| Message ID | 20210909143303.811171-1-johan.almbladh@anyfinetworks.com |
|---|---|
| Headers | show |
| Series | bpf/tests: Extend JIT test suite coverage | expand |
On 9/9/21 4:33 PM, Johan Almbladh wrote: > This patch adds a tail call limit test where the program also emits > a BPF_CALL to an external function prior to the tail call. Mainly > testing that JITed programs preserve its internal register state, for > example tail call count, across such external calls. > > Signed-off-by: Johan Almbladh <johan.almbladh@anyfinetworks.com> > --- > lib/test_bpf.c | 83 ++++++++++++++++++++++++++++++++++++++++++++++++-- > 1 file changed, 80 insertions(+), 3 deletions(-) > > diff --git a/lib/test_bpf.c b/lib/test_bpf.c > index 7475abfd2186..152193b4080f 100644 > --- a/lib/test_bpf.c > +++ b/lib/test_bpf.c > @@ -12202,6 +12202,30 @@ struct tail_call_test { > offset, TAIL_CALL_MARKER), \ > BPF_JMP_IMM(BPF_TAIL_CALL, 0, 0, 0) > > +/* > + * A test function to be called from a BPF program, clobbering a lot of > + * CPU registers in the process. A JITed BPF program calling this function > + * must save and restore any caller-saved registers it uses for internal > + * state, for example the current tail call count. > + */ > +BPF_CALL_1(bpf_test_func, u64, arg) > +{ > + char buf[64]; > + long a = 0; > + long b = 1; > + long c = 2; > + long d = 3; > + long e = 4; > + long f = 5; > + long g = 6; > + long h = 7; > + > + return snprintf(buf, sizeof(buf), > + "%ld %lu %lx %ld %lu %lx %ld %lu %x", > + a, b, c, d, e, f, g, h, (int)arg); > +} > +#define BPF_FUNC_test_func __BPF_FUNC_MAX_ID > + > /* > * Tail call tests. Each test case may call any other test in the table, > * including itself, specified as a relative index offset from the calling > @@ -12259,6 +12283,25 @@ static struct tail_call_test tail_call_tests[] = { > }, > .result = MAX_TAIL_CALL_CNT + 1, > }, > + { > + "Tail call count preserved across function calls", > + .insns = { > + BPF_ALU64_IMM(BPF_ADD, R1, 1), > + BPF_STX_MEM(BPF_DW, R10, R1, -8), > + BPF_CALL_REL(BPF_FUNC_get_numa_node_id), > + BPF_CALL_REL(BPF_FUNC_ktime_get_ns), > + BPF_CALL_REL(BPF_FUNC_ktime_get_boot_ns), > + BPF_CALL_REL(BPF_FUNC_ktime_get_coarse_ns), > + BPF_CALL_REL(BPF_FUNC_jiffies64), > + BPF_CALL_REL(BPF_FUNC_test_func), > + BPF_LDX_MEM(BPF_DW, R1, R10, -8), > + BPF_ALU32_REG(BPF_MOV, R0, R1), > + TAIL_CALL(0), > + BPF_EXIT_INSN(), From discussion with Johan, there'll be a v4 respin since assumption of R0 being valid before exit insn would not hold true when going through verifier. Fixing it confirmed the 33 limit for x86 JIT as well, so both interpreter and JIT is 33-aligned. > + }, > + .stack_depth = 8, > + .result = MAX_TAIL_CALL_CNT + 1, > + }, > { > "Tail call error path, NULL target", > .insns = { > @@ -12333,17 +12376,19 @@ static __init int prepare_tail_call_tests(struct bpf_array **pprogs) > /* Relocate runtime tail call offsets and addresses */ > for (i = 0; i < len; i++) { > struct bpf_insn *insn = &fp->insnsi[i]; > - > - if (insn->imm != TAIL_CALL_MARKER) > - continue; > + long addr = 0; > > switch (insn->code) { > case BPF_LD | BPF_DW | BPF_IMM: