Skip to contentSkip to conversation

btf_encoder: Fix handling of percpu symbols on s390

PATCH Applied Ilya Leoshkevich 2 participants 2 messages Oct 12, 2021 – Oct 15, 2021 BPF

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 btf_encoder: Fix handling of percpu symbols on s390

pahole does not generate VARs for percpu symbols on s390. A percpu symbol definition on a typical x86_64 kernel looks like this:

[33] .data..percpu PROGBITS 0000000000000000 01c00000

                                          ^^^^^^^^^^^^^^^^ sh_addr
  LOAD           0x0000000001c00000 0x0000000000000000 0x000000000286f000
                                    ^^^^^^^^^^^^^^^^^^ p_vaddr
 13559: 000000000001ba50     4 OBJECT  LOCAL  DEFAULT   33 cpu_profile_flip
        ^^^^^^^^^^^^^^^^ st_value

Most importantly, .data..percpu's sh_addr is 0, and this is what pahole is currently assuming. However, on s390 this is different:

[37] .data..percpu PROGBITS 00000000019cd000 018ce000

                                           ^^^^^^^^^^^^^^^^ sh_addr
  LOAD           0x000000000136e000 0x000000000146d000 0x000000000146d000
                                    ^^^^^^^^^^^^^^^^^^ p_vaddr
80377: 0000000001ba1440     4 OBJECT  WEAK   DEFAULT   37 cpu_profile_flip
       ^^^^^^^^^^^^^^^^ st_value

Fix by restructuring the code to always use section-relative offsets for symbols. Change the comment to focus on this invariant.

---
 btf_encoder.c | 21 ++++++++++++++-------
 1 file changed, 14 insertions(+), 7 deletions(-)
btf_encoder.c +14 -7
index c341f95..16e90c3 100644
@@ -56,7 +56,8 @@ struct btf_encoder {
 			  raw_output,
 			  verbose,
 			  force,
-			  gen_floats;
+			  gen_floats,
+			  is_rel;
 	uint32_t	  array_index_id;
 	struct {
 		struct var_info vars[MAX_PERCPU_VAR_CNT];
@@ -1104,6 +1105,13 @@ static int btf_encoder__collect_percpu_var(struct btf_encoder *encoder, GElf_Sym
 	if (encoder->verbose)
 		printf("Found per-CPU symbol '%s' at address 0x%" PRIx64 "\n", sym_name, addr);
 
+	/* Make sure addr is section-relative. For kernel modules (which are
+	 * ET_REL files) this is already the case. For vmlinux (which is an
+	 * ET_EXEC file) we need to subtract the section address.
+	 */
+	if (!encoder->is_rel)
+		addr -= encoder->percpu.base_addr;
+
 	if (encoder->percpu.var_cnt == MAX_PERCPU_VAR_CNT) {
 		fprintf(stderr, "Reached the limit of per-CPU variables: %d\n",
 			MAX_PERCPU_VAR_CNT);
@@ -1195,12 +1203,9 @@ static int btf_encoder__encode_cu_variables(struct btf_encoder *encoder, struct
 		addr = var->ip.addr;
 		dwarf_name = variable__name(var);
 
-		/* DWARF takes into account .data..percpu section offset
-		 * within its segment, which for vmlinux is 0, but for kernel
-		 * modules is >0. ELF symbols, on the other hand, don't take
-		 * into account these offsets (as they are relative to the
-		 * section start), so to match DWARF and ELF symbols we need
-		 * to negate the section base address here.
+		/* Make sure addr is section-relative. DWARF, unlike ELF,
+		 * always contains virtual symbol addresses, so subtract
+		 * the section address unconditionally.
 		 */
 		if (addr < encoder->percpu.base_addr || addr >= encoder->percpu.base_addr + encoder->percpu.sec_sz)
 			continue;
@@ -1322,6 +1327,8 @@ struct btf_encoder *btf_encoder__new(struct cu *cu, const char *detached_filenam
 			goto out_delete;
 		}
 
+		encoder->is_rel = ehdr.e_type == ET_REL;
+
 		switch (ehdr.e_ident[EI_DATA]) {
 		case ELFDATA2LSB:
 			btf__set_endianness(encoder->btf, BTF_LITTLE_ENDIAN);
Signature · 1 line
2.31.1
#1 lore raw

Em Tue, Oct 12, 2021 at 04:26:37AM +0200, Ilya Leoshkevich escreveu:

Quoted 20 lines

pahole does not generate VARs for percpu symbols on s390. A percpu symbol definition on a typical x86_64 kernel looks like this:

[33] .data..percpu PROGBITS 0000000000000000 01c00000

                                          ^^^^^^^^^^^^^^^^ sh_addr
  LOAD           0x0000000001c00000 0x0000000000000000 0x000000000286f000
                                    ^^^^^^^^^^^^^^^^^^ p_vaddr
 13559: 000000000001ba50     4 OBJECT  LOCAL  DEFAULT   33 cpu_profile_flip
        ^^^^^^^^^^^^^^^^ st_value

Most importantly, .data..percpu's sh_addr is 0, and this is what pahole is currently assuming. However, on s390 this is different:

   [37] .data..percpu     PROGBITS         00000000019cd000  018ce000
                                           ^^^^^^^^^^^^^^^^ sh_addr
  LOAD           0x000000000136e000 0x000000000146d000 0x000000000146d000
                                    ^^^^^^^^^^^^^^^^^^ p_vaddr
80377: 0000000001ba1440     4 OBJECT  WEAK   DEFAULT   37 cpu_profile_flip
       ^^^^^^^^^^^^^^^^ st_value
Quoted 2 lines

Fix by restructuring the code to always use section-relative offsets for symbols. Change the comment to focus on this invariant.

Thanks, applied.

- Arnaldo

Quoted 60 lines
---
 btf_encoder.c | 21 ++++++++++++++-------
 1 file changed, 14 insertions(+), 7 deletions(-)
btf_encoder.c +14 -7
index c341f95..16e90c3 100644
@@ -56,7 +56,8 @@ struct btf_encoder {
 			  raw_output,
 			  verbose,
 			  force,
-			  gen_floats;
+			  gen_floats,
+			  is_rel;
 	uint32_t	  array_index_id;
 	struct {
 		struct var_info vars[MAX_PERCPU_VAR_CNT];
@@ -1104,6 +1105,13 @@ static int btf_encoder__collect_percpu_var(struct btf_encoder *encoder, GElf_Sym
 	if (encoder->verbose)
 		printf("Found per-CPU symbol '%s' at address 0x%" PRIx64 "\n", sym_name, addr);
 
+	/* Make sure addr is section-relative. For kernel modules (which are
+	 * ET_REL files) this is already the case. For vmlinux (which is an
+	 * ET_EXEC file) we need to subtract the section address.
+	 */
+	if (!encoder->is_rel)
+		addr -= encoder->percpu.base_addr;
+
 	if (encoder->percpu.var_cnt == MAX_PERCPU_VAR_CNT) {
 		fprintf(stderr, "Reached the limit of per-CPU variables: %d\n",
 			MAX_PERCPU_VAR_CNT);
@@ -1195,12 +1203,9 @@ static int btf_encoder__encode_cu_variables(struct btf_encoder *encoder, struct
 		addr = var->ip.addr;
 		dwarf_name = variable__name(var);
 
-		/* DWARF takes into account .data..percpu section offset
-		 * within its segment, which for vmlinux is 0, but for kernel
-		 * modules is >0. ELF symbols, on the other hand, don't take
-		 * into account these offsets (as they are relative to the
-		 * section start), so to match DWARF and ELF symbols we need
-		 * to negate the section base address here.
+		/* Make sure addr is section-relative. DWARF, unlike ELF,
+		 * always contains virtual symbol addresses, so subtract
+		 * the section address unconditionally.
 		 */
 		if (addr < encoder->percpu.base_addr || addr >= encoder->percpu.base_addr + encoder->percpu.sec_sz)
 			continue;
@@ -1322,6 +1327,8 @@ struct btf_encoder *btf_encoder__new(struct cu *cu, const char *detached_filenam
 			goto out_delete;
 		}
 
+		encoder->is_rel = ehdr.e_type == ET_REL;
+
 		switch (ehdr.e_ident[EI_DATA]) {
 		case ELFDATA2LSB:
 			btf__set_endianness(encoder->btf, BTF_LITTLE_ENDIAN);
Signature · 1 line
2.31.1
Signature · 1 line
- Arnaldo