From mboxrd@z Thu Jan 1 00:00:00 1970 From: James Bottomley Subject: RE: [PATCH] updated: suspending I/Os to a device Date: 19 Aug 2004 13:42:47 -0400 Sender: linux-scsi-owner@vger.kernel.org Message-ID: <1092937373.2378.697.camel@mulgrave> References: <0B1E13B586976742A7599D71A6AC733C02D85A@xbl3.ma.emulex.com> Mime-Version: 1.0 Content-Type: text/plain Content-Transfer-Encoding: 7bit Return-path: Received: from stat16.steeleye.com ([209.192.50.48]:53739 "EHLO hancock.sc.steeleye.com") by vger.kernel.org with ESMTP id S266596AbUHSRm5 (ORCPT ); Thu, 19 Aug 2004 13:42:57 -0400 In-Reply-To: <0B1E13B586976742A7599D71A6AC733C02D85A@xbl3.ma.emulex.com> List-Id: linux-scsi@vger.kernel.org To: James.Smart@Emulex.Com Cc: SCSI Mailing List On Thu, 2004-08-19 at 13:24, James.Smart@Emulex.Com wrote: > The one con: > Depends on well-behaved LLDD's to not leave the device blocked indefinitely. > But this is no different from scsi_block_requests(). But that's the big one. A significant number of driver bugs have come to us from incorrect timer handling (not understanding the difference between del_timer and del_timer_sync being the biggest). The whole reason for moving the code up into the mid-layer is to avoid the LLD handling the timer. After the rewrite to simplify the code, the added overhead of having the mid-layer (and transport classes) do the timer is scarcely sevently lines longer than the one you currently attach. James