This is the mail archive of the
binutils@sourceware.org
mailing list for the binutils project.
Re: [gold patch] PR 22969: Add support for R_AARCH64_TLSLE_LDST8_TPREL_LO12, etc.
- From: Cary Coutant <ccoutant at gmail dot com>
- To: Binutils <binutils at sourceware dot org>, Peter Smith <peter dot smith at linaro dot org>, Han Shen <shenhan at google dot com>
- Date: Tue, 27 Mar 2018 18:23:52 -0700
- Subject: Re: [gold patch] PR 22969: Add support for R_AARCH64_TLSLE_LDST8_TPREL_LO12, etc.
- References: <CAJimCsF6v6d9S4E93H4GHans00kcJpJU04phS_yQcfP+tBvEtA@mail.gmail.com>
> Can you please look over this patch and give it a test? Let me know if
> it's OK to check in.
While I was looking in relocate_tls(), I found the following, which
looks suspicious to me:
case elfcpp::R_AARCH64_TLSLE_MOVW_TPREL_G2:
case elfcpp::R_AARCH64_TLSLE_MOVW_TPREL_G1:
case elfcpp::R_AARCH64_TLSLE_MOVW_TPREL_G1_NC:
case elfcpp::R_AARCH64_TLSLE_MOVW_TPREL_G0:
case elfcpp::R_AARCH64_TLSLE_MOVW_TPREL_G0_NC:
case elfcpp::R_AARCH64_TLSLE_ADD_TPREL_HI12:
case elfcpp::R_AARCH64_TLSLE_ADD_TPREL_LO12:
case elfcpp::R_AARCH64_TLSLE_ADD_TPREL_LO12_NC:
{
gold_assert(tls_segment != NULL);
AArch64_address value = psymval->value(object, 0);
if (!parameters->options().shared())
{
AArch64_address aligned_tcb_size =
align_address(target->tcb_size(),
tls_segment->maximum_alignment());
value += aligned_tcb_size;
switch (r_type)
{
case elfcpp::R_AARCH64_TLSLE_MOVW_TPREL_G2:
case elfcpp::R_AARCH64_TLSLE_MOVW_TPREL_G1:
case elfcpp::R_AARCH64_TLSLE_MOVW_TPREL_G0:
return aarch64_reloc_funcs::movnz(view, value + addend,
reloc_property);
default:
return aarch64_reloc_funcs::template
rela_general<32>(view,
value,
addend,
reloc_property);
}
}
For TLSLE_MOVW_TPREL_G1_NC and TLSLE_MOVW_TPREL_G0_NC, shouldn't we
also use movnz instead of the default rela_general?
-cary