[PATCH bpf-next v4 00/10] bpf: enhancements for multi-function programs

[1] Support for bpf-to-bpf function calls in the powerpc64 JIT compiler.

[2] Provide a way for resolving function calls because of the way JITed
    images are allocated in powerpc64.

[3] Fix to get JITed instruction dumps for multi-function programs from
    the bpf system call.

[4] Fix for bpftool to show delimited multi-function JITed image dumps.

v4:
 - Incorporate review comments from Jakub.
 - Fix JSON output for bpftool.

v3:
 - Change base tree tag to bpf-next.
 - Incorporate review comments from Alexei, Daniel and Jakub.
 - Make sure that the JITed image does not grow or shrink after
   the last pass due to the way the instruction sequence used
   to load a callee's address maybe optimized.
 - Make additional changes to the bpf system call and bpftool to
   make multi-function JITed dumps easier to correlate.

v2:
 - Incorporate review comments from Jakub.

Sandipan Das (10):
  bpf: support 64-bit offsets for bpf function calls
  bpf: powerpc64: pad function address loads with NOPs
  bpf: powerpc64: add JIT support for multi-function programs
  bpf: get kernel symbol addresses via syscall
  tools: bpf: sync bpf uapi header
  tools: bpftool: resolve calls without using imm field
  bpf: fix multi-function JITed dump obtained via syscall
  bpf: get JITed image lengths of functions via syscall
  tools: bpf: sync bpf uapi header
  tools: bpftool: add delimiters to multi-function JITed dumps

 arch/powerpc/net/bpf_jit_comp64.c | 110 ++++++++++++++++++++++++++++++--------
 include/uapi/linux/bpf.h          |   4 ++
 kernel/bpf/syscall.c              |  82 ++++++++++++++++++++++++++--
 kernel/bpf/verifier.c             |  22 +++++---
 tools/bpf/bpftool/prog.c          |  97 ++++++++++++++++++++++++++++++++-
 tools/bpf/bpftool/xlated_dumper.c |  14 +++--
 tools/bpf/bpftool/xlated_dumper.h |   3 ++
 tools/include/uapi/linux/bpf.h    |   4 ++
 8 files changed, 301 insertions(+), 35 deletions(-)

-- 
2.14.3

The imm field of a bpf instruction is a signed 32-bit integer.
For JITed bpf-to-bpf function calls, it holds the offset of the
start address of the callee's JITed image from __bpf_call_base.

For some architectures, such as powerpc64, this offset may be
as large as 64 bits and cannot be accomodated in the imm field
without truncation.

We resolve this by:

[1] Additionally using the auxiliary data of each function to
    keep a list of start addresses of the JITed images for all
    functions determined by the verifier.

[2] Retaining the subprog id inside the off field of the call
    instructions and using it to index into the list mentioned
    above and lookup the callee's address.

To make sure that the existing JIT compilers continue to work
without requiring changes, we keep the imm field as it is.

Signed-off-by: Sandipan Das <redacted>
---
 kernel/bpf/verifier.c | 15 ++++++++++++++-
 1 file changed, 14 insertions(+), 1 deletion(-)

diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index a9e4b1372da6..559cb74ba29e 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c

@@ -5383,11 +5383,24 @@ static int jit_subprogs(struct bpf_verifier_env *env)
 			    insn->src_reg != BPF_PSEUDO_CALL)
 				continue;
 			subprog = insn->off;
-			insn->off = 0;
 			insn->imm = (u64 (*)(u64, u64, u64, u64, u64))
 				func[subprog]->bpf_func -
 				__bpf_call_base;
 		}
+
+		/* we use the aux data to keep a list of the start addresses
+		 * of the JITed images for each function in the program
+		 *
+		 * for some architectures, such as powerpc64, the imm field
+		 * might not be large enough to hold the offset of the start
+		 * address of the callee's JITed image from __bpf_call_base
+		 *
+		 * in such cases, we can lookup the start address of a callee
+		 * by using its subprog id, available from the off field of
+		 * the call instruction, as an index for this list
+		 */
+		func[i]->aux->func = func;
+		func[i]->aux->func_cnt = env->subprog_cnt;
 	}
 	for (i = 0; i < env->subprog_cnt; i++) {
 		old_bpf_func = func[i]->bpf_func;

-- 
2.14.3

For multi-function programs, loading the address of a callee
function to a register requires emitting instructions whose
count varies from one to five depending on the nature of the
address.

Since we come to know of the callee's address only before the
extra pass, the number of instructions required to load this
address may vary from what was previously generated. This can
make the JITed image grow or shrink.

To avoid this, we should generate a constant five-instruction
when loading function addresses by padding the optimized load
sequence with NOPs.

Signed-off-by: Sandipan Das <redacted>
---
 arch/powerpc/net/bpf_jit_comp64.c | 34 +++++++++++++++++++++++-----------
 1 file changed, 23 insertions(+), 11 deletions(-)

diff --git a/arch/powerpc/net/bpf_jit_comp64.c b/arch/powerpc/net/bpf_jit_comp64.c
index 1bdb1aff0619..e4582744a31d 100644
--- a/arch/powerpc/net/bpf_jit_comp64.c
+++ b/arch/powerpc/net/bpf_jit_comp64.c

@@ -167,25 +167,37 @@ static void bpf_jit_build_epilogue(u32 *image, struct codegen_context *ctx)
 
 static void bpf_jit_emit_func_call(u32 *image, struct codegen_context *ctx, u64 func)
 {
+	unsigned int i, ctx_idx = ctx->idx;
+
+	/* Load function address into r12 */
+	PPC_LI64(12, func);
+
+	/* For bpf-to-bpf function calls, the callee's address is unknown
+	 * until the last extra pass. As seen above, we use PPC_LI64() to
+	 * load the callee's address, but this may optimize the number of
+	 * instructions required based on the nature of the address.
+	 *
+	 * Since we don't want the number of instructions emitted to change,
+	 * we pad the optimized PPC_LI64() call with NOPs to guarantee that
+	 * we always have a five-instruction sequence, which is the maximum
+	 * that PPC_LI64() can emit.
+	 */
+	for (i = ctx->idx - ctx_idx; i < 5; i++)
+		PPC_NOP();
+
 #ifdef PPC64_ELF_ABI_v1
-	/* func points to the function descriptor */
-	PPC_LI64(b2p[TMP_REG_2], func);
-	/* Load actual entry point from function descriptor */
-	PPC_BPF_LL(b2p[TMP_REG_1], b2p[TMP_REG_2], 0);
-	/* ... and move it to LR */
-	PPC_MTLR(b2p[TMP_REG_1]);
 	/*
 	 * Load TOC from function descriptor at offset 8.
 	 * We can clobber r2 since we get called through a
 	 * function pointer (so caller will save/restore r2)
 	 * and since we don't use a TOC ourself.
 	 */
-	PPC_BPF_LL(2, b2p[TMP_REG_2], 8);
-#else
-	/* We can clobber r12 */
-	PPC_FUNC_ADDR(12, func);
-	PPC_MTLR(12);
+	PPC_BPF_LL(2, 12, 8);
+	/* Load actual entry point from function descriptor */
+	PPC_BPF_LL(12, 12, 0);
 #endif
+
+	PPC_MTLR(12);
 	PPC_BLRL();
 }
 

-- 
2.14.3

This adds new two new fields to struct bpf_prog_info. For
multi-function programs, these fields can be used to pass
a list of kernel symbol addresses for all functions in a
given program to userspace using the bpf system call with
the BPF_OBJ_GET_INFO_BY_FD command.

When bpf_jit_kallsyms is enabled, we can get the address
of the corresponding kernel symbol for a callee function
and resolve the symbol's name. The address is determined
by adding the value of the call instruction's imm field
to __bpf_call_base. This offset gets assigned to the imm
field by the verifier.

For some architectures, such as powerpc64, the imm field
is not large enough to hold this offset.

We resolve this by:

[1] Assigning the subprog id to the imm field of a call
    instruction in the verifier instead of the offset of
    the callee's symbol's address from __bpf_call_base.

[2] Determining the address of a callee's corresponding
    symbol by using the imm field as an index for the
    list of kernel symbol addresses now available from
    the program info.

Suggested-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Sandipan Das <redacted>
---
v3:
 - Copy addresses to jited_ksyms only if bpf_dump_raw_ok()
   is true.
 - Move new fields to the end of bpf_prog_info to avoid
   breaking userspace.
---
 include/uapi/linux/bpf.h |  2 ++
 kernel/bpf/syscall.c     | 25 +++++++++++++++++++++++++
 kernel/bpf/verifier.c    |  7 +------
 3 files changed, 28 insertions(+), 6 deletions(-)

diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
index c3e502d06bc3..0be90965867d 100644
--- a/include/uapi/linux/bpf.h
+++ b/include/uapi/linux/bpf.h

@@ -2205,6 +2205,8 @@ struct bpf_prog_info {
 	__u32 gpl_compatible:1;
 	__u64 netns_dev;
 	__u64 netns_ino;
+	__u32 nr_jited_ksyms;
+	__aligned_u64 jited_ksyms;
 } __attribute__((aligned(8)));
 
 struct bpf_map_info {

diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
index 0b4c94551001..068a4fc79ddb 100644
--- a/kernel/bpf/syscall.c
+++ b/kernel/bpf/syscall.c

@@ -1933,6 +1933,7 @@ static int bpf_prog_get_info_by_fd(struct bpf_prog *prog,
 	if (!capable(CAP_SYS_ADMIN)) {
 		info.jited_prog_len = 0;
 		info.xlated_prog_len = 0;
+		info.nr_jited_ksyms = 0;
 		goto done;
 	}
 

@@ -1981,6 +1982,30 @@ static int bpf_prog_get_info_by_fd(struct bpf_prog *prog,
 		}
 	}
 
+	ulen = info.nr_jited_ksyms;
+	info.nr_jited_ksyms = prog->aux->func_cnt;
+	if (info.nr_jited_ksyms && ulen) {
+		if (bpf_dump_raw_ok()) {
+			u64 __user *user_ksyms;
+			ulong ksym_addr;
+			u32 i;
+
+			/* copy the address of the kernel symbol
+			 * corresponding to each function
+			 */
+			ulen = min_t(u32, info.nr_jited_ksyms, ulen);
+			user_ksyms = u64_to_user_ptr(info.jited_ksyms);
+			for (i = 0; i < ulen; i++) {
+				ksym_addr = (ulong) prog->aux->func[i]->bpf_func;
+				ksym_addr &= PAGE_MASK;
+				if (put_user((u64) ksym_addr, &user_ksyms[i]))
+					return -EFAULT;
+			}
+		} else {
+			info.jited_ksyms = 0;
+		}
+	}
+
 done:
 	if (copy_to_user(uinfo, &info, info_len) ||
 	    put_user(info_len, &uattr->info.info_len))

diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c
index 559cb74ba29e..8c4d9d0fd3ab 100644
--- a/kernel/bpf/verifier.c
+++ b/kernel/bpf/verifier.c

@@ -5426,17 +5426,12 @@ static int jit_subprogs(struct bpf_verifier_env *env)
 	 * later look the same as if they were interpreted only.
 	 */
 	for (i = 0, insn = prog->insnsi; i < prog->len; i++, insn++) {
-		unsigned long addr;
-
 		if (insn->code != (BPF_JMP | BPF_CALL) ||
 		    insn->src_reg != BPF_PSEUDO_CALL)
 			continue;
 		insn->off = env->insn_aux_data[i].call_imm;
 		subprog = find_subprog(env, i + insn->off + 1);
-		addr  = (unsigned long)func[subprog]->bpf_func;
-		addr &= PAGE_MASK;
-		insn->imm = (u64 (*)(u64, u64, u64, u64, u64))
-			    addr - __bpf_call_base;
+		insn->imm = subprog;
 	}
 
 	prog->jited = 1;

-- 
2.14.3

This adds support for bpf-to-bpf function calls in the powerpc64
JIT compiler. The JIT compiler converts the bpf call instructions
to native branch instructions. After a round of the usual passes,
the start addresses of the JITed images for the callee functions
are known. Finally, to fixup the branch target addresses, we need
to perform an extra pass.

Because of the address range in which JITed images are allocated
on powerpc64, the offsets of the start addresses of these images
from __bpf_call_base are as large as 64 bits. So, for a function
call, we cannot use the imm field of the instruction to determine
the callee's address. Instead, we use the alternative method of
getting it from the list of function addresses in the auxiliary
data of the caller by using the off field as an index.

Signed-off-by: Sandipan Das <redacted>
---
v3:
 - Fix memory leak for jit_data when we fail to allocated addrs.
 - Remove unnecessary bpf_jit_binary_lock_ro() call.
---
 arch/powerpc/net/bpf_jit_comp64.c | 76 +++++++++++++++++++++++++++++++++------
 1 file changed, 66 insertions(+), 10 deletions(-)

diff --git a/arch/powerpc/net/bpf_jit_comp64.c b/arch/powerpc/net/bpf_jit_comp64.c
index e4582744a31d..f1c95779843b 100644
--- a/arch/powerpc/net/bpf_jit_comp64.c
+++ b/arch/powerpc/net/bpf_jit_comp64.c

@@ -268,7 +268,7 @@ static void bpf_jit_emit_tail_call(u32 *image, struct codegen_context *ctx, u32
 /* Assemble the body code between the prologue & epilogue */
 static int bpf_jit_build_body(struct bpf_prog *fp, u32 *image,
 			      struct codegen_context *ctx,
-			      u32 *addrs)
+			      u32 *addrs, bool extra_pass)
 {
 	const struct bpf_insn *insn = fp->insnsi;
 	int flen = fp->len;

@@ -724,11 +724,25 @@ static int bpf_jit_build_body(struct bpf_prog *fp, u32 *image,
 			break;
 
 		/*
-		 * Call kernel helper
+		 * Call kernel helper or bpf function
 		 */
 		case BPF_JMP | BPF_CALL:
 			ctx->seen |= SEEN_FUNC;
-			func = (u8 *) __bpf_call_base + imm;
+
+			/* bpf function call */
+			if (insn[i].src_reg == BPF_PSEUDO_CALL)
+				if (!extra_pass)
+					func = NULL;
+				else if (fp->aux->func && off < fp->aux->func_cnt)
+					/* use the subprog id from the off
+					 * field to lookup the callee address
+					 */
+					func = (u8 *) fp->aux->func[off]->bpf_func;
+				else
+					return -EINVAL;
+			/* kernel helper call */
+			else
+				func = (u8 *) __bpf_call_base + imm;
 
 			bpf_jit_emit_func_call(image, ctx, (u64)func);
 

@@ -876,6 +890,14 @@ static int bpf_jit_build_body(struct bpf_prog *fp, u32 *image,
 	return 0;
 }
 
+struct powerpc64_jit_data {
+	struct bpf_binary_header *header;
+	u32 *addrs;
+	u8 *image;
+	u32 proglen;
+	struct codegen_context ctx;
+};
+
 struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *fp)
 {
 	u32 proglen;

@@ -883,6 +905,7 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *fp)
 	u8 *image = NULL;
 	u32 *code_base;
 	u32 *addrs;
+	struct powerpc64_jit_data *jit_data;
 	struct codegen_context cgctx;
 	int pass;
 	int flen;

@@ -890,6 +913,7 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *fp)
 	struct bpf_prog *org_fp = fp;
 	struct bpf_prog *tmp_fp;
 	bool bpf_blinded = false;
+	bool extra_pass = false;
 
 	if (!fp->jit_requested)
 		return org_fp;

@@ -903,11 +927,32 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *fp)
 		fp = tmp_fp;
 	}
 
+	jit_data = fp->aux->jit_data;
+	if (!jit_data) {
+		jit_data = kzalloc(sizeof(*jit_data), GFP_KERNEL);
+		if (!jit_data) {
+			fp = org_fp;
+			goto out;
+		}
+		fp->aux->jit_data = jit_data;
+	}
+
 	flen = fp->len;
+	addrs = jit_data->addrs;
+	if (addrs) {
+		cgctx = jit_data->ctx;
+		image = jit_data->image;
+		bpf_hdr = jit_data->header;
+		proglen = jit_data->proglen;
+		alloclen = proglen + FUNCTION_DESCR_SIZE;
+		extra_pass = true;
+		goto skip_init_ctx;
+	}
+
 	addrs = kzalloc((flen+1) * sizeof(*addrs), GFP_KERNEL);
 	if (addrs == NULL) {
 		fp = org_fp;
-		goto out;
+		goto out_addrs;
 	}
 
 	memset(&cgctx, 0, sizeof(struct codegen_context));

@@ -916,10 +961,10 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *fp)
 	cgctx.stack_size = round_up(fp->aux->stack_depth, 16);
 
 	/* Scouting faux-generate pass 0 */
-	if (bpf_jit_build_body(fp, 0, &cgctx, addrs)) {
+	if (bpf_jit_build_body(fp, 0, &cgctx, addrs, false)) {
 		/* We hit something illegal or unsupported. */
 		fp = org_fp;
-		goto out;
+		goto out_addrs;
 	}
 
 	/*

@@ -937,9 +982,10 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *fp)
 			bpf_jit_fill_ill_insns);
 	if (!bpf_hdr) {
 		fp = org_fp;
-		goto out;
+		goto out_addrs;
 	}
 
+skip_init_ctx:
 	code_base = (u32 *)(image + FUNCTION_DESCR_SIZE);
 
 	/* Code generation passes 1-2 */

@@ -947,7 +993,7 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *fp)
 		/* Now build the prologue, body code & epilogue for real. */
 		cgctx.idx = 0;
 		bpf_jit_build_prologue(code_base, &cgctx);
-		bpf_jit_build_body(fp, code_base, &cgctx, addrs);
+		bpf_jit_build_body(fp, code_base, &cgctx, addrs, extra_pass);
 		bpf_jit_build_epilogue(code_base, &cgctx);
 
 		if (bpf_jit_enable > 1)

@@ -973,10 +1019,20 @@ struct bpf_prog *bpf_int_jit_compile(struct bpf_prog *fp)
 	fp->jited_len = alloclen;
 
 	bpf_flush_icache(bpf_hdr, (u8 *)bpf_hdr + (bpf_hdr->pages * PAGE_SIZE));
+	if (!fp->is_func || extra_pass) {
+out_addrs:
+		kfree(addrs);
+		kfree(jit_data);
+		fp->aux->jit_data = NULL;
+	} else {
+		jit_data->addrs = addrs;
+		jit_data->ctx = cgctx;
+		jit_data->proglen = proglen;
+		jit_data->image = image;
+		jit_data->header = bpf_hdr;
+	}
 
 out:
-	kfree(addrs);
-
 	if (bpf_blinded)
 		bpf_jit_prog_release_other(fp, fp == org_fp ? tmp_fp : org_fp);
 

-- 
2.14.3

Syncing the bpf.h uapi header with tools so that struct
bpf_prog_info has the two new fields for passing on the
addresses of the kernel symbols corresponding to each
function in a program.

Signed-off-by: Sandipan Das <redacted>
---
v3:
 - Move new fields to the end of bpf_prog_info to avoid
   breaking userspace.
---
 tools/include/uapi/linux/bpf.h | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/tools/include/uapi/linux/bpf.h b/tools/include/uapi/linux/bpf.h
index c3e502d06bc3..0be90965867d 100644
--- a/tools/include/uapi/linux/bpf.h
+++ b/tools/include/uapi/linux/bpf.h

@@ -2205,6 +2205,8 @@ struct bpf_prog_info {
 	__u32 gpl_compatible:1;
 	__u64 netns_dev;
 	__u64 netns_ino;
+	__u32 nr_jited_ksyms;
+	__aligned_u64 jited_ksyms;
 } __attribute__((aligned(8)));
 
 struct bpf_map_info {

-- 
2.14.3

Currently, we resolve the callee's address for a JITed function
call by using the imm field of the call instruction as an offset
from __bpf_call_base. If bpf_jit_kallsyms is enabled, we further
use this address to get the callee's kernel symbol's name.

For some architectures, such as powerpc64, the imm field is not
large enough to hold this offset. So, instead of assigning this
offset to the imm field, the verifier now assigns the subprog
id. Also, a list of kernel symbol addresses for all the JITed
functions is provided in the program info. We now use the imm
field as an index for this list to lookup a callee's symbol's
address and resolve its name.

Suggested-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Sandipan Das <redacted>
Reviewed-by: Jakub Kicinski <redacted>
---
v3:
 - Avoid using redundant pointers.
 - Fix indentation.

v2:
 - Order variables from longest to shortest.
 - Make sure that ksyms_ptr and ksyms_len are always initialized.
 - Simplify code.
---
 tools/bpf/bpftool/prog.c          | 24 ++++++++++++++++++++++++
 tools/bpf/bpftool/xlated_dumper.c | 10 +++++++++-
 tools/bpf/bpftool/xlated_dumper.h |  2 ++
 3 files changed, 35 insertions(+), 1 deletion(-)

diff --git a/tools/bpf/bpftool/prog.c b/tools/bpf/bpftool/prog.c
index 9bdfdf2d3fbe..e05ab58d39e2 100644
--- a/tools/bpf/bpftool/prog.c
+++ b/tools/bpf/bpftool/prog.c

@@ -420,7 +420,9 @@ static int do_show(int argc, char **argv)
 
 static int do_dump(int argc, char **argv)
 {
+	unsigned long *func_ksyms = NULL;
 	struct bpf_prog_info info = {};
+	unsigned int nr_func_ksyms;
 	struct dump_data dd = {};
 	__u32 len = sizeof(info);
 	unsigned int buf_size;

@@ -496,10 +498,22 @@ static int do_dump(int argc, char **argv)
 		return -1;
 	}
 
+	nr_func_ksyms = info.nr_jited_ksyms;
+	if (nr_func_ksyms) {
+		func_ksyms = malloc(nr_func_ksyms * sizeof(__u64));
+		if (!func_ksyms) {
+			p_err("mem alloc failed");
+			close(fd);
+			goto err_free;
+		}
+	}
+
 	memset(&info, 0, sizeof(info));
 
 	*member_ptr = ptr_to_u64(buf);
 	*member_len = buf_size;
+	info.jited_ksyms = ptr_to_u64(func_ksyms);
+	info.nr_jited_ksyms = nr_func_ksyms;
 
 	err = bpf_obj_get_info_by_fd(fd, &info, &len);
 	close(fd);

@@ -513,6 +527,11 @@ static int do_dump(int argc, char **argv)
 		goto err_free;
 	}
 
+	if (info.nr_jited_ksyms > nr_func_ksyms) {
+		p_err("too many addresses returned");
+		goto err_free;
+	}
+
 	if ((member_len == &info.jited_prog_len &&
 	     info.jited_prog_insns == 0) ||
 	    (member_len == &info.xlated_prog_len &&

@@ -558,6 +577,9 @@ static int do_dump(int argc, char **argv)
 			dump_xlated_cfg(buf, *member_len);
 	} else {
 		kernel_syms_load(&dd);
+		dd.nr_jited_ksyms = info.nr_jited_ksyms;
+		dd.jited_ksyms = (__u64 *) info.jited_ksyms;
+
 		if (json_output)
 			dump_xlated_json(&dd, buf, *member_len, opcodes);
 		else

@@ -566,10 +588,12 @@ static int do_dump(int argc, char **argv)
 	}
 
 	free(buf);
+	free(func_ksyms);
 	return 0;
 
 err_free:
 	free(buf);
+	free(func_ksyms);
 	return -1;
 }
 

diff --git a/tools/bpf/bpftool/xlated_dumper.c b/tools/bpf/bpftool/xlated_dumper.c
index 7a3173b76c16..efdc8fecf2bb 100644
--- a/tools/bpf/bpftool/xlated_dumper.c
+++ b/tools/bpf/bpftool/xlated_dumper.c

@@ -174,7 +174,11 @@ static const char *print_call_pcrel(struct dump_data *dd,
 				    unsigned long address,
 				    const struct bpf_insn *insn)
 {
-	if (sym)
+	if (!dd->nr_jited_ksyms)
+		/* Do not show address for interpreted programs */
+		snprintf(dd->scratch_buff, sizeof(dd->scratch_buff),
+			"%+d", insn->off);
+	else if (sym)
 		snprintf(dd->scratch_buff, sizeof(dd->scratch_buff),
 			 "%+d#%s", insn->off, sym->name);
 	else

@@ -203,6 +207,10 @@ static const char *print_call(void *private_data,
 	unsigned long address = dd->address_call_base + insn->imm;
 	struct kernel_sym *sym;
 
+	if (insn->src_reg == BPF_PSEUDO_CALL &&
+	    (__u32) insn->imm < dd->nr_jited_ksyms)
+		address = dd->jited_ksyms[insn->imm];
+
 	sym = kernel_syms_search(dd, address);
 	if (insn->src_reg == BPF_PSEUDO_CALL)
 		return print_call_pcrel(dd, sym, address, insn);

diff --git a/tools/bpf/bpftool/xlated_dumper.h b/tools/bpf/bpftool/xlated_dumper.h
index b34affa7ef2d..eafbb49c8d0b 100644
--- a/tools/bpf/bpftool/xlated_dumper.h
+++ b/tools/bpf/bpftool/xlated_dumper.h

@@ -49,6 +49,8 @@ struct dump_data {
 	unsigned long address_call_base;
 	struct kernel_sym *sym_mapping;
 	__u32 sym_count;
+	__u64 *jited_ksyms;
+	__u32 nr_jited_ksyms;
 	char scratch_buff[SYM_MAX_NAME + 8];
 };
 

-- 
2.14.3

Currently, for multi-function programs, we cannot get the JITed
instructions using the bpf system call's BPF_OBJ_GET_INFO_BY_FD
command. Because of this, userspace tools such as bpftool fail
to identify a multi-function program as being JITed or not.

With the JIT enabled and the test program running, this can be
verified as follows:

  # cat /proc/sys/net/core/bpf_jit_enable
  1

Before applying this patch:

  # bpftool prog list
  1: kprobe  name foo  tag b811aab41a39ad3d  gpl
          loaded_at 2018-05-16T11:43:38+0530  uid 0
          xlated 216B  not jited  memlock 65536B
  ...

  # bpftool prog dump jited id 1
  no instructions returned

After applying this patch:

  # bpftool prog list
  1: kprobe  name foo  tag b811aab41a39ad3d  gpl
          loaded_at 2018-05-16T12:13:01+0530  uid 0
          xlated 216B  jited 308B  memlock 65536B
  ...

  # bpftool prog dump jited id 1
     0:   nop
     4:   nop
     8:   mflr    r0
     c:   std     r0,16(r1)
    10:   stdu    r1,-112(r1)
    14:   std     r31,104(r1)
    18:   addi    r31,r1,48
    1c:   li      r3,10
  ...

Signed-off-by: Sandipan Das <redacted>
---
v4:
 - If the image allocated in userspace is not large enough,
   fill it up till the end even if the last JITed image can
   only be copied partially.
---
 kernel/bpf/syscall.c | 37 ++++++++++++++++++++++++++++++++++---
 1 file changed, 34 insertions(+), 3 deletions(-)

diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
index 068a4fc79ddb..c8e987a612b5 100644
--- a/kernel/bpf/syscall.c
+++ b/kernel/bpf/syscall.c

@@ -1970,13 +1970,44 @@ static int bpf_prog_get_info_by_fd(struct bpf_prog *prog,
 	 * for offload.
 	 */
 	ulen = info.jited_prog_len;
-	info.jited_prog_len = prog->jited_len;
+	if (prog->aux->func_cnt) {
+		u32 i;
+
+		info.jited_prog_len = 0;
+		for (i = 0; i < prog->aux->func_cnt; i++)
+			info.jited_prog_len += prog->aux->func[i]->jited_len;
+	} else {
+		info.jited_prog_len = prog->jited_len;
+	}
+
 	if (info.jited_prog_len && ulen) {
 		if (bpf_dump_raw_ok()) {
 			uinsns = u64_to_user_ptr(info.jited_prog_insns);
 			ulen = min_t(u32, info.jited_prog_len, ulen);
-			if (copy_to_user(uinsns, prog->bpf_func, ulen))
-				return -EFAULT;
+
+			/* for multi-function programs, copy the JITed
+			 * instructions for all the functions
+			 */
+			if (prog->aux->func_cnt) {
+				u32 len, free, i;
+				u8 *img;
+
+				free = ulen;
+				for (i = 0; i < prog->aux->func_cnt; i++) {
+					len = prog->aux->func[i]->jited_len;
+					len = min_t(u32, len, free);
+					img = (u8 *) prog->aux->func[i]->bpf_func;
+					if (copy_to_user(uinsns, img, len))
+						return -EFAULT;
+					uinsns += len;
+					free -= len;
+					if (!free)
+						break;
+				}
+			} else {
+				if (copy_to_user(uinsns, prog->bpf_func, ulen))
+					return -EFAULT;
+			}
 		} else {
 			info.jited_prog_insns = 0;
 		}

-- 
2.14.3

This adds new two new fields to struct bpf_prog_info. For
multi-function programs, these fields can be used to pass
a list of the JITed image lengths of each function for a
given program to userspace using the bpf system call with
the BPF_OBJ_GET_INFO_BY_FD command.

This can be used by userspace applications like bpftool
to split up the contiguous JITed dump, also obtained via
the system call, into more relatable chunks corresponding
to each function.

Signed-off-by: Sandipan Das <redacted>
---
 include/uapi/linux/bpf.h |  2 ++
 kernel/bpf/syscall.c     | 20 ++++++++++++++++++++
 2 files changed, 22 insertions(+)

diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
index 0be90965867d..344d2ddcef49 100644
--- a/include/uapi/linux/bpf.h
+++ b/include/uapi/linux/bpf.h

@@ -2206,7 +2206,9 @@ struct bpf_prog_info {
 	__u64 netns_dev;
 	__u64 netns_ino;
 	__u32 nr_jited_ksyms;
+	__u32 nr_jited_func_lens;
 	__aligned_u64 jited_ksyms;
+	__aligned_u64 jited_func_lens;
 } __attribute__((aligned(8)));
 
 struct bpf_map_info {

diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
index c8e987a612b5..788456c18617 100644
--- a/kernel/bpf/syscall.c
+++ b/kernel/bpf/syscall.c

@@ -2037,6 +2037,26 @@ static int bpf_prog_get_info_by_fd(struct bpf_prog *prog,
 		}
 	}
 
+	ulen = info.nr_jited_func_lens;
+	info.nr_jited_func_lens = prog->aux->func_cnt;
+	if (info.nr_jited_func_lens && ulen) {
+		if (bpf_dump_raw_ok()) {
+			u32 __user *user_lens;
+			u32 func_len, i;
+
+			/* copy the JITed image lengths for each function */
+			ulen = min_t(u32, info.nr_jited_func_lens, ulen);
+			user_lens = u64_to_user_ptr(info.jited_func_lens);
+			for (i = 0; i < ulen; i++) {
+				func_len = prog->aux->func[i]->jited_len;
+				if (put_user(func_len, &user_lens[i]))
+					return -EFAULT;
+			}
+		} else {
+			info.jited_func_lens = 0;
+		}
+	}
+
 done:
 	if (copy_to_user(uinfo, &info, info_len) ||
 	    put_user(info_len, &uattr->info.info_len))

-- 
2.14.3

Syncing the bpf.h uapi header with tools so that struct
bpf_prog_info has the two new fields for passing on the
JITed image lengths of each function in a multi-function
program.

Signed-off-by: Sandipan Das <redacted>
---
 tools/include/uapi/linux/bpf.h | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/tools/include/uapi/linux/bpf.h b/tools/include/uapi/linux/bpf.h
index 0be90965867d..344d2ddcef49 100644
--- a/tools/include/uapi/linux/bpf.h
+++ b/tools/include/uapi/linux/bpf.h

@@ -2206,7 +2206,9 @@ struct bpf_prog_info {
 	__u64 netns_dev;
 	__u64 netns_ino;
 	__u32 nr_jited_ksyms;
+	__u32 nr_jited_func_lens;
 	__aligned_u64 jited_ksyms;
+	__aligned_u64 jited_func_lens;
 } __attribute__((aligned(8)));
 
 struct bpf_map_info {

-- 
2.14.3

This splits up the contiguous JITed dump obtained via the bpf
system call into more relatable chunks for each function in
the program. If the kernel symbols corresponding to these are
known, they are printed in the header for each JIT image dump
otherwise the masked start address is printed.

Before applying this patch:

  # bpftool prog dump jited id 1

     0:	push   %rbp
     1:	mov    %rsp,%rbp
  ...
    70:	leaveq
    71:	retq
    72:	push   %rbp
    73:	mov    %rsp,%rbp
  ...
    dd:	leaveq
    de:	retq

  # bpftool -p prog dump jited id 1

  [{
          "pc": "0x0",
          "operation": "push",
          "operands": ["%rbp"
          ]
      },{
  ...
      },{
          "pc": "0x71",
          "operation": "retq",
          "operands": [null
          ]
      },{
          "pc": "0x72",
          "operation": "push",
          "operands": ["%rbp"
          ]
      },{
  ...
      },{
          "pc": "0xde",
          "operation": "retq",
          "operands": [null
          ]
      }
  ]

After applying this patch:

  # echo 0 > /proc/sys/net/core/bpf_jit_kallsyms
  # bpftool prog dump jited id 1

  0xffffffffc02c7000:
     0:	push   %rbp
     1:	mov    %rsp,%rbp
  ...
    70:	leaveq
    71:	retq

  0xffffffffc02cf000:
     0:	push   %rbp
     1:	mov    %rsp,%rbp
  ...
    6b:	leaveq
    6c:	retq

  # bpftool -p prog dump jited id 1

  [{
          "name": "0xffffffffc02c7000",
          "insns": [{
                  "pc": "0x0",
                  "operation": "push",
                  "operands": ["%rbp"
                  ]
              },{
  ...
              },{
                  "pc": "0x71",
                  "operation": "retq",
                  "operands": [null
                  ]
              }
          ]
      },{
          "name": "0xffffffffc02cf000",
          "insns": [{
                  "pc": "0x0",
                  "operation": "push",
                  "operands": ["%rbp"
                  ]
              },{
  ...
              },{
                  "pc": "0x6c",
                  "operation": "retq",
                  "operands": [null
                  ]
              }
          ]
      }
  ]

  # echo 1 > /proc/sys/net/core/bpf_jit_kallsyms
  # bpftool prog dump jited id 1

  bpf_prog_b811aab41a39ad3d_foo:
     0:	push   %rbp
     1:	mov    %rsp,%rbp
  ...
    70:	leaveq
    71:	retq

  bpf_prog_cf418ac8b67bebd9_F:
     0:	push   %rbp
     1:	mov    %rsp,%rbp
  ...
    6b:	leaveq
    6c:	retq

  # bpftool -p prog dump jited id 1

  [{
          "name": "bpf_prog_b811aab41a39ad3d_foo",
          "insns": [{
                  "pc": "0x0",
                  "operation": "push",
                  "operands": ["%rbp"
                  ]
              },{
  ...
              },{
                  "pc": "0x71",
                  "operation": "retq",
                  "operands": [null
                  ]
              }
          ]
      },{
          "name": "bpf_prog_cf418ac8b67bebd9_F",
          "insns": [{
                  "pc": "0x0",
                  "operation": "push",
                  "operands": ["%rbp"
                  ]
              },{
  ...
              },{
                  "pc": "0x6c",
                  "operation": "retq",
                  "operands": [null
                  ]
              }
          ]
      }
  ]

Signed-off-by: Sandipan Das <redacted>
---
v4:
 - Fix JSON output.
---
 tools/bpf/bpftool/prog.c          | 73 ++++++++++++++++++++++++++++++++++++++-
 tools/bpf/bpftool/xlated_dumper.c |  4 +--
 tools/bpf/bpftool/xlated_dumper.h |  1 +
 3 files changed, 75 insertions(+), 3 deletions(-)

diff --git a/tools/bpf/bpftool/prog.c b/tools/bpf/bpftool/prog.c
index e05ab58d39e2..39b88e760367 100644
--- a/tools/bpf/bpftool/prog.c
+++ b/tools/bpf/bpftool/prog.c

@@ -422,7 +422,9 @@ static int do_dump(int argc, char **argv)
 {
 	unsigned long *func_ksyms = NULL;
 	struct bpf_prog_info info = {};
+	unsigned int *func_lens = NULL;
 	unsigned int nr_func_ksyms;
+	unsigned int nr_func_lens;
 	struct dump_data dd = {};
 	__u32 len = sizeof(info);
 	unsigned int buf_size;

@@ -508,12 +510,24 @@ static int do_dump(int argc, char **argv)
 		}
 	}
 
+	nr_func_lens = info.nr_jited_func_lens;
+	if (nr_func_lens) {
+		func_lens = malloc(nr_func_lens * sizeof(__u32));
+		if (!func_lens) {
+			p_err("mem alloc failed");
+			close(fd);
+			goto err_free;
+		}
+	}
+
 	memset(&info, 0, sizeof(info));
 
 	*member_ptr = ptr_to_u64(buf);
 	*member_len = buf_size;
 	info.jited_ksyms = ptr_to_u64(func_ksyms);
 	info.nr_jited_ksyms = nr_func_ksyms;
+	info.jited_func_lens = ptr_to_u64(func_lens);
+	info.nr_jited_func_lens = nr_func_lens;
 
 	err = bpf_obj_get_info_by_fd(fd, &info, &len);
 	close(fd);

@@ -532,6 +546,11 @@ static int do_dump(int argc, char **argv)
 		goto err_free;
 	}
 
+	if (info.nr_jited_func_lens > nr_func_lens) {
+		p_err("too many values returned");
+		goto err_free;
+	}
+
 	if ((member_len == &info.jited_prog_len &&
 	     info.jited_prog_insns == 0) ||
 	    (member_len == &info.xlated_prog_len &&

@@ -569,7 +588,57 @@ static int do_dump(int argc, char **argv)
 				goto err_free;
 		}
 
-		disasm_print_insn(buf, *member_len, opcodes, name);
+		if (info.nr_jited_func_lens && info.jited_func_lens) {
+			struct kernel_sym *sym = NULL;
+			char sym_name[SYM_MAX_NAME];
+			unsigned char *img = buf;
+			__u64 *ksyms = NULL;
+			__u32 *lens;
+			__u32 i;
+
+			if (info.nr_jited_ksyms) {
+				kernel_syms_load(&dd);
+				ksyms = (__u64 *) info.jited_ksyms;
+			}
+
+			if (json_output)
+				jsonw_start_array(json_wtr);
+
+			lens = (__u32 *) info.jited_func_lens;
+			for (i = 0; i < info.nr_jited_func_lens; i++) {
+				if (ksyms) {
+					sym = kernel_syms_search(&dd, ksyms[i]);
+					if (sym)
+						sprintf(sym_name, "%s", sym->name);
+					else
+						sprintf(sym_name, "0x%016llx", ksyms[i]);
+				} else {
+					strcpy(sym_name, "unknown");
+				}
+
+				if (json_output) {
+					jsonw_start_object(json_wtr);
+					jsonw_name(json_wtr, "name");
+					jsonw_string(json_wtr, sym_name);
+					jsonw_name(json_wtr, "insns");
+				} else {
+					printf("%s:\n", sym_name);
+				}
+
+				disasm_print_insn(img, lens[i], opcodes, name);
+				img += lens[i];
+
+				if (json_output)
+					jsonw_end_object(json_wtr);
+				else
+					printf("\n");
+			}
+
+			if (json_output)
+				jsonw_end_array(json_wtr);
+		} else {
+			disasm_print_insn(buf, *member_len, opcodes, name);
+		}
 	} else if (visual) {
 		if (json_output)
 			jsonw_null(json_wtr);

@@ -589,11 +658,13 @@ static int do_dump(int argc, char **argv)
 
 	free(buf);
 	free(func_ksyms);
+	free(func_lens);
 	return 0;
 
 err_free:
 	free(buf);
 	free(func_ksyms);
+	free(func_lens);
 	return -1;
 }
 

diff --git a/tools/bpf/bpftool/xlated_dumper.c b/tools/bpf/bpftool/xlated_dumper.c
index efdc8fecf2bb..b97f1da60dd1 100644
--- a/tools/bpf/bpftool/xlated_dumper.c
+++ b/tools/bpf/bpftool/xlated_dumper.c

@@ -102,8 +102,8 @@ void kernel_syms_destroy(struct dump_data *dd)
 	free(dd->sym_mapping);
 }
 
-static struct kernel_sym *kernel_syms_search(struct dump_data *dd,
-					     unsigned long key)
+struct kernel_sym *kernel_syms_search(struct dump_data *dd,
+				      unsigned long key)
 {
 	struct kernel_sym sym = {
 		.address = key,

diff --git a/tools/bpf/bpftool/xlated_dumper.h b/tools/bpf/bpftool/xlated_dumper.h
index eafbb49c8d0b..33d86e2b369b 100644
--- a/tools/bpf/bpftool/xlated_dumper.h
+++ b/tools/bpf/bpftool/xlated_dumper.h

@@ -56,6 +56,7 @@ struct dump_data {
 
 void kernel_syms_load(struct dump_data *dd);
 void kernel_syms_destroy(struct dump_data *dd);
+struct kernel_sym *kernel_syms_search(struct dump_data *dd, unsigned long key);
 void dump_xlated_json(struct dump_data *dd, void *buf, unsigned int len,
 		      bool opcodes);
 void dump_xlated_plain(struct dump_data *dd, void *buf, unsigned int len,

-- 
2.14.3

On Thu, 24 May 2018 12:26:54 +0530, Sandipan Das wrote:

This splits up the contiguous JITed dump obtained via the bpf
system call into more relatable chunks for each function in
the program. If the kernel symbols corresponding to these are
known, they are printed in the header for each JIT image dump
otherwise the masked start address is printed.

...
 

Signed-off-by: Sandipan Das <redacted>

Reviewed-by: Jakub Kicinski <redacted>

Thank you!

On 05/24/2018 08:56 AM, Sandipan Das wrote:

[1] Support for bpf-to-bpf function calls in the powerpc64 JIT compiler.

[2] Provide a way for resolving function calls because of the way JITed
    images are allocated in powerpc64.

[3] Fix to get JITed instruction dumps for multi-function programs from
    the bpf system call.

[4] Fix for bpftool to show delimited multi-function JITed image dumps.

v4:
 - Incorporate review comments from Jakub.
 - Fix JSON output for bpftool.

v3:
 - Change base tree tag to bpf-next.
 - Incorporate review comments from Alexei, Daniel and Jakub.
 - Make sure that the JITed image does not grow or shrink after
   the last pass due to the way the instruction sequence used
   to load a callee's address maybe optimized.
 - Make additional changes to the bpf system call and bpftool to
   make multi-function JITed dumps easier to correlate.

v2:
 - Incorporate review comments from Jakub.

Sandipan Das (10):
  bpf: support 64-bit offsets for bpf function calls
  bpf: powerpc64: pad function address loads with NOPs
  bpf: powerpc64: add JIT support for multi-function programs
  bpf: get kernel symbol addresses via syscall
  tools: bpf: sync bpf uapi header
  tools: bpftool: resolve calls without using imm field
  bpf: fix multi-function JITed dump obtained via syscall
  bpf: get JITed image lengths of functions via syscall
  tools: bpf: sync bpf uapi header
  tools: bpftool: add delimiters to multi-function JITed dumps

 arch/powerpc/net/bpf_jit_comp64.c | 110 ++++++++++++++++++++++++++++++--------
 include/uapi/linux/bpf.h          |   4 ++
 kernel/bpf/syscall.c              |  82 ++++++++++++++++++++++++++--
 kernel/bpf/verifier.c             |  22 +++++---
 tools/bpf/bpftool/prog.c          |  97 ++++++++++++++++++++++++++++++++-
 tools/bpf/bpftool/xlated_dumper.c |  14 +++--
 tools/bpf/bpftool/xlated_dumper.h |   3 ++
 tools/include/uapi/linux/bpf.h    |   4 ++
 8 files changed, 301 insertions(+), 35 deletions(-)

Applied to bpf-next, thanks a lot Sandipan!

On 05/24/2018 08:56 AM, Sandipan Das wrote:

By the way, I think you can still optimize this. The nops are not really
needed in case of insn->src_reg != BPF_PSEUDO_CALL since the address of
a normal BPF helper call will always be at a fixed location and known a
priori.

 #ifdef PPC64_ELF_ABI_v1
-	/* func points to the function descriptor */
-	PPC_LI64(b2p[TMP_REG_2], func);
-	/* Load actual entry point from function descriptor */
-	PPC_BPF_LL(b2p[TMP_REG_1], b2p[TMP_REG_2], 0);
-	/* ... and move it to LR */
-	PPC_MTLR(b2p[TMP_REG_1]);
 	/*
 	 * Load TOC from function descriptor at offset 8.
 	 * We can clobber r2 since we get called through a
 	 * function pointer (so caller will save/restore r2)
 	 * and since we don't use a TOC ourself.
 	 */
-	PPC_BPF_LL(2, b2p[TMP_REG_2], 8);
-#else
-	/* We can clobber r12 */
-	PPC_FUNC_ADDR(12, func);
-	PPC_MTLR(12);
+	PPC_BPF_LL(2, 12, 8);
+	/* Load actual entry point from function descriptor */
+	PPC_BPF_LL(12, 12, 0);
 #endif
+
+	PPC_MTLR(12);
 	PPC_BLRL();
 }
 

On 05/24/2018 01:04 PM, Daniel Borkmann wrote:

On 05/24/2018 08:56 AM, Sandipan Das wrote:

quoted

For multi-function programs, loading the address of a callee
function to a register requires emitting instructions whose
count varies from one to five depending on the nature of the
address.

Since we come to know of the callee's address only before the
extra pass, the number of instructions required to load this
address may vary from what was previously generated. This can
make the JITed image grow or shrink.

To avoid this, we should generate a constant five-instruction
when loading function addresses by padding the optimized load
sequence with NOPs.

Signed-off-by: Sandipan Das <redacted>
---
 arch/powerpc/net/bpf_jit_comp64.c | 34 +++++++++++++++++++++++-----------
 1 file changed, 23 insertions(+), 11 deletions(-)

diff --git a/arch/powerpc/net/bpf_jit_comp64.c b/arch/powerpc/net/bpf_jit_comp64.c
index 1bdb1aff0619..e4582744a31d 100644
--- a/arch/powerpc/net/bpf_jit_comp64.c
+++ b/arch/powerpc/net/bpf_jit_comp64.c

@@ -167,25 +167,37 @@ static void bpf_jit_build_epilogue(u32 *image, struct codegen_context *ctx)
 
 static void bpf_jit_emit_func_call(u32 *image, struct codegen_context *ctx, u64 func)
 {
+	unsigned int i, ctx_idx = ctx->idx;
+
+	/* Load function address into r12 */
+	PPC_LI64(12, func);
+
+	/* For bpf-to-bpf function calls, the callee's address is unknown
+	 * until the last extra pass. As seen above, we use PPC_LI64() to
+	 * load the callee's address, but this may optimize the number of
+	 * instructions required based on the nature of the address.
+	 *
+	 * Since we don't want the number of instructions emitted to change,
+	 * we pad the optimized PPC_LI64() call with NOPs to guarantee that
+	 * we always have a five-instruction sequence, which is the maximum
+	 * that PPC_LI64() can emit.
+	 */
+	for (i = ctx->idx - ctx_idx; i < 5; i++)
+		PPC_NOP();

By the way, I think you can still optimize this. The nops are not really
needed in case of insn->src_reg != BPF_PSEUDO_CALL since the address of
a normal BPF helper call will always be at a fixed location and known a
priori.

Ah, true. Thanks for pointing this out. There are a few other things that
we are planning to do for the ppc64 JIT compiler. Will put out a patch for
this with that series.

- Sandipan

quoted

 #ifdef PPC64_ELF_ABI_v1
-	/* func points to the function descriptor */
-	PPC_LI64(b2p[TMP_REG_2], func);
-	/* Load actual entry point from function descriptor */
-	PPC_BPF_LL(b2p[TMP_REG_1], b2p[TMP_REG_2], 0);
-	/* ... and move it to LR */
-	PPC_MTLR(b2p[TMP_REG_1]);
 	/*
 	 * Load TOC from function descriptor at offset 8.
 	 * We can clobber r2 since we get called through a
 	 * function pointer (so caller will save/restore r2)
 	 * and since we don't use a TOC ourself.
 	 */
-	PPC_BPF_LL(2, b2p[TMP_REG_2], 8);
-#else
-	/* We can clobber r12 */
-	PPC_FUNC_ADDR(12, func);
-	PPC_MTLR(12);
+	PPC_BPF_LL(2, 12, 8);
+	/* Load actual entry point from function descriptor */
+	PPC_BPF_LL(12, 12, 0);
 #endif
+
+	PPC_MTLR(12);
 	PPC_BLRL();
 }
 

On 05/24/2018 10:25 AM, Sandipan Das wrote:

On 05/24/2018 01:04 PM, Daniel Borkmann wrote:

quoted

On 05/24/2018 08:56 AM, Sandipan Das wrote:

quoted

For multi-function programs, loading the address of a callee
function to a register requires emitting instructions whose
count varies from one to five depending on the nature of the
address.

Since we come to know of the callee's address only before the
extra pass, the number of instructions required to load this
address may vary from what was previously generated. This can
make the JITed image grow or shrink.

To avoid this, we should generate a constant five-instruction
when loading function addresses by padding the optimized load
sequence with NOPs.

Signed-off-by: Sandipan Das <redacted>
---
 arch/powerpc/net/bpf_jit_comp64.c | 34 +++++++++++++++++++++++-----------
 1 file changed, 23 insertions(+), 11 deletions(-)

diff --git a/arch/powerpc/net/bpf_jit_comp64.c b/arch/powerpc/net/bpf_jit_comp64.c
index 1bdb1aff0619..e4582744a31d 100644
--- a/arch/powerpc/net/bpf_jit_comp64.c
+++ b/arch/powerpc/net/bpf_jit_comp64.c

@@ -167,25 +167,37 @@ static void bpf_jit_build_epilogue(u32 *image, struct codegen_context *ctx)
 
 static void bpf_jit_emit_func_call(u32 *image, struct codegen_context *ctx, u64 func)
 {
+	unsigned int i, ctx_idx = ctx->idx;
+
+	/* Load function address into r12 */
+	PPC_LI64(12, func);
+
+	/* For bpf-to-bpf function calls, the callee's address is unknown
+	 * until the last extra pass. As seen above, we use PPC_LI64() to
+	 * load the callee's address, but this may optimize the number of
+	 * instructions required based on the nature of the address.
+	 *
+	 * Since we don't want the number of instructions emitted to change,
+	 * we pad the optimized PPC_LI64() call with NOPs to guarantee that
+	 * we always have a five-instruction sequence, which is the maximum
+	 * that PPC_LI64() can emit.
+	 */
+	for (i = ctx->idx - ctx_idx; i < 5; i++)
+		PPC_NOP();

By the way, I think you can still optimize this. The nops are not really
needed in case of insn->src_reg != BPF_PSEUDO_CALL since the address of
a normal BPF helper call will always be at a fixed location and known a
priori.

Ah, true. Thanks for pointing this out. There are a few other things that
we are planning to do for the ppc64 JIT compiler. Will put out a patch for
this with that series.

Awesome, thanks Sandipan!