aboutsummaryrefslogtreecommitdiffstats
path: root/meta/recipes-devtools/patchelf/patchelf/avoidholes.patch
blob: a273688b9883bcabfc7febf6c16613d34d8193c7 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
Different types of binaries create challenges for patchelf. In order to extend 
sections they need to be moved within the binary. The current approach to 
handling ET_DYN binaries is to move the INTERP section to the end of the file.
This means changing PT_PHDR to add an extra PT_LOAD section so that the new section
is mmaped into memory by the elf loader in the kernel. In order to extend PHDR,
this means moving it to the end of the file.

Its documented in patchelf there is a kernel 'bug' which means that if you have holes
in memory between the base load address and the PT_LOAD segment that contains PHDR,
it will pass an incorrect PHDR address to ld.so and fail to load the binary, segfaulting.

To avoid this, the code currently inserts space into the binary to ensure that when 
loaded into memory there are no holes between the PT_LOAD sections. This inflates the 
binaries by many MBs in some cases. Whilst we could make them sparse, there is a second
issue which is that strip can fail to process these binaries:

$ strip fixincl
Not enough room for program headers, try linking with -N
[.note.ABI-tag]: Bad value

This turns out to be due to libbfd not liking the relocated PHDR section either 
(https://github.com/NixOS/patchelf/issues/10).

Instead this patch implements a different approach, leaving PHDR where it is but extending
it in place to allow addition of a new PT_LOAD section. This overwrites sections in the 
binary but those get moved to the end of the file in the new PT_LOAD section.

This is based on patches linked from the above github issue, however whilst the idea
was good, the implementation wasn't correct and they've been rewritten here.

RP
2017/3/7

Upstream-Status: Pending

Index: patchelf-0.9/src/patchelf.cc
===================================================================
--- patchelf-0.9.orig/src/patchelf.cc
+++ patchelf-0.9/src/patchelf.cc
@@ -146,6 +146,8 @@ private:
     string & replaceSection(const SectionName & sectionName,
         unsigned int size);
 
+    bool haveReplacedSection(const SectionName & sectionName);
+
     void writeReplacedSections(Elf_Off & curOff,
         Elf_Addr startAddr, Elf_Off startOffset);
 
@@ -497,6 +499,16 @@ unsigned int ElfFile<ElfFileParamNames>:
     return 0;
 }
 
+template<ElfFileParams>
+bool ElfFile<ElfFileParamNames>::haveReplacedSection(const SectionName & sectionName)
+{
+    ReplacedSections::iterator i = replacedSections.find(sectionName);
+
+    if (i != replacedSections.end())
+        return true;
+    return false;
+}
+
 
 template<ElfFileParams>
 string & ElfFile<ElfFileParamNames>::replaceSection(const SectionName & sectionName,
@@ -595,52 +607,52 @@ void ElfFile<ElfFileParamNames>::rewrite
 
     debug("last page is 0x%llx\n", (unsigned long long) startPage);
 
+    /* Because we're adding a new section header, we're necessarily increasing
+       the size of the program header table.  This can cause the first section
+       to overlap the program header table in memory; we need to shift the first
+       few segments to someplace else. */
+    /* Some sections may already be replaced so account for that */
+    unsigned int i = 1;
+    Elf_Addr pht_size = sizeof(Elf_Ehdr) + (phdrs.size() + 1)*sizeof(Elf_Phdr);
+    while( shdrs[i].sh_addr <= pht_size && i < rdi(hdr->e_shnum) ) {
+        if (not haveReplacedSection(getSectionName(shdrs[i])))
+            replaceSection(getSectionName(shdrs[i]), shdrs[i].sh_size);
+        i++;
+    }
 
-    /* Compute the total space needed for the replaced sections and
-       the program headers. */
-    off_t neededSpace = (phdrs.size() + 1) * sizeof(Elf_Phdr);
+    /* Compute the total space needed for the replaced sections */
+    off_t neededSpace = 0;
     for (ReplacedSections::iterator i = replacedSections.begin();
          i != replacedSections.end(); ++i)
         neededSpace += roundUp(i->second.size(), sectionAlignment);
     debug("needed space is %d\n", neededSpace);
 
-
     size_t startOffset = roundUp(fileSize, getPageSize());
 
     growFile(startOffset + neededSpace);
 
-
     /* Even though this file is of type ET_DYN, it could actually be
        an executable.  For instance, Gold produces executables marked
-       ET_DYN.  In that case we can still hit the kernel bug that
-       necessitated rewriteSectionsExecutable().  However, such
-       executables also tend to start at virtual address 0, so
+       ET_DYN as does LD when linking with pie. If we move PT_PHDR, it
+       has to stay in the first PT_LOAD segment or any subsequent ones
+       if they're continuous in memory due to linux kernel constraints
+       (see BUGS). Since the end of the file would be after bss, we can't 
+       move PHDR there, we therefore choose to leave PT_PHDR where it is but
+       move enough following sections such that we can add the extra PT_LOAD
+       section to it. This PT_LOAD segment ensures the sections at the end of
+       the file are mapped into memory for ld.so to process.
+       We can't use the approach in rewriteSectionsExecutable()
+       since DYN executables tend to start at virtual address 0, so
        rewriteSectionsExecutable() won't work because it doesn't have
-       any virtual address space to grow downwards into.  As a
-       workaround, make sure that the virtual address of our new
-       PT_LOAD segment relative to the first PT_LOAD segment is equal
-       to its offset; otherwise we hit the kernel bug.  This may
-       require creating a hole in the executable.  The bigger the size
-       of the uninitialised data segment, the bigger the hole. */
+       any virtual address space to grow downwards into. */
     if (isExecutable) {
         if (startOffset >= startPage) {
             debug("shifting new PT_LOAD segment by %d bytes to work around a Linux kernel bug\n", startOffset - startPage);
-        } else {
-            size_t hole = startPage - startOffset;
-            /* Print a warning, because the hole could be very big. */
-            fprintf(stderr, "warning: working around a Linux kernel bug by creating a hole of %zu bytes in ‘%s’\n", hole, fileName.c_str());
-            assert(hole % getPageSize() == 0);
-            /* !!! We could create an actual hole in the file here,
-               but it's probably not worth the effort. */
-            growFile(fileSize + hole);
-            startOffset += hole;
         }
         startPage = startOffset;
     }
 
-
-    /* Add a segment that maps the replaced sections and program
-       headers into memory. */
+    /* Add a segment that maps the replaced sections into memory. */
     phdrs.resize(rdi(hdr->e_phnum) + 1);
     wri(hdr->e_phnum, rdi(hdr->e_phnum) + 1);
     Elf_Phdr & phdr = phdrs[rdi(hdr->e_phnum) - 1];
@@ -653,15 +665,12 @@ void ElfFile<ElfFileParamNames>::rewrite
 
 
     /* Write out the replaced sections. */
-    Elf_Off curOff = startOffset + phdrs.size() * sizeof(Elf_Phdr);
+    Elf_Off curOff = startOffset;
     writeReplacedSections(curOff, startPage, startOffset);
     assert(curOff == startOffset + neededSpace);
 
-
-    /* Move the program header to the start of the new area. */
-    wri(hdr->e_phoff, startOffset);
-
-    rewriteHeaders(startPage);
+    /* Write out the updated program and section headers */
+    rewriteHeaders(hdr->e_phoff);
 }