[PATCH] bfd/RISC-V: Repeat a single target relax pass instead of 3
Palmer Dabbelt
palmer@dabbelt.com
Fri Oct 8 16:24:41 GMT 2021
On Fri, 08 Oct 2021 01:52:12 PDT (-0700), nelson.chu@sifive.com wrote:
> Hi Lewis,
>
> This patch is actually option 3 rather than option 2 in the pr28410,
> but personally I prefer option 3, so the solution is what I expected,
> thank you. However, I need more time to run the regressions, and make
> sure the patch won't cause side effects. But before I start to run
> the regressions, I have a quick review, and do see some places that
> could be improved.
I haven't looked at the patch (and must have missed the earlier
discussion), but the specific case that I was trying to solve with the
R_RISCV_DELETE split was something like
r0: auipc t0, symbol0
r1: auipc t1, symbol1
ld t0, symbol0, r0 # reloc points to r0
ld t1, symbol1, r1 # reloc points to r1
in this case, if the auipc at r0 gets deleted then we end up with an
aliasing issue and need some way to disambiguate which of the two
symbols we're targeting. I don't remember if we ever put a test case in
there for this, but it did manifest in real code (IIRC it was a Linux
build, but you need to twiddle all the compiler flags so it'll schedule
the auipc-based pairs).
The idea behing the extra pass here was to split this into two phases,
so we could resolve the addresses before creating the ambiguities.
>
> On Fri, Oct 8, 2021 at 5:01 AM Lewis Revill <lewis.revill@embecosm.com> wrote:
>>
>> Commit abd20cb637008da9d32018b4b03973e119388a0a and
>> ebdcad3fddf6ec21f6d4dcc702379a12718cf0c4 introduced additional
>> complexity into the paths run by the RISC-V relaxation pass in order to
>> resolve the issue of accurately keeping track of pcrel_hi and pcrel_lo
>> pairs. The first commit split up relaxation of these relocs into a pass
>> which occurred after other relaxations in order to prevent the situation
>> where bytes were deleted in between a pcrel_lo/pcrel_hi pair, inhibiting
>> our ability to find the corresponding pcrel_hi relocation from the
>> address attached to the pcrel_lo.
>>
>> Since the relaxation was split into two passes the 'again' parameter
>> could not be used to perform the entire relaxation process again and so
>> the second commit added a way to restart ldelf_map_segments, thus
>> starting the whole process again.
>>
>> Unfortunately this process could not account for the fact that we were
>> not finished with the relaxation process so in some cases - such as the
>> case where code would not fit in a memory region before the
>> R_RISCV_ALIGN relocation was relaxed - sanity checks in generic code
>> would fail.
>>
>> This patch fixes all three of these concerns by reverting back to a
>> system of having only one target relax pass but updating entries in the
>> table of pcrel_hi/pcrel_lo relocs every time any bytes are deleted. Thus
>> we can keep track of the pairs accurately, and we can use the 'again'
>> parameter to restart the entire target relax pass, behaving in the way
>> that generic code expects.
>>
>> Fixes https://sourceware.org/bugzilla/show_bug.cgi?id=28410
>>
>> bfd/
>> * elfnn-riscv.c: Revert restart_relax changes.
>> (_bfd_riscv_relax_pc): Delete bytes immediately rather than
>> using R_RISCV_DELETE.
>> (_bfd_riscv_relax_delete): Remove.
>> (riscv_relax_delete_bytes): Update tpgp relocs whenever bytes
>> are deleted.
>> (riscv_update_pcgp_reloc_offs): Add function to update pcrel_hi
>> section offsets.
>> (riscv_update_pcgp_reloc_symvals): Add function to update
>> pcrel_hi symbol values.
>> (_bfd_riscv_relax_section): Use only one pass for all target
>> relaxations.
>> * elfxx-riscv.h: Revert restart_relax changes.
>>
>> ld/
>> * emultempl/riscvelf.em: Revert restart_relax changes and set
>> relax_pass to 2.
>> ---
>> bfd/ChangeLog | 16 +
>> bfd/elfnn-riscv.c | 403 +++++++++---------
>> bfd/elfxx-riscv.h | 5 -
>> ld/ChangeLog | 5 +
>> ld/emultempl/riscvelf.em | 8 +-
>> .../ld-riscv-elf/align-small-region.d | 12 +
>> .../ld-riscv-elf/align-small-region.ld | 12 +
>> .../ld-riscv-elf/align-small-region.s | 7 +
>> ld/testsuite/ld-riscv-elf/ld-riscv-elf.exp | 1 +
>> 9 files changed, 251 insertions(+), 218 deletions(-)
>> create mode 100644 ld/testsuite/ld-riscv-elf/align-small-region.d
>> create mode 100644 ld/testsuite/ld-riscv-elf/align-small-region.ld
>> create mode 100644 ld/testsuite/ld-riscv-elf/align-small-region.s
>>
>> diff --git a/bfd/ChangeLog b/bfd/ChangeLog
>> index 2a08ff7cfb4..7bcba3a090f 100644
>> --- a/bfd/ChangeLog
>> +++ b/bfd/ChangeLog
>> @@ -1,3 +1,19 @@
>> +2021-10-07 Lewis Revill <lewis.revill@embecosm.com>
>> +
>> + * elfnn-riscv.c: Revert restart_relax changes.
>> + (_bfd_riscv_relax_pc): Delete bytes immediately rather than
>> + using R_RISCV_DELETE.
>> + (_bfd_riscv_relax_delete): Remove.
>> + (riscv_relax_delete_bytes): Update tpgp relocs whenever bytes
>> + are deleted.
>> + (riscv_update_pcgp_reloc_offs): Add function to update pcrel_hi
>> + section offsets.
>> + (riscv_update_pcgp_reloc_symvals): Add function to update
>> + pcrel_hi symbol values.
>> + (_bfd_riscv_relax_section): Use only one pass for all target
>> + relaxations.
>> + * elfxx-riscv.h: Revert restart_relax changes.
>> +
>> 2021-09-27 Nick Alcock <nick.alcock@oracle.com>
>>
>> * configure: Regenerate.
>> diff --git a/bfd/elfnn-riscv.c b/bfd/elfnn-riscv.c
>> index 2e8df72fa2a..95e2d5536d1 100644
>> --- a/bfd/elfnn-riscv.c
>> +++ b/bfd/elfnn-riscv.c
>> @@ -38,9 +38,6 @@
>> #define CHAR_BIT 8
>> #endif
>>
>> -/* Internal relocations used exclusively by the relaxation pass. */
>> -#define R_RISCV_DELETE (R_RISCV_max + 1)
>> -
>
> This looks interesting since you are removing the relax pass, which
> deletes the bytes that pass 1 made obsolete. I'm not sure if the
> change will cause any side effects, so I would suggest keeping the
> delete relax pass separating, like what we always did before.
>
>> #define ARCH_SIZE NN
>>
>> #define MINUS_ONE ((bfd_vma)0 - 1)
>> @@ -131,9 +128,6 @@ struct riscv_elf_link_hash_table
>> /* The index of the last unused .rel.iplt slot. */
>> bfd_vma last_iplt_index;
>>
>> - /* Re-run the relaxations from relax pass 0 if TRUE. */
>> - bool restart_relax;
>> -
>> /* The data segment phase, don't relax the section
>> when it is exp_seg_relro_adjust. */
>> int *data_segment_phase;
>> @@ -405,7 +399,6 @@ riscv_elf_link_hash_table_create (bfd *abfd)
>> }
>>
>> ret->max_alignment = (bfd_vma) -1;
>> - ret->restart_relax = false;
>>
>> /* Create hash table for local ifunc. */
>> ret->loc_hash_table = htab_try_create (1024,
>> @@ -1714,9 +1707,6 @@ perform_relocation (const reloc_howto_type *howto,
>> case R_RISCV_TLS_DTPREL64:
>> break;
>>
>> - case R_RISCV_DELETE:
>> - return bfd_reloc_ok;
>> -
>> default:
>> return bfd_reloc_notsupported;
>> }
>> @@ -2326,7 +2316,6 @@ riscv_elf_relocate_section (bfd *output_bfd,
>> case R_RISCV_SET16:
>> case R_RISCV_SET32:
>> case R_RISCV_32_PCREL:
>> - case R_RISCV_DELETE:
>> /* These require no special handling beyond perform_relocation. */
>> break;
>>
>> @@ -3923,115 +3912,6 @@ _bfd_riscv_elf_merge_private_bfd_data (bfd *ibfd,
>> struct bfd_link_info *info)
>> return false;
>> }
>>
>> -/* Delete some bytes from a section while relaxing. */
>> -
>> -static bool
>> -riscv_relax_delete_bytes (bfd *abfd, asection *sec, bfd_vma addr, size_t
>> count,
>> - struct bfd_link_info *link_info)
>> -{
>> - unsigned int i, symcount;
>> - bfd_vma toaddr = sec->size;
>> - struct elf_link_hash_entry **sym_hashes = elf_sym_hashes (abfd);
>> - Elf_Internal_Shdr *symtab_hdr = &elf_tdata (abfd)->symtab_hdr;
>> - unsigned int sec_shndx = _bfd_elf_section_from_bfd_section (abfd, sec);
>> - struct bfd_elf_section_data *data = elf_section_data (sec);
>> - bfd_byte *contents = data->this_hdr.contents;
>> -
>> - /* Actually delete the bytes. */
>> - sec->size -= count;
>> - memmove (contents + addr, contents + addr + count, toaddr - addr -
>> count);
>> -
>> - /* Adjust the location of all of the relocs. Note that we need not
>> - adjust the addends, since all PC-relative references must be against
>> - symbols, which we will adjust below. */
>> - for (i = 0; i < sec->reloc_count; i++)
>> - if (data->relocs[i].r_offset > addr && data->relocs[i].r_offset <
>> toaddr)
>> - data->relocs[i].r_offset -= count;
>> -
>> - /* Adjust the local symbols defined in this section. */
>> - for (i = 0; i < symtab_hdr->sh_info; i++)
>> - {
>> - Elf_Internal_Sym *sym = (Elf_Internal_Sym *) symtab_hdr->contents +
>> i;
>> - if (sym->st_shndx == sec_shndx)
>> - {
>> - /* If the symbol is in the range of memory we just moved, we
>> - have to adjust its value. */
>> - if (sym->st_value > addr && sym->st_value <= toaddr)
>> - sym->st_value -= count;
>> -
>> - /* If the symbol *spans* the bytes we just deleted (i.e. its
>> - *end* is in the moved bytes but its *start* isn't), then we
>> - must adjust its size.
>> -
>> - This test needs to use the original value of st_value, otherwise
>> - we might accidentally decrease size when deleting bytes right
>> - before the symbol. But since deleted relocs can't span across
>> - symbols, we can't have both a st_value and a st_size decrease,
>> - so it is simpler to just use an else. */
>> - else if (sym->st_value <= addr
>> - && sym->st_value + sym->st_size > addr
>> - && sym->st_value + sym->st_size <= toaddr)
>> - sym->st_size -= count;
>> - }
>> - }
>> -
>> - /* Now adjust the global symbols defined in this section. */
>> - symcount = ((symtab_hdr->sh_size / sizeof (ElfNN_External_Sym))
>> - - symtab_hdr->sh_info);
>> -
>> - for (i = 0; i < symcount; i++)
>> - {
>> - struct elf_link_hash_entry *sym_hash = sym_hashes[i];
>> -
>> - /* The '--wrap SYMBOL' option is causing a pain when the object file,
>> - containing the definition of __wrap_SYMBOL, includes a direct
>> - call to SYMBOL as well. Since both __wrap_SYMBOL and SYMBOL reference
>> - the same symbol (which is __wrap_SYMBOL), but still exist as two
>> - different symbols in 'sym_hashes', we don't want to adjust
>> - the global symbol __wrap_SYMBOL twice.
>> -
>> - The same problem occurs with symbols that are versioned_hidden, as
>> - foo becomes an alias for foo@BAR, and hence they need the same
>> - treatment. */
>> - if (link_info->wrap_hash != NULL
>> - || sym_hash->versioned != unversioned)
>> - {
>> - struct elf_link_hash_entry **cur_sym_hashes;
>> -
>> - /* Loop only over the symbols which have already been checked. */
>> - for (cur_sym_hashes = sym_hashes; cur_sym_hashes < &sym_hashes[i];
>> - cur_sym_hashes++)
>> - {
>> - /* If the current symbol is identical to 'sym_hash', that means
>> - the symbol was already adjusted (or at least checked). */
>> - if (*cur_sym_hashes == sym_hash)
>> - break;
>> - }
>> - /* Don't adjust the symbol again. */
>> - if (cur_sym_hashes < &sym_hashes[i])
>> - continue;
>> - }
>> -
>> - if ((sym_hash->root.type == bfd_link_hash_defined
>> - || sym_hash->root.type == bfd_link_hash_defweak)
>> - && sym_hash->root.u.def.section == sec)
>> - {
>> - /* As above, adjust the value if needed. */
>> - if (sym_hash->root.u.def.value > addr
>> - && sym_hash->root.u.def.value <= toaddr)
>> - sym_hash->root.u.def.value -= count;
>> -
>> - /* As above, adjust the size if needed. */
>> - else if (sym_hash->root.u.def.value <= addr
>> - && sym_hash->root.u.def.value + sym_hash->size > addr
>> - && sym_hash->root.u.def.value + sym_hash->size <= toaddr)
>> - sym_hash->size -= count;
>> - }
>> - }
>> -
>> - return true;
>> -}
>> -
>> /* A second format for recording PC-relative hi relocations. This stores
>> the
>> information required to relax them to GP-relative addresses. */
>>
>> @@ -4162,6 +4042,166 @@ riscv_find_pcgp_lo_reloc (riscv_pcgp_relocs *p,
>> bfd_vma hi_sec_off)
>> return false;
>> }
>>
>> +static bool
>> +riscv_update_pcgp_reloc_offs (riscv_pcgp_relocs *p, bfd_vma
>> deleted_sec_off,
>> + size_t deleted_count)
>> +{
>> + riscv_pcgp_lo_reloc *l;
>> + riscv_pcgp_hi_reloc *h;
>> + bool updated = false;
>> +
>> + for (l = p->lo; l != NULL; l = l->next)
>> + if (l->hi_sec_off > deleted_sec_off)
>> + {
>> + l->hi_sec_off -= deleted_count;
>> + updated = true;
>> + }
>> +
>> + for (h = p->hi; h != NULL; h = h->next)
>> + if (h->hi_sec_off > deleted_sec_off)
>> + {
>> + h->hi_sec_off -= deleted_count;
>> + updated = true;
>> + }
>> +
>> + return updated;
>> +}
>
> The returned boolean `updated' looks useless and redundant, so we
> probably can remove it.
>
>> +static bool
>> +riscv_update_pcgp_reloc_symvals(riscv_pcgp_relocs *p, bfd_vma old_value,
>> + size_t deleted_count)
>> +{
>> + riscv_pcgp_hi_reloc *h;
>> + bool updated = false;
>> +
>> + for (h = p->hi; h != NULL; h = h->next)
>> + if (h->hi_addr == old_value)
>> + {
>> + h->hi_addr -= deleted_count;
>> + updated = true;
>> + }
>> + return updated;
>> +}
>
> Likewise. Besides, I suppose you also need to check if the `sym_sec'
> of riscv_pcgp_relocs is the same as the section in which we are
> deleting the bytes. That is - we can only modify the symbol value
> when the symbol section is the same as the deleting section.
>
>> +/* Delete some bytes from a section while relaxing. */
>> +
>> +static bool
>> +riscv_relax_delete_bytes (bfd *abfd, asection *sec, bfd_vma addr, size_t
>> count,
>> + struct bfd_link_info *link_info, riscv_pcgp_relocs *p)
>> +{
>> + unsigned int i, symcount;
>> + bfd_vma toaddr = sec->size;
>> + struct elf_link_hash_entry **sym_hashes = elf_sym_hashes (abfd);
>> + Elf_Internal_Shdr *symtab_hdr = &elf_tdata (abfd)->symtab_hdr;
>> + unsigned int sec_shndx = _bfd_elf_section_from_bfd_section (abfd, sec);
>> + struct bfd_elf_section_data *data = elf_section_data (sec);
>> + bfd_byte *contents = data->this_hdr.contents;
>> +
>> + /* Actually delete the bytes. */
>> + sec->size -= count;
>> + memmove (contents + addr, contents + addr + count, toaddr - addr -
>> count);
>> +
>> + /* Adjust the location of all of the relocs. Note that we need not
>> + adjust the addends, since all PC-relative references must be against
>> + symbols, which we will adjust below. */
>> + for (i = 0; i < sec->reloc_count; i++)
>> + if (data->relocs[i].r_offset > addr && data->relocs[i].r_offset <
>> toaddr)
>> + data->relocs[i].r_offset -= count;
>> +
>> + /* Adjust the hi_sec_off of entries in the pcgp relocs table. */
>> + riscv_update_pcgp_reloc_offs (p, addr, count);
>> +
>> + /* Adjust the local symbols defined in this section. */
>> + for (i = 0; i < symtab_hdr->sh_info; i++)
>> + {
>> + Elf_Internal_Sym *sym = (Elf_Internal_Sym *) symtab_hdr->contents +
>> i;
>> + if (sym->st_shndx == sec_shndx)
>> + {
>> + /* If the symbol is in the range of memory we just moved, we
>> + have to adjust its value. */
>> + if (sym->st_value > addr && sym->st_value <= toaddr)
>> + {
>> + riscv_update_pcgp_reloc_symvals (p, sym->st_value, count);
>
> In this place we make sure that only the symbols defined in the same
> `sec' will be modified. But you only pass the symbol value into the
> riscv_update_pcgp_reloc_symvals, and use the value to decide if the
> pcgp table's symbols are the same one. I think this is dangerous
> since you may modify the symbols which are actually defined in other
> sections.
>
> I think `riscv_update_pcgp_reloc_symvals' should be moved outside the
> if statement, and you only need to trace the whole pcgp tables once,
> with checking the `sym_sec' and `hi_addr', maybe something like this,
>
> if (sym_sec == sec && hi_addr > addr && hi_addr <= toaddr)
>
>> + sym->st_value -= count;
>> + }
>> +
>> + /* If the symbol *spans* the bytes we just deleted (i.e. its
>> + *end* is in the moved bytes but its *start* isn't), then we
>> + must adjust its size.
>> +
>> + This test needs to use the original value of st_value, otherwise
>> + we might accidentally decrease size when deleting bytes right
>> + before the symbol. But since deleted relocs can't span across
>> + symbols, we can't have both a st_value and a st_size decrease,
>> + so it is simpler to just use an else. */
>> + else if (sym->st_value <= addr
>> + && sym->st_value + sym->st_size > addr
>> + && sym->st_value + sym->st_size <= toaddr)
>> + sym->st_size -= count;
>> + }
>> + }
>> +
>> + /* Now adjust the global symbols defined in this section. */
>> + symcount = ((symtab_hdr->sh_size / sizeof (ElfNN_External_Sym))
>> + - symtab_hdr->sh_info);
>> +
>> + for (i = 0; i < symcount; i++)
>> + {
>> + struct elf_link_hash_entry *sym_hash = sym_hashes[i];
>> +
>> + /* The '--wrap SYMBOL' option is causing a pain when the object file,
>> + containing the definition of __wrap_SYMBOL, includes a direct
>> + call to SYMBOL as well. Since both __wrap_SYMBOL and SYMBOL reference
>> + the same symbol (which is __wrap_SYMBOL), but still exist as two
>> + different symbols in 'sym_hashes', we don't want to adjust
>> + the global symbol __wrap_SYMBOL twice.
>> +
>> + The same problem occurs with symbols that are versioned_hidden, as
>> + foo becomes an alias for foo@BAR, and hence they need the same
>> + treatment. */
>> + if (link_info->wrap_hash != NULL
>> + || sym_hash->versioned != unversioned)
>> + {
>> + struct elf_link_hash_entry **cur_sym_hashes;
>> +
>> + /* Loop only over the symbols which have already been checked. */
>> + for (cur_sym_hashes = sym_hashes; cur_sym_hashes < &sym_hashes[i];
>> + cur_sym_hashes++)
>> + {
>> + /* If the current symbol is identical to 'sym_hash', that means
>> + the symbol was already adjusted (or at least checked). */
>> + if (*cur_sym_hashes == sym_hash)
>> + break;
>> + }
>> + /* Don't adjust the symbol again. */
>> + if (cur_sym_hashes < &sym_hashes[i])
>> + continue;
>> + }
>> +
>> + if ((sym_hash->root.type == bfd_link_hash_defined
>> + || sym_hash->root.type == bfd_link_hash_defweak)
>> + && sym_hash->root.u.def.section == sec)
>> + {
>> + /* As above, adjust the value if needed. */
>> + if (sym_hash->root.u.def.value > addr
>> + && sym_hash->root.u.def.value <= toaddr)
>> + {
>> + riscv_update_pcgp_reloc_symvals (p, sym_hash->root.u.def.value,
>> + count);
>
> Likewise, just checking the symbol value looks dangerous.
>
>> + sym_hash->root.u.def.value -= count;
>> + }
>> +
>> + /* As above, adjust the size if needed. */
>> + else if (sym_hash->root.u.def.value <= addr
>> + && sym_hash->root.u.def.value + sym_hash->size > addr
>> + && sym_hash->root.u.def.value + sym_hash->size <= toaddr)
>> + sym_hash->size -= count;
>> + }
>> + }
>> +
>> + return true;
>> +}
>> +
>> typedef bool (*relax_func_t) (bfd *, asection *, asection *,
>> struct bfd_link_info *,
>> Elf_Internal_Rela *,
>> @@ -4179,7 +4219,7 @@ _bfd_riscv_relax_call (bfd *abfd, asection *sec,
>> asection *sym_sec,
>> bfd_vma max_alignment,
>> bfd_vma reserve_size ATTRIBUTE_UNUSED,
>> bool *again,
>> - riscv_pcgp_relocs *pcgp_relocs ATTRIBUTE_UNUSED,
>> + riscv_pcgp_relocs *pcgp_relocs,
>> bool undefined_weak ATTRIBUTE_UNUSED)
>> {
>> bfd_byte *contents = elf_section_data (sec)->this_hdr.contents;
>> @@ -4243,7 +4283,7 @@ _bfd_riscv_relax_call (bfd *abfd, asection *sec,
>> asection *sym_sec,
>> /* Delete unnecessary JALR. */
>> *again = true;
>> return riscv_relax_delete_bytes (abfd, sec, rel->r_offset + len, 8 - len,
>> - link_info);
>> + link_info, pcgp_relocs);
>> }
>>
>> /* Traverse all output sections and return the max alignment. */
>> @@ -4275,7 +4315,7 @@ _bfd_riscv_relax_lui (bfd *abfd,
>> bfd_vma max_alignment,
>> bfd_vma reserve_size,
>> bool *again,
>> - riscv_pcgp_relocs *pcgp_relocs ATTRIBUTE_UNUSED,
>> + riscv_pcgp_relocs *pcgp_relocs,
>> bool undefined_weak)
>> {
>> bfd_byte *contents = elf_section_data (sec)->this_hdr.contents;
>> @@ -4337,7 +4377,7 @@ _bfd_riscv_relax_lui (bfd *abfd,
>> rel->r_info = ELFNN_R_INFO (0, R_RISCV_NONE);
>> *again = true;
>> return riscv_relax_delete_bytes (abfd, sec, rel->r_offset, 4,
>> - link_info);
>> + link_info, pcgp_relocs);
>>
>> default:
>> abort ();
>> @@ -4370,7 +4410,7 @@ _bfd_riscv_relax_lui (bfd *abfd,
>>
>> *again = true;
>> return riscv_relax_delete_bytes (abfd, sec, rel->r_offset + 2, 2,
>> - link_info);
>> + link_info, pcgp_relocs);
>> }
>>
>> return true;
>> @@ -4388,7 +4428,7 @@ _bfd_riscv_relax_tls_le (bfd *abfd,
>> bfd_vma max_alignment ATTRIBUTE_UNUSED,
>> bfd_vma reserve_size ATTRIBUTE_UNUSED,
>> bool *again,
>> - riscv_pcgp_relocs *prcel_relocs ATTRIBUTE_UNUSED,
>> + riscv_pcgp_relocs *prcel_relocs,
>> bool undefined_weak ATTRIBUTE_UNUSED)
>> {
>> /* See if this symbol is in range of tp. */
>> @@ -4411,15 +4451,15 @@ _bfd_riscv_relax_tls_le (bfd *abfd,
>> /* We can delete the unnecessary instruction and reloc. */
>> rel->r_info = ELFNN_R_INFO (0, R_RISCV_NONE);
>> *again = true;
>> - return riscv_relax_delete_bytes (abfd, sec, rel->r_offset, 4,
>> link_info);
>> + return riscv_relax_delete_bytes (abfd, sec, rel->r_offset, 4,
>> link_info,
>> + prcel_relocs);
>>
>> default:
>> abort ();
>> }
>> }
>>
>> -/* Implement R_RISCV_ALIGN by deleting excess alignment NOPs.
>> - Once we've handled an R_RISCV_ALIGN, we can't relax anything else. */
>> +/* Implement R_RISCV_ALIGN by deleting excess alignment NOPs. */
>>
>> static bool
>> _bfd_riscv_relax_align (bfd *abfd, asection *sec,
>> @@ -4430,7 +4470,7 @@ _bfd_riscv_relax_align (bfd *abfd, asection *sec,
>> bfd_vma max_alignment ATTRIBUTE_UNUSED,
>> bfd_vma reserve_size ATTRIBUTE_UNUSED,
>> bool *again ATTRIBUTE_UNUSED,
>> - riscv_pcgp_relocs *pcrel_relocs ATTRIBUTE_UNUSED,
>> + riscv_pcgp_relocs *pcgp_relocs,
>> bool undefined_weak ATTRIBUTE_UNUSED)
>> {
>> bfd_byte *contents = elf_section_data (sec)->this_hdr.contents;
>> @@ -4442,6 +4482,9 @@ _bfd_riscv_relax_align (bfd *abfd, asection *sec,
>> bfd_vma aligned_addr = ((symval - 1) & ~(alignment - 1)) + alignment;
>> bfd_vma nop_bytes = aligned_addr - symval;
>>
>> + /* Once we've handled an R_RISCV_ALIGN, we can't relax anything else. */
>> + sec->sec_flg0 = true;
>> +
>> /* Make sure there are enough NOPs to actually achieve the alignment. */
>> if (rel->r_addend < nop_bytes)
>> {
>> @@ -4471,13 +4514,13 @@ _bfd_riscv_relax_align (bfd *abfd, asection *sec,
>>
>> /* Delete the excess bytes. */
>> return riscv_relax_delete_bytes (abfd, sec, rel->r_offset + nop_bytes,
>> - rel->r_addend - nop_bytes, link_info);
>> + rel->r_addend - nop_bytes, link_info, pcgp_relocs);
>> }
>
> The alignment relaxations are behind the pcgc relaxations, so we don't
> need to update the pcgp tables in fact.
>
>> /* Relax PC-relative references to GP-relative references. */
>>
>> static bool
>> -_bfd_riscv_relax_pc (bfd *abfd ATTRIBUTE_UNUSED,
>> +_bfd_riscv_relax_pc (bfd *abfd,
>> asection *sec,
>> asection *sym_sec,
>> struct bfd_link_info *link_info,
>> @@ -4485,7 +4528,7 @@ _bfd_riscv_relax_pc (bfd *abfd ATTRIBUTE_UNUSED,
>> bfd_vma symval,
>> bfd_vma max_alignment,
>> bfd_vma reserve_size,
>> - bool *again ATTRIBUTE_UNUSED,
>> + bool *again,
>> riscv_pcgp_relocs *pcgp_relocs,
>> bool undefined_weak)
>> {
>> @@ -4614,9 +4657,11 @@ _bfd_riscv_relax_pc (bfd *abfd ATTRIBUTE_UNUSED,
>> sym_sec,
>> undefined_weak);
>> /* We can delete the unnecessary AUIPC and reloc. */
>> - rel->r_info = ELFNN_R_INFO (0, R_RISCV_DELETE);
>> - rel->r_addend = 4;
>> - return true;
>> +
>> + rel->r_info = ELFNN_R_INFO (0, R_RISCV_NONE);
>> + *again = true;
>> + return riscv_relax_delete_bytes (abfd, sec, rel->r_offset, 4, link_info,
>> + pcgp_relocs);
>>
>> default:
>> abort ();
>> @@ -4626,29 +4671,6 @@ _bfd_riscv_relax_pc (bfd *abfd ATTRIBUTE_UNUSED,
>> return true;
>> }
>>
>> -/* Delete the bytes for R_RISCV_DELETE. */
>> -
>> -static bool
>> -_bfd_riscv_relax_delete (bfd *abfd,
>> - asection *sec,
>> - asection *sym_sec ATTRIBUTE_UNUSED,
>> - struct bfd_link_info *link_info,
>> - Elf_Internal_Rela *rel,
>> - bfd_vma symval ATTRIBUTE_UNUSED,
>> - bfd_vma max_alignment ATTRIBUTE_UNUSED,
>> - bfd_vma reserve_size ATTRIBUTE_UNUSED,
>> - bool *again,
>> - riscv_pcgp_relocs *pcgp_relocs ATTRIBUTE_UNUSED,
>> - bool undefined_weak ATTRIBUTE_UNUSED)
>> -{
>> - if (!riscv_relax_delete_bytes (abfd, sec, rel->r_offset, rel->r_addend,
>> - link_info))
>> - return false;
>> - rel->r_info = ELFNN_R_INFO (0, R_RISCV_NONE);
>> - *again = true;
>> - return true;
>> -}
>> -
>
> Likewise, I would suggest keeping the _bfd_riscv_relax_delete.
>
>> /* Called by after_allocation to set the information of data segment
>> before relaxing. */
>>
>> @@ -4660,35 +4682,10 @@ bfd_elfNN_riscv_set_data_segment_info (struct
>> bfd_link_info *info,
>> htab->data_segment_phase = data_segment_phase;
>> }
>>
>> -/* Called by after_allocation to check if we need to run the whole
>> - relaxations again. */
>> -
>> -bool
>> -bfd_elfNN_riscv_restart_relax_sections (struct bfd_link_info *info)
>> -{
>> - struct riscv_elf_link_hash_table *htab = riscv_elf_hash_table (info);
>> - bool restart = htab->restart_relax;
>> - /* Reset the flag. */
>> - htab->restart_relax = false;
>> - return restart;
>> -}
>> -
>> /* Relax a section.
>>
>> - Pass 0: Shortens code sequences for LUI/CALL/TPREL relocs.
>> - Pass 1: Shortens code sequences for PCREL relocs.
>> - Pass 2: Deletes the bytes that pass 1 made obsolete.
>> - Pass 3: Which cannot be disabled, handles code alignment directives.
>> -
>> - The `again` is used to determine whether the relax pass itself needs to
>> - run again. And the `restart_relax` is used to determine if we need to
>> - run the whole relax passes again from 0 to 2. Once we have deleted the
>> - code between relax pass 0 to 2, the restart_relax will be set to TRUE,
>> - and we should run the whole relaxations again to give them more chances
>> - to shorten the code.
>> -
>> - Since we can't relax anything else once we start to handle the
>> alignments,
>> - we will only enter into the relax pass 3 when the restart_relax is
>> FALSE. */
>> + Pass 0: Shortens code sequences for LUI/CALL/TPREL/PCREL relocs.
>> + Pass 1: Which cannot be disabled, handles code alignment directives. */
>>
>> static bool
>> _bfd_riscv_relax_section (bfd *abfd, asection *sec,
>> @@ -4707,12 +4704,11 @@ _bfd_riscv_relax_section (bfd *abfd, asection *sec,
>> *again = false;
>>
>> if (bfd_link_relocatable (info)
>> + || sec->sec_flg0
>> || (sec->flags & SEC_RELOC) == 0
>> || sec->reloc_count == 0
>> || (info->disable_target_specific_optimizations
>> - && info->relax_pass < 2)
>> - || (htab->restart_relax
>> - && info->relax_pass == 3)
>> + && info->relax_pass == 0)
>> /* The exp_seg_relro_adjust is enum phase_enum (0x4),
>> and defined in ld/ldexp.h. */
>> || *(htab->data_segment_phase) == 4)
>> @@ -4757,31 +4753,27 @@ _bfd_riscv_relax_section (bfd *abfd, asection *sec,
>> || type == R_RISCV_CALL_PLT)
>> relax_func = _bfd_riscv_relax_call;
>> else if (type == R_RISCV_HI20
>> - || type == R_RISCV_LO12_I
>> - || type == R_RISCV_LO12_S)
>> + || type == R_RISCV_LO12_I
>> + || type == R_RISCV_LO12_S)
>> relax_func = _bfd_riscv_relax_lui;
>> else if (type == R_RISCV_TPREL_HI20
>> - || type == R_RISCV_TPREL_ADD
>> - || type == R_RISCV_TPREL_LO12_I
>> - || type == R_RISCV_TPREL_LO12_S)
>> + || type == R_RISCV_TPREL_ADD
>> + || type == R_RISCV_TPREL_LO12_I
>> + || type == R_RISCV_TPREL_LO12_S)
>> relax_func = _bfd_riscv_relax_tls_le;
>> + else if (!bfd_link_pic (info) && (type == R_RISCV_PCREL_HI20
>> + || type == R_RISCV_PCREL_LO12_I
>> + || type == R_RISCV_PCREL_LO12_S))
>> + relax_func = _bfd_riscv_relax_pc;
>> else
>> continue;
>> }
>> - else if (info->relax_pass == 1
>> - && !bfd_link_pic (info)
>> - && (type == R_RISCV_PCREL_HI20
>> - || type == R_RISCV_PCREL_LO12_I
>> - || type == R_RISCV_PCREL_LO12_S))
>> - relax_func = _bfd_riscv_relax_pc;
>> - else if (info->relax_pass == 2 && type == R_RISCV_DELETE)
>> - relax_func = _bfd_riscv_relax_delete;
>> - else if (info->relax_pass == 3 && type == R_RISCV_ALIGN)
>> + else if (info->relax_pass == 1 && type == R_RISCV_ALIGN)
>> relax_func = _bfd_riscv_relax_align;
>> else
>> continue;
>>
>> - if (info->relax_pass < 2)
>> + if (info->relax_pass == 0)
>> {
>> /* Only relax this reloc if it is paired with R_RISCV_RELAX. */
>> if (i == sec->reloc_count - 1
>> @@ -4954,9 +4946,6 @@ _bfd_riscv_relax_section (bfd *abfd, asection *sec,
>> free (relocs);
>> riscv_free_pcgp_relocs (&pcgp_relocs, abfd, sec);
>>
>> - if (*again)
>> - htab->restart_relax = true;
>> -
>> return ret;
>> }
>>
>> diff --git a/bfd/elfxx-riscv.h b/bfd/elfxx-riscv.h
>> index e691a97ecda..3af8fd99d9a 100644
>> --- a/bfd/elfxx-riscv.h
>> +++ b/bfd/elfxx-riscv.h
>> @@ -92,11 +92,6 @@ riscv_estimate_digit (unsigned);
>> extern int
>> riscv_compare_subsets (const char *, const char *);
>>
>> -extern bool
>> -bfd_elf32_riscv_restart_relax_sections (struct bfd_link_info *);
>> -extern bool
>> -bfd_elf64_riscv_restart_relax_sections (struct bfd_link_info *);
>> -
>> extern void
>> bfd_elf32_riscv_set_data_segment_info (struct bfd_link_info *, int *);
>> extern void
>> diff --git a/ld/ChangeLog b/ld/ChangeLog
>> index 808191bd14e..9cfcdd88754 100644
>> --- a/ld/ChangeLog
>> +++ b/ld/ChangeLog
>> @@ -1,3 +1,8 @@
>> +2021-10-07 Lewis Revill <lewis.revill@embecosm.com>
>> +
>> + * emultempl/riscvelf.em: Revert restart_relax changes and set
>> + relax_pass to 2.
>> +
>> 2021-09-30 Dimitar Dimitrov <dimitar@dinux.eu>
>>
>> * scripttempl/pru.sc (.resource_table): Align the output
>> diff --git a/ld/emultempl/riscvelf.em b/ld/emultempl/riscvelf.em
>> index c625a631fe5..865e3abd494 100644
>> --- a/ld/emultempl/riscvelf.em
>> +++ b/ld/emultempl/riscvelf.em
>> @@ -42,7 +42,7 @@ riscv_elf_before_allocation (void)
>> ENABLE_RELAXATION;
>> }
>>
>> - link_info.relax_pass = 4;
>> + link_info.relax_pass = 2;
>> }
>>
>> static void
>> @@ -76,11 +76,7 @@ gld${EMULATION_NAME}_after_allocation (void)
>> enum phase_enum *phase = &(expld.dataseg.phase);
>> bfd_elf${ELFSIZE}_riscv_set_data_segment_info (&link_info, (int *)
>> phase);
>>
>> - do
>> - {
>> - ldelf_map_segments (need_layout);
>> - }
>> - while (bfd_elf${ELFSIZE}_riscv_restart_relax_sections (&link_info));
>> + ldelf_map_segments (need_layout);
>> }
>>
>> /* This is a convenient point to tell BFD about target specific flags.
>> diff --git a/ld/testsuite/ld-riscv-elf/align-small-region.d
>> b/ld/testsuite/ld-riscv-elf/align-small-region.d
>> new file mode 100644
>> index 00000000000..3799129c665
>> --- /dev/null
>> +++ b/ld/testsuite/ld-riscv-elf/align-small-region.d
>> @@ -0,0 +1,12 @@
>> +#source: align-small-region.s
>> +#as: -march=rv32i
>> +#ld: -melf32lriscv --relax -Talign-small-region.ld --defsym=_start=0x100
>> +#objdump: -d
>> +
>> +.*:[ ]+file format .*
>> +
>> +Disassembly of section \.entry:
>> +
>> +00000000 <_reset>:
>> +.*:[ ]+[0-9a-f]+[ ]+j[ ]+100[ ]+<_start>
>> +#pass
>> diff --git a/ld/testsuite/ld-riscv-elf/align-small-region.ld
>> b/ld/testsuite/ld-riscv-elf/align-small-region.ld
>> new file mode 100644
>> index 00000000000..6a3e6638b26
>> --- /dev/null
>> +++ b/ld/testsuite/ld-riscv-elf/align-small-region.ld
>> @@ -0,0 +1,12 @@
>> +MEMORY
>> +{
>> + reset : ORIGIN = 0x0, LENGTH = 32
>> +}
>> +
>> +SECTIONS
>> +{
>> + .entry :
>> + {
>> + KEEP (*(.entry))
>> + } > reset
>> +}
>> diff --git a/ld/testsuite/ld-riscv-elf/align-small-region.s
>> b/ld/testsuite/ld-riscv-elf/align-small-region.s
>> new file mode 100644
>> index 00000000000..ace81d640ed
>> --- /dev/null
>> +++ b/ld/testsuite/ld-riscv-elf/align-small-region.s
>> @@ -0,0 +1,7 @@
>> + .section .entry, "xa"
>> + .align 5
>> + .globl _reset
>> + .type _reset, @function
>> +_reset:
>> + tail _start
>> + .size _reset, . - _reset
>> diff --git a/ld/testsuite/ld-riscv-elf/ld-riscv-elf.exp
>> b/ld/testsuite/ld-riscv-elf/ld-riscv-elf.exp
>> index 11977611733..739cc04b889 100644
>> --- a/ld/testsuite/ld-riscv-elf/ld-riscv-elf.exp
>> +++ b/ld/testsuite/ld-riscv-elf/ld-riscv-elf.exp
>> @@ -119,6 +119,7 @@ proc run_relax_twice_test {} {
>> }
>>
>> if [istarget "riscv*-*-*"] {
>> + run_dump_test "align-small-region"
>> run_dump_test "call-relax"
>> run_dump_test "pcgp-relax"
>> run_dump_test "c-lui"
>> --
>> 2.25.1
More information about the Binutils
mailing list