* futex_atomic_cmpxchg_inatomic is missing
@ 2018-09-26 8:41 ` Andreas Schwab
0 siblings, 0 replies; 10+ messages in thread
From: Andreas Schwab @ 2018-09-26 8:41 UTC (permalink / raw)
To: linux-riscv
The riscv port does not define futex_atomic_cmpxchg_inatomic, but glibc
is assuming that set_robust_list always works.
Andreas.
--
Andreas Schwab, SUSE Labs, schwab at suse.de
GPG Key fingerprint = 0196 BAD8 1CE9 1970 F4BE 1748 E4D4 88E3 0EEA B9D7
"And now for something completely different."
^ permalink raw reply [flat|nested] 10+ messages in thread
* futex_atomic_cmpxchg_inatomic is missing
@ 2018-09-26 8:41 ` Andreas Schwab
0 siblings, 0 replies; 10+ messages in thread
From: Andreas Schwab @ 2018-09-26 8:41 UTC (permalink / raw)
To: linux-riscv
The riscv port does not define futex_atomic_cmpxchg_inatomic, but glibc
is assuming that set_robust_list always works.
Andreas.
--
Andreas Schwab, SUSE Labs, schwab@suse.de
GPG Key fingerprint = 0196 BAD8 1CE9 1970 F4BE 1748 E4D4 88E3 0EEA B9D7
"And now for something completely different."
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 10+ messages in thread
* futex_atomic_cmpxchg_inatomic is missing
@ 2018-09-29 2:17 ` Palmer Dabbelt
0 siblings, 0 replies; 10+ messages in thread
From: Palmer Dabbelt @ 2018-09-29 2:17 UTC (permalink / raw)
To: linux-riscv
On Wed, 26 Sep 2018 01:41:18 PDT (-0700), schwab at suse.de wrote:
> The riscv port does not define futex_atomic_cmpxchg_inatomic, but glibc
> is assuming that set_robust_list always works.
Yes. Someone has pointed this out before, but we haven't had time to figure
out the correct kernel implementation yet. Properly implementing the whole API is
probably only a hundred lines of code, but any of this lock stuff is always
very slow going.
Christoph published a simple Kconfig workaround a while ago to turn on the
generic support for non-SMP systems, which might help you work around your
issue. My plan was to construct a for-next branch over the weekend, but I
managed to do so today so I might have time to take a look -- no promises,
though :)
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: futex_atomic_cmpxchg_inatomic is missing
@ 2018-09-29 2:17 ` Palmer Dabbelt
0 siblings, 0 replies; 10+ messages in thread
From: Palmer Dabbelt @ 2018-09-29 2:17 UTC (permalink / raw)
To: schwab; +Cc: Christoph Hellwig, linux-riscv
On Wed, 26 Sep 2018 01:41:18 PDT (-0700), schwab@suse.de wrote:
> The riscv port does not define futex_atomic_cmpxchg_inatomic, but glibc
> is assuming that set_robust_list always works.
Yes. Someone has pointed this out before, but we haven't had time to figure
out the correct kernel implementation yet. Properly implementing the whole API is
probably only a hundred lines of code, but any of this lock stuff is always
very slow going.
Christoph published a simple Kconfig workaround a while ago to turn on the
generic support for non-SMP systems, which might help you work around your
issue. My plan was to construct a for-next branch over the weekend, but I
managed to do so today so I might have time to take a look -- no promises,
though :)
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 10+ messages in thread
* futex_atomic_cmpxchg_inatomic is missing
@ 2018-10-01 7:07 ` Andreas Schwab
0 siblings, 0 replies; 10+ messages in thread
From: Andreas Schwab @ 2018-10-01 7:07 UTC (permalink / raw)
To: linux-riscv
On Sep 28 2018, Palmer Dabbelt <palmer@sifive.com> wrote:
> Christoph published a simple Kconfig workaround a while ago to turn on the
> generic support for non-SMP systems, which might help you work around your
> issue.
What kind of workaround? If you disable SMP this should be
automatically enabled. But of course, being non-SMP, this is also
useless.
Andreas.
--
Andreas Schwab, SUSE Labs, schwab at suse.de
GPG Key fingerprint = 0196 BAD8 1CE9 1970 F4BE 1748 E4D4 88E3 0EEA B9D7
"And now for something completely different."
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: futex_atomic_cmpxchg_inatomic is missing
@ 2018-10-01 7:07 ` Andreas Schwab
0 siblings, 0 replies; 10+ messages in thread
From: Andreas Schwab @ 2018-10-01 7:07 UTC (permalink / raw)
To: Palmer Dabbelt; +Cc: Christoph Hellwig, linux-riscv
On Sep 28 2018, Palmer Dabbelt <palmer@sifive.com> wrote:
> Christoph published a simple Kconfig workaround a while ago to turn on the
> generic support for non-SMP systems, which might help you work around your
> issue.
What kind of workaround? If you disable SMP this should be
automatically enabled. But of course, being non-SMP, this is also
useless.
Andreas.
--
Andreas Schwab, SUSE Labs, schwab@suse.de
GPG Key fingerprint = 0196 BAD8 1CE9 1970 F4BE 1748 E4D4 88E3 0EEA B9D7
"And now for something completely different."
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 10+ messages in thread
* futex_atomic_cmpxchg_inatomic is missing
@ 2018-10-01 22:19 ` Palmer Dabbelt
0 siblings, 0 replies; 10+ messages in thread
From: Palmer Dabbelt @ 2018-10-01 22:19 UTC (permalink / raw)
To: linux-riscv
On Mon, 01 Oct 2018 00:07:22 PDT (-0700), schwab at suse.de wrote:
> On Sep 28 2018, Palmer Dabbelt <palmer@sifive.com> wrote:
>
>> Christoph published a simple Kconfig workaround a while ago to turn on the
>> generic support for non-SMP systems, which might help you work around your
>> issue.
>
> What kind of workaround? If you disable SMP this should be
> automatically enabled. But of course, being non-SMP, this is also
> useless.
I think we just need some sort of
CONFIG_WE_HAVE_GENERIC_FUTEX_CMPXCHG_FOR_NONSMP selected by default in
arch/riscv/Kconfig. I punted on it because having this support only for
non-SMP systems didn't seem so useful.
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: futex_atomic_cmpxchg_inatomic is missing
@ 2018-10-01 22:19 ` Palmer Dabbelt
0 siblings, 0 replies; 10+ messages in thread
From: Palmer Dabbelt @ 2018-10-01 22:19 UTC (permalink / raw)
To: schwab; +Cc: Christoph Hellwig, linux-riscv
On Mon, 01 Oct 2018 00:07:22 PDT (-0700), schwab@suse.de wrote:
> On Sep 28 2018, Palmer Dabbelt <palmer@sifive.com> wrote:
>
>> Christoph published a simple Kconfig workaround a while ago to turn on the
>> generic support for non-SMP systems, which might help you work around your
>> issue.
>
> What kind of workaround? If you disable SMP this should be
> automatically enabled. But of course, being non-SMP, this is also
> useless.
I think we just need some sort of
CONFIG_WE_HAVE_GENERIC_FUTEX_CMPXCHG_FOR_NONSMP selected by default in
arch/riscv/Kconfig. I punted on it because having this support only for
non-SMP systems didn't seem so useful.
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 10+ messages in thread
* futex_atomic_cmpxchg_inatomic is missing
@ 2018-10-09 16:15 ` Christoph Hellwig
0 siblings, 0 replies; 10+ messages in thread
From: Christoph Hellwig @ 2018-10-09 16:15 UTC (permalink / raw)
To: linux-riscv
On Fri, Sep 28, 2018 at 07:17:17PM -0700, Palmer Dabbelt wrote:
> Christoph published a simple Kconfig workaround a while ago to turn on the
> generic support for non-SMP systems, which might help you work around your
> issue. My plan was to construct a for-next branch over the weekend, but I
> managed to do so today so I might have time to take a look -- no promises,
> though :)
Can you pull that commit in at least? It has been dandling in various
riscv tree for more than half a year now..
^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: futex_atomic_cmpxchg_inatomic is missing
@ 2018-10-09 16:15 ` Christoph Hellwig
0 siblings, 0 replies; 10+ messages in thread
From: Christoph Hellwig @ 2018-10-09 16:15 UTC (permalink / raw)
To: Palmer Dabbelt; +Cc: Christoph Hellwig, schwab, linux-riscv
On Fri, Sep 28, 2018 at 07:17:17PM -0700, Palmer Dabbelt wrote:
> Christoph published a simple Kconfig workaround a while ago to turn on the
> generic support for non-SMP systems, which might help you work around your
> issue. My plan was to construct a for-next branch over the weekend, but I
> managed to do so today so I might have time to take a look -- no promises,
> though :)
Can you pull that commit in at least? It has been dandling in various
riscv tree for more than half a year now..
_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv
^ permalink raw reply [flat|nested] 10+ messages in thread
end of thread, other threads:[~2018-10-09 16:15 UTC | newest]
Thread overview: 10+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2018-09-26 8:41 futex_atomic_cmpxchg_inatomic is missing Andreas Schwab
2018-09-26 8:41 ` Andreas Schwab
2018-09-29 2:17 ` Palmer Dabbelt
2018-09-29 2:17 ` Palmer Dabbelt
2018-10-01 7:07 ` Andreas Schwab
2018-10-01 7:07 ` Andreas Schwab
2018-10-01 22:19 ` Palmer Dabbelt
2018-10-01 22:19 ` Palmer Dabbelt
2018-10-09 16:15 ` Christoph Hellwig
2018-10-09 16:15 ` Christoph Hellwig
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.