// script-sections.cc -- linker script SECTIONS for gold
-// Copyright 2008, 2009 Free Software Foundation, Inc.
+// Copyright 2008, 2009, 2010, 2011 Free Software Foundation, Inc.
// Written by Ian Lance Taylor <iant@google.com>.
// This file is part of gold.
namespace gold
{
+// A region of memory.
+class Memory_region
+{
+ public:
+ Memory_region(const char* name, size_t namelen, unsigned int attributes,
+ Expression* start, Expression* length)
+ : name_(name, namelen),
+ attributes_(attributes),
+ start_(start),
+ length_(length),
+ current_offset_(0),
+ vma_sections_(),
+ lma_sections_(),
+ last_section_(NULL)
+ { }
+
+ // Return the name of this region.
+ const std::string&
+ name() const
+ { return this->name_; }
+
+ // Return the start address of this region.
+ Expression*
+ start_address() const
+ { return this->start_; }
+
+ // Return the length of this region.
+ Expression*
+ length() const
+ { return this->length_; }
+
+ // Print the region (when debugging).
+ void
+ print(FILE*) const;
+
+ // Return true if <name,namelen> matches this region.
+ bool
+ name_match(const char* name, size_t namelen)
+ {
+ return (this->name_.length() == namelen
+ && strncmp(this->name_.c_str(), name, namelen) == 0);
+ }
+
+ Expression*
+ get_current_address() const
+ {
+ return
+ script_exp_binary_add(this->start_,
+ script_exp_integer(this->current_offset_));
+ }
+
+ void
+ increment_offset(std::string section_name, uint64_t amount,
+ const Symbol_table* symtab, const Layout* layout)
+ {
+ this->current_offset_ += amount;
+
+ if (this->current_offset_
+ > this->length_->eval(symtab, layout, false))
+ gold_error(_("section %s overflows end of region %s"),
+ section_name.c_str(), this->name_.c_str());
+ }
+
+ // Returns true iff there is room left in this region
+ // for AMOUNT more bytes of data.
+ bool
+ has_room_for(const Symbol_table* symtab, const Layout* layout,
+ uint64_t amount) const
+ {
+ return (this->current_offset_ + amount
+ < this->length_->eval(symtab, layout, false));
+ }
+
+ // Return true if the provided section flags
+ // are compatible with this region's attributes.
+ bool
+ attributes_compatible(elfcpp::Elf_Xword flags, elfcpp::Elf_Xword type) const;
+
+ void
+ add_section(Output_section_definition* sec, bool vma)
+ {
+ if (vma)
+ this->vma_sections_.push_back(sec);
+ else
+ this->lma_sections_.push_back(sec);
+ }
+
+ typedef std::vector<Output_section_definition*> Section_list;
+
+ // Return the start of the list of sections
+ // whose VMAs are taken from this region.
+ Section_list::const_iterator
+ get_vma_section_list_start() const
+ { return this->vma_sections_.begin(); }
+
+ // Return the start of the list of sections
+ // whose LMAs are taken from this region.
+ Section_list::const_iterator
+ get_lma_section_list_start() const
+ { return this->lma_sections_.begin(); }
+
+ // Return the end of the list of sections
+ // whose VMAs are taken from this region.
+ Section_list::const_iterator
+ get_vma_section_list_end() const
+ { return this->vma_sections_.end(); }
+
+ // Return the end of the list of sections
+ // whose LMAs are taken from this region.
+ Section_list::const_iterator
+ get_lma_section_list_end() const
+ { return this->lma_sections_.end(); }
+
+ Output_section_definition*
+ get_last_section() const
+ { return this->last_section_; }
+
+ void
+ set_last_section(Output_section_definition* sec)
+ { this->last_section_ = sec; }
+
+ private:
+
+ std::string name_;
+ unsigned int attributes_;
+ Expression* start_;
+ Expression* length_;
+ // The offset to the next free byte in the region.
+ // Note - for compatibility with GNU LD we only maintain one offset
+ // regardless of whether the region is being used for VMA values,
+ // LMA values, or both.
+ uint64_t current_offset_;
+ // A list of sections whose VMAs are set inside this region.
+ Section_list vma_sections_;
+ // A list of sections whose LMAs are set inside this region.
+ Section_list lma_sections_;
+ // The latest section to make use of this region.
+ Output_section_definition* last_section_;
+};
+
+// Return true if the provided section flags
+// are compatible with this region's attributes.
+
+bool
+Memory_region::attributes_compatible(elfcpp::Elf_Xword flags,
+ elfcpp::Elf_Xword type) const
+{
+ unsigned int attrs = this->attributes_;
+
+ // No attributes means that this region is not compatible with anything.
+ if (attrs == 0)
+ return false;
+
+ bool match = true;
+ do
+ {
+ switch (attrs & - attrs)
+ {
+ case MEM_EXECUTABLE:
+ if ((flags & elfcpp::SHF_EXECINSTR) == 0)
+ match = false;
+ break;
+
+ case MEM_WRITEABLE:
+ if ((flags & elfcpp::SHF_WRITE) == 0)
+ match = false;
+ break;
+
+ case MEM_READABLE:
+ // All sections are presumed readable.
+ break;
+
+ case MEM_ALLOCATABLE:
+ if ((flags & elfcpp::SHF_ALLOC) == 0)
+ match = false;
+ break;
+
+ case MEM_INITIALIZED:
+ if ((type & elfcpp::SHT_NOBITS) != 0)
+ match = false;
+ break;
+ }
+ attrs &= ~ (attrs & - attrs);
+ }
+ while (attrs != 0);
+
+ return match;
+}
+
+// Print a memory region.
+
+void
+Memory_region::print(FILE* f) const
+{
+ fprintf(f, " %s", this->name_.c_str());
+
+ unsigned int attrs = this->attributes_;
+ if (attrs != 0)
+ {
+ fprintf(f, " (");
+ do
+ {
+ switch (attrs & - attrs)
+ {
+ case MEM_EXECUTABLE: fputc('x', f); break;
+ case MEM_WRITEABLE: fputc('w', f); break;
+ case MEM_READABLE: fputc('r', f); break;
+ case MEM_ALLOCATABLE: fputc('a', f); break;
+ case MEM_INITIALIZED: fputc('i', f); break;
+ default:
+ gold_unreachable();
+ }
+ attrs &= ~ (attrs & - attrs);
+ }
+ while (attrs != 0);
+ fputc(')', f);
+ }
+
+ fprintf(f, " : origin = ");
+ this->start_->print(f);
+ fprintf(f, ", length = ");
+ this->length_->print(f);
+ fprintf(f, "\n");
+}
+
// Manage orphan sections. This is intended to be largely compatible
// with the GNU linker. The Linux kernel implicitly relies on
// something similar to the GNU linker's orphan placement. We
// Output_section_definition.
virtual const char*
output_section_name(const char*, const char*, Output_section***,
- Script_sections::Section_type*)
+ Script_sections::Section_type*, bool*)
{ return NULL; }
// Initialize OSP with an output section.
{ }
// Set section addresses. This includes applying assignments if the
- // the expression is an absolute value.
+ // expression is an absolute value.
virtual void
set_section_addresses(Symbol_table*, Layout*, uint64_t*, uint64_t*,
uint64_t*)
get_output_section() const
{ return NULL; }
+ // Set the section's memory regions.
+ virtual void
+ set_memory_region(Memory_region*, bool)
+ { gold_error(_("Attempt to set a memory region for a non-output section")); }
+
// Print the element for debugging purposes.
virtual void
print(FILE* f) const = 0;
set_section_addresses(Symbol_table* symtab, Layout* layout,
uint64_t* dot_value, uint64_t*, uint64_t*)
{
- this->assignment_.set_if_absolute(symtab, layout, true, *dot_value);
+ this->assignment_.set_if_absolute(symtab, layout, true, *dot_value, NULL);
}
// Print for debugging.
// We ignore the section of the result because outside of an
// output section definition the dot symbol is always considered
// to be absolute.
- Output_section* dummy;
*dot_value = this->val_->eval_with_dot(symtab, layout, true, *dot_value,
- NULL, &dummy, NULL);
+ NULL, NULL, NULL, false);
}
// Update the dot symbol while setting section addresses.
uint64_t* dot_value, uint64_t* dot_alignment,
uint64_t* load_address)
{
- Output_section* dummy;
*dot_value = this->val_->eval_with_dot(symtab, layout, false, *dot_value,
- NULL, &dummy, dot_alignment);
+ NULL, NULL, dot_alignment, false);
*load_address = *dot_value;
}
// Return whether this element matches FILE_NAME and SECTION_NAME.
// The only real implementation is in Output_section_element_input.
virtual bool
- match_name(const char*, const char*) const
+ match_name(const char*, const char*, bool *) const
{ return false; }
// Set section addresses. This includes applying assignments if the
- // the expression is an absolute value.
+ // expression is an absolute value.
virtual void
set_section_addresses(Symbol_table*, Layout*, Output_section*, uint64_t,
uint64_t*, uint64_t*, Output_section**, std::string*,
void
set_section_addresses(Symbol_table* symtab, Layout* layout, Output_section*,
uint64_t, uint64_t* dot_value, uint64_t*,
- Output_section**, std::string*, Input_section_list*)
+ Output_section** dot_section, std::string*,
+ Input_section_list*)
{
- this->assignment_.set_if_absolute(symtab, layout, true, *dot_value);
+ this->assignment_.set_if_absolute(symtab, layout, true, *dot_value,
+ *dot_section);
}
// Print for debugging.
: val_(val)
{ }
+ // An assignment to dot within an output section is enough to force
+ // the output section to exist.
+ bool
+ needs_output_section() const
+ { return true; }
+
// Finalize the symbol.
void
finalize_symbols(Symbol_table* symtab, const Layout* layout,
uint64_t* dot_value, Output_section** dot_section)
{
*dot_value = this->val_->eval_with_dot(symtab, layout, true, *dot_value,
- *dot_section, dot_section, NULL);
+ *dot_section, dot_section, NULL,
+ true);
}
// Update the dot symbol while setting section addresses.
void
set_section_addresses(Symbol_table* symtab, Layout* layout, Output_section*,
uint64_t, uint64_t* dot_value, uint64_t*,
- Output_section**, std::string*, Input_section_list*);
+ Output_section** dot_section, std::string*,
+ Input_section_list*);
// Print for debugging.
void
{
uint64_t next_dot = this->val_->eval_with_dot(symtab, layout, false,
*dot_value, *dot_section,
- dot_section, dot_alignment);
+ dot_section, dot_alignment,
+ true);
if (next_dot < *dot_value)
gold_error(_("dot may not move backward"));
if (next_dot > *dot_value && output_section != NULL)
void
Output_data_expression::do_write_to_buffer(unsigned char* buf)
{
- Output_section* dummy;
uint64_t val = this->val_->eval_with_dot(this->symtab_, this->layout_,
true, this->dot_value_,
- this->dot_section_, &dummy, NULL);
+ this->dot_section_, NULL, NULL,
+ false);
if (parameters->target().is_big_endian())
this->endian_write_to_buffer<true>(val, buf);
Output_section* fill_section;
uint64_t fill_val = this->val_->eval_with_dot(symtab, layout, false,
*dot_value, *dot_section,
- &fill_section, NULL);
+ &fill_section, NULL, false);
if (fill_section != NULL)
gold_warning(_("fill value is not absolute"));
// FIXME: The GNU linker supports fill values of arbitrary length.
Expression* val_;
};
-// Return whether STRING contains a wildcard character. This is used
-// to speed up matching.
-
-static inline bool
-is_wildcard_string(const std::string& s)
-{
- return strpbrk(s.c_str(), "?*[") != NULL;
-}
-
// An input section specification in an output section
class Output_section_element_input : public Output_section_element
*dot_section = this->final_dot_section_;
}
- // See whether we match FILE_NAME and SECTION_NAME as an input
- // section.
+ // See whether we match FILE_NAME and SECTION_NAME as an input section.
+ // If we do then also indicate whether the section should be KEPT.
bool
- match_name(const char* file_name, const char* section_name) const;
+ match_name(const char* file_name, const char* section_name, bool* keep) const;
// Set the section address.
void
Input_section_pattern(const char* patterna, size_t patternlena,
Sort_wildcard sorta)
: pattern(patterna, patternlena),
- pattern_is_wildcard(is_wildcard_string(this->pattern)),
+ pattern_is_wildcard(is_wildcard_string(this->pattern.c_str())),
sort(sorta)
{ }
};
if (spec->file.name.length != 1 || spec->file.name.value[0] != '*')
this->filename_pattern_.assign(spec->file.name.value,
spec->file.name.length);
- this->filename_is_wildcard_ = is_wildcard_string(this->filename_pattern_);
+ this->filename_is_wildcard_ = is_wildcard_string(this->filename_pattern_.c_str());
if (spec->input_sections.exclude != NULL)
{
p != spec->input_sections.exclude->end();
++p)
{
- bool is_wildcard = is_wildcard_string(*p);
+ bool is_wildcard = is_wildcard_string((*p).c_str());
this->filename_exclusions_.push_back(std::make_pair(*p,
is_wildcard));
}
return true;
}
-// See whether we match FILE_NAME and SECTION_NAME.
+// See whether we match FILE_NAME and SECTION_NAME. If we do then
+// KEEP indicates whether the section should survive garbage collection.
bool
Output_section_element_input::match_name(const char* file_name,
- const char* section_name) const
+ const char* section_name,
+ bool *keep) const
{
if (!this->match_file_name(file_name))
return false;
+ *keep = this->keep_;
+
// If there are no section name patterns, then we match.
if (this->input_section_patterns_.empty())
return true;
isi.set_section_name(relobj->section_name(shndx));
if (p->is_relaxed_input_section())
{
- // We use current data size because relxed section sizes may not
+ // We use current data size because relaxed section sizes may not
// have finalized yet.
isi.set_size(p->relaxed_input_section()->current_data_size());
isi.set_addralign(p->relaxed_input_section()->addralign());
// section name.
const char*
output_section_name(const char* file_name, const char* section_name,
- Output_section***, Script_sections::Section_type*);
+ Output_section***, Script_sections::Section_type*,
+ bool*);
// Initialize OSP with an output section.
void
Script_sections::Section_type
section_type() const;
+ // Store the memory region to use.
+ void
+ set_memory_region(Memory_region*, bool set_vma);
+
+ void
+ set_section_vma(Expression* address)
+ { this->address_ = address; }
+
+ void
+ set_section_lma(Expression* address)
+ { this->load_address_ = address; }
+
+ const std::string&
+ get_section_name() const
+ { return this->name_; }
+
private:
static const char*
script_section_type_name(Script_section_type);
uint64_t address = *dot_value;
if (this->address_ != NULL)
{
- Output_section* dummy;
address = this->address_->eval_with_dot(symtab, layout, true,
*dot_value, NULL,
- &dummy, NULL);
+ NULL, NULL, false);
}
if (this->align_ != NULL)
{
- Output_section* dummy;
uint64_t align = this->align_->eval_with_dot(symtab, layout, true,
- *dot_value,
- NULL,
- &dummy, NULL);
+ *dot_value, NULL,
+ NULL, NULL, false);
address = align_address(address, align);
}
*dot_value = address;
const char* file_name,
const char* section_name,
Output_section*** slot,
- Script_sections::Section_type *psection_type)
+ Script_sections::Section_type* psection_type,
+ bool* keep)
{
// Ask each element whether it matches NAME.
for (Output_section_elements::const_iterator p = this->elements_.begin();
p != this->elements_.end();
++p)
{
- if ((*p)->match_name(file_name, section_name))
+ if ((*p)->match_name(file_name, section_name, keep))
{
// We found a match for NAME, which means that it should go
// into this output section.
return NULL;
}
+// Return true if memory from START to START + LENGTH is contained
+// within a memory region.
+
+bool
+Script_sections::block_in_region(Symbol_table* symtab, Layout* layout,
+ uint64_t start, uint64_t length) const
+{
+ if (this->memory_regions_ == NULL)
+ return false;
+
+ for (Memory_regions::const_iterator mr = this->memory_regions_->begin();
+ mr != this->memory_regions_->end();
+ ++mr)
+ {
+ uint64_t s = (*mr)->start_address()->eval(symtab, layout, false);
+ uint64_t l = (*mr)->length()->eval(symtab, layout, false);
+
+ if (s <= start
+ && (s + l) >= (start + length))
+ return true;
+ }
+
+ return false;
+}
+
+// Find a memory region that should be used by a given output SECTION.
+// If provided set PREVIOUS_SECTION_RETURN to point to the last section
+// that used the return memory region.
+
+Memory_region*
+Script_sections::find_memory_region(
+ Output_section_definition* section,
+ bool find_vma_region,
+ Output_section_definition** previous_section_return)
+{
+ if (previous_section_return != NULL)
+ * previous_section_return = NULL;
+
+ // Walk the memory regions specified in this script, if any.
+ if (this->memory_regions_ == NULL)
+ return NULL;
+
+ // The /DISCARD/ section never gets assigned to any region.
+ if (section->get_section_name() == "/DISCARD/")
+ return NULL;
+
+ Memory_region* first_match = NULL;
+
+ // First check to see if a region has been assigned to this section.
+ for (Memory_regions::const_iterator mr = this->memory_regions_->begin();
+ mr != this->memory_regions_->end();
+ ++mr)
+ {
+ if (find_vma_region)
+ {
+ for (Memory_region::Section_list::const_iterator s =
+ (*mr)->get_vma_section_list_start();
+ s != (*mr)->get_vma_section_list_end();
+ ++s)
+ if ((*s) == section)
+ {
+ (*mr)->set_last_section(section);
+ return *mr;
+ }
+ }
+ else
+ {
+ for (Memory_region::Section_list::const_iterator s =
+ (*mr)->get_lma_section_list_start();
+ s != (*mr)->get_lma_section_list_end();
+ ++s)
+ if ((*s) == section)
+ {
+ (*mr)->set_last_section(section);
+ return *mr;
+ }
+ }
+
+ // Make a note of the first memory region whose attributes
+ // are compatible with the section. If we do not find an
+ // explicit region assignment, then we will return this region.
+ Output_section* out_sec = section->get_output_section();
+ if (first_match == NULL
+ && out_sec != NULL
+ && (*mr)->attributes_compatible(out_sec->flags(),
+ out_sec->type()))
+ first_match = *mr;
+ }
+
+ // With LMA computations, if an explicit region has not been specified then
+ // we will want to set the difference between the VMA and the LMA of the
+ // section were searching for to be the same as the difference between the
+ // VMA and LMA of the last section to be added to first matched region.
+ // Hence, if it was asked for, we return a pointer to the last section
+ // known to be used by the first matched region.
+ if (first_match != NULL
+ && previous_section_return != NULL)
+ *previous_section_return = first_match->get_last_section();
+
+ return first_match;
+}
+
// Set the section address. Note that the OUTPUT_SECTION_ field will
// be NULL if no input sections were mapped to this output section.
// We still have to adjust dot and process symbol assignments.
uint64_t* dot_alignment,
uint64_t* load_address)
{
+ Memory_region* vma_region = NULL;
+ Memory_region* lma_region = NULL;
+ Script_sections* script_sections =
+ layout->script_options()->script_sections();
uint64_t address;
uint64_t old_dot_value = *dot_value;
uint64_t old_load_address = *load_address;
- // Check for --section-start.
- bool is_address_set = false;
- if (this->output_section_ != NULL)
- is_address_set =
- parameters->options().section_start(this->output_section_->name(),
- &address);
- if (!is_address_set)
+ // If input section sorting is requested via --section-ordering-file or
+ // linker plugins, then do it here. This is important because we want
+ // any sorting specified in the linker scripts, which will be done after
+ // this, to take precedence. The final order of input sections is then
+ // guaranteed to be according to the linker script specification.
+ if (this->output_section_ != NULL
+ && this->output_section_->input_section_order_specified())
+ this->output_section_->sort_attached_input_sections();
+
+ // Decide the start address for the section. The algorithm is:
+ // 1) If an address has been specified in a linker script, use that.
+ // 2) Otherwise if a memory region has been specified for the section,
+ // use the next free address in the region.
+ // 3) Otherwise if memory regions have been specified find the first
+ // region whose attributes are compatible with this section and
+ // install it into that region.
+ // 4) Otherwise use the current location counter.
+
+ if (this->output_section_ != NULL
+ // Check for --section-start.
+ && parameters->options().section_start(this->output_section_->name(),
+ &address))
+ ;
+ else if (this->address_ == NULL)
{
- if (this->address_ == NULL)
- address = *dot_value;
+ vma_region = script_sections->find_memory_region(this, true, NULL);
+
+ if (vma_region != NULL)
+ address = vma_region->get_current_address()->eval(symtab, layout,
+ false);
else
- {
- Output_section* dummy;
- address = this->address_->eval_with_dot(symtab, layout, true,
- *dot_value, NULL, &dummy,
- dot_alignment);
- }
+ address = *dot_value;
}
-
+ else
+ address = this->address_->eval_with_dot(symtab, layout, true,
+ *dot_value, NULL, NULL,
+ dot_alignment, false);
uint64_t align;
if (this->align_ == NULL)
{
{
Output_section* align_section;
align = this->align_->eval_with_dot(symtab, layout, true, *dot_value,
- NULL, &align_section, NULL);
+ NULL, &align_section, NULL, false);
if (align_section != NULL)
gold_warning(_("alignment of section %s is not absolute"),
this->name_.c_str());
this->evaluated_address_ = address;
this->evaluated_addralign_ = align;
+ uint64_t laddr;
+
if (this->load_address_ == NULL)
- this->evaluated_load_address_ = address;
+ {
+ Output_section_definition* previous_section;
+
+ // Determine if an LMA region has been set for this section.
+ lma_region = script_sections->find_memory_region(this, false,
+ &previous_section);
+
+ if (lma_region != NULL)
+ {
+ if (previous_section == NULL)
+ // The LMA address was explicitly set to the given region.
+ laddr = lma_region->get_current_address()->eval(symtab, layout,
+ false);
+ else
+ {
+ // We are not going to use the discovered lma_region, so
+ // make sure that we do not update it in the code below.
+ lma_region = NULL;
+
+ if (this->address_ != NULL || previous_section == this)
+ {
+ // Either an explicit VMA address has been set, or an
+ // explicit VMA region has been set, so set the LMA equal to
+ // the VMA.
+ laddr = address;
+ }
+ else
+ {
+ // The LMA address was not explicitly or implicitly set.
+ //
+ // We have been given the first memory region that is
+ // compatible with the current section and a pointer to the
+ // last section to use this region. Set the LMA of this
+ // section so that the difference between its' VMA and LMA
+ // is the same as the difference between the VMA and LMA of
+ // the last section in the given region.
+ laddr = address + (previous_section->evaluated_load_address_
+ - previous_section->evaluated_address_);
+ }
+ }
+
+ if (this->output_section_ != NULL)
+ this->output_section_->set_load_address(laddr);
+ }
+ else
+ {
+ // Do not set the load address of the output section, if one exists.
+ // This allows future sections to determine what the load address
+ // should be. If none is ever set, it will default to being the
+ // same as the vma address.
+ laddr = address;
+ }
+ }
else
{
- Output_section* dummy;
- uint64_t laddr =
- this->load_address_->eval_with_dot(symtab, layout, true, *dot_value,
- this->output_section_, &dummy,
- NULL);
+ laddr = this->load_address_->eval_with_dot(symtab, layout, true,
+ *dot_value,
+ this->output_section_,
+ NULL, NULL, false);
if (this->output_section_ != NULL)
this->output_section_->set_load_address(laddr);
- this->evaluated_load_address_ = laddr;
}
+ this->evaluated_load_address_ = laddr;
+
uint64_t subalign;
if (this->subalign_ == NULL)
subalign = 0;
Output_section* subalign_section;
subalign = this->subalign_->eval_with_dot(symtab, layout, true,
*dot_value, NULL,
- &subalign_section, NULL);
+ &subalign_section, NULL,
+ false);
if (subalign_section != NULL)
gold_warning(_("subalign of section %s is not absolute"),
this->name_.c_str());
uint64_t fill_val = this->fill_->eval_with_dot(symtab, layout, true,
*dot_value,
NULL, &fill_section,
- NULL);
+ NULL, false);
if (fill_section != NULL)
gold_warning(_("fill of section %s is not absolute"),
this->name_.c_str());
gold_assert(input_sections.empty());
- if (this->load_address_ == NULL || this->output_section_ == NULL)
+ if (vma_region != NULL)
+ {
+ // Update the VMA region being used by the section now that we know how
+ // big it is. Use the current address in the region, rather than
+ // start_address because that might have been aligned upwards and we
+ // need to allow for the padding.
+ Expression* addr = vma_region->get_current_address();
+ uint64_t size = *dot_value - addr->eval(symtab, layout, false);
+
+ vma_region->increment_offset(this->get_section_name(), size,
+ symtab, layout);
+ }
+
+ // If the LMA region is different from the VMA region, then increment the
+ // offset there as well. Note that we use the same "dot_value -
+ // start_address" formula that is used in the load_address assignment below.
+ if (lma_region != NULL && lma_region != vma_region)
+ lma_region->increment_offset(this->get_section_name(),
+ *dot_value - start_address,
+ symtab, layout);
+
+ // Compute the load address for the following section.
+ if (this->output_section_ == NULL)
*load_address = *dot_value;
+ else if (this->load_address_ == NULL)
+ {
+ if (lma_region == NULL)
+ *load_address = *dot_value;
+ else
+ *load_address =
+ lma_region->get_current_address()->eval(symtab, layout, false);
+ }
else
*load_address = (this->output_section_->load_address()
+ (*dot_value - start_address));
// Return the name of a script section type.
const char*
-Output_section_definition::script_section_type_name (
+Output_section_definition::script_section_type_name(
Script_section_type script_section_type)
{
switch (script_section_type)
}
}
+void
+Output_section_definition::set_memory_region(Memory_region* mr, bool set_vma)
+{
+ gold_assert(mr != NULL);
+ // Add the current section to the specified region's list.
+ mr->add_section(this, set_vma);
+}
+
// An output section created to hold orphaned input sections. These
// do not actually appear in linker scripts. However, for convenience
// when setting the output section addresses, we put a marker to these
uint64_t address = *dot_value;
address = align_address(address, this->os_->addralign());
+ // If input section sorting is requested via --section-ordering-file or
+ // linker plugins, then do it here. This is important because we want
+ // any sorting specified in the linker scripts, which will be done after
+ // this, to take precedence. The final order of input sections is then
+ // guaranteed to be according to the linker script specification.
+ if (this->os_ != NULL
+ && this->os_->input_section_order_specified())
+ this->os_->sort_attached_input_sections();
+
+ // For a relocatable link, all orphan sections are put at
+ // address 0. In general we expect all sections to be at
+ // address 0 for a relocatable link, but we permit the linker
+ // script to override that for specific output sections.
+ if (parameters->options().relocatable())
+ {
+ address = 0;
+ *load_address = 0;
+ have_load_address = false;
+ }
+
if ((this->os_->flags() & elfcpp::SHF_ALLOC) != 0)
{
this->os_->set_address(address);
address += size;
}
- // An SHF_TLS/SHT_NOBITS section does not take up any address space.
- if (this->os_ == NULL
- || (this->os_->flags() & elfcpp::SHF_TLS) == 0
- || this->os_->type() != elfcpp::SHT_NOBITS)
+ if (parameters->options().relocatable())
{
+ // For a relocatable link, reset DOT_VALUE to 0.
+ *dot_value = 0;
+ *load_address = 0;
+ }
+ else if (this->os_ == NULL
+ || (this->os_->flags() & elfcpp::SHF_TLS) == 0
+ || this->os_->type() != elfcpp::SHT_NOBITS)
+ {
+ // An SHF_TLS/SHT_NOBITS section does not take up any address space.
if (!have_load_address)
*load_address = address;
else
fprintf(f, ";\n");
}
+// Add a memory region.
+
+void
+Script_sections::add_memory_region(const char* name, size_t namelen,
+ unsigned int attributes,
+ Expression* start, Expression* length)
+{
+ if (this->memory_regions_ == NULL)
+ this->memory_regions_ = new Memory_regions();
+ else if (this->find_memory_region(name, namelen))
+ {
+ gold_error(_("region '%.*s' already defined"), static_cast<int>(namelen),
+ name);
+ // FIXME: Add a GOLD extension to allow multiple regions with the same
+ // name. This would amount to a single region covering disjoint blocks
+ // of memory, which is useful for embedded devices.
+ }
+
+ // FIXME: Check the length and start values. Currently we allow
+ // non-constant expressions for these values, whereas LD does not.
+
+ // FIXME: Add a GOLD extension to allow NEGATIVE LENGTHS. This would
+ // describe a region that packs from the end address going down, rather
+ // than the start address going up. This would be useful for embedded
+ // devices.
+
+ this->memory_regions_->push_back(new Memory_region(name, namelen, attributes,
+ start, length));
+}
+
+// Find a memory region.
+
+Memory_region*
+Script_sections::find_memory_region(const char* name, size_t namelen)
+{
+ if (this->memory_regions_ == NULL)
+ return NULL;
+
+ for (Memory_regions::const_iterator m = this->memory_regions_->begin();
+ m != this->memory_regions_->end();
+ ++m)
+ if ((*m)->name_match(name, namelen))
+ return *m;
+
+ return NULL;
+}
+
+// Find a memory region's origin.
+
+Expression*
+Script_sections::find_memory_region_origin(const char* name, size_t namelen)
+{
+ Memory_region* mr = find_memory_region(name, namelen);
+ if (mr == NULL)
+ return NULL;
+
+ return mr->start_address();
+}
+
+// Find a memory region's length.
+
+Expression*
+Script_sections::find_memory_region_length(const char* name, size_t namelen)
+{
+ Memory_region* mr = find_memory_region(name, namelen);
+ if (mr == NULL)
+ return NULL;
+
+ return mr->length();
+}
+
+// Set the memory region to use for the current section.
+
+void
+Script_sections::set_memory_region(Memory_region* mr, bool set_vma)
+{
+ gold_assert(!this->sections_elements_->empty());
+ this->sections_elements_->back()->set_memory_region(mr, set_vma);
+}
+
// Class Script_sections.
Script_sections::Script_sections()
in_sections_clause_(false),
sections_elements_(NULL),
output_section_(NULL),
+ memory_regions_(NULL),
phdrs_elements_(NULL),
orphan_section_placement_(NULL),
data_segment_align_start_(),
Script_sections::start_output_section(
const char* name,
size_t namelen,
- const Parser_output_section_header *header)
+ const Parser_output_section_header* header)
{
Output_section_definition* posd = new Output_section_definition(name,
namelen,
const char* file_name,
const char* section_name,
Output_section*** output_section_slot,
- Script_sections::Section_type *psection_type)
+ Script_sections::Section_type* psection_type,
+ bool* keep)
{
for (Sections_elements::const_iterator p = this->sections_elements_->begin();
p != this->sections_elements_->end();
{
const char* ret = (*p)->output_section_name(file_name, section_name,
output_section_slot,
- psection_type);
+ psection_type, keep);
if (ret != NULL)
{
Script_sections::set_section_addresses(Symbol_table* symtab, Layout* layout)
{
gold_assert(this->saw_sections_clause_);
-
+
// Implement ONLY_IF_RO/ONLY_IF_RW constraints. These are a pain
// for our representation.
for (Sections_elements::iterator p = this->sections_elements_->begin();
p != this->sections_elements_->end();
++p)
{
- Output_section *os = (*p)->get_output_section();
+ Output_section* os = (*p)->get_output_section();
if (os != NULL && (os->flags() & elfcpp::SHF_TLS) != 0)
{
if (first_tls == NULL)
class Sort_output_sections
{
public:
+ Sort_output_sections(const Script_sections::Sections_elements* elements)
+ : elements_(elements)
+ { }
+
bool
operator()(const Output_section* os1, const Output_section* os2) const;
+
+ private:
+ int
+ script_compare(const Output_section* os1, const Output_section* os2) const;
+
+ private:
+ const Script_sections::Sections_elements* elements_;
};
bool
if (os1->address() != os2->address())
return os1->address() < os2->address();
- // Sort TLS sections to the end.
+ // If the linker script says which of these sections is first, go
+ // with what it says.
+ int i = this->script_compare(os1, os2);
+ if (i != 0)
+ return i < 0;
+
+ // Sort PROGBITS before NOBITS.
+ bool nobits1 = os1->type() == elfcpp::SHT_NOBITS;
+ bool nobits2 = os2->type() == elfcpp::SHT_NOBITS;
+ if (nobits1 != nobits2)
+ return nobits2;
+
+ // Sort PROGBITS TLS sections to the end, NOBITS TLS sections to the
+ // beginning.
bool tls1 = (os1->flags() & elfcpp::SHF_TLS) != 0;
bool tls2 = (os2->flags() & elfcpp::SHF_TLS) != 0;
if (tls1 != tls2)
- return tls2;
-
- // Sort PROGBITS before NOBITS.
- if (os1->type() == elfcpp::SHT_PROGBITS && os2->type() == elfcpp::SHT_NOBITS)
- return true;
- if (os1->type() == elfcpp::SHT_NOBITS && os2->type() == elfcpp::SHT_PROGBITS)
- return false;
+ return nobits1 ? tls1 : tls2;
// Sort non-NOLOAD before NOLOAD.
if (os1->is_noload() && !os2->is_noload())
return true;
if (!os1->is_noload() && os2->is_noload())
return true;
-
- // Otherwise we don't care.
- return false;
+
+ // The sections seem practically identical. Sort by name to get a
+ // stable sort.
+ return os1->name() < os2->name();
+}
+
+// Return -1 if OS1 comes before OS2 in ELEMENTS_, 1 if comes after, 0
+// if either OS1 or OS2 is not mentioned. This ensures that we keep
+// empty sections in the order in which they appear in a linker
+// script.
+
+int
+Sort_output_sections::script_compare(const Output_section* os1,
+ const Output_section* os2) const
+{
+ if (this->elements_ == NULL)
+ return 0;
+
+ bool found_os1 = false;
+ bool found_os2 = false;
+ for (Script_sections::Sections_elements::const_iterator
+ p = this->elements_->begin();
+ p != this->elements_->end();
+ ++p)
+ {
+ if (os2 == (*p)->get_output_section())
+ {
+ if (found_os1)
+ return -1;
+ found_os2 = true;
+ }
+ else if (os1 == (*p)->get_output_section())
+ {
+ if (found_os2)
+ return 1;
+ found_os1 = true;
+ }
+ }
+
+ return 0;
}
// Return whether OS is a BSS section. This is a SHT_NOBITS section.
return file_header_size + segment_headers_size;
}
-// Return the amount we have to subtract from the LMA to accomodate
+// Return the amount we have to subtract from the LMA to accommodate
// headers of the given size. The complication is that the file
// header have to be at the start of a page, as otherwise it will not
// be at the start of the file.
layout->get_allocated_sections(§ions);
// Sort the sections by address.
- std::stable_sort(sections.begin(), sections.end(), Sort_output_sections());
+ std::stable_sort(sections.begin(), sections.end(),
+ Sort_output_sections(this->sections_elements_));
this->create_note_and_tls_segments(layout, §ions);
is_current_seg_readonly = true;
}
- current_seg->add_output_section(*p, seg_flags, false);
+ current_seg->add_output_section_to_load(layout, *p, seg_flags);
if (((*p)->flags() & elfcpp::SHF_WRITE) != 0)
is_current_seg_readonly = false;
if (lma < subtract || vma < subtract)
return NULL;
+ // If memory regions have been specified and the address range
+ // we are about to use is not contained within any region then
+ // issue a warning message about the segment we are going to
+ // create. It will be outside of any region and so possibly
+ // using non-existent or protected memory. We test LMA rather
+ // than VMA since we assume that the headers will never be
+ // relocated.
+ if (this->memory_regions_ != NULL
+ && !this->block_in_region (NULL, layout, lma - subtract, subtract))
+ gold_warning(_("creating a segment to contain the file and program"
+ " headers outside of any MEMORY region"));
+
Output_segment* load_seg = layout->make_output_segment(elfcpp::PT_LOAD,
elfcpp::PF_R);
load_seg->set_addresses(vma - subtract, lma - subtract);
Layout::section_flags_to_segment((*p)->flags());
Output_segment* oseg = layout->make_output_segment(elfcpp::PT_NOTE,
seg_flags);
- oseg->add_output_section(*p, seg_flags, false);
+ oseg->add_output_section_to_nonload(*p, seg_flags);
// Incorporate any subsequent SHT_NOTE sections, in the
// hopes that the script is sensible.
&& (*pnext)->type() == elfcpp::SHT_NOTE)
{
seg_flags = Layout::section_flags_to_segment((*pnext)->flags());
- oseg->add_output_section(*pnext, seg_flags, false);
+ oseg->add_output_section_to_nonload(*pnext, seg_flags);
p = pnext;
++pnext;
}
Layout::section_flags_to_segment((*p)->flags());
Output_segment* oseg = layout->make_output_segment(elfcpp::PT_TLS,
seg_flags);
- oseg->add_output_section(*p, seg_flags, false);
+ oseg->add_output_section_to_nonload(*p, seg_flags);
Layout::Section_list::const_iterator pnext = p + 1;
while (pnext != sections->end()
&& ((*pnext)->flags() & elfcpp::SHF_TLS) != 0)
{
seg_flags = Layout::section_flags_to_segment((*pnext)->flags());
- oseg->add_output_section(*pnext, seg_flags, false);
+ oseg->add_output_section_to_nonload(*pnext, seg_flags);
p = pnext;
++pnext;
}
saw_tls = true;
}
+
+ // If we are making a shared library, and we see a section named
+ // .interp then put the .interp section in a PT_INTERP segment.
+ // This is for GNU ld compatibility.
+ if (strcmp((*p)->name(), ".interp") == 0)
+ {
+ elfcpp::Elf_Word seg_flags =
+ Layout::section_flags_to_segment((*p)->flags());
+ Output_segment* oseg = layout->make_output_segment(elfcpp::PT_INTERP,
+ seg_flags);
+ oseg->add_output_section_to_nonload(*p, seg_flags);
+ }
}
}
p != this->sections_elements_->end();
++p)
{
- bool orphan;
+ bool is_orphan;
String_list* old_phdr_names = phdr_names;
- Output_section* os = (*p)->allocate_to_segment(&phdr_names, &orphan);
+ Output_section* os = (*p)->allocate_to_segment(&phdr_names, &is_orphan);
if (os == NULL)
continue;
+ elfcpp::Elf_Word seg_flags =
+ Layout::section_flags_to_segment(os->flags());
+
if (phdr_names == NULL)
{
- gold_error(_("allocated section not in any segment"));
+ // Don't worry about empty orphan sections.
+ if (is_orphan && os->current_data_size() > 0)
+ gold_error(_("allocated section %s not in any segment"),
+ os->name());
+
+ // To avoid later crashes drop this section into the first
+ // PT_LOAD segment.
+ for (Phdrs_elements::const_iterator ppe =
+ this->phdrs_elements_->begin();
+ ppe != this->phdrs_elements_->end();
+ ++ppe)
+ {
+ Output_segment* oseg = (*ppe)->segment();
+ if (oseg->type() == elfcpp::PT_LOAD)
+ {
+ oseg->add_output_section_to_load(layout, os, seg_flags);
+ break;
+ }
+ }
+
continue;
}
// PT_INTERP segment will pick up following orphan sections,
// which does not make sense. If this is not an orphan section,
// we trust the linker script.
- if (orphan)
+ if (is_orphan)
{
// Enable PT_LOAD segments only filtering until we see another
// list of segment names.
&& r->second->type() != elfcpp::PT_LOAD)
continue;
- elfcpp::Elf_Word seg_flags =
- Layout::section_flags_to_segment(os->flags());
- r->second->add_output_section(os, seg_flags, false);
-
- if (r->second->type() == elfcpp::PT_LOAD)
+ if (r->second->type() != elfcpp::PT_LOAD)
+ r->second->add_output_section_to_nonload(os, seg_flags);
+ else
{
+ r->second->add_output_section_to_load(layout, os, seg_flags);
if (in_load_segment)
gold_error(_("section in two PT_LOAD segments"));
in_load_segment = true;
void
Script_sections::print(FILE* f) const
{
+ if (this->phdrs_elements_ != NULL)
+ {
+ fprintf(f, "PHDRS {\n");
+ for (Phdrs_elements::const_iterator p = this->phdrs_elements_->begin();
+ p != this->phdrs_elements_->end();
+ ++p)
+ (*p)->print(f);
+ fprintf(f, "}\n");
+ }
+
+ if (this->memory_regions_ != NULL)
+ {
+ fprintf(f, "MEMORY {\n");
+ for (Memory_regions::const_iterator m = this->memory_regions_->begin();
+ m != this->memory_regions_->end();
+ ++m)
+ (*m)->print(f);
+ fprintf(f, "}\n");
+ }
+
if (!this->saw_sections_clause_)
return;
(*p)->print(f);
fprintf(f, "}\n");
-
- if (this->phdrs_elements_ != NULL)
- {
- fprintf(f, "PHDRS {\n");
- for (Phdrs_elements::const_iterator p = this->phdrs_elements_->begin();
- p != this->phdrs_elements_->end();
- ++p)
- (*p)->print(f);
- fprintf(f, "}\n");
- }
}
} // End namespace gold.