From mboxrd@z Thu Jan 1 00:00:00 1970 From: Markus Lidel Subject: Request for comment on new I2O driver Date: Tue, 29 Jun 2004 01:21:30 +0200 Sender: linux-scsi-owner@vger.kernel.org Message-ID: <40E0A7FA.4050208@shadowconnect.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit Return-path: Received: from [213.239.201.226] ([213.239.201.226]:18883 "EHLO mail.shadowconnect.com") by vger.kernel.org with ESMTP id S265163AbUF1Xax (ORCPT ); Mon, 28 Jun 2004 19:30:53 -0400 List-Id: linux-scsi@vger.kernel.org To: linux-scsi@vger.kernel.org Hello, in the last few weeks i've partly rewritten the I2O subsystem to better suite the 2.6 kernel (sysfs support, PCI probing, dynamically allocated request queues and gendisks for block devices, ...). Although i'm not finished, it would be nice if someone could comment about it. Because the patch is 500K large, it's better to view the sources online: http://i2o.shadowconnect.com/svn/repos/kernel/trunk/drivers/message/i2o/ Login: anonymous without a password. The i2o_*.c files are still "old", so please skip those :-) Any advices/comments would be appreciated, but escpecially interesting for me would be: - sysfs * Is the way it is done okay? * Is there something important missing? * In the PCI probing function the device struct from pci_dev is simply copied to the device struct from I2O controller. Is this okay, or is there a better way to do it? - interrupt handling * are 4 (or maybe more) function calls inside an interrupt bearable * could / should the interrupt handler put incomming data in a workqueue? - workqueue * i've implemented a workqueue for event handling for each driver (e. g. block device, tape device, ...). Is this to expensive? - file split * i've split the huge i2o_core into 7 smaller ones. Is there a problem with it? - kmalloc/mempool in request queue prep function * i've used a kmalloc in the request queue prep. The scsi layer uses a mempool for it. At the moment the driver only need 47k memory. But is it better to pre-allocate the memory in this case? Thank you very much for any advice/comment in advance. Best regards, Markus Lidel ------------------------------------------ Markus Lidel (Senior IT Consultant) Shadow Connect GmbH Carl-Reisch-Weg 12 D-86381 Krumbach Germany Phone: +49 82 82/99 51-0 Fax: +49 82 82/99 51-11 E-Mail: Markus.Lidel@shadowconnect.com URL: http://www.shadowconnect.com