Re: *** buffer overflow detected ***: /usr/src/linux/tools/perf/perf terminated

From: Markus Trippelsdorf
Date: Mon Dec 19 2016 - 11:53:08 EST


On 2016.12.19 at 17:18 +0100, Markus Trippelsdorf wrote:
> Running the latest kernel git tree, I get buffer overflow warnings when
> I try to run "perf top":
>
> *** buffer overflow detected ***: /usr/src/linux/tools/perf/perf terminated
>
> __GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:51
> 51 }
> (gdb) bt
> #0 __GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:51
> #1 0x00007ffff5f29cbb in __GI_abort () at abort.c:89
> #2 0x00007ffff5f6f553 in __libc_message (do_abort=do_abort@entry=2, fmt=fmt@entry=0x7ffff6056455 "*** %s ***: %s terminated\n") at ../sysdeps/posix/libc_fatal.c:175
> #3 0x00007ffff600d8c7 in __GI___fortify_fail (msg=msg@entry=0x7ffff60563ec "buffer overflow detected") at fortify_fail.c:30
> #4 0x00007ffff600b790 in __GI___chk_fail () at chk_fail.c:28
> #5 0x00007ffff600bc35 in __read_chk (fd=fd@entry=3, buf=buf@entry=0x7fffffff48f0, nbytes=nbytes@entry=1572864, buflen=buflen@entry=8192) at read_chk.c:30
> #6 0x00000000004fafc4 in read (__nbytes=1572864, __buf=0x7fffffff48f0, __fd=3) at /usr/include/bits/unistd.h:39
> #7 sysfs__read_build_id (filename=filename@entry=0x7fffffff6940 "/sys/kernel/notes", build_id=build_id@entry=0x1db5e34, size=size@entry=20) at util/symbol-elf.c:540
> #8 0x00000000004a8c22 in dso__read_running_kernel_build_id (dso=dso@entry=0x1db5d70, machine=machine@entry=0x1db5ab8) at util/dso.c:1152
> #9 0x00000000004b875a in machine__get_kernel (machine=machine@entry=0x1db5ab8) at util/machine.c:715
> #10 0x00000000004ba0ab in machine__create_kernel_maps (machine=machine@entry=0x1db5ab8) at util/machine.c:1127
> #11 0x00000000004bf169 in perf_session__create_kernel_maps (session=session@entry=0x1db59d0) at util/session.c:70
> #12 0x00000000004bf43b in perf_session__new (file=file@entry=0x0, repipe=repipe@entry=false, tool=tool@entry=0x0) at util/session.c:154
> #13 0x0000000000441eca in __cmd_top (top=0x7fffffff8a70) at builtin-top.c:925
> #14 cmd_top (argc=<optimized out>, argv=<optimized out>, prefix=<optimized out>) at builtin-top.c:1330
> #15 0x0000000000487481 in run_builtin (p=p@entry=0x679d78 <commands+312>, argc=argc@entry=2, argv=argv@entry=0x7fffffffe4c0) at perf.c:358
> #16 0x0000000000487722 in handle_internal_command (argc=2, argv=0x7fffffffe4c0) at perf.c:420
> #17 0x0000000000487982 in run_argv (argv=0x7fffffffe250, argcp=0x7fffffffe25c) at perf.c:466
> #18 main (argc=<optimized out>, argv=<optimized out>) at perf.c:610
>
> (gdb) up
> #7 sysfs__read_build_id (filename=filename@entry=0x7fffffff6940 "/sys/kernel/notes", build_id=build_id@entry=0x1db5e34, size=size@entry=20) at util/symbol-elf.c:540
> 540 if (read(fd, bf, n) != n)
> (gdb) l
> 535 }
> 536 } else if (read(fd, bf, descsz) != (ssize_t)descsz)
> 537 break;
> 538 } else {
> 539 int n = namesz + descsz;
> 540 if (read(fd, bf, n) != n)
> 541 break;
> 542 }
> 543 }
> 544 close(fd);
>
> (gdb) p n
> $4 = 1572864
> (gdb) p BUFSIZ
> $5 = 8192

It only happens when I link my kernel with gold.

(with ld.bfd)
markus@x4 linux % readelf -n vmlinux

Displaying notes found in: .notes
Owner Data size Description
GNU 0x00000014 NT_GNU_BUILD_ID (unique build ID bitstring)
Build ID: 74b478c2194c0ba07481acf30f6f8781d4ffcdee

(with ld.gold)
markus@x4 linux % readelf -n vmlinux

Displaying notes found in: .note.gnu.build-id
Owner Data size Description
GNU 0x00000014 NT_GNU_BUILD_ID (unique build ID bitstring)
Build ID: 017ef6e6f8ce3b44332d0baa0767526bb3a77bf8

Displaying notes found in: .note.gnu.gold-version
Owner Data size Description
GNU 0x00000009 NT_GNU_GOLD_VERSION (gold version)
Version: gold 1.12

And perf than chokes when parsing /sys/kernel/notes.

--
Markus