From: Lukasz Majewski <lukma@denx.de>
To: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Cc: Joseph Myers <joseph@codesourcery.com>,
Paul Eggert <eggert@cs.ucla.edu>,
Alistair Francis <alistair23@gmail.com>,
Arnd Bergmann <arnd@arndb.de>,
Alistair Francis <alistair.francis@wdc.com>,
GNU C Library <libc-alpha@sourceware.org>,
Florian Weimer <fweimer@redhat.com>,
Carlos O'Donell <carlos@redhat.com>,
Stepan Golosunov <stepan@golosunov.pp.ru>,
Andreas Schwab <schwab@suse.de>, Zack Weinberg <zackw@panix.com>,
Jeff Law <law@redhat.com>
Subject: Re: [PATCH v2 2/3] Y2038: nptl: Provide futex_abstimed_wait64 supporting 64 bit time
Date: Wed, 30 Sep 2020 16:06:25 +0200 [thread overview]
Message-ID: <20200930160625.00f3ed4d@jawa> (raw)
In-Reply-To: <20200930153953.424a46c2@jawa>
[-- Attachment #1: Type: text/plain, Size: 9247 bytes --]
On Wed, 30 Sep 2020 15:39:53 +0200
Lukasz Majewski <lukma@denx.de> wrote:
> Hi Adhemerval,
>
> > On 30/09/2020 09:47, Lukasz Majewski wrote:
> > > Hi Adhemerval,
> > >
> > >> On 19/09/2020 10:07, Lukasz Majewski wrote:
> > >>> This is the helper function, which uses struct __timespec64
> > >>> to provide 64 bit absolute time to futex syscalls.
> > >>>
> > >>> The aim of this function is to move convoluted pre-processor
> > >>> macro code from sysdeps/nptl/lowlevellock-futex.h to C
> > >>> function in futex-internal.c
> > >>>
> > >>> The futex_abstimed_wait64 function has been put into a separate
> > >>> file on the purpose - to avoid issues apparent on the m68k
> > >>> architecture related to small number of available registers
> > >>> (there is not enough registers to put all necessary arguments
> > >>> in them if the above function would be added to
> > >>> futex-internal.h with __always_inline attribute).
> > >>>
> > >>> Additional precautions for m68k port have been taken - the
> > >>> futex-internal.c file will be compiled with -fno-inline flag.
> > >>>
> > >>
> > >> LGTM with a nit regarding style and a clarification about
> > >> '-fno-inline'.
> > >
> > > Please find the explanation below.
> > >
> > >>
> > >>>
> > >>> ---
> > >>> Changes for v2:
> > >>> - Handle the case when *abstime pointer is NULL
> > >>> ---
> > >>> sysdeps/nptl/futex-internal.c | 70
> > >>> +++++++++++++++++++++++++++ sysdeps/nptl/futex-internal.h
> > >>> | 6 +++ sysdeps/unix/sysv/linux/m68k/Makefile | 2 +
> > >>> 3 files changed, 78 insertions(+)
> > >>>
> > >>> diff --git a/sysdeps/nptl/futex-internal.c
> > >>> b/sysdeps/nptl/futex-internal.c index 3366aac162..3211b4c94f
> > >>> 100644 --- a/sysdeps/nptl/futex-internal.c
> > >>> +++ b/sysdeps/nptl/futex-internal.c
> > >>> @@ -45,6 +45,29 @@ __futex_abstimed_wait_cancelable32 (unsigned
> > >>> int* futex_word, abstime != NULL ? &ts32 : NULL,
> > >>> NULL /* Unused. */,
> > >>> FUTEX_BITSET_MATCH_ANY); }
> > >>> +
> > >>> +static int
> > >>> +__futex_abstimed_wait32 (unsigned int* futex_word,
> > >>> + unsigned int expected, clockid_t
> > >>> clockid,
> > >>> + const struct __timespec64* abstime,
> > >>> + int private)
> > >>> +{
> > >>> + struct timespec ts32;
> > >>> +
> > >>> + if (abstime != NULL && ! in_time_t_range (abstime->tv_sec))
> > >>> + return -EOVERFLOW;
> > >>> +
> > >>> + unsigned int clockbit = (clockid == CLOCK_REALTIME) ?
> > >>> + FUTEX_CLOCK_REALTIME : 0;
> > >>> + int op = __lll_private_flag (FUTEX_WAIT_BITSET | clockbit,
> > >>> private); +
> > >>> + if (abstime != NULL)
> > >>> + ts32 = valid_timespec64_to_timespec (*abstime);
> > >>> +
> > >>> + return INTERNAL_SYSCALL_CALL (futex, futex_word, op,
> > >>> expected,
> > >>> + abstime != NULL ? &ts32 : NULL,
> > >>> + NULL /* Unused. */,
> > >>> FUTEX_BITSET_MATCH_ANY); +}
> > >>> #endif
> > >>>
> > >>> int
> > >>
> > >> Ok.
> > >>
> > >>> @@ -97,3 +120,50 @@ __futex_abstimed_wait_cancelable64 (unsigned
> > >>> int* futex_word, futex_fatal_error ();
> > >>> }
> > >>> }
> > >>> +
> > >>> +int
> > >>> +__futex_abstimed_wait64 (unsigned int* futex_word, unsigned int
> > >>> expected,
> > >>> + clockid_t clockid,
> > >>> + const struct __timespec64* abstime,
> > >>> int private) +{
> > >>> + unsigned int clockbit;
> > >>> + int err;
> > >>> +
> > >>> + /* Work around the fact that the kernel rejects negative
> > >>> timeout values
> > >>> + despite them being valid. */
> > >>> + if (__glibc_unlikely ((abstime != NULL) && (abstime->tv_sec <
> > >>> 0)))
> > >>> + return ETIMEDOUT;
> > >>> +
> > >>> + if (! lll_futex_supported_clockid(clockid))
> > >>> + return EINVAL;
> > >>
> > >> Space before parentheses.
> > >
> > > Ok. Fixed.
> > >
> > >>
> > >>> +
> > >>> + clockbit = (clockid == CLOCK_REALTIME) ?
> > >>> FUTEX_CLOCK_REALTIME : 0;
> > >>> + int op = __lll_private_flag (FUTEX_WAIT_BITSET | clockbit,
> > >>> private); +
> > >>> + err = INTERNAL_SYSCALL_CALL (futex_time64, futex_word, op,
> > >>> expected,
> > >>> + abstime, NULL /* Unused. */,
> > >>> + FUTEX_BITSET_MATCH_ANY);
> > >>> +#ifndef __ASSUME_TIME64_SYSCALLS
> > >>> + if (err == -ENOSYS)
> > >>> + err = __futex_abstimed_wait32 (futex_word, expected,
> > >>> + clockid, abstime, private);
> > >>> +#endif
> > >>> + switch (err)
> > >>> + {
> > >>> + case 0:
> > >>> + case -EAGAIN:
> > >>> + case -EINTR:
> > >>> + case -ETIMEDOUT:
> > >>> + return -err;
> > >>> +
> > >>> + case -EFAULT: /* Must have been caused by a glibc or
> > >>> application bug. */
> > >>> + case -EINVAL: /* Either due to wrong alignment, unsupported
> > >>> + clockid or due to the timeout not being
> > >>> + normalized. Must have been caused by a
> > >>> glibc or
> > >>> + application bug. */
> > >>> + case -ENOSYS: /* Must have been caused by a glibc bug. */
> > >>> + /* No other errors are documented at this time. */
> > >>> + default:
> > >>> + futex_fatal_error ();
> > >>> + }
> > >>> +}
> > >>
> > >> Ok.
> > >>
> > >>> diff --git a/sysdeps/nptl/futex-internal.h
> > >>> b/sysdeps/nptl/futex-internal.h index 7f3910ad98..1ba0d61938
> > >>> 100644 --- a/sysdeps/nptl/futex-internal.h
> > >>> +++ b/sysdeps/nptl/futex-internal.h
> > >>> @@ -529,4 +529,10 @@ __futex_abstimed_wait_cancelable64
> > >>> (unsigned int* futex_word, const struct __timespec64* abstime,
> > >>> int private)
> > >>> attribute_hidden;
> > >>> +int
> > >>> +__futex_abstimed_wait64 (unsigned int* futex_word, unsigned int
> > >>> expected,
> > >>> + clockid_t clockid,
> > >>> + const struct __timespec64* abstime,
> > >>> + int private) attribute_hidden;
> > >>> +
> > >>> #endif /* futex-internal.h */
> > >>
> > >> Ok.
> > >>
> > >>> diff --git a/sysdeps/unix/sysv/linux/m68k/Makefile
> > >>> b/sysdeps/unix/sysv/linux/m68k/Makefile index
> > >>> be40fae68a..65164c5752 100644 ---
> > >>> a/sysdeps/unix/sysv/linux/m68k/Makefile +++
> > >>> b/sysdeps/unix/sysv/linux/m68k/Makefile @@ -21,3 +21,5 @@
> > >>> sysdep-dl-routines += dl-static sysdep-others += lddlibc4
> > >>> install-bin += lddlibc4
> > >>> endif
> > >>> +
> > >>> +CFLAGS-futex-internal.c += -fno-inline
> > >>>
> > >>
> > >> Do we still need it?
> > >
> > > Unfortunately, yes. The m68k architecture has the issue with
> > > properly compiling this code (in futex-internal.c) as it has very
> > > limited number of registers (8 x 32 bit IIRC).
> > >
> > > I did some investigation and it looks like static inline
> > > in_time_t_range() function affects the compiler capabilities to
> > > generate correct code.
> > >
> > > It can be easily inlined on any architecture but m68k.
> > >
> > > As m68k has issues with this code compilation - it is IMHO better
> > > to disable inlining (-fno-inline) on this particular port.
> > >
> > > As a result we would have slower execution only on relevant
> > > functions, but 64 bit time support would be provided.
> >
> > I recall this was need on first patch for the cancellable 64-bit
> > futex_abstimed_wait where it embedded the 32-bit fallback in the
> > __futex_abstimed_wait_cancelable64. And my suggestion to move it to
> > an external function seems to avoid the extra compiler flag.
> >
> > This patchset uses the same strategy and I checked both patches
> > and it seems that -fno-inline is not required to build m68k.
>
> As fair as I can tell (at the time I tested it) - it was necessary to
> have -fno-inline when this patch was applied on top of the one which
> is now in the -master branch (in futex-internal.c).
>
> I will double check it now with:
>
> glibc/glibc-many-build --keep failed glibcs
Indeed with current -master the issue is gone:
../src/scripts/build-many-glibcs.py
/home/lukma/work/glibc/glibc-many-build --keep failed glibcs
m68k-linux-gnu m68k-linux-gnu-coldfire m68k-linux-gnu-coldfire-soft
x86_64-linux-gnu
Is all OK with the -fno-inline removed.... Strange.
>
>
> Best regards,
>
> Lukasz Majewski
>
> --
>
> DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
> HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
> Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
> lukma@denx.de
Best regards,
Lukasz Majewski
--
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: lukma@denx.de
[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]
next prev parent reply other threads:[~2020-09-30 14:06 UTC|newest]
Thread overview: 20+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-09-19 13:07 [PATCH v2 1/3] nptl: futex: Move __NR_futex_time64 alias to beginning of futex-internal.h Lukasz Majewski
2020-09-19 13:07 ` [PATCH v2 2/3] Y2038: nptl: Provide futex_abstimed_wait64 supporting 64 bit time Lukasz Majewski
2020-09-21 17:00 ` Alistair Francis
2020-09-29 16:55 ` Lukasz Majewski
2020-09-30 11:52 ` Adhemerval Zanella
2020-09-30 12:47 ` Lukasz Majewski
2020-09-30 12:58 ` Andreas Schwab
2020-09-30 13:17 ` Lukasz Majewski
2020-09-30 13:14 ` Adhemerval Zanella
2020-09-30 13:39 ` Lukasz Majewski
2020-09-30 14:06 ` Lukasz Majewski [this message]
2020-09-19 13:07 ` [PATCH v2 3/3] y2038: nptl: Convert pthread_rwlock_{clock|timed}{rd|wr}lock to support " Lukasz Majewski
2020-09-29 16:56 ` Lukasz Majewski
2020-09-30 11:59 ` Adhemerval Zanella
2020-09-30 13:12 ` Lukasz Majewski
2020-09-30 13:19 ` Adhemerval Zanella
2020-10-02 12:15 ` Lukasz Majewski
2020-10-02 13:51 ` Adhemerval Zanella
2020-09-29 16:55 ` [PATCH v2 1/3] nptl: futex: Move __NR_futex_time64 alias to beginning of futex-internal.h Lukasz Majewski
2020-09-29 18:12 ` Adhemerval Zanella
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=20200930160625.00f3ed4d@jawa \
--to=lukma@denx.de \
--cc=adhemerval.zanella@linaro.org \
--cc=alistair.francis@wdc.com \
--cc=alistair23@gmail.com \
--cc=arnd@arndb.de \
--cc=carlos@redhat.com \
--cc=eggert@cs.ucla.edu \
--cc=fweimer@redhat.com \
--cc=joseph@codesourcery.com \
--cc=law@redhat.com \
--cc=libc-alpha@sourceware.org \
--cc=schwab@suse.de \
--cc=stepan@golosunov.pp.ru \
--cc=zackw@panix.com \
/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).