fix dumping big-endian bitfields
In reply to Ilya LeoshkevichOn big-endian arches not only bytes, but also bits are numbered in reverse order (see e.g. S/390 ELF ABI Supplement, but this is also true for other big-endian arches as well).
--- tools/lib/bpf/btf_dump.c | 15 +++++++-------- 1 file changed, 7 insertions(+), 8 deletions(-)
tools/lib/bpf/btf_dump.c +7 -8
@@ -1562,29 +1562,28 @@ static int btf_dump_get_bitfield_value(struct btf_dump *d, __u64 *value) {__u16 left_shift_bits, right_shift_bits; -__u8 nr_copy_bits, nr_copy_bytes;const __u8 *bytes = data; -int sz = t->size; +__u8 nr_copy_bits;__u64 num =0 ;int i;/* Maximum supported bitfield size is 64 bits */ -if (sz >8 ) { -pr_warn ("unexpected bitfield size %d\n" , sz); +if (t->size >8 ) { +pr_warn ("unexpected bitfield size %d\n" , t->size);return -EINVAL ; }/* Bitfield value retrieval is done in two steps; first relevant bytes are * stored in num, then we left/right shift num to eliminate irrelevant bits. */ - nr_copy_bits = bit_sz + bits_offset; - nr_copy_bytes = t->size;#if __BYTE_ORDER == __LITTLE_ENDIAN -for (i = nr_copy_bytes -1 ; i >=0 ; i--) +for (i = t->size -1 ; i >=0 ; i--) num = num *256 + bytes[i]; + nr_copy_bits = bit_sz + bits_offset;#elif __BYTE_ORDER == __BIG_ENDIAN -for (i =0 ; i < nr_copy_bytes; i++) +for (i =0 ; i < t->size; i++) num = num *256 + bytes[i]; + nr_copy_bits = t->size *8 - bits_offset;#else # error "Unrecognized __BYTE_ORDER__"#endif
Signature · 1 line
2.31.1
PATCH v2 3/4 libbpf: Fix dumping non-aligned __int128
In reply to Ilya LeoshkevichNon-aligned integers are dumped as bields, which is supported for at most 64-bit integers. Fix by using the same trick as btf_dump_float_data(): copy non-aligned values to the local buffer.
--- tools/lib/bpf/btf_dump.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-)
tools/lib/bpf/btf_dump.c +6 -3
@@ -1670,9 +1670,10 @@ static int btf_dump_int_data(struct btf_dump *d, {__u8 encoding =btf_int_encoding (t);bool sign = encoding &BTF_INT_SIGNED ; +char buf[16 ]__aligned(16) ;int sz = t->size; -if (sz ==0 ) { +if (sz ==0 || sz >sizeof (buf)) {pr_warn ("unexpected size %d for id [%u]\n" , sz, type_id);return -EINVAL ; }
@@ -1680,8 +1681,10 @@ static int btf_dump_int_data(struct btf_dump *d,/* handle packed int data - accesses of integers not aligned on * int boundaries can cause problems on some platforms. */ -if (!ptr_is_aligned (data, sz)) -return btf_dump_bitfield_data (d, t, data,0 ,0 ); +if (!ptr_is_aligned (data, sz)) { +memcpy (buf, data, sz); + data = buf; + }switch (sz) {case 16 : {
Signature · 1 line
2.31.1
PATCH v2 4/4 libbpf: Fix ptr_is_aligned() usages
In reply to Ilya LeoshkevichCurrently ptr_is_aligned() takes size, and not alignment, as a parameter, which may be overly pessimistic e.g. for __i128 on s390, which must be only 8-byte aligned. Fix by using btf__align_of() where possible - one notable exception is ptr_sz, for which there is no corresponding type.
--- tools/lib/bpf/btf_dump.c | 12 +++++------- 1 file changed, 5 insertions(+), 7 deletions(-)
tools/lib/bpf/btf_dump.c +5 -7
@@ -1657,9 +1657,9 @@ static int btf_dump_base_type_check_zero(struct btf_dump *d,return 0 ; } -static bool ptr_is_aligned (const void *data,int data_sz) +static bool ptr_is_aligned (const void *data,int alignment) { -return ((uintptr_t )data) % data_sz ==0 ; +return ((uintptr_t )data) % alignment ==0 ; } static int btf_dump_int_data(struct btf_dump *d,
@@ -1681,7 +1681,7 @@ static int btf_dump_int_data(struct btf_dump *d,/* handle packed int data - accesses of integers not aligned on * int boundaries can cause problems on some platforms. */ -if (!ptr_is_aligned (data, sz)) { +if (!ptr_is_aligned (data,btf__align_of (d->btf, type_id))) {memcpy (buf, data, sz); data = buf; }
@@ -1770,7 +1770,7 @@ static int btf_dump_float_data(struct btf_dump *d,int sz = t->size;/* handle unaligned data; copy to local union */ -if (!ptr_is_aligned (data, sz)) { +if (!ptr_is_aligned (data,btf__align_of (d->btf, type_id))) {memcpy (&fl, data, sz); flp = &fl; }
@@ -1953,10 +1953,8 @@ static int btf_dump_get_enum_value(struct btf_dump *d, __u32 id, __s64 *value) { - int sz = t->size; -/* handle unaligned enum value */ -if (!ptr_is_aligned (data, sz)) { +if (!ptr_is_aligned (data,btf__align_of (d->btf, id))) {__u64 val;int err;
Signature · 1 line
2.31.1
Re: [PATCH bpf-next v2 4/4] libbpf: Fix ptr_is_aligned() usages
Quoted 17 lines · Ilya Leoshkevich
Currently
ptr_is_aligned()takes size, and not alignment, as a parameter, which may be overly pessimistic e.g. for__i128on s390, which must be only 8-byte aligned. Fix by usingbtf__align_of()where possible - one notable exception isptr_sz, for which there is no corresponding type.Signed-off-by Ilya Leoshkevich <[email protected]>--- tools/lib/bpf/btf_dump.c | 12 +++++------- 1 file changed, 5 insertions(+), 7 deletions(-)tools/lib/bpf/btf_dump.c +0
-0
tools/lib/bpf/btf_dump.c static int btf_dump_base_type_check_zero(struct btf_dump *d, · line 1662
Quoted 8 lines
tools/lib/bpf/btf_dump.c +2
-2@@ -1657,9 +1657,9 @@ static int btf_dump_base_type_check_zero(struct btf_dump *d,return 0 ; } -static bool ptr_is_aligned (const void *data,int data_sz) +static bool ptr_is_aligned (const void *data,int alignment) { -return ((uintptr_t )data) % data_sz ==0 ; +return ((uintptr_t )data) % alignment ==0 ;
btf__align_of() can return 0 on error and this will be div by 0. I think the better approach would be for ptr_is_aligned to accept struct btf *btf and __u32 type_id, call btf__align_of() based on btf and type id, handle 0 case pessimistically (assume not aligned).
Quoted 36 lines
tools/lib/bpf/btf_dump.c +3
-6} static int btf_dump_int_data(struct btf_dump *d,@@ -1681,7 +1681,7 @@ static int btf_dump_int_data(struct btf_dump *d,/* handle packed int data - accesses of integers not aligned on * int boundaries can cause problems on some platforms. */ -if (!ptr_is_aligned (data, sz)) { +if (!ptr_is_aligned (data,btf__align_of (d->btf, type_id))) {memcpy (buf, data, sz); data = buf; }@@ -1770,7 +1770,7 @@ static int btf_dump_float_data(struct btf_dump *d,int sz = t->size;/* handle unaligned data; copy to local union */ -if (!ptr_is_aligned (data, sz)) { +if (!ptr_is_aligned (data,btf__align_of (d->btf, type_id))) {memcpy (&fl, data, sz); flp = &fl; }@@ -1953,10 +1953,8 @@ static int btf_dump_get_enum_value(struct btf_dump *d, __u32 id, __s64 *value) { - int sz = t->size; -/* handle unaligned enum value */ -if (!ptr_is_aligned (data, sz)) { +if (!ptr_is_aligned (data,btf__align_of (d->btf, id))) {__u64 val;int err; --2.31.1
Re: [PATCH bpf-next v2 4/4] libbpf: Fix ptr_is_aligned() usages
Quoted 24 lines · Andrii Nakryiko
Quoted 22 lines · Ilya Leoshkevich
Currently
ptr_is_aligned()takes size, and not alignment, as a parameter, which may be overly pessimistic e.g. for__i128on s390, which must be only 8-byte aligned. Fix by usingbtf__align_of()where
possible - one notable exception isptr_sz, for which there is no corresponding type.Signed-off-by Ilya Leoshkevich <[email protected]>---
tools/lib/bpf/btf_dump.c | 12 +++++-------1 file changed, 5 insertions(+), 7 deletions(-)tools/lib/bpf/btf_dump.c +0
-0@@ -1657,9 +1657,9 @@ static int btf_dump_base_type_check_zero(struct btf_dump *d, return 0; }
Quoted 12 lines
tools/lib/bpf/btf_dump.c static int btf_dump_base_type_check_zero(struct btf_dump *d, · line 1662
Quoted 5 lines
tools/lib/bpf/btf_dump.c +2
-2-static bool ptr_is_aligned (const void *data,int data_sz) +static bool ptr_is_aligned (const void *data,int alignment) { - return ((uintptr_t)data) % data_sz == 0; + return ((uintptr_t)data) % alignment == 0;
btf__align_of()can return 0 on error and this will be div by 0. I think the better approach would be forptr_is_alignedto accept struct
btf *btf and__u32type_id, callbtf__align_of()based on btf and type
id, handle 0 case pessimistically (assume not aligned).
I thought about this, but it won't cover the ptr_sz case. Maybe we just need two functions - I'll give it a try tomorrow.
Re: [PATCH bpf-next v2 4/4] libbpf: Fix ptr_is_aligned() usages
Quoted 39 lines · Ilya Leoshkevich
Quoted 24 lines · Andrii Nakryiko
Quoted 22 lines · Ilya Leoshkevich
Currently
ptr_is_aligned()takes size, and not alignment, as a parameter, which may be overly pessimistic e.g. for__i128on s390, which must be only 8-byte aligned. Fix by usingbtf__align_of()where
possible - one notable exception isptr_sz, for which there is no corresponding type.Signed-off-by Ilya Leoshkevich <[email protected]>--- tools/lib/bpf/btf_dump.c | 12 +++++------- 1 file changed, 5 insertions(+), 7 deletions(-)tools/lib/bpf/btf_dump.c +0
-0@@ -1657,9 +1657,9 @@ static int btf_dump_base_type_check_zero(struct btf_dump *d, return 0; }Quoted 12 lines
tools/lib/bpf/btf_dump.c static int btf_dump_base_type_check_zero(struct btf_dump *d, · line 1662
Quoted 5 lines
tools/lib/bpf/btf_dump.c +2
-2-static bool ptr_is_aligned (const void *data,int data_sz) +static bool ptr_is_aligned (const void *data,int alignment) { -return ((uintptr_t )data) % data_sz ==0 ; +return ((uintptr_t )data) % alignment ==0 ;
btf__align_of()can return 0 on error and this will be div by 0. I think the better approach would be forptr_is_alignedto accept struct
btf *btf and__u32type_id, callbtf__align_of()based on btf and type
id, handle 0 case pessimistically (assume not aligned).
Quoted 3 lines
I thought about this, but it won't cover the
ptr_szcase. Maybe we just need two functions - I'll give it a try tomorrow.
Sorry, what's the ptr_sz case? Is this about btf_ptr_sz() helper somehow?
Re: [PATCH bpf-next v2 4/4] libbpf: Fix ptr_is_aligned() usages
Quoted 43 lines · Andrii Nakryiko
Quoted 41 lines · Ilya Leoshkevich
Quoted 38 lines · Andrii Nakryiko
Quoted 19 lines · Ilya Leoshkevich
Currently
ptr_is_aligned()takes size, and not alignment, as a parameter, which may be overly pessimistic e.g. for__i128on s390,
which must be only 8-byte aligned. Fix by usingbtf__align_of()where
possible - one notable exception isptr_sz, for which there is no corresponding type.Signed-off-by Ilya Leoshkevich <[email protected]>---
tools/lib/bpf/btf_dump.c | 12 +++++-------1 file changed, 5 insertions(+), 7 deletions(-)tools/lib/bpf/btf_dump.c +0
-0tools/lib/bpf/btf_dump.c static int · line 1663
Quoted 9 lines
tools/lib/bpf/btf_dump.c +2
-2@@ -1657,9 +1657,9 @@ static int btf_dump_base_type_check_zero(struct btf_dump *d, return 0; } -static bool ptr_is_aligned (const void *data,int data_sz) +static bool ptr_is_aligned (const void *data,int alignment) { - return ((uintptr_t)data) % data_sz == 0; + return ((uintptr_t)data) % alignment == 0;
btf__align_of()can return 0 on error and this will be div by 0. I think the better approach would be forptr_is_alignedto accept struct
btf *btf and__u32type_id, callbtf__align_of()based on btf and type
id, handle 0 case pessimistically (assume not aligned).
Quoted 6 lines
Quoted 3 lines
I thought about this, but it won't cover the
ptr_szcase. Maybe we just need two functions - I'll give it a try tomorrow.Sorry, what's the
ptr_szcase? Is this aboutbtf_ptr_sz()helper somehow?
Almost, I'm referring to this check:
static int btf_dump_ptr_data(struct btf_dump *d,
const struct btf_type *t, __u32 id,
const void *data)
{
if (ptr_is_aligned(data, d->ptr_sz) && d->ptr_sz ==sizeof(void *)) {
...
Quoted 37 lines · Ilya Leoshkevich
v1: https://lore.kernel.org/bpf/[email protected]/
v1 -> v2:
- Remove redundant local variables, uset->sizedirectly instead.
- Addbtf__align_of()patch.
Pending questions:
- Can things likedefined(__i386__)break cross-compilation?
- Why exactly do we need bothcpu_numberandcpu_profile_flip? If we do, is there a suitable replacement forcpu_numberin common code?---
Hi,
This series along with [1] and [2] fixes all the failures in the
btf_dumptestsuite currently present on s390, in particular:* [1] fixes intermittent build bug causing "failed to encode tag ..."
* error messages.
* [2] fixes missing VAR entries on s390.
* Patch 1 disables Intel-specific code in a testcase.
* Patch 2 fixes an endianness-related bug.
* Patch 3 fixes an alignment-related bug.
* Patch 4 improves overly pessimistic alignment handling.[1] https://lore.kernel.org/bpf/[email protected]/
[2] https://lore.kernel.org/bpf/[email protected]/Best regards,
IlyaIlya Leoshkevich (4):
selftests/bpf: Usecpu_numberonly on arches that have it
libbpf: Fix dumping big-endian bitfields
libbpf: Fix dumping non-aligned__int128
libbpf: Fixptr_is_aligned()usages
Quoted 7 lines
tools/lib/bpf/btf_dump.c | 34 +++++++++---------- .../selftests/bpf/prog_tests/btf_dump.c | 2 ++ 2 files changed, 19 insertions(+), 17 deletions(-)--
2.31.1
I've dropped patch 4 for now, applied the first three to bpf-next, thanks.