Skip to contentSkip to conversation

powerpc64/bpf: various fixes

PATCH Applied Hari Bathini 2 participants 8 messages Mar 3 – Mar 11

Subsystems (4)
lore
Share & tools
lore

Hari Bathini

author maintainer
#0 lore raw

v4 cover powerpc64/bpf: various fixes

List of fixes for powerpc64 BPF JIT:
- Patches 1, 2 & 6 fix issues with existing powerpc64 JIT and are tagged for -stable.
- Patches 3 & 4 resolve issues with bpf2bpf tailcall support added recently.
- Patch 5 resolves an issue reported by [email protected] while adding exceptions support on powerpc64 BPF JIT.

Changes since v3:
- Fixed a minor issue in zero_extend() helper in patch 6 as pointed out by [email protected].

Changes since v2:
- Added Tested-by tag from Venkat on patches 1, 2 & 4.
- Patches 6 added to fix kfunc call ABI issue on powerpc.
- Fixed a few comments on Abhishek's suggestion.

Changes since v1:
- Addressed couple of review comments from [email protected] in patch 3 & 5.

Hari Bathini (6):
powerpc64/bpf: do not increment tailcall count when prog is NULL powerpc64/bpf: fix the address returned by bpf_get_func_ip powerpc64/bpf: use consistent tailcall offset in trampoline powerpc64/bpf: remove BPF redzone protection in trampoline stack powerpc64/bpf: fix handling of BPF stack in exception callback powerpc64/bpf: fix kfunc call support

 arch/powerpc/net/bpf_jit.h        |   5 -
 arch/powerpc/net/bpf_jit_comp.c   | 127 +++++++++------------
 arch/powerpc/net/bpf_jit_comp64.c | 181 +++++++++++++++++++++++-------
 3 files changed, 199 insertions(+), 114 deletions(-)
Signature · 1 line
2.53.0
author maintainer
#1 lore raw

PATCH v4 1/6 powerpc64/bpf: do not increment tailcall count when prog is NULL

Do not increment tailcall count, if tailcall did not succeed due to missing BPF program.

Fixes ce0761419fae ("powerpc/bpf: Implement support for tail calls")

---

* No changes in v4.

Changes in v3:
- Added Tested-by tag from Venkat.

 arch/powerpc/net/bpf_jit_comp64.c | 39 +++++++++++++++++--------------
 1 file changed, 21 insertions(+), 18 deletions(-)
arch/powerpc/net/bpf_jit_comp64.c +21 -18
index b1a3945ccc9f..44ce8a8783f9 100644
@@ -522,9 +522,30 @@ static int bpf_jit_emit_tail_call(u32 *image, struct codegen_context *ctx, u32 o
 
 	/*
 	 * tail_call_info++; <- Actual value of tcc here
+	 * Writeback this updated value only if tailcall succeeds.
 	 */
 	EMIT(PPC_RAW_ADDI(bpf_to_ppc(TMP_REG_1), bpf_to_ppc(TMP_REG_1), 1));
 
+	/* prog = array->ptrs[index]; */
+	EMIT(PPC_RAW_MULI(bpf_to_ppc(TMP_REG_2), b2p_index, 8));
+	EMIT(PPC_RAW_ADD(bpf_to_ppc(TMP_REG_2), bpf_to_ppc(TMP_REG_2), b2p_bpf_array));
+	EMIT(PPC_RAW_LD(bpf_to_ppc(TMP_REG_2), bpf_to_ppc(TMP_REG_2),
+			offsetof(struct bpf_array, ptrs)));
+
+	/*
+	 * if (prog == NULL)
+	 *   goto out;
+	 */
+	EMIT(PPC_RAW_CMPLDI(bpf_to_ppc(TMP_REG_2), 0));
+	PPC_BCC_SHORT(COND_EQ, out);
+
+	/* goto *(prog->bpf_func + prologue_size); */
+	EMIT(PPC_RAW_LD(bpf_to_ppc(TMP_REG_2), bpf_to_ppc(TMP_REG_2),
+			offsetof(struct bpf_prog, bpf_func)));
+	EMIT(PPC_RAW_ADDI(bpf_to_ppc(TMP_REG_2), bpf_to_ppc(TMP_REG_2),
+			  FUNCTION_DESCR_SIZE + bpf_tailcall_prologue_size));
+	EMIT(PPC_RAW_MTCTR(bpf_to_ppc(TMP_REG_2)));
+
 	/*
 	 * Before writing updated tail_call_info, distinguish if current frame
 	 * is storing a reference to tail_call_info or actual tcc value in
@@ -539,24 +560,6 @@ static int bpf_jit_emit_tail_call(u32 *image, struct codegen_context *ctx, u32 o
 	/* Writeback updated value to tail_call_info */
 	EMIT(PPC_RAW_STD(bpf_to_ppc(TMP_REG_1), bpf_to_ppc(TMP_REG_2), 0));
 
-	/* prog = array->ptrs[index]; */
-	EMIT(PPC_RAW_MULI(bpf_to_ppc(TMP_REG_1), b2p_index, 8));
-	EMIT(PPC_RAW_ADD(bpf_to_ppc(TMP_REG_1), bpf_to_ppc(TMP_REG_1), b2p_bpf_array));
-	EMIT(PPC_RAW_LD(bpf_to_ppc(TMP_REG_1), bpf_to_ppc(TMP_REG_1), offsetof(struct bpf_array, ptrs)));
-
-	/*
-	 * if (prog == NULL)
-	 *   goto out;
-	 */
-	EMIT(PPC_RAW_CMPLDI(bpf_to_ppc(TMP_REG_1), 0));
-	PPC_BCC_SHORT(COND_EQ, out);
-
-	/* goto *(prog->bpf_func + prologue_size); */
-	EMIT(PPC_RAW_LD(bpf_to_ppc(TMP_REG_1), bpf_to_ppc(TMP_REG_1), offsetof(struct bpf_prog, bpf_func)));
-	EMIT(PPC_RAW_ADDI(bpf_to_ppc(TMP_REG_1), bpf_to_ppc(TMP_REG_1),
-			FUNCTION_DESCR_SIZE + bpf_tailcall_prologue_size));
-	EMIT(PPC_RAW_MTCTR(bpf_to_ppc(TMP_REG_1)));
-
 	/* tear down stack, restore NVRs, ... */
 	bpf_jit_emit_common_epilogue(image, ctx);
 
Signature · 1 line
2.53.0
author maintainer
#2 lore raw

PATCH v4 2/6 powerpc64/bpf: fix the address returned by bpf_get_func_ip

In reply to Hari Bathini

bpf_get_func_ip() helper function returns the address of the traced function. It relies on the IP address stored at ctx - 16 by the bpf trampoline. On 64-bit powerpc, this address is recovered from LR accounting for OOL trampoline. But the address stored here was off by 4-bytes. Ensure the address is the actual start of the traced function.

Fixes d243b62b7bd3 ("powerpc64/bpf: Add support for bpf trampolines")

---

* No changes in v4.

Changes in v3:
- Added Tested-by tag from Venkat.
- Updated comments based on suggestions from Abhishek.

 arch/powerpc/net/bpf_jit_comp.c | 28 +++++++++++++++++++---------
 1 file changed, 19 insertions(+), 9 deletions(-)
arch/powerpc/net/bpf_jit_comp.c +19 -9
index 52162e4a7f84..95f208229b09 100644
@@ -785,9 +785,9 @@ static int __arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *rw_im
 	 *       retval_off             [ return value      ]
 	 *                              [ reg argN          ]
 	 *                              [ ...               ]
-	 *       regs_off               [ reg_arg1          ] prog ctx context
-	 *       nregs_off              [ args count        ]
-	 *       ip_off                 [ traced function   ]
+	 *       regs_off               [ reg_arg1          ] prog_ctx
+	 *       nregs_off              [ args count        ] ((u64 *)prog_ctx)[-1]
+	 *       ip_off                 [ traced function   ] ((u64 *)prog_ctx)[-2]
 	 *                              [ ...               ]
 	 *       run_ctx_off            [ bpf_tramp_run_ctx ]
 	 *                              [ reg argN          ]
@@ -895,7 +895,7 @@ static int __arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *rw_im
 
 	bpf_trampoline_save_args(image, ctx, func_frame_offset, nr_regs, regs_off);
 
-	/* Save our return address */
+	/* Save our LR/return address */
 	EMIT(PPC_RAW_MFLR(_R3));
 	if (IS_ENABLED(CONFIG_PPC_FTRACE_OUT_OF_LINE))
 		EMIT(PPC_RAW_STL(_R3, _R1, alt_lr_off));
@@ -903,24 +903,34 @@ static int __arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *rw_im
 		EMIT(PPC_RAW_STL(_R3, _R1, bpf_frame_size + PPC_LR_STKOFF));
 
 	/*
-	 * Save ip address of the traced function.
-	 * We could recover this from LR, but we will need to address for OOL trampoline,
-	 * and optional GEP area.
+	 * Derive IP address of the traced function.
+	 * In case of CONFIG_PPC_FTRACE_OUT_OF_LINE or BPF program, LR points to the instruction
+	 * after the 'bl' instruction in the OOL stub. Refer to ftrace_init_ool_stub() and
+	 * bpf_arch_text_poke() for OOL stub of kernel functions and bpf programs respectively.
+	 * Relevant stub sequence:
+	 *
+	 *               bl <tramp>
+	 *   LR (R3) =>  mtlr r0
+	 *               b <func_addr+4>
+	 *
+	 * Recover kernel function/bpf program address from the unconditional
+	 * branch instruction at the end of OOL stub.
 	 */
 	if (IS_ENABLED(CONFIG_PPC_FTRACE_OUT_OF_LINE) || flags & BPF_TRAMP_F_IP_ARG) {
 		EMIT(PPC_RAW_LWZ(_R4, _R3, 4));
 		EMIT(PPC_RAW_SLWI(_R4, _R4, 6));
 		EMIT(PPC_RAW_SRAWI(_R4, _R4, 6));
 		EMIT(PPC_RAW_ADD(_R3, _R3, _R4));
-		EMIT(PPC_RAW_ADDI(_R3, _R3, 4));
 	}
 
 	if (flags & BPF_TRAMP_F_IP_ARG)
 		EMIT(PPC_RAW_STL(_R3, _R1, ip_off));
 
-	if (IS_ENABLED(CONFIG_PPC_FTRACE_OUT_OF_LINE))
+	if (IS_ENABLED(CONFIG_PPC_FTRACE_OUT_OF_LINE)) {
 		/* Fake our LR for unwind */
+		EMIT(PPC_RAW_ADDI(_R3, _R3, 4));
 		EMIT(PPC_RAW_STL(_R3, _R1, bpf_frame_size + PPC_LR_STKOFF));
+	}
 
 	/* Save function arg count -- see bpf_get_func_arg_cnt() */
 	EMIT(PPC_RAW_LI(_R3, nr_regs));
Signature · 1 line
2.53.0
author maintainer
#3 lore raw

PATCH v4 3/6 powerpc64/bpf: use consistent tailcall offset in trampoline

In reply to Hari Bathini

Ideally, the offset used to load the tail call info field and to find the pass by reference address for tail call field should be the same. But while setting up the tail call info in the trampoline, this was not followed. This can be misleading and can lead to unpredictable results if and when bpf_has_stack_frame() ends up returning true for trampoline frame. Since commit 15513beeb673 ("powerpc64/bpf:
Moving tail_call_cnt to bottom of frame") and commit 2ed2d8f6fb38 ("powerpc64/bpf: Support tailcalls with subprogs") ensured tail call field is at the bottom of the stack frame for BPF programs as well as BPF trampoline, avoid relying on bpf_jit_stack_tailcallinfo_offset() and bpf_has_stack_frame() for trampoline frame and always calculate tail call field offset with reference to older frame.

Fixes 2ed2d8f6fb38 ("powerpc64/bpf: Support tailcalls with subprogs")

---

* No changes in v4.

Changes in v3:
- BPF_PPC_STACK_SAVE macro definition adjusted as suggested by Abhishek.

 arch/powerpc/net/bpf_jit.h        |  5 -----
 arch/powerpc/net/bpf_jit_comp.c   | 12 +++++-------
 arch/powerpc/net/bpf_jit_comp64.c |  5 ++++-
 3 files changed, 9 insertions(+), 13 deletions(-)
arch/powerpc/net/bpf_jit.h +0 -5
index 82bbf63f0e57..7354e1d72f79 100644
@@ -81,9 +81,6 @@
 
 #ifdef CONFIG_PPC64
 
-/* for gpr non volatile registers BPG_REG_6 to 10 */
-#define BPF_PPC_STACK_SAVE	(6 * 8)
-
 /* If dummy pass (!image), account for maximum possible instructions */
 #define PPC_LI64(d, i)		do {					      \
 	if (!image)							      \
@@ -219,8 +216,6 @@ int bpf_jit_emit_exit_insn(u32 *image, struct codegen_context *ctx, int tmp_reg,
 int bpf_add_extable_entry(struct bpf_prog *fp, u32 *image, u32 *fimage, int pass,
 			  struct codegen_context *ctx, int insn_idx,
 			  int jmp_off, int dst_reg, u32 code);
-
-int bpf_jit_stack_tailcallinfo_offset(struct codegen_context *ctx);
 #endif
 
 #endif
arch/powerpc/net/bpf_jit_comp.c +5 -7
index 95f208229b09..a4aa1e4c9f1c 100644
@@ -642,15 +642,13 @@ static void bpf_trampoline_setup_tail_call_info(u32 *image, struct codegen_conte
 						int bpf_dummy_frame_size, int r4_off)
 {
 	if (IS_ENABLED(CONFIG_PPC64)) {
-		/* See Generated stack layout */
-		int tailcallinfo_offset = BPF_PPC_TAILCALL;
-
 		/*
 		 * func_frame_offset =                                   ...(1)
 		 *      bpf_dummy_frame_size + trampoline_frame_size
 		 */
 		EMIT(PPC_RAW_LD(_R4, _R1, func_frame_offset));
-		EMIT(PPC_RAW_LD(_R3, _R4, -tailcallinfo_offset));
+		/* Refer to trampoline's Generated stack layout */
+		EMIT(PPC_RAW_LD(_R3, _R4, -BPF_PPC_TAILCALL));
 
 		/*
 		 * Setting the tail_call_info in trampoline's frame
@@ -658,7 +656,7 @@ static void bpf_trampoline_setup_tail_call_info(u32 *image, struct codegen_conte
 		 */
 		EMIT(PPC_RAW_CMPLWI(_R3, MAX_TAIL_CALL_CNT));
 		PPC_BCC_CONST_SHORT(COND_GT, 8);
-		EMIT(PPC_RAW_ADDI(_R3, _R4, bpf_jit_stack_tailcallinfo_offset(ctx)));
+		EMIT(PPC_RAW_ADDI(_R3, _R4, -BPF_PPC_TAILCALL));
 		/*
 		 * From ...(1) above:
 		 * trampoline_frame_bottom =                            ...(2)
@@ -666,14 +664,14 @@ static void bpf_trampoline_setup_tail_call_info(u32 *image, struct codegen_conte
 		 *
 		 * Using ...(2) derived above:
 		 * trampoline_tail_call_info_offset =                  ...(3)
-		 *      trampoline_frame_bottom - tailcallinfo_offset
+		 *      trampoline_frame_bottom - BPF_PPC_TAILCALL
 		 *
 		 * From ...(3):
 		 * Use trampoline_tail_call_info_offset to write reference of main's
 		 * tail_call_info in trampoline frame.
 		 */
 		EMIT(PPC_RAW_STL(_R3, _R1, (func_frame_offset - bpf_dummy_frame_size)
-								- tailcallinfo_offset));
+								- BPF_PPC_TAILCALL));
 	} else {
 		/* See bpf_jit_stack_offsetof() and BPF_PPC_TC */
 		EMIT(PPC_RAW_LL(_R4, _R1, r4_off));
arch/powerpc/net/bpf_jit_comp64.c +4 -1
index 44ce8a8783f9..7655be76f537 100644
@@ -42,6 +42,9 @@
  * exception boundary.
  */
 
+/* BPF non-volatile registers save area size */
+#define BPF_PPC_STACK_SAVE	(6 * 8)
+
 /* for bpf JIT code internal usage */
 #define BPF_PPC_STACK_LOCALS	24
 /*
@@ -148,7 +151,7 @@ static int bpf_jit_stack_local(struct codegen_context *ctx)
 	}
 }
 
-int bpf_jit_stack_tailcallinfo_offset(struct codegen_context *ctx)
+static int bpf_jit_stack_tailcallinfo_offset(struct codegen_context *ctx)
 {
 	return bpf_jit_stack_local(ctx) + BPF_PPC_STACK_LOCALS + BPF_PPC_STACK_SAVE;
 }
Signature · 1 line
2.53.0
author maintainer
#4 lore raw

PATCH v4 4/6 powerpc64/bpf: remove BPF redzone protection in trampoline stack

In reply to Hari Bathini

Since bpf2bpf tailcall support is enabled for 64-bit powerpc with kernel commit 2ed2d8f6fb38 ("powerpc64/bpf: Support tailcalls with subprogs"), 'tailcalls/tailcall_bpf2bpf_hierarchy_fexit' BPF selftest is triggering "corrupted stack end detected inside scheduler" with the config option CONFIG_SCHED_STACK_END_CHECK enabled. While reviewing the stack layout for BPF trampoline, observed that the dummy frame is trying to protect the redzone of BPF program. This is because tail call info and NVRs save area are in redzone at the time of tailcall as the current BPF program stack frame is teared down before the tailcall. But saving this redzone in the dummy frame of trampoline is unnecessary because of the follow reasons:

1) Firstly, trampoline can be attached to BPF entry/main program or subprog. But prologue part of the BPF entry/main program, where the trampoline attachpoint is, is skipped during tailcall. So, protecting the redzone does not arise when the trampoline is not even triggered in this scenario.
2) In case of subprog, the caller's stackframe is already setup and the subprog's stackframe is yet to be setup. So, nothing on the redzone to be protected.

Also, using dummy frame in BPF trampoline, wastes critically scarce kernel stack space, especially in tailcall sequence, for marginal benefit in stack unwinding. So, drop setting up the dummy frame. Instead, save return address in bpf trampoline frame and use it as appropriate. Pruning this unnecessary stack usage mitigates the likelihood of stack overflow in scenarios where bpf2bpf tailcalls and fexit programs are mixed.

Fixes 2ed2d8f6fb38 ("powerpc64/bpf: Support tailcalls with subprogs")

---

* No changes in v4.

Changes in v3:
- Added Tested-by tag from Venkat.

 arch/powerpc/net/bpf_jit_comp.c | 89 ++++++++++++---------------------
 1 file changed, 33 insertions(+), 56 deletions(-)
arch/powerpc/net/bpf_jit_comp.c +33 -56
index a4aa1e4c9f1c..85457bcb2040 100644
@@ -638,15 +638,10 @@ static int invoke_bpf_mod_ret(u32 *image, u32 *ro_image, struct codegen_context
  * for the traced function (BPF subprog/callee) to fetch it.
  */
 static void bpf_trampoline_setup_tail_call_info(u32 *image, struct codegen_context *ctx,
-						int func_frame_offset,
-						int bpf_dummy_frame_size, int r4_off)
+						int bpf_frame_size, int r4_off)
 {
 	if (IS_ENABLED(CONFIG_PPC64)) {
-		/*
-		 * func_frame_offset =                                   ...(1)
-		 *      bpf_dummy_frame_size + trampoline_frame_size
-		 */
-		EMIT(PPC_RAW_LD(_R4, _R1, func_frame_offset));
+		EMIT(PPC_RAW_LD(_R4, _R1, bpf_frame_size));
 		/* Refer to trampoline's Generated stack layout */
 		EMIT(PPC_RAW_LD(_R3, _R4, -BPF_PPC_TAILCALL));
 
@@ -657,21 +652,13 @@ static void bpf_trampoline_setup_tail_call_info(u32 *image, struct codegen_conte
 		EMIT(PPC_RAW_CMPLWI(_R3, MAX_TAIL_CALL_CNT));
 		PPC_BCC_CONST_SHORT(COND_GT, 8);
 		EMIT(PPC_RAW_ADDI(_R3, _R4, -BPF_PPC_TAILCALL));
+
 		/*
-		 * From ...(1) above:
-		 * trampoline_frame_bottom =                            ...(2)
-		 *      func_frame_offset - bpf_dummy_frame_size
-		 *
-		 * Using ...(2) derived above:
-		 * trampoline_tail_call_info_offset =                  ...(3)
-		 *      trampoline_frame_bottom - BPF_PPC_TAILCALL
-		 *
-		 * From ...(3):
-		 * Use trampoline_tail_call_info_offset to write reference of main's
-		 * tail_call_info in trampoline frame.
+		 * Trampoline's tail_call_info is at the same offset, as that of
+		 * any bpf program, with reference to previous frame. Update the
+		 * address of main's tail_call_info in trampoline frame.
 		 */
-		EMIT(PPC_RAW_STL(_R3, _R1, (func_frame_offset - bpf_dummy_frame_size)
-								- BPF_PPC_TAILCALL));
+		EMIT(PPC_RAW_STL(_R3, _R1, bpf_frame_size - BPF_PPC_TAILCALL));
 	} else {
 		/* See bpf_jit_stack_offsetof() and BPF_PPC_TC */
 		EMIT(PPC_RAW_LL(_R4, _R1, r4_off));
@@ -679,7 +666,7 @@ static void bpf_trampoline_setup_tail_call_info(u32 *image, struct codegen_conte
 }
 
 static void bpf_trampoline_restore_tail_call_cnt(u32 *image, struct codegen_context *ctx,
-						 int func_frame_offset, int r4_off)
+						 int bpf_frame_size, int r4_off)
 {
 	if (IS_ENABLED(CONFIG_PPC32)) {
 		/*
@@ -690,12 +677,12 @@ static void bpf_trampoline_restore_tail_call_cnt(u32 *image, struct codegen_cont
 	}
 }
 
-static void bpf_trampoline_save_args(u32 *image, struct codegen_context *ctx, int func_frame_offset,
-				     int nr_regs, int regs_off)
+static void bpf_trampoline_save_args(u32 *image, struct codegen_context *ctx,
+				     int bpf_frame_size, int nr_regs, int regs_off)
 {
 	int param_save_area_offset;
 
-	param_save_area_offset = func_frame_offset; /* the two frames we alloted */
+	param_save_area_offset = bpf_frame_size;
 	param_save_area_offset += STACK_FRAME_MIN_SIZE; /* param save area is past frame header */
 
 	for (int i = 0; i < nr_regs; i++) {
@@ -718,11 +705,11 @@ static void bpf_trampoline_restore_args_regs(u32 *image, struct codegen_context
 
 /* Used when we call into the traced function. Replicate parameter save area */
 static void bpf_trampoline_restore_args_stack(u32 *image, struct codegen_context *ctx,
-					      int func_frame_offset, int nr_regs, int regs_off)
+					      int bpf_frame_size, int nr_regs, int regs_off)
 {
 	int param_save_area_offset;
 
-	param_save_area_offset = func_frame_offset; /* the two frames we alloted */
+	param_save_area_offset = bpf_frame_size;
 	param_save_area_offset += STACK_FRAME_MIN_SIZE; /* param save area is past frame header */
 
 	for (int i = 8; i < nr_regs; i++) {
@@ -739,10 +726,10 @@ static int __arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *rw_im
 					 void *func_addr)
 {
 	int regs_off, nregs_off, ip_off, run_ctx_off, retval_off, nvr_off, alt_lr_off, r4_off = 0;
-	int i, ret, nr_regs, bpf_frame_size = 0, bpf_dummy_frame_size = 0, func_frame_offset;
 	struct bpf_tramp_links *fmod_ret = &tlinks[BPF_TRAMP_MODIFY_RETURN];
 	struct bpf_tramp_links *fentry = &tlinks[BPF_TRAMP_FENTRY];
 	struct bpf_tramp_links *fexit = &tlinks[BPF_TRAMP_FEXIT];
+	int i, ret, nr_regs, retaddr_off, bpf_frame_size = 0;
 	struct codegen_context codegen_ctx, *ctx;
 	u32 *image = (u32 *)rw_image;
 	ppc_inst_t branch_insn;
@@ -768,16 +755,11 @@ static int __arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *rw_im
 	 * Generated stack layout:
 	 *
 	 * func prev back chain         [ back chain        ]
-	 *                              [                   ]
-	 * bpf prog redzone/tailcallcnt [ ...               ] 64 bytes (64-bit powerpc)
-	 *                              [                   ] --
-	 * LR save area                 [ r0 save (64-bit)  ]   | header
-	 *                              [ r0 save (32-bit)  ]   |
-	 * dummy frame for unwind       [ back chain 1      ] --
 	 *                              [ tail_call_info    ] optional - 64-bit powerpc
 	 *                              [ padding           ] align stack frame
 	 *       r4_off                 [ r4 (tailcallcnt)  ] optional - 32-bit powerpc
 	 *       alt_lr_off             [ real lr (ool stub)] optional - actual lr
+	 *       retaddr_off            [ return address    ]
 	 *                              [ r26               ]
 	 *       nvr_off                [ r25               ] nvr save area
 	 *       retval_off             [ return value      ]
@@ -841,6 +823,10 @@ static int __arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *rw_im
 	nvr_off = bpf_frame_size;
 	bpf_frame_size += 2 * SZL;
 
+	/* Save area for return address */
+	retaddr_off = bpf_frame_size;
+	bpf_frame_size += SZL;
+
 	/* Optional save area for actual LR in case of ool ftrace */
 	if (IS_ENABLED(CONFIG_PPC_FTRACE_OUT_OF_LINE)) {
 		alt_lr_off = bpf_frame_size;
@@ -867,16 +853,8 @@ static int __arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *rw_im
 	/* Padding to align stack frame, if any */
 	bpf_frame_size = round_up(bpf_frame_size, SZL * 2);
 
-	/* Dummy frame size for proper unwind - includes 64-bytes red zone for 64-bit powerpc */
-	bpf_dummy_frame_size = STACK_FRAME_MIN_SIZE + 64;
-
-	/* Offset to the traced function's stack frame */
-	func_frame_offset = bpf_dummy_frame_size + bpf_frame_size;
-
-	/* Create dummy frame for unwind, store original return value */
+	/*  Store original return value */
 	EMIT(PPC_RAW_STL(_R0, _R1, PPC_LR_STKOFF));
-	/* Protect red zone where tail call count goes */
-	EMIT(PPC_RAW_STLU(_R1, _R1, -bpf_dummy_frame_size));
 
 	/* Create our stack frame */
 	EMIT(PPC_RAW_STLU(_R1, _R1, -bpf_frame_size));
@@ -891,14 +869,14 @@ static int __arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *rw_im
 	if (IS_ENABLED(CONFIG_PPC32) && nr_regs < 2)
 		EMIT(PPC_RAW_STL(_R4, _R1, r4_off));
 
-	bpf_trampoline_save_args(image, ctx, func_frame_offset, nr_regs, regs_off);
+	bpf_trampoline_save_args(image, ctx, bpf_frame_size, nr_regs, regs_off);
 
 	/* Save our LR/return address */
 	EMIT(PPC_RAW_MFLR(_R3));
 	if (IS_ENABLED(CONFIG_PPC_FTRACE_OUT_OF_LINE))
 		EMIT(PPC_RAW_STL(_R3, _R1, alt_lr_off));
 	else
-		EMIT(PPC_RAW_STL(_R3, _R1, bpf_frame_size + PPC_LR_STKOFF));
+		EMIT(PPC_RAW_STL(_R3, _R1, retaddr_off));
 
 	/*
 	 * Derive IP address of the traced function.
@@ -925,9 +903,9 @@ static int __arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *rw_im
 		EMIT(PPC_RAW_STL(_R3, _R1, ip_off));
 
 	if (IS_ENABLED(CONFIG_PPC_FTRACE_OUT_OF_LINE)) {
-		/* Fake our LR for unwind */
+		/* Fake our LR for BPF_TRAMP_F_CALL_ORIG case */
 		EMIT(PPC_RAW_ADDI(_R3, _R3, 4));
-		EMIT(PPC_RAW_STL(_R3, _R1, bpf_frame_size + PPC_LR_STKOFF));
+		EMIT(PPC_RAW_STL(_R3, _R1, retaddr_off));
 	}
 
 	/* Save function arg count -- see bpf_get_func_arg_cnt() */
@@ -966,20 +944,19 @@ static int __arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *rw_im
 	/* Call the traced function */
 	if (flags & BPF_TRAMP_F_CALL_ORIG) {
 		/*
-		 * The address in LR save area points to the correct point in the original function
+		 * retaddr on trampoline stack points to the correct point in the original function
 		 * with both PPC_FTRACE_OUT_OF_LINE as well as with traditional ftrace instruction
 		 * sequence
 		 */
-		EMIT(PPC_RAW_LL(_R3, _R1, bpf_frame_size + PPC_LR_STKOFF));
+		EMIT(PPC_RAW_LL(_R3, _R1, retaddr_off));
 		EMIT(PPC_RAW_MTCTR(_R3));
 
 		/* Replicate tail_call_cnt before calling the original BPF prog */
 		if (flags & BPF_TRAMP_F_TAIL_CALL_CTX)
-			bpf_trampoline_setup_tail_call_info(image, ctx, func_frame_offset,
-								bpf_dummy_frame_size, r4_off);
+			bpf_trampoline_setup_tail_call_info(image, ctx, bpf_frame_size, r4_off);
 
 		/* Restore args */
-		bpf_trampoline_restore_args_stack(image, ctx, func_frame_offset, nr_regs, regs_off);
+		bpf_trampoline_restore_args_stack(image, ctx, bpf_frame_size, nr_regs, regs_off);
 
 		/* Restore TOC for 64-bit */
 		if (IS_ENABLED(CONFIG_PPC64_ELF_ABI_V2) && !IS_ENABLED(CONFIG_PPC_KERNEL_PCREL))
@@ -993,7 +970,7 @@ static int __arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *rw_im
 
 		/* Restore updated tail_call_cnt */
 		if (flags & BPF_TRAMP_F_TAIL_CALL_CTX)
-			bpf_trampoline_restore_tail_call_cnt(image, ctx, func_frame_offset, r4_off);
+			bpf_trampoline_restore_tail_call_cnt(image, ctx, bpf_frame_size, r4_off);
 
 		/* Reserve space to patch branch instruction to skip fexit progs */
 		if (ro_image) /* image is NULL for dummy pass */
@@ -1045,7 +1022,7 @@ static int __arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *rw_im
 		EMIT(PPC_RAW_LD(_R2, _R1, 24));
 	if (flags & BPF_TRAMP_F_SKIP_FRAME) {
 		/* Skip the traced function and return to parent */
-		EMIT(PPC_RAW_ADDI(_R1, _R1, func_frame_offset));
+		EMIT(PPC_RAW_ADDI(_R1, _R1, bpf_frame_size));
 		EMIT(PPC_RAW_LL(_R0, _R1, PPC_LR_STKOFF));
 		EMIT(PPC_RAW_MTLR(_R0));
 		EMIT(PPC_RAW_BLR());
@@ -1053,13 +1030,13 @@ static int __arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void *rw_im
 		if (IS_ENABLED(CONFIG_PPC_FTRACE_OUT_OF_LINE)) {
 			EMIT(PPC_RAW_LL(_R0, _R1, alt_lr_off));
 			EMIT(PPC_RAW_MTLR(_R0));
-			EMIT(PPC_RAW_ADDI(_R1, _R1, func_frame_offset));
+			EMIT(PPC_RAW_ADDI(_R1, _R1, bpf_frame_size));
 			EMIT(PPC_RAW_LL(_R0, _R1, PPC_LR_STKOFF));
 			EMIT(PPC_RAW_BLR());
 		} else {
-			EMIT(PPC_RAW_LL(_R0, _R1, bpf_frame_size + PPC_LR_STKOFF));
+			EMIT(PPC_RAW_LL(_R0, _R1, retaddr_off));
 			EMIT(PPC_RAW_MTCTR(_R0));
-			EMIT(PPC_RAW_ADDI(_R1, _R1, func_frame_offset));
+			EMIT(PPC_RAW_ADDI(_R1, _R1, bpf_frame_size));
 			EMIT(PPC_RAW_LL(_R0, _R1, PPC_LR_STKOFF));
 			EMIT(PPC_RAW_MTLR(_R0));
 			EMIT(PPC_RAW_BCTR());
Signature · 1 line
2.53.0
author maintainer
#5 lore raw

PATCH v4 5/6 powerpc64/bpf: fix handling of BPF stack in exception callback

In reply to Hari Bathini

Exception callback reuses the stack frame of exception boundary. When exception boundary and exception callback programs have different BPF stack depth, the current stack unwind in exception callback will fail. Adjust the stack frame size of exception callback, in its prologue, if its BPF stack depth is different from that of exception boundary.

---

* No changes in v4.

Changes in v3:
- Fixed a few comments.

 arch/powerpc/net/bpf_jit_comp64.c | 36 +++++++++++++++++++++----------
 1 file changed, 25 insertions(+), 11 deletions(-)
arch/powerpc/net/bpf_jit_comp64.c +25 -11
index 7655be76f537..04e76440d1ad 100644
@@ -32,14 +32,15 @@
  *
  *		[	prev sp		] <-------------
  *		[    tail_call_info	] 8		|
- *		[   nv gpr save area	] 6*8 + (12*8)	|
+ *		[   nv gpr save area	] (6 * 8)	|
+ *		[ addl. nv gpr save area] (12 * 8)	| <--- exception boundary/callback program
  *		[    local_tmp_var	] 24		|
  * fp (r31) -->	[   ebpf stack space	] upto 512	|
  *		[     frame header	] 32/112	|
  * sp (r1) --->	[    stack pointer	] --------------
  *
- * Additional (12*8) in 'nv gpr save area' only in case of
- * exception boundary.
+ * Additional (12 * 8) in 'nv gpr save area' only in case of
+ * exception boundary/callback.
  */
 
 /* BPF non-volatile registers save area size */
@@ -51,7 +52,7 @@
  * for additional non volatile registers(r14-r25) to be saved
  * at exception boundary
  */
-#define BPF_PPC_EXC_STACK_SAVE (12*8)
+#define BPF_PPC_EXC_STACK_SAVE (12 * 8)
 
 /* stack frame excluding BPF stack, ensure this is quadword aligned */
 #define BPF_PPC_STACKFRAME	(STACK_FRAME_MIN_SIZE + \
@@ -128,12 +129,13 @@ static inline bool bpf_has_stack_frame(struct codegen_context *ctx)
  *		[	  ...       	] 		|
  * sp (r1) --->	[    stack pointer	] --------------
  *		[    tail_call_info	] 8
- *		[   nv gpr save area	] 6*8 + (12*8)
+ *		[   nv gpr save area	] (6 * 8)
+ *		[ addl. nv gpr save area] (12 * 8) <--- exception boundary/callback program
  *		[    local_tmp_var	] 24
  *		[   unused red zone	] 224
  *
- * Additional (12*8) in 'nv gpr save area' only in case of
- * exception boundary.
+ * Additional (12 * 8) in 'nv gpr save area' only in case of
+ * exception boundary/callback.
  */
 static int bpf_jit_stack_local(struct codegen_context *ctx)
 {
@@ -240,10 +242,6 @@ void bpf_jit_build_prologue(u32 *image, struct codegen_context *ctx)
 
 	if (bpf_has_stack_frame(ctx) && !ctx->exception_cb) {
 		/*
-		 * exception_cb uses boundary frame after stack walk.
-		 * It can simply use redzone, this optimization reduces
-		 * stack walk loop by one level.
-		 *
 		 * We need a stack frame, but we don't necessarily need to
 		 * save/restore LR unless we call other functions
 		 */
@@ -287,6 +285,22 @@ void bpf_jit_build_prologue(u32 *image, struct codegen_context *ctx)
 		 * program(main prog) as third arg
 		 */
 		EMIT(PPC_RAW_MR(_R1, _R5));
+		/*
+		 * Exception callback reuses the stack frame of exception boundary.
+		 * But BPF stack depth of exception callback and exception boundary
+		 * don't have to be same. If BPF stack depth is different, adjust the
+		 * stack frame size considering BPF stack depth of exception callback.
+		 * The non-volatile register save area remains unchanged. These non-
+		 * volatile registers are restored in exception callback's epilogue.
+		 */
+		EMIT(PPC_RAW_LD(bpf_to_ppc(TMP_REG_1), _R5, 0));
+		EMIT(PPC_RAW_SUB(bpf_to_ppc(TMP_REG_2), bpf_to_ppc(TMP_REG_1), _R1));
+		EMIT(PPC_RAW_ADDI(bpf_to_ppc(TMP_REG_2), bpf_to_ppc(TMP_REG_2),
+			-BPF_PPC_EXC_STACKFRAME));
+		EMIT(PPC_RAW_CMPLDI(bpf_to_ppc(TMP_REG_2), ctx->stack_size));
+		PPC_BCC_CONST_SHORT(COND_EQ, 12);
+		EMIT(PPC_RAW_MR(_R1, bpf_to_ppc(TMP_REG_1)));
+		EMIT(PPC_RAW_STDU(_R1, _R1, -(BPF_PPC_EXC_STACKFRAME + ctx->stack_size)));
 	}
 
 	/*
Signature · 1 line
2.53.0
author maintainer
#6 lore raw

PATCH v4 6/6 powerpc64/bpf: fix kfunc call support

In reply to Hari Bathini

Commit 61688a82e047 ("powerpc/bpf: enable kfunc call") inadvertently enabled kfunc call support for 32-bit powerpc but that support will not be possible until ABI mismatch between 32-bit powerpc and eBPF is handled in 32-bit powerpc JIT code. Till then, advertise support only for 64-bit powerpc. Also, in powerpc ABI, caller needs to extend the arguments properly based on signedness. The JIT code is responsible for handling this explicitly for kfunc calls as verifier can't handle this for each architecture-specific ABI needs. But this was not taken care of while kfunc call support was enabled for powerpc. Fix it by handling this with bpf_jit_find_kfunc_model() and using zero_extend() & sign_extend() helper functions.

Fixes 61688a82e047 ("powerpc/bpf: enable kfunc call")

---

Changes in v4:
- Fixed src_reg usage in zero_extend() helper - reported by [email protected].

Changes in v3:
- New patch to fix kfunc call ABI issue on powerpc while passing arguments.

 arch/powerpc/net/bpf_jit_comp.c   |   2 +-
 arch/powerpc/net/bpf_jit_comp64.c | 101 +++++++++++++++++++++++++++---
 2 files changed, 94 insertions(+), 9 deletions(-)
arch/powerpc/net/bpf_jit_comp.c +1 -1
index 85457bcb2040..a62a9a92b7b5 100644
@@ -450,7 +450,7 @@ bool bpf_jit_supports_subprog_tailcalls(void)
 
 bool bpf_jit_supports_kfunc_call(void)
 {
-	return true;
+	return IS_ENABLED(CONFIG_PPC64);
 }
 
 bool bpf_jit_supports_arena(void)
arch/powerpc/net/bpf_jit_comp64.c +93 -8

119 lines not on this page · raw message

Signature · 1 line
2.53.0
maintainer Applied
#7 lore raw
In reply to Hari Bathini
Quoted 9 lines · Hari Bathini

List of fixes for powerpc64 BPF JIT:
- Patches 1, 2 & 6 fix issues with existing powerpc64 JIT and are tagged for -stable.
- Patches 3 & 4 resolve issues with bpf2bpf tailcall support added recently.
- Patch 5 resolves an issue reported by [email protected] while adding exceptions support on powerpc64 BPF JIT.

[...]

Applied to powerpc/fixes.

[1/6] powerpc64/bpf: do not increment tailcall count when prog is NULL

      https://git.kernel.org/powerpc/c/521bd39d9d28ce54cbfec7f9b89c94ad4fdb8350
[2/6] powerpc64/bpf: fix the address returned by bpf_get_func_ip
      https://git.kernel.org/powerpc/c/157820264ac3dadfafffad63184b883eb28f9ae0
[3/6] powerpc64/bpf: use consistent tailcall offset in trampoline
      https://git.kernel.org/powerpc/c/3727d6ec13665c1d99bf6dedb107104368ba42b4
[4/6] powerpc64/bpf: remove BPF redzone protection in trampoline stack
      https://git.kernel.org/powerpc/c/2d347d10f8e20e28a9eab52edf55079ae1ec0aae
[5/6] powerpc64/bpf: fix handling of BPF stack in exception callback
      https://git.kernel.org/powerpc/c/51b8de4b3d27ec12128fa2405e526c527a77ae65
[6/6] powerpc64/bpf: fix kfunc call support
      https://git.kernel.org/powerpc/c/01b6ac72729610ae732ca2a66e3a642e23f6cd60

cheers