follow up for __jited test tag
Subsystems (3)
Eduard Zingerman
maintainercover follow up for __jited test tag
This patch-set is a collection of follow-ups for "__jited test tag to check disassembly after jit" series (see [1]).
First patch is most important:
as it turns out, I broke all test_loader based tests for s390 CI. E.g. see log [2] for s390 execution of test_progs,
note all 'verivier_*' tests being skipped.
This happens because of incorrect handling of corner case when get_current_arch() does not know which architecture to return.
Second patch makes matching of function return sequence in verifier_tailcall_jit more flexible:
-__jited(" retq")
+__jited(" {{(retq|jmp 0x)}}")The difference could be seen with and w/o mitigations=off boot parameter for test VM (CI runs with mitigations=off, hence it generates retq).
Third patch addresses Alexei's request to add #define and a comment in jit_disasm_helpers.c.
[1] https://lore.kernel.org/bpf/[email protected]/
[2] https://github.com/kernel-patches/bpf/actions/runs/10518445973/job/29144511595
Eduard Zingerman (3):
selftests/bpf: test_loader.c:get_current_arch() should not return 0 selftests/bpf: match both retq/rethunk in verifier_tailcall_jit selftests/bpf: #define LOCAL_LABEL_LEN for jit_disasm_helpers.c
.../testing/selftests/bpf/jit_disasm_helpers.c | 17 ++++++++++++++--- .../selftests/bpf/progs/verifier_tailcall_jit.c | 4 ++-- tools/testing/selftests/bpf/test_loader.c | 9 +++++---- 3 files changed, 21 insertions(+), 9 deletions(-)
Signature · 1 line
2.46.0
Eduard Zingerman
maintainerPATCH 1/3 selftests/bpf: test_loader.c:get_current_arch() should not return 0
At the moment, when test_loader.c:get_current_arch() can't determine the arch, it returns 0. The arch check in run_subtest() looks as follows:
if ((get_current_arch() & spec->arch_mask) == 0) {
test__skip();
return;
}Which means that all test_loader based tests would be skipped if arch could not be determined. get_current_arch() recognizes x86_64, arm64 and riscv64. Which means that CI skips test_loader tests for s390.
Fix this by making sure that get_current_arch() always returns non-zero value. In combination with default spec->arch_mask == -1 this should cover all possibilities.
--- tools/testing/selftests/bpf/test_loader.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-)
tools/testing/selftests/bpf/test_loader.c +5 -4
@@ -336,9 +336,10 @@ static const char *skip_dynamic_pfx(const char *s, const char *pfx) }enum arch { -ARCH_X86_64 =0x1 , -ARCH_ARM64 =0x2 , -ARCH_RISCV64 =0x4 , +ARCH_UNKNOWN =0x1 , +ARCH_X86_64 =0x2 , +ARCH_ARM64 =0x4 , +ARCH_RISCV64 =0x8 , };static int get_current_arch (void )
@@ -350,7 +351,7 @@ static int get_current_arch(void)#elif defined(__riscv) && __riscv_xlen ==64 return ARCH_RISCV64 ;#endif -return 0 ; +return ARCH_UNKNOWN ; }/* Uses btf_decl_tag attributes to describe the expected test
Signature · 1 line
2.46.0
Eduard Zingerman
maintainerPATCH 2/3 selftests/bpf: match both retq/rethunk in verifier_tailcall_jit
In reply to Eduard ZingermanDepending on kernel parameters, x86 jit generates either retq or jump to rethunk for 'exit' instruction. The difference could be seen when kernel is booted with and without mitigations=off parameter. Relax the verifier_tailcall_jit test case to match both variants.
--- tools/testing/selftests/bpf/progs/verifier_tailcall_jit.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
tools/testing/selftests/bpf/progs/verifier_tailcall_jit.c +2 -2
@@ -59,7 +59,7 @@ __jited(" movq -0x10(%rbp), %rax") __jited(" callq 0x{{.*}}") /* call to sub() */ __jited(" xorl %eax, %eax") __jited(" leave") -__jited(" retq") +__jited(" {{(retq|jmp 0x)}}") /* return or jump to rethunk */__jited ("..." ) /* subprogram entry for sub(), regular function prologue */__jited (" endbr64" )
@@ -89,7 +89,7 @@ __jited(" popq %rax") __jited(" popq %rax") __jited(" jmp {{.*}}") /* jump to tail call tgt */ __jited("L0: leave") -__jited(" retq") +__jited(" {{(retq|jmp 0x)}}") /* return or jump to rethunk */ SEC("tc")__naked int main (void ) {
Signature · 1 line
2.46.0
Eduard Zingerman
maintainerPATCH 3/3 selftests/bpf: #define LOCAL_LABEL_LEN for jit_disasm_helpers.c
In reply to Eduard ZingermanExtract local label length as a #define directive and elaborate why 'i % MAX_LOCAL_LABELS' expression is needed for local labels array initialization.
--- .../testing/selftests/bpf/jit_disasm_helpers.c | 17 ++++++++++++++--- 1 file changed, 14 insertions(+), 3 deletions(-)
tools/testing/selftests/bpf/jit_disasm_helpers.c +14 -3
@@ -16,6 +16,11 @@*/ #define MAX_LOCAL_LABELS 32 +/* Local labels are encoded as 'L42', this requires 4 bytes of storage: +* 3 characters + zero byte +*/ +#define LOCAL_LABEL_LEN 4 +static bool llvm_initialized;struct local_labels {
@@ -23,7 +28,7 @@ struct local_labels {__u32 prog_len;__u32 cnt;__u32 pcs[MAX_LOCAL_LABELS ]; -char names[MAX_LOCAL_LABELS ][4 ]; +char names[MAX_LOCAL_LABELS ][LOCAL_LABEL_LEN ]; }; static const char *lookup_symbol(void *data, uint64_t ref_value, uint64_t *ref_type,
@@ -118,8 +123,14 @@ static int disasm_one_func(FILE *text_out, uint8_t *image, __u32 len) }qsort (labels.pcs, labels.cnt,sizeof (*labels.pcs), cmp_u32);for (i =0 ; i < labels.cnt; ++i) -/* use (i % 100) to avoid format truncation warning */ -snprintf (labels.names[i],sizeof (labels.names[i]),"L%d" , i %100 ); +/* gcc is unable to infer upper bound for labels.cnt and assumes +* it to be U32_MAX. U32_MAX takes 10 decimal digits. +* snprintf below prints into labels.names[*], +* which has space only for two digits and a letter. +* To avoid truncation warning use (i % MAX_LOCAL_LABELS), +* which informs gcc about printed value upper bound. +*/ +snprintf (labels.names[i],sizeof (labels.names[i]),"L%d" , i %MAX_LOCAL_LABELS );/* now print with labels */ labels.print_phase = true;
Signature · 1 line
2.46.0
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 11 lines · On Fri, 23 Aug 2024 01:06:41 -0700 you wrote:
This patch-set is a collection of follow-ups for "
__jitedtest tag to check disassembly after jit" series (see [1]).First patch is most important:
as it turns out, I broke alltest_loaderbased tests for s390 CI. E.g. see log [2] for s390 execution oftest_progs,
note all 'verivier_*' tests being skipped.
This happens because of incorrect handling of corner case whenget_current_arch()does not know which architecture to return.[...]
Here is the summary with links:
- [bpf-next,1/3] selftests/bpf: test_loader.c:get_current_arch() should not return 0
https://git.kernel.org/bpf/bpf-next/c/ec1f77f6557b - [bpf-next,2/3] selftests/bpf: match both retq/rethunk in verifier_tailcall_jit https://git.kernel.org/bpf/bpf-next/c/c52a1e6eb74f - [bpf-next,3/3] selftests/bpf: #define LOCAL_LABEL_LEN for jit_disasm_helpers.c https://git.kernel.org/bpf/bpf-next/c/21a56fc503fa
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html