[llvm-objcopy]Preserve data in segments not covered by sections

[jhenderson]

Summary

llvm-objcopy previously knew nothing about data in segments that wasn't covered by section headers, meaning that it wrote zeroes instead of what was there. As it is possible for this data to be useful to the loader, this patch causes llvm-objcopy to start preserving this data. Data in sections that are explicitly removed continues to be written as zeroes.

This fixes https://bugs.llvm.org/show_bug.cgi?id=41005.

The test is possibly more complicated than the current implementation warrants, but did pick up several problems with an earlier implementation, so I'd prefer to keep the test cases rather than reduce it down.

[jhenderson]

Fix the test to use a correct regex pattern, and add content to the "blob" sections to show that sections are set to zero post-removal, as desired.

[jakehehrlich]

The code looks good to me but the test is going to take me a while to review. I don't think I'll get to this today.

[jhenderson]

In D59483#1433555, @jakehehrlich wrote:

The code looks good to me but the test is going to take me a while to review. I don't think I'll get to this today.

Yeah, sorry about that. Let me know if you need any help in deciphering it!

[rupprecht]

test/tools/llvm-objcopy/ELF/preserve-segment-contents.test

26

You could probably make this shorter just this in a loop, e.g.

with open('%/t.in', 'r+') as input:
  for offset in [
    0x2000, 0x2008, ...,
    ..., 0x5038]:
      input.seek(offset); input.write('\xDE\xAD\xBE\xEF');

[jhenderson]

Update test as suggested.

[jakehehrlich]

Ok, I think how the test functions is fine and it's probably a good test. It certainly tests every case I'd consider. I haven't checked that all the offsets are correct but they don't look wrong at a glance.

LGTM.

[jhenderson]

Fix an issue where the ELF header and program header table were not being updated if they were covered by a segment. This was fixed by doing the writing of segment data before writing the ELF header and program header table. Also add a test dedicated to showing that this is no longer an issue.

[jhenderson]

@rupprecht, @jakehehrlich, would you take a look at this again, please. I found a bug in it whilst testing that resulted in the ELF header and program header table not being updated.

test/tools/llvm-objcopy/ELF/preserve-segment-contents.test

+# We want to preserve areas in segments that are not covered by section headers.
+# This test shows that we do this for areas at the start of a segment, between
+# sections in a segment, and after all sections in a segment.
+# To create inputs with arbitrary data in segments, not covered by sections, we
+# use yaml2obj to create segments with sections covering all areas, then remove
+# some sections in those segments, and finally write over the areas of the
+# removed sections using python.
+
+# blob* sections are the sections that will be removed to create unlabelled
+# areas and then overwritten with data to show we preserve the data.
+
+# RUN: yaml2obj %s -o %t.base
+# RUN: llvm-objcopy %t.base %t.stripped --regex -R blob.*
+# Show that the removal leaves the bytes as zeroes, as desired, for all our
+# test cases.
+# RUN: od -t x1 -j 0x2000 -N 24 %t.stripped | FileCheck %s --check-prefix=CHECK1 -DPATTERN="00 00 00 00"
+# RUN: od -t x1 -j 0x2100 -N 12 %t.stripped | FileCheck %s --check-prefix=CHECK2 -DPATTERN="00 00 00 00"
+# RUN: od -t x1 -j 0x2200 -N 4  %t.stripped | FileCheck %s --check-prefix=CHECK3 -DPATTERN="00 00 00 00"
+# RUN: od -t x1 -j 0x2300 -N 12 %t.stripped | FileCheck %s --check-prefix=CHECK4 -DPATTERN="00 00 00 00"
+# RUN: od -t x1 -j 0x3000 -N 68 %t.stripped | FileCheck %s --check-prefix=CHECK5 -DPATTERN="00 00 00 00"
+# RUN: od -t x1 -j 0x4000 -N 60 %t.stripped | FileCheck %s --check-prefix=CHECK6 -DPATTERN="00 00 00 00"
+# RUN: od -t x1 -j 0x5000 -N 60 %t.stripped | FileCheck %s --check-prefix=CHECK7 -DPATTERN="00 00 00 00"
+
+# RUN: cp %t.stripped %t.in
+# RUN: %python -c "with open('%/t.in', 'r+') as input:          \
+# RUN:     input.seek(0x2000); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x2008); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x200C); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x2014); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x2104); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x2300); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x3008); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x3010); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x3018); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x3020); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x3028); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x302C); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x3034); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x303C); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x4000); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x4008); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x4010); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x4014); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x401C); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x4024); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x4034); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x5000); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x5008); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x5010); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x501C); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x5024); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x502C); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x5030); input.write('\xDE\xAD\xBE\xEF'); \
+# RUN:     input.seek(0x5038); input.write('\xDE\xAD\xBE\xEF');"
+# RUN: llvm-objcopy %t.in %t.out
+# RUN: od -t x1 -j 0x2000 -N 24 %t.out | FileCheck %s --check-prefix=CHECK1 -DPATTERN="de ad be ef"
+# RUN: od -t x1 -j 0x2100 -N 12 %t.out | FileCheck %s --check-prefix=CHECK2 -DPATTERN="de ad be ef"
+# RUN: od -t x1 -j 0x2200 -N 4  %t.out | FileCheck %s --check-prefix=CHECK3 -DPATTERN="de ad be ef"
+# RUN: od -t x1 -j 0x2300 -N 12 %t.out | FileCheck %s --check-prefix=CHECK4 -DPATTERN="de ad be ef"
+# RUN: od -t x1 -j 0x3000 -N 68 %t.out | FileCheck %s --check-prefix=CHECK5 -DPATTERN="de ad be ef"
+# RUN: od -t x1 -j 0x4000 -N 60 %t.out | FileCheck %s --check-prefix=CHECK6 -DPATTERN="de ad be ef"
+# RUN: od -t x1 -j 0x5000 -N 60 %t.out | FileCheck %s --check-prefix=CHECK7 -DPATTERN="de ad be ef"
+
+# CHECK1:      [[PATTERN]] 11 22 33 44 [[PATTERN]] [[PATTERN]]
+# CHECK1-NEXT: 55 66 77 88 [[PATTERN]]
+# CHECK2:      99 00 aa bb [[PATTERN]] cc dd ee ff
+# CHECK3:      fe fe fe fe
+# CHECK4:      [[PATTERN]] 00 00 00 00 00 00 00 00
+# CHECK5:      ff ff ee ee dd dd cc cc [[PATTERN]] bb bb aa aa
+# CHECK5-NEXT: [[PATTERN]] 00 00 99 99 [[PATTERN]] 88 88 77 77
+# CHECK5-NEXT: [[PATTERN]] 66 66 55 55 [[PATTERN]] [[PATTERN]]
+# CHECK5-NEXT: 44 44 33 33 [[PATTERN]] 22 22 11 11 [[PATTERN]]
+# CHECK5-NEXT: 00 11 22 33
+# CHECK6:      [[PATTERN]] 44 55 66 77 [[PATTERN]] 88 99 aa bb
+# CHECK6-NEXT: [[PATTERN]] [[PATTERN]] cc dd ee ff [[PATTERN]]
+# CHECK6-NEXT: ff ee dd cc [[PATTERN]] bb aa 99 88 77 66 55 44
+# CHECK6-NEXT: 33 22 11 00 [[PATTERN]] 11 11 11 11
+# CHECK7:      [[PATTERN]] 12 34 56 78 [[PATTERN]] 90 ab cd ef
+# CHECK7-NEXT: [[PATTERN]] fe dc ba 09 87 65 43 21 [[PATTERN]]
+# CHECK7-NEXT: 22 22 22 22 [[PATTERN]] 33 33 33 33 [[PATTERN]]
+# CHECK7-NEXT: [[PATTERN]] 44 44 44 44 [[PATTERN]]
+
+--- !ELF
+FileHeader:
+  Class:   ELFCLASS64
+  Data:    ELFDATA2LSB
+  Type:    ET_EXEC
+  Machine: EM_X86_64
+Sections:
+  - Name: blob1
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x2000
+    AddressAlign: 0x2000
+  - Name: section1
+    Type: SHT_PROGBITS
+    Address: 0x2004
+    Content: '11223344'
+  - Name: blob2
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x2008
+  - Name: section2
+    Type: SHT_NOBITS
+    Size: 4
+    Address: 0x200C
+  - Name: blob3
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x2010
+  - Name: section3
+    Type: SHT_PROGBITS
+    Content: '55667788'
+    Address: 0x2014
+  - Name: blob4
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x2018
+  - Name: section4
+    Type: SHT_PROGBITS
+    Content: '9900aabb'
+    Address: 0x2100
+    AddressAlign: 0x100
+  - Name: blob5
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x2104
+  - Name: section5
+    Type: SHT_PROGBITS
+    Address: 0x2108
+    Content: 'ccddeeff'
+  - Name: section6
+    Type: SHT_PROGBITS
+    Content: 'fefefefe'
+    Address: 0x2200
+    AddressAlign: 0x100
+  - Name: blob6
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x2300
+    AddressAlign: 0x100
+  - Name: sectionA
+    Type: SHT_PROGBITS
+    Content: 'ffffeeee'
+    Address: 0x3000
+    AddressAlign: 0x1000
+  - Name: sectionB
+    Type: SHT_PROGBITS
+    Content: 'ddddcccc'
+    Address: 0x3004
+  - Name: blobA
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x3008
+  - Name: sectionC
+    Type: SHT_PROGBITS
+    Content: 'bbbbaaaa'
+    Address: 0x300C
+  - Name: blobB
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x3010
+  - Name: sectionD
+    Type: SHT_PROGBITS
+    Content: '00009999'
+    Address: 0x3014
+  - Name: blobC
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x3018
+  - Name: sectionE
+    Type: SHT_PROGBITS
+    Content: '88887777'
+    Address: 0x301C
+  - Name: blobD
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x3020
+  - Name: sectionF
+    Type: SHT_PROGBITS
+    Content: '66665555'
+    Address: 0x3024
+  - Name: blobE
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x3028
+  - Name: blobF
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x302C
+  - Name: sectionG
+    Type: SHT_PROGBITS
+    Content: '44443333'
+    Address: 0x3030
+  - Name: blobG
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x3034
+  - Name: sectionH
+    Type: SHT_PROGBITS
+    Content: '22221111'
+    Address: 0x3038
+  - Name: blobH
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x303C
+  - Name: sectionI
+    Type: SHT_PROGBITS
+    Content: '00112233'
+    Address: 0x3040
+  - Name: blobz
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x4000
+    AddressAlign: 0x1000
+  - Name: sectionz
+    Type: SHT_PROGBITS
+    Content: '44556677'
+    Address: 0x4004
+  - Name: bloby
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x4008
+  - Name: sectiony
+    Type: SHT_PROGBITS
+    Content: '8899aabb'
+    Address: 0x400C
+  - Name: blobx
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x4010
+  - Name: blobw
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x4014
+  - Name: sectionx
+    Type: SHT_PROGBITS
+    Content: 'ccddeeff'
+    Address: 0x4018
+  - Name: blobv
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x401C
+  - Name: sectionw
+    Type: SHT_PROGBITS
+    Content: 'ffeeddcc'
+    Address: 0x4020
+  - Name: blobu
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x4024
+  - Name: sectionv
+    Type: SHT_PROGBITS
+    Content: 'bbaa9988'
+    Address: 0x4028
+  - Name: sectionu
+    Type: SHT_PROGBITS
+    Content: '77665544'
+    Address: 0x402C
+  - Name: sectiont
+    Type: SHT_PROGBITS
+    Content: '33221100'
+    Address: 0x4030
+  - Name: blobt
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x4034
+  - Name: sections
+    Type: SHT_PROGBITS
+    Content: '11111111'
+    Address: 0x4038
+  - Name: bloba
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x5000
+    AddressAlign: 0x1000
+  - Name: sectiona
+    Type: SHT_PROGBITS
+    Content: '12345678'
+    Address: 0x5004
+  - Name: blobb
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x5008
+  - Name: sectionb
+    Type: SHT_PROGBITS
+    Content: '90abcdef'
+    Address: 0x500C
+  - Name: blobc
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x5010
+  - Name: sectionc
+    Type: SHT_PROGBITS
+    Content: 'fedcba09'
+    Address: 0x5014
+  - Name: sectiond
+    Type: SHT_PROGBITS
+    Content: '87654321'
+    Address: 0x5018
+  - Name: blobd
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x501C
+  - Name: sectione
+    Type: SHT_PROGBITS
+    Content: '22222222'
+    Address: 0x5020
+  - Name: blobe
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x5024
+  - Name: sectionf
+    Type: SHT_PROGBITS
+    Content: '33333333'
+    Address: 0x5028
+  - Name: blobf
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x502C
+  - Name: blobg
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x5030
+  - Name: sectiong
+    Type: SHT_PROGBITS
+    Content: '44444444'
+    Address: 0x5034
+  - Name: blobh
+    Type: SHT_PROGBITS
+    Content: 'abbababa'
+    Address: 0x5038
+ProgramHeaders:
+  # First segment has unlabelled space at start and end.
+  - Type:  0x6ABCDEF0 # Non-specific segment type.
+    VAddr: 0x2000
+    PAddr: 0x2000
+    Align: 0x2000
+    Sections:
+      - Section: blob1
+      - Section: section1
+      - Section: blob2
+      - Section: section2 # nobits
+      - Section: blob3
+      - Section: section3
+      - Section: blob4
+  # Second segment has sections at start and end.
+  - Type:  0x6ABCDEF0
+    VAddr: 0x2100
+    PAddr: 0x2100
+    Align: 0x100
+    Sections:
+      - Section: section4
+      - Section: blob5
+      - Section: section5
+  # Third segment is all covered by a section.
+  - Type:  0x6ABCDEF0
+    VAddr: 0x2200
+    PAddr: 0x2200
+    Align: 0x100
+    Sections:
+      - Section: section6
+  # Fourth segment has no sections (after removing blob headers).
+  - Type:  0x6ABCDEF0
+    VAddr: 0x2300
+    PAddr: 0x2300
+    Align: 0x100
+    Sections:
+      - Section: blob6
+  # Fifth segment is empty.
+  - Type:   0x6ABCDEF0
+    VAddr:  0x2308
+    PAddr:  0x2308
+    Offset: 0x2308
+
+  # The next few segments test behaviour of fully nested segments.
+  # Sixth segment is the "parent" segment.
+  - Type:  0x6ABCDEF0
+    VAddr: 0x3000
+    PAddr: 0x3000
+    Align: 0x1000
+    Sections:
+      - Section: sectionA
+      - Section: sectionB
+      - Section: blobA
+      - Section: sectionC
+      - Section: blobB
+      - Section: sectionD
+      - Section: blobC
+      - Section: sectionE
+      - Section: blobD
+      - Section: sectionF
+      - Section: blobE
+      - Section: blobF
+      - Section: sectionG
+      - Section: blobG
+      - Section: sectionH
+      - Section: blobH
+      - Section: sectionI
+  # Seventh segment is empty and nested.
+  - Type:   0x6ABCDEF0
+    VAddr:  0x3002
+    PAddr:  0x3002
+    Offset: 0x3002
+  # Eighth segment contains only a section and is nested.
+  - Type:  0x6ABCDEF0
+    VAddr: 0x3004
+    PAddr: 0x3004
+    Sections:
+      - Section: sectionB
+  # Ninth segment contains only unlabelled space and is nested.
+  - Type:  0x6ABCDEF0
+    VAddr: 0x3008
+    PAddr: 0x3008
+    Sections:
+      - Section: blobA
+  # Tenth segment contains two sections with space between and is nested.
+  - Type:  0x6ABCDEF0
+    VAddr: 0x300C
+    PAddr: 0x300C
+    Sections:
+      - Section: sectionC
+      - Section: blobB
+      - Section: sectionD
+  # Eleventh segment contains two sections with space between and at ends and is nested.
+  - Type:  0x6ABCDEF0
+    VAddr: 0x3018
+    PAddr: 0x3018
+    Sections:
+      - Section: blobC
+      - Section: sectionE
+      - Section: blobD
+      - Section: sectionF
+      - Section: blobE
+  # Twelfth segment contains one section with space at ends adjacent to space in parent segment.
+  - Type:     0x6ABCDEF0
+    VAddr:    0x302E
+    PAddr:    0x302E
+    Offset:   0x302E
+    FileSize: 8
+    Sections:
+      - Section: sectionG
+  # Thirteenth segment contains overlaps sections at either end in parent segment.
+  - Type:     0x6ABCDEF0
+    VAddr:    0x303A
+    PAddr:    0x303A
+    Offset:   0x303A
+    FileSize: 0x8
+    Sections:
+      - Section: blobH
+
+  # The next batch of segments are segments that only partially overlap other segments.
+
+  # Segment14: |-unlabelled-|-Sec-|
+  # Segment15:           |--|-Sec-|-unlabelled-|
+  - Type:  0x6ABCDEF0
+    VAddr: 0x4000
+    PAddr: 0x4000
+    Sections:
+      - Section: blobz
+      - Section: sectionz
+  - Type:   0x6ABCDEF0
+    VAddr:  0x4002
+    PAddr:  0x4002
+    Offset: 0x4002
+    Sections:
+      - Section: sectionz
+      - Section: bloby
+
+  # Segment16: |-Sec-|--|
+  # Segment17:    |--|----unlabelled---|
+  - Type:  0x6ABCDEF0
+    VAddr: 0x400C
+    PAddr: 0x400C
+    FileSize: 6
+    Sections:
+      - Section: sectiony
+  - Type:   0x6ABCDEF0
+    VAddr:  0x400E
+    PAddr:  0x400E
+    Offset: 0x400E
+    Sections:
+      - Section: blobx
+
+  # Segment18: |-unlabelled-|-Sec-|
+  # Segment19:              |-Sec-|-unlabelled-|
+  - Type:  0x6ABCDEF0
+    VAddr: 0x4014
+    PAddr: 0x4014
+    Sections:
+      - Section: blobw
+      - Section: sectionx
+  - Type:  0x6ABCDEF0
+    VAddr: 0x4018
+    PAddr: 0x4018
+    Sections:
+      - Section: sectionx
+      - Section: blobv
+
+  # Segment20: |-Sec-|
+  # Segment21:    |--|-unlabelled-|-Sec-|
+  - Type:  0x6ABCDEF0
+    VAddr: 0x4020
+    PAddr: 0x4020
+    Sections:
+      - Section: sectionw
+  - Type:   0x6ABCDEF0
+    VAddr:  0x4022
+    PAddr:  0x4022
+    Offset: 0x4022
+    Sections:
+      - Section: blobu
+      - Section: sectionv
+
+  # Segment22: |-Sec-|
+  # Segment23:    |--|-Sec-|
+  - Type:  0x6ABCDEF0
+    VAddr: 0x402C
+    PAddr: 0x402C
+    Sections:
+      - Section: sectionu
+  - Type:   0x6ABCDEF0
+    VAddr:  0x402E
+    PAddr:  0x402E
+    Offset: 0x402E
+    Sections:
+      - Section: sectiont
+
+  # Segment24: |-unlabelled-|--|
+  # Segment25:              |--Sec--|
+  - Type:  0x6ABCDEF0
+    VAddr: 0x4034
+    PAddr: 0x4034
+    FileSize: 6
+    Sections:
+      - Section: blobt
+  - Type:  0x6ABCDEF0
+    VAddr: 0x4038
+    PAddr: 0x4038
+    Sections:
+      - Section: sections
+
+  # The next batch of segments represent groups of three nested/overlapping segments,
+  # with one parent segment containing two overlapping segments.
+
+  # Segment26: |-unlabelled-|-Sec-|-unlabelled-|
+  # Segment27: |------------|--|
+  # Segment28:              |-Sec-|------------|
+  - Type:  0x6ABCDEF0
+    VAddr: 0x5000
+    PAddr: 0x5000
+    Align: 0x1000
+    Sections:
+      - Section: bloba
+      - Section: sectiona
+      - Section: blobb
+  - Type:  0x6ABCDEF0
+    VAddr: 0x5000
+    PAddr: 0x5000
+    FileSize: 6
+    Sections:
+      - Section: bloba
+  - Type:  0x6ABCDEF0
+    VAddr: 0x5004
+    PAddr: 0x5004
+    Sections:
+      - Section: sectiona
+      - Section: blobb
+
+  # Segment29: |-Sec-|-unlabelled-|-Sec-|
+  # Segment30: |-Sec-|--------|
+  # Segment31:          |---------|-Sec-|
+  - Type:  0x6ABCDEF0
+    VAddr: 0x500C
+    PAddr: 0x500C
+    Sections:
+      - Section: sectionb
+      - Section: blobc
+      - Section: sectionc
+  - Type:  0x6ABCDEF0
+    VAddr: 0x500C
+    PAddr: 0x500C
+    FileSize: 7
+    Sections:
+      - Section: sectionb
+  - Type:   0x6ABCDEF0
+    VAddr:  0x5011
+    PAddr:  0x5011
+    Offset: 0x5011
+    Sections:
+      - Section: sectionc
+
+  # Segment32: |-Sec-|-unlabelled-|-Sec-|
+  # Segment33: |-Sec-|------------|
+  # Segment34:       |------------|-Sec-|
+  - Type:  0x6ABCDEF0
+    VAddr: 0x5018
+    PAddr: 0x5018
+    Sections:
+      - Section: sectiond
+      - Section: blobd
+      - Section: sectione
+  - Type:  0x6ABCDEF0
+    VAddr: 0x5018
+    PAddr: 0x5018
+    Sections:
+      - Section: sectiond
+      - Section: blobd
+  - Type:  0x6ABCDEF0
+    VAddr: 0x501C
+    PAddr: 0x501C
+    Sections:
+      - Section: blobd
+      - Section: sectione
+
+  # Segment35: |-unlabelled-|-Sec-|-unlabelled-|
+  # Segment36: |------------|-Sec-|
+  # Segment37:              |-Sec-|------------|
+  - Type:  0x6ABCDEF0
+    VAddr: 0x5024
+    PAddr: 0x5024
+    Sections:
+      - Section: blobe
+      - Section: sectionf
+      - Section: blobf
+  - Type:  0x6ABCDEF0
+    VAddr: 0x5024
+    PAddr: 0x5024
+    Sections:
+      - Section: blobe
+      - Section: sectionf
+  - Type:  0x6ABCDEF0
+    VAddr: 0x5028
+    PAddr: 0x5028
+    Sections:
+      - Section: sectionf
+      - Section: blobf
+
+  # Segment38: |-unlabelled-|-Sec-|-unlabelled-|
+  # Segment39: |------------|---|
+  # Segment40:                |---|------------|
+  - Type:  0x6ABCDEF0
+    VAddr: 0x5030
+    PAddr: 0x5030
+    Sections:
+      - Section: blobg
+      - Section: sectiong
+      - Section: blobh
+  - Type:  0x6ABCDEF0
+    VAddr: 0x5030
+    PAddr: 0x5030
+    FileSize: 7
+    Sections:
+      - Section: blobg
+  - Type:   0x6ABCDEF0
+    VAddr:  0x5035
+    PAddr:  0x5035
+    Offset: 0x5035
+    Sections:
+      - Section: blobh

tools/llvm-objcopy/ELF/Object.h

private:
 
   void writeEhdr();
   void writePhdr(const Segment &Seg);
   void writeShdr(const SectionBase &Sec);
 
   void writePhdrs();
   void writeShdrs();
   void writeSectionData();
+  void writeSegmentData();
 
   void assignOffsets();
 
   std::unique_ptr<ELFSectionWriter<ELFT>> SecWriter;
 
   size_t totalSize() const;
 
 public:
public:
   uint64_t PAddr;
   uint64_t FileSize;
   uint64_t MemSize;
   uint64_t Align;
 
   uint32_t Index;
   uint64_t OriginalOffset;
   Segment *ParentSegment = nullptr;
+  ArrayRef<uint8_t> Contents;
+
+  explicit Segment(ArrayRef<uint8_t> Data) : Contents(Data) {}
+  Segment() {}
 
   const SectionBase *firstSection() const {
     if (!Sections.empty())
       return *Sections.begin();
     return nullptr;
   }
 
   void removeSection(const SectionBase *Sec) { Sections.erase(Sec); }
   void addSection(const SectionBase *Sec) { Sections.insert(Sec); }
+
+  ArrayRef<uint8_t> getContents() const { return Contents; }
 };
 
 class Section : public SectionBase {
   MAKE_SEC_WRITER_FRIEND
 
   ArrayRef<uint8_t> Contents;
   SectionBase *LinkSection = nullptr;
 

 
 class Object {
 private:
   using SecPtr = std::unique_ptr<SectionBase>;
   using SegPtr = std::unique_ptr<Segment>;
 
   std::vector<SecPtr> Sections;
   std::vector<SegPtr> Segments;
+  std::vector<SecPtr> RemovedSections;
 
 public:
   template <class T>
   using Range = iterator_range<
       pointee_iterator<typename std::vector<std::unique_ptr<T>>::iterator>>;
 
   template <class T>
   using ConstRange = iterator_range<pointee_iterator<
public:
   ConstRange<SectionBase> sections() const {
     return make_pointee_range(Sections);
   }
   SectionBase *findSection(StringRef Name) {
     auto SecIt =
         find_if(Sections, [&](const SecPtr &Sec) { return Sec->Name == Name; });
     return SecIt == Sections.end() ? nullptr : SecIt->get();
   }
+  SectionTableRef removedSections() { return SectionTableRef(RemovedSections); }
+
   Range<Segment> segments() { return make_pointee_range(Segments); }
   ConstRange<Segment> segments() const { return make_pointee_range(Segments); }
 
   Error removeSections(std::function<bool(const SectionBase &)> ToRemove);
   Error removeSymbols(function_ref<bool(const Symbol &)> ToRemove);
   template <class T, class... Ts> T &addSection(Ts &&... Args) {
     auto Sec = llvm::make_unique<T>(std::forward<Ts>(Args)...);
     auto Ptr = Sec.get();
     Sections.emplace_back(std::move(Sec));
     Ptr->Index = Sections.size();
     return *Ptr;
   }
-  Segment &addSegment() {
-    Segments.emplace_back(llvm::make_unique<Segment>());
+  Segment &addSegment(ArrayRef<uint8_t> Data) {
+    Segments.emplace_back(llvm::make_unique<Segment>(Data));
     return *Segments.back();
   }
 };
 
 } // end namespace elf
 } // end namespace objcopy
 } // end namespace llvm
 
 #endif // LLVM_TOOLS_OBJCOPY_OBJECT_H

tools/llvm-objcopy/ELF/Object.cpp

if (&Child != &Parent && segmentOverlapsSegment(Child, Parent)) {
         }
     }
   }
 }
 
 template <class ELFT> void ELFBuilder<ELFT>::readProgramHeaders() {
   uint32_t Index = 0;
   for (const auto &Phdr : unwrapOrError(ElfFile.program_headers())) {
-    Segment &Seg = Obj.addSegment();
+    ArrayRef<uint8_t> Data{ElfFile.base() + Phdr.p_offset,
+                           (size_t)Phdr.p_filesz};
+    Segment &Seg = Obj.addSegment(Data);
     Seg.Type = Phdr.p_type;
     Seg.Flags = Phdr.p_flags;
     Seg.OriginalOffset = Phdr.p_offset;
     Seg.Offset = Phdr.p_offset;
     Seg.VAddr = Phdr.p_vaddr;
     Seg.PAddr = Phdr.p_paddr;
     Seg.FileSize = Phdr.p_filesz;
     Seg.MemSize = Phdr.p_memsz;
template <class ELFT> void ELFWriter<ELFT>::writeShdrs() {
   Shdr.sh_entsize = 0;
 
   for (auto &Sec : Obj.sections())
     writeShdr(Sec);
 }
 
 template <class ELFT> void ELFWriter<ELFT>::writeSectionData() {
   for (auto &Sec : Obj.sections())
+    // Segments are responsible for writing their contents, so only write the
+    // section data if the section is not in a segment. Note that this renders
+    // sections in segments effectively immutable.
+    if (Sec.ParentSegment == nullptr)
       Sec.accept(*SecWriter);
 }
 
+template <class ELFT> void ELFWriter<ELFT>::writeSegmentData() {
+  for (Segment &Seg : Obj.segments()) {
+    uint8_t *B = Buf.getBufferStart() + Seg.Offset;
+    assert(Seg.FileSize == Seg.getContents().size() &&
+           "Segment size must match contents size");
+    std::memcpy(B, Seg.getContents().data(), Seg.FileSize);
+  }
+
+  // Iterate over removed sections and overwrite their old data with zeroes.
+  for (auto &Sec : Obj.removedSections()) {
+    Segment *Parent = Sec.ParentSegment;
+    if (Parent == nullptr || Sec.Type == SHT_NOBITS || Sec.Size == 0)
+      continue;
+    uint64_t Offset =
+        Sec.OriginalOffset - Parent->OriginalOffset + Parent->Offset;
+    uint8_t *B = Buf.getBufferStart();
+    std::memset(B + Offset, 0, Sec.Size);
+  }
+}
+
 Error Object::removeSections(
     std::function<bool(const SectionBase &)> ToRemove) {
 
   auto Iter = std::stable_partition(
       std::begin(Sections), std::end(Sections), [=](const SecPtr &Sec) {
         if (ToRemove(*Sec))
           return false;
         if (auto RelSec = dyn_cast<RelocationSectionBase>(Sec.get())) {
Error Object::removeSections(
   for (auto &KeepSec : make_range(std::begin(Sections), Iter)) {
     if (Error E = KeepSec->removeSectionReferences(
             [&RemoveSections](const SectionBase *Sec) {
               return RemoveSections.find(Sec) != RemoveSections.end();
             }))
       return E;
   }
 
-  // Now finally get rid of them all togethor.
+  // Transfer removed sections into the Object RemovedSections container for use
+  // later.
+  std::move(Iter, Sections.end(), std::back_inserter(RemovedSections));
+  // Now finally get rid of them all together.
   Sections.erase(Iter, std::end(Sections));
   return Error::success();
 }
 
 Error Object::removeSymbols(function_ref<bool(const Symbol &)> ToRemove) {
   if (SymbolTable)
     for (const SecPtr &Sec : Sections)
       if (Error E = Sec->removeSymbols(ToRemove))
template <class ELFT> size_t ELFWriter<ELFT>::totalSize() const {
   auto NullSectionSize = WriteSectionHeaders ? sizeof(Elf_Shdr) : 0;
   return Obj.SHOffset + size(Obj.sections()) * sizeof(Elf_Shdr) +
          NullSectionSize;
 }
 
 template <class ELFT> Error ELFWriter<ELFT>::write() {
   writeEhdr();
   writePhdrs();
+  writeSegmentData();
   writeSectionData();
   if (WriteSectionHeaders)
     writeShdrs();
   return Buf.commit();
 }
 
 template <class ELFT> Error ELFWriter<ELFT>::finalize() {
   // It could happen that SectionNames has been removed and yet the user wants