gcc -B/home/dave/gnu/binutils/objdir/ld/tmpdir/ld/ -I/home/dave/gnu/binutils/src/ld/testsuite/ld-elf -g -O2 -c -g -O2 -fno-PIE -c /home/dave/gnu/binutils/src/ld/testsuite/ld-elf/pr22393-2b.c -o tmpdir/pr22393-2b.o Executing on host: sh -c {gcc -B/home/dave/gnu/binutils/objdir/ld/tmpdir/ld/ -I/home/dave/gnu/binutils/src/ld/testsuite/ld-elf -g -O2 -c -g -O2 -fno-PIE -c /home/dave/gnu/binutils/src/ld/testsuite/ld-elf/pr22393-2b.c -o tmpdir/pr22393-2b.o 2>&1} /dev/null ld.tmp (timeout = 300) spawn [open ...] gcc -B/home/dave/gnu/binutils/objdir/ld/tmpdir/ld/ -L=/home/dave/opt/gnu/hppa-unknown-linux-gnu/lib -L=/home/dave/opt/gnu/lib -L=/usr/local/lib -L=/lib -L=/usr/lib -o tmpdir/pr22393-2 -L/home/dave/gnu/binutils/src/ld/testsuite/ld-elf -no-pie -Wl,-z,separate-code,--no-as-needed tmpdir/pr22393-2a.so tmpdir/pr22393-2b.o Executing on host: sh -c {gcc -B/home/dave/gnu/binutils/objdir/ld/tmpdir/ld/ -L=/home/dave/opt/gnu/hppa-unknown-linux-gnu/lib -L=/home/dave/opt/gnu/lib -L=/usr/local/lib -L=/lib -L=/usr/lib -o tmpdir/pr22393-2 -L/home/dave/gnu/binutils/src/ld/testsuite/ld-elf -no-pie -Wl,-z,separate-code,--no-as-needed tmpdir/pr22393-2a.so tmpdir/pr22393-2b.o 2>&1} /dev/null ld.tmp (timeout = 300) spawn [open ...] Running: tmpdir/pr22393-2 > tmpdir/pr22393-2.out child killed: segmentation violation FAIL: Run pr22393-2 Similar fail: FAIL: Run pr22393-2 (PIE)
Hi John, Please could you narrow this down to the linker command line and a tarball of the object files & libraries involved ? (I do not have easy access to an hppa based build system). Cheers Nick
Hi Nick, On 2018-01-29 9:18 AM, nickc at redhat dot com wrote: > Please could you narrow this down to the linker command line and a tarball > of the object files & libraries involved ? (I do not have easy access to > an hppa based build system). I'll try to look at this as soon as I can. Alan has an account on a hppa-linux system. Dave
Let me add that we have the same couple of failures on SPARC64/Linux.
Here are pr22393-2 program headers and section to header mapping: PHDR 0x000034 0x00010034 0x00010034 0x00140 0x00140 R 0x4 INTERP 0x000174 0x00010174 0x00010174 0x0000d 0x0000d R 0x1 [Requesting program interpreter: /lib/ld.so.1] LOAD 0x000000 0x00010000 0x00010000 0x0043c 0x0043c R 0x1000 LOAD 0x001000 0x00011000 0x00011000 0x00554 0x00554 R E 0x1000 LOAD 0x002000 0x00012000 0x00012000 0x00118 0x00118 R 0x1000 LOAD 0x003000 0x00013000 0x00013000 0x000f8 0x000f8 RW 0x1000 LOAD 0x0040f8 0x000130f8 0x000130f8 0x00064 0x00064 RWE 0x1000 LOAD 0x00515c 0x0001315c 0x0001315c 0x0001c 0x0002c RW 0x1000 DYNAMIC 0x003020 0x00013020 0x00013020 0x000d0 0x000d0 RW 0x4 NOTE 0x000184 0x00010184 0x00010184 0x00044 0x00044 R 0x4 00 01 .interp 02 .interp .note.ABI-tag .note.gnu.build-id .hash .gnu.hash .dynsym .dynstr .gnu.version .gnu.version_r .rela.plt 03 .init .text .fini 04 .rodata .PARISC.unwind .eh_frame 05 .init_array .ctors .dtors .data.rel.ro .dynamic .data 06 .plt 07 .got .bss 08 .dynamic 09 .note.ABI-tag .note.gnu.build-id There's a problem with the last 3 PT_LOAD headers. They can't be mapped from different disk pages (0x3000, 0x4000 and 0x5000) into the same memory page (0x13000).
The master branch has been updated by Alan Modra <amodra@sourceware.org>: https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;h=76cb3a89a6615cf3418fa1efe8268bf6673a5c8a commit 76cb3a89a6615cf3418fa1efe8268bf6673a5c8a Author: Alan Modra <amodra@gmail.com> Date: Tue Jan 30 16:02:32 2018 +1030 PR22758, FAIL: Run pr22393-2 We can't map different disk pages into the same memory page; The last page mapped will simply overwrite any previous pages. The executable/non-executable new_segment test ignored this fact, leading to a ld.so segfault on hppa when .dynamic is overwritten with zeros. This patch moves existing tests for demand paging with lma on the same memory page, to a new test performed before any case where we want a new segment due to protection or loadable conflicts. PR 22758 * elf.c (_bfd_elf_map_sections_to_segments): Don't start a new segment when demand paged with lma on the same page. Test this before load/non-load, executable/non-executable, writable/non-writable tests and simplify. Delete bogus relro condition in writable/non-writable test. Delete outdated comment. Formatting.
Fixed master.
Hi, Long story short, I was trying to build clang in PGO+LTO bootstrap mode the last few days, and the bootstrap fails with a segfault at the moment clang is ran on the profiling data set. I hunted down the issue to the fact that in some LLVM libraries, "atexit" function is inlined and included in the loaded shared libraries, but placed in a section that ld.so mapped in a RW segment rather than a RE segment, and both segment are not correctly aligned to the boundary of a 2MB page like it should. When applying this patch, the bootstrap of clang works. Given that this might potentially happen arbitrarily to any user of binutils 2.30 and it is not so easy to understand that section placement is wrong here, would you please backport this to the 2.30 branch ? Thanks, Romain
The binutils-2_30-branch branch has been updated by Nick Clifton <nickc@sourceware.org>: https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;h=355d8ed79b1e830d5216a70026f751f5a38df9e9 commit 355d8ed79b1e830d5216a70026f751f5a38df9e9 Author: Nick Clifton <nickc@redhat.com> Date: Thu Feb 15 16:21:02 2018 +0000 Import patch from mainline to fix a bug that would place executable and non-executables pages in the same segment. PR 22758 bfd * elf.c (_bfd_elf_map_sections_to_segments): Don't start a new segment when demand paged with lma on the same page. Test this before load/non-load, executable/non-executable, writable/non-writable tests and simplify. Delete bogus relro condition in writable/non-writable test. Delete outdated comment. Formatting.
(In reply to Romain Geissler from comment #7) > Given that this > might potentially happen arbitrarily to any user of binutils 2.30 and it is > not so easy to understand that section placement is wrong here, would you > please backport this to the 2.30 branch ? Done.