[PATCH] [ELF][Writer] Refactor Section to not have atoms. Move atoms into AtomSection.
Michael Spencer
bigcheesegs at gmail.com
Wed Feb 13 11:43:31 PST 2013
================
Comment at: lib/ReaderWriter/ELF/Chunk.h:40
@@ -39,2 +39,3 @@
K_ELFSegment, // Segment
K_ELFSection, // Section
+ K_AtomSection, //< A section containing atoms.
----------------
Shankar Kalpathi Easwaran wrote:
> Is K_ELFSection going to be set for sections where the sections dont have atoms in them ?
Yes.
================
Comment at: lib/ReaderWriter/ELF/DefaultLayout.h:159-161
@@ -158,5 +158,5 @@
/// \brief Returns the section to be created
- virtual Section<ELFT> *getSection(StringRef name, const int32_t contentType,
- const int32_t contentPermissions,
- const int32_t sectionOrder);
+ virtual AtomSection<ELFT> *getSection(
+ StringRef name, int32_t contentType,
+ DefinedAtom::ContentPermissions contentPermissions);
----------------
Shankar Kalpathi Easwaran wrote:
> Shankar Kalpathi Easwaran wrote:
> > I was thinking that we could have the function take DefinedAtom here, as the function need to access other attributes of the DefinedAtom in the future.
> This also needs corresponding changes in the ELFTargetHandler.
This can be done in a separate patch.
================
Comment at: lib/ReaderWriter/ELF/DefaultLayout.h:415-431
@@ -418,9 +414,19 @@
template <class ELFT>
-Section<ELFT> *DefaultLayout<ELFT>::getSection(
- StringRef sectionName, int32_t contentType, int32_t permissions,
- int32_t sectionOrder) {
- return new (_allocator) Section<ELFT>(_targetInfo, sectionName, contentType,
- permissions, sectionOrder);
+AtomSection<ELFT> *DefaultLayout<ELFT>::getSection(
+ StringRef sectionName, int32_t contentType,
+ DefinedAtom::ContentPermissions permissions) {
+ const SectionKey sectionKey(sectionName, permissions);
+ auto sec = _sectionMap.find(sectionKey);
+ if (sec != _sectionMap.end())
+ return sec->second;
+ SectionOrder sectionOrder =
+ getSectionOrder(sectionName, contentType, permissions);
+ AtomSection<ELFT> *newSec = new (_allocator) AtomSection<ELFT>(
+ _targetInfo, sectionName, contentType, permissions, sectionOrder);
+ newSec->setOrder(sectionOrder);
+ _sections.push_back(newSec);
+ _sectionMap.insert(std::make_pair(sectionKey, newSec));
+ return newSec;
}
----------------
Shankar Kalpathi Easwaran wrote:
> I would prefer this function just create the section, as targets can just override this function and just create whatever section they want. It would be hard for targets to add sections this way as they need to create the section and add ito the _sections, and the _sectionMap, which makes it harder.
k
================
Comment at: lib/ReaderWriter/ELF/SectionChunks.h:113-147
@@ +112,37 @@
+ _contentType(contentType), _contentPermissions(permissions) {
+ this->setOrder(order);
+ switch (contentType) {
+ case DefinedAtom::typeCode:
+ case DefinedAtom::typeData:
+ case DefinedAtom::typeConstant:
+ case DefinedAtom::typeGOT:
+ case DefinedAtom::typeStub:
+ case DefinedAtom::typeResolver:
+ case DefinedAtom::typeTLVInitialData:
+ this->_type = SHT_PROGBITS;
+ break;
+ case DefinedAtom::typeZeroFill:
+ case DefinedAtom::typeTLVInitialZeroFill:
+ this->_type = SHT_NOBITS;
+ break;
+ }
+
+ switch (permissions) {
+ case DefinedAtom::permR__:
+ this->_flags = SHF_ALLOC;
+ break;
+ case DefinedAtom::permR_X:
+ this->_flags = SHF_ALLOC | SHF_EXECINSTR;
+ break;
+ case DefinedAtom::permRW_:
+ case DefinedAtom::permRW_L:
+ this->_flags = SHF_ALLOC | SHF_WRITE;
+ if (_contentType == DefinedAtom::typeTLVInitialData ||
+ _contentType == DefinedAtom::typeTLVInitialZeroFill)
+ this->_flags |= SHF_TLS;
+ break;
+ case DefinedAtom::permRWX:
+ this->_flags = SHF_ALLOC | SHF_WRITE | SHF_EXECINSTR;
+ break;
+ }
+ }
----------------
Shankar Kalpathi Easwaran wrote:
> Can we have these as two seperate functions, so that targets can override functionality which derive from AtomSection, and they just want to set different flags.
This constructor is run before derived constructors. They can change the flags.
http://llvm-reviews.chandlerc.com/D400
More information about the llvm-commits
mailing list