From: "H.J. Lu" <hjl.tools@gmail.com>
To: Florian Weimer <fweimer@redhat.com>
Cc: GNU C Library <libc-alpha@sourceware.org>
Subject: Re: [PATCH 1/5] Linux: Implement a useful version of _startup_fatal
Date: Tue, 3 May 2022 08:54:41 -0700 [thread overview]
Message-ID: <CAMe9rOo3iPYVT0F2mGkrSht13KDt_Hm9TXD+8EQSe0L7gdrwBw@mail.gmail.com> (raw)
In-Reply-To: <87bkwfxkhg.fsf@oldenburg.str.redhat.com>
On Mon, May 2, 2022 at 10:18 PM Florian Weimer <fweimer@redhat.com> wrote:
>
> * Florian Weimer:
>
> > * H. J. Lu:
> >
> >> On Mon, May 2, 2022 at 12:53 PM Florian Weimer <fweimer@redhat.com> wrote:
> >>>
> >>> * H. J. Lu:
> >>>
> >>> >> +#include_next <startup.h>
> >>> >> diff --git a/sysdeps/unix/sysv/linux/startup.h b/sysdeps/unix/sysv/linux/startup.h
> >>> >> new file mode 100644
> >>> >> index 0000000000..a5de941759
> >>> >> --- /dev/null
> >>> >> +++ b/sysdeps/unix/sysv/linux/startup.h
> >>> >> @@ -0,0 +1,38 @@
> >>> >> +/* Linux definitions of functions used by static libc main startup.
> >>> >> + Copyright (C) 2017-2022 Free Software Foundation, Inc.
> >>> > Shouldn't it just be 2022?
> >>> >> + This file is part of the GNU C Library.
> >>> >> +
> >>> >> + The GNU C Library is free software; you can redistribute it and/or
> >>> >> + modify it under the terms of the GNU Lesser General Public
> >>> >> + License as published by the Free Software Foundation; either
> >>> >> + version 2.1 of the License, or (at your option) any later version.
> >>> >> +
> >>> >> + The GNU C Library is distributed in the hope that it will be useful,
> >>> >> + but WITHOUT ANY WARRANTY; without even the implied warranty of
> >>> >> + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> >>> >> + Lesser General Public License for more details.
> >>> >> +
> >>> >> + You should have received a copy of the GNU Lesser General Public
> >>> >> + License along with the GNU C Library; if not, see
> >>> >> + <https://www.gnu.org/licenses/>. */
> >>> >> +
> >>> >> +#ifdef SHARED
> >>> >> +# include_next <startup.h>
> >>> >> +#else
> >>> >> +# include <sysdep.h>
> >>> >> +
> >>> >> +/* Avoid a run-time invocation of strlen. */
> >>> >> +#define _startup_fatal(message) \
> >>> >
> >>> > Is this needed only for static PIE?
> >>>
> >>> I don't know. It's used from csu/libc-tls.c, some targets may need it
> >>> for ET_EXEC static linking, too. I haven't checked. The additional
> >>> code size is minimal.
> >>
> >> _startup_fatal was added for i386 static PIE. I don't think it is
> >> needed for normal static executables.
> >
> > Based on the current initialization ordering, it seems to me that we
> > need it for non-PIE static programs, too. Maybe I should check this
> > tomorrow.
>
> It does not work on x86-64 because cancellation handling needs the TCB:
>
> Breakpoint 1, _dl_early_allocate (size=size@entry=3520) at ../sysdeps/unix/sysv/linux/dl-early_allocate.c:41
> 41 if (__curbrk != NULL)
> (gdb) finish
> Run till exit from #0 _dl_early_allocate (size=size@entry=3520)
> at ../sysdeps/unix/sysv/linux/dl-early_allocate.c:41
> __libc_setup_tls () at libc-tls.c:149
> 149 if (tlsblock == NULL)
> Value returned is $1 = (void *) 0x4eb000
> (gdb) set $rax = 0
> (gdb) c
> Continuing.
>
> Program received signal SIGSEGV, Segmentation fault.
> __writev (fd=fd@entry=2, iov=iov@entry=0x7fffffffddb0, iovcnt=iovcnt@entry=1) at ../sysdeps/unix/sysv/linux/writev.c:26
> 26 return SYSCALL_CANCEL (writev, fd, iov, iovcnt);
> (gdb) disas
> Dump of assembler code for function __writev:
> => 0x000000000044e3c0 <+0>: mov %fs:0x18,%eax
> 0x000000000044e3c8 <+8>: test %eax,%eax
> 0x000000000044e3ca <+10>: jne 0x44e3e0 <__writev+32>
> 0x000000000044e3cc <+12>: mov $0x14,%eax
> 0x000000000044e3d1 <+17>: syscall
> 0x000000000044e3d3 <+19>: cmp $0xfffffffffffff000,%rax
> 0x000000000044e3d9 <+25>: ja 0x44e430 <__writev+112>
> 0x000000000044e3db <+27>: ret
> 0x000000000044e3dc <+28>: nopl 0x0(%rax)
> 0x000000000044e3e0 <+32>: sub $0x28,%rsp
> 0x000000000044e3e4 <+36>: mov %edx,0x1c(%rsp)
> 0x000000000044e3e8 <+40>: mov %rsi,0x10(%rsp)
> 0x000000000044e3ed <+45>: mov %edi,0x8(%rsp)
> 0x000000000044e3f1 <+49>: call 0x46d2b0 <__pthread_enable_asynccancel>
> 0x000000000044e3f6 <+54>: mov 0x1c(%rsp),%edx
> 0x000000000044e3fa <+58>: mov 0x10(%rsp),%rsi
> 0x000000000044e3ff <+63>: mov %eax,%r8d
> 0x000000000044e402 <+66>: mov 0x8(%rsp),%edi
> 0x000000000044e406 <+70>: mov $0x14,%eax
> 0x000000000044e40b <+75>: syscall
> 0x000000000044e40d <+77>: cmp $0xfffffffffffff000,%rax
> 0x000000000044e413 <+83>: ja 0x44e448 <__writev+136>
> 0x000000000044e415 <+85>: mov %r8d,%edi
> 0x000000000044e418 <+88>: mov %rax,0x8(%rsp)
> 0x000000000044e41d <+93>: call 0x46d330 <__pthread_disable_asynccancel>
> 0x000000000044e422 <+98>: mov 0x8(%rsp),%rax
> 0x000000000044e427 <+103>: add $0x28,%rsp
> 0x000000000044e42b <+107>: ret
> 0x000000000044e42c <+108>: nopl 0x0(%rax)
> 0x000000000044e430 <+112>: mov $0xffffffffffffffa8,%rdx
> 0x000000000044e437 <+119>: neg %eax
> 0x000000000044e439 <+121>: mov %eax,%fs:(%rdx)
> 0x000000000044e43c <+124>: mov $0xffffffffffffffff,%rax
> 0x000000000044e443 <+131>: ret
> 0x000000000044e444 <+132>: nopl 0x0(%rax)
> 0x000000000044e448 <+136>: mov $0xffffffffffffffa8,%rdx
> 0x000000000044e44f <+143>: neg %eax
> 0x000000000044e451 <+145>: mov %eax,%fs:(%rdx)
> 0x000000000044e454 <+148>: mov $0xffffffffffffffff,%rax
> 0x000000000044e45b <+155>: jmp 0x44e415 <__writev+85>
> End of assembler dump.
> (gdb) bt
> #0 __writev (fd=fd@entry=2, iov=iov@entry=0x7fffffffddb0,
> iovcnt=iovcnt@entry=1) at ../sysdeps/unix/sysv/linux/writev.c:26
> #1 0x00000000004077aa in writev_for_fatal (fd=<optimized out>,
> total=<optimized out>, niov=<optimized out>, iov=<optimized out>)
> at ../sysdeps/posix/libc_fatal.c:44
> #2 __libc_message (action=action@entry=do_abort, fmt=fmt@entry=0x4c70ad "%s")
> at ../sysdeps/posix/libc_fatal.c:124
> #3 0x000000000040789c in __libc_fatal (
> message=message@entry=0x4adb60 "Fatal glibc error: Cannot allocate TLS block\n") at ../sysdeps/posix/libc_fatal.c:164
> #4 0x000000000040a3c0 in __libc_setup_tls () at libc-tls.c:150
> #5 0x0000000000409709 in __libc_start_main_impl (main=0x401740 <main>,
> argc=1, argv=0x7fffffffe008, init=<optimized out>, fini=<optimized out>,
> rtld_fini=0x0, stack_end=0x7fffffffdff8) at ../csu/libc-start.c:301
> #6 0x0000000000401da1 in _start () at ../sysdeps/x86_64/start.S:115
> (gdb)
>
> That might be a different bug.
The startup codes can't use cancellation before it is ready. Do we need to
check SHARED in this case?
--
H.J.
next prev parent reply other threads:[~2022-05-03 15:55 UTC|newest]
Thread overview: 23+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-05-02 19:15 [PATCH 0/5] Linux: Fall back to mmap if early sbrk fails Florian Weimer
2022-05-02 19:15 ` [PATCH 1/5] Linux: Implement a useful version of _startup_fatal Florian Weimer
2022-05-02 19:45 ` H.J. Lu
2022-05-02 19:52 ` Florian Weimer
2022-05-02 20:06 ` H.J. Lu
2022-05-02 20:37 ` Florian Weimer
2022-05-03 5:18 ` Florian Weimer
2022-05-03 15:54 ` H.J. Lu [this message]
2022-05-03 16:00 ` Florian Weimer
2022-05-03 16:36 ` H.J. Lu
2022-05-03 16:43 ` Florian Weimer
2022-05-03 16:46 ` H.J. Lu
2022-05-03 16:57 ` Florian Weimer
2022-05-03 17:01 ` Adhemerval Zanella
2022-05-03 20:16 ` Florian Weimer
2022-05-02 19:15 ` [PATCH 2/5] Linux: Define MMAP_CALL_INTERNAL Florian Weimer
2022-05-03 18:30 ` Florian Weimer
2022-05-04 13:10 ` Stefan Liebler
2022-05-02 19:15 ` [PATCH 3/5] i386: Remove OPTIMIZE_FOR_GCC_5 from Linux libc-do-syscall.S Florian Weimer
2022-05-02 19:20 ` H.J. Lu
2022-05-02 19:15 ` [PATCH 4/5] i386: Honor I386_USE_SYSENTER for 6-argument Linux system calls Florian Weimer
2022-05-02 19:24 ` H.J. Lu
2022-05-02 19:15 ` [PATCH 5/5] csu: Implement and use _dl_early_allocate during static startup Florian Weimer
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=CAMe9rOo3iPYVT0F2mGkrSht13KDt_Hm9TXD+8EQSe0L7gdrwBw@mail.gmail.com \
--to=hjl.tools@gmail.com \
--cc=fweimer@redhat.com \
--cc=libc-alpha@sourceware.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).