Allocate bss section within proper range of other sectionswip/fix-i386-2

Allocate bss section within proper range of other sections:

* when `+RTS -xp` is passed, allocate it contiguously as we did for
  jump islands
* when we mmap the code to lower 2Gb, we should allocate bss section
  there too

This depends on {D5195}

Test Plan:
1. `./validate`

2.

with

```
DYNAMIC_GHC_PROGRAMS = NO
DYNAMIC_BY_DEFAULT = NO
```

`TEST="T15729" make test` passed in both linux and macos.

3.

Also test in a use case where we used to encouter error like:

```
ghc-iserv-prof: R_X86_64_PC32 relocation out of range: (noname) =
b90282ba
```

and now, everything works fine.

Reviewers: simonmar, bgamari, angerman, erikd

Reviewed By: simonmar

Subscribers: rwbarton, carter

GHC Trac Issues: #15729

Differential Revision: https://phabricator.haskell.org/D5219

1 files changed, 14 insertions, 17 deletions

diff --git a/rts/linker/SymbolExtras.c b/rts/linker/SymbolExtras.c
index 4c40b10877..a9e4c37967 100644
--- a/rts/linker/SymbolExtras.c
+++ b/rts/linker/SymbolExtras.c
@@ -31,10 +31,11 @@
 #endif /* RTS_LINKER_USE_MMAP */
 
 /*
-  ocAllocateSymbolExtras
+  ocAllocateExtras
 
   Allocate additional space at the end of the object file image to make room
-  for jump islands (powerpc, x86_64, arm) and GOT entries (x86_64).
+  for jump islands (powerpc, x86_64, arm), GOT entries (x86_64) and
+  bss sections.
 
   PowerPC relative branch instructions have a 24 bit displacement field.
   As PPC code is always 4-byte-aligned, this yields a +-32MB range.
@@ -49,12 +50,11 @@
   filled in by makeSymbolExtra below.
 */
 
-int ocAllocateSymbolExtras( ObjectCode* oc, int count, int first )
+int ocAllocateExtras(ObjectCode* oc, int count, int first, int bssSize)
 {
-  size_t n;
   void* oldImage = oc->image;
 
-  if (count > 0) {
+  if (count > 0 || bssSize > 0) {
     if (!RTS_LINKER_USE_MMAP) {
 
       // round up to the nearest 4
@@ -65,16 +65,15 @@ int ocAllocateSymbolExtras( ObjectCode* oc, int count, int first )
       oc->image = stgReallocBytes( oc->image,
                                misalignment +
                                aligned + sizeof (SymbolExtra) * count,
-                               "ocAllocateSymbolExtras" );
+                               "ocAllocateExtras" );
       oc->image += misalignment;
 
       oc->symbol_extras = (SymbolExtra *) (oc->image + aligned);
     } else if (USE_CONTIGUOUS_MMAP || RtsFlags.MiscFlags.linkerAlwaysPic) {
-      n = roundUpToPage(oc->fileSize);
-
-      /* Keep image and symbol_extras contiguous */
-
-      size_t allocated_size = n + (sizeof(SymbolExtra) * count);
+      /* Keep image, bssExtras and symbol_extras contiguous */
+      size_t n = roundUpToPage(oc->fileSize);
+      bssSize = roundUpToAlign(bssSize, 8);
+      size_t allocated_size = n + bssSize + (sizeof(SymbolExtra) * count);
       void *new = mmapForLinker(allocated_size, MAP_ANONYMOUS, -1, 0);
       if (new) {
           memcpy(new, oc->image, oc->fileSize);
@@ -83,12 +82,10 @@ int ocAllocateSymbolExtras( ObjectCode* oc, int count, int first )
           }
           oc->image = new;
           oc->imageMapped = true;
-          oc->fileSize = n + (sizeof(SymbolExtra) * count);
-          oc->symbol_extras = (SymbolExtra *) (oc->image + n);
-          if (mprotect(new, allocated_size,
-                       PROT_READ | PROT_WRITE | PROT_EXEC) != 0) {
-            sysErrorBelch("unable to protect memory");
-          }
+          oc->fileSize = allocated_size;
+          oc->symbol_extras = (SymbolExtra *) (oc->image + n + bssSize);
+          oc->bssBegin = oc->image + n;
+          oc->bssEnd = oc->image + n + bssSize;
       }
       else {
           oc->symbol_extras = NULL;