[lld] Add target support for SystemZ (s390x) #75643
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
@llvm/pr-subscribers-lld-elf Author: Ulrich Weigand (uweigand) ChangesThis patch adds full support for linking SystemZ (ELF s390x) object files. Support should be generally complete:
In addition to new platform code and the obvious changes, there were a few additional changes to common code:
This was tested by building and testing >1500 Fedora packages, with only a handful of failures; as these also have issues when building with LLD on other architectures, they seem unrelated. Patch is 88.34 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/75643.diff 38 Files Affected:
diff --git a/lld/ELF/Arch/SystemZ.cpp b/lld/ELF/Arch/SystemZ.cpp
new file mode 100644
index 00000000000000..80a65cf32b6c32
--- /dev/null
+++ b/lld/ELF/Arch/SystemZ.cpp
@@ -0,0 +1,581 @@
+//===- SystemZ.cpp --------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "OutputSections.h"
+#include "Symbols.h"
+#include "SyntheticSections.h"
+#include "Target.h"
+#include "lld/Common/ErrorHandler.h"
+#include "llvm/BinaryFormat/ELF.h"
+#include "llvm/Support/Endian.h"
+
+using namespace llvm;
+using namespace llvm::object;
+using namespace llvm::support::endian;
+using namespace llvm::ELF;
+using namespace lld;
+using namespace lld::elf;
+
+namespace {
+class SystemZ : public TargetInfo {
+public:
+ SystemZ();
+ RelExpr getRelExpr(RelType type, const Symbol &s,
+ const uint8_t *loc) const override;
+ RelType getDynRel(RelType type) const override;
+ void writeGotHeader(uint8_t *buf) const override;
+ void writeGotPlt(uint8_t *buf, const Symbol &s) const override;
+ void writeIgotPlt(uint8_t *buf, const Symbol &s) const override;
+ void writePltHeader(uint8_t *buf) const override;
+ void writePlt(uint8_t *buf, const Symbol &sym,
+ uint64_t pltEntryAddr) const override;
+ RelExpr adjustTlsExpr(RelType type, RelExpr expr) const override;
+ RelExpr adjustGotPcExpr(RelType type, int64_t addend,
+ const uint8_t *loc) const override;
+ bool relaxOnce(int pass) const override;
+ void relocate(uint8_t *loc, const Relocation &rel,
+ uint64_t val) const override;
+ int64_t getImplicitAddend(const uint8_t *buf, RelType type) const override;
+private:
+ void relaxGot(uint8_t *loc, const Relocation &rel, uint64_t val) const;
+ void relaxTlsGdToIe(uint8_t *loc, const Relocation &rel, uint64_t val) const;
+ void relaxTlsGdToLe(uint8_t *loc, const Relocation &rel, uint64_t val) const;
+ void relaxTlsLdToLe(uint8_t *loc, const Relocation &rel, uint64_t val) const;
+};
+} // namespace
+
+SystemZ::SystemZ() {
+ copyRel = R_390_COPY;
+ gotRel = R_390_GLOB_DAT;
+ pltRel = R_390_JMP_SLOT;
+ relativeRel = R_390_RELATIVE;
+ iRelativeRel = R_390_IRELATIVE;
+ symbolicRel = R_390_64;
+ tlsGotRel = R_390_TLS_TPOFF;
+ tlsModuleIndexRel = R_390_TLS_DTPMOD;
+ tlsOffsetRel = R_390_TLS_DTPOFF;
+ gotHeaderEntriesNum = 3;
+ gotPltHeaderEntriesNum = 0;
+ gotEntrySize = 8;
+ pltHeaderSize = 32;
+ pltEntrySize = 32;
+ ipltEntrySize = 32;
+
+ // This "trap instruction" is used to fill gaps between sections.
+ // On SystemZ, the behavior of the GNU ld is to fill those gaps
+ // with nop instructions instead - and unfortunately the default
+ // glibc crt object files (used to) rely on that behavior since
+ // they use an alignment on the .init section fragments that causes
+ // gaps which must be filled with nops as they are being executed.
+ // Therefore, we provide a nop instruction as "trapInstr" here.
+ trapInstr = {0x07, 0x07, 0x07, 0x07};
+
+ defaultImageBase = 0x1000000;
+}
+
+RelExpr SystemZ::getRelExpr(RelType type, const Symbol &s,
+ const uint8_t *loc) const {
+ switch (type) {
+ case R_390_NONE:
+ return R_NONE;
+ // Relocations targeting the symbol value.
+ case R_390_8:
+ case R_390_12:
+ case R_390_16:
+ case R_390_20:
+ case R_390_32:
+ case R_390_64:
+ return R_ABS;
+ case R_390_PC16:
+ case R_390_PC32:
+ case R_390_PC64:
+ case R_390_PC12DBL:
+ case R_390_PC16DBL:
+ case R_390_PC24DBL:
+ case R_390_PC32DBL:
+ return R_PC;
+ case R_390_GOTOFF16:
+ case R_390_GOTOFF: // a.k.a. R_390_GOTOFF32
+ case R_390_GOTOFF64:
+ return R_GOTREL;
+ // Relocations targeting the PLT associated with the symbol.
+ case R_390_PLT32:
+ case R_390_PLT64:
+ case R_390_PLT12DBL:
+ case R_390_PLT16DBL:
+ case R_390_PLT24DBL:
+ case R_390_PLT32DBL:
+ return R_PLT_PC;
+ case R_390_PLTOFF16:
+ case R_390_PLTOFF32:
+ case R_390_PLTOFF64:
+ return R_PLT_GOTREL;
+ // Relocations targeting the GOT entry associated with the symbol.
+ case R_390_GOTENT:
+ return R_GOT_PC;
+ case R_390_GOT12:
+ case R_390_GOT16:
+ case R_390_GOT20:
+ case R_390_GOT32:
+ case R_390_GOT64:
+ return R_GOT_OFF;
+ // Relocations targeting the GOTPLT entry associated with the symbol.
+ case R_390_GOTPLTENT:
+ return R_GOTPLT_PC;
+ case R_390_GOTPLT12:
+ case R_390_GOTPLT16:
+ case R_390_GOTPLT20:
+ case R_390_GOTPLT32:
+ case R_390_GOTPLT64:
+ return R_GOTPLT_OFF;
+ // Relocations targeting _GLOBAL_OFFSET_TABLE_.
+ case R_390_GOTPC:
+ case R_390_GOTPCDBL:
+ return R_GOTONLY_PC;
+ // TLS-related relocations.
+ case R_390_TLS_LOAD:
+ return R_NONE;
+ case R_390_TLS_GDCALL:
+ return R_TLSGD_PC;
+ case R_390_TLS_LDCALL:
+ return R_TLSLD_PC;
+ case R_390_TLS_GD32:
+ case R_390_TLS_GD64:
+ return R_TLSGD_GOT;
+ case R_390_TLS_LDM32:
+ case R_390_TLS_LDM64:
+ return R_TLSLD_GOT;
+ case R_390_TLS_LDO32:
+ case R_390_TLS_LDO64:
+ return R_DTPREL;
+ case R_390_TLS_LE32:
+ case R_390_TLS_LE64:
+ return R_TPREL;
+ case R_390_TLS_IE32:
+ case R_390_TLS_IE64:
+ return R_GOT;
+ case R_390_TLS_GOTIE12:
+ case R_390_TLS_GOTIE20:
+ case R_390_TLS_GOTIE32:
+ case R_390_TLS_GOTIE64:
+ return R_GOT_OFF;
+ case R_390_TLS_IEENT:
+ return R_GOT_PC;
+
+ default:
+ error(getErrorLocation(loc) + "unknown relocation (" + Twine(type) +
+ ") against symbol " + toString(s));
+ return R_NONE;
+ }
+}
+
+void SystemZ::writeGotHeader(uint8_t *buf) const {
+ // _GLOBAL_OFFSET_TABLE_[0] holds the value of _DYNAMIC.
+ // _GLOBAL_OFFSET_TABLE_[1] and [2] are reserved.
+ write64be(buf, mainPart->dynamic->getVA());
+}
+
+void SystemZ::writeGotPlt(uint8_t *buf, const Symbol &s) const {
+ write64be(buf, s.getPltVA() + 14);
+}
+
+void SystemZ::writeIgotPlt(uint8_t *buf, const Symbol &s) const {
+ if (config->writeAddends)
+ write64be(buf, s.getVA());
+}
+
+void SystemZ::writePltHeader(uint8_t *buf) const {
+ const uint8_t pltData[] = {
+ 0xe3, 0x10, 0xf0, 0x38, 0x00, 0x24, // stg %r1,56(%r15)
+ 0xc0, 0x10, 0x00, 0x00, 0x00, 0x00, // larl %r1,_GLOBAL_OFFSET_TABLE_
+ 0xd2, 0x07, 0xf0, 0x30, 0x10, 0x08, // mvc 48(8,%r15),8(%r1)
+ 0xe3, 0x10, 0x10, 0x10, 0x00, 0x04, // lg %r1,16(%r1)
+ 0x07, 0xf1, // br %r1
+ 0x07, 0x00, // nopr
+ 0x07, 0x00, // nopr
+ 0x07, 0x00, // nopr
+ };
+ memcpy(buf, pltData, sizeof(pltData));
+ uint64_t got = in.got->getVA();
+ uint64_t plt = in.plt->getVA();
+ write32be(buf + 8, (got - plt - 6) >> 1);
+}
+
+void SystemZ::writePlt(uint8_t *buf, const Symbol &sym,
+ uint64_t pltEntryAddr) const {
+ const uint8_t inst[] = {
+ 0xc0, 0x10, 0x00, 0x00, 0x00, 0x00, // larl %r1,<.got.plt slot>
+ 0xe3, 0x10, 0x10, 0x00, 0x00, 0x04, // lg %r1,0(%r1)
+ 0x07, 0xf1, // br %r1
+ 0x0d, 0x10, // basr %r1,%r0
+ 0xe3, 0x10, 0x10, 0x0c, 0x00, 0x14, // lgf %r1,12(%r1)
+ 0xc0, 0xf4, 0x00, 0x00, 0x00, 0x00, // jg <plt header>
+ 0x00, 0x00, 0x00, 0x00, // <relocation offset>
+ };
+ memcpy(buf, inst, sizeof(inst));
+
+ write32be(buf + 2, (sym.getGotPltVA() - pltEntryAddr) >> 1);
+ write32be(buf + 24, (in.plt->getVA() - pltEntryAddr - 22) >> 1);
+ write32be(buf + 28, in.relaPlt->entsize * sym.getPltIdx());
+}
+
+int64_t SystemZ::getImplicitAddend(const uint8_t *buf, RelType type) const {
+ switch (type) {
+ case R_390_8:
+ return SignExtend64<8>(*buf);
+ case R_390_16:
+ case R_390_PC16:
+ return SignExtend64<16>(read16be(buf));
+ case R_390_PC16DBL:
+ return SignExtend64<16>(read16be(buf)) << 1;
+ case R_390_32:
+ case R_390_PC32:
+ return SignExtend64<32>(read32be(buf));
+ case R_390_PC32DBL:
+ return SignExtend64<32>(read32be(buf)) << 1;
+ case R_390_64:
+ case R_390_PC64:
+ case R_390_TLS_DTPMOD:
+ case R_390_TLS_DTPOFF:
+ case R_390_TLS_TPOFF:
+ case R_390_GLOB_DAT:
+ case R_390_RELATIVE:
+ case R_390_IRELATIVE:
+ return read64be(buf);
+ case R_390_COPY:
+ case R_390_JMP_SLOT:
+ case R_390_NONE:
+ // These relocations are defined as not having an implicit addend.
+ return 0;
+ default:
+ internalLinkerError(getErrorLocation(buf),
+ "cannot read addend for relocation " + toString(type));
+ return 0;
+ }
+}
+
+RelType SystemZ::getDynRel(RelType type) const {
+ if (type == R_390_64 || type == R_390_PC64)
+ return type;
+ return R_390_NONE;
+}
+
+RelExpr SystemZ::adjustTlsExpr(RelType type, RelExpr expr) const {
+ if (expr == R_RELAX_TLS_GD_TO_IE)
+ return R_RELAX_TLS_GD_TO_IE_GOT_OFF;
+ return expr;
+}
+
+void SystemZ::relaxTlsGdToIe(uint8_t *loc, const Relocation &rel,
+ uint64_t val) const {
+ // The general-dynamic code sequence for a global `x`:
+ //
+ // Instruction Relocation Symbol
+ // ear %rX,%a0
+ // sllg %rX,%rX,32
+ // ear %rX,%a1
+ // larl %r12,_GLOBAL_OFFSET_TABLE_ R_390_GOTPCDBL _GLOBAL_OFFSET_TABLE_
+ // lgrl %r2,.LC0 R_390_PC32DBL .LC0
+ // brasl %r14,__tls_get_offset@plt R_390_TLS_GDCALL x
+ // :tls_gdcall:x R_390_PLT32DBL __tls_get_offset
+ // la %r2,0(%r2,%rX)
+ //
+ // .LC0:
+ // .quad x@TLSGD R_390_TLS_GD64 x
+ //
+ // Relaxing to initial-exec entails:
+ // 1) Replacing the call by a load from the GOT.
+ // 2) Replacing the relocation on the constant LC0 by R_390_TLS_GOTIE64.
+
+ switch (rel.type) {
+ case R_390_TLS_GDCALL:
+ /* brasl %r14,__tls_get_offset@plt -> lg %r2,0(%r2,%r12) */
+ write16be(loc, 0xe322);
+ write16be(loc + 2, 0xc000);
+ write16be(loc + 4, 0x0004);
+ break;
+ case R_390_TLS_GD64:
+ relocateNoSym(loc, R_390_TLS_GOTIE64, val);
+ break;
+ default:
+ llvm_unreachable("unsupported relocation for TLS GD to IE relaxation");
+ }
+}
+
+void SystemZ::relaxTlsGdToLe(uint8_t *loc, const Relocation &rel,
+ uint64_t val) const {
+ // The general-dynamic code sequence for a global `x`:
+ //
+ // Instruction Relocation Symbol
+ // ear %rX,%a0
+ // sllg %rX,%rX,32
+ // ear %rX,%a1
+ // larl %r12,_GLOBAL_OFFSET_TABLE_ R_390_GOTPCDBL _GLOBAL_OFFSET_TABLE_
+ // lgrl %r2,.LC0 R_390_PC32DBL .LC0
+ // brasl %r14,__tls_get_offset@plt R_390_TLS_GDCALL x
+ // :tls_gdcall:x R_390_PLT32DBL __tls_get_offset
+ // la %r2,0(%r2,%rX)
+ //
+ // .LC0:
+ // .quad x@tlsgd R_390_TLS_GD64 x
+ //
+ // Relaxing to local-exec entails:
+ // 1) Replacing the call by a nop.
+ // 2) Replacing the relocation on the constant LC0 by R_390_TLS_LE64.
+
+ switch (rel.type) {
+ case R_390_TLS_GDCALL:
+ /* brasl %r14,__tls_get_offset@plt -> brcl 0,. */
+ write16be(loc, 0xc004);
+ write16be(loc + 2, 0x0000);
+ write16be(loc + 4, 0x0000);
+ break;
+ case R_390_TLS_GD64:
+ relocateNoSym(loc, R_390_TLS_LE64, val);
+ break;
+ default:
+ llvm_unreachable("unsupported relocation for TLS GD to LE relaxation");
+ }
+}
+
+void SystemZ::relaxTlsLdToLe(uint8_t *loc, const Relocation &rel,
+ uint64_t val) const {
+ // The local-dynamic code sequence for a global `x`:
+ //
+ // Instruction Relocation Symbol
+ // ear %rX,%a0
+ // sllg %rX,%rX,32
+ // ear %rX,%a1
+ // larl %r12,_GLOBAL_OFFSET_TABLE_ R_390_GOTPCDBL _GLOBAL_OFFSET_TABLE_
+ // lgrl %r2,.LC0 R_390_PC32DBL .LC0
+ // brasl %r14,__tls_get_offset@plt R_390_TLS_LDCALL <sym>
+ // :tls_ldcall:<sym> R_390_PLT32DBL __tls_get_offset
+ // la %r2,0(%r2,%rX)
+ // lgrl %rY,.LC1 R_390_PC32DBL .LC1
+ // la %r2,0(%r2,%rY)
+ //
+ // .LC0:
+ // .quad <sym>@tlsldm R_390_TLS_LDM64 <sym>
+ // .LC1:
+ // .quad x@dtpoff R_390_TLS_LDO64 x
+ //
+ // Relaxing to local-exec entails:
+ // 1) Replacing the call by a nop.
+ // 2) Replacing the constant LC0 by 0 (i.e. ignoring the relocation).
+ // 3) Replacing the relocation on the constant LC1 by R_390_TLS_LE64.
+
+ switch (rel.type) {
+ case R_390_TLS_LDCALL:
+ /* brasl %r14,__tls_get_offset@plt -> brcl 0,. */
+ write16be(loc, 0xc004);
+ write16be(loc + 2, 0x0000);
+ write16be(loc + 4, 0x0000);
+ break;
+ case R_390_TLS_LDM64:
+ break;
+ case R_390_TLS_LDO64:
+ relocateNoSym(loc, R_390_TLS_LE64, val);
+ break;
+ default:
+ llvm_unreachable("unsupported relocation for TLS LD to LE relaxation");
+ }
+}
+
+RelExpr SystemZ::adjustGotPcExpr(RelType type, int64_t addend,
+ const uint8_t *loc) const {
+ // Only R_390_GOTENT with addend 2 can be relaxed.
+ if (!config->relax || addend != 2 || type != R_390_GOTENT)
+ return R_GOT_PC;
+ const uint16_t op = read16be(loc - 2);
+
+ // lgrl rx,sym@GOTENT -> larl rx, sym
+ // This relaxation is legal if "sym" binds locally (which was
+ // already verified by our caller) and is in-range and properly
+ // aligned for a LARL instruction. We cannot verify the latter
+ // constraint here, so we assume it is true and revert the decision
+ // later on in relaxOnce if necessary.
+ if ((op & 0xff0f) == 0xc408)
+ return R_RELAX_GOT_PC;
+
+ return R_GOT_PC;
+}
+
+bool SystemZ::relaxOnce(int pass) const {
+ // If we decided in adjustGotPcExpr to relax a R_390_GOTENT,
+ // we need to validate the target symbol is in-range and aligned.
+ SmallVector<InputSection *, 0> storage;
+ bool changed = false;
+ for (OutputSection *osec : outputSections) {
+ if (!(osec->flags & SHF_EXECINSTR))
+ continue;
+ for (InputSection *sec : getInputSections(*osec, storage)) {
+ for (Relocation &rel : sec->relocs()) {
+ if (rel.expr != R_RELAX_GOT_PC)
+ continue;
+
+ uint64_t v = sec->getRelocTargetVA(
+ sec->file, rel.type, rel.addend,
+ sec->getOutputSection()->addr + rel.offset, *rel.sym, rel.expr);
+ if (isInt<33>(v) && !(v & 1))
+ continue;
+ if (rel.sym->auxIdx == 0) {
+ rel.sym->allocateAux();
+ addGotEntry(*rel.sym);
+ changed = true;
+ }
+ rel.expr = R_GOT_PC;
+ }
+ }
+ }
+ return changed;
+}
+
+void SystemZ::relaxGot(uint8_t *loc, const Relocation &rel,
+ uint64_t val) const {
+ assert(isInt<33>(val) &&
+ "R_390_GOTENT should not have been relaxed if it overflows");
+ assert(!(val & 1) &&
+ "R_390_GOTENT should not have been relaxed if it is misaligned");
+ const uint16_t op = read16be(loc - 2);
+
+ // lgrl rx,sym@GOTENT -> larl rx, sym
+ if ((op & 0xff0f) == 0xc408) {
+ write16be(loc - 2, 0xc000 | (op & 0x00f0));
+ write32be(loc, val >> 1);
+ return;
+ }
+}
+
+void SystemZ::relocate(uint8_t *loc, const Relocation &rel, uint64_t val) const {
+ switch (rel.expr) {
+ case R_RELAX_GOT_PC:
+ return relaxGot(loc, rel, val);
+ case R_RELAX_TLS_GD_TO_IE_GOT_OFF:
+ return relaxTlsGdToIe(loc, rel, val);
+ case R_RELAX_TLS_GD_TO_LE:
+ return relaxTlsGdToLe(loc, rel, val);
+ case R_RELAX_TLS_LD_TO_LE:
+ return relaxTlsLdToLe(loc, rel, val);
+ default:
+ break;
+ }
+ switch (rel.type) {
+ case R_390_8:
+ checkIntUInt(loc, val, 8, rel);
+ *loc = val;
+ break;
+ case R_390_12:
+ case R_390_GOT12:
+ case R_390_GOTPLT12:
+ case R_390_TLS_GOTIE12:
+ checkUInt(loc, val, 12, rel);
+ write16be(loc, (read16be(loc) & 0xF000) | val);
+ break;
+ case R_390_PC12DBL:
+ case R_390_PLT12DBL:
+ checkInt(loc, val, 13, rel);
+ checkAlignment(loc, val, 2, rel);
+ write16be(loc, (read16be(loc) & 0xF000) | (val >> 1));
+ break;
+ case R_390_16:
+ case R_390_GOT16:
+ case R_390_GOTPLT16:
+ case R_390_GOTOFF16:
+ case R_390_PLTOFF16:
+ checkIntUInt(loc, val, 16, rel);
+ write16be(loc, val);
+ break;
+ case R_390_PC16:
+ checkInt(loc, val, 16, rel);
+ write16be(loc, val);
+ break;
+ case R_390_PC16DBL:
+ case R_390_PLT16DBL:
+ checkInt(loc, val, 17, rel);
+ checkAlignment(loc, val, 2, rel);
+ write16be(loc, val >> 1);
+ break;
+ case R_390_20:
+ case R_390_GOT20:
+ case R_390_GOTPLT20:
+ case R_390_TLS_GOTIE20:
+ checkInt(loc, val, 20, rel);
+ write32be(loc, (read32be(loc) & 0xF00000FF)
+ | ((val & 0xFFF) << 16) | ((val & 0xFF000) >> 4));
+ break;
+ case R_390_PC24DBL:
+ case R_390_PLT24DBL:
+ checkInt(loc, val, 25, rel);
+ loc[0] = val >> 17;
+ loc[1] = val >> 9;
+ loc[2] = val >> 1;
+ break;
+ case R_390_32:
+ case R_390_GOT32:
+ case R_390_GOTPLT32:
+ case R_390_GOTOFF:
+ case R_390_PLTOFF32:
+ case R_390_TLS_IE32:
+ case R_390_TLS_GOTIE32:
+ case R_390_TLS_GD32:
+ case R_390_TLS_LDM32:
+ case R_390_TLS_LDO32:
+ case R_390_TLS_LE32:
+ checkIntUInt(loc, val, 32, rel);
+ write32be(loc, val);
+ break;
+ case R_390_PC32:
+ case R_390_PLT32:
+ checkInt(loc, val, 32, rel);
+ write32be(loc, val);
+ break;
+ case R_390_PC32DBL:
+ case R_390_PLT32DBL:
+ case R_390_GOTPCDBL:
+ case R_390_GOTENT:
+ case R_390_GOTPLTENT:
+ case R_390_TLS_IEENT:
+ checkInt(loc, val, 33, rel);
+ checkAlignment(loc, val, 2, rel);
+ write32be(loc, val >> 1);
+ break;
+ case R_390_64:
+ case R_390_PC64:
+ case R_390_PLT64:
+ case R_390_GOT64:
+ case R_390_GOTPLT64:
+ case R_390_GOTOFF64:
+ case R_390_PLTOFF64:
+ case R_390_GOTPC:
+ case R_390_TLS_IE64:
+ case R_390_TLS_GOTIE64:
+ case R_390_TLS_GD64:
+ case R_390_TLS_LDM64:
+ case R_390_TLS_LDO64:
+ case R_390_TLS_LE64:
+ case R_390_TLS_DTPMOD:
+ case R_390_TLS_DTPOFF:
+ case R_390_TLS_TPOFF:
+ write64be(loc, val);
+ break;
+ case R_390_TLS_LOAD:
+ case R_390_TLS_GDCALL:
+ case R_390_TLS_LDCALL:
+ break;
+ default:
+ error(getErrorLocation(loc) + "unknown relocation (" + Twine(rel.type) + ")");
+ llvm_unreachable("unknown relocation");
+ }
+}
+
+static TargetInfo *getTargetInfo() {
+ static SystemZ t;
+ return &t;
+}
+
+TargetInfo *elf::getSystemZTargetInfo() { return getTargetInfo(); }
diff --git a/lld/ELF/CMakeLists.txt b/lld/ELF/CMakeLists.txt
index 475f7dea1dd7e9..83d816ddb0601e 100644
--- a/lld/ELF/CMakeLists.txt
+++ b/lld/ELF/CMakeLists.txt
@@ -33,6 +33,7 @@ add_lld_library(lldELF
Arch/PPC64.cpp
Arch/RISCV.cpp
Arch/SPARCV9.cpp
+ Arch/SystemZ.cpp
Arch/X86.cpp
Arch/X86_64.cpp
ARMErrataFix.cpp
diff --git a/lld/ELF/Driver.cpp b/lld/ELF/Driver.cpp
index 6bef09eeca015a..d1514556f013f9 100644
--- a/lld/ELF/Driver.cpp
+++ b/lld/ELF/Driver.cpp
@@ -199,6 +199,7 @@ static std::tuple<ELFKind, uint16_t, uint8_t> parseEmulation(StringRef emul) {
.Case("msp430elf", {ELF32LEKind, EM_MSP430})
.Case("elf64_amdgpu", {ELF64LEKind, EM_AMDGPU})
.Case("elf64loongarch", {ELF64LEKind, EM_LOONGARCH})
+ .Case("elf64_s390", {ELF64BEKind, EM_S390})
.Default({ELFNoneKind, EM_NONE});
if (ret.first == ELFNoneKind)
@@ -1171,7 +1172,7 @@ static bool getIsRela(opt::InputArgList &args) {
uint16_t m = config->emachine;
return m == EM_AARCH64 || m == EM_AMDGPU || m == EM_HEXAGON ||
m == EM_LOONGARCH || m == EM_PPC || m == EM_PPC64 || m == EM_RISCV ||
- m == EM_X86_64;
+ m == EM_S390 || m == EM_X86_64;
}
static void parseClangOption(StringRef opt, const Twine &msg) {
diff --git a/lld/ELF/InputFiles.cpp b/lld/ELF/InputFiles.cpp
index cc2c5916e05c22..574e9c59fc6c9b 100644
--- a/lld/ELF/InputFiles.cpp
+++ b/lld/ELF/InputFiles.cpp
@@ -1612,6 +1612,8 @@ static uint16_t getBitcodeMachineKind(StringRef path, const Triple &t) {
return EM_RISCV;
case Triple::sparcv9:
return EM_SPARCV9;
+ case Triple::systemz:
+ return EM_S39...
[truncated]
|
|
@llvm/pr-subscribers-lld Author: Ulrich Weigand (uweigand) ChangesThis patch adds full support for linking SystemZ (ELF s390x) object files. Support should be generally complete:
In addition to new platform code and the obvious changes, there were a few additional changes to common code:
This was tested by building and testing >1500 Fedora packages, with only a handful of failures; as these also have issues when building with LLD on other architectures, they seem unrelated. Patch is 88.34 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/75643.diff 38 Files Affected:
diff --git a/lld/ELF/Arch/SystemZ.cpp b/lld/ELF/Arch/SystemZ.cpp
new file mode 100644
index 00000000000000..80a65cf32b6c32
--- /dev/null
+++ b/lld/ELF/Arch/SystemZ.cpp
@@ -0,0 +1,581 @@
+//===- SystemZ.cpp --------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "OutputSections.h"
+#include "Symbols.h"
+#include "SyntheticSections.h"
+#include "Target.h"
+#include "lld/Common/ErrorHandler.h"
+#include "llvm/BinaryFormat/ELF.h"
+#include "llvm/Support/Endian.h"
+
+using namespace llvm;
+using namespace llvm::object;
+using namespace llvm::support::endian;
+using namespace llvm::ELF;
+using namespace lld;
+using namespace lld::elf;
+
+namespace {
+class SystemZ : public TargetInfo {
+public:
+ SystemZ();
+ RelExpr getRelExpr(RelType type, const Symbol &s,
+ const uint8_t *loc) const override;
+ RelType getDynRel(RelType type) const override;
+ void writeGotHeader(uint8_t *buf) const override;
+ void writeGotPlt(uint8_t *buf, const Symbol &s) const override;
+ void writeIgotPlt(uint8_t *buf, const Symbol &s) const override;
+ void writePltHeader(uint8_t *buf) const override;
+ void writePlt(uint8_t *buf, const Symbol &sym,
+ uint64_t pltEntryAddr) const override;
+ RelExpr adjustTlsExpr(RelType type, RelExpr expr) const override;
+ RelExpr adjustGotPcExpr(RelType type, int64_t addend,
+ const uint8_t *loc) const override;
+ bool relaxOnce(int pass) const override;
+ void relocate(uint8_t *loc, const Relocation &rel,
+ uint64_t val) const override;
+ int64_t getImplicitAddend(const uint8_t *buf, RelType type) const override;
+private:
+ void relaxGot(uint8_t *loc, const Relocation &rel, uint64_t val) const;
+ void relaxTlsGdToIe(uint8_t *loc, const Relocation &rel, uint64_t val) const;
+ void relaxTlsGdToLe(uint8_t *loc, const Relocation &rel, uint64_t val) const;
+ void relaxTlsLdToLe(uint8_t *loc, const Relocation &rel, uint64_t val) const;
+};
+} // namespace
+
+SystemZ::SystemZ() {
+ copyRel = R_390_COPY;
+ gotRel = R_390_GLOB_DAT;
+ pltRel = R_390_JMP_SLOT;
+ relativeRel = R_390_RELATIVE;
+ iRelativeRel = R_390_IRELATIVE;
+ symbolicRel = R_390_64;
+ tlsGotRel = R_390_TLS_TPOFF;
+ tlsModuleIndexRel = R_390_TLS_DTPMOD;
+ tlsOffsetRel = R_390_TLS_DTPOFF;
+ gotHeaderEntriesNum = 3;
+ gotPltHeaderEntriesNum = 0;
+ gotEntrySize = 8;
+ pltHeaderSize = 32;
+ pltEntrySize = 32;
+ ipltEntrySize = 32;
+
+ // This "trap instruction" is used to fill gaps between sections.
+ // On SystemZ, the behavior of the GNU ld is to fill those gaps
+ // with nop instructions instead - and unfortunately the default
+ // glibc crt object files (used to) rely on that behavior since
+ // they use an alignment on the .init section fragments that causes
+ // gaps which must be filled with nops as they are being executed.
+ // Therefore, we provide a nop instruction as "trapInstr" here.
+ trapInstr = {0x07, 0x07, 0x07, 0x07};
+
+ defaultImageBase = 0x1000000;
+}
+
+RelExpr SystemZ::getRelExpr(RelType type, const Symbol &s,
+ const uint8_t *loc) const {
+ switch (type) {
+ case R_390_NONE:
+ return R_NONE;
+ // Relocations targeting the symbol value.
+ case R_390_8:
+ case R_390_12:
+ case R_390_16:
+ case R_390_20:
+ case R_390_32:
+ case R_390_64:
+ return R_ABS;
+ case R_390_PC16:
+ case R_390_PC32:
+ case R_390_PC64:
+ case R_390_PC12DBL:
+ case R_390_PC16DBL:
+ case R_390_PC24DBL:
+ case R_390_PC32DBL:
+ return R_PC;
+ case R_390_GOTOFF16:
+ case R_390_GOTOFF: // a.k.a. R_390_GOTOFF32
+ case R_390_GOTOFF64:
+ return R_GOTREL;
+ // Relocations targeting the PLT associated with the symbol.
+ case R_390_PLT32:
+ case R_390_PLT64:
+ case R_390_PLT12DBL:
+ case R_390_PLT16DBL:
+ case R_390_PLT24DBL:
+ case R_390_PLT32DBL:
+ return R_PLT_PC;
+ case R_390_PLTOFF16:
+ case R_390_PLTOFF32:
+ case R_390_PLTOFF64:
+ return R_PLT_GOTREL;
+ // Relocations targeting the GOT entry associated with the symbol.
+ case R_390_GOTENT:
+ return R_GOT_PC;
+ case R_390_GOT12:
+ case R_390_GOT16:
+ case R_390_GOT20:
+ case R_390_GOT32:
+ case R_390_GOT64:
+ return R_GOT_OFF;
+ // Relocations targeting the GOTPLT entry associated with the symbol.
+ case R_390_GOTPLTENT:
+ return R_GOTPLT_PC;
+ case R_390_GOTPLT12:
+ case R_390_GOTPLT16:
+ case R_390_GOTPLT20:
+ case R_390_GOTPLT32:
+ case R_390_GOTPLT64:
+ return R_GOTPLT_OFF;
+ // Relocations targeting _GLOBAL_OFFSET_TABLE_.
+ case R_390_GOTPC:
+ case R_390_GOTPCDBL:
+ return R_GOTONLY_PC;
+ // TLS-related relocations.
+ case R_390_TLS_LOAD:
+ return R_NONE;
+ case R_390_TLS_GDCALL:
+ return R_TLSGD_PC;
+ case R_390_TLS_LDCALL:
+ return R_TLSLD_PC;
+ case R_390_TLS_GD32:
+ case R_390_TLS_GD64:
+ return R_TLSGD_GOT;
+ case R_390_TLS_LDM32:
+ case R_390_TLS_LDM64:
+ return R_TLSLD_GOT;
+ case R_390_TLS_LDO32:
+ case R_390_TLS_LDO64:
+ return R_DTPREL;
+ case R_390_TLS_LE32:
+ case R_390_TLS_LE64:
+ return R_TPREL;
+ case R_390_TLS_IE32:
+ case R_390_TLS_IE64:
+ return R_GOT;
+ case R_390_TLS_GOTIE12:
+ case R_390_TLS_GOTIE20:
+ case R_390_TLS_GOTIE32:
+ case R_390_TLS_GOTIE64:
+ return R_GOT_OFF;
+ case R_390_TLS_IEENT:
+ return R_GOT_PC;
+
+ default:
+ error(getErrorLocation(loc) + "unknown relocation (" + Twine(type) +
+ ") against symbol " + toString(s));
+ return R_NONE;
+ }
+}
+
+void SystemZ::writeGotHeader(uint8_t *buf) const {
+ // _GLOBAL_OFFSET_TABLE_[0] holds the value of _DYNAMIC.
+ // _GLOBAL_OFFSET_TABLE_[1] and [2] are reserved.
+ write64be(buf, mainPart->dynamic->getVA());
+}
+
+void SystemZ::writeGotPlt(uint8_t *buf, const Symbol &s) const {
+ write64be(buf, s.getPltVA() + 14);
+}
+
+void SystemZ::writeIgotPlt(uint8_t *buf, const Symbol &s) const {
+ if (config->writeAddends)
+ write64be(buf, s.getVA());
+}
+
+void SystemZ::writePltHeader(uint8_t *buf) const {
+ const uint8_t pltData[] = {
+ 0xe3, 0x10, 0xf0, 0x38, 0x00, 0x24, // stg %r1,56(%r15)
+ 0xc0, 0x10, 0x00, 0x00, 0x00, 0x00, // larl %r1,_GLOBAL_OFFSET_TABLE_
+ 0xd2, 0x07, 0xf0, 0x30, 0x10, 0x08, // mvc 48(8,%r15),8(%r1)
+ 0xe3, 0x10, 0x10, 0x10, 0x00, 0x04, // lg %r1,16(%r1)
+ 0x07, 0xf1, // br %r1
+ 0x07, 0x00, // nopr
+ 0x07, 0x00, // nopr
+ 0x07, 0x00, // nopr
+ };
+ memcpy(buf, pltData, sizeof(pltData));
+ uint64_t got = in.got->getVA();
+ uint64_t plt = in.plt->getVA();
+ write32be(buf + 8, (got - plt - 6) >> 1);
+}
+
+void SystemZ::writePlt(uint8_t *buf, const Symbol &sym,
+ uint64_t pltEntryAddr) const {
+ const uint8_t inst[] = {
+ 0xc0, 0x10, 0x00, 0x00, 0x00, 0x00, // larl %r1,<.got.plt slot>
+ 0xe3, 0x10, 0x10, 0x00, 0x00, 0x04, // lg %r1,0(%r1)
+ 0x07, 0xf1, // br %r1
+ 0x0d, 0x10, // basr %r1,%r0
+ 0xe3, 0x10, 0x10, 0x0c, 0x00, 0x14, // lgf %r1,12(%r1)
+ 0xc0, 0xf4, 0x00, 0x00, 0x00, 0x00, // jg <plt header>
+ 0x00, 0x00, 0x00, 0x00, // <relocation offset>
+ };
+ memcpy(buf, inst, sizeof(inst));
+
+ write32be(buf + 2, (sym.getGotPltVA() - pltEntryAddr) >> 1);
+ write32be(buf + 24, (in.plt->getVA() - pltEntryAddr - 22) >> 1);
+ write32be(buf + 28, in.relaPlt->entsize * sym.getPltIdx());
+}
+
+int64_t SystemZ::getImplicitAddend(const uint8_t *buf, RelType type) const {
+ switch (type) {
+ case R_390_8:
+ return SignExtend64<8>(*buf);
+ case R_390_16:
+ case R_390_PC16:
+ return SignExtend64<16>(read16be(buf));
+ case R_390_PC16DBL:
+ return SignExtend64<16>(read16be(buf)) << 1;
+ case R_390_32:
+ case R_390_PC32:
+ return SignExtend64<32>(read32be(buf));
+ case R_390_PC32DBL:
+ return SignExtend64<32>(read32be(buf)) << 1;
+ case R_390_64:
+ case R_390_PC64:
+ case R_390_TLS_DTPMOD:
+ case R_390_TLS_DTPOFF:
+ case R_390_TLS_TPOFF:
+ case R_390_GLOB_DAT:
+ case R_390_RELATIVE:
+ case R_390_IRELATIVE:
+ return read64be(buf);
+ case R_390_COPY:
+ case R_390_JMP_SLOT:
+ case R_390_NONE:
+ // These relocations are defined as not having an implicit addend.
+ return 0;
+ default:
+ internalLinkerError(getErrorLocation(buf),
+ "cannot read addend for relocation " + toString(type));
+ return 0;
+ }
+}
+
+RelType SystemZ::getDynRel(RelType type) const {
+ if (type == R_390_64 || type == R_390_PC64)
+ return type;
+ return R_390_NONE;
+}
+
+RelExpr SystemZ::adjustTlsExpr(RelType type, RelExpr expr) const {
+ if (expr == R_RELAX_TLS_GD_TO_IE)
+ return R_RELAX_TLS_GD_TO_IE_GOT_OFF;
+ return expr;
+}
+
+void SystemZ::relaxTlsGdToIe(uint8_t *loc, const Relocation &rel,
+ uint64_t val) const {
+ // The general-dynamic code sequence for a global `x`:
+ //
+ // Instruction Relocation Symbol
+ // ear %rX,%a0
+ // sllg %rX,%rX,32
+ // ear %rX,%a1
+ // larl %r12,_GLOBAL_OFFSET_TABLE_ R_390_GOTPCDBL _GLOBAL_OFFSET_TABLE_
+ // lgrl %r2,.LC0 R_390_PC32DBL .LC0
+ // brasl %r14,__tls_get_offset@plt R_390_TLS_GDCALL x
+ // :tls_gdcall:x R_390_PLT32DBL __tls_get_offset
+ // la %r2,0(%r2,%rX)
+ //
+ // .LC0:
+ // .quad x@TLSGD R_390_TLS_GD64 x
+ //
+ // Relaxing to initial-exec entails:
+ // 1) Replacing the call by a load from the GOT.
+ // 2) Replacing the relocation on the constant LC0 by R_390_TLS_GOTIE64.
+
+ switch (rel.type) {
+ case R_390_TLS_GDCALL:
+ /* brasl %r14,__tls_get_offset@plt -> lg %r2,0(%r2,%r12) */
+ write16be(loc, 0xe322);
+ write16be(loc + 2, 0xc000);
+ write16be(loc + 4, 0x0004);
+ break;
+ case R_390_TLS_GD64:
+ relocateNoSym(loc, R_390_TLS_GOTIE64, val);
+ break;
+ default:
+ llvm_unreachable("unsupported relocation for TLS GD to IE relaxation");
+ }
+}
+
+void SystemZ::relaxTlsGdToLe(uint8_t *loc, const Relocation &rel,
+ uint64_t val) const {
+ // The general-dynamic code sequence for a global `x`:
+ //
+ // Instruction Relocation Symbol
+ // ear %rX,%a0
+ // sllg %rX,%rX,32
+ // ear %rX,%a1
+ // larl %r12,_GLOBAL_OFFSET_TABLE_ R_390_GOTPCDBL _GLOBAL_OFFSET_TABLE_
+ // lgrl %r2,.LC0 R_390_PC32DBL .LC0
+ // brasl %r14,__tls_get_offset@plt R_390_TLS_GDCALL x
+ // :tls_gdcall:x R_390_PLT32DBL __tls_get_offset
+ // la %r2,0(%r2,%rX)
+ //
+ // .LC0:
+ // .quad x@tlsgd R_390_TLS_GD64 x
+ //
+ // Relaxing to local-exec entails:
+ // 1) Replacing the call by a nop.
+ // 2) Replacing the relocation on the constant LC0 by R_390_TLS_LE64.
+
+ switch (rel.type) {
+ case R_390_TLS_GDCALL:
+ /* brasl %r14,__tls_get_offset@plt -> brcl 0,. */
+ write16be(loc, 0xc004);
+ write16be(loc + 2, 0x0000);
+ write16be(loc + 4, 0x0000);
+ break;
+ case R_390_TLS_GD64:
+ relocateNoSym(loc, R_390_TLS_LE64, val);
+ break;
+ default:
+ llvm_unreachable("unsupported relocation for TLS GD to LE relaxation");
+ }
+}
+
+void SystemZ::relaxTlsLdToLe(uint8_t *loc, const Relocation &rel,
+ uint64_t val) const {
+ // The local-dynamic code sequence for a global `x`:
+ //
+ // Instruction Relocation Symbol
+ // ear %rX,%a0
+ // sllg %rX,%rX,32
+ // ear %rX,%a1
+ // larl %r12,_GLOBAL_OFFSET_TABLE_ R_390_GOTPCDBL _GLOBAL_OFFSET_TABLE_
+ // lgrl %r2,.LC0 R_390_PC32DBL .LC0
+ // brasl %r14,__tls_get_offset@plt R_390_TLS_LDCALL <sym>
+ // :tls_ldcall:<sym> R_390_PLT32DBL __tls_get_offset
+ // la %r2,0(%r2,%rX)
+ // lgrl %rY,.LC1 R_390_PC32DBL .LC1
+ // la %r2,0(%r2,%rY)
+ //
+ // .LC0:
+ // .quad <sym>@tlsldm R_390_TLS_LDM64 <sym>
+ // .LC1:
+ // .quad x@dtpoff R_390_TLS_LDO64 x
+ //
+ // Relaxing to local-exec entails:
+ // 1) Replacing the call by a nop.
+ // 2) Replacing the constant LC0 by 0 (i.e. ignoring the relocation).
+ // 3) Replacing the relocation on the constant LC1 by R_390_TLS_LE64.
+
+ switch (rel.type) {
+ case R_390_TLS_LDCALL:
+ /* brasl %r14,__tls_get_offset@plt -> brcl 0,. */
+ write16be(loc, 0xc004);
+ write16be(loc + 2, 0x0000);
+ write16be(loc + 4, 0x0000);
+ break;
+ case R_390_TLS_LDM64:
+ break;
+ case R_390_TLS_LDO64:
+ relocateNoSym(loc, R_390_TLS_LE64, val);
+ break;
+ default:
+ llvm_unreachable("unsupported relocation for TLS LD to LE relaxation");
+ }
+}
+
+RelExpr SystemZ::adjustGotPcExpr(RelType type, int64_t addend,
+ const uint8_t *loc) const {
+ // Only R_390_GOTENT with addend 2 can be relaxed.
+ if (!config->relax || addend != 2 || type != R_390_GOTENT)
+ return R_GOT_PC;
+ const uint16_t op = read16be(loc - 2);
+
+ // lgrl rx,sym@GOTENT -> larl rx, sym
+ // This relaxation is legal if "sym" binds locally (which was
+ // already verified by our caller) and is in-range and properly
+ // aligned for a LARL instruction. We cannot verify the latter
+ // constraint here, so we assume it is true and revert the decision
+ // later on in relaxOnce if necessary.
+ if ((op & 0xff0f) == 0xc408)
+ return R_RELAX_GOT_PC;
+
+ return R_GOT_PC;
+}
+
+bool SystemZ::relaxOnce(int pass) const {
+ // If we decided in adjustGotPcExpr to relax a R_390_GOTENT,
+ // we need to validate the target symbol is in-range and aligned.
+ SmallVector<InputSection *, 0> storage;
+ bool changed = false;
+ for (OutputSection *osec : outputSections) {
+ if (!(osec->flags & SHF_EXECINSTR))
+ continue;
+ for (InputSection *sec : getInputSections(*osec, storage)) {
+ for (Relocation &rel : sec->relocs()) {
+ if (rel.expr != R_RELAX_GOT_PC)
+ continue;
+
+ uint64_t v = sec->getRelocTargetVA(
+ sec->file, rel.type, rel.addend,
+ sec->getOutputSection()->addr + rel.offset, *rel.sym, rel.expr);
+ if (isInt<33>(v) && !(v & 1))
+ continue;
+ if (rel.sym->auxIdx == 0) {
+ rel.sym->allocateAux();
+ addGotEntry(*rel.sym);
+ changed = true;
+ }
+ rel.expr = R_GOT_PC;
+ }
+ }
+ }
+ return changed;
+}
+
+void SystemZ::relaxGot(uint8_t *loc, const Relocation &rel,
+ uint64_t val) const {
+ assert(isInt<33>(val) &&
+ "R_390_GOTENT should not have been relaxed if it overflows");
+ assert(!(val & 1) &&
+ "R_390_GOTENT should not have been relaxed if it is misaligned");
+ const uint16_t op = read16be(loc - 2);
+
+ // lgrl rx,sym@GOTENT -> larl rx, sym
+ if ((op & 0xff0f) == 0xc408) {
+ write16be(loc - 2, 0xc000 | (op & 0x00f0));
+ write32be(loc, val >> 1);
+ return;
+ }
+}
+
+void SystemZ::relocate(uint8_t *loc, const Relocation &rel, uint64_t val) const {
+ switch (rel.expr) {
+ case R_RELAX_GOT_PC:
+ return relaxGot(loc, rel, val);
+ case R_RELAX_TLS_GD_TO_IE_GOT_OFF:
+ return relaxTlsGdToIe(loc, rel, val);
+ case R_RELAX_TLS_GD_TO_LE:
+ return relaxTlsGdToLe(loc, rel, val);
+ case R_RELAX_TLS_LD_TO_LE:
+ return relaxTlsLdToLe(loc, rel, val);
+ default:
+ break;
+ }
+ switch (rel.type) {
+ case R_390_8:
+ checkIntUInt(loc, val, 8, rel);
+ *loc = val;
+ break;
+ case R_390_12:
+ case R_390_GOT12:
+ case R_390_GOTPLT12:
+ case R_390_TLS_GOTIE12:
+ checkUInt(loc, val, 12, rel);
+ write16be(loc, (read16be(loc) & 0xF000) | val);
+ break;
+ case R_390_PC12DBL:
+ case R_390_PLT12DBL:
+ checkInt(loc, val, 13, rel);
+ checkAlignment(loc, val, 2, rel);
+ write16be(loc, (read16be(loc) & 0xF000) | (val >> 1));
+ break;
+ case R_390_16:
+ case R_390_GOT16:
+ case R_390_GOTPLT16:
+ case R_390_GOTOFF16:
+ case R_390_PLTOFF16:
+ checkIntUInt(loc, val, 16, rel);
+ write16be(loc, val);
+ break;
+ case R_390_PC16:
+ checkInt(loc, val, 16, rel);
+ write16be(loc, val);
+ break;
+ case R_390_PC16DBL:
+ case R_390_PLT16DBL:
+ checkInt(loc, val, 17, rel);
+ checkAlignment(loc, val, 2, rel);
+ write16be(loc, val >> 1);
+ break;
+ case R_390_20:
+ case R_390_GOT20:
+ case R_390_GOTPLT20:
+ case R_390_TLS_GOTIE20:
+ checkInt(loc, val, 20, rel);
+ write32be(loc, (read32be(loc) & 0xF00000FF)
+ | ((val & 0xFFF) << 16) | ((val & 0xFF000) >> 4));
+ break;
+ case R_390_PC24DBL:
+ case R_390_PLT24DBL:
+ checkInt(loc, val, 25, rel);
+ loc[0] = val >> 17;
+ loc[1] = val >> 9;
+ loc[2] = val >> 1;
+ break;
+ case R_390_32:
+ case R_390_GOT32:
+ case R_390_GOTPLT32:
+ case R_390_GOTOFF:
+ case R_390_PLTOFF32:
+ case R_390_TLS_IE32:
+ case R_390_TLS_GOTIE32:
+ case R_390_TLS_GD32:
+ case R_390_TLS_LDM32:
+ case R_390_TLS_LDO32:
+ case R_390_TLS_LE32:
+ checkIntUInt(loc, val, 32, rel);
+ write32be(loc, val);
+ break;
+ case R_390_PC32:
+ case R_390_PLT32:
+ checkInt(loc, val, 32, rel);
+ write32be(loc, val);
+ break;
+ case R_390_PC32DBL:
+ case R_390_PLT32DBL:
+ case R_390_GOTPCDBL:
+ case R_390_GOTENT:
+ case R_390_GOTPLTENT:
+ case R_390_TLS_IEENT:
+ checkInt(loc, val, 33, rel);
+ checkAlignment(loc, val, 2, rel);
+ write32be(loc, val >> 1);
+ break;
+ case R_390_64:
+ case R_390_PC64:
+ case R_390_PLT64:
+ case R_390_GOT64:
+ case R_390_GOTPLT64:
+ case R_390_GOTOFF64:
+ case R_390_PLTOFF64:
+ case R_390_GOTPC:
+ case R_390_TLS_IE64:
+ case R_390_TLS_GOTIE64:
+ case R_390_TLS_GD64:
+ case R_390_TLS_LDM64:
+ case R_390_TLS_LDO64:
+ case R_390_TLS_LE64:
+ case R_390_TLS_DTPMOD:
+ case R_390_TLS_DTPOFF:
+ case R_390_TLS_TPOFF:
+ write64be(loc, val);
+ break;
+ case R_390_TLS_LOAD:
+ case R_390_TLS_GDCALL:
+ case R_390_TLS_LDCALL:
+ break;
+ default:
+ error(getErrorLocation(loc) + "unknown relocation (" + Twine(rel.type) + ")");
+ llvm_unreachable("unknown relocation");
+ }
+}
+
+static TargetInfo *getTargetInfo() {
+ static SystemZ t;
+ return &t;
+}
+
+TargetInfo *elf::getSystemZTargetInfo() { return getTargetInfo(); }
diff --git a/lld/ELF/CMakeLists.txt b/lld/ELF/CMakeLists.txt
index 475f7dea1dd7e9..83d816ddb0601e 100644
--- a/lld/ELF/CMakeLists.txt
+++ b/lld/ELF/CMakeLists.txt
@@ -33,6 +33,7 @@ add_lld_library(lldELF
Arch/PPC64.cpp
Arch/RISCV.cpp
Arch/SPARCV9.cpp
+ Arch/SystemZ.cpp
Arch/X86.cpp
Arch/X86_64.cpp
ARMErrataFix.cpp
diff --git a/lld/ELF/Driver.cpp b/lld/ELF/Driver.cpp
index 6bef09eeca015a..d1514556f013f9 100644
--- a/lld/ELF/Driver.cpp
+++ b/lld/ELF/Driver.cpp
@@ -199,6 +199,7 @@ static std::tuple<ELFKind, uint16_t, uint8_t> parseEmulation(StringRef emul) {
.Case("msp430elf", {ELF32LEKind, EM_MSP430})
.Case("elf64_amdgpu", {ELF64LEKind, EM_AMDGPU})
.Case("elf64loongarch", {ELF64LEKind, EM_LOONGARCH})
+ .Case("elf64_s390", {ELF64BEKind, EM_S390})
.Default({ELFNoneKind, EM_NONE});
if (ret.first == ELFNoneKind)
@@ -1171,7 +1172,7 @@ static bool getIsRela(opt::InputArgList &args) {
uint16_t m = config->emachine;
return m == EM_AARCH64 || m == EM_AMDGPU || m == EM_HEXAGON ||
m == EM_LOONGARCH || m == EM_PPC || m == EM_PPC64 || m == EM_RISCV ||
- m == EM_X86_64;
+ m == EM_S390 || m == EM_X86_64;
}
static void parseClangOption(StringRef opt, const Twine &msg) {
diff --git a/lld/ELF/InputFiles.cpp b/lld/ELF/InputFiles.cpp
index cc2c5916e05c22..574e9c59fc6c9b 100644
--- a/lld/ELF/InputFiles.cpp
+++ b/lld/ELF/InputFiles.cpp
@@ -1612,6 +1612,8 @@ static uint16_t getBitcodeMachineKind(StringRef path, const Triple &t) {
return EM_RISCV;
case Triple::sparcv9:
return EM_SPARCV9;
+ case Triple::systemz:
+ return EM_S39...
[truncated]
|
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
MaskRay
reviewed
Dec 20, 2023
lld/test/ELF/basic-systemz.s
Outdated
| # REQUIRES: systemz | ||
| # RUN: llvm-mc -filetype=obj -triple=s390x-unknown-linux %s -o %t.o | ||
| # RUN: ld.lld --hash-style=sysv -discard-all -shared %t.o -o %t.so | ||
| # RUN: llvm-readobj --file-headers --sections --section-data -l %t.so | FileCheck %s |
There was a problem hiding this comment.
I know that existing basic-* use llvm-readobj, which has a very verbose output. Consider llvm-readelf for new tests
There was a problem hiding this comment.
Switched to using llvm-readelf everywhere.
lld/ELF/Arch/SystemZ.cpp
Outdated
| } | ||
| } | ||
|
|
||
| static TargetInfo *getTargetInfo() { |
lld/test/ELF/systemz-pie.s
Outdated
|
|
||
| ## Default is no PIE. | ||
| # RUN: ld.lld %t1.o -o %t | ||
| # RUN: llvm-readobj --file-headers --sections -l --symbols -r %t \ |
There was a problem hiding this comment.
Suggest that this is added to basic-* and switch to llvm-readelf
lld/test/ELF/systemz-pie.s
Outdated
| # RUN: ld.lld -pie %t1.o -o %t | ||
| # RUN: llvm-readobj --file-headers --sections -l -d --symbols -r %t | FileCheck %s | ||
|
|
||
| ## Test --pic-executable alias |
There was a problem hiding this comment.
Delete --pic-executable test. Having this in pie.s is sufficient.
ppc32 is a good example for test organization.
lld/test/ELF/systemz-plt.s
Outdated
| @@ -0,0 +1,73 @@ | |||
| // REQUIRES: systemz | |||
There was a problem hiding this comment.
Test more symbols like riscv-plt.s
There was a problem hiding this comment.
Done. This actually uncovered a bug - it was possible for .got to be removed completely in a file that uses .plt - but the PLT header refers to _GLOBAL_OFFSET_TABLE_ which points to the start of .got on SystemZ ... Fixed by enforcing the presence of .got in SystemZ::addPltHeaderSymbols.
MaskRay
reviewed
Dec 20, 2023
lld/ELF/Relocations.cpp
Outdated
| // defined. | ||
| if (toExecRelax && isLocalInExecutable) { | ||
| // defined. This is not supported on SystemZ. | ||
| if (toExecRelax && isLocalInExecutable && config->emachine != EM_S390) { |
There was a problem hiding this comment.
Update bool toExecRelax declaration instead?
There was a problem hiding this comment.
But that would disable all the relaxations. We do have Local-Dynamic to Local-Exec and Gobal-Dynamic to Initial/Local-Exec. We just cannot do Initial-Exec to Local-Exec relaxation.
MaskRay
reviewed
Dec 20, 2023
lld/ELF/Relocations.cpp
Outdated
| // (i.e. TLS relaxation applies), and the relocation *after* the current one | ||
| // is a TLS call marker instruction matching the current instruction, then | ||
| // skip this relocation. | ||
| if (config->emachine == EM_S390 && !config->shared) { |
There was a problem hiding this comment.
This adds overhead for other architectures and the ppc64 is not a good example to follow. Is it possible to not make other architectures pay the branch cost ?
There was a problem hiding this comment.
I tried to find a better way, but didn't see anything obvious. Given how large and complex this function already is, would that single (normally very well-predicted) branch really be noticeable?
There was a problem hiding this comment.
Perhaps an alternative is to sort relocations at Relocations.cpp:1574 Sort relocations by offset for more efficient searching for, similar to PPC64 and RISC-V.
After sorting, R_390_TLS_LDCALL will come before R_390_PLT32DBL
There was a problem hiding this comment.
This does indeed seem to work. (However, we need to sort before the scanOne calls, not afterwards.) Patch updated accordingly.
MaskRay
reviewed
Dec 20, 2023
lld/test/ELF/systemz-reloc-disp20.s
Outdated
| _start: | ||
| .reloc .+2, R_390_20, DISP | ||
| lg %r6, 0(%r7,%r8) | ||
|
|
There was a problem hiding this comment.
delete blank line in the end of file. fix this everywhere
uweigand
force-pushed
the
lld-s390x
branch
3 times, most recently
from
fd92a1b
to
166483a
Compare
|
Fixed merge conflict. Ping? |
|
This is a very large patch. There are quite few other large changes as well and my spare time has been scarce. It may take a while for me to come to this... and I will need to learn some SystemZ basics. |
|
This is great! Thanks for all of this work!
Any chance you tried linking the Linux kernel with LLD for SystemZ? |
@tuliom this would be a question for you ... |
Thanks! If you'd like to review the SystemZ specifics, you'll primarily need the ABI document here: https://github.com/IBM/s390x-abi (you should check current mainline, which has some fixes against the latest published version - we actually noticed those bugs while implementing LLD support ...) |
|
Thanks for the link!
Can you ask the team to vendor a new release? The last published pdf was Nov '21. |
|
I took this for a spin against the Linux kernel and immediately found two problems, although they do not appears to be strictly related to this pull request. It appears that I introduced the first issue in commit 2b2a25845d53 (" I think the solution is just to remove this altogether: diff --git a/arch/s390/kernel/vdso64/Makefile b/arch/s390/kernel/vdso64/Makefile
index e3c9085f8fa7..caa4ebff8a19 100644
--- a/arch/s390/kernel/vdso64/Makefile
+++ b/arch/s390/kernel/vdso64/Makefile
@@ -26,7 +26,7 @@ KBUILD_AFLAGS_64 += -m64
KBUILD_CFLAGS_64 := $(filter-out -m64,$(KBUILD_CFLAGS))
KBUILD_CFLAGS_64 := $(filter-out -mno-pic-data-is-text-relative,$(KBUILD_CFLAGS_64))
KBUILD_CFLAGS_64 += -m64 -fPIC -fno-common -fno-builtin
-ldflags-y := -fPIC -shared -soname=linux-vdso64.so.1 \
+ldflags-y := -shared -soname=linux-vdso64.so.1 \
--hash-style=both --build-id=sha1 -T
$(targets:%=$(obj)/%.dbg): KBUILD_CFLAGS = $(KBUILD_CFLAGS_64)I am unsure about the second issue, which is that the This was changed by commit e21f8baf8d9a (" |
|
I have sent the patch to remove https://lore.kernel.org/20240130-s390-vdso-drop-fpic-from-ldflags-v1-1-094ad104fc55@kernel.org/ I have worked around the other reported issue above with Unfortunately, I immediately hit an issue that appears to be related to Working around that by turning off the configurations that add Perhaps it is possible that the kernel should be compiled with |
@nathanchance I don't think so. I'm afraid that lld is actually pointing at a real error here, but I'd love if somebody with more experience would review this.
With the information that @nathanchance provided, I decided to test the build with gcc-14 + lld-17. Then, after applying the same suggestions from @nathanchance , I hit the following errors: It looks like that I decided to try to confirm if vmlinux would link after removing the following line from The build progressed, but it hits another issue: |
Version 1.6.1 is now out: https://github.com/IBM/s390x-abi/releases/tag/v1.6.1 - thanks @aarnez ! |
Well, this just emit a ".quad symbol", which should certainly be supported with I suspect the reason lld doesn't do that is because the section may not be writable, and so it thinks it cannot put dynamic relocations there. Does |
@uweigand You're right. It does help! Thanks! This is what I did: After applying this change, I hit the same issue reported in my previous comment: |
Hmm, this looks like a real problem. Looking at the linker script (boot/vmlinux.lds.S): the Not sure why lld doesn't appear to respect the |
This appears to be related to differences in the orphan section placement between If I add diff --git a/arch/s390/boot/Makefile b/arch/s390/boot/Makefile
index c7c81e5f9218..f99407b5430c 100644
--- a/arch/s390/boot/Makefile
+++ b/arch/s390/boot/Makefile
@@ -73,11 +73,11 @@ $(obj)/bzImage: $(obj)/vmlinux $(obj)/section_cmp.boot.data $(obj)/section_cmp.b
$(obj)/section_cmp%: vmlinux $(obj)/vmlinux FORCE
$(call if_changed,section_cmp)
-LDFLAGS_vmlinux := --oformat $(LD_BFD) -e startup $(if $(CONFIG_VMLINUX_MAP),-Map=$(obj)/vmlinux.map) --build-id=sha1 -T
+LDFLAGS_vmlinux := --orphan-handling=warn --oformat $(LD_BFD) -e startup $(if $(CONFIG_VMLINUX_MAP),-Map=$(obj)/vmlinux.map) --build-id=sha1 -T
$(obj)/vmlinux: $(obj)/vmlinux.lds $(OBJECTS_ALL) FORCE
$(call if_changed,ld)
-LDFLAGS_vmlinux.syms := --oformat $(LD_BFD) -e startup -T
+LDFLAGS_vmlinux.syms := --orphan-handling=warn --oformat $(LD_BFD) -e startup -T
$(obj)/vmlinux.syms: $(obj)/vmlinux.lds $(OBJECTS) FORCE
$(call if_changed,ld)
It reveals: So handling those sections in the linker script (this may be entirely incorrect for where to place things): diff --git a/arch/s390/boot/vmlinux.lds.S b/arch/s390/boot/vmlinux.lds.S
index 389df0e0d9e5..303928263f46 100644
--- a/arch/s390/boot/vmlinux.lds.S
+++ b/arch/s390/boot/vmlinux.lds.S
@@ -31,6 +31,7 @@ SECTIONS
_text = .; /* Text */
*(.text)
*(.text.*)
+ INIT_TEXT
_etext = . ;
}
.rodata : {
@@ -74,6 +75,10 @@ SECTIONS
_ebss = .;
}
+ .got : {
+ *(.got)
+ }
+
/*
* uncompressed image info used by the decompressor it should match
* struct vmlinux_info. It comes from .vmlinux.info section of
@@ -118,8 +123,12 @@ SECTIONS
}
_end = .;
+ DWARF_DEBUG
+ ELF_DETAILS
+
/* Sections to be discarded */
/DISCARD/ : {
+ COMMON_DISCARDS
*(.eh_frame)
*(__ex_table)
*(*__ksymtab*)
and adding diff --git a/arch/s390/Makefile b/arch/s390/Makefile
index 73873e451686..994f9b3d575f 100644
--- a/arch/s390/Makefile
+++ b/arch/s390/Makefile
@@ -15,7 +15,7 @@ KBUILD_CFLAGS_MODULE += -fPIC
KBUILD_AFLAGS += -m64
KBUILD_CFLAGS += -m64
KBUILD_CFLAGS += -fPIE
-LDFLAGS_vmlinux := -pie
+LDFLAGS_vmlinux := -pie -z notext
aflags_dwarf := -Wa,-gdwarf-2
KBUILD_AFLAGS_DECOMPRESSOR := $(CLANG_FLAGS) -m64 -D__ASSEMBLY__
ifndef CONFIG_AS_IS_LLVMand disabling All and all, not too bad! I think most of those fixes can be sent upstream, I'll see if I can talk to the s390 kernel folks about the bug tables issue, maybe the kernel can workaround that as well... |
Avenger-285714
pushed a commit
to Avenger-285714/WLIB-Kernel
that referenced
this pull request
Mar 27, 2024
[ Upstream commit 0628c03 ] '-fPIC' as an option to the linker does not do what it seems like it should. With ld.bfd, it is treated as '-f PIC', which does not make sense based on the meaning of '-f': -f SHLIB, --auxiliary SHLIB Auxiliary filter for shared object symbol table When building with ld.lld (currently under review in a GitHub pull request), it just errors out because '-f' means nothing and neither does '-fPIC': ld.lld: error: unknown argument '-fPIC' '-fPIC' was blindly copied from CFLAGS when the vDSO stopped being linked with '$(CC)', it should not be needed. Remove it to clear up the build failure with ld.lld. Fixes: 2b2a258 ("s390/vdso: Use $(LD) instead of $(CC) to link vDSO") Link: llvm/llvm-project#75643 Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Fangrui Song <maskray@google.com> Link: https://lore.kernel.org/r/20240130-s390-vdso-drop-fpic-from-ldflags-v1-1-094ad104fc55@kernel.org Signed-off-by: Heiko Carstens <hca@linux.ibm.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
jpuhlman
pushed a commit
to MontaVista-OpenSourceTechnology/linux-mvista
that referenced
this pull request
Apr 7, 2024
Source: Kernel.org MR: 132997 Type: Integration Disposition: Backport from git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable linux-5.4.y ChangeID: 4ae191effbc1427b1400b2ff3cbc23f82a0e5fad Description: commit e3a9ee9 upstream. Commit 90ceddc ("bpf: Support llvm-objcopy for vmlinux BTF") changed the ELF type of .btf.vmlinux.bin.o to ET_REL via dd, which works fine for little endian platforms: 00000000 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 03 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................| +00000010 01 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................| However, for big endian platforms, it changes the wrong byte, resulting in an invalid ELF file type, which ld.lld rejects: 00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| +00000010 01 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| Type: <unknown>: 103 ld.lld: error: .btf.vmlinux.bin.o: unknown file type Fix this by updating the entire 16-bit e_type field rather than just a single byte, so that everything works correctly for all platforms and linkers. 00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| +00000010 00 01 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| Type: REL (Relocatable file) While in the area, update the comment to mention that binutils 2.35+ matches LLD's behavior of rejecting an ET_EXEC input, which occurred after the comment was added. Cc: stable@vger.kernel.org Fixes: 90ceddc ("bpf: Support llvm-objcopy for vmlinux BTF") Link: llvm/llvm-project#75643 Suggested-by: Masahiro Yamada <masahiroy@kernel.org> Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Fangrui Song <maskray@google.com> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Reviewed-by: Kees Cook <keescook@chromium.org> Reviewed-by: Justin Stitt <justinstitt@google.com> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> [nathan: Fix silent conflict due to lack of 7d15369 in older trees] Signed-off-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Armin Kuster <akuster@mvista.com>
jpuhlman
pushed a commit
to MontaVista-OpenSourceTechnology/linux-mvista
that referenced
this pull request
Apr 7, 2024
Source: Kernel.org MR: 132369 Type: Integration Disposition: Backport from git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable linux-5.10.y ChangeID: 5abf3e8af2e34dd8b21f6a35f9a7672eea0e3abc Description: commit e3a9ee9 upstream. Commit 90ceddc ("bpf: Support llvm-objcopy for vmlinux BTF") changed the ELF type of .btf.vmlinux.bin.o to ET_REL via dd, which works fine for little endian platforms: 00000000 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 03 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................| +00000010 01 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................| However, for big endian platforms, it changes the wrong byte, resulting in an invalid ELF file type, which ld.lld rejects: 00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| +00000010 01 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| Type: <unknown>: 103 ld.lld: error: .btf.vmlinux.bin.o: unknown file type Fix this by updating the entire 16-bit e_type field rather than just a single byte, so that everything works correctly for all platforms and linkers. 00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| +00000010 00 01 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| Type: REL (Relocatable file) While in the area, update the comment to mention that binutils 2.35+ matches LLD's behavior of rejecting an ET_EXEC input, which occurred after the comment was added. Cc: stable@vger.kernel.org Fixes: 90ceddc ("bpf: Support llvm-objcopy for vmlinux BTF") Link: llvm/llvm-project#75643 Suggested-by: Masahiro Yamada <masahiroy@kernel.org> Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Fangrui Song <maskray@google.com> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Reviewed-by: Kees Cook <keescook@chromium.org> Reviewed-by: Justin Stitt <justinstitt@google.com> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> [nathan: Fix silent conflict due to lack of 7d15369 in older trees] Signed-off-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Armin Kuster <akuster@mvista.com>
jpuhlman
pushed a commit
to MontaVista-OpenSourceTechnology/linux-mvista
that referenced
this pull request
Apr 7, 2024
Source: Kernel.org MR: 132369 Type: Integration Disposition: Backport from git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable linux-5.10.y ChangeID: 5abf3e8af2e34dd8b21f6a35f9a7672eea0e3abc Description: commit e3a9ee9 upstream. Commit 90ceddc ("bpf: Support llvm-objcopy for vmlinux BTF") changed the ELF type of .btf.vmlinux.bin.o to ET_REL via dd, which works fine for little endian platforms: 00000000 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 03 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................| +00000010 01 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................| However, for big endian platforms, it changes the wrong byte, resulting in an invalid ELF file type, which ld.lld rejects: 00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| +00000010 01 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| Type: <unknown>: 103 ld.lld: error: .btf.vmlinux.bin.o: unknown file type Fix this by updating the entire 16-bit e_type field rather than just a single byte, so that everything works correctly for all platforms and linkers. 00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| +00000010 00 01 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| Type: REL (Relocatable file) While in the area, update the comment to mention that binutils 2.35+ matches LLD's behavior of rejecting an ET_EXEC input, which occurred after the comment was added. Cc: stable@vger.kernel.org Fixes: 90ceddc ("bpf: Support llvm-objcopy for vmlinux BTF") Link: llvm/llvm-project#75643 Suggested-by: Masahiro Yamada <masahiroy@kernel.org> Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Fangrui Song <maskray@google.com> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Reviewed-by: Kees Cook <keescook@chromium.org> Reviewed-by: Justin Stitt <justinstitt@google.com> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> [nathan: Fix silent conflict due to lack of 7d15369 in older trees] Signed-off-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Armin Kuster <akuster@mvista.com>
jpuhlman
pushed a commit
to MontaVista-OpenSourceTechnology/linux-mvista
that referenced
this pull request
Apr 7, 2024
Source: Kernel.org MR: 132369 Type: Integration Disposition: Backport from git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable linux-5.10.y ChangeID: 5abf3e8af2e34dd8b21f6a35f9a7672eea0e3abc Description: commit e3a9ee9 upstream. Commit 90ceddc ("bpf: Support llvm-objcopy for vmlinux BTF") changed the ELF type of .btf.vmlinux.bin.o to ET_REL via dd, which works fine for little endian platforms: 00000000 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 03 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................| +00000010 01 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................| However, for big endian platforms, it changes the wrong byte, resulting in an invalid ELF file type, which ld.lld rejects: 00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| +00000010 01 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| Type: <unknown>: 103 ld.lld: error: .btf.vmlinux.bin.o: unknown file type Fix this by updating the entire 16-bit e_type field rather than just a single byte, so that everything works correctly for all platforms and linkers. 00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| +00000010 00 01 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| Type: REL (Relocatable file) While in the area, update the comment to mention that binutils 2.35+ matches LLD's behavior of rejecting an ET_EXEC input, which occurred after the comment was added. Cc: stable@vger.kernel.org Fixes: 90ceddc ("bpf: Support llvm-objcopy for vmlinux BTF") Link: llvm/llvm-project#75643 Suggested-by: Masahiro Yamada <masahiroy@kernel.org> Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Fangrui Song <maskray@google.com> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Reviewed-by: Kees Cook <keescook@chromium.org> Reviewed-by: Justin Stitt <justinstitt@google.com> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> [nathan: Fix silent conflict due to lack of 7d15369 in older trees] Signed-off-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Armin Kuster <akuster@mvista.com>
phhgsi
pushed a commit
to phhgsi/android_kernel_oneplus_sm6375
that referenced
this pull request
Apr 8, 2024
commit e3a9ee963ad8ba677ca925149812c5932b49af69 upstream.
Commit 90ceddcb4950 ("bpf: Support llvm-objcopy for vmlinux BTF")
changed the ELF type of .btf.vmlinux.bin.o to ET_REL via dd, which works
fine for little endian platforms:
00000000 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 03 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................|
+00000010 01 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................|
However, for big endian platforms, it changes the wrong byte, resulting
in an invalid ELF file type, which ld.lld rejects:
00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
+00000010 01 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
Type: <unknown>: 103
ld.lld: error: .btf.vmlinux.bin.o: unknown file type
Fix this by updating the entire 16-bit e_type field rather than just a
single byte, so that everything works correctly for all platforms and
linkers.
00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
+00000010 00 01 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
Type: REL (Relocatable file)
While in the area, update the comment to mention that binutils 2.35+
matches LLD's behavior of rejecting an ET_EXEC input, which occurred
after the comment was added.
Cc: stable@vger.kernel.org
Fixes: 90ceddcb4950 ("bpf: Support llvm-objcopy for vmlinux BTF")
Link: llvm/llvm-project#75643
Suggested-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Reviewed-by: Fangrui Song <maskray@google.com>
Reviewed-by: Nicolas Schier <nicolas@fjasle.eu>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Justin Stitt <justinstitt@google.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
[nathan: Fix silent conflict due to lack of 7d153696e5db in older trees]
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
oraclelinuxkernel
pushed a commit
to oracle/linux-uek
that referenced
this pull request
Apr 19, 2024
[ Upstream commit 0628c03 ] '-fPIC' as an option to the linker does not do what it seems like it should. With ld.bfd, it is treated as '-f PIC', which does not make sense based on the meaning of '-f': -f SHLIB, --auxiliary SHLIB Auxiliary filter for shared object symbol table When building with ld.lld (currently under review in a GitHub pull request), it just errors out because '-f' means nothing and neither does '-fPIC': ld.lld: error: unknown argument '-fPIC' '-fPIC' was blindly copied from CFLAGS when the vDSO stopped being linked with '$(CC)', it should not be needed. Remove it to clear up the build failure with ld.lld. Fixes: 2b2a258 ("s390/vdso: Use $(LD) instead of $(CC) to link vDSO") Link: llvm/llvm-project#75643 Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Fangrui Song <maskray@google.com> Link: https://lore.kernel.org/r/20240130-s390-vdso-drop-fpic-from-ldflags-v1-1-094ad104fc55@kernel.org Signed-off-by: Heiko Carstens <hca@linux.ibm.com> Signed-off-by: Sasha Levin <sashal@kernel.org> (cherry picked from commit a6a50788b46b1dff9e848de9221270d4a80308a9) Signed-off-by: Vijayendra Suman <vijayendra.suman@oracle.com>
linkjumper
pushed a commit
to linkjumper/linux-fslc
that referenced
this pull request
Apr 20, 2024
commit e3a9ee9 upstream. Commit 90ceddc ("bpf: Support llvm-objcopy for vmlinux BTF") changed the ELF type of .btf.vmlinux.bin.o to ET_REL via dd, which works fine for little endian platforms: 00000000 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 03 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................| +00000010 01 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................| However, for big endian platforms, it changes the wrong byte, resulting in an invalid ELF file type, which ld.lld rejects: 00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| +00000010 01 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| Type: <unknown>: 103 ld.lld: error: .btf.vmlinux.bin.o: unknown file type Fix this by updating the entire 16-bit e_type field rather than just a single byte, so that everything works correctly for all platforms and linkers. 00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| +00000010 00 01 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| Type: REL (Relocatable file) While in the area, update the comment to mention that binutils 2.35+ matches LLD's behavior of rejecting an ET_EXEC input, which occurred after the comment was added. Cc: stable@vger.kernel.org Fixes: 90ceddc ("bpf: Support llvm-objcopy for vmlinux BTF") Link: llvm/llvm-project#75643 Suggested-by: Masahiro Yamada <masahiroy@kernel.org> Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Fangrui Song <maskray@google.com> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Reviewed-by: Kees Cook <keescook@chromium.org> Reviewed-by: Justin Stitt <justinstitt@google.com> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> [nathan: Fix silent conflict due to lack of 7d15369 in older trees] Signed-off-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
linkjumper
pushed a commit
to linkjumper/linux-fslc
that referenced
this pull request
Apr 20, 2024
[ Upstream commit 0628c03 ] '-fPIC' as an option to the linker does not do what it seems like it should. With ld.bfd, it is treated as '-f PIC', which does not make sense based on the meaning of '-f': -f SHLIB, --auxiliary SHLIB Auxiliary filter for shared object symbol table When building with ld.lld (currently under review in a GitHub pull request), it just errors out because '-f' means nothing and neither does '-fPIC': ld.lld: error: unknown argument '-fPIC' '-fPIC' was blindly copied from CFLAGS when the vDSO stopped being linked with '$(CC)', it should not be needed. Remove it to clear up the build failure with ld.lld. Fixes: 2b2a258 ("s390/vdso: Use $(LD) instead of $(CC) to link vDSO") Link: llvm/llvm-project#75643 Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Fangrui Song <maskray@google.com> Link: https://lore.kernel.org/r/20240130-s390-vdso-drop-fpic-from-ldflags-v1-1-094ad104fc55@kernel.org Signed-off-by: Heiko Carstens <hca@linux.ibm.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
linkjumper
pushed a commit
to linkjumper/linux-fslc
that referenced
this pull request
Apr 23, 2024
commit e3a9ee9 upstream. Commit 90ceddc ("bpf: Support llvm-objcopy for vmlinux BTF") changed the ELF type of .btf.vmlinux.bin.o to ET_REL via dd, which works fine for little endian platforms: 00000000 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 03 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................| +00000010 01 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................| However, for big endian platforms, it changes the wrong byte, resulting in an invalid ELF file type, which ld.lld rejects: 00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| +00000010 01 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| Type: <unknown>: 103 ld.lld: error: .btf.vmlinux.bin.o: unknown file type Fix this by updating the entire 16-bit e_type field rather than just a single byte, so that everything works correctly for all platforms and linkers. 00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| +00000010 00 01 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| Type: REL (Relocatable file) While in the area, update the comment to mention that binutils 2.35+ matches LLD's behavior of rejecting an ET_EXEC input, which occurred after the comment was added. Cc: stable@vger.kernel.org Fixes: 90ceddc ("bpf: Support llvm-objcopy for vmlinux BTF") Link: llvm/llvm-project#75643 Suggested-by: Masahiro Yamada <masahiroy@kernel.org> Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Fangrui Song <maskray@google.com> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Reviewed-by: Kees Cook <keescook@chromium.org> Reviewed-by: Justin Stitt <justinstitt@google.com> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> [nathan: Fix silent conflict due to lack of 7d15369 in older trees] Signed-off-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
linkjumper
pushed a commit
to linkjumper/linux-fslc
that referenced
this pull request
Apr 23, 2024
[ Upstream commit 0628c03 ] '-fPIC' as an option to the linker does not do what it seems like it should. With ld.bfd, it is treated as '-f PIC', which does not make sense based on the meaning of '-f': -f SHLIB, --auxiliary SHLIB Auxiliary filter for shared object symbol table When building with ld.lld (currently under review in a GitHub pull request), it just errors out because '-f' means nothing and neither does '-fPIC': ld.lld: error: unknown argument '-fPIC' '-fPIC' was blindly copied from CFLAGS when the vDSO stopped being linked with '$(CC)', it should not be needed. Remove it to clear up the build failure with ld.lld. Fixes: 2b2a258 ("s390/vdso: Use $(LD) instead of $(CC) to link vDSO") Link: llvm/llvm-project#75643 Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Fangrui Song <maskray@google.com> Link: https://lore.kernel.org/r/20240130-s390-vdso-drop-fpic-from-ldflags-v1-1-094ad104fc55@kernel.org Signed-off-by: Heiko Carstens <hca@linux.ibm.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
linkjumper
pushed a commit
to linkjumper/linux-fslc
that referenced
this pull request
Apr 23, 2024
[ Upstream commit 0628c03 ] '-fPIC' as an option to the linker does not do what it seems like it should. With ld.bfd, it is treated as '-f PIC', which does not make sense based on the meaning of '-f': -f SHLIB, --auxiliary SHLIB Auxiliary filter for shared object symbol table When building with ld.lld (currently under review in a GitHub pull request), it just errors out because '-f' means nothing and neither does '-fPIC': ld.lld: error: unknown argument '-fPIC' '-fPIC' was blindly copied from CFLAGS when the vDSO stopped being linked with '$(CC)', it should not be needed. Remove it to clear up the build failure with ld.lld. Fixes: 2b2a258 ("s390/vdso: Use $(LD) instead of $(CC) to link vDSO") Link: llvm/llvm-project#75643 Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Fangrui Song <maskray@google.com> Link: https://lore.kernel.org/r/20240130-s390-vdso-drop-fpic-from-ldflags-v1-1-094ad104fc55@kernel.org Signed-off-by: Heiko Carstens <hca@linux.ibm.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Closed
Hanamizaki
pushed a commit
to Hanamizaki/android_kernel_oneplus_sm8350
that referenced
this pull request
Jun 1, 2024
commit e3a9ee963ad8ba677ca925149812c5932b49af69 upstream.
Commit 90ceddcb4950 ("bpf: Support llvm-objcopy for vmlinux BTF")
changed the ELF type of .btf.vmlinux.bin.o to ET_REL via dd, which works
fine for little endian platforms:
00000000 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 03 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................|
+00000010 01 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................|
However, for big endian platforms, it changes the wrong byte, resulting
in an invalid ELF file type, which ld.lld rejects:
00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
+00000010 01 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
Type: <unknown>: 103
ld.lld: error: .btf.vmlinux.bin.o: unknown file type
Fix this by updating the entire 16-bit e_type field rather than just a
single byte, so that everything works correctly for all platforms and
linkers.
00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
+00000010 00 01 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
Type: REL (Relocatable file)
While in the area, update the comment to mention that binutils 2.35+
matches LLD's behavior of rejecting an ET_EXEC input, which occurred
after the comment was added.
Cc: stable@vger.kernel.org
Fixes: 90ceddcb4950 ("bpf: Support llvm-objcopy for vmlinux BTF")
Link: llvm/llvm-project#75643
Suggested-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Reviewed-by: Fangrui Song <maskray@google.com>
Reviewed-by: Nicolas Schier <nicolas@fjasle.eu>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Justin Stitt <justinstitt@google.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
[nathan: Fix silent conflict due to lack of 7d153696e5db in older trees]
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Hanamizaki
pushed a commit
to Hanamizaki/android_kernel_oneplus_sm8350
that referenced
this pull request
Jun 1, 2024
commit e3a9ee963ad8ba677ca925149812c5932b49af69 upstream.
Commit 90ceddcb4950 ("bpf: Support llvm-objcopy for vmlinux BTF")
changed the ELF type of .btf.vmlinux.bin.o to ET_REL via dd, which works
fine for little endian platforms:
00000000 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 03 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................|
+00000010 01 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................|
However, for big endian platforms, it changes the wrong byte, resulting
in an invalid ELF file type, which ld.lld rejects:
00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
+00000010 01 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
Type: <unknown>: 103
ld.lld: error: .btf.vmlinux.bin.o: unknown file type
Fix this by updating the entire 16-bit e_type field rather than just a
single byte, so that everything works correctly for all platforms and
linkers.
00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
+00000010 00 01 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
Type: REL (Relocatable file)
While in the area, update the comment to mention that binutils 2.35+
matches LLD's behavior of rejecting an ET_EXEC input, which occurred
after the comment was added.
Cc: stable@vger.kernel.org
Fixes: 90ceddcb4950 ("bpf: Support llvm-objcopy for vmlinux BTF")
Link: llvm/llvm-project#75643
Suggested-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Reviewed-by: Fangrui Song <maskray@google.com>
Reviewed-by: Nicolas Schier <nicolas@fjasle.eu>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Justin Stitt <justinstitt@google.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
[nathan: Fix silent conflict due to lack of 7d153696e5db in older trees]
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
EviraKernel
pushed a commit
to EviraKernel/android_kernel_google_zuma
that referenced
this pull request
Jun 6, 2024
commit e3a9ee963ad8ba677ca925149812c5932b49af69 upstream. Commit 90ceddc ("bpf: Support llvm-objcopy for vmlinux BTF") changed the ELF type of .btf.vmlinux.bin.o to ET_REL via dd, which works fine for little endian platforms: 00000000 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 03 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................| +00000010 01 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................| However, for big endian platforms, it changes the wrong byte, resulting in an invalid ELF file type, which ld.lld rejects: 00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| +00000010 01 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| Type: <unknown>: 103 ld.lld: error: .btf.vmlinux.bin.o: unknown file type Fix this by updating the entire 16-bit e_type field rather than just a single byte, so that everything works correctly for all platforms and linkers. 00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| +00000010 00 01 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| Type: REL (Relocatable file) While in the area, update the comment to mention that binutils 2.35+ matches LLD's behavior of rejecting an ET_EXEC input, which occurred after the comment was added. Cc: stable@vger.kernel.org Fixes: 90ceddc ("bpf: Support llvm-objcopy for vmlinux BTF") Link: llvm/llvm-project#75643 Suggested-by: Masahiro Yamada <masahiroy@kernel.org> Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Fangrui Song <maskray@google.com> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Reviewed-by: Kees Cook <keescook@chromium.org> Reviewed-by: Justin Stitt <justinstitt@google.com> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> [nathan: Fix silent conflict due to lack of 7d153696e5db in older trees] Signed-off-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Hanamizaki
pushed a commit
to Hanamizaki/android_kernel_oneplus_sm8350
that referenced
this pull request
Jun 6, 2024
commit e3a9ee963ad8ba677ca925149812c5932b49af69 upstream.
Commit 90ceddcb4950 ("bpf: Support llvm-objcopy for vmlinux BTF")
changed the ELF type of .btf.vmlinux.bin.o to ET_REL via dd, which works
fine for little endian platforms:
00000000 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 03 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................|
+00000010 01 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................|
However, for big endian platforms, it changes the wrong byte, resulting
in an invalid ELF file type, which ld.lld rejects:
00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
+00000010 01 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
Type: <unknown>: 103
ld.lld: error: .btf.vmlinux.bin.o: unknown file type
Fix this by updating the entire 16-bit e_type field rather than just a
single byte, so that everything works correctly for all platforms and
linkers.
00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
+00000010 00 01 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
Type: REL (Relocatable file)
While in the area, update the comment to mention that binutils 2.35+
matches LLD's behavior of rejecting an ET_EXEC input, which occurred
after the comment was added.
Cc: stable@vger.kernel.org
Fixes: 90ceddcb4950 ("bpf: Support llvm-objcopy for vmlinux BTF")
Link: llvm/llvm-project#75643
Suggested-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Reviewed-by: Fangrui Song <maskray@google.com>
Reviewed-by: Nicolas Schier <nicolas@fjasle.eu>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Justin Stitt <justinstitt@google.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
[nathan: Fix silent conflict due to lack of 7d153696e5db in older trees]
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
edward0181
pushed a commit
to edward0181/upstream_kernel_sm6375
that referenced
this pull request
Jun 18, 2024
commit e3a9ee963ad8ba677ca925149812c5932b49af69 upstream.
Commit 90ceddcb4950 ("bpf: Support llvm-objcopy for vmlinux BTF")
changed the ELF type of .btf.vmlinux.bin.o to ET_REL via dd, which works
fine for little endian platforms:
00000000 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 03 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................|
+00000010 01 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................|
However, for big endian platforms, it changes the wrong byte, resulting
in an invalid ELF file type, which ld.lld rejects:
00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
+00000010 01 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
Type: <unknown>: 103
ld.lld: error: .btf.vmlinux.bin.o: unknown file type
Fix this by updating the entire 16-bit e_type field rather than just a
single byte, so that everything works correctly for all platforms and
linkers.
00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
+00000010 00 01 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
Type: REL (Relocatable file)
While in the area, update the comment to mention that binutils 2.35+
matches LLD's behavior of rejecting an ET_EXEC input, which occurred
after the comment was added.
Cc: stable@vger.kernel.org
Fixes: 90ceddcb4950 ("bpf: Support llvm-objcopy for vmlinux BTF")
Link: llvm/llvm-project#75643
Suggested-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Reviewed-by: Fangrui Song <maskray@google.com>
Reviewed-by: Nicolas Schier <nicolas@fjasle.eu>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Justin Stitt <justinstitt@google.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
[nathan: Fix silent conflict due to lack of 7d153696e5db in older trees]
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
ScrubMcDub69
pushed a commit
to Matrixx-Devices/kernel_google_gs201
that referenced
this pull request
Jun 20, 2024
commit e3a9ee963ad8ba677ca925149812c5932b49af69 upstream. Commit 90ceddc ("bpf: Support llvm-objcopy for vmlinux BTF") changed the ELF type of .btf.vmlinux.bin.o to ET_REL via dd, which works fine for little endian platforms: 00000000 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 03 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................| +00000010 01 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................| However, for big endian platforms, it changes the wrong byte, resulting in an invalid ELF file type, which ld.lld rejects: 00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| +00000010 01 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| Type: <unknown>: 103 ld.lld: error: .btf.vmlinux.bin.o: unknown file type Fix this by updating the entire 16-bit e_type field rather than just a single byte, so that everything works correctly for all platforms and linkers. 00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| +00000010 00 01 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| Type: REL (Relocatable file) While in the area, update the comment to mention that binutils 2.35+ matches LLD's behavior of rejecting an ET_EXEC input, which occurred after the comment was added. Cc: stable@vger.kernel.org Fixes: 90ceddc ("bpf: Support llvm-objcopy for vmlinux BTF") Link: llvm/llvm-project#75643 Suggested-by: Masahiro Yamada <masahiroy@kernel.org> Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Fangrui Song <maskray@google.com> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Reviewed-by: Kees Cook <keescook@chromium.org> Reviewed-by: Justin Stitt <justinstitt@google.com> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> [nathan: Fix silent conflict due to lack of 7d153696e5db in older trees] Signed-off-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
ScrubMcDub69
pushed a commit
to Matrixx-Devices/kernel_google_gs201
that referenced
this pull request
Jun 20, 2024
commit e3a9ee963ad8ba677ca925149812c5932b49af69 upstream. Commit 90ceddc ("bpf: Support llvm-objcopy for vmlinux BTF") changed the ELF type of .btf.vmlinux.bin.o to ET_REL via dd, which works fine for little endian platforms: 00000000 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 03 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................| +00000010 01 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................| However, for big endian platforms, it changes the wrong byte, resulting in an invalid ELF file type, which ld.lld rejects: 00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| +00000010 01 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| Type: <unknown>: 103 ld.lld: error: .btf.vmlinux.bin.o: unknown file type Fix this by updating the entire 16-bit e_type field rather than just a single byte, so that everything works correctly for all platforms and linkers. 00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| +00000010 00 01 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| Type: REL (Relocatable file) While in the area, update the comment to mention that binutils 2.35+ matches LLD's behavior of rejecting an ET_EXEC input, which occurred after the comment was added. Cc: stable@vger.kernel.org Fixes: 90ceddc ("bpf: Support llvm-objcopy for vmlinux BTF") Link: llvm/llvm-project#75643 Suggested-by: Masahiro Yamada <masahiroy@kernel.org> Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Fangrui Song <maskray@google.com> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Reviewed-by: Kees Cook <keescook@chromium.org> Reviewed-by: Justin Stitt <justinstitt@google.com> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> [nathan: Fix silent conflict due to lack of 7d153696e5db in older trees] Signed-off-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
tuxedo-bot
pushed a commit
to tuxedocomputers/linux
that referenced
this pull request
Jun 24, 2024
BugLink: https://bugs.launchpad.net/bugs/2061814 commit e3a9ee9 upstream. Commit 90ceddc ("bpf: Support llvm-objcopy for vmlinux BTF") changed the ELF type of .btf.vmlinux.bin.o to ET_REL via dd, which works fine for little endian platforms: 00000000 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 03 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................| +00000010 01 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................| However, for big endian platforms, it changes the wrong byte, resulting in an invalid ELF file type, which ld.lld rejects: 00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| +00000010 01 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| Type: <unknown>: 103 ld.lld: error: .btf.vmlinux.bin.o: unknown file type Fix this by updating the entire 16-bit e_type field rather than just a single byte, so that everything works correctly for all platforms and linkers. 00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| +00000010 00 01 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| Type: REL (Relocatable file) While in the area, update the comment to mention that binutils 2.35+ matches LLD's behavior of rejecting an ET_EXEC input, which occurred after the comment was added. Cc: stable@vger.kernel.org Fixes: 90ceddc ("bpf: Support llvm-objcopy for vmlinux BTF") Link: llvm/llvm-project#75643 Suggested-by: Masahiro Yamada <masahiroy@kernel.org> Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Fangrui Song <maskray@google.com> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Reviewed-by: Kees Cook <keescook@chromium.org> Reviewed-by: Justin Stitt <justinstitt@google.com> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Portia Stephens <portia.stephens@canonical.com> Signed-off-by: Roxana Nicolescu <roxana.nicolescu@canonical.com>
forkposix
pushed a commit
to forkposix/amagas_kernel_lemonade
that referenced
this pull request
Jun 25, 2024
commit e3a9ee963ad8ba677ca925149812c5932b49af69 upstream.
Commit 90ceddcb4950 ("bpf: Support llvm-objcopy for vmlinux BTF")
changed the ELF type of .btf.vmlinux.bin.o to ET_REL via dd, which works
fine for little endian platforms:
00000000 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 03 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................|
+00000010 01 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................|
However, for big endian platforms, it changes the wrong byte, resulting
in an invalid ELF file type, which ld.lld rejects:
00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
+00000010 01 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
Type: <unknown>: 103
ld.lld: error: .btf.vmlinux.bin.o: unknown file type
Fix this by updating the entire 16-bit e_type field rather than just a
single byte, so that everything works correctly for all platforms and
linkers.
00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
+00000010 00 01 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
Type: REL (Relocatable file)
While in the area, update the comment to mention that binutils 2.35+
matches LLD's behavior of rejecting an ET_EXEC input, which occurred
after the comment was added.
Cc: stable@vger.kernel.org
Fixes: 90ceddcb4950 ("bpf: Support llvm-objcopy for vmlinux BTF")
Link: llvm/llvm-project#75643
Suggested-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Reviewed-by: Fangrui Song <maskray@google.com>
Reviewed-by: Nicolas Schier <nicolas@fjasle.eu>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Justin Stitt <justinstitt@google.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
[nathan: Fix silent conflict due to lack of 7d153696e5db in older trees]
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
1xtAsh
pushed a commit
to 1xtAsh/kernel-xiaomi-sm8350
that referenced
this pull request
Jul 26, 2024
commit e3a9ee963ad8ba677ca925149812c5932b49af69 upstream.
Commit 90ceddcb4950 ("bpf: Support llvm-objcopy for vmlinux BTF")
changed the ELF type of .btf.vmlinux.bin.o to ET_REL via dd, which works
fine for little endian platforms:
00000000 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 03 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................|
+00000010 01 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................|
However, for big endian platforms, it changes the wrong byte, resulting
in an invalid ELF file type, which ld.lld rejects:
00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
+00000010 01 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
Type: <unknown>: 103
ld.lld: error: .btf.vmlinux.bin.o: unknown file type
Fix this by updating the entire 16-bit e_type field rather than just a
single byte, so that everything works correctly for all platforms and
linkers.
00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
+00000010 00 01 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
Type: REL (Relocatable file)
While in the area, update the comment to mention that binutils 2.35+
matches LLD's behavior of rejecting an ET_EXEC input, which occurred
after the comment was added.
Cc: stable@vger.kernel.org
Fixes: 90ceddcb4950 ("bpf: Support llvm-objcopy for vmlinux BTF")
Link: llvm/llvm-project#75643
Suggested-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Reviewed-by: Fangrui Song <maskray@google.com>
Reviewed-by: Nicolas Schier <nicolas@fjasle.eu>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Justin Stitt <justinstitt@google.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
[nathan: Fix silent conflict due to lack of 7d153696e5db in older trees]
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
xt0032rus
pushed a commit
to xt0032rus/kernel_xiaomi_sm8550
that referenced
this pull request
Aug 30, 2024
[ Upstream commit 0628c03934187be33942580e10bb9afcc61adeed ] '-fPIC' as an option to the linker does not do what it seems like it should. With ld.bfd, it is treated as '-f PIC', which does not make sense based on the meaning of '-f': -f SHLIB, --auxiliary SHLIB Auxiliary filter for shared object symbol table When building with ld.lld (currently under review in a GitHub pull request), it just errors out because '-f' means nothing and neither does '-fPIC': ld.lld: error: unknown argument '-fPIC' '-fPIC' was blindly copied from CFLAGS when the vDSO stopped being linked with '$(CC)', it should not be needed. Remove it to clear up the build failure with ld.lld. Fixes: 2b2a258 ("s390/vdso: Use $(LD) instead of $(CC) to link vDSO") Link: llvm/llvm-project#75643 Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Fangrui Song <maskray@google.com> Link: https://lore.kernel.org/r/20240130-s390-vdso-drop-fpic-from-ldflags-v1-1-094ad104fc55@kernel.org Signed-off-by: Heiko Carstens <hca@linux.ibm.com> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: xt0032rus <andrey.denikin9@mail.com>
fluffball3
pushed a commit
to fluffball3/android_kernel_samsung_m33x
that referenced
this pull request
Oct 2, 2024
commit e3a9ee963ad8ba677ca925149812c5932b49af69 upstream.
Commit 90ceddcb4950 ("bpf: Support llvm-objcopy for vmlinux BTF")
changed the ELF type of .btf.vmlinux.bin.o to ET_REL via dd, which works
fine for little endian platforms:
00000000 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 03 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................|
+00000010 01 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................|
However, for big endian platforms, it changes the wrong byte, resulting
in an invalid ELF file type, which ld.lld rejects:
00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
+00000010 01 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
Type: <unknown>: 103
ld.lld: error: .btf.vmlinux.bin.o: unknown file type
Fix this by updating the entire 16-bit e_type field rather than just a
single byte, so that everything works correctly for all platforms and
linkers.
00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
+00000010 00 01 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
Type: REL (Relocatable file)
While in the area, update the comment to mention that binutils 2.35+
matches LLD's behavior of rejecting an ET_EXEC input, which occurred
after the comment was added.
Cc: stable@vger.kernel.org
Fixes: 90ceddcb4950 ("bpf: Support llvm-objcopy for vmlinux BTF")
Link: llvm/llvm-project#75643
Suggested-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Reviewed-by: Fangrui Song <maskray@google.com>
Reviewed-by: Nicolas Schier <nicolas@fjasle.eu>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Justin Stitt <justinstitt@google.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
[nathan: Fix silent conflict due to lack of 7d153696e5db in older trees]
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
OpenHarmonySCM-noreply
pushed a commit
to openharmony/kernel_linux_5.10
that referenced
this pull request
Oct 29, 2024
stable inclusion from stable-5.10.210 commit 5abf3e8af2e34dd8b21f6a35f9a7672eea0e3abc category: bugfix issue: #IAOD4C CVE: NA Signed-off-by: yaowenrui <yaowenrui2@huawei.com> --------------------------------------- commit e3a9ee963ad8ba677ca925149812c5932b49af69 upstream. Commit 90ceddc ("bpf: Support llvm-objcopy for vmlinux BTF") changed the ELF type of .btf.vmlinux.bin.o to ET_REL via dd, which works fine for little endian platforms: 00000000 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 03 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................| +00000010 01 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................| However, for big endian platforms, it changes the wrong byte, resulting in an invalid ELF file type, which ld.lld rejects: 00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| +00000010 01 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| Type: <unknown>: 103 ld.lld: error: .btf.vmlinux.bin.o: unknown file type Fix this by updating the entire 16-bit e_type field rather than just a single byte, so that everything works correctly for all platforms and linkers. 00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............| -00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| +00000010 00 01 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................| Type: REL (Relocatable file) While in the area, update the comment to mention that binutils 2.35+ matches LLD's behavior of rejecting an ET_EXEC input, which occurred after the comment was added. Cc: stable@vger.kernel.org Fixes: 90ceddc ("bpf: Support llvm-objcopy for vmlinux BTF") Link: llvm/llvm-project#75643 Suggested-by: Masahiro Yamada <masahiroy@kernel.org> Signed-off-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Fangrui Song <maskray@google.com> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Reviewed-by: Kees Cook <keescook@chromium.org> Reviewed-by: Justin Stitt <justinstitt@google.com> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> [nathan: Fix silent conflict due to lack of 7d153696e5db in older trees] Signed-off-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: yaowenrui <yaowenrui2@huawei.com>
Ksawlii
pushed a commit
to Ksawlii/android_kernel_samsung_a53x-FireAsf
that referenced
this pull request
Nov 18, 2024
commit e3a9ee963ad8ba677ca925149812c5932b49af69 upstream.
Commit 90ceddcb4950 ("bpf: Support llvm-objcopy for vmlinux BTF")
changed the ELF type of .btf.vmlinux.bin.o to ET_REL via dd, which works
fine for little endian platforms:
00000000 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 03 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................|
+00000010 01 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................|
However, for big endian platforms, it changes the wrong byte, resulting
in an invalid ELF file type, which ld.lld rejects:
00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
+00000010 01 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
Type: <unknown>: 103
ld.lld: error: .btf.vmlinux.bin.o: unknown file type
Fix this by updating the entire 16-bit e_type field rather than just a
single byte, so that everything works correctly for all platforms and
linkers.
00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
+00000010 00 01 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
Type: REL (Relocatable file)
While in the area, update the comment to mention that binutils 2.35+
matches LLD's behavior of rejecting an ET_EXEC input, which occurred
after the comment was added.
Cc: stable@vger.kernel.org
Fixes: 90ceddcb4950 ("bpf: Support llvm-objcopy for vmlinux BTF")
Link: llvm/llvm-project#75643
Suggested-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Reviewed-by: Fangrui Song <maskray@google.com>
Reviewed-by: Nicolas Schier <nicolas@fjasle.eu>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Justin Stitt <justinstitt@google.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
[nathan: Fix silent conflict due to lack of 7d153696e5db in older trees]
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fus0g
pushed a commit
to sm8350-martini/kernel_oneplus_sm8350
that referenced
this pull request
Nov 29, 2024
commit e3a9ee963ad8ba677ca925149812c5932b49af69 upstream.
Commit 90ceddcb4950 ("bpf: Support llvm-objcopy for vmlinux BTF")
changed the ELF type of .btf.vmlinux.bin.o to ET_REL via dd, which works
fine for little endian platforms:
00000000 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 03 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................|
+00000010 01 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................|
However, for big endian platforms, it changes the wrong byte, resulting
in an invalid ELF file type, which ld.lld rejects:
00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
+00000010 01 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
Type: <unknown>: 103
ld.lld: error: .btf.vmlinux.bin.o: unknown file type
Fix this by updating the entire 16-bit e_type field rather than just a
single byte, so that everything works correctly for all platforms and
linkers.
00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
+00000010 00 01 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
Type: REL (Relocatable file)
While in the area, update the comment to mention that binutils 2.35+
matches LLD's behavior of rejecting an ET_EXEC input, which occurred
after the comment was added.
Cc: stable@vger.kernel.org
Fixes: 90ceddcb4950 ("bpf: Support llvm-objcopy for vmlinux BTF")
Link: llvm/llvm-project#75643
Suggested-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Reviewed-by: Fangrui Song <maskray@google.com>
Reviewed-by: Nicolas Schier <nicolas@fjasle.eu>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Justin Stitt <justinstitt@google.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
[nathan: Fix silent conflict due to lack of 7d153696e5db in older trees]
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
guckesh
pushed a commit
to exynos2100-dev/android_kernel_samsung_universal2100
that referenced
this pull request
Nov 30, 2024
commit e3a9ee963ad8ba677ca925149812c5932b49af69 upstream.
Commit 90ceddcb4950 ("bpf: Support llvm-objcopy for vmlinux BTF")
changed the ELF type of .btf.vmlinux.bin.o to ET_REL via dd, which works
fine for little endian platforms:
00000000 7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 03 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................|
+00000010 01 00 b7 00 01 00 00 00 00 00 00 80 00 80 ff ff |................|
However, for big endian platforms, it changes the wrong byte, resulting
in an invalid ELF file type, which ld.lld rejects:
00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
+00000010 01 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
Type: <unknown>: 103
ld.lld: error: .btf.vmlinux.bin.o: unknown file type
Fix this by updating the entire 16-bit e_type field rather than just a
single byte, so that everything works correctly for all platforms and
linkers.
00000000 7f 45 4c 46 02 02 01 00 00 00 00 00 00 00 00 00 |.ELF............|
-00000010 00 03 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
+00000010 00 01 00 16 00 00 00 01 00 00 00 00 00 10 00 00 |................|
Type: REL (Relocatable file)
While in the area, update the comment to mention that binutils 2.35+
matches LLD's behavior of rejecting an ET_EXEC input, which occurred
after the comment was added.
Cc: stable@vger.kernel.org
Fixes: 90ceddcb4950 ("bpf: Support llvm-objcopy for vmlinux BTF")
Link: llvm/llvm-project#75643
Suggested-by: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Reviewed-by: Fangrui Song <maskray@google.com>
Reviewed-by: Nicolas Schier <nicolas@fjasle.eu>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Justin Stitt <justinstitt@google.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
[nathan: Fix silent conflict due to lack of 7d153696e5db in older trees]
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This patch adds full support for linking SystemZ (ELF s390x) object files. Support should be generally complete:
In addition to new platform code and the obvious changes, there were a few additional changes to common code:
Add three new RelExpr members (R_GOTPLT_OFF, R_GOTPLT_PC, and R_PLT_GOTREL) needed to support certain s390x relocations. I chose not to use a platform-specific name since nothing in the definition of these relocs is actually platform-specific; it is well possible that other platforms will need the same.
A couple of tweaks to TLS relocation handling, as the particular semantics of the s390x versions differ slightly. See comments in the code.
This was tested by building and testing >1500 Fedora packages, with only a handful of failures; as these also have issues when building with LLD on other architectures, they seem unrelated.