Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 25 Dec 2015 17:21:11 -0800
From:      Mark Millard <markmi@dsl-only.net>
To:        Warner Losh <imp@bsdimp.com>
Cc:        freebsd-arm <freebsd-arm@freebsd.org>, FreeBSD Toolchain <freebsd-toolchain@freebsd.org>, Ian Lepore <ian@FreeBSD.org>, mat@FreeBSD.org, sbruno@FreeBSD.org
Subject:   Re: 11.0-CURRENT (r292413) on a rpi2b: arm-gnueabi-freebsd/bin/ar, _fseeko, and memset vs memory alignment (SCTRL bit[1]=1?): Explains the Bus error?
Message-ID:  <BBAAE33E-BD65-40A3-A0B3-F3346FC08112@dsl-only.net>
In-Reply-To: <8B52074F-FDEF-4119-BB04-630F9BE9E6DB@bsdimp.com>
References:  <4CC6220D-72FB-45AD-AE70-5EB4EF0BCF5C@dsl-only.net> <DB75F0D6-86CB-4383-8653-6017C76729F9@dsl-only.net> <A338272B-982F-4E1F-B87D-1E33815EA212@dsl-only.net> <0D81C2CA-BF1C-4C14-B816-A8C5F68715B5@bsdimp.com> <51EB4AAB-BC81-4282-BA4D-D329C41D660B@dsl-only.net> <8B52074F-FDEF-4119-BB04-630F9BE9E6DB@bsdimp.com>

next in thread | previous in thread | raw e-mail | index | archive | help
> On 2015-Dec-25, at 3:42 PM, Warner Losh <imp@bsdimp.com> wrote:
>=20
>=20
>> On Dec 25, 2015, at 3:14 PM, Mark Millard <markmi@dsl-only.net> =
wrote:
>>=20
>> [I'm going to break much of the earlier "original material" text to =
tail of the message.]
>>=20
>>> On 2015-Dec-25, at 11:53 AM, Warner Losh <imp@bsdimp.com> wrote:
>>>=20
>>> So what happens if we actually fix the underlying bug?
>>>=20
>>> I see two ways of doing this. In findfp.c, we allocate an array of =
FILE * today like:
>>>      g =3D (struct glue *)malloc(sizeof(*g) + ALIGNBYTES + n * =
sizeof(FILE));
>>> but that assumes that FILE just has normal pointer alignment =
requirements. However,
>>> due to the mbstate having int64_t alignment requirements, this is =
wrong. Maybe we
>>> need to do something like
>>> 	g =3D (struct glue *)malloc(sizeof(*g) + =
max(sizeof(int64_t),ALIGNBYTES) + n * sizeof(FILE));
>>> which wouldn=E2=80=99t change anything on LP64 systems, but would =
result in proper alignment
>>> for ILP32 systems. We=E2=80=99d have to fix the loop that uses ALIGN =
afterwards to use
>>> roundup. Instead, we=E2=80=99d need to round up to the neared 8-byte =
aligned offset (or technically,
>>> the max of ALIGNBYTES and 8, but that=E2=80=99s always 8 on =
today=E2=80=99s systems. If we do this,
>>> we can make sure that each file is 8-byte aligned or better. We may =
need to round up
>>> sizeof(FILE) to a multiple of 8 as well. I believe that since it has =
the 8-byte alignment
>>> for a member, its size must be a multiple of 8, but I=E2=80=99ve not =
chased that belief to ground.
>>> If not, we may need another decorator (__aligned(8), I think, =
spelled with the ugly
>>> max expression above). That way, the contract we=E2=80=99re making =
with the compiler will
>>> always be true. ALIGN BYTES is 4 on Arm anyway, so that bit is =
clearly wrong.
>>>=20
>>> This wouldn=E2=80=99t be an ABI change, since you can only get a =
valid FILE * from fopen (and
>>> friends), plus stdin, stdout, and stderr. Those addresses aren=E2=80=99=
t hard coded into binaries,
>>> so even if we have to tweak the last three and deal with some =
=E2=80=98fake=E2=80=99 FILE abuse in libc
>>> (which I don=E2=80=99t think suffers from this issue, btw, given the =
alignment requirements that would
>>> naturally follow from something on the stack), we=E2=80=99d still be =
ahead. At least for all CONFORMING
>>> implementations[*]...
>>>=20
>>> TL;DR: Why not make FILE * always 8-byte aligned? The compiler =
options are a band-aide.
>>>=20
>>> Warner
>>>=20
>>> [*] There=E2=80=99s at least on popular package that has a copy of =
the FILE structure in one of its
>>> .h files and uses that to do unnatural optimization things, but even =
that=E2=80=99s cool, I think,
>>> since it never allocates a new one.
>>>=20
>>=20
>> The ARM documentation mentions cases of 16 byte alignment =
requirements. I've no clue if the clang code generation ever creates =
such code. There might be wider requirements possible in arm code as =
well. (I'm not an arm expert.) As an example of an implication: "The =
malloc() function returns a pointer to a block of at least size bytes =
suitably aligned for any use." In other words: aligned to some figure =
that is a multiple of *every* alignment requirement that the code =
generator can produce, possibly being the least common multiple.
>>=20
>> "-fmax-type-align=3D. . ." is a means of controlling/limiting the =
range of potential alignments to no more than a fixed, predefined value. =
Above that and the code generation has to work in small size accesses =
and build-up/split-up bigger values. Using "-fmax-type-align=3D. . ." =
allows defining a figure as part of an ABI that is then not subject to =
code generator updates that could increase the maximum alignment figure =
and break things: It turns off such new capabilities. Other options need =
not work that way to preserve the ABI.
>=20
> That=E2=80=99s true, as far as it goes=E2=80=A6 But I=E2=80=99m not =
sure it goes far enough. The premise here is that the problem is =
wide-spread, when in fact I think it is quite narrow.
>=20
>> But in the most fundamental terms process wise as far as I can tell. =
. .
>>=20
>> While the FILE case that occurred is a specific example, every =
memory-allocation-like operation is at a potential issue for all such =
"allocated" objects where the related code generation requires alignment =
to avoid Bus Error (given the SCTLR bit[1] in use).
>=20
> The problem isn=E2=80=99t general. The problem isn=E2=80=99t malloc. =
Malloc will generally return the right thing on arm (and if it =
doesn=E2=80=99t,
> then we need to make sure it does).
>=20
> The problem is we get a boatload of FILEs from the system all at once, =
and those are misaligned because of a bug in the code. One that=E2=80=99s =
fixed, I believe, in https://reviews.freebsd.org/D4708.
>=20
>=20
>> How many other places in FreeBSD might sometimes return mis-aligned =
pointers for the existing code generation and ABI combination?
>=20
> It isn=E2=80=99t an ABI thing, just a code bug thing. The only reason =
it was an issue was due to the optimizing nature of clang.
>=20
> We=E2=80=99ve had to deal with the arm alignment issues for years. I =
wager there are very few indeed. The only reason this was was brought to =
light was better code-gen from clang.
>=20
>> How many other places are subject to breakage when "internal" =
structs/unions/fields involved are changed to be of a different size =
because the code is not fully auto-adjusting to match the code =
generation yet --even if right now "it works"? How fragile will things =
be for future work?
>=20
> If there are others, I=E2=80=99ll bet they could be counted on one =
hand since very few things do the =E2=80=98slab=E2=80=99 allocator that =
FILE does.
>=20
>> What would it take to find out and deal with them all? (I do not have =
the background knowledge to span much.)
>>=20
>> My experiment avoided potentially changing parts of the ABI and also =
avoided dealing with such a "lots of code to investigate" issue. It may =
not be the long term 11.0-RELEASE solution. Even if not, it may be =
appropriate for various temporary purposes that need to avoid Bus Errors =
in the process. For example if Ian has a good reason to use clang 3.7 =
instead of gcc 4.2.1.
>=20
> The review above doesn=E2=80=99t change the ABI either.
>=20
>> Other notes:
>>=20
>>> I believe that since it has the 8-byte alignment
>>> for a member, its size must be a multiple of 8
>>=20
>> There are some C/C++ language rules about the address of a structure =
equalling the address of the first field, uniformity of the offsets, and =
the like. But. . .
>>=20
>> The C and C++ languages specify no specific numerical alignment =
figures, not even relative to specific sizeof(...) expressions. To use =
an old example: a 68010 only needs alignment for >=3D 2 byte things and =
even alignment is all that is then required. Some other contexts take a =
lot more to meet the specifications. There are some implications of the =
modern memory model(s) created to cover concurrency explicitly, such as =
avoiding interactions that can happen via, for example, separate objects =
(in part) sharing a cache line. (I've only looked at C++ for this, and =
only to a degree.)
>>=20
>> The detailed alignment rules are more "implementation defined" than =
"predefined by the standard". But the definition is trying to meet =
language criteria. It is not a fully independent choice.
>=20
> Many of them are actually defined by a combination of the standard =
language definition, as well as the ABI standard. This is why we know =
that mbstate_t must be 8 byte aligned.
>=20
>> May be some other standards that FreeBSD is tied to specify more =
specifics, such as a N byte integer always aligns to some multiple of N =
(a waste on the 68010), including the alignment for union or struct that =
it may be a part of tracking. But such rules force padding that may or =
may not be required to meet the language's more abstract criteria and =
such rules may not match the existing/in-use ABI.
>=20
> It is all spelled out in the ARM EABI docs.
>=20
>> So far as I can tell explicitly declared alignments may well be =
necessary. If that one "popular package", say, formed an array of FILE =
copies then the resultant alignments need not all match the ones =
produced by your example code unless the FILE declaration forces the =
compiler to match, causing sizeof(FILE) to track as well. FILE need not =
be the only such issue.
>=20
> Arrays of FILEs isn=E2=80=99t an issue (except that it encodes the =
size of FILE into the app). It=E2=80=99s the specifically quirky way =
that libc does it that=E2=80=99s the problem.
>=20
>> My background and reference material are mostly tied the languages =
--and so my notes tend to be limited to that much context.
>=20
> Understood. While there may be issues with alignment still, tossing a =
big hammer at the problem because they might exist will likely mean they =
will persist far longer than fixing them one at a time. When we first =
ported to arm, there were maybe half a dozen places that needed fixing. =
I doubt there=E2=80=99s more now.
>=20
> Can you try the patch in the above code review w/o the -f switch and =
let me know if it works for you?
>=20
> Warner

buildworld/buildkernel has been started on amd64 for a rpi2 target. That =
and install kernel/world and starting up a port rebuild on the rpi2 and =
waiting for it means it will be a few hours even if I start the next =
thing just as each prior thing finishes. I may give up and go to sleep =
first.

As for presumptions: I'll take your word on expected status of things. =
I've no clue. But absent even the hear-say status information at the =
time I did not presume that what was in front of me was all there is to =
worry about --nor did I try to go figure it all out on my own. I took a =
path to cover both possibilities for local-only vs. more-wide-spread (so =
long as that path did not force a split-up of some larger form of atomic =
action).

In my view "-mno-unaligned-access" is an even bigger hammer than I used. =
I find no clang statement about what its ABI consequences would be, =
unlike for what I did: What mix of more padding for alignment vs. more =
but smaller accesses? But as I remember I've seen =
"-mno-unaligned-access" in use in ports and the like so its consequences =
may be familiar material for some folks.

Absent any questions about ABI consequences "-mno-unaligned-access" does =
well mark the expected SCTLR bit[1] status, far better than what I did. =
Again: I was covering my ignorance while making any significant =
investigation/debugging as unlikely as I could.


> Original material:
>=20
>> On Dec 25, 2015, at 7:24 AM, Mark Millard <markmi@dsl-only.net> =
wrote:
>>=20
>> [Good News Summary: Rebuilding buildworld/buildkernel for rpi2 =
11.0-CURRENT 292413 from amd64 based on adding -fmax-type-align=3D4 has =
so far removed the crashes during the toolchain activity: no more =
misaligned accesses in libc's _fseeko or elsewhere.]
>>=20
>> On 2015-Dec-25, at 12:31 AM, Mark Millard <markmi@dsl-only.net> =
wrote:
>>=20
>>> On 2015-Dec-24, at 10:39 PM, Mark Millard <markmi@dsl-only.net> =
wrote:
>>>=20
>>>> [I do not know if this partial crash analysis related to on-arm =
clang-associated activity is good enough and appropriate to submit or =
not.]
>>>>=20
>>>> The /usr/local/arm-gnueabi-freebsd/bin/ar on the rpi2b involved =
below came from pkg install activity instead of port building. Used =
as-is.
>>>>=20
>>>> When I just tried my first from-rpi2b builds (ports for a rpi2b), =
/usr/local/arm-gnueabi-freebsd/bin/ar crashed. I believe that the =
following suggests an alignment error for the type of instructions that =
memset for 128 bytes was translated to (sizeof(mbstate_t)) in the code =
used by /usr/local/arm-gnueabi-freebsd/bin/ar. (But I do not know how to =
check SCTLR bit[1] to be directly sure that alignment was being =
enforced.)
>>>>=20
>>>> The crash was a Bus error in /usr/local/arm-gnueabi-freebsd/bin/ar =
:
>>>>=20
>>>>> libtool: link: /usr/local/arm-gnueabi-freebsd/bin/ar cru =
.libs/libgnuintl.a  bindtextdom.o dcgettext.o dgettext.o gettext.o =
finddomain.o hash-string.o loadmsgcat.o localealias.o textdomain.o =
l10nflist.o explodename.o dcigettext.o dcngettext.o dngettext.o =
ngettext.o pluralx.o plural-exp.o localcharset.o threadlib.o lock.o =
relocatable.o langprefs.o localename.o log.o printf.o setlocale.o =
version.o xsize.o osdep.o intl-compat.o
>>>>> Bus error (core dumped)
>>>>> *** [libgnuintl.la] Error code 138
>>>>=20
>>>> It failed in _fseeko doing a memset that turned into uses of =
"vst1.64	{d16-d17}, [r0]" instructions, for an address in =
register r0 that ended in 0xa4, so was not aligned to 8 byte boundaries. =
=46rom what I read such "VSTn (multiple n-element structures)" that have =
.64 require 8 byte alignment. The evidence of the code and register =
value follow.
>>>>=20
>>>>> # gdb /usr/local/arm-gnueabi-freebsd/bin/ar =
/usr/obj/portswork/usr/ports/devel/gettext-tools/work/gettext-0.19.6/gette=
xt-tools/intl/ar.core
>>>>> . . .
>>>>> #0  0x2033adcc in _fseeko (fp=3D0x20651dcc, offset=3D<value =
optimized out>, whence=3D<value optimized out>, ltest=3D<value optimized =
out>) at /usr/src/lib/libc/stdio/fseek.c:299
>>>>> 299		memset(&fp->_mbstate, 0, sizeof(mbstate_t));
>>>>> . . .
>>>>> (gdb) x/24i 0x2033adb0
>>>>> 0x2033adb0 <_fseeko+836>:	vmov.i32	q8, #0	; 0x00000000
>>>>> 0x2033adb4 <_fseeko+840>:	movw	r1, #65503	; 0xffdf
>>>>> 0x2033adb8 <_fseeko+844>:	stm	r4, {r0, r7}
>>>>> 0x2033adbc <_fseeko+848>:	ldrh	r0, [r4, #12]
>>>>> 0x2033adc0 <_fseeko+852>:	and	r0, r0, r1
>>>>> 0x2033adc4 <_fseeko+856>:	strh	r0, [r4, #12]
>>>>> 0x2033adc8 <_fseeko+860>:	add	r0, r4, #216	; 0xd8
>>>>> 0x2033adcc <_fseeko+864>:	vst1.64	{d16-d17}, [r0]
>>>>> 0x2033add0 <_fseeko+868>:	add	r0, r4, #200	; 0xc8
>>>>> 0x2033add4 <_fseeko+872>:	vst1.64	{d16-d17}, [r0]
>>>>> 0x2033add8 <_fseeko+876>:	add	r0, r4, #184	; 0xb8
>>>>> 0x2033addc <_fseeko+880>:	vst1.64	{d16-d17}, [r0]
>>>>> 0x2033ade0 <_fseeko+884>:	add	r0, r4, #168	; 0xa8
>>>>> 0x2033ade4 <_fseeko+888>:	vst1.64	{d16-d17}, [r0]
>>>>> 0x2033ade8 <_fseeko+892>:	add	r0, r4, #152	; 0x98
>>>>> 0x2033adec <_fseeko+896>:	vst1.64	{d16-d17}, [r0]
>>>>> 0x2033adf0 <_fseeko+900>:	add	r0, r4, #136	; 0x88
>>>>> 0x2033adf4 <_fseeko+904>:	vst1.64	{d16-d17}, [r0]
>>>>> 0x2033adf8 <_fseeko+908>:	add	r0, r4, #120	; 0x78
>>>>> 0x2033adfc <_fseeko+912>:	vst1.64	{d16-d17}, [r0]
>>>>> 0x2033ae00 <_fseeko+916>:	add	r0, r4, #104	; 0x68
>>>>> 0x2033ae04 <_fseeko+920>:	vst1.64	{d16-d17}, [r0]
>>>>> 0x2033ae08 <_fseeko+924>:	b	0x2033b070 <_fseeko+1540>
>>>>> 0x2033ae0c <_fseeko+928>:	cmp	r5, #0	; 0x0
>>>>> (gdb) info all-registers
>>>>> r0             0x20651ea4	543497892
>>>>> r1             0xffdf	65503
>>>>> r2             0x0	0
>>>>> r3             0x0	0
>>>>> r4             0x20651dcc	543497676
>>>>> r5             0x0	0
>>>>> r6             0x0	0
>>>>> r7             0x0	0
>>>>> r8             0x20359df4	540384756
>>>>> r9             0x0	0
>>>>> r10            0x0	0
>>>>> r11            0xbfbfb948	-1077954232
>>>>> r12            0x2037b208	540520968
>>>>> sp             0xbfbfb898	-1077954408
>>>>> lr             0x2035a004	540385284
>>>>> pc             0x2033adcc	540257740
>>>>> f0             0	(raw 0x000000000000000000000000)
>>>>> f1             0	(raw 0x000000000000000000000000)
>>>>> f2             0	(raw 0x000000000000000000000000)
>>>>> f3             0	(raw 0x000000000000000000000000)
>>>>> f4             0	(raw 0x000000000000000000000000)
>>>>> f5             0	(raw 0x000000000000000000000000)
>>>>> f6             0	(raw 0x000000000000000000000000)
>>>>> f7             0	(raw 0x000000000000000000000000)
>>>>> fps            0x0	0
>>>>> cpsr           0x60000010	1610612752
>>>>=20
>>>> The syntax in use for vst1.64 instructions does not explicitly have =
the alignment notation. Presuming that the decoding is correct then from =
what I read the following applies:
>>>>=20
>>>>> Home > NEON and VFP Programming > NEON load and store element and =
structure instructions > Alignment restrictions in load and store, =
element and structure instructions
>>>>>=20
>>>>> . . . When the alignment is not specified in the instruction, the =
alignment restriction is controlled by the A bit (SCTLR bit[1]):
>>>>> 	=E2=80=A2	if the A bit is 0, there are no alignment =
restrictions (except for strongly ordered or device memory, where =
accesses must be element aligned or the result is unpredictable)
>>>>> 	=E2=80=A2	if the A bit is 1, accesses must be element =
aligned.
>>>>> If an address is not correctly aligned, an alignment fault occurs.
>>>>=20
>>>> So if at the time the "A bit" (SCTLR bit[1]) is 1 then the Bus =
error would have the context to happen because of the mis-alignment.
>>>>=20
>>>> The following shows the make.conf context that explains how =
/usr/local/arm-gnueabi-freebsd/bin/ar came to be invoked:
>>>>=20
>>>>> # more /etc/make.conf
>>>>> WRKDIRPREFIX=3D/usr/obj/portswork
>>>>> WITH_DEBUG=3D
>>>>> WITH_DEBUG_FILES=3D
>>>>> MALLOC_PRODUCTION=3D
>>>>> #
>>>>> TO_TYPE=3Darmv6
>>>>> TOOLS_TO_TYPE=3Darm-gnueabi
>>>>> CROSS_BINUTILS_PREFIX=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/
>>>>> .if ${.MAKE.LEVEL} =3D=3D 0
>>>>> CC=3D/usr/bin/clang -target ${TO_TYPE}--freebsd11.0-gnueabi =
-march=3Darmv7a
>>>>> CXX=3D/usr/bin/clang++ -target ${TO_TYPE}--freebsd11.0-gnueabi =
-march=3Darmv7a
>>>>> CPP=3D/usr/bin/clang-cpp -target ${TO_TYPE}--freebsd11.0-gnueabi =
-march=3Darmv7a
>>>>> .export CC
>>>>> .export CXX
>>>>> .export CPP
>>>>> AS=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/as
>>>>> AR=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/ar
>>>>> LD=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/ld
>>>>> NM=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/nm
>>>>> OBJCOPY=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/objcopy
>>>>> OBJDUMP=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/objdump
>>>>> RANLIB=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/ranlib
>>>>> SIZE=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/size
>>>>> #NO-SUCH: STRINGS=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/strings
>>>>> STRINGS=3D/usr/local/bin/${TOOLS_TO_TYPE}-freebsd-strings
>>>>> .export AS
>>>>> .export AR
>>>>> .export LD
>>>>> .export NM
>>>>> .export OBJCOPY
>>>>> .export OBJDUMP
>>>>> .export RANLIB
>>>>> .export SIZE
>>>>> .export STRINGS
>>>>> .endif
>>>>=20
>>>>=20
>>>> Other context:
>>>>=20
>>>>> # freebsd-version -ku; uname -aKU
>>>>> 11.0-CURRENT
>>>>> 11.0-CURRENT
>>>>> FreeBSD rpi2 11.0-CURRENT FreeBSD 11.0-CURRENT #0 r292413M: Tue =
Dec 22 22:02:21 PST 2015     =
root@FreeBSDx64:/usr/obj/clang/arm.armv6/usr/src/sys/RPI2-NODBG  arm =
1100091 1100091
>>>>=20
>>>>=20
>>>>=20
>>>> I will note that world and kernel are my own build of -r292413 =
(earlier experiment) --a build made from an amd64 host context and put =
in place via DESTDIR=3D. My expectation would be that the amd64 context =
would not be likely to have similar alignment restrictions involved in =
its ar activity (or other activity). That would explain how I got this =
far using such a clang 3.7 related toolchain for targeting an rpi2 =
before finding such a problem.
>>>=20
>>>=20
>>> I realized re-reading the all above that it seems to suggest that =
the _fseeko code involved is from /usr/local/arm-gnueabi-freebsd/bin/ar =
but that was not my intent.
>>>=20
>>> libc.so.7 is from my buildworld, including the fseeko =
implementation:
>>>=20
>>> Reading symbols from /lib/libc.so.7...Reading symbols from =
/usr/lib/debug//lib/libc.so.7.debug...done.
>>> done.
>>> Loaded symbols for /lib/libc.so.7
>>>=20
>>>=20
>>> head/sys/sys/_types.h has:
>>>=20
>>> /*
>>> * mbstate_t is an opaque object to keep conversion state during =
multibyte
>>> * stream conversions.
>>> */
>>> typedef union {
>>>    char            __mbstate8[128];
>>>    __int64_t       _mbstateL;      /* for alignment */
>>> } __mbstate_t;
>>>=20
>>> suggesting an implicit alignment of the union to whatever the =
implementation defines for __int64_t --which need not be 8 byte =
alignment (in the abstract, general case). But 8 byte alignment is a =
possibility as well (in the abstract).
>>>=20
>>> But printing *fp in gdb for the fp argument to _fseeko reports the =
same not-8-byte aligned address for __mbstate8 that was in r0:
>>>=20
>>>> (gdb) bt
>>>> #0  0x2033adcc in _fseeko (fp=3D0x20651dcc, offset=3D<value =
optimized out>, whence=3D<value optimized out>, ltest=3D<value optimized =
out>) at /usr/src/lib/libc/stdio/fseek.c:299
>>>> #1  0x2033b108 in fseeko (fp=3D0x20651dcc, offset=3D18571438587904, =
whence=3D0) at /usr/src/lib/libc/stdio/fseek.c:82
>>>> #2  0x00016138 in ?? ()
>>>> (gdb) print fp
>>>> $2 =3D (FILE *) 0x20651dcc
>>>> (gdb) print *fp
>>>> $3 =3D {_p =3D 0x2069a240 "", _r =3D 0, _w =3D 0, _flags =3D 5264, =
_file =3D 36, _bf =3D {_base =3D 0x2069a240 "", _size =3D 32768}, =
_lbfsize =3D 0, _cookie =3D 0x20651dcc, _close =3D 0x20359dfc =
<__sclose>,
>>>> _read =3D 0x20359de4 <__sread>, _seek =3D 0x20359df4 <__sseek>, =
_write =3D 0x20359dec <__swrite>, _ub =3D {_base =3D 0x0, _size =3D 0}, =
_up =3D 0x0, _ur =3D 0, _ubuf =3D 0x20651e0c "", _nbuf =3D 0x20651e0f =
"", _lb =3D {
>>>> _base =3D 0x0, _size =3D 0}, _blksize =3D 32768, _offset =3D 0, =
_fl_mutex =3D 0x0, _fl_owner =3D 0x0, _fl_count =3D 0, _orientation =3D =
0, _mbstate =3D {__mbstate8 =3D 0x20651e34 "", _mbstateL =3D 0}, _flags2 =
=3D 0}
>>>=20
>>> The overall FILE struct containing the _mbstate field is also not =
8-byte aligned. But the offset from the start of the FILE struct to =
__mbstate8 is a multiple of 8 bytes.
>>>=20
>>> It is my interpretation that there is nothing here to justify the =
memset implementation combination:
>>>=20
>>> SCTLR bit[1]=3D=3D1
>>>=20
>>> mixed with
>>>=20
>>> vst1.64 instructions
>>>=20
>>> I.e.: one or both needs to change unless some way for forcing 8-byte =
alignment is introduced.
>>>=20
>>> I have not managed to track down anything that would indicate =
FreeBSD's intent for SCTLR bit[1]. I do not even know if it is required =
by the design to be constant (once initialized).
>>=20
>>=20
>> I have (so far) removed the build tool crashes based on adding =
-fmax-type-align=3D4 to avoid the misaligned accesses. Details follow.
>>=20
>> src.conf on amd64 for the rpi2 targeting buildworld/buildkernel now =
looks like:
>>=20
>>> # more ~/src.configs/src.conf.rpi2-clang.amd64-host
>>> TO_TYPE=3Darmv6
>>> TOOLS_TO_TYPE=3Darm-gnueabi
>>> FROM_TYPE=3Damd64
>>> TOOLS_FROM_TYPE=3Dx86_64
>>> VERSION_CONTEXT=3D11.0
>>> #
>>> KERNCONF=3DRPI2-NODBG
>>> TARGET=3Darm
>>> .if ${.MAKE.LEVEL} =3D=3D 0
>>> TARGET_ARCH=3D${TO_TYPE}
>>> .export TARGET_ARCH
>>> .endif
>>> #
>>> WITHOUT_CROSS_COMPILER=3D
>>> #
>>> # For WITH_BOOT=3D . . .
>>> # arm-gnueabi-freebsd/bin/ld reports bootinfo.o: relocation =
R_ARM_MOVW_ABS_NC against `a local symbol' can not be used when making a =
shared object; recompile with -fPIC
>>> WITHOUT_BOOT=3D
>>> #
>>> WITH_FAST_DEPEND=3D
>>> WITH_LIBCPLUSPLUS=3D
>>> WITH_CLANG=3D
>>> WITH_CLANG_IS_CC=3D
>>> WITH_CLANG_FULL=3D
>>> WITH_LLDB=3D
>>> WITH_CLANG_EXTRAS=3D
>>> #
>>> WITHOUT_LIB32=3D
>>> WITHOUT_GCC=3D
>>> WITHOUT_GNUCXX=3D
>>> #
>>> NO_WERROR=3D
>>> MALLOC_PRODUCTION=3D
>>> #CFLAGS+=3D -DELF_VERBOSE
>>> #
>>> WITH_DEBUG=3D
>>> WITH_DEBUG_FILES=3D
>>> #
>>> # TOOLS_TO_TYPE based on ${TO_TYPE}-xtoolchain-gcc related =
bintutils...
>>> #
>>> #CROSS_TOOLCHAIN=3D${TO_TYPE}-gcc
>>> X_COMPILER_TYPE=3Dclang
>>> CROSS_BINUTILS_PREFIX=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/
>>> .if ${.MAKE.LEVEL} =3D=3D 0
>>> XCC=3D/usr/bin/clang -target ${TO_TYPE}--freebsd11.0-gnueabi =
-march=3Darmv7a -fmax-type-align=3D4
>>> XCXX=3D/usr/bin/clang++ -target ${TO_TYPE}--freebsd11.0-gnueabi =
-march=3Darmv7a -fmax-type-align=3D4
>>> XCPP=3D/usr/bin/clang-cpp -target ${TO_TYPE}--freebsd11.0-gnueabi =
-march=3Darmv7a -fmax-type-align=3D4
>>> .export XCC
>>> .export XCXX
>>> .export XCPP
>>> XAS=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/as
>>> XAR=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/ar
>>> XLD=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/ld
>>> XNM=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/nm
>>> XOBJCOPY=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/objcopy
>>> XOBJDUMP=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/objdump
>>> XRANLIB=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/ranlib
>>> XSIZE=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/size
>>> #NO-SUCH: XSTRINGS=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/strings
>>> XSTRINGS=3D/usr/local/bin/${TOOLS_TO_TYPE}-freebsd-strings
>>> .export XAS
>>> .export XAR
>>> .export XLD
>>> .export XNM
>>> .export XOBJCOPY
>>> .export XOBJDUMP
>>> .export XRANLIB
>>> .export XSIZE
>>> .export XSTRINGS
>>> .endif
>>> #
>>> # Host compiler stuff:
>>> .if ${.MAKE.LEVEL} =3D=3D 0
>>> CC=3D/usr/bin/clang -B/usr/local/${TOOLS_FROM_TYPE}-freebsd/bin
>>> CXX=3D/usr/bin/clang++ -B/usr/local/${TOOLS_FROM_TYPE}-freebsd/bin
>>> CPP=3D/usr/bin/clang-cpp -B/usr/local/${TOOLS_FROM_TYPE}-freebsd/bin
>>> .export CC
>>> .export CXX
>>> .export CPP
>>> AS=3D/usr/local/${TOOLS_FROM_TYPE}-freebsd/bin/as
>>> AR=3D/usr/local/${TOOLS_FROM_TYPE}-freebsd/bin/ar
>>> LD=3D/usr/local/${TOOLS_FROM_TYPE}-freebsd/bin/ld
>>> NM=3D/usr/local/${TOOLS_FROM_TYPE}-freebsd/bin/nm
>>> OBJCOPY=3D/usr/local/${TOOLS_FROM_TYPE}-freebsd/bin/objcopy
>>> OBJDUMP=3D/usr/local/${TOOLS_FROM_TYPE}-freebsd/bin/objdump
>>> RANLIB=3D/usr/local/${TOOLS_FROM_TYPE}-freebsd/bin/ranlib
>>> SIZE=3D/usr/local/${TOOLS_FROM_TYPE}-freebsd/bin/size
>>> #NO-SUCH: STRINGS=3D/usr/local/${TOOLS_FROM_TYPE}-freebsd/bin/strings
>>> STRINGS=3D/usr/local/bin/${TOOLS_FROM_TYPE}-freebsd-strings
>>> .export AS
>>> .export AR
>>> .export LD
>>> .export NM
>>> .export OBJCOPY
>>> .export OBJDUMP
>>> .export RANLIB
>>> .export SIZE
>>> .export STRINGS
>>> .endif
>>=20
>> make.conf for during the on-rpi2 port builds now looks like:
>>=20
>>> $ more /etc/make.conf
>>> WRKDIRPREFIX=3D/usr/obj/portswork
>>> WITH_DEBUG=3D
>>> WITH_DEBUG_FILES=3D
>>> MALLOC_PRODUCTION=3D
>>> #
>>> TO_TYPE=3Darmv6
>>> TOOLS_TO_TYPE=3Darm-gnueabi
>>> CROSS_BINUTILS_PREFIX=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/
>>> .if ${.MAKE.LEVEL} =3D=3D 0
>>> CC=3D/usr/bin/clang -target ${TO_TYPE}--freebsd11.0-gnueabi =
-march=3Darmv7a -fmax-type-align=3D4
>>> CXX=3D/usr/bin/clang++ -target ${TO_TYPE}--freebsd11.0-gnueabi =
-march=3Darmv7a -fmax-type-align=3D4
>>> CPP=3D/usr/bin/clang-cpp -target ${TO_TYPE}--freebsd11.0-gnueabi =
-march=3Darmv7a -fmax-type-align=3D4
>>> .export CC
>>> .export CXX
>>> .export CPP
>>> AS=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/as
>>> AR=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/ar
>>> LD=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/ld
>>> NM=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/nm
>>> OBJCOPY=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/objcopy
>>> OBJDUMP=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/objdump
>>> RANLIB=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/ranlib
>>> SIZE=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/size
>>> #NO-SUCH: STRINGS=3D/usr/local/${TOOLS_TO_TYPE}-freebsd/bin/strings
>>> STRINGS=3D/usr/local/bin/${TOOLS_TO_TYPE}-freebsd-strings
>>> .export AS
>>> .export AR
>>> .export LD
>>> .export NM
>>> .export OBJCOPY
>>> .export OBJDUMP
>>> .export RANLIB
>>> .export SIZE
>>> .export STRINGS
>>> .endif
>>=20
>>=20
>>=20
>> =3D=3D=3D
>> Mark Millard
>> markmi at dsl-only.net
>>=20
>>=20
>>=20
>> _______________________________________________
>> freebsd-toolchain@freebsd.org mailing list
>> https://lists.freebsd.org/mailman/listinfo/freebsd-toolchain
>> To unsubscribe, send any mail to =
"freebsd-toolchain-unsubscribe@freebsd.org"





Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?BBAAE33E-BD65-40A3-A0B3-F3346FC08112>