Le 24/05/2022 à 13:00, Sathvika Vasireddy a écrit :

[Vous ne recevez pas souvent de courriers de la part de 
sv@linux.vnet.ibm.com. Découvrez pourquoi cela peut être important à 
l’adresse https://aka.ms/LearnAboutSenderIdentification.]

On 24/05/22 15:05, Christophe Leroy wrote:

quoted

Le 23/05/2022 à 19:55, Sathvika Vasireddy a écrit :

quoted

This patch enables objtool --mcount on powerpc, and
adds implementation specific to powerpc.

Signed-off-by: Sathvika Vasireddy <redacted>
---
   arch/powerpc/Kconfig                |  1 +
   tools/objtool/arch/powerpc/decode.c | 14 ++++++++++++++
   tools/objtool/check.c               | 12 +++++++-----
   tools/objtool/elf.c                 | 13 +++++++++++++
   tools/objtool/include/objtool/elf.h |  1 +
   5 files changed, 36 insertions(+), 5 deletions(-)

diff --git a/arch/powerpc/Kconfig b/arch/powerpc/Kconfig
index 732a3f91ee5e..3373d44a1298 100644
--- a/arch/powerpc/Kconfig
+++ b/arch/powerpc/Kconfig

@@ -233,6 +233,7 @@ config PPC

     select HAVE_NMI                         if PERF_EVENTS || (PPC64 
&& PPC_BOOK3S)
     select HAVE_OPTPROBES
     select HAVE_OBJTOOL                     if PPC64
+    select HAVE_OBJTOOL_MCOUNT              if HAVE_OBJTOOL
     select HAVE_PERF_EVENTS
     select HAVE_PERF_EVENTS_NMI             if PPC64
     select HAVE_PERF_REGS

diff --git a/tools/objtool/arch/powerpc/decode.c 

b/tools/objtool/arch/powerpc/decode.c
index e3b77a6ce357..ad3d79fffac2 100644

--- a/tools/objtool/arch/powerpc/decode.c
+++ b/tools/objtool/arch/powerpc/decode.c

@@ -40,12 +40,26 @@ int arch_decode_instruction(struct objtool_file 

*file, const struct section *sec
                         struct list_head *ops_list)
   {
     u32 insn;
+    unsigned int opcode;

     *immediate = 0;
     memcpy(&insn, sec->data->d_buf+offset, 4);
     *len = 4;
     *type = INSN_OTHER;

+    opcode = (insn >> 26);

You dont need the brackets here.

quoted

+
+    switch (opcode) {
+    case 18: /* bl */
+            if ((insn & 3) == 1) {
+                    *type = INSN_CALL;
+                    *immediate = insn & 0x3fffffc;
+                    if (*immediate & 0x2000000)
+                            *immediate -= 0x4000000;
+            }
+            break;
+    }
+
     return 0;
   }

diff --git a/tools/objtool/check.c b/tools/objtool/check.c
index 056302d58e23..fd8bad092f89 100644
--- a/tools/objtool/check.c
+++ b/tools/objtool/check.c

@@ -832,7 +832,7 @@ static int create_mcount_loc_sections(struct 

objtool_file *file)

             if (elf_add_reloc_to_insn(file->elf, sec,
                                       idx * sizeof(unsigned long),
-                                      R_X86_64_64,
+                                      elf_reloc_type_long(file->elf),
                                       insn->sec, insn->offset))
                     return -1;

@@ -2183,7 +2183,7 @@ static int classify_symbols(struct objtool_file 

*file)
                     if (arch_is_retpoline(func))
                             func->retpoline_thunk = true;

-                    if (!strcmp(func->name, "__fentry__"))
+                    if ((!strcmp(func->name, "__fentry__")) || 
(!strcmp(func->name, "_mcount")))
                             func->fentry = true;

                     if (is_profiling_func(func->name))

@@ -2259,9 +2259,11 @@ static int decode_sections(struct objtool_file 

*file)
      * Must be before add_jump_destinations(), which depends on 'func'
      * being set for alternatives, to enable proper sibling call 
detection.
      */
-    ret = add_special_section_alts(file);
-    if (ret)
-            return ret;
+    if (opts.stackval || opts.orc || opts.uaccess || opts.noinstr) {
+            ret = add_special_section_alts(file);
+            if (ret)
+                    return ret;
+    }

I think this change should be a patch by itself, it's not related to
powerpc.

Makes sense. I'll make this a separate patch in the next revision.

Great. Can you base your next revision on the one I just sent out ?

I will now start looking at inline static calls for PPC32.

quoted

quoted

     ret = add_jump_destinations(file);
     if (ret)

diff --git a/tools/objtool/elf.c b/tools/objtool/elf.c
index c25e957c1e52..95763060d551 100644
--- a/tools/objtool/elf.c
+++ b/tools/objtool/elf.c

@@ -793,6 +793,19 @@ elf_create_section_symbol(struct elf *elf, 

struct section *sec)
     return sym;
   }

+int elf_reloc_type_long(struct elf *elf)

Not sure it's a good name, because for 32 bits we have to use 'int'.

Sure, I'll rename it to elf_reloc_type() or some such.

quoted

quoted

+{
+    switch (elf->ehdr.e_machine) {
+    case EM_X86_64:
+            return R_X86_64_64;
+    case EM_PPC64:
+            return R_PPC64_ADDR64;
+    default:
+            WARN("unknown machine...");
+            exit(-1);
+    }
+}

Wouldn't it be better to make that function arch specific ?

This is so that we can support cross architecture builds.

I'm not sure I follow you here.

This is only based on the target, it doesn't depend on the build host so 
I can't the link with cross arch builds.

The same as you have arch_decode_instruction(), you could have 
arch_elf_reloc_type_long()
It would make sense indeed, because there is no point in supporting X86 
relocation when you don't support X86 instruction decoding.

Christophe