[ELF] Change (NOLOAD) type mismatch to use SHT_NOBITS instead of SHT_PROGBITS

[MaskRay]

Summary

Placing a non-SHT_NOBITS input section in an output section specified with
(NOLOAD) is fishy but used by some projects. D118840 changed the output type to
SHT_PROGBITS, but using the specified type seems to make more sense and improve
GNU ld compatibility: (NOLOAD) seems to change the output section type
regardless of input.

I think we should keep the current type mismatch warning as it does indicate an
error-prone usage.

[manojgupta]

Thanks fangrui.

[manojgupta]

lld/ELF/OutputSections.cpp

127

please remove tab.

[MaskRay]

lld/ELF/OutputSections.cpp

127

There is no tab. Phabricator may display increasing indent with >>.

[peter.smith]

From doing some reading about NOLOAD, there is a use case for SHT_PROGBITS in NOLOAD. In an embedded system it can represent something like a ROM that is always present on the device at that address. This gets modelled as SHT_NOBITS so that addresses are calculated but not written to the file. Whomever writes the ZI initialization code will obviously have to make sure not to try and write that ZI.

References:
https://community.st.com/s/question/0D50X0000A7UjmzSQC/understanding-of-noload-directive-for-gnuld
https://software-dl.ti.com/ccs/esd/documents/sdto_cgt_linker_special_section_types.html (Strictly speaking this is TI's linker which has some extensions)
https://www.embeddedrelated.com/showthread/comp.arch.embedded/276222-1.php

lld/ELF/OutputSections.cpp

if (hasInputSections || typeIsSet) {
         // them. Other types get an error.
         auto diagnose = type == SHT_NOBITS ? warn : errorOrWarn;
         diagnose("section type mismatch for " + isec->name + "\n>>> " +
                  toString(isec) + ": " +
                  getELFSectionTypeName(config->emachine, isec->type) +
                  "\n>>> output section " + name + ": " +
                  getELFSectionTypeName(config->emachine, type));
       }
+      if (!typeIsSet)
         type = SHT_PROGBITS;
     } else {
       type = isec->type;
     }
   }
   if (!hasInputSections) {
     // If IS is the first section to be added to this section,
     // initialize type, entsize and flags from isec.
     hasInputSections = true;

lld/test/ELF/linkerscript/custom-section-type.s

 # ERR1-NEXT: >>> output section expr: Unknown
 
 # RUN: ld.lld -T %t/a.lds %t/a.o %t/mismatch.o -o %t/mismatch --noinhibit-exec
 # RUN: llvm-readelf -S %t/mismatch | FileCheck %s --check-prefix=MISMATCH
 
 ## Mismatched progbits and expr are changed to SHT_PROGBITS.
 # MISMATCH: progbits PROGBITS
 # MISMATCH: note     NOTE
-# MISMATCH: expr     PROGBITS
+# MISMATCH: expr     0x42: <unknown>
 
 # RUN: not ld.lld -T %t/unknown1.lds %t/a.o -o /dev/null 2>&1 | FileCheck %s --check-prefix=UNKNOWN1
 # UNKNOWN1: error: {{.*}}.lds:1: symbol not found: foo
 
 ## For a symbol named SHT_*, give a better diagnostic.
 # RUN: not ld.lld -T %t/unknown2.lds %t/a.o -o /dev/null 2>&1 | FileCheck %s --check-prefix=UNKNOWN2
 # UNKNOWN2: error: {{.*}}.lds:1: unknown section type SHT_DYNAMIC
 

lld/test/ELF/linkerscript/noload.s

 
 ## The output SHT_PROBITS is contrary to the user expectation of SHT_NOBITS.
 ## Issue a warning. See https://github.com/ClangBuiltLinux/linux/issues/1597
 # RUN: ld.lld --script %t/lds %t.o %t/mismatch.o -o %t/out 2>&1 | FileCheck %s --check-prefix=WARN
 # RUN: llvm-readelf -S -l %t/out | FileCheck %s --check-prefix=CHECK2
 
 # WARN:   warning: section type mismatch for .data_noload_a
 # CHECK2:      Name                 Type     Address          Off               Size
-# CHECK2:      .data_noload_a       PROGBITS 0000000000000000 [[OFF:[0-9a-f]+]] 001001
+# CHECK2:      .data_noload_a       NOBITS   0000000000000000 [[OFF:[0-9a-f]+]] 001001
 
 #--- asm
 .section .text,"ax",@progbits
   nop
 
 .section .data_noload_a,"aw",@nobits
 .zero 4096