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 for Android: free password hash cracker in your pocket
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20130903145719.GB14221@dhcp-26-207.brq.redhat.com>
Date:	Tue, 3 Sep 2013 16:57:19 +0200
From:	Alexander Gordeev <agordeev@...hat.com>
To:	Tejun Heo <tj@...nel.org>
Cc:	linux-kernel@...r.kernel.org, x86@...nel.org,
	linux-pci@...r.kernel.org, linux-ide@...r.kernel.org,
	Ingo Molnar <mingo@...nel.org>, Joerg Roedel <joro@...tes.org>,
	Jan Beulich <JBeulich@...e.com>,
	Bjorn Helgaas <bhelgaas@...gle.com>
Subject: Re: [PATCH 0/4] AHCI: Conserve interrupts with
 pci_enable_msi_block_part() interface

On Tue, Sep 03, 2013 at 09:55:41AM -0400, Tejun Heo wrote:
> Hello,
> 
> On Mon, Sep 02, 2013 at 10:59:07AM +0200, Alexander Gordeev wrote:
> > This series is aimed to conserve on othewise wasted interrupt
> > resources for 10 of 16 unused MSI vectors for AHCI devices on
> > Intel chipsets.
> 
> Hmmm.... I've been looking at the code and and a curiosity.  Why does
> multiple MSI support implicitly enabled threaded IRQ handling?  Why
> are those two linked?  Also, do you have any numbers to show that this
> actually is better?  Handling the processing off to a thread isn't a
> light operation.

Just to emphasize the purpose of this series - it is aimed on configuring
(allocating x86 vectors, irq_desc's etc.) 6/16 interrupts instead of current
16/16 - while 10/16 are unused. Nothing more.

Multiple MSI support enables threaded IRQ handling, because at the time of
posting I did not want to intrude into the existing single-MSI codebase while
multiple MSI/multipe CPU approach gained good numbers (below).

Besides, the ports interrupt handling moved from the interrupt-disabled
hardware context to the interrupt-enabled threaded context and also ports
got per-port locks instead of the single host lock. AFAIR I did not post
the numbers, but it was something close to no-contention.


Host-wide lock statistics summary:

  Taken using command 'if=/dev/sd{a,b,c} of=/dev/null' running in parallel
  on three SATA HDDs:
	ata1: SATA link up 1.5 Gbps (SStatus 113 SControl 300)
	ata2: SATA link up 1.5 Gbps (SStatus 113 SControl 300)
	ata3: SATA link up 3.0 Gbps (SStatus 123 SControl 300)

  ahci_interrupt()
  Capabilities: [80] MSI: Enable+ Count=1/16 Maskable- 64bit-

Test	holdtime-total	waittime-total	acquis.	holdtime-avg	waittime-avg
#
01.	22565801.27	93056.41	6377094	3.538571216	0.014592291
02.	20358324.12	60825.47	6411012	3.175524257	0.009487655
03.	21190730.38	63610.38	6384508	3.319085884	0.009963239
04.	22491064.54	80624.96	6398390	3.515113105	0.012600820
05.	21986193.10	78034.38	6379092	3.446602291	0.012232835
06.	20556437.70	62314.64	6287673	3.269323596	0.009910604
07.	20477137.55	66190.92	6388507	3.205308776	0.010360937
08.	21442240.03	69306.80	6402109	3.349246323	0.010825620
						-----------	-----------
avg						3.352346931	0.011246750

  ahci_hw_interrupt()
  Capabilities: [80] MSI: Enable+ Count=16/16 Maskable- 64bit-

Test	holdtime-total	waittime-total	acquis.	holdtime-avg	waittime-avg
#
09.	8936643.32	54559.79	6505606	1.373683454	0.008386581
10.	8579972.36	51496.56	6496233	1.320761180	0.007927142
11.	8138708.47	54061.46	6494647	1.253141005	0.008324003
12.	8322068.81	60427.51	6509975	1.278356493	0.009282295
13.	8527745.18	65978.83	6515589	1.308821839	0.010126303
14.	8662461.99	57291.39	6510702	1.330495850	0.008799572
15.	8054911.26	69186.41	6514262	1.236504037	0.010620759
16.	9618631.17	39726.37	6517385	1.475842101	0.006095446
						-----------	-----------
avg						1.322200745	0.008695263


> Thanks.
> 
> -- 
> tejun

-- 
Regards,
Alexander Gordeev
agordeev@...hat.com
--
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/

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ