linux-kernel - [Patch/RFC 0/4] Allow inlined spinlocks (again)

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-ID: <20090810081324.GA30276@osiris.boeblingen.de.ibm.com>
Date:	Mon, 10 Aug 2009 10:13:24 +0200
From:	Heiko Carstens <heiko.carstens@...ibm.com>
To:	Ingo Molnar <mingo@...e.hu>, linux-kernel@...r.kernel.org
Cc:	Horst Hartmann <horsth@...ux.vnet.ibm.com>,
	Christian Ehrhardt <ehrhardt@...ux.vnet.ibm.com>,
	Martin Schwidefsky <schwidefsky@...ibm.com>
Subject: [Patch/RFC 0/4] Allow inlined spinlocks (again)

This patch set allows to have inlined spinlocks (again).

The rationale behind this is that function calls on at least s390 are
expensive.
If one considers that server kernels are usually compiled with
!CONFIG_PREEMPT a simple spin_lock is just a compare and swap loop.
The extra overhead for a function call is significant.
With inlined spinlocks overall cpu usage gets reduced by 1%-5% on s390.
These numbers were taken with some network benchmarks. However I expect
any workload that calls frequently into the kernel and which grabs a few
spinlocks to perform better.

Please note: I'm aware that this implementation is a bit ;) ugly and
might also contain bugs. But the performance improvements show that its
worth to do something in this area.

The actual implemtation is quite simple: move the spinlock C code to
a header file and add some wrappers. Configuration dependent this
file gets included into a C file and normal spinlock functions like
we them currently will generated. Otherwise the inline variants
will be generated as we had them before the out of line code was
introduced:

A standard spinlock function like this

void __lockfunc _spin_lock(spinlock_t *lock)
{
...
}
EXPORT_SYMBOL(_spin_lock);

gets converted to

LOCKFUNC void _spin_lock(spinlock_t *lock)
{
...
}
SPIN_EXPORT_SYMBOL(_spin_lock);

where LOCKFUNC and SPIN_EXPORT_SYMBOL get expanded dependent
on CONFIG_SPINLOCK_INLINE:

#ifdef CONFIG_SPINLOCK_INLINE

#define LOCKFUNC                 static inline
#define SPIN_EXPORT_SYMBOL(func)

#else /* CONFIG_SPINLOCK_INLINE */

#define LOCKFUNC                 __lockfunc
#define SPIN_EXPORT_SYMBOL(func) EXPORT_SYMBOL(func)

#endif /* CONFIG_SPINLOCK_INLINE */
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/