Date: Mon, 22 Jul 2024 04:38:00 -0700 From: Mark Millard <marklmi@yahoo.com> To: John F Carr <jfc@mit.edu>, Konstantin Belousov <kostikbel@gmail.com>, Baptiste Daroussin <bapt@FreeBSD.org> Cc: arm@freebsd.org, current@freebsd.org Subject: Re: armv7-on-aarch64 stuck at urdlck: I got a replication of the "ampere2" bulk build hangup problem on a Windows DevKit 2023 Message-ID: <0FD31A4D-DE2C-400A-B6DB-FA3EAB94E374@yahoo.com> In-Reply-To: <3D644314-7F84-4AA8-81A5-DCD29D4ABBC1@yahoo.com> References: <A3BBDBA3-FF7E-47B8-8B45-C3351918FC08@yahoo.com> <8214703E-AB28-4FB3-A3DD-03C87363D8C6@yahoo.com> <E40B106D-30B7-49DC-8540-911A56E53FCA@yahoo.com> <BFCA474F-78D4-4E98-8A9B-B92C2DB6B0DE@yahoo.com> <E40627DC-35A2-4D5E-A85C-73D7FFD426DF@yahoo.com> <A208F7C8-0D9E-402C-BB85-A5022B37E3EC@yahoo.com> <Zpt79jpCFDSiffcy@kib.kiev.ua> <8E9579B7-2ABF-4446-B65E-E993E7B67C5C@yahoo.com> <CBF08B5A-A701-4D8E-AE96-3CB54137B14E@yahoo.com> <D02F1ADE-4E1B-4DAF-81C0-8EBD5315FC66@yahoo.com> <9E98F6F6-6896-4958-9D88-FF68C4AB57F2@mit.edu> <DA56EE85-F50E-46F0-A8B4-538C698A4058@yahoo.com> <705AE360-C1C3-4B54-B6EE-FC81548D46B8@yahoo.com> <3D644314-7F84-4AA8-81A5-DCD29D4ABBC1@yahoo.com>
index | next in thread | previous in thread | raw e-mail
On Jul 21, 2024, at 21:08, Mark Millard <marklmi@yahoo.com> wrote:
> On Jul 21, 2024, at 20:58, Mark Millard <marklmi@yahoo.com> wrote:
>
>> I found a significant difference in my failing vs. working
>> armv7 contexts as installed: Presence vs. Lack of a .symtab
>> entry for the symbol _rtld_get_stack_prot in
>> /libexec/ld-elf.so.1 .
>>
>> gdb inspection of operation shows distinctions based on
>> the difference.
>>
>> This is related to the code:
>>
>> (gdb) list 140
>> 135 void
>> 136 _thr_stack_fix_protection(struct pthread *thrd)
>> 137 {
>> 138
>> 139 mprotect((char *)thrd->attr.stackaddr_attr +
>> 140 round_up(thrd->attr.guardsize_attr),
>> 141 round_up(thrd->attr.stacksize_attr),
>> 142 _rtld_get_stack_prot());
>> 143 }
>>
>>
>> Working context (Personal build):
>>
>> NOTE THE .symtab ENTRY BELOW. It allows the gdb run to work:
>>
>> # readelf -a /libexec/ld-elf.so.1 | grep -E "(^[^ 0-9]|.*_rtld_get_stack_prot)" | less
>> ELF Header:
>> Elf file type is DYN (Shared object file)
>> Entry point 0x14548
>> There are 10 program headers, starting at offset 52
>> Program Headers:
>> There are 24 section headers, starting at offset 0x1f2b8:
>> Section Headers:
>> Key to Flags:
>> Dynamic section at offset 0x199f8 contains 15 entries:
>> Relocation section (.rel.dyn):
>> r_offset r_info r_type st_value st_name
>> Symbol table '.dynsym' contains 27 entries:
>> 5: 000000000001b9ac 16 FUNC GLOBAL DEFAULT 11 _rtld_get_stack_prot@@FBSDprivate_1.0 (11)
>> Symbol table '.symtab' contains 911 entries:
>> 903: 000000000001b9ac 16 FUNC GLOBAL DEFAULT 11 _rtld_get_stack_prot
>> Notes at offset 0x00000174 with length 0x00000018:
>> Histogram for bucket list length (total of 6 buckets):
>> Histogram for bucket list length (total of 27 buckets):
>> Version symbol section (.gnu.version):
>> Version definition section (.gnu.version_d):
>> Attribute Section: aeabi
>> File Attributes
>>
>>
>> Breakpoint 8.3, _thr_stack_fix_protection (thrd=0x2006f000) at /usr/main-src/lib/libthr/thread/thr_stack.c:139
>> 139 mprotect((char *)thrd->attr.stackaddr_attr +
>> (gdb) si
>> 141 round_up(thrd->attr.stacksize_attr),
>> (gdb)
>> 140 round_up(thrd->attr.guardsize_attr),
>> (gdb)
>> round_up (size=4096) at /usr/main-src/lib/libthr/thread/thr_stack.c:129
>> 129 if (size % _thr_page_size != 0)
>> (gdb)
>> 0x201110b8 129 if (size % _thr_page_size != 0)
>> 130 size = ((size / _thr_page_size) + 1) *
>> (gdb)
>> 0x201110c0 130 size = ((size / _thr_page_size) + 1) *
>> (gdb)
>> 0x201110c4 in round_up (size=4096) at /usr/main-src/lib/libthr/thread/thr_stack.c:130
>> 130 size = ((size / _thr_page_size) + 1) *
>> (gdb)
>> 0x201110c8 130 size = ((size / _thr_page_size) + 1) *
>> (gdb)
>> round_up (size=67108864) at /usr/main-src/lib/libthr/thread/thr_stack.c:129
>> 129 if (size % _thr_page_size != 0)
>> (gdb)
>> 0x201110d0 in round_up (size=4096) at /usr/main-src/lib/libthr/thread/thr_stack.c:129
>> 129 if (size % _thr_page_size != 0)
>> (gdb)
>> 0x201110d4 in round_up (size=67108864) at /usr/main-src/lib/libthr/thread/thr_stack.c:129
>> 129 if (size % _thr_page_size != 0)
>> (gdb)
>> 0x201110d8 129 if (size % _thr_page_size != 0)
>> (gdb)
>> 0x201110dc in round_up (size=4096) at /usr/main-src/lib/libthr/thread/thr_stack.c:129
>> 129 if (size % _thr_page_size != 0)
>> (gdb)
>> 0x201110e0 129 if (size % _thr_page_size != 0)
>> (gdb)
>> _thr_stack_fix_protection (thrd=0x2006f000) at /usr/main-src/lib/libthr/thread/thr_stack.c:139
>> 139 mprotect((char *)thrd->attr.stackaddr_attr +
>> (gdb)
>> 142 _rtld_get_stack_prot());
>> (gdb)
>> 0x20114880 in ?? () from /lib/libthr.so.3
>> (gdb)
>> 0x20114884 in ?? () from /lib/libthr.so.3
>> (gdb)
>> 0x20114888 in ?? () from /lib/libthr.so.3
>> (gdb)
>>
>> Breakpoint 9.1, _rtld_get_stack_prot () at /usr/main-src/libexec/rtld-elf/rtld.c:5884
>> 5884 return (stack_prot);
>> (gdb)
>> 0x2005b9b0 5884 return (stack_prot);
>> (gdb)
>> 0x2005b9b4 5884 return (stack_prot);
>>
>>
>>
>> Failing context (Official PkgBase build):
>>
>> NOTE THE *LACK OF* THE .symtab ENTRY ABOVE.
>
> Not "ABOVE": BELOW! Sorry.
>
>> _rtld_bind_start ends
>> up in use instead, which looks to lead to the gdb run not working.
>>
>> IN FACT, NOTE THE LACK OF ANY "Symbol table '.symtab' contains"
>> TEXT AT ALL!
>>
>> # readelf -a /libexec/ld-elf.so.1 | grep -E "(^[^ 0-9]|.*_rtld_get_stack_prot)" | less
>> ELF Header:
>> Elf file type is DYN (Shared object file)
>> Entry point 0x147b0
>> There are 10 program headers, starting at offset 52
>> Program Headers:
>> There are 22 section headers, starting at offset 0x1a960:
>> Section Headers:
>> Key to Flags:
>> Dynamic section at offset 0x1a4cc contains 15 entries:
>> Relocation section (.rel.dyn):
>> r_offset r_info r_type st_value st_name
>> Symbol table '.dynsym' contains 27 entries:
>> 5: 000000000001bcd8 16 FUNC GLOBAL DEFAULT 11 _rtld_get_stack_prot@@FBSDprivate_1.0 (11)
>> Notes at offset 0x00000174 with length 0x00000018:
>> Histogram for bucket list length (total of 6 buckets):
>> Histogram for bucket list length (total of 27 buckets):
>> Version symbol section (.gnu.version):
>> Version definition section (.gnu.version_d):
>> Attribute Section: aeabi
>> File Attributes
>>
>>
>> Breakpoint 2.3, _thr_stack_fix_protection (thrd=0x20070000) at /home/pkgbuild/worktrees/main/lib/libthr/thread/thr_stack.c:140
>> 140 round_up(thrd->attr.guardsize_attr),
>> (gdb) si
>> 139 mprotect((char *)thrd->attr.stackaddr_attr +
>> (gdb)
>> 141 round_up(thrd->attr.stacksize_attr),
>> (gdb)
>> round_up (size=4096) at /home/pkgbuild/worktrees/main/lib/libthr/thread/thr_stack.c:129
>> 129 if (size % _thr_page_size != 0)
>> (gdb)
>> 130 size = ((size / _thr_page_size) + 1) *
>> (gdb)
>> 129 if (size % _thr_page_size != 0)
>> (gdb)
>> 130 size = ((size / _thr_page_size) + 1) *
>> (gdb)
>> 0x20112ef8 130 size = ((size / _thr_page_size) + 1) *
>> (gdb)
>> 0x20116b60 in ?? () from /lib/libthr.so.3
>> (gdb)
>> 0x20116b64 in ?? () from /lib/libthr.so.3
>> (gdb)
>> 0x20116b68 in ?? () from /lib/libthr.so.3
>> (gdb)
>> 0x20116760 in ?? () from /lib/libthr.so.3
>> (gdb)
>> 0x20116764 in ?? () from /lib/libthr.so.3
>> (gdb)
>> 0x20116768 in ?? () from /lib/libthr.so.3
>> (gdb)
>> 0x2011676c in ?? () from /lib/libthr.so.3
>> (gdb)
>> _rtld_bind_start () at /home/pkgbuild/worktrees/main/libexec/rtld-elf/arm/rtld_start.S:78
>> 78 stmdb sp!,{r0-r5,sl,fp}
>> (gdb) si
>> 80 sub r1, ip, lr /* r1 = 4 * (n + 1) */
>> (gdb)
>> 81 sub r1, r1, #4 /* r1 = 4 * n */
>> (gdb)
>> 82 add r1, r1, r1 /* r1 = 8 * n */
>> (gdb)
>> 84 ldr r0, [lr, #-4] /* get obj ptr from GOT[1] */
>> (gdb)
>> 85 mov r4, ip /* save GOT location */
>> (gdb)
>> 87 mov r5, sp /* Save the stack pointer */
>> (gdb)
>> 88 bic sp, sp, #7 /* Align the stack pointer */
>> (gdb)
>> _rtld_bind_start () at /home/pkgbuild/worktrees/main/libexec/rtld-elf/arm/rtld_start.S:89
>> 89 bl _rtld_bind /* Call the binder */
>>
>>
>> I have not checked for other .symtab entry problems.
>>
>> Nor have I figured out why the installed materials are
>> different for Symbol table '.symtab' . So this is not
>> yet root-cause information.
I know why. My personal FreeBSD builds have long had my
equivalents of src.conf (under normal naming) contain:
#
# Avoid stripping but do not control host -g status as well:
DEBUG_FLAGS+=
This likely replaces my earlier -mcpu=cortex-a7 hypothesis
for what a systematic difference might be that could
contribute to the personal builds not having the problem.
===
Mark Millard
marklmi at yahoo.com
help
Want to link to this message? Use this
URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?0FD31A4D-DE2C-400A-B6DB-FA3EAB94E374>